capabilities of modern numerical differential … · design numerical differential protections with...

CAPABILITIES OF MODERN NUMERICAL DIFFERENTIAL PROTECTIONS

Zoran Gajic, Janez Zakonjšek

ABB Automation Technology Products AB Control and Force Measurement / SA

Västerås, Sweden e-mail: [email protected]

e-mail: [email protected]

1. ABSTRACT The paper presents a fact that it is now possible to design numerical differential protections with simi-lar or better performances than the best available in previously known analogue differential relays. For transformer and busbar differential protection ap-plications, it is extremely important to built-in good security since an unwanted operation might result in severe consequences for a complete power sys-tem. On the other hand, the protection must be highly dependable as well. Failure to operate or even slow operation of the differential relay, in case of an actual internal fault, can have fatal conse-quences. These two main requirements are contra-dictory to each other. The characteristic of the con-ventional instrument transformers during transient conditions is an important input for the design. The analogue relays have been successfully used for these types of applications for years. Can their nu-merical counterparts offer the same performance? The report focuses on the numerical differential protection design, with reference to the transient performance of the protected elements together with the associated conventional instrument trans-formers.

2. INTRODUCTION For several decades, the power system protection relaying has experienced many important changes, from purely electromechanical type to the mixture of electronic and electromechanical type, then to fully static, and now fully numerical protectionss based on microprocessors. In the transformer and busbar protection area, similar changes can be seen. With the fast development in the area of digital technology, the hardware and software used must provide economic solutions regarding the cost of functions and hardware elements for the dedicated diffe rential protection applications.

2.1 Terminal Hardware Platform The basic idea in building the new protection sys-tem platform is coming from two points:

• The design of new hardware and software platforms must be possible to cover quite wide range of applications, and

• It should be open for future upgrading as in presented in reference [5].

According to these requirements, the Pentium fa m-ily of microprocessors has been selected for the protection applications. The main advantages of the Pentium processor are due to its high floating per-formance, low power consumption and low cost. For communication with high speed modules such as analogue input modules and high speed serial interface, the Pentium is equipped with a Compact-PCI bus. For communication with low speed mo d-ules such as binary input and output modules, a CAN bus interface is used. The remote HMI can be connected via different communication buses such as SPA and LON. The system is provided with con-tinuous self-monitoring and diagnostics. In order to fulfill the application requirements for different types of flexible protection, control, and monitoring purposes, the following modules are available within the hardware platform:

• Up to two analogue input modules (AIMs), each of them including 10 analogue inputs with built -in analogue to digital converters (A/D-converter)

• A number of binary input/output modules can be selected from the standard hard-ware library. Among these modules, 24 output relay contacts are available in one binary output module (BOM); 16 binary inputs are available in one binary input module (BIM), and 12 contact outputs and 8 binary inputs are provided in one module known as binary input and output module (BIOM)

• The mA -analogue input modules (MIM) used e.g. for tap position indication for on load tap changer control and for tempera-ture reading via different measuring con-verters are another alternative.

• Power supply module (PSM) • Main CPU module with high performance

Pentium processor

3. DIFFERENTIAL PROTECTIONS The basic concept of any differential protection is that the sum of all currents, which flow into the protection zone, must be equal to the sum of all currents, which flow out of the protection zone. If that is not the case, an internal fault has occurred. This is practically a direct use of Kirchoff’s first law, which is taught in the “Basics of Electricity” during the first year of electrical engineering stud-ies. Unfortunately, practice is often different and more difficult than the theory in the books.

3.1 Magnetic Core Current Transformers Protective devices in general do not measure di-rectly the primary currents in the high voltage con-ductors, but the secondary currents of conventional magnetic core current transformers, which are in-stalled in all high-voltage bays. Because the current transformer is a non-linear measuring device, the secondary CT current can differ drastically from the original primary current under high current condi-tions in the primary CT circuit. This is caused by CT saturation, a phenomenon that is well known to protection engineers, references [8-11]. It is espe-cially relevant for bus differential protection appli-cations, because it has the tendency to cause un-wanted operation of the bus differential protection. Remanence in the magnetic core of a current trans-former is an additional factor, which can influence the secondary CT current. It can improve or reduce the capability of the current transformer to properly transform the primary current to the secondary side. However the CT remanence is a random parameter and it is not possible in practice to precisely deter-mine it. Another, and maybe less known, transient phe-nomenon appears in the CT secondary circuit at the instant when a high primary current is interrupted. It is particularly dominant if the HV circuit breaker interrupts the primary current before it’s natural zero crossing (current chopping). This phenomenon is manifested as an exponentially decaying dc cur-rent component in the CT secondary circuit. This secondary dc current has no corresponding primary current in the power system. The phenomenon can be simply explained as a discharge of the magnetic energy stored in the magnetic core of the current transformer during the high primary current condi-tion. Depending on the type and design of the cur-rent transformer this discharging current can have a time constant in the order of a hundred millisec-onds. Therefore, it has to be considered during the design stage of a differential protection in order to prevent its unwanted operation at the instant when a heavy external fault is cleared by the faulty line’s circuit breaker.

3.2 Analogue differential protection One of the most common types of analogue differ-ential relay design is the high impedance differen-tial protection. This design solves all practical prob-lems caused by the CT non-linear characteristics by using the galvanic connection between the secon-dary circuits of all CTs connected to the protected zone. The scheme is designed in such a way that the current distribution through the differential branch during all transient conditions caused by non-linearity of the CTs will not cause the unwanted operation of the differential protection. This is achieved by connecting a high-impedance (usually resistance) in series with the operating element of the differential relay. This impedance will then limit the level of false differential current through the differential branch. To obtain the optimum relay performance, the resistive burden in the individual CT secondary circuits must be kept low and should have a similar value in all bays. Therefore no other device can be connected to the same CT core. At the same time, due to the lack of any other restraint quantity, it is strongly required that all current transformers within one differential zone have to have the same ratio and the same magnetizing char-acteristic. All these requirements impose additional expense to the power utility in order to purchase specially made CTs dedicated only to the differen-tial protection. However if all of the above conditions are met, this scheme is quite reliable and very sensitive. Usual operating times for internal faults are around one power system cycle. The percentage restrained differential protection scheme, based on a special analogue circuit, de-pends as well on the galvanic connection between the secondary circuits of all CTs, connected to the protected zone, to remain stable for all transient conditions caused by the non-linearity of the main CTs. The galvanic connection is made via a special diode circuit arrangement as shown in Figure 1. This diode circuit creates the rectified incoming current IT3 and the rectified outgoing current IL. The difference between these two currents is the differ-ential current Id1. By this approach, a very simple but effective design of the relay is obtained. All relay decisions are based only on these three quanti-ties, and the operation of the relay does not depend on the number of connected HV bays to the protec-tion zone. Stability of this protection is guaranteed, regardless of the primary fault current level and CT saturation, if the total CT secondary circuit loop resistance, transferred across its intermediate current trans-former to the relay side, is less than or equal to the resistance Rd11 in the differential relay branch (for operating slope of 0.5). Due to this special design feature, the relay allows much bigger resistance to be included in the secondary circuits of the individ-ual main current transformers than in the original

high impedance scheme. It can accommodate dif-ferent CT ratios by use of intermediate current transformers. The CT requirements are very moder-ate and the relay can tolerate other relays on the same CT core. At the same time, by the use of high-speed reed relay (dR on Figure 1), this differential

protection scheme reliably detects internal faults within 1 to 3 milliseconds and issues the trip signal to the high voltage circuit breakers within 9 to 13 milliseconds from the occurrence of the internal fault.

Figure 1: Principal schematic of the analogue percentage restrained differential protection

Although successful in practice, both of the above mentioned analogue differential relays have some shortcomings, such as the need for intermediate CTs to match the different CT ratios, the lack of self-supervision, secondary CT switching for dou-ble or multiple busbar arrangements and relatively complicated scheme engineering. Their application is limited to protection of power system elements, which have galvanically con-nected all input and output terminals, like:

• Stator windings of power generators and big motors

• Galvanically connected terminals of power transformers (autotransformers)

• Shunt and serial reactors • Capacitors, etc.

They cannot be applied for transformer protection in cases where their protection area is extended over galvanically separated windings.

3.3 Numerical differential busbar protection The first generation of digital (not numerical) dif-ferential protection was developed in order to cover for the known shortcomings and to fulfill the addi-tional requirements. However, these new devices could not match the performance of the analogue

relays regarding the speed of operation. The de-mands on the main current transformers were as well much higher than in the case of the analogue percentage restrained differential relay. Therefore, the analogue busbar differential protection schemes continued to be used worldwide. Digital static and microprocessor-based devices require special preventive actions against the influ-ence of non-linear current transformers for each current input separately. This made the hardware of static relays very expensive and causes the algo-rithms in digital terminals to be complex and de-manding on computer capacity. ABB has decided for this reason to re-use the im-portant quantities from the analog percentage re-strained differential protection (see Figure 1), like incoming, outgoing and differential currents, in the new numerical design. These three quantities can be easily calculated numerically from the raw sample values from all analog CT inputs connected to the differential zone. At the same time, they have ex-tremely valuable physical meaning, which clearly describes the condition of the protected zone during all operating conditions. Much better operating characteristics and perform-ances of numerical differential protections are fea-sible compared to conventional designs due to the possibility to:

• Memorize all characteristic quantities cal-culated in the past

• Perform practically unlimited number of logical operations on the above mentioned quantities

• Process the same quantity on a number of different ways

In new numerical protection devices, all CT and VT inputs are galvanicaly separated from each other. All analogue input quantities are sampled with a constant sampling rate and these discreet values are then transferred to corresponding numerical values (i.e. AD conversion). After these conversions, only the numbers are used in the protection algorithms. It is for this reason impossible to directly re-use and copy the operating principles from the analogue protection schemes, because there is not any gal-vanic connection between the CTs and VTs.

3.4 Numerical differential transformer protection The most important protection function in modern transformer protection, control, and monitoring terminals is the differential protection function. Two trip criteria are used. The first trip criterion is based on unrestrained differential trip algorithm for heavy internal faults, which will produce a very high differential current so that it is not necessary to check if it is inrush conditions. The second trip criterion is the through fault current restrained dif-ferential algorithm which also includes the wave-form blocking criterion in combination with criteria of second harmonic restrain for inrush and the fifth harmonic restrain for over-excitation. The function is designed to consider stabilization from up to five groups of three-phase current inputs, which covers most of applications for the transformer differential protection. The vector group and amplitude com-pensation and the zero sequence current deduction are made internally by software according to the connection group and rated data information for a protected power transformer.

4. ADAPTIVE FEATURES OF NU-MERICAL DIFFERENTIAL ALGO-RITHMS The adaptive relaying has been formally defined in CIGRE working group report, reference [12] as follows: "Adaptive protection is a protection phi-losophy which permits and seeks to make adjust-ments in various protection functions automatically in order to make them more attuned to prevailing power system conditions." In the differential pro-tection terminals the following adaptive functions are provided.

4.1 Through Fault Current Stabilization for T Configurations The application of transformer differential protec-tion function in one and half breaker scheme or

double breaker scheme is known as T configuration application. The differential protection function will meet stability problem for an ext ernal fault F1 in Figure 2, if two groups of current inputs (CT1 and CT2) are simply summed externally and connected into the differential protection as one group of cur-rents.

Figure 2: Power Transformer with T configuration on HV side On the other hand, there is a risk that the sensitivity of the differential protection will considerably be reduced for an internal fault F2, if all current inputs are connected separately in a traditional way as restrain signals. The modern transformer differential protection function has an adaptive feature that considers both stability and sensitivity by investigating the original current inputs so that a reasonable restrained current is calculated adaptively for each individual fault. Finally, a special adaptive feature in this algorithm is that the sensitivity will be automatically reduced in case when heavy external faults are detected.

4.2 On-Load-Tap-Changer Compensation for Transformer Differential Protection The transformer differential protection function has a spill current due to the on load tap changer influ-ence. Normally, the control range of an on load tap changer might be around ±15% of rated voltage so that the contribution of spill current can often be around 15%. In order to keep the security and avoid unwanted operations of the conventional differen-tial protection, it is common to set the minimum sensitivity level above 30% of rated current. The sensitivity of the differential protection will be poor and in mo st cases it is not possible to detect inter-turn faults in transformer windings. In the new differential protection function, the tap changer position is monitored continuously and used in the algorithm so that the transformer turns ratio can be calculated. Because of the adaptive calculation of the differential current, it is possible

CT1 CT2

CT3

F1

F2

PowerTrans-former

to considerably increase the sensitivity and the pos-sibility to detect inter-turn faults in the transformer. To prevent an unwanted operation if the reading of the tap changer position is lost during normal op-eration, the differential protection function will change its sensitivity and will not issue any trip signal when the calculated differential current is less than 30% of rated current.

4.3 Inrush Blocking for Transformer Differential Protection Usually the inrush current stabilization has been based on the second harmonic blocking method. In most countries this is a generally accepted method with good operational experiences. However, there is a risk of delayed operations, in case of heavy internal faults, due to the presence of a second har-monic secondary current caused by saturation of the current transformer. With a combination of a condi-tionally second harmonic restraint method and a permanent waveform monitoring restraint method, it is now possible to gain a protection with both, high security and stability against inrush effects and at the same time maintain high performance in case of heavy internal faults. The second harmonic blocking criterion will be enabled both when the transformer is not energized and during the first period after the power trans-former is energized. The second harmonic blocking criterion will also be automatically enabled for some seconds once a heavy external fault is de-tected. This will further reduce the risk of an un-wanted operation due to recovery inrush current, when the heavy external fault is cleared. The wave-form-blocking criterion is always activated to detect initial inrush, sympathetic inrush, and recovery in-rush.

4.4 Adaptive Differential Current Calculation for Bus Differential Protection Differential protections do not measure directly the primary currents in the high voltage conductors, but the secondary currents of magnetic core current transformers, which are installed in all high-voltage bays. Because the current transformer is a non-linear measuring device, under high current condi-tions in the primary CT circuit the secondary CT current can be drastically different from the original primary current. This is caused by CT saturation, a phenomenon that is well known to protection engi-neers. This phenomenon is especially relevant for bus differential protection applications, because it has the tendency to cause unwanted operation of the differential protection. Another difficulty is the large number of main CTs, which can be connected to the bus differential relay. If the CT saturation has to be checked and preven-tive measures taken for every HV CT connected to the protection zone on one-by-one basis, the differ-ential protection algorithm would be slow and quite

complex. Therefore in RED 521 design only the properties of incoming, outgoing and differential currents are used in order to cope with CT satura-tion of any main CT connected to RED 521. This CT saturation compensation logic effectively suppresses the false differential current by looking into properties of the incoming, outgoing and dif-ferential currents. Output of the logic is a modified RMS value of the differential current which has quite small value during external faults followed by CT saturation or full differential current (Id) value in case of an internal fault. This logic incorporates a memory feature as well in order to cope with full CT remanence in the faulty overhead line bay in case of a high-speed automatic reclosing onto a permanent fault. By this approach a new, patent pending differential algorithm has been formed which is completely stable for all external faults and operates very fast in case of all internal faults. All problems caused by the non-linearity of the CTs are solved in an innovative numerical way on the basic principles described above.

4.5 Open CT Detection for Bus Differential Pro-tection Quite a number of main CTs can be connected to one bus differential protection. When a CT circuit is open circuited by a mistake the differential current increases and the protection might maloperate and disconnect all circuits connected to the protected bus. This might have serious consequences for the power utility. Due to this reason a special algorithm is implemented inside the bus differential protection in order to prevent the maloperation in case of an open CT circuit condition. The open CT detection logic will instantly detect the moment when a healthy CT secondary circuit carrying the load current is accidentally opened (i.e. current interrupted to the differential protection). The logic is based on the perception that the total through-load current is the same before and after that CT is open circuited, but the differential current suddenly appears. In order to prevent false opera-tion of this logic in case of a fault or disturbance in the power system, the total through-load current must not experience big changes for three seconds before the open CT condition is detected. When open CT condition is declared, the trip output of the affected phase is blocked and alarm is given to the operator.

5. TESTING OF NEW DIFFEREN-TIAL PROTECTIONS Newly developed protection functions must be tested extremely intensive under different known system conditions, which can influence their de-pendability as well as their security. Different methods are available for testing, but dynamic modeling on different types of models, which cor-

respond the best to real system conditions, are a must. Different utilities worldwide have also their own bad operational experiences with older types of protection devices, which are often transferred to corresponding tests, and specified as obligatory for testing the new protection devices, before they can be applied in their power system. Results of two dynamic test simulations are pre-sented in paper. The first one was executed on a model of big EHV power transformer at ECEPA in PR China. The second one is a repetition of tests, which have been performed on analogue busbar

differential protection type RADSS for more than 25 years ago.

5.1 System set-up for testing of transformer dif-ferential protection and the test results ECEPA’s analogue power system simulator was used to perform all tests. The simulator can incor-porate the analogue models of lines, transformers, generators etc. For the autotransformer protection testing, the simulated test system was therefore set-up in accordance with single-phase transformer units. The tested power system with the tested in-ternal and external fault positions is shown in Figure 3.

Figure 3: Test system set-up for testing the transformer differential protection on a model of big power transformer

5.1.1 Autotransformer Model

The common and serial windings of the autotrans-former model consist of more than twenty coils with independently connectable taps in order to facilitate autotransformer connection with different turns ratio, faults close to the neutral, turn-to-turn faults involving any number of short-circuited turns etc. The LV winding has the similar construction. By changing the relative position of the coils and the size of the air-gap of the iron core, the leakage impedance and magnetizing impedance of the auto-transformer model can be adjusted. The three sin-gle-phase transformers were connected in such a way to form a three-phase autotransformer with Yy0d11 vector group connection. The equivalent data for three-phase autotransformer model are 1082/1082/541MVA, 500/250/115kV.

5.1.2 Protection System Connections and Set-tings

Overall three winding biased differential protection function (ANSI device No 87T) with additional unrestrained operating level (ANSI device No 87H) and low impedance restricted earth-fault protection function (ANSI device No 87N) were activated within protection terminal RET 521 during all tests. All ten currents connected to the protection system were recorded during all tests, as shown in Figure 3. The differential protection function (87T & 87H) has measured the three-phase currents from all three sides of the protected autotransformer. Ho wever it shall be noted that currents from the LV side were located inside of the delta winding. Therefore measured currents on all sides were in phase. Due

to that fact, autotransformer vector group was set as Yy0y0 in the protection system. Automatic deduc-tions of the zero-sequence currents for the differen-tial function were enabled on all three sides within the protection system. This was done in order to prevent the unwanted operation of the transformer differential function during external single-phase to ground faults on 500kV or 250kV side. The cross blocking between the phases was always disabled. Therefore during any inrush conditions each phase individually had to restrain and prevent unwanted operation only on the measurement available from that phase. The minimum pick-up current for the transformer differential function was set to 30% on the protected autotransformer base of 1082MVA. The differential function overall operating charac-teristic is given in Figure 4.

Figure 4: Restrained (87T) & unrestrained (87H) operat-ing characteristics

Figure 5: Restricted Earth-fault (87N) operating charac-teristic

The restricted earth-fault function (87N) measured two sets of three-phase currents from 500kV and 250kV sides and one single-phase current from the autotransformer common neutral point. The mini-mum pick-up current for the restricted earth-fault function was set to 30% on the base of rated current

for the 250kV winding. Its overall operating charac-teristic is given in Figure 5.

5.1.3 Internal faults followed by CT satura-tion

Traditionally the second harmonic blocking is used in order to restrain the transformer differential re-lays during inrush condition. However it is known that the second harmonic blocking can prevent or delay the operation of the differential relay for in-ternal faults followed by CT saturation. The behavior of the protection system during such operating conditions was tested (see Figure 6). First the protection system was set-up in the traditional way. The second harmonic blocking was always active, and the set level for this restrain criterion was 15%.

Figure 6: Late 87T trip due to traditional use of 2nd har-monic blocking criteria

Figure 7: Fast 87T trip

After this test, the unique feature of the protection system, to adaptively use the second harmonic blocking was enabled. No any other setting parame-ter was changed. In this case the operation of the differential function for internal faults was not ef-fected at all by presence of second harmonic due to distorted CT secondary current, as shown in Figure 7.

These tests show that modern numerical protection system can adaptively use the second harmonic blocking criteria. The protection system can utilize it as restrain quantity during inrush conditions, but disregard its delaying influence during internal faults. This feature ensures much quicker operation of the differential function within numerical protec-tion for internal faults followed by CT saturation.

5.1.4 Internal turn-to-turn faults

The turn-to-turn fault is a unique type of fault ap-pearing only in electrical machines (i.e. transform-ers, generators and motors). The main problem with this type of fault is that the terminal currents are almost not affected at all, but the currents in the short-circuited turns can be many times higher than the rated winding. Table 1: Summary for Turn-to-turn Faults

Fault Position

Percentage of Short-Circuited

turns

Differential current

Common Winding

1% 0.12 pu

Common Winding 2% 0.62 pu

Serial Winding 1% 0.11 pu

Serial Winding

2% 0.55 pu

LV Winding 6% 1.42 pu

The differential protection function is the only elec-trical protection, which can detect this type of fault. Turn-to-turn faults were simulated in serial, com-

mon and LV autotransformer windings. The influ-ence of the different fault locations within the wind-ing was checked. The following Table 1 summa-rizes the test results for the several test cases. During these tests, the differential protection func-tion detected and tripped all turn-to-turn faults, which had two or more percents of short-circuited turns.

5.1.5 Internal winding-to-earth faults

The internal winding-to-earth fault is a unique type of fault appearing only in electrical machines. The fault voltage varies in proportion to the fault loca-tion within the winding. It has the full phase-to-earth voltage value for faults close to the winding bushing and is equal almost to zero volts for fault close to the autotransformer neutral point. The most difficult faults to detect are the faults close to the neutral point. For these faults the terminal currents are almost not affected at all, but the current in the autotransformer neutral has very high value irre-spective of the fault location in the winding. This is practically the major difference in comparison with the turn-to-turn faults. The restricted-earth-fault protection function and the differential protection function are the main electrical type protections for this type of faults. However due to the fact that the neutral current is always high for this type of faults, restricted-earth-fault protection function has shown distinct advan-tages for fast disconnection of autotransformer for this type of internal fault. Winding-to-earth faults were simulated in serial and common autotrans-former windings. The following Table 2 summa-rizes the test results for several test cases.

Table 2: Summary for Winding-to-earth Faults

Fault Posi-tion

Percent of involved

turns

Fault Resistance

Neutral Current

Common Winding

1% from neutral 0 Ohms 8.4 kA

Common Winding


Common Winding

2% from neutral

6.5 Ohms 0.75 kA

Common Winding

4.3% from neutral 6.5 Ohms 1.6 kA

250kV Bushing


500kV Bushing

100% from neutral

0 Ohms 9.9 kA

500kV Bushing


500kV Bushing


The restricted earth fault function successfully de-tected and tripped all winding-to-earth faults re-gardless of the percentage of involved turns and fault resistance.

5.1.6 Energizing of faulty autotransformer

After a couple of incidences in China with energiz-ing of faulty transformer, followed by failure of the protection system to operate for these conditions, the local utilities have set-up very strict require-ments on new transformer protection system for this type of operating conditions. Therefore, all types of internal faults including turn-to-turn and winding to earth faults must be tested during dy-namic simulations. The numerical protection sys-tem successfully detected and cleared all of these faults.

Figure 8: Energizing of faulty autotransformer with inter-nal Phase L3-to-earth fault

Figure 9: Energizing of faulty autotransformer with internal 2% winding-to-earth fault

For simplicity, the results of only two test cases are presented in this paper. Both test cases represent switching-on the autotransformer from the 500kV side. However, in both cases the transformer had internal earth fault, which was quickly detected and tripped by the numerical protection system. For

these two test cases direct pictures from the distur-bance evaluation tool are shown in Figure 8and Figure 9.

5.1.7 Practical case

Finally the proper operation of the differential pro-tection function in substation Tumbri, Croatia will be presented. Internal fault was located on the 31.5kV bushings of the three-winding autotrans-former with the follo wing nominal data: 400/115/31,5kV; 300/300/100MVA; vector group Yy0d5.

Figure 10: Disturbance recording from Tumbri substation

The fault current was limited by transformer self impedance, but as it can be seen from Figure 10 the differential function has operated in only 30ms from the moment of fault inception.

5.2 Heavy current testing of a busbar differential protection When a new protection is designed, it is extremely important to properly test it in order to verify the performance and check its limits. The performance of the analogue percentage restrained differential protection was checked in the heavy current labora-tory about twenty-five years ago, with outstanding results, reference [1 & 2]. Similar tests were per-formed on the new numerical differential protec-tion.

5.2.1 Stability tests for numerical differential protections

When the design and the implementation of the new numerical differential protection were finished, the following heavy current testing facility was set-up to test its performances: • Short circuit current generator with the capabil-

ity of producing currents of 26kA or 50kA

RMS, and all other necessary heavy current equipment

• Calibrated resistive shunt for measurement of the primary current

• Digital recorder with the sampling interval of nine microseconds with ten input channels

• Four identical dual ratio CTs with the data as specified in Table 3.

Table 3: Data for dual ratio CTs used during testing

Ratio Class Rating

[VA] Secondary Resistance

[Ohms]

Knee-point

Voltage [V]

200/1 5P20 10 1.7 280 400/1 5P20 20 3.4 580

These four current transformers are designated (Figure 11) as TA, TB, TC and TX. The three cur-rent transformers TA, TB and TC were used with the ratio 400/1 in all tests. However for current transformer TX the ratio 200/1 was used. It was as well possible to add additional resistance RX in the secondary circuit of the TX current transformer in order to obtain the very heavy CT saturation.

In addition, a 200 mA dc current source was avail-able to pre-magnetize the TX current transformer in order to obtain the maximum possible remanence in the magnetic core for some tests. The following quantities were recorded during all tests: • Itot – total primary current of the short circuit

generator • IA – current on the secondary side of the TA

current transformer • IB – current on the secondary side of the TB

current transformer • IX – current on the secondary side of the TX

current transformer • Trip – status of the binary trip output contact

from the numerical differential protection However, it should be noted that all currents re-corded on the secondary side of the current trans-formers (IA, IB and IX) are scaled, in all of the following figures, with the relevant CT ratio in or-der to be compared with the primary current Itot. As can be seen from the Figure 11, the test circuit was arranged in such a way that it was possible to test external faults with or without the preload. During all the tests a primary current level of 26kA RMS was used.

G

IX

IA

IB

TripShunt RXTX

TATBTC

E

C

RED521ABB Relays

NumericalDifferential

Relay

L R

CB1

CB2

Itot

Figure 11: Simplified test circuit for external fault testing of the numerical differential protection

If the circuit breaker CB2 was closed before the main circuit breaker CB1, then the preload of 200A was established. On closing of the main circuit breaker CB1 the external fault condition was ap-plied to the numerical differential protection. It was as well possible to control the closing instant of the main circuit breaker in order to control the dc offset of the primary current.

The total incoming current to the differential zone was supplied to the numerical protection via three parallel-connected current transformers (TA, TB and TC). However, the total outgoing current from the differential zone was supplied to the numerical protection via only the weak current transformer TX. Therefore, during the external fault, the meas-ured current IX on the secondary side of the current

transformer TX had to balance the differential pro-tection and prevent any unwanted operation. For simplicity only two external fault test cases are presented in this paper. In test case No 20 (Figure 12), the current transformer TX was pre-magnetized with the dc current in order to get maximum possi-ble remanence. Resistance RX had a value of 30 ohms and the fault was applied without any pre-load. The switching angle was chosen to get the maximum possible dc offset of the fault current. The first peak value of the primary current was 65kA. Current transformer TX saturated within 2 ms, but the numerical differential protection re-mained fully stable as can be seen from Figure 12.

Figure 12: Test case No 20 - External Fault

Figure 13: Test case No 75 - External Fault

In test case No 75 (Figure 13), the current trans-former was not pre-magnetized, and the secondary

resistance RX had a value of 48 Ohms. The fault was applied after a preload current of 200A. The switching angle was chosen to get the maximum possible dc offset of the fault current. The first peak value of the primary current was 65kA. Current transformer TX saturated within 1.8 ms, but the numerical protection remained fully stable as can be seen from Figure 13. In all other external fault test cases the numerical differential protection remained fully stable as long as the time to saturation of the weakest current transformer was above 1.2 ms. These tests showed also that the protection is properly designed with respect to transients in the CT secondary circuits caused by clearing of high primary fault current. Such CT transients, captured during these tests in the secondary circuits of the TA and TB current transformers, are represented in Figure 14.

Figure 14: CT transients when external fault is cleared

5.2.2 Dependability tests for the numerical differential protection

During testing of the internal faults supplied with strong input CTs, the test circuit was arranged as shown in Figure 15. During these tests, primary current levels of 26kA RMS and 50kA RMS were used. As it can be seen from Figure 15, the test circuit was arranged in such a way that it was only possi-ble to test the internal faults without any preload. After closing of the main circuit breaker CB1 the internal fault condition is applied to the numerical differential protection. It was as well possible to control the closing instant of the main circuit breaker in order to control the fault inception angle. The protection was tested for the complete possible span of fault inception angles, from 0 to 360 de-grees, in steps of 10 degrees.

GIA

IB

TripShunt

TATBTC

CB1E

C

RED521ABB Relays


Relay

Itot

Figure 15: Test circuit for internal fault testing of the numerical differential protection with strong CTs

The total incoming current to the differential zone was supplied to the numerical protection terminal via three parallel-connected current transformers (TA, TB and TC). The main purpose of these tests was to check the speed of operation of the protec-tion for internal faults. For some of these test cases the primary current level was increased to 50kA RMS in order to check the capability of the protec-tion to operate properly for extremely big secondary CT currents, which flow through the protection. For simplicity only two test cases with the 50kA fault level are presented in this paper. In test case No 43 (Figure 16), the internal fault was applied without any preload. The fault inception angle was chosen in to give completely symmetri-cal fault current without any dc offset. The first peak value of the total primary current was 70kA. The numerical differential protection correctly op-erated in 12ms, as can be seen from Figure 16.

Figure 16: Test case No 43 - Internal Fault

In test case No 45 (Figure 17) the internal fault was applied as well without any preload. The fault in-ception angle was chosen to get the maximum pos-sible dc offset of the fault current. The first peak value of the total primary current was 125kA. Cur-rent transformer TA saturated within 4 ms, but the numerical differential protection correctly operated in <12 ms, as can be seen from Figure 17.


It is as well extremely important to test the opera-tion of the bus differential protection for internal faults with heavy incoming CT saturation. There-fore, the test circuit was re-arranged as shown in Figure 18, and the fault current was supplied to the protection via only one weak current transformer. During these tests a primary current level of 26kA RMS was used.

G

IX TripRXTX

E

C

RED521ABB Relays


Relay

Shunt

CB1Itot

Figure 18: Test circuit for internal fault testing of the numerical differential protection with one weak input CT

For simplicity only one of these test cases is pre-sented here. In test case No 66 (Figure 19), the cur-rent transformer TX was pre-magnetized with the dc current in order to get maximum possible rema-nence. Resistance RX had a value of 30 Ohms.


The internal fault was then applied without any preload. The fault inception angle was chosen to get the maximum possible dc offset of the fault current. The first peak value of the primary current was 65kA. Current transformer TX saturated within 1,2 ms, but the numerical differential protection cor-rectly operated in less than 12 ms, as can be seen from Figure 19. For all other internal fault test cases the numerical differential protection operated as expected with the trip time span from 12 to 16 ms.

6. CONCLUSION The tests and practical fault cases have verified that the modern numerical differential protections can fulfill very demanding requirements set by power utilities. Modern numerical bus and transformer differential protection terminals showed the follow-ing capabilities:

• To remain stable for all external faults re-gardless the heavy CT saturation

• To detect and trip with high speed all in-ternal phase-to-ground & phase-to-phase faults

• To detect and trip all transformer turn-to turn faults with two or more percent of short circuited turns

• To detect and trip all transformer winding-to-earth faults, even the one located only one percent away from the neutral point

• To detect and trip all evolving internal faults

• To detect and trip all types of internal faults during switching of the faulty power transformer

• To remain stable for all magnetic in rushes This means that the new numerical differential pro-tections can practically match the performances of the previous analogue generation of transformer and busbar differential protection devices. They have high speed of operation and very low require-ments on the main current transformers. At the same time they offer all the other benefits of the numerical technology such as communication, self-supervision, no need for auxiliary CTs to match the different CT ratios, software CT switching for dou-ble or multiple busbar arrangements, easy scheme engineering etc.

7. REFERENCES 1. T. Forford and J.R. Linders, 1974, “A Half Cycle

Bus Differential Relay and its Application”, IEEE Transaction on Power Apparatus and Systems, Vol.PAS-93

2. T. Forford, 1980, “High Speed Differential Pro-

tection for Large Generators”, IEE Confer-ence on Developments in Power System Pro -tection, London, UK

3. C.H. Einvall and J.R. Linders, 1975, “A Three-

Phase Differential Relay for Transformer Pro-tection”, IEEE Transaction on Power Appara-tus and Systems, Vol.PAS-94

4. J. Wang, Z. Gajic and N. Brandt, 2001, “High

Performance Differential Protection, Analog versus Numerical”, 55th Georgia Tech Protec-tive Relaying Conference, Atlanta, Georgia-USA

5. Z. Gajic, G. Z. Shen, J. M. Chen and Z. F. Xiang,

2001, “Verification of Utility Requirements on Modern Numerical Transformer Protection by Dynamic Simulation”, IEE Conference on Developments in Power System Protection, Amsterdam, Netherlands

6. B. Hillstrom, N. Cosic and I. Brncic, 1998, "Ad-

vances in Power System Protection", 11th In-ternational Conference on Power System Pro-tection PSP 98, Bled, Slovenia

7. Xiang CF, 1998, “Power Transformer Protection,

Dynamic Simulation Test Program”, English Version of Internal ECEPA Document.

8. Bin L, 2000, “ECEPA Report about the Dynamic

Testing”, English Translation of ECEPTRI Document No 19992221967

9. Sidhu T S, Sachdev M S, Wood H C and Nagpal

M, 1992, “Design, Implementation and Test-

ing of a Microprocessor-Based High-Speed Relay for Detecting Transformer Winding Faults”, IEEE Transaction on Power Delivery, Vol. 7 No 1, pp 108-117.

10. Mikrut M, Winkler W and Witek B, 1989, “Per-

formance of Differential Protection for Three-winding Power Transformers During Tran-sient CT’s Saturation”, 4th International Con-ference on Developments in Power System Protection, IEE Pub. No 302, pp 40-44.

11. J. Esztergalyos, J. Bertsch and M. Ilar, 1997,

“Performance of a Busbar Differential Protec-tion Based on EMTP Simulation and Digital System Tests”, 24th Annual Western Protec-tive Relay Conference, Spokane, Washington

12. Elmore W A, 1995, “Protective Relaying The-

ory and Applications”, ABB Power T&D 13. CIGRE Working Group 34.02, August 1995, "

Adaptive Protection and Control", CIGRE Fi-nal Report

14. IEEE, 1990, “Standard C37.91-1985” 15. IEEE, 1990, “Standard C37.97-1979”

Current Differential Relaying - Coping with Communications Channel Asymmetry 1

V I S I M P O S I O I B E R O A M E R I C A N O S O B R E P R O T E C C I O N D E S I S T E M A S E L E C T R I C O S D E P O T E N C I A

Monterrey, Nuevo Leon, Mexico – 17 – 20 de Noviembre 2002

Current Differential Relaying Coping with Communications Channel Asymmetry

Gustavo Brunello, M.Eng, P.Eng Ilia Voloh [email protected] [email protected]

GE Multilin

215 Anderson Av Markham, On., Canada, L6E 1B3

Phone 1 905 201 2009

ABSTRACT Recent years have seen an increased interest in the application of line differential relays to very long transmission lines by utilities all over the world. This renewed attention to this well-known principle is due mainly to the availability of digital diff relays operating over high-speed digital communication channels. Current differential relays compare the current entering and leaving the protected circuit. In the case of a line differential scheme these currents can be located far apart and the currents measured by a relay at one terminal have to be moved to the other terminal in order to make the comparison. To evaluate the differential function samples have to be taken at the same time on both terminals. This requires that the relay clocks be synchronized; any time difference between the relays clocks will translate into a differential current that may cause a relay to maloperate. When using digital line differential relays, many users utilize state of the art Sonet/SDH optical communications systems to move current values between line terminals multiplexed using. The performance of the communications channel is critical to the reliable operation of the current differential system. Most of these comm. systems have a ring configuration with self-healing capabilities; during normal operating conditions, line differential relays on the system communicate over the main path (usually the shortest between the relays). In this conditions the channels are said to be symmetric meaning that the transmission and receiving paths have equal propagation delays. In case of a failure of either the Tx or Rx channel, the failed signal is automatically re-routed by the communications system over the protective path (the remaining part of the ring). The new operating conditions have asymmetric channel delays with one signal having a short propagation delay and the other a much longer, with upto 10ms difference. This condition is also known as split path communication.


Synchronizing the relay clocks and calculating the propagation delay having symmetric channels and operating under normal conditions is relatively easy with techniques such as ping-pong. Line differential relays operating on a split path with asymmetrical communications channels present a more difficult task as, if not compensated for the different comm. delays, they will have their clocks not synchronized, samples not taken at the same time and as a result the differential function will maloperate for steady state conditions. This paper describes a GPS based mechanism to guarantee reliable operation of current differential relays under channel asymmetry conditions. Other topics also discussed in the paper include:

?? Types of digital communication used for line differential relays: direct optic fiber, multiplexers and Sonet/ SDH communications systems

?? Relay clock synchronization when the communication channel is symmetrical ?? The use of GPS/ Irig-B time signal for sample synchronization during

asymmetrical communication conditions ?? Fallback strategies for line differential relays under GPS/ communications

contingencies 1) INTRODUCTION The basic operating principle of current differential relaying is to calculate the difference between the currents entering and leaving the protected zone. The protection operates when the current difference exceed a set threshold. Line differential relaying requires that the information of the current flowing through each line terminal is made known to all other terminals. Earlier line differential relays used analogue communication channels, namely pilot wires to exchange the current values from one terminal to the other. The application of line differential scheme over pilot wires was limited to a maximum distance of about 8/10km due to several factors such as: pilot wire capacitance/ resistance, extraneous induced voltages (ground potential rise), etc. The arrival of digital communications allowed relay manufacturers to code the information exchanged between the terminals as zeros and ones. The first choice of media for digital differential relays was a direct fiber optic connection as it provides security and noise immunity (Fig 1). Synchronization of data sampling clocks is required in a digital differential protection because measurements must be made at the same time otherwise the protection scheme may maloperate. For best results, samples should be taken simultaneously at all terminals.


Fig 1 – Digital Line Differential Scheme Using Direct Fiber Optic

2) CURRENT DIFFERENTIAL SIGNALS CARRIED OVER OPTICAL MULTIPLEXERS With the advent of the digital communications revolution, utilities have seen the opportunity to improve the utilization of the installed fiber - taking advantage of its almost unlimited bandwidth – deploying optical multiplexers across the network to carry additional traffic from different applications in addition to the line differential signals. Optical multiplexers are communications equipment where similar digital signals (relays, voice, video, data, Ethernet) from several sources meet to be converted in an optical bit stream of higher rate. They are like a large river were small tributaries meet to construct larger ones and several of them confluence to become a river. There are 2 well established, similar standards for optical multiplexers: a) SONET (Synchronous Optical Networks) mainly used in North America; and, b) SDH (Synchronous Digital Hierarchy) as defined by the ITU (International Telecommunications Union) and used in the rest of the world. The introduction of this type of higher order communications systems has brought new challenges to the relay engineer. Normally, Sonet/ SDH optical networks have ring topologies (Fig 2) where nodes are connected by a pair of fibers; one for transmit (Tx) and the other for receive (Rx). The traffic between any 2 nodes on the network is usually carried over the shortest fiber path. There is an alternate (back up) communication path on the complement of the ring. Under healthy conditions relay at node 1 communicates with relay at node 3 via node 2 and viceversa. Thus, the communications delay for the transmit and receive paths are the same. Because of its ring topology, Sonet/ SDH networks have some unique features such as self-healing capabilities. Upon failure of either the Tx of Rx signal, utility grade optical multiplexers have a special built-in function called “switch on yellow” that automatically switches both Tx and Rx channels over the alternate comm. path achieving a new symmetrical operation condition. This switching takes place in a very short time (3ms), it is transparent to the line differential schemes and no further action or special algorithm is required in the relay to compensate asymmetry.


2.1) Split Path Communications - Channel Asymmetry Utilities with no fiber optic infrastructure lease digital channels from telecomm companies. The performance of these channels are critical to the reliable operation of the differential protection; however, there is no guarantee as to the path of data messages within the network and the communications time delays, as the links are no longer under the direct management and supervision of the power utility. This is also the case where the utility uses non-utility grade multiplexers as they generally do not support the “switch on yellow” function. Let’s assume that only the transmit channel from relay 1 to relay 3 fails. Under these conditions, a Sonet/ SDH based network will automatically switch the route; messages from relay 1 to relay 3 will travel via nodes 5, 4 and then to 3. However, the Tx delay increases significantly due to its longer path. .

Fiber Pair

Diff Relay

Primary path

Alternate path (in less than 3msec)

Multiplexer

Diff Relay

Multiplexer

MultiplexerMultiplexer

MultiplexerDamaged

Fiber

1

2

3

45

Fiber Pair

Diff Relay

Primary path


Multiplexer

Diff Relay

Multiplexer


MultiplexerDamaged

Fiber

Fiber Pair

Diff RelayDiff Relay

Primary path



Diff RelayDiff Relay


MultiplexerMultiplexerMultiplexerMultiplexer

MultiplexerMultiplexerDamaged

Fiber

1

2

3

45

Fig. 2 – Digital Differential Relays Communicating Over a Sonet/ SDH System The new configuration has now asymmetrical communication channel delays, as the time delay for the Tx signal is different from that of the Rx signal. This channel delay asymmetry can be as high as 8/10 ms depending of the routing and length of the alternate path. The channel asymmetry causes incorrect synchronization between relays, as ping-pong clock synchronization scheme operates based on the assumption that transmit-receive delays are the equal. The result is an fictitious differential current, proportional to the value of the channel asymmetry (Fig. 3). If currents and channel asymmetry are high enough the relay will miss-operate. For these split path situations a special mechanism


has to be built-in the relays to cope with the differences between the transmit-receive paths.

IDIFFIA IB

Half of the channel asymmetry in

electrical degrees

IDIFFIA IB

Half of the channel asymmetry in

electrical degrees Fig. 3 – Spurious Differential Current Introduced by Asymmetric Comm Channels

3) CHANNEL DELAY CALCULATIONS AND SAMPLE CLOCK SYNCHRONIZATION 3.1) Ping-Pong Mechanism for Channel Delay Calculation To calculate the differential current flow between line terminals when digital relays are used, it is necessary that the current samples at each terminal are taken at the same instant of time requiring synchronization of data sampling clocks. Synchronization errors manifest as phase angle and transient errors in phasor measurements at the terminals. By phase angle errors, we mean that identical currents produce phasors with different phase angles. By transient errors, we mean that when currents change at the same time, the effect is seen at different times at different measurement points. For best results, samples should be taken simultaneously at all terminals; this creates a challenge when data is taken at remote locations. Modern line differential relays approach to clock synchronization relies upon distributed synchronization. Distributed synchronization is accomplished by synchronizing the clocks to each other rather than to a master clock. Clocks are phase synchronized to each other and frequency synchronized to the power system frequency. Each relay compares the phase of its clock to the phase of the other clocks and compares the frequency of its clock to the power system frequency and makes appropriate adjustments. The frequency and phase tracking algorithm keeps the measurements at all relays within plus or minus 25 microsecond error during normal conditions for a 2 or 3 terminal relay system. These relays use a peer-to-peer architecture in which the relays at every terminal are identical. The peer to peer architecture is based on two main concepts that reduce the dependence of the system on the communication channels: replication of protection and distributed synchronization. Each relay computes differential current and clocks are synchronized to each other in a distributed fashion.


The most reliable technique for clock synchronization uses a variation of the "ping-pong" algorithm in which clock error is computed from 4 time measurements collected during a round trip pair of messages (Fig. 4). Once the nodes are synchronized the "ping-pong" mechanism keeps exchanging round trip messages between nodes on the network to maintain synchronism. The algorithm assumes that the communications channel delay is the same in each direction. If it is not, the timing error is equal to one half of the difference between the delays in each direction.

Tx

Rx

Rx

TxL90#1

L90#2

CommunicationsChannel

t0

t3

t2

t1

tf

tr

tf = forward travel time

tr = return travel time

Current differential requires both relays to be in time synchronization for differential calculations.

tf = tr = t3 - t0 - (t2 - t1)2

“Ping-Pong” Synchronization Technique:

Tx

Rx

Rx

TxL90#1

L90#2

CommunicationsChannel

t0

t3

t2

t1

tf

tr

tf = forward travel time

tr = return travel time

Current differential requires both relays to be in time synchronization for differential calculations.

tf = tr = t3 - t0 - (t2 - t1)2


tf = tr = t3 - t0 - (t2 - t1)2


Fig, 4 – Calculating Channel Delay Using Ping-Pong The basic approach to frequency locking is to compute frequency deviation from the apparent rotation of phasors in the complex plane, and adjust the sampling frequency accordingly. The rotational rate of phasors is equal to the difference between the power system frequency and the ratio of the sampling frequency divided by the number of samples per cycle. This difference is used to correct the sampling clocks to synchronize the sampling with the power system frequency. The amount of time synchronization error depends on the accuracy of the local clocks, how often the "ping-pong" is executed, channel delay, and other factors. The "ping-pong" must be executed often enough to compensate for the drift in the local clocks, which are moderately accurate crystal clocks. A small amount of channel delay itself is not critical, mainly affecting only system transient response, provided the channel delay is the same in each direction. If it is not the same, the difference between the delays causes a differential error between the clocks being synchronized over the channels. The ping-pong mechanism is well documented in the literature and is used for internet time synchronization and in several commercially available systems.


3.2) Using GPS Technology to Cope with Communications Channel Asymmetry Modern differential relays have the their sampling clocks synchronized to each other in phase and to the power system in frequency driven by hybrid phase-frequency locked loops with inputs from phase deviation (relay sampling clock) and system frequency measured from local and remote currents. Synchronizing the phase of the sampling clocks to each other reduces the residual fictitious differential current that arises from phase errors in the clocks. Synchronizing the frequency of the sampling clocks to the power system improves the rejection of non-fundamental frequency components that would otherwise leak into the measurement of fundamental frequency voltages and currents eliminating the error effects of asynchronous sampling. The phase locked loop is a simple, linear, proportional plus integral controller that drives both frequency deviation and phase deviation to zero. The weight of phase deviation is larger than frequency deviation in the controller, meaning that phase deviation has a higher priority and is suppressed much faster than frequency deviation. The phase deviation between each relay is computed from 4 time stamps collected during an exchange of messages between relays, consisting of the time the messages were sent and received according to the local and remote clocks. Upon the occurrence of a channel asymmetry condition, if no special means to detect it is provided, the sampling clocks will start to drift apart at a speed defined by the value of asymmetry and coefficients of the phase and frequency locked loop and as a consequence maloperation may occur. The higher the channel asymmetry, the higher the apparent differential current appearing on the relay causing conventional line current differential system to misoperate. Typically, depending on the 87L settings, conventional line current differential system can tolerate up to 1.5-2.5 ms of asymmetry. Above that value it becomes dangerous to operate 87L relays without a dedicated algorithm to cope with split path conditions. The GPS time reference delivered to the relay as an IRIG_B signal is used to compute and compensate for channel asymmetry. A number of considerations made GPS the system of choice: its very high degree of available, reliability and accuracy. It was designed to provide navigation throughout the world without failure under the worst conceivable scenarios, including all weather, system degradation, and interference conditions. Modern digital relays are compatible with most of GPS clock receivers available on the market providing IRIG-B signal. There are 2 clocks at each relay: the local sampling clock and a local GPS clock, the latter embedded in the relay box (Fig.5). Each of these clocks uses an independent “ping pong” mechanism to exchange time messages with the remote relay. The sampling clocks provide synchronous and accurate data sampling for all relays on the system. The local GPS clocks are synchronized to an external time reference: the Global Positioning System (GPS) absolute time using an external a GPS receiver.


Fig. 5 – Sampling Clock Synchronization Using GPS time reference

As indicated above, the sampling clock deviation is computed from 4 time stamps collected from local and remote sampling clock messages using the ping-pong technique. The computed deviation is equal to the actual deviation plus half of the channel asymmetry, as shown is Fig 4. To measure and compute channel asymmetry, 4 additional time stamps taken from the local GPS clock are exchanged. The deviation, computed from the GPS time stamps is equal to half of the channel asymmetry, which is then subtracted from the computed sampling clock deviation. Therefore, the computed value of asymmetry compensates for the error caused on the sampling clocks by comm. channel asymmetry, preventing them to drift apart under split path conditions. Asymmetry is subtracted prior to the phase locked loop providing perfect tracking for the communications transients. The resulting clock deviation is then used to drive the phase locked loop. To note is that this approach does not time stamp the messages carrying the power system current values but rather uses two different Ping-Pong algorithms: a) for the sample clock synchronization; and, b) for the GPS phase deviation. This approach allows for a very flexible strategy to fall back modes in case of contingencies in the GPS receiver or communication channel operation mode. The following example will illustrate the principle outlined above. Let’s assume that prior to channel asymmetry conditions the system was synchronized with the Tx delay equal to the Rx delay and equal to 5ms. Due to switching of communications paths, an


asymmetry was introduced resulting in that Tx delay is 8ms and Rx delay equal to 5ms for relay 1. The sampling clock of relay 1 computes the clock deviation of Dev_sampling for relay 1 equal to (8-5)/2= +1.5ms while relay 2 computes the value of Dev_sampling2 as (5-8)/2= -1.5ms. Without compensation, the relay 1 estimates that it is ahead of relay 2 clock and will try to slow down. In the meantime, relay 2 estimates that it's behind relay 1 clock and will try to move ahead. However, using GPS signal, relay 1 computes that asymmetry of Dev_GPS1=(5-8)/2= -1.5ms and relay 2 computes that asymmetry of Dev_GPS2=(8-5)/2= +1.5ms. The total deviation is equal to Dev_total=Dev_sampling minus Dev_GPS which is zero at both relays thus preventing the sampling clocks to drift. Precise synchronization of all sampling clocks provides for best transient response both for communication channel impairments and GPS clock anomalies. Channel asymmetry is calculated exactly and immediately after it changes in any direction. The channel asymmetry is measured directly and can be presented on the relay display. There is no need for a dedicated GPS source for the channel asymmetry algorithm/ line differential element and the regular IRIG-B input of the relay provides both channel asymmetry compensation and event time stamping. Modern line differential relays have the conventional line current differential element and the mechanism to cope with channel asymmetry in one box. This approach allows the user to apply the same relay on a line differential system over direct fiber optic at the beginning and later use the same relay over multiplexed channels (SONET/SDH rings) with potential channel asymmetry conditions. Further, the customer can dynamically switch asymmetry compensation on and off and continue to run 87L to suit a variety of protection philosophies. 4) MODE OF OPERATION UNDER CONTINGENCIES The addition of an algorithm that rely on an external source to handle the asymmetry in communication channels incorporates a new potential point of failure that has to be addressed. The external signal source includes the GPS receiver that converts the GPS signal received from the satellite into an IRIG_B signal to the relay. There are 2 main points of failure in the GPS system: loss of the GPS signal and failure of the GPS receiver. Other failure mode to confront, although very improbable, is the complete loss of the communication channel. Loss of the GPS signal from the satellite is extremely unlikely; however, the failure of the GPS receiver antenna is possible. In this event, the time reference for the GPS ping-pong mechanism can be kept running for a long time as GPS receivers are built with high quality, very stable clock crystals. Channel asymmetry can be permanently monitored in the relay and an indication signaled if channel asymmetry exceeds preset values. Modern line differential relays allow several mechanisms/ strategies to handle contingencies on the GPS or communications systems and a few of them are reviewed.

Block GPS time reference: Digital relays have a setting to indicate that the time reference is not valid. The time reference may be not accurate due to problems with the GSP receiver. User has to be


aware when GPS satellite receiver loses signal from the satellite and reverts to it is own calibrated crystal oscillator. When the satellite signal is not available, time accuracy degrades in time and eventually might cause the relay to misoperate. It is recommended to verify with the GPS receiver manufacturer for how long the receiver maintains the accuracy upon loss of the satellite signal. In addition, the GPS receiver shall have an alarm contact indicating loss of the satellite signal. When the accuracy of the time reference is not guaranteed any longer, a hard wired signal from the receiver output contact shall be connected to the relay via contact input and the GPS compensation should be effectively blocked. Typically this setting will be a signal from the GPS receiver signaling problems or time inaccuracy.

Maximum Channel Asymmetry Relays have a setting (in ms) for the max asymmetry expected and used to detect when it is exceeded. The output operand can be used to alarm problems with communication equipment, and/or to decide whether the channel asymmetry compensation shall remain in operation should the GPS-based time reference be lost. Channel asymmetry is measured if both terminals of a given channel have a valid time reference

Round Trip Time Change This setting is used to detect changes in round trip time. The function has an output operand that can be used to alarm on problems with communication equipment, and/or to decide whether the channel asymmetry compensation shall remain operation should the GPS-based time reference be lost. Fall back strategy 1 On the loss of the GPS/ Irig-B signal at any terminal block 87L after certain time. This is a simple, but conservative way of using GPS feature. It is recommended to block the 87L after a time set not higher than 10 seconds. This time allows for a window of opportunity for the return of the GPS signal if the loss was temporary meanwhile not jeopardizing the security of the system

Fall back strategy 2 Upon loss of the GPS signal at any terminal the line differential scheme continues to operate the 87L element until a sadden change in the channel round-trip delay is detected. If GPS is enabled at all terminals and the GPS signal is present, the differential relays compensate channel asymmetry. On the loss of the GPS signal, the differential relays have stored the last measured value of the channel asymmetry and compensates for the asymmetry until the GPS/ Irig-B clock signal is reestablished. This is possible due to the GPS receiver clock crystal stability. However, if during the lack of GPS/ Irig-B signal the channel asymmetry changes because it is switched to another physical path, the 87L element must be blocked, since the channel asymmetry cannot be measured and system is no longer accurately synchronized. The value of the step change in the channel asymmetry can be set in modern line differential relays (see Round Trip Time Change above). 5) CONCLUSIONS The introduction of multiplexed digital communications enabled utilities to apply current differential relaying schemes to long transmission lines. Sonet/ SDH communication networks with its self healing capabilities can operate in split path producing fictitious current differential that may cause maloperation. A dedicated GPS based mechanism


has been presented that handled communications asymmetric delays. Failure modes and contingent operation conditions has been analyzed. Biographies Gustavo Brunello received his Engineering Degree from National University in Argentina and a Master in Engineering from University of Toronto. After graduation he worked for the National Electrical Power Board in Argentina where he participated in the designed and commissioning of the 500 kV transmission system. For several years he worked with ABB Relays and Network Control both in Canada and Italy where he became Engineering Manager for protection and control systems. In 1999, he joined GE Power Management as an application engineer where he is responsible for the application and design of protective relays and control systems. Gustavo is a Professional Engineer of the Province of Ontario and a member of the IEEE. Ilia Voloh graduated from Ivanovo State Power University, Soviet Union. He then was for many years with Moldova Power Company in various roles in the Protection and Control area. Currently Ilia is a Protection and Control Advisor with GE Power Management. His areas of interest are current differential and phase comparison line protection, and communications for protective relaying. Ilia is a Member of IEEE.

V I S I M P O S I O I B E R O A M E R I C A N O S O B R E P R O T E C C I O N D ES I S T E M A S E L E C T R I C O S D E P O T E N C I A

DISTANCE PROTECTION OF SERIES COMPENSATED LINES –PROBLEMS AND SOLUTIONS


Bogdan [email protected]

GE Power Management215 Anderson Avenue

Markham, OntarioCanada L6E 1B3

Gustavo [email protected]



Distance Protection of Series-Compensated Lines – Problems and Solutions

Page 2 of 35

1. Introduction

Increased transmittable power, improved system stability, reduced transmissionlosses, enhanced voltage control and more flexible power flow control are technical rea-sons behind installing Series Capacitors (SCs) on long transmission lines. Environmentalconcerns and direct cost benefits stand for that too.

SCs and their overvoltage protection devices (typically Metal Oxide Varistors, MOVsand/or air gaps), when installed on a transmission line, create, however, several problemsfor its protective relays and fault locators. Operating conditions for protective relays be-come unfavorable and include such phenomena as voltage and/or current inversion, sub-harmonic oscillations, and additional transients caused by the air gaps triggered by ther-mal protection of the MOVs. Overreaching of distance elements due to series compensa-tion is probably the most critical and known consequence of SCs. The opposite may hap-pen as well: a distance function may fail to pick up a low-current fault on the protectedline.

This paper focuses on several phenomena specific to series compensated lines. In-sights of these singularities help to understand limitations of distance relays put in opera-tion on series compensated lines and to apply the relays more efficiently.

The analysis part of the paper starts with Section 2 where the equivalent phase andsymmetrical impedances of the SCs and MOVs are examined. The problem of a zero-sequence compensating factor is addressed as well.

Section 3 analyses a low-current single-line-to-ground (SLG) fault in a sample seriescompensated system. The phenomena of voltage and current inversions are explained andillustrated using this sample system. Consequences of several combinations of inversionof various signals on typical distance comparators are presented. The signals includephase, positive-, negative- and zero-sequence voltages and currents. The comparatorsanalyzed are memory-polarized mho, zero-sequence polarized reactance, memory-polarized negative- and zero-sequence directional.

Section 4 addresses transients associated with series compensation.Section 5 presents a solution that ensures directional integrity of protection on series

compensated lines and in a vicinity of series compensated lines: an offset impedance fornegative- and zero-sequence directional elements. Setting calculation rules for the offsetimpedance are provided.

Section 6 presents a concept of a current-controlled adaptive reach for distance pro-tection as a solution to an overreaching problem.

Section 7 discusses phase selection problems for single-pole tripping on series com-pensated lines.

Section 8 focuses on fault location in series compensated lines.


Page 3 of 35

2. Series Capacitors and MOVs – Equivalent Impedance

Vast majority of microprocessor-based distance relays respond to more (security) orless (speed) accurately filtered fundamental frequency components. Therefore, it be-comes important to understand relations between the fundamental frequency voltage andcurrent of a typical arrangement of SCs and their overvoltage protection devices.

Three single-phase banks of capacitors are used for series compensation. Each ca-pacitor must be protected against overvoltages by air gaps or Metal Oxide Varistors(MOVs) or both. Under load conditions or low-current faults, the voltage drop across theSCs is below the voltage protection level: neither the air gaps nor the MOVs conduct anycurrent. Therefore, the SC bank is equivalent to a pure reactance equal the reactance ofthe actual (physical) capacitor. Under very high current faults the voltage drop would befar above the protection level: the gaps and/or MOVs conduct majority of the throughcurrent, practically by-passing the SCs. Therefore, for large through currents the SC bankis equivalent to a small resistance.

Between the two extremes there are situations when a comparable amount of currentflows through the SCs and the MOVs. Fig.1 illustrates such a case. As the through cur-rent becomes larger, the voltage drop across the bank (Fig.1a) assumes more rectangularshape, being limited to the voltage protection level. The capacitors conduct the currentduring initial half-cycles (Fig.1b), while the MOVs conduct during the remaining halves(Fig.1c). The through current being a sum of the two is not distorted as compared with itstwo contributors, and is shifted in a leading direction with respect to the voltage dropacross the bank. Relation between the fundamental frequency components of the voltagedrop across the bank and the through current is a resistive-capacitive impedance knownas a Goldsworhty’s equivalent [1].

0 100 200 300 400 500 600 700 800-2

-1.5

-1

-0.5

0

0.5

1

1.5

2x 105

post-fault time [ms]

[V]

Voltage Drop Across MOV : vMOV(a)

0 100 200 300 400 500 600 700 800-8000

-6000

-4000

-2000

0

2000

4000

6000

8000


[A]

Current in Series Capacitor : iC

0 100 200 300 400 500 600 700 800-8000

-6000

-4000

-2000

0

2000

4000

6000

8000


[A]

Current in MOV : iMOV(b) (c)

Figure 1. Voltage drop across a series capacitor with a conducting MOV (a);capacitor (b) and MOV (c) currents, respectively.

2.1. Equivalent Phase ImpedanceConsider a parallel arrangement of an ideal capacitor (SC) and a non-linear resistor

(MOV) shown in Fig.2a. Approximation of the MOV characteristic – accurate enough forprotective relaying analysis – is given by the following equation:


Page 4 of 35

q

REFVvPi !!

"

#$$%

&⋅= (1)

where P and VREF are coordinates of the knee-point and q is an exponent of the charac-teristic (Fig.2c shows a sample MOV characteristic).

For any operating condition of the parallel connection of Fig.2a one may calculateanalytically, simulate using a transient program, or measure the fundamental componentsof the voltage across the bank and the through current. The ratio of such voltage and cur-rent phasors is an equivalent impedance (Fig.2b). It is obvious that the equivalent resis-tance and reactance are dependent on the through current (Fig.2d). For currents producingthe voltage drop below the voltage protection level (2.4kA in this example), the resis-tance is zero and the reactance equals the actual reactance of the SC (65 ohms in this ex-ample). For higher currents (3-4kA in this example) the resistance increases while the re-actance consistently decreases. For very high currents, the reactance approaches zero sodoes the resistance.

0 20 40 60 80 100 120 140 160iMOV/P

vv /VREF

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0

40

-40

0

-80|Iv|

Rv_ana

Rv_EMTP

Xv_ana Xv_EMTP

2000 4000 6000 8000p.u.

(d)

(b)

IMOV

VV

XV(|IV|)RV(|IV|)

SC

MOV

IV

VV(a)

(c)Ω

A

p.u.

IV

Figure 2. Series capacitors with MOVs (a); equivalent fundamental frequency impedance (b); sample MOVcharacteristic (c); sample current-dependent parameters of the equivalent impedance.

Fig.2d presents the equivalent resistance and reactance derived using two methods:First, the EMTP simulations in a sample system have been performed and the voltage

and current waveforms with their natural distortions have been recorded. Second, theFourier Transform has been used to calculate the phasors and derive the impedance. Theprocedure has been repeated for various levels of the through current, resulting in a char-acteristic.

Second, a sinusoidal voltage drop has been assumed, while the through current, andsubsequently the impedance has been calculated analytically.

The differences visible in Fig.2d result from various assumptions to calculating theGoldsworthy’s model. Generally, the EMTP-type equivalent is more adequate for protec-tive relaying studies, particularly if the phasors were derived using the actual filteringtechniques used by a protective relay under consideration.


Page 5 of 35

The concept of an equivalent impedance allows grasping the basics of the distanceoverreaching phenomenon. If the SCs are located between the fault and the relay poten-tial point, the fault loop contains the line-to-fault impedance, fault resistance (if any) andthe equivalent SC&MOV impedance. The latter being resistive-capacitive shifts the ap-parent impedance down and to the right as shown in Fig.3.

R

X far-end busbar (B)

local busbar (A)

RF

ZSC&MOV

Zseen

Zactual

high-currentfaults

low-currentfaults

Figure 3. Distance element overreaching due to series compensation.

Overreach is the primary consequence of the situation depicted in Fig.3. In the worstcase – for low-current faults – the equivalent SC&MOV impedance is a pure reactanceshifting the apparent impedance down by the entire reactance of the physical capacitors.As the lines are typically compensated at the 50-70% rate, the overreach may be as highas 50-70%. For high-current faults, though, the equivalent SC&MOV impedance shiftsthe apparent impedance only slightly to the right. There is no danger of overreaching.This observation leads to a concept of a current-controlled adaptive reach described inSection 6.

During medium-current faults on the line, the apparent impedance may be shifted tothe right by more than half the reactance of the capacitors. This relocation may be highenough to push the apparent impedance outside the operating characteristic, particularlyif a lens, conservatively set blinders, or load encroachment characteristics are used.

Another observation that can be derived from this simplified model is a failure of adistance function to respond to a low-current close-in fault. Under such a fault, the appar-ent impedance moves to the fourth quadrant of the impedance plane resulting in problemswith directional discrimination.

The SC&MOV bank acts as a “fault current stabilizer”: for larger currents the capaci-tive reactance is smaller while the resistance is larger – this reduces the current as com-pared with a fully compensated circuit; for smaller currents the capacitive reactance islarger – this reduces the net impedance and increases the current as compared with a non-compensated circuit. As a result, the fault current versus fault location characteristic isflatter for series-compensated lines comparing with non-compensated lines.

The Goldsworthy’s model is useful in understanding some phenomena, but its practi-cal applications – other than fault studies using numerical packages – are very difficult.


Page 6 of 35

First, the equivalent resistance and reactance are non-linear functions of the throughcurrent. This non-linearity makes all the analytical calculations practically impossible.

Second, the through current depends on a number of factors such as system condi-tions, fault location, fault type and fault resistance.

Third, both traditional power system analysis and principles of protective relaying arebased on sequence networks and symmetrical components rather than phase componentsand physical three-phase networks. For example, the reactance characteristic is typicallypolarized from the zero-sequence or negative-sequence currents for increased security onheavily loaded lines. The impact of series compensation on the angle of such polarizingsignal is equally important as the impact on the phase current or voltage.

For practical outcome the sequence networks of the SC bank must be analyzed talk-ing into account both asymmetries: the parallel asymmetry caused by a fault must be con-sidered together with the series asymmetry caused by various operating conditions of theSCs and MOVs in particular phases of the three-phase compensating bank. The next sub-section addresses this issue in more detail.

2.2. Equivalent Sequence ImpedancesIn general, voltage drops across a linear three-phase element are proportional to

through currents. The proportionality factor is called an impedance. For three-phase sys-tems this could be written in a compact matrix form as follows:

!!!

"

#

$$$

%

&

⋅!!!

"

#

$$$

%

&

=!!!

"

#

$$$

%

&

C

B

A

CCCBCA

BCBBBA

ACABAA

C

B

A

III

ZZZZZZZZZ

VVV

(2)

Or for the symmetrical components:

!!!

"

#

$$$

%

&

⋅!!!

"

#

$$$

%

&

=!!!

"

#

$$$

%

&

2

1

0

222120

121110

020100

2

1

0

III

ZZZZZZZZZ

VVV

(3)

Under normal circumstances, the equivalent Z012 impedance matrix becomes diago-nal. This means that there is no mutual coupling between the symmetrical components orsequence networks:

!!!

"

#

$$$

%

&

⋅!!!

"

#

$$$

%

&

=!!!

"

#

$$$

%

&

2

1

0

2

1

0

2

1

0

000000

III

ZZ

Z

VVV

(4)

The resulting decoupling of the sequence networks is the prime enabler for the tradi-tional short circuit calculations.

This is not, however, the case for SC and MOVs: the phase impedances are not cou-pled, while the sequence impedances may get coupled if the series asymmetry occurs, i.e.if the Goldsworthy’s impedances differ between the phases.


Page 7 of 35

The following applies for the banks of three SCs and MOVs:

!!!

"

#

$$$

%

&

⋅!!!

"

#

$$$

%

&

=!!!

"

#

$$$

%

&

C

B

A

C

B

A

C

B

A

III

ZZ

Z

VVV

000000

(5)

where ZA is a Goldsworthy’s equivalent for phase A (dependent on the magnitude of thephase A current), ZB is an equivalent for phase B, and ZC is an equivalent for phase C.

Talking into account the relations between the phase and symmetrical currents andvoltages, the following Z012 impedance matrix is calculated from equation (5) for thethree banks of SCs and MOVs:

!!!

"

#

$$$

%

&

++++++

++++++

++++++

=

CBACBACBA

CBACBACBA

CBACBACBA

ZZZZaaZZaZZaZaZZaZZZZZaaZZZaaZZaZZaZZZZ

Z22

22

22

012 31 (6)

where a is a 120deg shift operator as per theory of symmetrical components.In general, the three impedances A, B and C are different. This results in the off-

diagonal elements of the matrix (6). Electrically, it means that there are mutual couplingsbetween the sequence networks representing the SC bank: the positive sequence voltagedepends on the negative- and zero-sequence currents, for example.

Let us consider only the extreme cases of either very high or very low currents forvarious fault types.Low-current faults and load conditions (none of the MOVs conduct)

SCCBA jXZZZ −=== (7a)

Using (6) one calculates:

!!!

"

#

$$$

%

&

−

−

−

=

SC

SC

SC

jXjX

jXZ

000000

012 (7b)

For this case equation (7b) means that there is no mutual coupling between the se-quence networks. The zero-, positive- and negative-sequence impedances of the SC bankequal the physical reactance of the capacitors.High-current SLG faults (phase A MOV conducts all the current)

SCCBA jXZZZ −=== ,0 (8a)


!!!

"

#

$$$

%

&

−

−

−

=

SCSCSC

SCSCSC

SCSCSC

jXjXjXjXjXjXjXjXjX

Z2

22

31

012 (8b)


Page 8 of 35

For this case equation (8b) means that there is a mutual coupling between the se-quence networks. The self impedances (00,11,22) are capacitive and equal 2/3rd of thephysical reactance of the capacitors. The mutual impedances (01,02,12) are inductive andequal 1/3rd of the capacitor reactance. The latter fact means for example that the positive-sequence current causes a leading voltage drop for the zero-sequence voltage.High-current LL and LLG faults (phase A and B MOVs conduct all the current)

SCCBA jXZZZ −=== ,0 (9a)


!!!

"

#

$$$

%

&

−−−

−−−

−−−

=

SCSCSC

SCSCSC

SCSCSC

jXjXjXjXjXjXjXjXjX

Z31

012 (9b)

For this case equation (9b) means that there is a mutual coupling between the se-quence networks due to the series asymmetry.

Equations (6) through (9) flag a complex condition: mathematically the impedancematrices are not diagonal, or electrically there are mutual couplings between the sequencenetworks. The situation, however, is not that complex when considering the following:

First, for high-current SLG fault, the faulted phase SC is by-passed and the voltagedrop across that phase is small as compared with the voltage drop across the line. Thevoltage drops in the two healthy phases are small as well despite the series capacitors, be-cause the currents in those phases are very small as compared with the current in thefaulted phase. For fault at the far end busbar being a concern as far as the overreaching isconsidered, the voltage drops across the compensating bank in all three phases may bedisregarded comparing with the voltage drops across the line (Fig.4a).

This means that the sequence impedances of the SC and MOVs (Z0, Z1 and Z2) maybe assumed zero for the first approximation analysis of high-current SLG faults.

The same result is obtained using equation (8b) and assuming I1=I2=I0 (relation be-tween the symmetrical currents for a SLG fault).

Second, the LL and LLG faults are very unlikely to be high-resistance faults. It isjustified to assume that the SC will be effectively by-passed in the faulted phases. Thehealthy phase will conduct much lower current. Consequently, the voltage drops in thefaulted phases will be small as compared with the voltage drop along the line because ofthe low Goldsworthy’s equivalent impedance in those phases. The voltage drop in thehealthy phase is low as well despite the SC being in place because the current in thatphase is low.

Similarly to the previous case (Fig.4b) the sequence impedances of the SC and MOVs(Z0, Z1 and Z2) may be assumed zero for the first approximation analysis of high-currentLL, LLG and 3P faults.

The same result is obtained using equation (9b) and assuming I1=–I2.


Page 9 of 35

Third, the low-current faults may happen either in very weak systems (large equiva-lent impedance of the systems) and/or during high-resistance faults. The first alternativedoes not impose much danger on protective relaying, as the series capacitors will besmall compared to inductive system reactances. The effective compensation level will besmall and the major problems associated with series compensation will not occur. Thesecond alternative calls for a high fault resistance. Practically, however, high fault resis-tance could be encountered only for SLG faults.

Summarizing: for all the high-current faults it is justified to neglect all three banks ofSCs and MOVs regardless of the fault type. For lower current levels the effective seriesreactance is smaller as compared with the reactance of the physical capacitors. The entirephysical reactance (the one causing the maximum overreach) shall be considered only forhigh-resistance SLG faults. Under the latter condition, however, the analysis is simple asthe symmetrical networks are decoupled and Z0=Z1=Z2=–jXSC.

LINEA

B

C

A

B

C

SCs & MOVsIA high

VC = 0

IB = 0

IC = 0

VB = 0

VA = 0

(a)

LINEA

B

C

A

B

C

SCs & MOVsIA high

VC = 0

IB high

IC = 0

VA = 0

(b)

VB = 0

Figure 4. Voltage drops across the SC banks during high-current faults: SLG faults (a) and LL/LLGfaults (b).

2.3. Zero-Sequence Compensating FactorThe zero-sequence compensating factor as a setting may be defined differently by

various vendors, but as per principle of ground distance relaying the proper compensation(accounting for the difference between the positive- and zero-sequence impedances) callsfor the following operation:

Current for the 21G function, phase A: ( ) !!"

#$$%

& −⋅+++=

1

10

31

ZZZIIIII CBAAAG (10)

The (Z0-Z1)/Z1 is a complex number and depends on the ratio between the zero- andpositive-sequence impedances from the potential point of the relay to the fault. Normally,the protected circuits are homogenous or only slightly non-homogenous: the Z0/Z1 ratio isconstant along the line or changes only slightly depending on the fault position. Typi-cally, distance relays provide for a separate K0 setting for the first ground zone and allthe remaining zones.

In the case of series compensated lines, if the series capacitors are between the faultand the potential point of the relay, the Z0/Z1 ratio may be drastically affected by the faultposition and fault current level.

In general, for medium current faults, when the Goldsworthy’s equivalent impedancesare neither zeros nor pure reactances, mutual coupling between the sequence voltages andcurrents occurs, and the principle of traditional ground distance protection is violated. In


Page 10 of 35

other words, the accurate loop current for phase A of ground protection is not given bythe classical equation (10). The apparent positive sequence impedance still can be – atleast theoretically – calculated by the relay but the equations become non-linear and im-practical. Some fault location algorithms for series compensated lines, such as the algo-rithm [2], take into account the current-dependent Goldsworthy’s parameters. For dis-tance relaying, however, the approach is not practical.

Majority of the protection engineers would probably use the Z0/Z1 data of a plain line,or the “K0” factor calculated by a short-circuit program for a fault at the intended reachpoint. The latter may or may not include the series capacitors. If included, the capacitorsare treated as for low-current faults, i.e. ZA=ZB=ZC=Z0=Z1=Z2=–jXSC.

Consider a low-current SLG fault. The accurate Z0/Z1 ratio shall reflect the imped-ances from the potential point to the fault. For example, if the SCs are located in the sub-station, the VTs are on the bus side of the SCs, and the fault location is α from the sub-station with the SCs, the following applies:

11

0

1

10 −−⋅

−⋅=

−

SCLINE

SCLINE

jXZjXZ

ZZZ

αα (11)

Surprisingly, the accurate value of the zero-sequence compensation setting would de-pend on the fault location. None of the available relays provide for such as a feature. Dueto other factors limiting the reach accuracy of the ground distance protection, dependencylike the one given by equation (11) is not considered practical.

Fig.5 presents a plot of the (Z0-Z1)/Z1 number for the case given by equation (11). Forfaults just beyond SCs, the fault loop consists of the SCs alone. Because Z0=Z1 for theSCs under low-current faults, no compensation is required because (Z0-Z1)/Z1 = 0. Forfaults further into the line, some compensation is required, but the sign of it is opposite tothe one for uncompensated lines. For faults at about 70% (in this case), the Z1 value isvery small (the capacitive and inductive reactances cancel mutually, and Z1 equals thecircuit resistance), and consequently the (Z0-Z1)/Z1 number is very large. As the faultmoves away from the relaying point, the (Z0-Z1)/Z1 number converges asymptotically onthe line value.

In this example, when using the (Z0-Z1)/Z1 as for the plain line, the relay would becompensated in a wrong direction (the accurate (Z0-Z1)/Z1 number is negative – the angleis close to 180 degrees). The relay would be overcompensated for faults up to some 30%,and significantly undercompensated for faults close to 70% of the line length.

For plain lines it is quite straightforward to evaluate the consequences of over- andunder-compensation: overcompensation increases the loop current, thus reduces the ap-parent impedance and causes overreach; undercompensation takes the opposite effect.

For series compensated lines, the consequences are not that straightforward as the re-lation between the sequence currents may be quite different due to signal inversions.


Page 11 of 35

0 10 20 30 40 50 60 70 80 90 1000

5

10

15

20

25

Fault Location, %

(Z0-

Z1)/Z

1

0 10 20 30 40 50 60 70 80 90 100-50

0

50

100

150

200

Fault Location, %

(Z0-

Z1)/Z

1,de

g

SC+LINE

SC+LINE

LINE

LINE

Figure 5. Zero-sequence compensating factor for a relay looking through SCs (bus potential source) at alow-current SLG fault as a function of fault location.

Consequently, the plain line zero-sequence compensating factor is probably a reason-able practical choice. The application of setting is limited anyway as the relays allow theuser to enter a fixed number.

If there are no SCs between the intended reach and the relay potential point, the zero-sequence compensating factor is not affected. For external faults, however, the requiredcompensating factors may be quite different as compared with the value for the protectedline. Consequently, the practice to reduce the reach of ground distance elements (as com-pared with phase elements) is even more valid for series compensation applications.

3. Low-Current SLG faults

As indicated in the previous section, a low-current SLG fault is the type of situationthat could cause real problems for distance protection of series compensated lines.

If the low-current fault results from weak systems, the capacitive reactances are can-celled with a very large margin by source inductive reactances and problems related toseries compensation do not manifest themselves. A high-resistance fault in a relativelystrong system is thus the event to consider.

This section analyzes in detail signals at the relaying points of a sample system. Sub-sequently, response of traditional distance comparators is studied for the same system.Although using a single system, this section illustrates in a practical way several particu-larities of distance applications on series compensated lines.


Page 12 of 35

3.1. Protection Signals in a Sample SystemConsider the system shown in Fig.6. In order to analyze faults on the protected line

and its terminals, the system can be equivalenced as shown in Fig.7.

Protected Line

Z1=

15oh

m, Z

0=

45oh

m

Z1=

8oh

m,Z

0=

25oh

m

Z1 = 2 ohm, Z0 = 4 ohm

Z1 = 2 ohm, Z0 = 3 ohm Z1 = 1 ohm, Z0 = 1.5 ohm

Z1 = 15 ohm

Z0 = 35 ohm

-12 ohm-7 ohm

Z1 = 10 ohm, Z0 = 30 ohm

-5oh

m

Figure 6. Sample system with series-compensated lines (secondary ohms).

Z1 = 1.43 ohm, 84.2 degZ0 = 2.68 ohm, 82.6 deg

-7 ohmZ1 = 10 ohm, 84.3 degZ0 = 30 ohm, 80.0 deg

Z1 = 3.44 ohm, 71.5 degZ0 = 16.4 ohm, 82.5 deg

fault location (k)

Left

Term

inal

Righ

tTer

min

al

Figure 7. Equivalent of the sample system for internal faults.

Assume a SLG fault on the line at the distance k [pu] from the left terminal. As thefault is a low-current fault, the MOVs do not conduct, the sequence networks for the SCsare decoupled and the equivalent diagram for the short-circuit calculations is shown inFig.8. All the signals of interest can be easily calculated from this network. In the fol-lowing examples a 10-ohm (secondary) fault resistance is assumed.

Fig.9 shows the magnitude and phase angle (all phase angles referenced to the phaseA pre-fault voltage at the left terminal). For faults closer than some 35% of the line, thefault current is capacitive. This results from the fact that the sum of the zero-, positive-and negative-sequence impedances seen from the fault point is capacitive. For a fault atsome 35% of the line, the inductive and capacitive reactances cancel and the fault currentis purely resistive limited only by the total resistance of the fault loop of Fig.8. Conse-quently, the fault current is maximal for this fault.


Page 13 of 35

In a sense fault current “inversion” happens for faults closer than 35%. However, theterm “inversion” is not precise: what happens in series compensated networks is rather an“unnatural” phase shift with respect to the “natural” position of a given phasor. A 180-degree “inversion” is practically never happens.

Z0(ri

ght)

(1-k)*Z0Lk*Z0L I0 (right)

Z0(le

ft)

I0 (left)

V0F

V0(li

ne)

V0(b

us)

V0(ri

ght)

Z2(ri

ght)

(1-k)*Z2Lk*Z2L I2 (right)

Z2(le

ft)

I2 (left)

V2F

V2(li

ne)

V2(b

us)

V2(ri

ght)

Z1(ri

ght)

(1-k)*Z1Lk*Z1L I1F (right)

Z1(le

ft)

I1F (left)

V1(li

ne)

V1(b

us)

V1(ri

ght)

VTh

even

in

3*RF

IF

Figure 8. Sequence networks and symmetrical components for low-current SLG faults.

Fig.10 shows the currents in the faulted phase at the left and right terminals. Typi-cally, the magnitude of the fault current decreases as the fault moves away from the ter-minal. In the case of Fig.10, the right terminal current reaches its minimum for a fault atabout 55% from the right end of the line and starts increasing for more distant faults.What is also quite unusual is that the two currents are practically out of phase for faultsclose to the left terminal. Both the currents are unnaturally shifted: the left-terminal cur-rent is capacitive; the right-terminal current is practically resistive and out of phase withthe phase A voltage.

This situation may cause potential problems to the phase comparison protection. Thecurrent differential relay shall still perform correctly, as the currents – although out ofphase – differ significantly as to the magnitude. This would result is a large differentialcurrent and a trip.


Page 14 of 35

0 10 20 30 40 50 60 70 80 90 1005

5.5

6

6.5

Fault Location, %

Faul

tCur

rent

,A

0 10 20 30 40 50 60 70 80 90 100-40

-20

0

20

40

Fault Location, %

Faul

tCur

rent

,deg

Figure 9. Fault current: magnitude (top) and phase angle (bottom).

0 10 20 30 40 50 60 70 80 90 1000

2

4

6

8

Fault Location, %

Faul

ted

Phas

eC

urre

nt,A

0 10 20 30 40 50 60 70 80 90 100-150

-100

-50

0

50

Fault Location, %

Faul

ted

Phas

eCu

rrent

,deg

Left Terminal

Right Terminal

Right Terminal

Left Terminal

Figure 10. Faulted phase currents: magnitude (top) and phase angle (bottom).

Fig.11 presents the faulted phase voltages at the right and left (both bus-side and line-side VTs) terminals. For faults beyond some 25% of the line, the line-side voltage is sig-nificantly above the nominal value. This results from the fact that the inductive phasecurrent is flowing through a capacitive impedance causing a voltage drop out of phasewith the pre-fault voltage. The negative voltage drop subtracted from the bus voltage in-creases the voltage on the line side of the SCs above the pre-fault value. The bus-sidevoltage, in turn, rises above its nominal value for close in faults. This results from a ca-pacitive phase current flowing through an inductive system impedance and causing a


Page 15 of 35

voltage drop out of phase with the system electromotive force (e.m.f.). Subtracting thenegative value form the latter results in an overvoltage condition on the bus side.

0 10 20 30 40 50 60 70 80 90 10050

55

60

65

70

75

80

Fault Location, %

Faul

ted

Phas

eVo

ltage

,V

0 10 20 30 40 50 60 70 80 90 100-40

-20

0

20

40

Fault Location, %

Faul

ted

Phas

eVo

ltage

,deg

Left Terminal (bus voltage)

Right Terminal

Right Terminal

Left Terminal (line voltage)

Left Terminal (bus voltage)

Left Terminal (line voltage)

Figure 11. Faulted phase voltages: magnitude (top) and phase angle (bottom).

0 10 20 30 40 50 60 70 80 90 1000.5

1

1.5

2

2.5

3

Fault Location, %

Sym

met

rical

Cur

rent

s,A

0 10 20 30 40 50 60 70 80 90 100-40

-20

0

20

40

60

Fault Location, %

Sym

met

rical

Curre

nts,

deg

Left Terminal (I2,I1F)

Left Terminal (I0)

Left Terminal (I2,I1F)

Left Terminal (I0)

Figure 12. Symmetrical currents at the left terminal: magnitude (top) and phase angle (bottom).

Fig.12 presents symmetrical components of the current at the left terminal. Bothnegative- and zero-sequence currents are capacitive for faults within the 35-40% range.Due to differences between the zero- and negative-sequence impedances of the system,the switchover points for the two currents are different. This results in a spot between35% and 40% of the line, where the zero-sequence at the left terminal current is inductivewhile the negative-sequence current is capacitive.


Page 16 of 35

0 10 20 30 40 50 60 70 80 90 1002

4

6

8

10

12

Fault Location, %

AGLo

opC

urre

nt,A

0 10 20 30 40 50 60 70 80 90 100-40

-20

0

20

40

Fault Location, %

AGLo

opC

urre

nt,d

eg

Figure 13. Fault loop current at the left terminal: magnitude (top) and phase angle (bottom).

0 10 20 30 40 50 60 70 80 90 1000

0.5

1

1.5

Fault Location, %

Sym

met

rical

Cur

rent

s,A

0 10 20 30 40 50 60 70 80 90 100-150

-100

-50

0

Fault Location, %

Sym

met

rical

Curre

nts,

deg

Right Terminal (I2,I1F)

Right Terminal (I0)

Right Terminal (I2,I1F)

Right Terminal (I0)

Figure 14. Symmetrical currents at the left terminal: magnitude (top) and phase angle (bottom).

Fig.13 shows a fault loop current (equation (10)) at the left terminal obtained usingthe zero-sequence compensation setting as for a plain line. The current is capacitive forfaults closer than 40% of the line. This will cause directionality problems for the dynamicmho characteristic. The characteristic expands for forward faults (inductive currents) andcontracts for reverse faults (capacitive currents).

Fig.14 presents symmetrical currents at the right terminal. The magnitudes exhibittheir minima at 45% and 85% of the line measured from the right terminal. Within thisinterval the negative- and zero-sequence currents are practically out of phase, while forfaults close to the terminals, the currents are in phase. Normally the fault component of


Page 17 of 35

the positive-sequence current and the zero- and negative-sequence currents are all inphase for a SLG (AG) fault. Therefore, in the sample system faults within the 45% to85% interval may cause potential phase selection problems at the right terminal for algo-rithms that use relationships between the zero- and negative-sequence currents (the twocannot be out of phase for any known fault type on an uncompensated line).

0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

Fault Location, %

AGLo

opC

urre

nt,A

0 10 20 30 40 50 60 70 80 90 100-140

-120

-100

-80

-60

-40

-20

Fault Location, %

AGLo

opCu

rrent

,deg

Figure 15. Fault loop current at the left terminal: magnitude (top) and phase angle (bottom).

Fig.15 presents the fault loop current (with the zero-sequence compensation setting asfor a plain line) for the right-terminal relay. The current is significantly shifted from itsnatural position for faults close to the left terminal of the line.

Figs.16 and 17 show symmetrical voltages at the left terminal. Normally, for a SLGfault, the two voltages are out of phase as compared with the pre-fault positive sequencevoltage. Due to series compensation the two voltages are significantly shifted from thenatural positions. The line-side voltages are close to their natural position for close-infaults, but are some 110 degrees off for remote faults (Fig.16). The bus-side voltages areclose to their natural position for remote faults (40 degrees off), but are significantlyshifted for close-in faults (120 degrees off – Fig.17).

Fig.18 presents the symmetrical voltages at the right terminal. Their angular positionis approximately correct: both zero- and negative-sequence voltages are within some 40-degree limit at the –180-degree position. Magnitudes of the two voltages exhibit an inter-esting pattern: for faults at 45% and 85% from the right terminal, respectively, the volt-ages measured by the right-terminal relay are very small (practically zero). This wouldcreate some problems for negative- and zero-sequence directional elements, unless theelements are biased towards operation if the polarizing voltage is very low (known andpreferred solution; see Section 5 for more details).


Page 18 of 35

0 10 20 30 40 50 60 70 80 90 1000

5

10

15

20

Fault Location, %

Sym

met

rical

Volta

ges,

V

0 10 20 30 40 50 60 70 80 90 10060

80

100

120

140

Fault Location, %

Sym

met

rical

Volta

ges,

deg

LeftTerminal - Line Voltages (V2,V1F)

Left Terminal - Line Voltages (V0)

Left Terminal - Line Voltages (V2,V1F)

Left Terminal - Line Voltages (V0)

Figure 16. Symmetrical voltages – left terminal, line side VTs: magnitude (top) and phase angle (bottom).

0 10 20 30 40 50 60 70 80 90 1001

2

3

4

5

6

Fault Location, %

Sym

met

rical

Volta

ges,

V

0 10 20 30 40 50 60 70 80 90 100-140

-120

-100

-80

-60

-40

Fault Location, %

Sym

met

rical

Volta

ges,

deg

LeftTerminal - Bus Voltages (V2,V1F)

Left Terminal - Bus Voltages (V0)

Left Terminal - Bus Voltages (V2,V1F)

Left Terminal - Bus Voltages (V0)

Figure 17. Symmetrical voltages – left terminal, bus side VTs: magnitude (top) and phase angle (bottom).

3.2. Voltage and Current InversionAs practically illustrated in the previous subsection, various currents (fault, phase,

fault loop, sequence) and voltages (phase, sequence) may exhibit significant phase shiftsas compared with their natural positions. Terms “voltage inversion” or “current inver-sion” are not meaningful unless the “voltage” and “current” terms are defined as well asthe term “inversion”. In this paper, an inversion is defined as a shift by more than 90 de-grees from a “natural” position of a given phasor.


Page 19 of 35

0 10 20 30 40 50 60 70 80 90 1000

5

10

15

20

Fault Location, %

Sym

met

rical

Volta

ges,

V

0 10 20 30 40 50 60 70 80 90 100-200

-100

0

100

200

Fault Location, %

Sym

met

rical

Volta

ges,

deg

Right Terminal (V2,V1F)

Righ Terminal (V0)

Right Terminal (V2,V1F)

Righ Terminal (V0)

Figure 18. Symmetrical voltages – left terminal: magnitude (top) and phase angle (bottom).

Applying the above definition to the right terminal relay, it is concluded that none ofthe currents or voltages is inverted. Several signal exhibit significant phase displace-ments, but none is truly inverted.

The left-terminal relay experiences current and voltage inversions. Fig.19 illustratesthe fault positions that cause inversion of particular signals. Specifically:

The zero-, negative-sequence (and the fault component of the positive-sequence) cur-rents are inverted for faults beyond the SCs up to some 30-40% of the line length. Conse-quently, the faulted phase current and the ground loop currents are inverted as well. Asseen in Fig.19a, there are fault locations for which only some of the aforementioned cur-rents are inverted.

If the left relay is fed from the bus-side VTs (Fig.19b), the zero- and negative-sequence voltages are inverted for faults beyond the SCs, up to some 30-40% of the linelength. If the relay is fed from the line-side VTs, the voltages are also inverted but forfaults close to the right terminal.

The latter phenomenon may seem unexpected. It could be explained as follows. Forfaults close to the right terminal, the negative- and zero-sequence symmetrical networksfrom the fault point to the left electromotive force (e.m.f.) are inductive (Fig.8). So arethe networks from the fault point to the right e.m.f. Consequently, the zero-, negative-and positive-sequence equivalent impedances seen from the fault point are inductive. Thefault current is inductive and the sequence voltages at the fault point are not inverted.Since the negative- and zero-sequence voltages at the fault point are not inverted, and thenetworks are not capacitive, the negative- and zero-sequence currents at the left terminalare not inverted. The impedance, however, from the potential point (line-side VTs) backto the left e.m.f. is capacitive. The inductive negative-sequence current flowing throughthat capacitive impedance causes inversion of the negative-sequence voltage. This phe-nomenon does not happen for close in faults, when the negative-sequence current is in-


Page 20 of 35

verted (Fig.19a), as the inverted current and capacitive impedance cause make the nega-tive-sequence voltage to appear in its natural position.

The above simple method can be used to analyze inversions of all the relevant sig-nals. What is interesting is that practically all the system parameters influence any par-ticular inversion. For example, the zero-sequence impedance affects inversion of the faultcurrent, thus inversion of the sequence voltage at the fault point, thus inversion of thenegative-sequence current at a given terminal, thus inversion of the negative-sequencevoltage at that terminal.

(a)

I0 inversionI2 inversionIA inversionIAG (loop current) inv.

(b)V0 inversionV2 inversion

(c)V0 inversion

V2 inversion

VTs

VTs

Figure 19. Fault locations causing current and voltage inversions at the left terminal relay: currents (a),voltages while using the bus-side VTs (b), and voltages while using the line-side VTs (c).

3.3. Response of Selected Comparators to Signal InversionAs illustrated in the previous subsection various signals may invert or shift considera-

bly for various fault locations. A practical “distance element” of a modern relay is builtout of several mutually supervising comparators. Those comparators involve several dif-ferent currents and voltages. The actual equations must be used in order to predict re-sponse of any particular comparator to faults on series compensated lines.

This subsection continues the previous example and examines responses of the fol-lowing distance comparators:

100% memory-polarized dynamic mho

IAG Z – VAG versus V1A_mem (12)


Page 21 of 35

Zero-sequence current-polarized reactance

IAG Z – VAG versus I0 Z (13)

100% memory-polarized negative-sequence directional

I2 Z versus V1A_mem (14)

100% memory-polarized zero-sequence directional

I0 Z versus V1A_mem (15)

where Z is the reach impedance.One possible mho distance function could include conditions (12), (13), (14) and

(15). One possible quadrilateral function could include conditions (13), (14) and (15).Fig.20 presents the coverage of the comparators (12) through (15) for the right-

terminal relay and the left-terminal relay fed form the bus-side VTs. The reach is set attwice the line impedance (uncompensated) and the overreaching zone is analyzed.

The mho comparator of the right-terminal relay covers only about 10% of the line.Combination of high fault resistance and signal inversions does not allow covering moredespite the fact the reach is set at twice the line impedance.

The reactance characteristic has a gap in its coverage: it will not pickup for faultsfrom about 60% to 90% of the line length measured from the right terminal.

The negative- and zero-sequence memory-polarized directional functions cover theentire line.

The response of both the mho and reactance comparators is unexpected as at the rightterminal none of the signals is truly inverted. It is enough, however, for the signals to beshifted from their natural positions to cause problems.

mho

VTs

mho

reactance

reactancerea

memory polarized I0 directionalmemory polarized I2 directional


Coverage of the Right Terminal Relay

Coverage of the Left Terminal Relay

Figure 20. Coverage of various comparators while using the bus-side VTs.

The mho comparator of the left-terminal relay fed with the bus-side voltage willcover only some 10% of the line. The covered spot is unexpectedly in the middle of the


Page 22 of 35

line. The comparator will not pickup close-in faults due to signal inversions. It will notpickup faults located further due to limited resistive coverage of the mho towards thereach point.

Surprisingly, the reactance function covers the entire line (despite the signal inver-sions). The negative- and zero-sequence memory-polarized directional elements exhibitblind spots for close-in faults because the currents are inverted (Fig.19a).

Fig.21 presents the coverage of the same line, but for the case when the left-terminalrelay is fed from the line-side VTs. Situation is very similar except the mho comparator:it does not cover the line at all (again, the reach is set at twice the line impedance).

VTs

mho

reactance

reactancerea



Coverage of the Right Terminal Relay

Coverage of the Left Terminal Relay

Figure 21. Coverage of various comparators while using the line-side VTs.

As illustrated in this section, several unusual phenomena may occur on series com-pensated lines under low-current fault conditions. Transients – specific to series compen-sation – complicate the situation even further.

4. Transients

Most transient problems for relays installed on series compensated lines are caused bysub-synchronous oscillations. Series capacitors, if not effectively by-passed by the gapsand/or MOVs during a fault, make the fault loop capacitive-inductive-resistive. Such acircuit has its own resonant frequency that is very close to the nominal system frequency.The resonant frequency depends on fault location and fault type as the equivalent circuitand its parameters vary. In any case, the sub-synchronous oscillations can be only 5-10Hzaway from the system nominal frequency.

The sub-synchronous oscillations cannot be effectively filtered out by a relay unlessvery long data windows are applied for digital filters and the relay is significantlyslowed-down. Typically, relays apply the same filtering techniques as for regular appli-cations. As a consequence the phasors estimated by the relay follow the signal envelopes.

Sub-synchronous oscillations occur also in conjunction with high-current faults: dur-ing the fault the MOVs conduct and the circuit is well controlled. After the fault iscleared, however, the current drops, the MOVs stop conducting and the SCs are effec-tively re-inserted into the system. This results in large post-fault oscillations. High energy


Page 23 of 35

stored in the line inductance during the fault contributes to the magnitude of those oscil-lations.

As an example, Figs.22 and 23 present a low-current SLG fault in a sample seriescompensated system.

0.2 0.25 0.3 0.35 0.4 0.45

-100

0

100

Phas

eA

Volta

ge,V

0.2 0.25 0.3 0.35 0.4 0.45

-100

0

100

Phas

eB

Volta

ge,V

0.2 0.25 0.3 0.35 0.4 0.45

-100

0

100

time, sec

Phas

eC

Volta

ge,V

fault fault cleared

Figure 22. Sample fault on a series-compensated line: line-side VT voltages.

0.2 0.25 0.3 0.35 0.4 0.45

-10-505

10

Phas

eA

Cur

rent

,A

0.2 0.25 0.3 0.35 0.4 0.45

-10

0

10

Phas

eB

Cur

rent

,A

0.2 0.25 0.3 0.35 0.4 0.45

-10

0

10

time, sec

Phas

eC

Cur

rent

,A

fault fault cleared

Figure 23. Sample fault on a series-compensated line: currents.

Heavy oscillations are visible in both currents and voltages. Relay frequency trackingmay be one of the features severely affected by this kind of signal distortions. The zero-crossings of the waveforms fluctuate in time. If the frequency tracking is based on zero-crossing detection and is too fast, it may cause serious problems. If the frequency track-


Page 24 of 35

ing is based on phasors, it is not free from errors either as the phasors will move duringthe fault as well due to the sub-synchronous oscillations. The frequency tracking does nothave to be too fast. Therefore, averaging and/or other post-filtering such as median fil-tering solves the problem.

Figs.24 and 25 present plots of the voltage and current magnitudes measured usingone-cycle digital estimators (variations of the Fourier algorithm). As seen from the fig-ures, the estimated magnitudes follow the envelopes of their waveforms.

0.2 0.25 0.3 0.35 0.4 0.450

20

40

60

80

100

120

140

Volta

geM

agni

tude

s,V

time, sec

fault fault cleared

Figure 24. Sample fault on a series-compensated line: voltage magnitudes.

0.2 0.25 0.3 0.35 0.4 0.450

2

4

6

8

10

12

Cur

rent

Mag

nitu

des,

A

time, sec

fault fault cleared

Figure 25. Sample fault on a series-compensated line: current magnitudes.


Page 25 of 35

The sub-synchronous oscillations take significant effect on transient accuracy of dis-tance functions. In addition to the steady-state overreach resulting from the degree ofcompensation, significant transient overreach must be taken into account when settingunderreaching distance functions.

Taking into account all the sources of error (50-70% of steady-state overreach, 10-20% due to VT/CT/impedance errors, and 10-30% of the transient overreach), it turns outan underreaching distance zone cannot be practically set. Section 6 presents a solution tothis problem.

5. Negative- and Zero-Sequence Directional Elements

As illustrated in Sections 3 and 4 an overreaching distance element may not pickupall low-current faults on the protected line. In addition, an underreaching distance zonecannot be set to cover a sizeable portion of the protected line if the SCs are located be-tween the relay potential source and the far-end busbar, or there are other series compen-sated lines originating from the far-end busbar that have their SCs installed in the far-endsubstation.

One solution to this problem relies on ground directional elements such as negative-sequence or zero-sequence for pilot-aided schemes and on a current-controlled adaptivereach approach to underreaching distance elements. This section describes the first, whilethe next section describes the second part of the solution.

5.1. Offset Impedance Approach to Ground Directional ElementsFig.26a shows a negative-sequence voltage profile for a forward fault, while Fig.26b

shows the vector diagram for the involved negative-sequence signals. Normally, the cur-rent lags the inverted voltage by the angle of the negative-sequence impedance measuredfrom the potential source back to the local equivalent system. Therefore, the polarizingand operating quantities are defined as:

ECAISVS oppol ∠⋅=−= 1, 22 (16)

where ECA is an Element Characteristic Angle, set as explained above.For applications on plain lines, there is a danger that the polarizing signal (negative-

sequence voltage) may be too low to ensure reliable operation for forward faults. Thismay happen if the local equivalent system is very strong. Adding a small voltage to thepolarizing signal in phase with the operating signal solves the problem. Consequently, theoffset-impedance directional element is defined as follows.

Forward-looking (“tripping”) negative-sequence directional element:ECAISECAZIVS opoffsetpol ∠⋅=∠⋅⋅+−= 1,1 222 (17a)

Reverse-looking (“blocking”) negative-sequence directional element:ECAISECAZIVS opoffsetpol ∠⋅−=∠⋅⋅+−= 1,1 222 (17b)


Page 26 of 35

As shown in Fig.26b, the offset impedance increases the polarizing signal and ensuresproper operation.

Fig.26c shows the voltage profile for a reverse fault. During a reverse fault, the volt-age at the relaying point is at least the line impedance times the current at the relayingpoint. The offset impedance reduces the polarizing signal, but the element responds cor-rectly as long as the offset impedance is lower than the line impedance plus the imped-ance of the remote system (Fig.26d).

For applications on non-compensated lines, the offset impedance is set at the level of10-25% of the line impedance. This improves speed of operation, and guarantees detec-tion of forward faults even in very strong systems.

Relay

V_2

(a)

Relay

V_2 >I_2 * Z_2line

(c)

I_2

ECA

I_2 x ZV_2 -V_2

S_polS_op

(b)

I_2

ECA

I_2 x Z

V_2-V_2

S_pol S_op

(d)

-V_2

Figure 26. Negative-sequence currents and voltages for a forward ((a) and (b))and reverse ((c) and (d)) faults.

This known principle can be used on series-compensated lines resulting in a very ro-bust, dependable and secure directional element.

If, on a series compensated application, the negative-sequence voltage gets inverted, awrong directional indication will be given. However, if the offset impedance is highenough, the wrong direction is counterbalanced and the element responds correctly.

The same principle applies to the zero-sequence directional element.

5.2. Setting Recommendations for the Offset-Impedance Directional ElementsThe relation between the negative-sequence voltage and current at the relaying point

is entirely determined by the negative-sequence impedance measured from the potentialsource of the relay back to the e.m.f. of the local equivalent system. If this impedance isinductive, the negative-sequence voltage and current are in a “natural” relationship. If thisimpedance is capacitive, the relationship is such that the directional identification iswrong. The above results in very straightforward setting recommendations:

1. If the net impedance between the relay potential source and the e.m.f. of the lo-cal equivalent system is inductive, then no offset impedance is needed. If theimpedance is capacitive, then the offset shall be at least the reactance of such


Page 27 of 35

net capacitive impedance. This ensures that the element operates on all forwardfaults within its reach.

2. The offset impedance cannot be higher than the net inductive impedance be-tween the relay potential point and the e.m.f. of the remote equivalent system.Otherwise, the element is overcompensated and will pick up on reverse faults.

Consider the system of Fig.7. Neglecting the angle differences between the involvedimpedances one calculates the following settings:

Negative-sequence Directional Element of the Right Relay:Zoffset > 0 (no need for offset as the impedance behind the relay

is inductive)Zoffset < 10-7+1.43 = 4.43 ohm (the offset cannot be higher than the net impedance

from the potential source to the forward e.m.f.)Zero-sequence Directional Element of the Right Relay:Zoffset > 0Zoffset < 30-7+2.68 = 25.6 ohmNegative-sequence Directional Element of the Left Relay (line-side VTs):Zoffset > -(-7+1.43) = 5.57 ohm (offset higher than 5.57 ohm must be applied other-

wise forward faults will not be detected)Zoffset < 10+3.44 = 13.44 ohm (the offset cannot be higher than the net impedance

from the potential source to the forward e.m.f.)Zero-sequence Directional Element of the Left Relay (line-side VTs):Zoffset > -(-7+2.68) = 4.32 ohmZoffset < 30+16.4 = 46.4 ohmNegative-sequence Directional Element of the Left Relay (bus-side VTs):Zoffset > 0 (no need for offset as the impedance behind the relay

is inductive)Zoffset <-7+10+3.44 = 6.44 ohm (the offset cannot be higher than the net impedance

from the potential source to the forward e.m.f.)Zero-sequence Directional Element of the Left Relay (bus-side VTs):Zoffset > 0Zoffset < -7+30+16.4 = 39.4 ohm

If the system of Fig.6 may change configuration significantly, the original systemshall be considered rather then the equivalent shown in Fig.7. Consequently, the settingsmay be different.

Offset negative- and zero-sequence directional elements offer an excellent directionalintegrity for pilot aided schemes. They not only guarantee correct operation on series


Page 28 of 35

compensated lines, regardless of the location of the VTs and SCs, but also are fast andsensitive [3].

As long as the protected network is not overcompensated, the element can always beset, i.e. the two setting rules can always be satisfied.

6. Adaptive Distance Reach

As illustrated in section 3 distance functions do not offer reliable protection for low-current faults when applied to series compensated lines. First, they may fail to pick upinternal faults, even when significantly overreaching the protected line. Second, for anunderreaching direct tripping operation they must be set very conservatively, sometimesas low as 10-20% of the line in order to avoid overreaching due to sub-synchronous os-cillations.

One solution to the problem of effective application of distance protection to seriescompensated lines is to use a current-controlled adaptive reach method. In this solutionthe reach of a distance function is reduced by the following value:

( ) ( )SETLIM

SETeffective ZangleI

VZIZ ∠⋅−= 1 (18a)

The effective reach is a function of the magnitude of the loop current, I. The reduc-tion of reach takes place along the maximum torque angle line: the higher the current, thesmaller the reduction. For very large currents, the reach is pulled back by a very smallpercentage. On the other hand, for currents below certain value, the distance element iseffectively blocked:

blockZVI

SET

LIM →< (18b)

The VLIM value is a setting and shall be set as a sum of voltage protection levels of allthe SCs from the relay potential point up to the point where the zone must not reach.

Fig.27 illustrates the adaptive reach approach: for currents that cause the voltage dropbelow the voltage protection level across any SC within the intended reach, the reach ispulled back entirely and the zone is practically blocked. For larger currents that cause theMOVs and/or air gaps to conduct some current, stabilizing sub-synchronous oscillationsand reducing an effective overreach due to the SCs, the reach is reduced accordingly. Forvery large currents, when the SCs are practically completely by-passed, the reach is notreduced at all.

This approach gives maximum security, but at the same time allows covering up totraditional 80-85% of the line with the underreaching distance zone if the fault current ishigh enough to ensure suitable operating conditions for the distance protection principle.Low-current faults are covered by the ground directional functions.

Fig.28 illustrates the adaptive reach approach. Low-current faults at the reach pointwill cause significant steady-state overreach and extra transient overreach due to sub-


Page 29 of 35

synchronous oscillations. In such conditions (Fig.28a), the reach is reduced considerablyso that the zone does not overreach. During high-current faults at the reach point(Fig.28b), the zone reach is not reduced, but at the same time there is no need to reduce itas the effective Goldsworthy’s reactance is small and there are no sub-synchronous os-cillations. High-current internal faults are covered as in traditional applications becausethe reach is not reduced (Fig.28c).

THE REACH ISDYNAMICALLYREDUCED BY

VL/abs(I)

SET REACH (ZR)

ACTUAL REACH ISA FUNCTION OFCURRENTMAGNITUDE

ACTUAL REACHFOR VERY HIGHCURRENTS

ACTUAL REACHFOR VERY SMALLCURRENTS R

X FAR-ENDBUSBAR

Figure 27. Current-dependent adaptive reach.

Steady-stateapparentimpedance

R

X

FAR-ENDBUSBAR

MO

V

SC

RF

ZSC&M

OV

ImpedanceTrajectory

The reach is safelyreduced

(a)

Set reach

LOW-CURRENT EXTERNALFAULT


R

X

FAR-ENDBUSBAR

MO

V

SC

RF

ImpedanceTrajectory (nosubsynchronousoscillations)

The reach is notreduced

Set reach

(b)HIGH-CURRENT EXTERNAL

FAULT


R

X

FAR-ENDBUSBARM

OV

SC

RF

ImpedanceTrajectory (nosubsynchronousoscillations)

The reach is notreduced

Set reach

(c) HIGH-CURRENT INTERNALFAULT

Figure 28. Illustration of the current-dependent adaptive reach.

For overreaching zones, the VLIM setting shall be set at zero.For an underreaching zone, the VLIM setting shall be set as a sum of the voltage pro-

tection levels for the SCs between the relay potential point and the intended reach point.For example, for the system of Fig.6, the following settings apply:Right-terminal Relay:VLIM = sum of the voltage protection levels of the 7-ohm and 5-ohm SCs (the right termi-

nal Zone-1 must not overreach for faults beyond the 5-ohm SC when the 2-ohmequivalent system is disconnected or becomes very weak).


Page 30 of 35

Left-terminal Relay (bus-side VTs):VLIM = sum of the voltage protection levels of the 7-ohm and 12-ohm SCs (the left termi-

nal Zone-1 must not overreach for faults beyond the 12-ohm SC).Left-terminal Relay (line-side VTs):VLIM = the voltage protection level of 12-ohm SCs.

Using the current-controlled adaptive reach approach one may set the underreachingzone very aggressively, at 80-85% of the impedance of a non-compensated line.

7. Phase Selection for Single-Pole Tripping

Series compensated lines are strong links in transmission backbones of power sys-tems. Many utilities choose to apply single-pole tripping in order to keep the two healthyphases in services on SLG faults and maintain reduced power transfer.

Fast and accurate phase selection is critical for single-pole tripping. The following is-sues need to be considered for accurate phase selection.

Signal inversions during low-current faults may lead to wrong fault identification.The impact depends on a particular phase selection method used by a relay.

Sub-synchronous oscillations may cause additional problems because the signalsmeasured by the relay – typically phasors – are estimated with large errors.

For high-current faults, however, phase selection on series compensated does not im-pose more problems as compared with plain non-compensated lines.

One phase selecting algorithm uses a hierarchical approach to the fault type recogni-tion [4]. Firstly currents are used for phase selection. If the currents fail to recognize thefault, the voltages are utilized. Both currents and voltages are analyzed in terms of angu-lar relationships between the positive-, negative- and zero-sequence signals.

In the first step magnitudes of the fault component of the positive-sequence, negative-sequence and zero-sequence signals are checked. The thresholds – intended for confirm-ing if a given component is large enough to provide extra information on the fault type –are adaptive: carefully selected portions of both positive- and zero-sequence currents areused as adaptive thresholds. Once a given symmetrical component is validated, its angu-lar position with respect to the two other components is checked using the well-knownfault patterns shown in Fig.29.

The relation between the positive- and negative-sequence currents is always checked.If the zero-sequence current is also present, the relation between the negative- and zero-sequence currents is checked as well. If both positive-to-negative and negative-to-zerofault patterns are checked, they must agree in order to identify a given fault type.

If the currents fail to recognize the fault, the voltages are used in exactly same man-ner. The currents may fail during weak infeed conditions, when the zero-sequence currentfrom close-in transformers may dominate the positive- and negative-sequence currentspractically destroying any asymmetry in the phase currents and making phase selectionimpossible. If this is a case, the voltages show a great deal of asymmetry (weak system)


Page 31 of 35

allowing the relay to recognize the fault type. If the voltages fail (cross country faultswith the internal fault away from the relaying point, for example), the algorithm uses anunderreaching distance zone to determine the fault type. The latter is likely to be correctin such circumstances.

By utilizing two fault patterns, the algorithm is more secure. As explained and illus-trated in [4], the algorithm is very fast because it uses phase angles of signals that are ze-ros in the pre-fault state, and as such are not biased in any particular direction. For exam-ple, despite the oscillations and similar level of magnitudes of current in all three phases,the waveforms shown in Fig.23 are recognized correctly as an AG fault using 2-msecworth of data.

I1FI2F

AG

AB, ABGBG

BC, BCG

CG AC, ACG

(a)

I0FI2F

(b)

AG, BCG

CG, ABG

BG, CAG

Figure 29. Negative-sequence vs. positive-sequence fault signature (a) andnegative-sequence vs. zero-sequence fault signature (b).

During cross-country faults and/or due to problems with series compensation phaseselection may present some problems. However, it is very unlikely that phase selectors atboth terminals would misoperate. If one relay fails to recognize the fault correctly, multi-bit communication schemes as the ones presented in [4] ensure proper operation in ma-jority of troublesome cases. In a multi-bit scheme a relay uses more than one bit to send apermissive or blocking signal to its peer. In this way extra information on the fault typecan be exchanged between the relays. Nowadays, with advances in relay communicationssystems, using more than one bit for signaling schemes is not very difficult nor expen-sive, and shall be considered for single-pole tripping applications in difficult conditionssuch as series compensation.

8. Fault Location

Because series compensated lines are typically long, accurate fault location is essen-tial. Fault location in series compensated lines presents, however, even bigger challengethan distance protection. While reach accuracy, directional integrity and dependability arethe only requirements for a distance function, ability to pinpoint accurately any fault onthe line is the requirement for a fault locator.


Page 32 of 35

Several families of fault locating algorithms have been proposed in the past rangingfrom single-ended impedance-based approaches to traveling-wave-based techniques.Majority of commercially available fault locators either provided as a part of a DigitalFault Recorder (DFR) software package, stand alone device or integrated with digital re-lays, are based on fundamental frequency voltages and currents, thus on the apparent im-pedance.

Single-ended impedance-based algorithms make use of a set of equations that is un-der-stated, e.g. one equation is missing in order to perform fault location calculations.Various assumptions are made (such as the fault components of the currents from bothends of the line being in phase, fault impedance being pure resistance, or local and re-mote system equivalent impedances known) in order to generate the missing equation andderive a fault location algorithm. Various assumptions yield various fault locating meth-ods.

Double-ended algorithms use an over-stated set of equations, e.g. depending on theamount of data measured at both terminals one may generate more equations than un-knowns. This allows the locator to avoid using data that may not be accurate (such as thezero-sequence impedance of the line) and improve accuracy of fault location.

Regardless of the approach taken, the impedance algorithms solve a set of equationsthat describes a simplified electrical circuit for the unknown being the fault point (and asa by-product – the fault resistance). If the series capacitors and MOVs are not part ofthose equations, fault location cannot be accurate. Traditional algorithms may show up to20% of error when applied on series compensated lines [2].

There are several difficulties as far as a successful fault location in series compen-sated lines is considered.

First, the non-linear current-dependent impedance of the Goldsworthy’s model mustbe taken into account. This impedance depends on the through current and makes theequations non-linear. In addition, the sequence networks representing a three-phase bankof SCs and MOVs are mutually coupled. The coupling cannot be neglected when locatingfaults due to accuracy requirements. Consequently, the equations cannot be solved whendesigning an algorithm, and the fault locator must solve them numerically.

Second, there may be one or two compensating banks installed on the line at variouslocations. This calls for several parallel algorithms, or “fault locators” that assume a faulton a separate section of the line. Each such sub-algorithm delivers its own fault locationestimate. A separate selection procedure must be developed to decide on the most likelyfault location, or at least prioritize the results for the inspection crew.

Third, single-ended methods do not measure the currents at the remote terminal. Ifthere is series capacitor between the fault and the remote terminal, the Goldsworthy’smodel cannot be easily applied in a single-ended method because the through current ofthe compensating bank cannot be measured directly. Still, the operating point of the SCsand MOVs affects the accuracy of fault location. One solution to this problem is to solvefor the remote currents interactively [2,5]. The solution, however, requires the equivalentimpedance of the remote system.


Page 33 of 35

Fourth, if relays protecting the line trip fast, there may be a very short window of dataavailable to the fault locator. Presence of sub-synchronous oscillations may affect consid-erably accuracy of phasor estimation, and consequently, fault location.

Fifth, for slow relay/breaker operations, the MOVs may accumulate their maximumallowable energy and may get by-passed by the gaps triggered from the MOV thermalprotection. If this happens, the available data window, although long, is divided into sev-eral sections, each corresponding to a different network topology (MOVs on and by-passed). Each section of the recording may be too short to perform accurate fault loca-tion.

One impedance-based fault-locating algorithm developed specifically for series com-pensated lines [2,5] applies the Goldsworthy’s equivalent impedances. Instead of usingfault-loop approach, the algorithm solves the equations in the natural three-phase coordi-nates (i.e. for the phase voltages and currents). For a line with one SC bank on the line,the algorithm runs two internal “fault locators”: one assumes the fault between the relayand the SCs, the other assumes the fault beyond the SCs. A separate selection algorithmis proposed to indicate the correct alternative.

9. Conclusions

Protective relaying problems specific to series compensated lines are encountered totheir fullest extent only during low-current faults. Low-current faults can happen either inweak systems or due to large fault resistances. In the former case, the effect of seriescompensation is practically canceled by large inductive impedance of the system. Thelatter case can practically happen only during SLG faults. While a general analysis of aseries compensated system is quite difficult (Goldsworthy’s equivalent, mutual couplingbetween the sequence networks), the primary case to be considered for practical reasonsis a straightforward situation of a low-current SLG fault.

Generally, an optimal zero-sequence compensating factor for ground distance protec-tion is a complex function of fault location and fault current. Practically, the value calcu-lated from the line parameters may be used as other factors affecting accuracy of grounddistance protection are of much higher significance.

Both bus-side and line-side locations of the relay potential source create their ownproblems. Using line-side voltages does not simplify series compensation applications.The series capacitors are still there, either in the forward or reverse direction.

Distance protection faces serious difficulties during low-current faults on, or in a vi-cinity of series compensated lines. Distance protection overreaches due to series compen-sation, is exposed to large errors due to sub-synchronous oscillations, and may fail to de-tect internal faults due to signal inversions. An underreaching zone must be pulled backdrastically protecting only small portion of the line. Overreaching zones face dependabil-ity limitations even if their reach is very large.

Ground (negative-sequence and neutral) directional functions offer an excellent alter-native to the overreaching distance protection. The offset impedance approach ensures


Page 34 of 35

very good directional integrity regardless of the degree of series compensation, as well aslocation of series capacitors and voltage transformers.

Current-controlled adaptive reach is a preferred solution to the steady-state and tran-sient overreaching problems for the directly tripping distance zone. The solution offersmaximum security and at the same time allows tripping high-current faults within tradi-tional 80-85% of the line length.

Single-pole tripping is also exposed to problems associated with series compensation.Advanced phase selecting algorithms and multiple-bit pilot-aided schemes improve per-formance of the protection systems.

Fault location on series compensated lines is affected to a great extent. Traditional al-gorithms may exhibit errors at the level of 20%. Traveling wave-based algorithms andnovel impedance-based methods developed specifically for series compensation are muchmore accurate.

Despite constantly increasing power demand, series compensated lines are still quiterare. It is worth to undertake an effort to simulate transient conditions for series compen-sated applications, fine-tune and verify settings, and finally test the relaying system usingdigital simulators and/or play back systems.

10. References

[1] GOLDSWORTHY D.L., “A linearized model for MOV-protected series capacitors”, IEEETransactions on Power Systems, Vol.2, No.4, pp.953-958, November 1987.

[2] SAHA M.M., IZYKOWSKI J., ROSOLOWSKI E., KASZTENNY B., "A new accuratefault locating algorithm for series compensated lines", IEEE Transactions on Power Deliv-ery, Vol.14, No.3, July 1999, pp.789-797.

[3] KASZTENNY B., SHARPLES D., CAMPBELL B., POZZUOLI M., “Fast Ground Direc-tional Overcurrent Protection–Limitations and Solutions”, Proceedings of the 27th AnnualWestern Protective Relay Conference, Spokane, WA, October 24-26, 2000.

[4] KASZTENNY B., CAMPBELL B., MAZEREEUW B., “Phase Selection for Single-PoleTripping – Weak Infeed Conditions and Cross Country Faults”, ”, Proceedings of the 27th

Annual Western Protective Relay Conference, Spokane, WA, October 24-26, 2000.[5] SAHA M.M., HILLSTROM B., IZYKOWSKI J., ROSOLOWSKI E., KASZTENNY B.,

"A method of fault location for series-compensated lines", Patent WO99/32894, issued July1, 1999.

! ! !

Biographies

Bogdan Kasztenny received his M.Sc. and Ph.D. degrees from the Wroclaw University of Technology (WUT), Po-land. He joined the Department of Electrical Engineering of WUT after his graduation. Later he was with the SouthernIllinois University in Carbondale and Texas A&M University in College Station. From 1989 till 1999 Dr. Kasztennywas involved in a number of research projects for utilities, relay vendors and science foundations. Since 1999 Bogdanworks for GE Power Management as a Chief Application Engineer. Bogdan is a Senior Member of IEEE, has publishedmore than 100 technical papers, and is an inventor of 5 patents. His interests focus on advanced protection and controlalgorithms for microprocessor-based relays, power system modeling and analysis, and digital signal processing.

Gustavo Brunello received his Engineering Degree from the National University in Argentina and a Master in Engi-neering from the University of Toronto. He also attended a 2 year post-graduate course in Power Systems Engineering


Page 35 of 35

at Polytechnic of Turin (Italy). After graduation he worked for the National Electrical Power Board in Argentina wherehe was involved in testing and commissioning the 500 kV backbone transmission system. From 1990 until 1999 heworked with ABB Relays and Network Control both in Canada and Italy where he became Engineering Manager for

protection and control systems. In 1999, he joined GE Power Management as an application engineer. He is responsi-ble for the application and design of protective relays and control systems. He is a Professional Engineer in the Prov-

ince of Ontario and a member of PES of the IEEE.

!""=

=

( )[ ] #"!""

[ ]( )[ ]

""$%

"="#"

( )[ ]( )[ ]( )[ ]

&"$%"&=&

&"$%"&=&

&"$%"&=&

'

'

( )[ ] ( )[ ] !"+""+=

'

Tim e (Seconds)

A-P

hase

Vol

tage

(pu)

0 0 .1 0.2 0.3 0.4

-1

0

1

(

(

))

)

)

)

)

!!!!!

'

Tim e (Seconds)

A-P

hase

Vol

tage

(pu

)

0 0 .1 0.2 0.3 0.4

-1

0

1

0 0.1 0.2 0.3 0.40

10

20

30

40

50

60

Tim e (Seconds)

Freq

uenc

y (H

z)

* *

"+,-

./0

!!=*

"1""= !!

"1""=!

( )

+"=

2222 3333

4

4

)2

3

2

3

0 0.1 0 .2 0.3 0 .4 0.5 0.6 0 .7 0 .8 0 .9 1-1 .5

-1

-0 .5

0

0.5

1

1.5

T im e (Seconds)

A-P

hase

Vol

tage

(pu

)

0 0.1 0 .2 0.3 0 .4 0.5 0.6 0 .7 0 .8 0 .9 1-200

-100

0

100

200

300

T im e (Seconds)

Fau

lt C

urre

nt (

pu)

0 0.2 0.4 0.6 0 .8 1 1.2 1.4 1.6 1 .8 20

0.5

1

1.5

T im e (Seconds)

A-P

hase

Vol

t Mag

(pu)

0 0.2 0.4 0.6 0 .8 1 1.2 1.4 1.6 1 .8 2-250

-200

-150

-100

-50

0

50

T im e (Seconds)

A-P

hase

Vol

t Ang

(de

g)

24

0 .5

1

1.5

30

210

60

240

90

270

120

300

150

330

180 0

1

Abstract- This article presents a study about the behavior of the EHV (Extra High Voltage) lines protection system installed in UTE. Our aim is to be able to detect errors on the distance algorithms of the protection equipment in order to consider them in the setting calculation stage. Usually this error is considered like a fixed percentage of the longitude of the line without taking in account neither the algorithm of the relay nor the characteristics of the electric system.

At the same time we developed a set of programs using MATLAB® and its SIMULINK® library for studying phase to ground faults in a transmission line and the algorithms of the distance protection relays. When we could not get the algorithm we used voltage and current values which were obtained in the simulation for testing a real distance protection relay.

I. NOMENCLATURE UTE - Administración Nacional de Usinas y Trasmisiones

Eléctricas, Uruguay. EHV- Extra High Voltage.

II. INTRODUCTION hen a phase to ground fault with resistance is produced in a line with bilateral feeding, the impedance seen by

the relay can differ a lot from the impedance of the section from the line to the fault. This can cause a difference in the resistance as well as in the reactance that must be compensated by the relay calculus when estimating the distance to the fault. It is important to know this compensation for determining the settings of the distance relay zone so not to overestimate or underestimate the error of the algorithm. This was the main reason for our work.

III. DEVELOPMENT In order to determine the behavior of distance relay’s

algorithms the first thing we did was to simulate a transmission line with bilateral feeding, where values of voltage and current seen by the relay during the fault are obtained. The protection system of lines of the EHV of UTE is mostly made up of distance relays from two different generations and manufac1turers: the first are of solid-state type and correspond to the PDTS 1453D model of ENERTEC Schlumberger [2], while the latter are microprocessor based

This work belongs to the Final Project corresponding to the career of Electric Engineering, and it was directed by Jorge Alonso. It came to be because of the state made by the Protection Department of UTE about their necessity of having a more profound knowledge concerning the behavior of the Protection system for the EHV power system.

relays of ABB, REL531 model [4]. According to the information brought up by the manufacturer about the relay algorithm we proceeded in different ways as it will be considered later.

A. Simulation of the fault We used a simple model of the power system for the

simulation of faults to ground in an EHV line. This model is made up by a distributed parameters line with two generators in both independent extremes (Fig 1), this was implemented in Simulink of Matlab®.

Real EHV lines were modeled. In order to do so we had to set the power system model in Simulink® with a short-circuit and power-flow program that is used nowadays in UTE and which has already modeled the whole Uruguayan electric system. The approximated Thevenin equivalent impedances were found with the short-circuit program, there we took off the line and we obtained the equivalents in the extremes. The values of the Thevenin sources, modules and phase angles were adjusted in such a way that the pre-fault voltages and currents would become similar between the results of the short-circuit program and the regime values of our Simulink® model.

Fig. 1. Power system model used for the simulation in Simulink Matlab®.

Although the model considers two independent sources for

simulating the feeding in both extremes (which is not correct for a meshed net) the comparison among the results we got allowed us to take this model like an acceptable one for our aims. The type of fault simulated was a phase to ground with constant resistance during the transient. Several simulations were made: varying the resistance of each fault, varying the location of it in the line and the previous conditions of charge, in order to obtain the currents and voltages of polarization of each relay and therefore determine the behavior of the protection system in the line investigated. To be able to perform these simulations we developed a set of applications in the GUI Modelred program (a software tool developed using Matlab 5.2). In the graphic user interface it is possible to

Behavior of the protection system of Extra High Voltage lines in faults with resistance to ground

Fernando García, UTE, Francisco Ashfield, Verónica Azevedo, UTE, and Jorge Alonso, IIE Univ. Mayor, UTE,Uruguay.

W

2

enter in a direct form all the parameters of the different components of the modeled power system as well as to obtain the currents and instantaneous voltages during each fault, moreover, it allows us to store the results in COMTRADE format. Fig. 2 shows the user’s interface, where it is possible to appreciate some of the parameters and results obtained (voltages and currents).

Fig. 2. User Interface of the application GUI Modelred.

The output of the GUI Modelred application are several files with the currents and voltages seen in both extremes. This allows us to separate the part of the net simulation and the protection system, and by doing so the data of currents and voltages are now easy to obtain either by this or any other application, or by using data obtained in registers installed in the electric net. It also allows the obtained files to be used in different simulations of the protection system or even to test a real relay.

B. Relay PDTS After simulating the power system we tried to determine the

behavior of the algorithm of the relay when facing resistive faults to ground with previous current. We particularly studied the possible over or under-reach that can appear in this kind of faults. To do so we did not take into account the characteristic in resistive direction (Fig. 3). In addition, only the behavior of the first and the second zone were studied. From this relay model we obtained abundant information about the algorithm being used and the way in which it is implemented. For the successful fulfillment of the analysis of this algorithm we followed two ways, the analytical and the dynamic simulation.

The analytic way, from the equations described in the relay manual, in which one can only work with phasors and not with instantaneous values. The electric power circuit was resolved analytically in order to obtain the currents and voltages that the relay sees during a phase to ground fault with a Rf resistance varying from 0 to 100 ohm. Afterwards these magnitudes were entered in the relay equations for determining if the relay trips in the first or second zone.

R

X

Z1

Z4

Z3

Z2

Fig. 3. PDTS distance relay operating principles (Static characteristic).

The algorithm used by the relay is the following: take the

instantaneous voltage of phase A, for instance: va, is compared with the voltage wa in the passage of current i0 through 0, where: wa = y [ r (ia-i0) + l (ia'-i0') + l0 i0' ] , with r, l y l0 the parameters of the line and y the setting of the relay. Where ia and i0 are phase A and zero sequence currents, ia' and i0' are the time derivatives.

The passage from i0 through 0 deletes the influence of the fault resistance, we take as an hypothesis that the current in the fault has the same phase than i0 seen by the relay, this applies if the factors of distribution of the zero sequence current have the same argument, which happens approximately in EHV nets.

If the va voltage is bigger than wa in the passage of i0 from the negative semi-cycle to the positive one, then the fault is placed out from the setting zone and vice versa.

It is seen that for low resistance the relay algorithm operates correctly. (Fig. 4). For high resistance in some cases we have wrong operations, although we must bare in mind that we are just modeling the reactive part of the characteristic of the

Fig. 4. Impedance seen in the fault, ‘+’ does not operate, ‘o’ shoots.

relay. In order to get the total behavior we should also model

Resistance ( Ohm )

Re

acta

nce

(

Oh

m )

3

the resistive reach, which will limit the operation in the resistive axis.

The dynamic way. During the second stage we studied the dynamic behavior of the relay in the same power system used for the first analysis, but instead of using phasors we used instantaneous values. These values were obtained from the simulation of faults in the power system, just like it was explained before in A. The tool used was the Simulink Matlab with its power and digital library.

The relay operation principle was simulated based on the description from the manual (Fig. 5).

Fig. 5. Schematic model of the PDTS used in Simulink.

The relay works with squared waves, which are obtained from the analogous signals filtered. The filter generates a digital sign which value is "1" when the analogous sign of entrance is positive and "0" in the opposite case.

For the relay to recognize a fault in zone 1 the coming out of the flip-flops must be in logic level "1". In the case from the graphic, in Fig. 6, the relay recognizes the fault approximately about 30ms after the production of the fault in t = 80ms. In the same graphic we can appreciate the variation of the current and voltage in the phase A when the fault appears, in addition the binary signal of Trip is superimposed.

Fig. 6. PDTS trip, RMS current and voltage.

We evaluated the operation or no-operation of the relay

when varying the resistance of fault and the previous current (importation and exportation). Moreover, we studied the validity of the following approximation, which was proposed by L. Mouton and M. Souillard [3].

What we did in this approximation was to correct the static characteristic of the relay with the angle that exists between the current of phase Ia and the one that circulates through the resistance of the fault If.

Table I shows one case of fault simulation (fault in 70% of line, exporting power), and the approximation to dynamic characteristic with zone 1 setting in 80% of the line. There we see the fault resistance, zone 1 trip, the fault impedance seen by the relay, the angle used to correct the static characteristic, and the limit of zone 1 in the reactance direction (X corrected).

In Fig. 7 is represented the impedance seen by the relay (a triangle if the relay has operated in Zone 1) and X corrected (a circle and a line). We can see that the relay tripped in all cases, although some of the fault impedances are over the limit proposed (X corrected).

TABLE I

FAULT F-T IN 70% OF LINE, EXPORTING POWER

Rf Z1 Trip Z seen (x + j y) arg(If)-arg(Ia) X corrected

0,01 Yes 4,8 42,7 -26,75 48,7

10 Yes 18,5 40,2 -25,5 42,2

20 Yes 30,2 37,9 -24,5 37,1

30 Yes 41,1 35,7 -23,7 32,7

50 Yes 58,9 32,0 -22,5 26,3

75 Yes 76,7 27,8 -20,9 21,2

100 Yes 90,6 24,3 -20,1 17,3

200 Yes 125,9 14,6 -18,3 8,6

0,0

10,0

20,0

30,0

40,0

50,0

60,0

0,0 20,0 40,0 60,0 80,0 100,0

relay operation

Dinamic Characteristic

approximation

Resistence ( Ohm )

Reac

tanc

e (

Ohm

)

Fig. 7. Impedance fault seen by de relay and the approximated characteristic proposed by Mouton and Souillard.

We developed the application PDTS with the aim of evaluating the calculus characteristics of the impedance seen of the distance relay PDTS 1453D of ENERTEC Schlumberger. This is made up on one hand by the simulation file Simulink, and on the other hand by the graphic interface (Fig. 8), from which the data and adjustments for performing

4

the simulation will be entered

Fig. 8. User’s interface of the application PDTS.

This application needs, in order to perform their own simulations or calculus, the data of instantaneous voltages and currents during a fault. It is for this reason that before its usage it is necessary to run first a simulation in GUI Modelred.

As we saw above, the algorithm of the PDTS stays immune against the variations of the seen impedance caused by the resistance of fault and the previous current to the fault. This is proved for the rank of values of the interest resistance: 0 to 100 ohms approximately.

It is also observed that the correction used for the static characteristic of the relay proposed by L. Mouton and M. Souillard, can be used like a first approximation only for values of low Rf..

C. Relay REL531 Because the poor information available about the distance

algorithm of the REL5XX relay of ABB we made some tests on a relay model REL531 for determining its characteristic of operation.

The working procedure was the following: • We simulated different faults F+T in a line of 500kV and

we obtained the currents and voltages seen by the relay. The tool used was Simulink of Matlab.

• With the obtained data in the previous stage we generated files within an IEEE Comtrade format.

• We set the relay according to the data of the line that was simulated.

• By using a test system we injected in the relay the voltages and currents contained in the Comtrade IEEE files.

• We analyzed the relay registers for determining the relay behavior.

The faults were simulated up to 75 % and 85 % of the line varying the resistance of fault and the extreme that exported

power, while the relay was set up to 80 %. By doing this we were searching to determine how far the relay compensated the effect of the previous current when varying the fault resistance.

Table II is a summary of fault simulated. There we see the previous condition of power flow, the fault resistance, the fault impedance calculated in the moment of the fault following the formula ZV = Ua/ (Ia + Ko * 3Io), and zone 1 trip.

In Fig. 9 and Fig. 10 is represented the impedance seen by the relay (a cross if the relay operated in Zone 1 and a circle in the opposite case). We also used dashed lines for the interpolation of the impedance seen for faults up to 75 % and 85 % of the line.

TABLE II

Previous condition

Fault location

(%)

R fault Z seen (R,X) Zone 1 Trip

Exporting 85 1 7,42 51,46 NO

Exporting 85 10 24,54 47,22 NO

Exporting 85 100 109,50 22,71 NO

Exporting 75 1 6,36 45,43 Yes

Exporting 75 10 19,97 42,73 Yes

Exporting 75 100 96,22 24,34 Yes

Importing 85 1 7,55 52,64 NO

Importing 85 10 27,10 61,46 NO

Importing 85 100 -112,41 747,74 NO

Importing 75 1 6,42 46,25 Yes

Importing 75 10 21,22 52,52 NO

Importing 75 100 217,36 438,45 NO

0 20 40 60 80 100 1200

10

20

30

40

50

60

70

Resistance (ohm)

Reactance (ohm )

Fig. 9. Line exporting power.

5

0 5 10 15 20 25 30 35 40 450

10

20

30

40

50

60

70

Resistance (ohm)

Reactance (ohm )

Fig. 10. Line importing power.

It is possible to see that whenever the relay is placed on the side of the exportation source of energy there are neither problems of under-reach nor problems of over-reach, because all the faults up to 85% stay out from zone 1, and all the faults up to 75% fall within zone 1. However, when we are in conditions for importation of energy the behavior of the relay is not that good any longer, it under-reaches for resistive faults. For the case of faults up to 75% with resistance of fault 10 and 100 Ω the relay sees them out of Zone 1, and for the faults up to 85% it behaves correctly.

With this research we determined that the relay algorithm only compensates the reactance when it exports energy, and not in the opposite case. As a result this is the model that will be used in future studies.

IV. CONCLUSIONS The programs developed in Matlab are a tool for the study

of the behavior of the protection system, and they allow us to anticipate to any possible miss functioning of the protections and therefore to correct them if possible. In these programs an operation of different distance relays is simulated, and in this way we can evaluate their behavior with different settings. This allows us to quantify the algorithm errors during different fault conditions in order to consider the error in the adjustment stage of the protections, or in an analysis of a real fault.

V. ACKNOWLEDGMENTS We thank for the help given by the Protection Department of UTE. We also want to thank Engineer Mr. Juan Zorrilla de San Martin for his contributions.

VI. REFERENCES Technical Reports:

[1] L. P. Cavero, "Analysis of complex distance relay characteristics taking load into account", General Electric Company, USA.

[2] "Protection de distance statique contre les défauts phase-terre PDTS 1453D, Description et fonctionnement’, Département Réseaux Electriques, ENERTEC Schlumberger.

[3] L. Mouton and M. Souillard, "High speed static relays for distance measurements", Compagnie des Compteurs, Electricity Departament, Protective Systes and Relay Division, France.

[4] "REL 531*2.0, Line distance protection terminal, Technical reference manual", ABB Network Partner AB, September 1998.

[5] Jay Gosalia and Dennis Tierney, "Tutorial: Using COMTRADE Files for Relay Testing", Doble Engineering Company.

VII. BIOGRAPHIES

Fernando García was born in Montevideo, Uruguay, in 1973. He received his Electrical Engineer degree from the Universidad Mayor de la República Oriental del Uruguay, in 2001.

He joined the Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.) in 1995 and worked in protection systems area since then. His main activity has been power system protection studies and projects.

He is a member of IEEE since 1999.

Francisco Ashfield was born in Santa Isabel, Guinea Ecuatorial, in 1973. He received his Electrical Engineer degree from the Universidad Mayor de la República Oriental del Uruguay, in 2001.

His main activity has been electrical engineer projects. He is a member of IEEE since 2000.

Verónica Azevedo was born in Montevideo, Uruguay, in 1971. She received her degree in Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, in 2001.

She joined the Administración Nacional de Usinas y Transmisiones Eléctricas (UTE) in 1994, where she held the position of engineering in the System Protection Engineering section. Her main activity has been power system protection studies and disturbance analysis.

She is a member of IEEE since 2000.

Jorge Alonso was born in Montevideo, Uruguay, in 1956. He received the Engineer and MSc. degrees all in Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, in 1979 and 1998, respectively. Since 1980, he has been with the Institute of Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, now as an Aggregate Professor. Since 1979, he has been working for the Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.), where he is currently a general manager of control and protection power transmission system section. His main research and development activity has been on the simulation of numerical relays and the dynamic performance of electrical machines.

He is a member of the IEEE since 1987.

— 1 —

Protección de Interconexiones de Generadores de IPPUsando Tecnología Digital

Autor y Expositor: Charles J. MozinaBeckwith Electric Co., Inc.

Gerente de Aplicaciónes, Productos y Sistemas de Protección6190-118th Ave. North, Largo, FL 33773-3724 U.S.A.

Tel. (727) 544-2326 Fax (727) 546-0121E-mail: [email protected]

IntroducciónMucha de la nueva capacidad de generación con que se contará en el nuevo milenio se alcanzará

construyendo instalaciones de IPP (Independent Power Producer = Productor Independiente deEnergía Eléctrica). Dichas instalaciones pueden consistir en pequeñas unidades generadoras dispersas,o en plantas de gran capacidad de propiedad de (y generalmente operadas por) personal ajeno a lasempresas eléctricas. Este artículo analiza los requisitos de protección para interconectar estos generadoresa los sistemas de las empresas eléctricas, así como los métodos para reconectar estos generadoresluego del disparo de la protección de la interconexión. El artículo comenta asimismo las limitacionesde los métodos actuales de protección de interconexiones en aspectos tales como el respaldo delsistema de generación durante perturbaciones importantes en el sistema de la empresa eléctrica.

Es necesario proteger los generadores de IPP no sólo contra los cortocircuitos, sino contra lascondiciones anormales de operación. Muchas de estas condiciones anormales pueden ser impuestasen el generador de IPP por el sistema de la empresa eléctrica. Algunos ejemplos de dichas condicionesanormales: sobreexcitación, sobrevoltaje, corrientes desequilibradas, frecuencia anormal y esfuerzotorsional del eje debido al recierre automático de un interruptor de la empresa eléctrica. Al estarsometidos a estas condiciones, los generadores pueden, en pocos segundos, sufrir daños o fallacompleta. Los daños a las máquinas debidos a estas causas son una gran preocupación de lospropietarios de generadores de IPP.

Las empresas eléctricas, por su parte, se preocupan porque la instalación de generadores de IPPpuede resultar en daños a sus equipos o a los equipos de sus clientes. Los generadores pequeñosdispersos están conectados al sistema de la empresa eléctrica en los niveles de distribución ysubtransmisión. Estos circuitos de la empresa eléctrica están diseñados para alimentar cargas radiales.La introducción de generadores constituye una fuente indeseada de redistribución de corrientes decarga y de falla, así como una posible fuente de sobrevoltaje. Por lo general no se permite laoperación en isla [con formación o fraccionamiento en islas] de generadores de IPP dispersos concargas de la empresa eléctrica externas al sitio del IPP, por dos razones importantes:

1. La empresa eléctrica debe restaurar los circuitos interrumpidos, y este esfuerzo se complicamucho cuando hay generadores en isla con cargas de la empresa eléctrica. El recierre automáticoes generalmente el primer método que se intenta para restaurar energía eléctrica a los usuarios.Al haber generadores en isla, se complica el recierre automático y también la conmutaciónmanual que requiere sincronizar el generador/carga en isla al sistema de la empresa eléctrica.

2. La calidad de la energía (los niveles de voltaje y frecuencia, así como las armónicas) puede noser mantenida por los generadores de IPP en isla al nivel ofrecido por la empresa eléctrica, loque puede resultar en daños a los equipos de los usuarios.

La protección de interconexiones correctamente diseñada debe atender los factores que preocupanal propietario del IPP así como a la empresa eléctrica - al menor costo posible. La función principalde la protección de interconexiones es evitar la formación de islas en el sistema detectando laoperación asincrónica de generadores dispersos — en otras palabras, deberá determinar si el generadorha dejado de operar en paralelo con el sistema de la empresa eléctrica. La detección y el disparodeberán ser lo suficientemente rápidos para permitir el recierre automático en el sistema de laempresa eléctrica.

— 2 —

Los grandes generadores de IPP están por lo general conectados a sistemas de transmisión deempresas eléctricas. En la mayoría de los casos, las configuraciones unifilares del sistema y de laprotección son idénticas a las de los generadores de la empresa eléctrica. Cuánto mayor sea elgenerador, más probable será que esté integrado al sistema de la empresa eléctrica bajo las mismasreglas que los generadores de la misma. Estos grandes generadores pueden proporcionar respaldo degeneración a la empresa eléctrica durante perturbaciones importantes en el sistema. Para los IPPsgrandes y medianos se requiere telemetría, que proporciona a la empresa eléctrica información básicasobre la operación del generador. Cuanto mayor sea el generador, más información va a requerir laempresa eléctrica.

La tecnología disponible para ofrecer protección a los IPP ha evolucionado desde los reléselectromecánicos de función única a los relés estáticos (electrónicos) y ahora a los relés digitales. Laaparición de las tecnologías de microprocesadores de bajo costo ha posibilitado desarrollar los relésdigitales de multifunción, que combinan numerosas funciones de protección en un conjunto único derelés. Esta tecnología de relés ofrece ventajas de importancia sobre los antiguos relés electromecánicosy estáticos. Este artículo enfatiza el uso de dicha tecnología para ofrecer protección de interconexiones.Los otros tópicos específicos a tratar son los siguientes:

Breve Historia de la Generación de IPP en los Estados UnidosInfluencia del PURPAEstado actual de la generación de IPPTecnología de microturbinas

Protección de Interconexiones versus Protección de GeneradoresGrandes Generadores de IPP

Configuraciones de transformadores de interconexiónDiagramas unifilares de interconexiones

Pequeños Generadores de IPP DispersosImportante impacto de las conexiones de transformadores de interconexión sobre los requisitos deprotecciónSobrevoltajes transitorios producidos por IPPs en los sistemas de distribución de las empresaseléctricas y medidas atenuantesMétodos de detección de operaciones asincrónicas de IPPs con sistemas de las empresas eléctricasLimitaciones de los procedimientos actuales para permitir a los generadores de IPP dispersosproporcionar respaldo de generación durante perturbaciones de gran magnitud en el sistemaProcedimientos de restauración automática y recierre automático de las empresas eléctricas

Métodos y Procedimientos de Protección de Interconexiones de Pequeños Generadores DispersosDetección de la pérdida de operación en paralelo con la empresa eléctricaDetección de contraalimentación de fallasDetección de condiciones perjudiciales en el sistemaFlujo de potencia anormalRestauración

Uso de la Tecnología Digital para la Protección de InterconexionesVentajas de esta tecnologíaFuncionalidad seleccionable por el usuarioAutodiagnósticoCapacidad de comunicacionesCapacidad oscilográfica

Breve Historia de la Generación de IPP en los Estados UnidosHasta los últimos años de la década del 70, las empresas eléctricas no estaban obligadas a comprar la

energía eléctrica generada por entes ajenos a las empresas eléctricas dentro de sus áreas de servicio. Sinembargo, existían industrias, tales como las de la pulpa y el papel y la siderúrgica, así como las instalacionespetroquímicas, que contaban en sus instalaciones eléctricas con generación de IPP interna y queoperaban en paralelo con el sistema de la empresa eléctrica. Estos “cogeneradores” producíanelectricidad a partir de fuentes de calor, como ser los vapores originados en procesos fabriles.

— 3 —

Típicamente, estos generadores alimentaban parte de la carga en dichas instalaciones industriales ysuministraban energía de emergencia a las mismas durante interrupciones del servicio de la empresaeléctrica.

Luego del embargo petrolero de principios de la década de 1970, el gobierno federal estadounidensedecidió que las fuentes de energía convencionales, especialmente el petróleo y el gas, debían ser conservadasde modo de reducir nuestra dependencia de fuentes extranjeras. El gobierno federal deseaba promover lageneración de electricidad mediante fuentes de combustible renovables usando generadores ajenos a lasempresas eléctricas. Ello hizo que se pasara el Public Utility Regulatory Policies Act (PURPA) [Ley sobreNormas Regulatorias de Empresas de Servicios Públicos] de 1978. El PURPA exigió a las empresas eléctricas,por primera vez, que se interconectaran con IPPs calificados y que adquieran electricidad a un costo quereflejara el costo ahorrado por la empresa eléctrica al no tener que generar una magnitud equivalente deelectricidad por sí misma.

Para recibir los beneficios del PURPA, el IPP debía calificar el sitio de generación propuesto ya seamediante autocalificación o por certificación de FERC (Federal Energy Regulatory Commission) [ComisiónFederal Reguladora de la Energía]. La autocalificación era el método de aprobación más fácil ya que sólorequería una carta a FERC documentando que las instalaciones propuestas cumplían con los requisitos deelegibilidad del PURPA. Si bien la intención del PURPA era conservar recursos de petróleo y de gas, ciertascláusulas en la ley permitían que tales combustibles fueran usados como combustible primario por lasQualified Facilities (QF) [Instalaciones Calificadas]. El gas natural, en particular, se convirtió en un com-bustible muy usado por las instalaciones IPP calificadas.

El PURPA creó también un segundo tipo de IPP, el generador ajeno a la empresa pública, cuyo úniconegocio era vender energía a la empresa eléctrica con fines de ganancia. En las décadas de 1980 y 1990, almenguar los márgenes de reserva en las empresas eléctricas, algunas de estas empresas comenzaron a invitara IPPs a suministrar capacidad adicional para sus sistemas. Dichas empresas eléctricas consideran a los IPPscomo una alternativa viable respecto a la construcción de sus propias plantas generadoras, evitando así laasignación de un monto considerable de su capital con ganancias inciertas—dadas las dificultades ocasionadaspor el proceso regulatorio.

En los 1990s, al haber una mayor desregulación de las empresas eléctricas, el rol de los IPPs se hizo aúnmás importante. Algunas empresas eléctricas decidieron convertirse en compañías de transmisión y/o distribucióny se deshicieron de sus sistemas de generación—adquiriendo desde entonces la energía eléctrica de los IPPs.Otras empresas eléctricas se dedicaron al negocio de los IPPs. El acceso más abierto a la transmisiónpermite a los IPPs en una determinada zona de servicio vender energía a empresas eléctricas fuera de dichazona. En algunos casos, las empresas eléctricas actúan como agentes comerciales de energía para facilitarestas ventas de energía eléctrica y beneficiarse con ellas. Los Independent Power Operators (IPOs) [Explotadoresde Energía Independientes] regionales, como los que hay en California y en Alberta, Canadá, cuentan con“tableros de anuncios computarizados” que ofrecen capacidad generadora disponible y precios en base horaria.Se listan las capacidades de generación disponibles tanto de las empresas eléctricas como de los IPPs.

Como los costos del gas natural han caído, hay un mayor interés en las pequeñas microturbinas dispersas.Algunos de estos generadores de imán permanente de alta velocidad utilizan turbinas de tecnología avanzadadesarrolladas originalmente para vehículos militares. Actualmente se fabrican en el rango de 20 a 200 KVA,y son conectadas al sistema eléctrico de clientes comerciales para “repartir la carga en picos de demanda”con la empresa eléctrica. El reparto en picos permite a la instalación comercial reducir sus cargos pordemanda. Algunos expertos de la industria piensan que las microturbinas tendrán un rol importante atendiendolas demandas de carga en el próximo milenio. Algunas empresas eléctricas, aprovechando las nuevas leyesdesregulatorias, están involucradas en la comercialización de generadores de microturbina dispersos. Con elcorrer del tiempo se comprobará si estos tipos de generadores dispersos serán una fuente de energía viablepara usuarios comerciales e incluso residenciales.

Los requisitos de protección de las interconexiones también han evolucionado a través de los años.En los 1980s, el IEEE participó en el desarrollo de recomendaciones y directivas para la interconexiónde generadores de IPP. La Norma ANSI/IEEE 1001-1988 [1] proporcionó las directivas básicas queadoptaron muchas empresas eléctricas. Hacia 1990, la mayoría de las empresas eléctricas de losEE.UU. había publicado directivas específicas para la conexión de pequeños generadores de IPP

— 4 —

(generalmente de menos de 5 MW) a sus sistemas. Estas directivas casi siempre especifican proteccióncon relés de voltaje y frecuencia y requieren que los relés sean de “calidad [grado] tipo empresaeléctrica”—cumpliendo con las normas de diseño IEEE/ANSI C37.90. A través de los años, estosrelés han evolucionado de electromecánicos a estáticos, y finalmente a dispositivos de proteccióndigitales.

Protección de Interconexiones versus Protección de GeneradoresLa protección de interconexiones permite al IPP operar en paralelo con la red de la empresa

eléctrica. Típicamente, los requisitos de protección para conectar un IPP a la red son establecidos porcada empresa eléctrica específica. Estas normas cubren por lo general generadores de menor capacidad.Los generadores grandes se evalúan individualmente y por lo general se conectan al sistema detransmisión de la empresa eléctrica. Estos grandes generadores de IPP típicamente no requierenprotección específica de interconexión ya que están integrados al sistema de protección de laempresa eléctrica. Los pequeños generadores de IPP (de 5 MW o menos) habitualmente se conectana los sistemas de subtransmisión y distribución de la empresa eléctrica. Estos circuitos de la empresaestán diseñados para alimentar cargas radiales. Por ende, la incorporación del generador ofrece unafuente para redistribuir la corriente de falla y la carga del circuito alimentador, y es también unafuente potencial de sobrevoltaje. Típicamente, la protección de interconexiones para estos generadoresse establece en el punto de acoplamiento común entre la red de la empresa eléctrica y el IPP. Estepuede estar en el secundario del transformador de interconexión, como indica la Figura 1a, o en elprimario del transformador, como indica la Figura 1b, dependiendo de los requisitos de interconexiónde la empresa eléctrica y del propietario.

!"#$%&'()&*+,()-./$+"/'

!"#$%&'(0%-(122

34

54

GG

6%-.(0%17$%+/87%9":7

4+'7#;8+&'08+0%(17$%+/87%9":7

<-(!"#$%&'(0%(-'()&*+,()-./$+"/'

3'+='#(>8/'-%#GG

!"#$%&'(0%-(122

3'+='#(>8/'-%#

6%-.(0%17$%+/87%9":7

34

54

!"#$%&'()&*+,()-./$+"/'

<-(!"#$%&'(0%(-'()&*+,()-./$+"/'

4+'7#;8+&'08+(0%17$%+/87%9":7

Figura 1a Protección de Interconexión Típica Figura 1b Protección deInterconexión Típica

La protección de las interconexiones debe satisfacer los requisitos de la empresa eléctrica parapermitir que el generador sea conectado a la red. Su función es triple:

1. desconecta el generador cuando ha dejado de operar en paralelo con el sistema de la empresaeléctrica;

2. protege el sistema de la empresa eléctrica contra los daños ocasionados por la conexión delgenerador, incluyendo la corriente de falla que suministra el generador para fallas del sistema dela empresa y sobrevoltajes transitorios;

3. protege el generador contra daños producidos por el sistema de la empresa eléctrica, especialmentemediante el recierre automático.

— 5 —

La protección del generador típicamente se conecta en los terminales del generador, tal como semuestra en la Figura 2.

6%-.(0%2+8$%//":7(0%?%7%+'08+

34

54

G G3'+='#(>8/'-%#

4+'7#;8+&'08+(0%17$%+/87%9":7

<-(!"#$%&'(0%(-')&*+,()-./$+"/'

Figura 2 Típica Protección de Generador

La protección del generador permite detectar:1. cortocircuitos internos del generador;2. condiciones anormales de operación (pérdida de campo, potencia inversa, sobreexcitación y

corrientes desequilibradas).En el caso de los pequeños generadores dispersos, la mayor parte de las empresas eléctricas de los

EE.UU. dejan a los propietarios de IPPs y sus consultores la responsabilidad de seleccionar el nivel deprotección del generador que ellos consideran apropiado. Sin embargo, las empresas eléctricas, están participandoactivamente en especificar la protección de interconexiones. Los siguientes aspectos de la interconexión sontípicamente especificados por las empresas eléctricas:

1. configuración de los devanados del transformador de interconexión;2. requisitos generales para relés de interconexión de calidad tipo empresa eléctrica;3. requisitos para CTs (transformadores de corriente) y VTs (transformadores de voltaje);4. requisitos de protección funcional — por ejemplo, 81O/U, 27 y 59;5. ajustes de algunas funciones de interconexión;6. velocidad de operación.

Grandes Generadores de IPPLos grandes generadores de IPP se conectan a los sistemas de transmisión de las empresas eléctricas.

Estos grandes generadores están típicamente “conectados en unidad”—esto es, el generador alimenta directamenteun transformador elevador de generador (GSU), que es un transformador conectado en triángulo [delta] conpuesta a tierra en estrella [Y] como indica la Figura 3. El generador está típicamente conectado a tierra dealta impedancia para limitar la corriente de falla a tierra del estator.

— 6 —

!"#$%&'()&*+,()-./$+"/'

4+'7#;8+&'08+?!@

G

R

4+'7#;8+&'08+(0%<++'7AB%C<B9"-"'+

2B%#$'('(4"%++'(0%<-$'(1&*%0'7/"'

Figura 3 Típico Generador de IPP Conectado en Unidad

Hoy en día, muchos de estos grandes generadores consisten en turbinas a gas “integradas” en elrango de 100-350 MW. En algunos casos, son parte de una planta de ciclo combinado donde hay otrogenerador de vapor instalado en el mismo sitio y el escape de la(s) turbina(s) de gas se usa paracalentar el vapor y aumentar así la eficiencia del ciclo térmico general. El generador de vapor estípicamente más pequeño que la(s) turbina(s) de gas y también está conectado en unidad. La protecciónde estos generadores es generalmente proporcionada por el complemento normal de la protección delgenerador, como lo describe la “IEEE Guide for AC Generator Protection” (Guía del IEEE para laProtección de Generadores de CA) [5].

La conexión de estos grandes generadores de IPP al sistema de la empresa eléctrica varíasubstancialmente. Cuanto mayor sea el generador, más probable será que se lo conecte al sistema detransmisión de la empresa eléctrica de igual manera que un equipo generador perteneciente a dichaempresa. Estos grandes generadores no utilizan relés de frecuencia y voltaje para detectar la pérdidade la operación en paralelo con la empresa eléctrica. Estos equipos están completamente integradosal sistema de protección de la empresa eléctrica mediante canales de telecomunicaciones. La Figura4 muestra una típica configuración unifilar de un generador de IPP de tamaño mediano con tomas auna línea de transmisión de la empresa eléctrica.

— 7 —

!BD%#$'/":7(< !BD%#$'/":7(E

!"#$%&'(0%-(122

122

!"#$%&'()&*+,()-./$+"/'

A B

GFigura 4 Generador de IPP con Tomas a una Línea de Transmisión

Hay tres esquemas de protección por piloto de terminales, así como de disparos de transferencia,que se usan frecuentemente para la protección de alta velocidad contra fallas de línea. El recierreautomático de interruptores de subestación puede ser supervisado por relés de bajo voltaje y decomprobación de sincronismo, como indica la Figura 13. Los grandes IPPs están conectados alsistema de la empresa eléctrica mediante líneas múltiples. La Figura 5 ilustra una configuración en“barra recta” que proporciona múltiples entradas al sistema de la empresa eléctrica.

G

!"#$%&'(0%-(122

!"#$%&'()&*+,()-./$+"/'

Figura 5 Interconexión en “Barra Recta”

— 8 —

Otra configuración popular para grandes generadores de IPP consiste en la interconexión alsistema de la empresa eléctrica por medio de una subestación con un interruptor y medio o barra enanillo, como indica la Figura 6. La protección de estas diversas configuraciones unifilares es por logeneral idéntica a la protección que se emplearía si el generador fuera de propiedad de la empresaeléctrica.

G

FGB%H8(17$%++B*$8+

!"#$%&'(0%-(122

!"#$%&'()&*+,()-./$+"/'

Figura 6 Interconexión con Barra en Anillo

Pequeños Generadores de IPP DispersosTipos de Generadores Pequeños

Hay dos tipos tradicionales de pequeños generadores de IPP que operan interconectados con el sistema dela empresa eléctrica. Ellos son los generadores de inducción y los generadores síncronicos [síncronos]. Lasmáquinas de inducción son típicamente pequeñas—de menos de 500 KVA. Estas máquinas son de tamañorestringido porque su excitación es provista por una fuente externa de VArs, como muestra la Figura 7a. Losgeneradores de inducción son similares a los motores de inducción y se arrancan como motor (no requierenequipo de sincronización). Los generadores de inducción son menos costosos que los generadores sincrónicosporque no tienen devanados [arrollamientos] de campo. Las máquinas de inducción pueden suministrarpotencia real (watts) a la empresa eléctrica, pero requieren una fuente de potencia reactiva (VArs) que enalgunos casos es proporcionada por el sistema de la empresa eléctrica.

Los generadores sincrónicos tienen un devanado de campo de CC que proporciona una fuente de excitacióna la máquina. Pueden ser una fuente de watts y de VArs para el sistema de la empresa eléctrica, comomuestra la Figura 7b, y requieren equipo de sincronización para la puesta en paralelo con la red eléctrica.Ambos tipos de máquinas requieren protección de interconexión. La protección de interconexión pertinentea los generadores de inducción por lo general requiere únicamente relés de sobre/bajo voltaje y de frecuencia.

— 9 —

GI<44!

5<+#

J)(1GJ@331KGL((%9/"$'/":7(*+8*8+/"87'0'(%9$%+7'&%7$%L(('++'7/'(/8&8(B7(&8$8+((((M8(#%'N(78(+%AB"%+%(%AB"*8(0%(#"7/+87"O'/":7PL((&%78#(/8#$8#8(AB%(-'#(&QAB"7'#(#"7/+:7"/'#

!"#$%&'

RB%7$%(0%(5<+#

RB%7$%(0%(5<+#

GI<44!

5<+#

!1G36KG13SL((%-(/'&*8(0%(33(*+8*8+/"87'(-'(%9/"$'/":7L((#%(+%AB"%+%(#"7/+87"O'+-8('-(#"#$%&'

0%(-'(%&*+%#'(%-./$+"/'

!"#$%&'

L((((((T

RB%7$%(0%(33

J%H'7'08(0%(3'&*8

Figura 7a Generador de Inducción Figura 7b Generador Sincrónico

Los pequeños generadores de IPP de tipo no tradicional, especialmente los de nueva tecnología demicroturbina, son considerados cada vez más frecuentemente como una fuente de energía para elpróximo milenio. La mayoría de estas máquinas se conectan asincrónicamente al sistema de energíapor medio de Convertidores Estáticos de Potencia (SPCs). Estos SPCs on dispositivos tiristorizadoscontrolados por microprocesador que convierten voltaje de CA en una dada frecuencia en voltaje desistema de 60 Hz. El control electrónico digital del SPC regula la salida de potencia del generador ydetiene la máquina cuando el sistema de la empresa eléctrica no está disponible. No se determinó aúnsi es necesario contar con protección independiente para evitar la formación de islas en el sistema,pero está ello está siendo considerado por el Standards Coordinating Committee 21 (SCC-21) [ComitéCoordinador de Normas] del IEEE. Al ir aumentando el tamaño de estas máquinas, quizás se haránecesario considerar la protección independiente de las interconexiones. La Figura 7c muestra undiagrama unifilar típico para estos tipos de generadores.

GI<44!

<!1G36KG13SL((%-(/87H%+$"08+(%#$Q$"/8(0%(*8$%7/"'(M!23P(/87H"%+$%

-'(;+%/B%7/"' 0%-(=%7%+'08+(%7(;+%/B%7/"'(0%-(#"#$%&'L((%-(=%7%+'08+(%#$Q(/87%/$'08('#"7/+:7"/'&%7$%

'-(#"#$%&'(0%(%7%+=U'

!"#$%&'!23

)7-'/%(<#"7/+:7"/8

Figura 7c Generador Asincrónico

Principal Impacto de las Conexiones de Transformadores de Interconexión en la Protección de InterconexionesComo se ha mencionado en la sección anterior, la función principal de la protección de interconexiones es

desconectar el generador cuando ha dejado de operar en paralelo con el sistema de la empresa eléctrica. LosIPPs pequeños se conectan generalmente al sistema de la empresa eléctrica a nivel de distribución. En losEstados Unidos, los sistemas de distribución cubren un rango de 4 a 34.5 KV y son sistemas de 4 hilos conmúltiples conexiones a tierra. El uso de este tipo de sistema permite que los transformadores monofásicosmontados en poste, que típicamente constituyen la mayor parte de la carga del alimentador, sean clasificadospara voltaje de fase a neutro. Así, en un sistema de distribución de 13.8 KV, los transformadores monofásicosestarían clasificados a 13.8 KV/√ 3~8 KV. La Figura 8 muestra un circuito alimentador típico.

— 10 —

G

G

<

1G!4<><31SG)!(J)>(122!@E)!4<31SG(J)(J1!461E@31SGV)W26,()>)34613<

3<6?<

3<6?<

3<6?<

<

E

34+'7#;8+&'08+(%7(28#$%(387%/$'08R'#%LG%B$+8

!"#$%&'(0%(3B'$+8(X"-8#

3870B/$8+(G%B$+8

Figura 8 Típico Circuito Alimentador de Distribución de 4 Hilos

Hay cinco conexiones de transformador que se emplean frecuentemente para interconectar generadoresdispersos al sistema de la empresa eléctrica. Cada una de estas conexiones de transformadores tiene susventajas y sus desventajas. La Figura 9 muestra varias posibles opciones y algunas de las ventajas y problemasrelativos a cada tipo de conexión.

— 11 —

G122

3<6?<

3<6?<

RY

RZ

3<6?<

3<6?<

R[

\387%9"87%#(0%-4+'7#;8+&'08+(0%17$%+/87%9":7

A

2B%0%( #B&"7"#$+'+( '-/"+/B"$8('-"&%7$'08+(0%#0%B7'( ;B%7$%( 78( *B%#$'( '$"%++'(-B%=8(0%-(0"#*'+8(0%-"7$%++B*$8+(0%(#BD%#$'/":7< N ( 8 / ' # " 8 7 ' 7 0 8#8D+%H8-$']%,

2+8*8+/"87'(B7'(/8++"%7$%0%( $"%++'( "70%#%'0'( '7$%;' - -'#( 0%- ( / " +/B" $8( 0%#B&"7"#$+8( %7( RZ( ^( RY,

2% +& " $ % ( ' - ( + % - . ( 0% -'-"&%7$'08+(*+"&'+"8(%7(<+%#*870%+('(B7'(;'--'('($"%++'0%-( #%/B70'+"8( %7( R[,

>8_58-$'=%M!%/P

X"=`58-$'=%M2+"P

2+8D-%&'#

G 8 ( * + 8 * 8 + / " 8 7 '/87$+''-"&%7$'/":7(0%(;'--''($"%++'('7$%(;'--'#(%7(RZ(^RY,( G8( `'^( /8++"%7$%( 0%$"%++'(0%#0%(%-("7$%++B*$8+(<'7$%( B7'( ;'--'( %7( R[,

5%7$']'#

G8(̀ '^(#8D+%H8-$']%('7$%(B7';'--'('($"%++'(%7(RZ,

G8(`'^(/8++"%7$%(0%( $"%++'0%#0%(%-("7$%++B*$8+(<('7$%;'- -'#( %7( R[,( G8( `'^#8D+%H8-$']%( '7$%( ;'--'#( '$ "%++'( %7( RZ,( G8( `'^#8D+%H8-$']%('7$%(B7'(;'--'('$"%++'(%7(RZ,

F

F

F

!@E)!4<31SG(J)J1!461E@31SGV)W26,()>)34613<

Figura 9 Conexiones del Transformador de Interconexión

Conexiones del Transformador de Interconexión en Triángulo (Prim.)/Triángulo (Sec.),Triángulo (Prim.)/Estrella Puesta a Tierra (Sec.) y Estrella no Puesta a Tierra (Prim.)/Triángulo (Sec.)

La principal preocupación relativa a un transformador de interconexión con devanado primario no puestoa tierra es que luego que el interruptor de subestación A dispare ante una falla a tierra en la localización F1,el sistema con múltiples conexiones a tierra no esté conectado a tierra, sometiendo al transformador montadoen poste clasificado como L-N (fase a neutro), en las fases sin falla, a un sobrevoltaje cercano al voltaje L-L(entre fases). Ello puede ocurrir si el generador de IPP está cerca de la capacidad de la carga en el alimentadorcuando el interruptor A dispara. Los sobrevoltajes resultantes van a saturar el transformador montado enposte, que normalmente opera en el codo de la curva de saturación, como indica la Figura 10.

— 12 —

!"##$%&'%()%(*+,$'-,$.&

/"0'-1%

%2 %

Z,ab(*,B,Z,a(*,B,

/3455>LG

4+'7#;8+&'08+W87$'08(%7(28#$%

X"-8(0%(R'#%

G%B$+8

Figura 10 Curva de Saturación de Transformadores Montados en Poste

Muchas empresas eléctricas usan transformadores de interconexión sin puesta a tierra únicamentesi se experimenta una sobrecarga de 200% o más en el generador cuando el interruptor A dispara.Durante las fallas a tierra, este nivel de sobrecarga no permitirá que el voltaje en las fases sin falla seeleve en exceso del voltaje L-N normal, evitando la saturación del transformador montado en poste.Por esta razón, los devanados primarios sin puesta a tierra deberán por lo general reservarse parapequeños IPPs para los que se esperan sobrecargas de por lo menos 200% ante la formación de islas.

Conexiones del Transformador de Interconexión en Estrella Puesta a Tierra (Prim.)/Triángulo (Sec.)La principal desventaja de esta conexión es que proporciona una corriente de falla a tierra indeseada ante

las fallas del circuito de suministro en F1. Las Figuras 11a y 11b ilustran este aspecto para un circuito dedistribución típico.

— 13 —

VS

X

VIPP

F1 (Falla L-T)

XL1

XL2

Xsistema

Subestación-Empr. Eléctrica

Xd",X2,X0

IPP

XTSub.

Sub.A

XTIPP

F

!%/,(28#"$"H'

XL1 XsistemaXTsubXL2XTIPP

X"dgen

122 !BD,

F

!%/,(G%='$"H'

XsistemaXTsubX"2gen

122 !BD,

F

!%/,(3%+8

X0L1

XTsub

X0L2X"0gen

122 !BD,XTIPP

RZ

A

A

A

XL2XL1XTIPP

RZ

RZ

VSVIPP

Figura 11a Diagrama Unifilar para un Figura 11bCircuito de Componentes Simétricos

Transformador de Interconexión para un Transformador deInterconexión

con Estrella Puesta a Tierra con Estrella Puesta a Tierra(Prim.)/Triángulo (Sec.) (Prim.)/Triángulo (Sec.)

El análisis del circuito de componentes simétricos en la Figura 11b indica también que aún cuando elgenerador de IPP está fuera de la línea (el interruptor del generador está abierto), la corriente de falla a tierraseguirá siendo proporcionada al sistema de la empresa eléctrica si el transformador de interconexión del IPPpermanece conectado. Este será el caso habitual, ya que la protección de la interconexión típicamente disparael interruptor del generador. El transformador en el sitio del IPP actúa como un transformador de puesta atierra con corriente de secuencia cero circulando en los devanados del secundario en triángulo. Además deestos problemas, la corriente de carga desequilibrada en el sistema, que antes de añadirse el transformadordel IPP retornaba a tierra a través del neutro del transformador de subestación principal, ahora se divideentre los neutros del transformador del IPP y de la subestación. Esto puede reducir la capacidad de conducciónde carga del transformador del IPP y puede crear problemas cuando la corriente del alimentador está desequilibradacomo consecuencia de la operación de dispositivos de protección monofásicos, como ser los recerradores[reconectadores] de línea y fusibles. Si bien la conexión del transformador en estrella puesta a tierra/triángulose usa generalmente para grandes generadores conectados al sistema de transmisión de la empresa eléctrica,la misma presenta algunos problemas de importancia cuando se usa en sistemas de distribución de 4 hilos.Al considerar su posible uso, la empresa eléctrica deberá evaluar éstos aspectos.

— 14 —

Conexiones del Transformador de Interconexión en Estrella Puesta a Tierra (Prim.)/ Estrella Puesta a Tierra (Sec.)La principal preocupación respecto a un transformador de interconexión con devanados primarios y

secundarios puestos a tierra es que también proporciona una fuente de corriente a tierra indeseadaante fallas del alimentador de la empresa eléctrica, de modo similar a lo descrito en la secciónanterior. Asimismo, permite que los relés alimentadores de tierra con ajuste sensible en la subestaciónrespondan a las fallas a tierra en el secundario del transformador del IPP (F3). Las Figuras 12a y 12bilustran este aspecto mediante el análisis del circuito de componentes simétricos.

ConclusionesLa selección del transformador de interconexión juega un papel importante en establecer cómo va a

interactuar el IPP con el sistema de la empresa eléctrica. No hay una conexión “óptima” universalmenteaceptada. Todas las conexiones tienen sus ventajas y desventajas, que la empresa eléctrica deberá consideraren sus directivas de interconexión a los IPPs. Las selecciones de conexión del transformador ejercen asimismoun profundo impacto en los requisitos de protección de las interconexiones.

X

VIPP

F3

XL

Xsistema


X"d,X2,X0

IPP

XTSub.

51N/50NASub.

XTIPP (Falla L-T)

VS

F

!%/,(3%+8

X0L XTsubX"0gen

122 !BD,

XTIPP

R[

XTIPP

R[

XTIPP

R[

F

VIPP

!%/,(28#"$"H'

XsistemaXTsubXL

X"dgen

122 !BD,

F

!%/,(G%='$"H'

XsistemaXTsubXLX"2gen

122 !BD,

VS

A

A

A

Figura 12a Diagrama Unifilar para un Figura 12b Circuito de Componentes SimétricosTransformador de Interconexión para un Transformador de Interconexióncon Estrella Puesta a Tierra con Estrella Puesta a Tierra(Prim.)/Estrella Puesta a Tierra (Sec.) (Prim.)/Estrella Puesta a Tierra (Sec.)

Métodos y Procedimientos de Protección de Interconexiones de Pequeños Generadores DispersosLos niveles funcionales de la protección de interconexiones varían substancialmente dependiendo de

factores como: tamaño del generador, punto de interconexión con el sistema de la empresa eléctrica (distribución

— 15 —

o transmisión), tipo de generador (de inducción, sincrónico, asincrónico) y configuración del transformadorde interconexión (ver la sección previa de este artículo). Como se muestra en la Tabla 1, se pueden listar losobjetivos específicos de un sistema de protección de interconexiones así como los requisitos funcionales delrelé para lograr cada objetivo.

Tabla 1 Áreas de la Protección de InterconexionesObjetivo de la Protección Función de Protección a Usarde InterconexionesDetección de la pérdida de operación en 81O/U, 81R*, 27/59, 59I, TT**paralelo con el sistema de la empresa eléctrica

Detección de contraalimentación de fallas Fallas de Fase: 51V, 67, 21Fallas a Tierra: 51N, 67N, 59N, 27N

Detección de condiciones perjudiciales 47, 46en el sistema

Detección de flujo de potencia anormal 32

Restauración 25

* Tasa de cambio** Disparo de Transferencia

Detección de la Pérdida de Operación en Paralelo con el Sistema de la Empresa EléctricaEl medio más básico y universal de detectar la pérdida de operación en paralelo con la empresa eléctrica,

consiste en establecer una “ventana” de sobre/baja frecuencia (81O/U) y sobre/bajo voltaje (27/59) dentro dela cuál se le permite operar al generador de IPP. Cuando el generador de IPP está “en isla” con el sistema dela empresa eléctrica, debido ya sea a una falla o a otra condición anormal, la frecuencia y el voltaje saldráncon rapidez fuera de la ventana de operación si existe una diferencia significativa entre los niveles de lacarga y el generador de IPP.

En algunas aplicaciones de cogeneración, tales como las de la industria petroquímica y las de la pulpa y elpapel, se utilizan relés de tasa de cambio de la frecuencia (81R) para poder detectar más rápidamente lapérdida del suministro de la empresa eléctrica. La función 81R separa de la empresa eléctrica las instalacionesde la planta. En muchos casos, se produce internamente el rechazo de cargas por baja frecuencia en la plantay las cargas críticas son aisladas y alimentadas por los generadores de IPP de la planta.

Si la carga y el generador están casi en equilibrio al momento de la separación, el voltaje y la frecuenciapueden permanecer dentro de la ventana de operación normal y puede no producirse el disparo por baja/sobrefrecuencia y sobre/bajo voltaje. De existir esta posibilidad, quizás se necesite contar con disparo detransferencia (TT) usando un medio confiable de comunicación. Cuando los generadores de inducción estánen isla con capacitores montados en poste y la capacidad del generador es cercana a la de la carga en isla,puede ocurrir una condición resonante que produzca un sobrevoltaje no sinusoidal [5]. Para estos casos, sepuede utilizar un relé de sobrevoltaje instantáneo (59I) que responda a picos de sobrevoltaje para permitirdetectar esta situación.

Cuando se detecte la pérdida de operación en paralelo, el generador de IPP deberá ser separado delsistema de la empresa eléctrica con rapidez suficiente para permitir el recierre automático del interruptor enla subestación de la empresa eléctrica. El recierre de alta velocidad del sistema de la empresa eléctrica puedeocurrir tan pronto como en 15 a 20 ciclos luego del disparo del interruptor. La empresa eléctrica deberáindicar al propietario del IPP la velocidad de separación que se requiere.

El uso de relés de baja frecuencia en conjunción con la necesidad de separar el generador de IPPantes del recierre del interruptor de la empresa eléctrica, impide a la mayor parte de los pequeñosgeneradores de IPP dispersos suministrar energía de respaldo a la empresa eléctrica duranteperturbaciones importantes en el sistema. Cuando la frecuencia decrece a causa de una perturbaciónimportante en el sistema, estos generadores disparan quedando fuera de línea. Quizás sea posible

— 16 —

reducir los ajustes de baja frecuencia cumpliendo con los requisitos del Consejo de Confiabilidadregional, pero generalmente no se puede extender el tiempo de disparo requerido excediendo eltiempo para recierre automático. Este problema del sistema se hará más crítico si el porcentaje degeneración total en el sistema suministrado por pequeños generadores dispersos aumenta en lospróximos diez años, como lo predicen algunos expertos en la industria.

Si se extienden los tiempos de disparo por baja frecuencia, quizás resultará necesario modificar elmétodo de recierre de subestación, utilizando supervisión del voltaje de fuente conjuntamente conrecierre con comprobación de sincronismo. Este tipo de esquema, indicado en la Figura 13, ofreceseguridad contra el recierre previo a la desconexión del generador de IPP.

*A

* Puede requerir detección de ángulo y de deslizamiento


54

54IPP

2527

Figura 13 Esquema de Subestación de la Empresa Eléctrica

La Figura 14 muestra un típico esquema básico de sobre/bajo voltaje y sobre/baja frecuencia enuna pequeña instalación de IPP. Estas funciones de protección pueden todas incluirse en un sólo relédigital de multifunción.

>U7%'(0%(J"#$+"DB/":7

81/O

81/U27 5954

6%-.(0%(WB-$";B7/":7

4+'7#;8+&'08+(0%(17$%+/87%9":7

122GFigura 14 Típica Protección de Interconexión de un Generador Pequeño

Detección de Contraalimentación de FallasEn muchos generadores pequeños de IPP, no se proporciona por lo general detección de contraalimentación

de fallas. Los generadores de inducción suministran tan sólo dos o tres ciclos de corriente de falla para lasfallas externas, similarmente a los motores de inducción.

— 17 —

Las pequeñas máquinas sincrónicas están generalmente tan sobrecargadas luego que dispara elinterruptor de subestación de la empresa eléctrica, que su contribución de corriente de falla es muybaja. Para estos pequeños generadores, la detección de la pérdida de operación en paralelo por mediode los relés 81O/U y 27/59 es toda la protección de interconexión que se necesita.

Cuanto mayor sea el generador de IPP, mayor es la posibilidad que contribuirá una magnitudsignificativa de corriente a una falla del sistema de la empresa eléctrica. Para cubrir ésta situación, seproporciona detección de contraalimentación de fallas además de la protección contra pérdida deoperación en paralelo. Debe reconocerse que cuanto más prolongado sea el tiempo en que el generadorestá sometido a una falla, menor será la corriente que el generador sincrónico proporciona a la falla.La Figura 15 muestra la curva de decremento del generador. El nivel de corriente de falla a diversosintervalos luego de producirse la falla depende de las reactancias del generador (Xd", Xd'). La rapidezde decaimiento depende de las constantes de tiempo del campo del circuito abierto (Tdo", Tdo').

MF

2%+U808(!BD$+'7#"$8+"8MF0c(408cP

2%+U808(4+'7#"$8+"8MF0d(408dP

2%+U808(%7)#$'08()#$'/"87'+"8(MF0P

4"%&*8

)7H8-H%7$%5%+0'0%+'

)9$+'*8-'/":7(0%-5'-8+()#$'/"87'+"8

)9$+'*8-'/":7(0%(-')7H8-H%7$%(4+'7#"$8+"'

38++"%7$%(0%(38+$8/"+/B"$8(M1P

a

Figura 15 Curva de Corriente de Cortocircuito del Generador

Al desarrollar un sistema de protección con supresión de contraalimentación, es necesario considerarel decaimiento de la corriente ante fallas externas. Típicamente, se emplean funciones de relé talescomo la 67, la 21 o la 51V para detección de la contraalimentación de fallas de fase. Al establecer losajustes para los relés 67 y 21, el ajuste de arranque del relé deberá definirse excediendo el nivel decorriente de generación que el IPP está suministrando al sistema de la empresa eléctrica. Algunasempresas eléctricas supervisan un relé de sobrecorriente controlado con restricción de voltaje (51V)junto con la función 67 para incrementar la sensibilidad del arranque.

La supresión de la contraalimentación de fallas a tierra depende de la conexión del devanadoprimario del transformador de interconexión. Para devanados de transformador con primario conectadoa tierra se utiliza un relé de sobrecorriente de neutro 51N, o en algunos casos, un relé direccional detierra 67N. Las Figuras 16 y 17 muestran una típica protección de interconexiones para instalacionescon transformadores de interconexión con devanado primario conectado a tierra.

Para los transformadores de interconexión no puestos a tierra, los relés de sobrevoltaje de neutro(59N, 27N) proporcionan la detección de fallas a tierra del suministro. Los VTs (transformadores devoltaje) que alimentan estos relés tienen sus devanados primarios conectados fase a tierra. Estosdevanados primarios están generalmente clasificados para pleno voltaje entre fases. Muchas empresaseléctricas utilizan conexiones de transformadores de voltaje utilizando un sólo VT con relés 59N y27N o tres VTs conectados en configuración de triángulo abierto. La Figura 18 exhibe una protección

— 18 —

de interconexión típica para un IPP con configuración de transformador de interconexión no puesto atierra.

>U7%'(0%(J"#$+"DB/":7

!BD%#$'/":7

6

4

51N

G

[L34

47 27 59 81/O

81/U

51V 46 3267

Supresión deContraalimentaciónde Fallas

CondicionesPerjudiciales

Flujo dePotenciaAnormal Pérdida de Paralelo

* o Función 21** puede requerirse dependiendo

del tamaño del IPP

Disparo deTransferencia **

6%-.(WB-$";B7/"87'-

Y(:([((54d#

122(0%-)&*+,()-./$+"/'

122

*

Control

Figura 16 Protección Típica para un IPP de Tamaño Mediano con Transformador deInterconexión con Estrella Puesta a Tierra (Prim.)

— 19 —

6

4

51N

G

[L34

47 27 59 81/O

81/U

51V 46 3267

122

67N*

Control

ZL34

>U7%'(0%(J"#$+"DB/":7

!BD%#$'/":7





del tamaño del IPP

Disparo deTransferencia **

6%-.(WB-$";B7/"87'-

Y(:([((54d#

Figura 17 Protección Típica Alternativa para un IPP de Tamaño Mediano conTransformador de Interconexión con Estrella Puesta a Tierra (Prim.)

— 20 —

3"+/B"$8(0%(!B&"7"#$+8

6

4

G

[L34

47 27 59 81/O

81/U

51V 46 3267

122

*

27N59Ne o Δ

Control

!BD%#$'/":7





del tamaño del IPP

Disparo deTransferencia ** 6%-.(WB-$";B7/"87'-

Y(:([((54d#

122(0%-)&*+,()-./$+"/'

Z(:([((54d#

Figura 18 Protección Típica para un IPP de Tamaño Mediano con Transformador deInterconexióncon (Prim.) no Puesta a Tierra

Detección de Condiciones Perjudiciales en el SistemaLas condiciones de corriente desequilibrada producidas por conductores abiertos o inversiones de fase en

el circuito de suministro de la empresa eléctrica pueden someter al generador de IPP a un alto nivel decorriente de secuencia negativa. Esta alta corriente de secuencia negativa resulta en un rápido calentamientodel rotor, lo que provoca daños en el generador de IPP. Muchas empresas eléctricas proporcionan la proteccióncontra estas corrientes desequilibradas como parte del conjunto de protección de interconexiones, utilizandoun relé de sobrecorriente de secuencia negativa (46). Para ofrecer protección contra inversiones de fasedebidas al “intercambio de fases” inadvertido luego de la restauración de la potencia, se utiliza también unrelé de voltaje de secuencia negativa (47). Estas funciones se exhiben en las Figuras 16, 17 y 18.

Flujo de Potencia AnormalAlgunos contratos de interconexión entre IPPs que cogeneran y empresas eléctricas prohiben al IPP

suministrar potencia a la empresa eléctrica. El IPP cogenerador suministra potencia únicamente a la cargalocal en las instalaciones del IPP y reduce los costos de demanda de la empresa eléctrica mediante la“reducción de picos de demanda” (peak shaving). El procedimiento frecuente de las empresas eléctricasconsiste en instalar un relé de potencia direccional (32) para disparar el generador del IPP si se producenflujos inadvertidos de potencia al sistema de la empresa eléctrica durante un período predeterminado detiempo, en violación del contrato de interconexión. Las Figuras 16, 17 y 18 ilustran este tipo de detección deflujos de potencia anormales.

— 21 —

Procedimientos de Disparo/Restauración del IPPUna vez que el generador del IPP ha sido separado del sistema de la empresa eléctrica luego que haya

operado la protección de la interconexión, será necesario restaurar dicha interconexión. Hay dos procedimientosde disparo/restauración del IPP que se usan mucho en la industria. El primer método de restauración (caso 1)se utiliza en aplicaciones donde la generación en las instalaciones del IPP no cubre la carga local. En estoscasos, la protección de la interconexión generalmente dispara los interruptores del generador del IPP, comoindica la Figura 19. Al restaurarse el sistema de la empresa eléctrica, los generadores del IPP típicamenteson resincronizados en forma automática. Muchas empresas eléctricas requieren un relé de comprobación desincronismo (25) en el principal interruptor de entrada para supervisar el recierre, como medida de seguridadpara evitar el cierre no sincronizado. El relé de comprobación de sincronismo por lo general está equipadocon lógica de bajo voltaje de barra muerta [inactiva] para permitir el recierre desde el sistema de la empresaeléctrica ante una condición de barra muerta en las instalaciones del IPP.

6%-.(0%17$%+/87%9":734

54

<-(!"#$%&'(0%(-'()&*+,)-./$+"/'

G G

A

B C

Yb

3'+='(>8/'-

54

<B$8#"7/,

J"#*'+'E(^(3\

!B*%+H"#'+%/"%++%(0%(<

6%-.(J"="$'-(0%(WB-$";B7/":7

\(J"#*'+'(-8#("7$%++B*$8+%#E(^(3(#"f)-(=%7%+'08+(78(/BD+%(-'/'+='(-8/'-!%(+%#"7/+87"O'7(-8#"7$%++B*$8+%#(E(^(3(-B%=80%(+%#$'B+'+#%(%-(#"#$%&'0%(-'(%&*+%#'(%-./$+"/'

54 54

<B$8#"7/,

Figura 19 Restauración luego de un Disparo en la Interconexión—Caso 1

El segundo método de restauración (caso 2) se utiliza donde el generador de IPP cubre aproximadamentela carga local. En estos casos, la protección de la interconexión dispara el interruptor principal de llegada(interruptor A) como se muestra en la Figura 20. A menudo, las instalaciones del IPP pueden contar internamentecon rechazo [separación] de cargas por baja frecuencia, tal como es el procedimiento en las instalacionespetroquímica y de pulpa y papel, para adaptar la carga local a la generación de IPP disponible luego de laseparación de la empresa eléctrica. Para resincronizar las instalaciones del IPP al sistema de la empresaeléctrica, se requiere un relé de comprobación de sincronismo más sofisticado, que mide no sólo el ángulode fase (Δθ) sino el deslizamiento (ΔF) y la diferencia de voltaje (ΔV) entre los sistemas de la empresaeléctrica y del IPP. Típicamente, dichos relés supervisan el recierre manual local y remoto.

— 22 —

34

54

G G

A

B C

Yb

54

\(6%AB"#"$8#(0%-(6%-.(Yb

(((ΔΦ(L(J";,(0%(R'#%

(( ΔR(L(J%#-"O'&"%7$8

(((Δ5(L(J";,(0%(58-$']%

\6%-.(0%

17$%+/87%9":7

<-(!"#$%&'(0%(-'()&*+,)-./$+"/'

3'+='(>8/'-

J"#*'+'((((((<

6%/"%++'(<

6%-.(J"="$'-(0%(WB-$";B7/":7

Figura 20 Restauración luego de un Disparo en la Interconexión—Caso 2

Uso de la Tecnología Digital para Protección de InterconexionesLos modernos relés digitales de multifunción cuentan con diversas características que los hacen

ideales para la protección de interconexiones de generadores de IPP. Las más importantes de dichascaracterísticas son la funcionalidad seleccionable por el usuario, el autodiagnóstico, la capacidad decomunicaciones y el monitoreo oscilográfico.

Funcionalidad Seleccionable por el Usuario (“Selección Cuidadosa”)Como se ha indicado en este artículo, la funcionalidad de la protección de interconexiones varía mucho

según el tamaño del generador, el punto de interconexión al sistema de la empresa eléctrica, el tipo degenerador (de inducción o sincrónico) y la configuración del transformador de interconexión. Estas variableshacen que la funcionalidad seleccionable (“selección cuidadosa”) sea una característica de gran importancia.Dicha característica permite que la configuración específica del relé digital de multifunción sea controladapor el usuario, no por el fabricante. El costo es proporcional al nivel de funcionalidad que se requiera. Elusuario que adquiere un costoso conjunto de multifunción para interconexiones e inhabilita numerosas funcionesporque no son apropiadas para su aplicación específica, diluye las ventajas económicas de la protección tipomultifunción. Al utilizar un relé con las funciones básicas necesarias para la mayoría de las aplicaciones yhacer su selección adicional en una “biblioteca” de funciones opcionales, el usuario configura el equipo deprotección para la aplicación específica y minimiza su costo. La Figura 21 muestra una típica aplicación eninterconexiones que emplea este enfoque.

— 23 —

675685

9:

;95 6:5

ZL54:

[L54#\

[L34#

[(:(Y54#

;7 <; 67/ =9 =8>3 ?= 68

ZL54

?9 ;9

;6

!"#$%&'(0%(-'()&*+,()-./$+"/'

\>'(/87%9":7(/87(%-(4+'7#;8+&'08+(0%(58-$']%(M54P(*B%0%#%+(0%($+"Q7=B-8('D"%+$8(8(0%(;'#%('($"%++'(/87(B7(#:-8(54N0%*%70"%708(0%(-'('*-"/'/":7,

@72@7ABC6:D6:

EFG$,"(2%0H(IJ0'$KJ&,$.&()%(D&'%#,"&%+$.&

>J&,$.&(*L'M&)-#

>J&,$.&(AG,$"&-0

ZL34

!%(B#'(*'+'(387;"=B+'/":7(0%(4+'7#;8+&'08+(0%17$%+/87%9":7(/87(*+"&'+"8(78('($"%++'

4+'7#;8+&'08+0%(17$%+/87%9":7

Figura 21 Diagrama Unifilar de un Relé Digital de Multifunción para Interconexiones

AutodiagnósticoEl autodiagnóstico de un relé digital de multifunción permite la detección inmediata de fallas en el relé.

Si no se cuenta con protección de interconexiones, el generador de IPP así como el sistema de la empresaeléctrica pueden verse sometidos a condiciones perjudiciales tales como las corrientes de falla no detectadas,los sobrevoltajes y el alto esfuerzo torsional del eje del generador de IPP debido al recierre automático. Porestas razones, el autodiagnóstico adquiere cada vez mayor importancia. Muchas empresas eléctricas optanpor disparar el generador de IPP cuando falla el conjunto de protección de la interconexión, para evitar

— 24 —

dichas condiciones perjudiciales. El autodiagnóstico da a la empresa eléctrica cierto grado de seguridadsobre el buen funcionamiento de la protección de la interconexión. Las antiguas tecnologías electrónicas oelectromecánicas no ofrecían este margen de seguridad.

Capacidad de ComunicacionesTodos los relés digitales de multifunción cuentan con puertos de comunicaciones. Por lo general son del

tipo RS-232, RS-485, o en algunos casos, conexiones de fibra óptica. La mayoría de los IPPs medianos ograndes están obligados a proporcionar a la empresa eléctrica datos telemétricos continuos sobre la operacióndel generador. Típicamente se les exige información tal como el estado (abierto o cerrado) de interruptoresclave de generación e interconexión, así como la salida instantánea en MW o MVAr del generador. Granparte de esta información puede obtenerse mediante el conjunto de relés de interconexión, eliminando lanecesidad de contar con transductores y medición adicional. Asimismo, la capacidad de interrogar al relé deprotección de interconexión desde un lugar remoto para determinar los eventos de operación del relé permitedisponer de información esencial para restaurar la unidad de IPP al servicio.

Monitoreo OscilográficoEl monitoreo oscilográfico de las entradas del relé (corrientes y voltajes) proporciona información sobre

la causa de la operación del relé de interconexión e indica si el relé ha funcionado de acuerdo a lo planeado.Como la protección de interconexiones se aplica en el punto de acoplamiento común entre la empresaeléctrica y las instalaciones del IPP, el monitoreo oscilográfico ofrece información valiosa sobre cuál es elsistema que ha provocado el disparo. La información oscilográfica ha permitido resolver desacuerdos entreempresas eléctricas y propietarios de IPPs sobre la causa de eventos de disparo específicos.

ConclusionesLa protección de interconexiones tendrá mucha mayor importancia en el próximo milenio, si se materializan

las predicciones de numerosos expertos en la industria. La protección de interconexiones bien diseñadadeberá atender los aspectos que preocupan a los propietarios de generadores de IPP así como a las empresaseléctricas. En este artículo se ha intentado resumir los puntos principales que las empresas eléctricas y lospropietarios de IPPs deben considerar al establecer requisitos de interconexión. Uno de los aspectos másimportantes y que se ignora con mayor frecuencia, es la configuración del transformador de interconexión.Esto juega un rol fundamental en la minimización del sobrevoltaje potencial en el sistema de la empresaeléctrica, así como en la determinación de los requisitos para la protección de la interconexión.

Los requisitos funcionales de la protección de interconexiones varían considerablemente. Los factoresque determinan los requisitos de la protección incluyen: el tamaño del generador, el punto de interconexióncon el sistema de la empresa eléctrica, el tipo de generador (de inducción o sincrónico), y los niveles decontraalimentación de fallas. Estas variables hacen que la funcionalidad seleccionable por el usuario o de“selección cuidadosa” sea una característica muy importante en los modernos relés digitales de multifunciónpara interconexiones. Además de la lógica de disparo se requiere restauración automática, que puede serincorporada en un conjunto de relés digitales para interconexión. Esperamos que los aspectos enfatizados eneste artículo sean de utilidad para las empresas eléctricas a la hora de evaluar sus procedimientos de interconexión.

Referencias[1] ANSI/IEEE Std. 1001-1988, “Guide for Interfacing Dispersed Storage and Generation Facilities

with Electric Utility Systems” (Guía para el Interfaz de Instalaciones Dispersas de Almacenamientoy Generación con Sistemas de Empresas Eléctricas).

[2] Donahue, K.E., “Relay Protection Interface and Telemetry Requirements for Non-Utility Genera-tors and Electric Utilities” (Requisitos de Telemetría e Interfaz de Relés de Protección para Generadoresno de Empresas Eléctricas y Empresas Públicas de Electricidad), 1998 Power Generation Confer-ence, Orlando, Florida.

[3] Mozina, C.J., “Protecting Generator Sets Using Digital Technology” (Protegiendo Grupos GeneradoresMediante Tecnología Digital), Consulting/Specifying Engineer Magazine, EGSA Supplement, No-vember 1997.

— 25 —

[4] Feero, Gish, Wagner and Jones, “Relay Performance in DGS Islands” (Desempeño de losRelés en Islas de DGS”, IEEE Transactions on Power Delivery, January 1999.

[5] ANSI/IEEE C37.102-1995, “IEEE Guide for AC Generator Protection” (Guía del IEEE parala Protección de Generadores de CA).

[6] Yalla, Hornak, “A Digital Multifunction Relay for Intertie and Generator Protection” (Un ReléDigital de Multifunción para la Protección de Interconexiones y Generadores), Canadian ElectricalAssociation Conference, March 1992.

Acerca del AutorChuck Mozina es el Gerente de Aplicaciónes, Productos y Sistemas de Protección de Beckwith

Electric Co. Es responsable de la aplicación de productos y sistemas Beckwith que se utilizan enprotección de generadores y en esquemas de protección de interconexiones, sincronización y transferenciade barras.

Chuck es miembro activo del IEEE Power System Relay Committee (PSRC) (Comité sobre Relésde Sistemas de Energía Eléctrica del IEEE) y fue presidente del Rotating Machinery Subcommittee(Subcomité sobre Máquinas Rotativas). Es también miembro activo del comité IAS I&CPS delIEEE, que se ocupa de la protección de sistemas industriales. Chuck es el representante de EE.UU.al CIGRE Study Committee 34 on System Protection (Comité 34 de Estudios del CIGRE sobreProtección de Sistemas) y dirige un grupo de trabajo del CIGRE sobre protección de generadores.Ha dirigido también el equipo de trabajo del IEEE que produjo el instructivo “The Protection ofSynchronous Generators” (La Protección de Generadores Sincrónicos), que ganó el OutstandingWorking Group Award (Distinción al Grupo de Trabajo Sobresaliente) del PSRC en 1995. Chuckobtuvo en 1993 la distinción Career Service Award (Distinción al Servicio Profesional) del PSRC.

Chuck se graduó como Bachiller en Ciencias en Ingeniería Eléctrica en la Purdue University, y esautor de diversos ensayos y artículos en revistas sobre protección con relés. Tiene más de 25 años deexperiencia como ingeniero de protecciones en Centerior Energy, una importante empresa eléctricaprivada en Cleveland, Ohio, donde fue Gerente de la Sección de Protección de Sistemas. Se desempeñótambién como profesor en la Escuela de Graduados de Ingeniería Eléctrica en Cleveland StateUniversity y es un Ingeniero Profesional registrado en el estado de Ohio, EE.UU.

1

RELIABILITY ANALYSIS FOR A GENERATION SHEDDING SCHEME ON THE CFE MAIN TRANSMISSION NETWORK

Antulio Jarquin, Eduardo Mora Alcaraz, and Elizabeth Godoy Alcantar

Comisión Federal de Electricidad CFE

Jean León Eternod, Schweitzer Engineering Laboratories

Mexico

ABSTRACT Lack of transmission network capacity because of right-of-way restrictions and limited investment requires system operation at close to stability margins. Wide-area network protection schemes (also called system-wide remedial action schemes) are commonly used in these systems to prevent transient instability or voltage collapse problems and their associated operational and economical consequences. These schemes combine protection, control, and communications devices located at different geographic areas in the power system. CFE has used auxiliary relays, protection relays, industrial programmable logic controllers (PLCs), high-level language programming on control center master SCADA computers, and combinations of these solutions to solve problems at different power system locations.

This paper analyzes a proposal for increasing reliability in a real-life case study of a 1200 MW generation-shedding scheme for CFE�’s southeast hydroelectric power plants. The scheme includes devices that integrate protection, control, and direct digital communications between devices and substations in a common reliable platform. We look at the root cause of limits to generation shedding in the area of the case study and determine some practical options for improving reliability.

INTRODUCTION There are many contingencies for which several areas of the CFE network can become weak or unstable. The system requires wide-area network protection to avoid real and reactive power instability.

The power stability problems are common in networks with long transmission links between generation and load. Strong networks are susceptible to power instability after contingencies such as breaker failure or bus protection operations, where more than one line or transformer is lost.

Based on contingency studies, one can determine the needs of wide-area network protection. These types of protection can require power flow measurement, open line detection, communications, different tripping times, etc. depending on the kind of transmission limitation found. This paper shows some critical information from power system studies, but the focus is on field implementation rather than power system studies or trip decision algorithms.

In this paper, we examine the CFE southeast 400 kV transmission link between the Grijalva Hydroelectric System (about 50 percent of total national hydroelectric capacity) and the Central Region (which includes Mexico City and has the largest load). The transmission link, shown in Figure 1, is about 1000 km long, with transmission of as much as 3000 MW and as much as 1200 MW total generation shedding.

2

Figure 1 Mexico 400 kV Network

Several events on February 2002 led to major disturbances. The system lost three lines and became unstable after three oscillations. The system split into two areas covering six states. In these areas, load was lost for several hours.

Previous study of this contingency led to a shedding scheme designed to use generation shedding and some low-frequency load shedding as needed to prevent blackout. Simulations showed the scheme acting correctly to save the system from collapse. Field implementation failed, however, because of a bad signal from an auxiliary relay. This was the first dependability failure of significant consequence on this scheme; similar network schemes have had similar problems. Since scheme implementation some years ago, communication difficulties and bad signals from auxiliary relays have caused a number of security failures with incorrect generation trips.

CFE uses auxiliary relays, transducers, PLCs, SCADA master, conventional microwave, and carrier communications channels for field implementation of this kind of system. In this paper, we study a method that uses different technology to improve reliability in the case study scheme.

The improved system includes simple and reliable methods for open line detection, power measurement on each line, the sum of power from all lines on a transmission link, contingency grouping and logic processing capabilities, and communications signals. Other research topics included in the CFE project include ways to improve flexibility in configuring systems and new substations, increase redundancy at some points in the network, and obtain automatic supervision for the entire system.

3

POWER TRANSMISSION LIMITS On any power system, operation should meet equality constraints, the most important being that power generation must equal power load plus system losses.

Operation should also meet the following inequality criteria:

Voltage regulation of 0.9 pu < V < 1.1 pu

Frequency regulation of 59.8 Hz < f < 60.2 Hz

Current in line < conductor thermal limit

Figure 2 [1] defines possible system states based on equality and inequality constraints.

Normal E, I Safe ModeEconomic DispatchSystem Coordination

ExtremeLoad ShedEquipment Protection

UnsafeMode

WithoutSafety

Segregated System Intact System

Areas SeparationLoad Shed

Operation MarginsReduction

Emergency EEmergency Actions(most of the time fast)

Recovery ILoad RecoveryAreas Synchronization

Alert E, IPreventive Control(may be slow)

Unequality RestrictionsViolation

Figure 2 Power System Operation Under Stress Conditions,

Based Upon A Diagram by Fink and Carlsen

In the normal state, the equality and inequality constraints are met, there are no overload elements on the network, and the security margins are normal. Control objectives are economic optimization of all system operations and system coordination with good security margins.

In the alarm state, the equality and inequality constraints are still met, but the security margins have been reduced or eliminated. Control objectives are increased security margins either through generation and load reallocation or through voltage and frequency regulation. Actions could be automatic or initiated by a control center operator, which can take from minutes to hours. Some weak systems may operate permanently in the alarm state during daily peak hours.

In the emergency state, the system is still intact, but the inequality constraints have been violated. Control or protection actions should be fast enough to relieve overloads and return the system to the alert state. It is better still to attempt normal control actions before the system proceeds to the emergency state, which may trigger wide-area network protection actions. Normal control actions can include fast valving, excitation control, single-pole tripping and reclose, capacitor or

4

reactor switching, or use of flexible alternate current transmission system (FACTS) or high-voltage direct current (HVDC) controls.

In the extreme state, the system loses integrity; some areas can become isolated from the rest of the system, the system sheds load, and some system devices can operate outside normal operational limits. Action should be as fast as possible, with the protection or control objective being the retention of as much load as possible. Normally, some load or generation can be shed intentionally at selected locations, and some areas can be isolated from the system. Changes from the normal, alarm, or emergency states to the extreme state must trigger wide-area network protection actions.

In the recovery state, the system meets the inequality restrictions, but some load remains out of service. Control actions in this state include startup or synchronization of generators to the network, load recovery, and segregated area synchronization.

Some equality or inequality constraints can be obtained transiently, while others can be steady-state conditions. Steady-state limits are the following:

!" Conductor, switchgear, and current transformer or power transformer thermal limits !" Voltage regulation !" Reactive power operational margins !" Voltage collapse security margins

Thermal limits depend upon ambient conditions such as temperature, wind, and sunlight. Information is available from suppliers of conductors and switchgear equipment, from standards such as IEEE C.57 for transformers, and from generator capability curves. Because thermal problems can take seconds or even minutes to cause damage, protection can take a long time to operate and tripping times are not critical under most conditions. For thermal problems, operators or automatic control can generally take action before automatic protection schemes operate.

Voltage regulation and reactive operational margins depend upon contracts with customers or internal rules governing interconnection points and reactive capacity at power plants and other reactive power sources. You would normally not need automatic protection action, only automatic control.

Strictly speaking, voltage collapse is a transient problem, but several analyses assume steady-state conditions. Reactive power security margins decrease to a point where voltage could collapse for a slight increase in reactive power demand. Generally, there is low voltage in this condition. Then, automatic voltage regulators on some distribution or transmission transformers change taps, and the reactive power demand increases in either a few seconds or minutes, according to the time delay for the automatic load tap changers. Maximum excitation limiters can contribute to voltage collapse by reducing the reactive power after an initial fast voltage regulator response. Maximum excitation limiters operate with slower time constants than voltage regulators, changing generator terminal voltage in several seconds. In cases where you need to improve security margins, a corrective action for voltage collapse can occur after several seconds without consequences. A second contingency could develop a faster collapse, but it also causes transient stability problems, and will be discussed with similar problems later in the paper.

5

Tripping time for automatic wide-area network protection against steady-state limit violations is normally not a critical constraint. Dynamic limits include the following:

!" Transient or first cycle stability limit !" Dynamic instability with low or negative damping of oscillations !" Slow dynamic limit, normally for frequency or voltage regulation after disturbance

Transient instability causes severe damage to generators, as well as to other equipment such as transmission transformers. Voltage, power, and torque variations are high, and the tripping mode is difficult to control after the first cycle of oscillation. Fast action is imperative.

Fast action is also necessary for dynamic instability, which causes low or negative damping of oscillations. The system might support the first cycle of operation, but oscillations will persist for several seconds until some line or generator protection operates and the system becomes unstable.

Transient and dynamic stability limits cause an oscillation frequency between 0.5 and 2 Hz and oscillation cycles from 0.5 to 2 seconds. Protection and control actions should occur before completion of the first oscillation cycle. Contingency studies can show the time limits for scheme operation.

Utilities use contingency analysis to determine how, when, and where a network reaches limits. For such tasks as increasing the size of a transmission network, planning departments use n-1 (single) contingency analysis. Where investment is low, this first contingency analysis does not include breaker failure or bus failure analysis.

Operation departments should use n-2 (double contingency) or n-3 (triple contingency) analysis. This is important because, although a multiple contingency failure is unlikely to occur, such a failure would have catastrophic consequences if there were no action. Planning work should be evaluated with (n-1) analysis without loss of load. Operation work should be evaluated with the amount of load system loss after (n-1), (n-2) or (n-3) contingencies and the cost of operation in the normal state.

If the system operation takes into consideration double or triple contingencies, real power transmission limits will be very low and the operation costs can increase because power must be obtained from another source that could be more expensive. If the system operation takes into consideration only a single contingency, real power transmission limits can be high and operation expenses low. Any case of a double or triple contingency failure, however, would be costly. It is possible to use (n-1) analysis to define normal operation transmission limits with low operation cost and use (n-2) or (n-3) analysis to define wide-area network protection scheme needs and settings.

CASE STUDY The case under study involved 19 transmission lines and 11 substations as the online diagram in Figure 3 shows. We use three-letter codes to name substations; we use the names of substations at each line end to define a line. CFE has done complete contingency analyses for each single and double contingency for different operation scenarios.

6

PBD

TCL

5 X 180 MW

OJP TMD MID MPS

225 KM 144 KM

6 X 180 MW

80 KM

ANG

90 KM

7 X 300 MW

MMTJUI

225 KM

190 KM

155 KM 243 KM

CHMCTS

G

G

G

Figure 3 Online Diagram

CFE finds that almost all double contingency combinations cause transient stability problems with loss of synchronism or voltage problems near voltage collapse. Table 1 describes the way some contingencies group together on certain parts of the link.

Table 1 Double Contingency Groups and Limits

Link Double Contingencies Transmission limit

found Reason to limit

MMT�–JUI�–TMD MMT-JUI & MMT-JUI JUI-TMD & JUI-TMD

1600 MW from MID to TMD

Near to voltage collapse in central area

MMT-MPS MMT-MPS & MMT-MPS 1200 MW from MMT to JUI

Transient stability limit for 3-phase failure

MPS-MID MPD-CTS-MID

MPS-MID & MPD-MID MPS-MID & MPD-CTS MPS-MID & CTS-MID

1800 MW from JUI to TMD

Near to voltage collapse in central area and dynamic stability

MID-TMD MID-CHM-TMD

MID-TMD & MID-CHMMID-TMD & CHM-TMD

1200 MW from TMD to central area

Transient stability

TMD-OJP-PBD TMD-TCL

TMD-PBD & TMD-OJP TMD-PBD & OJP-PBD OJP-PBD & TMT-TCL TMD-PBD & TMT-TCL TMT-TCL & TMD-OJP

1100 MW from TMD to central area

Near to voltage collapse in central area and dynamic stability

7

There are several other possible combinations. There can be, for example, double contingencies that involve lines over different links. Most of the double contingency limits were found with simulation of failure and simultaneous tripping of both lines, but the results were reviewed against a steady-state initial single contingency condition and against a second failure with very similar results.

With these combinations, we notice some important points about the scheme. !" Some double contingencies can be grouped together, but there are many different limits

depending on which pair of lines is lost. Each limit imposes a different amount of generation shedding. Several pilot protection channels are necessary to send different kinds of contingencies from places where a trip occurred to generation units.

!" There are many different combinations of single or double contingencies to be transmitted to the next substation. These combinations should be handled by logic operations (AND, OR, NOT) with inputs from open line detection. Logic schemes could be implemented with auxiliary relays, PLCs, or logic processors.

!" Transmission limitation could be measured at locations other than those where lines trip. Transmission limits are based on link total power and not on line power. Some device or system should provide a sum of all power from the different lines.

!" Some of the limits are for transient stability. Fast trip is needed for these cases but could be used for other conditions.

!" Fast trip means fast and reliable open pole detection; a fast and reliable direct transmission channel is needed.

DYNAMIC SIMULATION RESULTS AND EVENT ANALYSIS With the help of Figure 4, we can provide a simplified explanation of what happened during the severe disturbance on February 2002. We have divided the system into three areas. Area 1 includes the central region of Mexico, which has a very high concentration of load. Area 2 is the Grijalva Hydroelectric System, which has a high concentration of generation and low load. Area 3 is a small network on the Yucatan Peninsula with a weak link to Area 2 that contributes some inter-area oscillations.

Before the disturbance, the power flow was about 1600 MW from Area 2 to Area 1, with three lines from the TMD substation to the central region. During the February 2002 event, two parallel lines (TMD-PBD and TMD-OJP) and another line (TMD-CHM) were lost. One line (TMD-TCL) supported the link for about 7 seconds until it also tripped, causing the collapse of Area 2 and Area 3 and a great amount of load shedding from frequency relays in Area 1.

8

Area 1 Area2

Area 3

TMD-OJPTMD-PBD

TMD-TCL

Figure 4 Simplified Version of Transmission Links and Areas

Conditions and measurements from dynamic recorders reporting this event helped provide validation of simulation models. Dynamic recorders are important for diagnosing problems during the event, validating simulation models, and evaluating the results of the new scheme [2], [3]. From the dynamic recorders, we can obtain the following useful information:

!" Real and reactive power on each line and links !" Frequency !" Voltage phasors, including magnitude and angle from synchronized measurement

Recording of this information should be at a slower acquisition rate than that provided by line protection recorders. The dynamic event recorders make better use of memory than the line protection recorders and, as can be seen in Table 2, provide a higher acquisition rate than SCADA master recorders, from which dynamic effects cannot be seen. The total time to record each event is also an important variable; line protection recorders normally record 1 to 2 seconds, while dynamic events can last from 10 seconds to 2 or 3 minutes.

Table 2 Features on Different Types of Event Recorders

Type of recorder Acquisition frequency Total recording time

Line protection recorders 240 Hz to 2 kHZ 0.1 to 2 seconds

SCADA master recorders 0.1 to 0.5 Hz Continuous recording from hours to weeks

Dynamic recorders 30 to 10 Hz 10 to 180 seconds

For validation purposes, we compare the dynamic recorder records that were obtained during the disturbance versus the simulation results.

9

Figure 5, Figure 6, Figure 7, and Figure 8 show the frequencies for the three areas, the records of total link power from dynamic recorders for areas 1, 2, and 3, and the simulation results for the same variables and fault conditions.

Figure 5 Record of Frequencies in Three Areas.

Figure 6 Simulated Frequencies

- 4 0 0

- 3 0 0

- 2 0 0

- 1 0 0

0

1 0 0

2 0 0

3 0 0

4 0 0

1 7 : 3 0 : 0 8 1 7 :3 0 :1 3 1 7 :3 0 :1 8 1 7 : 3 0 : 2 3 1 7 :3 0 :H R S

MW

E S A - K L V E

Figure 7 Record of Real Power Exchange Between Areas 2 and 3

Figure 8 Real Power Exchange From

Simulation

The simulation results closely match the records, so we can assume that the model is valid and that we can simulate several contingencies and conditions to analyze the factors that influence the correct performance of a generation shedding scheme.

10

With this model, CFE simulated the effects of correct operation for a generation shedding scheme. Figure 9 compares simulation results for operation or no operation of generation shedding.

DAG

Sin DAG

Figure 9 Real Power on TMD-TCL With and Without Generation Shedding

One of the most important factors to consider for these schemes is time delay of the communications channels. Although wide-area network protection problems do not need short time delays such as those for pilot protection tripping, some time delays are critical, and very slow generator tripping does not cause necessary stabilization.

Consider also that the channel time delays are cumulative; signals traveling from substations to generators, over lines open for some specific contingency, often go through other substations where communications devices serve as �“repeaters.�”

Figure 10 shows real power flow on Line TMD-TCL (the only line that does not trip after the initial disturbance) for different channel time delays and tripping times for generators.

900 ms 1000ms 1100 ms

400 ms

Figure 10 Real Power Flow on Line TMD-TCL for Different Tripping Times

11

We can see that oscillations increase according to increases in tripping times. This effect is more apparent when looking at the angular difference between areas 2 and 3, as in Figure 11.

900 ms 1000 ms 1100 ms

400 ms

Figure 11 Angular Difference Between Areas 2 and 3 for Different Tripping Times

The critical time for generator tripping is 1100 ms for this specific case. We need a complete analysis to determine the critical tripping time for each contingency and the maximum acceptable channel delays, but this example shows some typical characteristics:

!" Although the maximum tripping time is 1100 ms, it is clear that faster operation reduces oscillations and the possibility of other generator trips or line protection operations. A good real life maximum time could be near 400 ms.

!" Tripping times of 200 to 400 ms or faster are handled easily with pilot protection direct channels including four or five repeaters, but these tripping times are not feasible with SCADA-type communications channels. This limitation is important to remember because, in the parts of a wide-area system dependent upon a SCADA regional master, a trip signal must travel directly between substations, with a great economic impact on the scheme.

!" Power oscillations are large for some stable disturbances. If we try to measure power after such a disturbance and decide to trip with this information, it is not possible to use an instantaneous power level detector; there are other events where the system will not need generator shedding, and oscillations beyond the power limit will occur for a short time.

!" If we try to use a definite-time delay characteristic for power level detection, we should wait some time to trip. This time delay, related to the first oscillation time, should be longer than 400 ms in most cases. This time delay is not acceptable for transient instability problems, but could be used for problems such as voltage collapse or thermal limits.

!" Because this part of the network is almost radial, total real power flow is almost the same after and before the initial disturbance. It is possible to measure real power, determine a power level, and pre-select a tripping mode, all before the disturbance. Then, the only delay between initial disturbance and generator trip comes from open line detection, logic schemes that make decisions about double contingencies at each location, and channel delay.

12

From the simulation and contingency analysis, we can determine that the scheme needs the following features:

!" There must be a measure of total link power for three substations, MMT, TMD, and MPS. Some device or system must determine the sum of real power flow from as many as three lines and send this information to the power plants. Only a total link power limit decision from TMD to the central region must be transmitted to generators. MMT and MPS must each determine locally the total flow for their own power plants.

!" To pre-select or prepare a generation shedding scheme, there must be a power limit decision before the disturbance. There is no specific time limit for this because power flow changes slowly before a disturbance.

!" After preparation or pre-selection of a generation shedding scheme, open line detectors must detect tripping of certain specific combinations of lines. PLCs, logic processors, or auxiliary control circuits must then logically combine these lines to form double contingency groups as Table 1 shows, and send a tripping decision to generator units.

!" Several communications channels are necessary to transmit trip decisions to generators. For the worst-case scenario, a repeater (channel repeater, tone-to-tone equipment connection, logic processor, auxiliary control circuit, etc.) must retransmit these decisions through four substations (four pilot protection channel services for each trip decision). Because one line can trip only at one end, either more channels are necessary to send open line information to the remote end, or each line must have direct transfer trip. Tripping time must be less than 400 ms.

PRESENT GENERATION SHEDDING SCHEME Figure 12 shows an overview of the present scheme. The regional control center at Puebla has all the network measurement information for SCADA and energy management systems. The information comes from both digital meters and transducers. A program at the regional master station adds the real power for the entire link, compares this total real power against specific levels shown in Table 1, and sends pre-selection instructions to generation plants. Information is available also for system operators who telephone the power plants and order manual pre-selection of generators in case of any problem with SCADA control signals. Unless a control signal fails to arrive at a power plant, there is no alarm for a control signal problem. Single contingency pre-selections are always made by phone between the system and the plant operators.

13

Present Scheme

CD ControlCircuits

RX

RX

CD ControlCircuits

CD ControlCircuits

CD ControlCircuits

TX

CD ControlCircuits

METERS ORTRANSDUCERS

RTU

SCADA MASTER !P ANDLIMITS PRE-SELECTION

AUXILIARY RELAYSTRIP DECISION

PLCTRIP DECISION

COMMUNICATIONSEQUIPMENT

AUXILIARY RELAYSCONTINGENCY LOGIC

AUXILIARY RELAYSOPEN LINE DETECTOR

PLCCONTINGENCY LOGIC

TONE EQUIPMENT

SCADA ChannelControl Signals

SCADA ChannelAnalog Values

TRIP UNITS

Microwave orCarrier Channel �—

one for each tripdecision

CD ControlCircuits

MICROWAVEEQUIPMENT

MICROWAVEEQUIPMENTRTU

43

ControlSwitches

Manual UnitPre-Selection

MicrowaveChannel

MW

TONE EQUIPMENT

B2

B1

CD ControlCircuits

G

I V

OPEN LINE DETECTOR

Figure 12 Present Scheme

14

On each line control circuit, there are auxiliary relays in parallel with the trip coils. When a line trips, an auxiliary relay sends notification of this trip through the control circuit to hardwired single and double contingency logic schemes. Other sets of auxiliary relays connect by means of the control circuit to digital inputs in a PLC that has single and double contingency logic schemes.

Substation configurations include breaker-and-a-half and double bus with double breaker. For open line detection, both line breakers must trip. Under maintenance conditions, switch disconnectors can isolate one breaker, while the remaining breaker activates line trip detection. Auxiliary relays with hardwired logic schemes for each line (also known as line detectors) feed information from all such conditions and from trip bus and disconnector auxiliary relays to PLCs or other hardwired logic schemes for contingency identification.

Both PLC and auxiliary relay logic schemes use timers to prevent double contingencies where only a few seconds separate trips on different lines. Analysis shows this as a weakness of the present system. Some line could be open for several minutes, but a second failure will cause exactly the same effects as if both lines had tripped together for contingencies resulting in voltage collapse problems. A similar effect could occur for stability problems.

PLC or auxiliary relay logic schemes send decisions for a generation trip over conventional microwave or carrier channels without power limit supervision; only the master can send this information to the power plant substations. Mean tripping times are very good, as fast as 14 ms from channels without repeaters and nearly 60 ms total tripping time for the worst case situation involving four repeaters. Table 3 shows the total number of channels needed, without redundancy. Full redundancy on communications channels is costly, so we implement full redundancy only on certain links.

Table 3 Communications Channels Used by the Present Scheme

From To Channels

TCL TMD 1

OJP TMD 1

CHM TMD 1

TMD MID 2

MID MPS 5

MPS MMT 4

MMT ANG 2

CHM MID 1

CTS MID 1

JUI MMT 2

TOTAL 20

Auxiliary relays at power plant substations trip after receiving generation trip signals, whereas PLCs or hardwired logic schemes must take into account the trip signals from remote contingencies, local contingencies, power limits from SCADA, and local generation unit selection

15

before making a trip decision. Selection of generation units must follow several considerations including which units are manually selected and which units you want to trip first.

The present scheme allows the following failures, all of which have occurred in the past:

Security failures: !" Some line detectors and associated auxiliary relays have sent incorrect trip signals during

line relay maintenance. !" Some line detectors have sent incorrect trip signals during vibration and during such work

as changes or upgrades to relays and wiring. !" Some remote incorrect trip signals have been caused by confusion over terminals during

maintenance work on protection or communications panels. !" Some remote incorrect trip signals have been caused by transitory signals on the pilot

protection channel. !" Manual selection of generators to trip or temporary SCADA channel failure to send

information deselecting generation units during a decrease in power flow can cause overpower trips.

Dependability failures: !" An auxiliary relay failure on the line detector can cause a failure to send a generator trip

signal when needed. !" An open dc circuit can cause a failure to send a generator trip signal when needed. !" SCADA channel failures can cause a failure to select generation units at the necessary

time. This failure consists of two parts: a failure to receive power information and a failure to send pre-selection signals to power plants.

!" A channel failure can cause a trip signal to not be sent. The present system has channel redundancy on some links but not on the entire system.

Since January 2001, this part of the transmission network has had 47 events with line tripping, and CFE has experienced three security failures with incorrect tripping of generation. One such failure was the result of an incorrect signal from the SCADA system. The second failure was an incorrect generator trip reception caused by channel noise problems, and the last failure resulted from a dc circuit transient. The system has experienced one major dependability failure, as we described before. Analysis and operation statistics show the need for a more reliable system.

PROPOSAL FOR NEW GENERATION SHEDDING SCHEME For our proposal, we analyzed the best options for providing open line detection, line power measurement, measurement of total link power, limit detection, and digital local and remote signal communication. The following discussion includes the results of this analysis.

Open Line Detection and Line Power Measurement

Line protection makes use of multifunction digital relays, such as distance relays and directional overcurrent relays, to provide primary and backup protection. These relays already receive all the signals necessary for each line.

16

We take advantage of the ability of these multifunction digital relays to provide power measurement and transmit such measurement information through serial communication to other scheme devices. Current detectors in these relays, together with internal relay logic that detects switch-onto-fault conditions, will provide us with pole open detection. Current signals are more reliable than auxiliary signals and provide us with more dependable open line detection than that available from auxiliary relays. On some lines where special low load conditions could fall below the sensitivity limits of the current detector, line side potentials or circuit breaker auxiliary relays together with current signals could provide open line detection. As we will see later, in an unusual case where the current is so low that a current detector alone sends incorrect open line detection information, the system will use power limit detection to block transmission. The proposed system does not use disconnector position signals.

Total Link Power Measurement and Limit Detection

A local communications processor receives real power measurement by means of a direct serial connection from line relays. The communications processor adds power from different lines to obtain a measurement of the power on the entire link. It then compares this sum against the limits from Table 1. The power limit detection for the communications processor is available locally, through contacts or serial communications, and remotely, through serial communications. This is possible because the limits in use are from substations where all link power goes through only one substation, as Figure 13 shows. The communications processor must have analog operators (for providing power summation) and analog comparators (to provide level detection).

Serial Communications

COMMUNICATIONS

PROCESSOR

Line Relay

Line Relay

Line Relay

Figure 13 Total Link Power Limit Detection

If two or more substations are involved in providing total link power, then a channel is necessary. This channel could be a dedicated channel between substations (Figure 14), which is preferable to having a SCADA channel to the regional control center.

17

Serial Communications Line Relay

Line Relay

Line Relay

COMMUNICATIONS

PROCESSOR

Dedicated Channel forRemote Power Measurement

Figure 14 Total Link Power Limit Detection From Two Substations

This configuration has several advantages over the present system: !" A SCADA channel failure does not affect the system. !" Local testing is easier. !" It is easier to add power from new lines with digital meters or relays. !" Power limit pre-selection is available locally, and pre-selection messages could be sent

with generator signals to other substations and power plant logic schemes. This may appear at first to require more dedicated channels, but the proposed solution for channels eliminates this as a problem.

!" Power limit pre-selection can be used locally to supervise local trip transmission signals for added security.

!" While the present system has four points of failure: meter or transducer�—remote thermal unit (RTU)�—SCADA channel�—master, the proposed system has two points of failure: digital relay�—communications processor.

!" The proposed system time response, although not critical to total link power measurement and limit detection, is better than for the present system. The maximum time response we measured in testing the proposed system was 2.6 seconds, and the minimum time was 0.9 seconds. The present system time response is about 10 seconds.

Digital Local and Remote Signal Communications and Channels

Technical paper [4] discusses technology capable of transmitting digital information directly between relays for the pilot protection functions and other high-speed signals over such low bandwidth channels as a microwave analog channel. This technology takes advantage of the capability of logic processors and digital relays to communicate directly over serial port connections without external communications equipment (see Figure 15).

18

Relay 1 Relay 2

Station 1 Station 2

TXRX52a

+

�–

52a

+

�–

Relay 1 Relay 2

CommunicationChannel

TCTC

PTPT

TXRX

Trip Key Key Trip

Relay 1 Relay 2

Station 1 Station 2

TRIP TXRX

TC

52a

Relay 1

+

�–

Relay 2TRIPTXRX

TC

52a

+

�–

EIA-232 DigitalCommunication

Figure 15 Relay-to-Relay Communications Technology Versus Traditional Pilot Protection Communications

Through the application of relay-to-relay communication, as shown in Figure 16, we can use a serial port connection to obtain a two-way virtual transmission of eight bits of high-speed control signals. Message security is continuously monitored, and alarm signals are available if any connection fails.

19

Relay 1 Relay 2

1

0 0

0

0

RECEIVE

TRANSMIT

RMB1

RMB8

RMB2

TMB1

TMB8

TMB2

RMB1

RMB8

RMB2

TMB1

TMB8

TMB2

TRANSMIT

RECEIVE

0

0

0

0

0

0

1

Station 1 Station 2

Figure 16 Virtual Connection of Eight High-Speed Control Signals

The proposed scheme uses this relay-to-relay communications technology for two different purposes. The first purpose is to provide communication of open line detection from digital relays to a control processor, while the second purpose is to provide transmission of remote signals.

For the first purpose, use of a control processor is similar to a PLC in the sense that one can program logic schemes. A control processor, however, receives serial-supervised messages, rather than the hardwired dc input signals available to PLCs. Communications between relays and a control processor occur in a quarter cycle, or about 4 ms. A control processor has logic schemes to determine each kind or group of single or double contingencies and to send generator trip signals through contacts or serial connections.

For transmission of remote signals, logic processors send generator trip decisions, power limit detection signals, and pre-selection signals to power plant processors in some substations. Because eight control points are available on each channel through relay-to-relay communications, we can multiply by eight the service each channel provides, use fewer channels than those assigned to the present scheme, and provide redundancy at a reasonable cost.

20

Table 4 Communications Channels Used By the Proposed Scheme

From To Channels Generator trip signals Power limit and unit pre-

selection signals

TCL TMD 1

OJP TMD 1

CHM TMD 1

TMD MID 1 2 3

MID MPS 1 5 3

MPS MMT 1 4 3

MMT ANG 1 2

CHM MID 1 1 3

CTS MID 1 1 3

JUI MMT 1 2 3

TOTAL 10

The advantages of this improvement include the following: !" Signals between open line detectors and equipment for contingency logic schemes are

supervised in real time; any interruption can be monitored and resolved quickly. There are no possible auxiliary relay failures in the proposed scheme.

!" Communications processor reliability, similar to that for a digital relay, is much better than industrial PLCs.

!" Communications processors at each location receive contingency signals and power limit signals. Power limit signals are used as supervision to send or repeat a received generator trip signal.

!" The total number of channels in use decreases from 20 on the present scheme to 10, cutting channel costs by half.

!" Communications are continuously monitored, and alarms are available for the failure of any channel. While the present scheme can accomplish this, there are two differences: records and statistics from any channel failure belong to the generator shedding scheme instead of to the communications scheme, and power limits for generator pre-selection are monitored continuously. The present scheme uses a SCADA channel to send pre-selection signals to power plants without automatic supervision.

Electric cables or multimode fiber-optic cables provide the physical serial connections between relays in use as open line detectors and logic processors. Serial connection between substations is through a specially designed modem that converts digital serial signals to analog signals at a transmission speed of 9600 baud. To avoid an external power source, a communications processor serial port provides power to the modem, which is also designed to avoid any error correction or other extra delays common to conventional modems. The four-wire analog signal from the modem connects directly to an analog microwave channel. The proposed scheme does not use an electromechanical contact interface (tone equipment), and so reduces one point of failure (Figure 15).

21

The modem causes a longer delay than direct conventional pilot protection channels. With this modem, we measured a one-way channel delay of 28 ms. A worst case trip signal through four repeaters would take 112 ms, compared to 60 ms for conventional pilot protection channels. Simulation results show that any time less than 200 ms is good, and that any time less than 400 ms is acceptable. With 1.1 seconds as the critical trip time, the longer delay with the modem poses no problem.

To integrate with system signals from other vendor devices, control switches, and conventional communications channels with electromechanical contact interfaces, the proposed scheme uses contacts to serial communications over fiber-optic converters. This allows both integration of other devices lacking direct serial communications technology and gradual upgrading of the scheme. CFE installed and tested in the field the proposed scheme for two of 11 substations. The proposed scheme integrates easily with the present scheme and allows future upgrading of the entire scheme with contacts to a serial communications converter. The prototype CFE scheme and general architecture are shown in Figure 17 and Figure 18. Other advantages of the new scheme are reduced maintenance and built-in event recording of scheme operations.

22

TX

TXLimitsPre-Selection

SerialConnection

DirectConnection

TXOpen LineDetection

SerialConnection

LINE RELAYSOPEN LINE DETECTOR

MW MEASUREMENT

COMMUNICATIONSPROCESSOR!P AND LIMITS

PRE-SELECTION

MODEM

LINE RELAYSOPEN LINE DETECTORS

RX

CONTACTS TO FIBERCONVERTER

Proposed Scheme

MWAnalog Value

B1

B2

LOGIC PROCESSORCONTINGENCY LOGIC

HIGH-SPEED PROCESSINGCAPABILITIES

MODEM

Microwave Channel

ContingenciesFrom OtherSubstations

MODEM

43

ControlSwitches

Manual UnitPre-SelectionTRIP UNITS

G

(One for 8 tripdecisions orpre-selection)

DIRECT SERIALSUPERVISEDCONNECTIONCABLE OR FIBER

52A (OPTIONAL)

RX

MicrowaveChannel

RX

LOGIC PROCESSORTRIP DECISION

CD ControlCircuits

I V

1 2 3

1

2

3

Figure 17 Proposed Scheme Architecture

23

Line RelayJUI-TMT

CommunicationsProcessor

Line RelayJUI-TMT

Line RelayMID-TMT

Line RelayCHM-TMT

Logic Processor

Line RelayTMT-TCL

Line RelayTMT-PBD

Line RelayTMT-OJP

Line RelayWithout Relay-to-Relay

CommunicationMID-CHM

Line RelayWithout Relay-to-Relay

CommunicationMID-CTS

I/O Contacts toFiber Logic Processor

P (MW)

P (MW)

P (MW)

P (MW)

Opened or closed

Opened or closed

Opened or closed

Opened or closed

Opened or closed

Opened or closed

Opened or closed TEMASCALTMD

MINATITLANMID

MODEM

MODEM

Signals forremote trip

Open or closedinformation from2 lines

Output Contacts toPresent System

Figure 18 Block Diagram of Prototype

FAULT TREE RELIABILITY ANALYSIS To numerically evaluate reliability improvements, we use fault tree reliability analysis. This method is easy to apply, and use of the method for protection and automation reliability estimates has been previously documented [5] [6].

The scheme failure of concern is called the top event. The probability that the scheme fails for the top event is a combination of the failure probabilities of the components in the scheme. For an OR gate, any inputs to that OR gate can contribute to scheme failure. Total probability is the sum of the input events. For an AND gate, any inputs to that gate must fail together to cause scheme failure. The upper level probability for scheme failure from an AND gate is the product of input probabilities.

24

To estimate the failure probability for each device in the scheme, we can use the device failure rate. One industry practice is to provide failure rates as Mean Time Between Failures (MTBF). MTBF could be based on field failure data or on assumptions about complexity and exposure of equipment. If we have 200 auxiliary relays and only one such relay fails per year, we can assume a failure rate of 1/200 failure per year or an MTBF of 200 years from field experience. Some communications equipment vendors, however, estimating failure rates based upon complexity, could publish an MTBF of 80 years.

To use this information to estimate probability, we should know or assume the fraction of time that a device cannot perform. Unavailability, as calculated in the following equation, provides us with this information.

q = #T = T / MTBF

where:

q is unavailability

# is failure rate

T is average down time per failure

MTBF is mean time between failures

T is the fraction of time MTBF when the device is either not useful or has failed. If a communications channel has a guard signal and a failure alarm, we could easily detect a failure on these devices. Then, T could equal two days for detection, analysis of the failure, and repair or replacement before the device is again in service and useful. Unavailability with this example information is 2 / (80 x 365) = 0.0000684 or 0.025 days/year.

One of the principal weaknesses of the present scheme is the dependency of the scheme on several auxiliary relays for open line detection, contingency logic schemes, tone equipment, etc., MTBF could work well for auxiliary relays, but T is always large because of a lack of automatic supervision. Failure of an auxiliary relay could go unnoticed until the next maintenance period or until operation of that relay is required. If the maintenance or testing period is each year, a failure could occur the day following a maintenance test or one day before the next period, an average time of six months. Unavailability with this example is (6 x 30) / (200 x 365) = 0.002465 or 0.9 days/year, 36 times worse than the example with the communications channel, even considering the much better MTBF.

Unavailability gives direct information about the probability that a device on the scheme will fail and contribute to scheme failure. From references [5] and [6], we obtain unavailability or MTBF for devices used in the present and the proposed scheme. These numbers, although the approximations are subject to dispute, provide valuable information for checking the degree of magnitude improvements and for estimating the impact of redundancy or other changes on the scheme configuration. We were unable to obtain failure rates for the auxiliary relays, so we used an optimistic estimate to compare with a proposed solution that does not use an auxiliary relay.

25

Table 5 Unavailability Indices for Devices Used on Present and Proposed Schemes

Devices or basic events MTBF T Unavailability

x 106

Current transformers 500 years 2 10

Potential transformers 500 years 2 10

Transducer 70 2 78

RTU 100,000 hours 48 hours 480

SCADA channel (2 modems plus microwave or carrier channel) 660

SCADA master computer (with redundancy)

22,500 hours each workstation 48 hours 4.5

Microwave transmission channel 600**

Analog microwave equipment 200

Tone equipment 100

Auxiliary relay 500 years 6 months 986

CD wiring 500 years 6 months 986

Monitored CD battery 50

PLC 17 years 2 days 320

Line relay 168 years* 2 days 32.6

Communications processor 200 years 2 days 27

Logic processor 200 years 2 days 27

Modem for protection 30

Serial supervised direct connections 100 years 2 days 54 * MTBF observed by CFE from a total population of about 12,000 relays with similar technology. ** Unavailability observed by CFE during a 5-day test on TMD-JUI microwave link with direct relay-to-relay communications and modems.

With the data from Table 2 as input, we can develop the fault tree in Figure 19 for a top event (Failure to Trip Units at MPS for (n-2) Contingency at TMD) similar to the February 2002 event. This first analysis shows the scheme as it presently exists at some locations, with a single channel but redundant trip and contingency logic schemes.

26

OR

FAIL

UR

E T

O T

RIP

UN

ITS

AT M

PSFO

R (

N-2

) CO

NTI

NG

ENC

Y AT

TM

D

1325

2

6.77

CO

MM

UN

ICAT

ION

S

Mic

row

ave

Cha

nnel

600

X 2

Bat

tery

50 X

3To

ne E

q.10

0 X4

Mic

row

ave

Equ

ipm

ent

200

X 4

OR

(CO

MM

UN

ICAT

ION

S TM

D�–M

ID�–M

PS)

OR

320

DC

Con

nect

ions

From

RTU

And

Tone

Eq.

986

X 2

DC

PLC

Bat

tery

50

OR

Auxi

liary

Logi

c98

6

DC

Batte

ry50

DC

Con

nect

ions

From

RTU

Or T

one

Eq.

986

X 2

TRIP

LO

GIC

AT

MPS

AN

D

RTU

AN

ALO

GM

EASU

RE

5010

X 1

8Tr

ansd

ucer

78 X

348

0

MW

RTU

TMD

RTU

OR

RTU

PR

E-SE

LEC

TIO

NS

IGN

AL A

T M

PS

OR

Anal

ogFr

om T

MD

660

4.5

Con

trol

Sign

al T

oM

PS

660

480

50

RTU

MPS

2798

.5

RTU

SC

AD

AC

HAN

NEL

SCA

DA

CH

AN

NE

LSC

ADA

MAS

TER

3008

2342

949

2550

7897

.35

10,4

47

(n-2

) C

ON

TIN

GEN

CY

SIG

NAL

AT

MPS

OR

(n-2

) LO

GIC

AT

TMD

(con

tinue

d)

Figure 19 Fault Tree for Present System�—Case 1: Trip and Contingency Logic Redundancy

Single Channel (continued)

27

OR

FAIL

UR

E O

N (n

-2)

LOG

IC A

T TM

D

7897

.35

9.35

OP

EN L

INE

2

DET

ECTO

R

3994

3944

3944 OP

EN L

INE

1

DET

ECTO

R

OR

AN

D

OR

Wiri

ng F

rom

Ope

n Li

ne98

6 X

2

DC

-W

986

X 2

Batte

ry50

AUXI

LIAR

Y R

ELAY

SLO

GIC

OR

320

Wiri

ng F

rom

Ope

n Li

ne98

6 X

2

DC

-WP

LC

50

PLC

LO

GIC

2342

Brea

ker 2

Trip

Auxi

liary

986

DC

-W

Wiri

ngFr

omTr

ip B

us98

6 X

2

Brea

ker

1 Tr

ipAu

xilia

ry98

6

Figure 19 Fault Tree for Present System�—Case 1: Trip and Contingency Logic Redundancy

Single Channel

Unavailability of this scheme for the specific top event is 13252 x 10-6 or 1.32 percent. Our analysis involved hardware failures only, but the scheme can experience such other failures as human errors with settings or testing. We can use some results from this fault tree to analyze two other cases with minor variations: no redundancy (Figure 20), or redundancy for trip and contingency logic and for channels (Figure 21).

28

OR

TRIP UNITS AT MPSFOR (n-2) AT TMD

RTU PRE-SELECTIONSIGNAL AT MPS

(n-2) CONTINGENCYSIGNAL AT MPS

OR

PLC Logic2342

Communications2550

3944 X 2PLC Option

17920

2798.5 4892

OPEN LINE DETECTORTRIP LOGIC AT MPS

2342 7888

Figure 20 Fault Tree for Present System Without Redundancy

OR

TRIP UNITS AT MPSFOR (n-2) AT TMD

(n-2) CONTINGENCYSIGNAL AT MPS

OR

Comm 22 Channels

2550

Comm 12 Channels

2550

RTUPre-selection

Signal At MPS

Trip LogicAt MPS

9909.12

15.851998.56.77

Open LineDetector3944 X 2

Auxiliaries3994

ANDAND

6.5 9.35

PLC2342

(n-2) Logic

Figure 21 Fault Tree for Present System With Full Redundancy

29

We can observe, and operation experience shows, that in all configurations using the present technology, the main causes of failure include the SCADA remote scheme and communications channels and the open line detectors with auxiliary relays.

We can develop a fault tree analysis for the proposed scheme without redundancy, as in Figure 22, and with full scheme redundancy, as in Figure 23.

PRE-SELECTION AND(n-2) CONTINGENCY

SIGNALS AT MPS

PRE-SELECTION AND(n-2) SIGNALS AT TMD

LogicProcessorAt TMD

27

Line RelayMW And

Open LineDetection

32.6

ComProcessor

27

SerialConnection

54 X 3

CTsAnd PTs10 X 18

BatteryTMD50

COMMUNICATIONS

OR OR

Battery MID50

LogicProcessor

At MID27

µWEquipment

200 X 4

µWChannel600 X 2

Modem30 X 4

BatteryMPS50

LogicProcessorAt MPS

27

2675

4782197

OR

FAILURE TO TRIP UNITS ATMPS FOR (n-2) CONTINGENCY

AT TMD

2752.6

OR

Figure 22 Fault Tree for Proposed System Without Redundancy

Unavailability for the proposed scheme without redundancy is 2752.6 x 10-6, or 0.27 percent, 4.81 times better than the present scheme with partial redundancy (Figure 3). Unavailability for the proposed scheme with full redundancy is 186.6 x 10-6 or 0.0186 percent, 71 times better than the present scheme in Figure 3. The main sources for enhancement include the use of open line current detection instead of auxiliary relay trip detection, local power limit processing instead of

30

SCADA master power limit detection, and the possibility of channel multiplication providing full redundancy at reasonable cost.

180

OR

FAILURE TO TRIP UNITS ATMPS FOR (n-2) CONTINGENCY

AT TMD

186.6

2572.6 2572.6

IDEMCase 1

IDEM Case 1Redundancy

CTs And PTs10 X 18

Figure 23 Fault Tree for Proposed System With Full Redundancy

Table 6 shows a chart correlating all options with reliability. It shows unavailability results only for a top event similar to the February 2002 event. A complete analysis is necessary for all possible modes of operation.

Table 6 Correlations Between Reliability and Equipment Used for Different Generation Scheme Options

Solution Unavailability

x106 Related

enhancement Total solution equipment as related cost

information

Present solution with trip and contingency logic redundancy, single channel

13252 Reference (Figure 3�–present system for this top event with partial redundancy)

7 PLCs 7 sets of auxiliary relays for contingency

or trip logic 36 sets of auxiliary relays for line trip

detection 20 pilot protection communications

channels, 40 analog microwave equipment services and 40 tone equipment services

3 SCADA channels, 3 RTUs, 2 master stations and 8 real power transducers are involved in the system (but are not part of the generation scheme direct cost)

Present solution without any redundancy

17920 35% worse than reference

7 PLCs 36 sets of auxiliary relays for line trip


channels, 40 analog microwave services and 40 tone services

31

Solution Unavailability

x106 Related

enhancement Total solution equipment as related cost

information

Present solution with trip and contingency logic redundancy and channel redundancy

9909 33% better than reference

7 PLCs 7 sets of auxiliary relays for contingency

or trip logics 37 sets of auxiliary relays for line trip


channels, 80 analog microwave services and 80 tone services

Proposed solution without any redundancy

2752 4.81 times better than reference

4 communications processors 10 logic processors 10 pilot protection communications

channels, 22 modems, 22 analog microwave services

36 line relays are involved (but are not part of the generation scheme direct cost)

Proposed solution with full redundancy

186.6 71 times better than reference

8 communications processors 20 logic processors 20 pilot protection communications

channels, 44 modems, 44 analog microwave services

FUTURE CONSIDERATIONS A lack of transmission network capacity, because of right-of-way restrictions and limited investment, is making wide-area network protection schemes increasingly more important. References [7] [8] are among several works dealing with new algorithms that improve the ability to detect conditions that can affect power system integrity. Such works discuss synchronized phasor measurement schemes for transient stability or voltage collapse estimates and neural networks to identify unsafe conditions and emergency control actions on complex meshed networks. From the development of algorithms to field implementation, all new and improved performance schemes should take reliability issues into consideration.

Communications will continue to become cheaper and safer with the increased use of digital fiber networks by most utilities. New communications will allow schemes that use more information to make better decisions. Intelligent use of these lower cost digital networks will have a strong impact on the development of wide-area network protection schemes.

To implement wide-area network protection schemes, more facilities will require analog operators and comparators such as those used in the proposed solution. Analog operators for new algorithms should include sum, multiplication, division, sine, and cosine functions. These features could be part of line relays or processors at every location to allow future implementation and changes as schemes develop and change. Some local control stations have high-level programming languages that allow mathematical operations, but the reliability of these is low and the processing speed is not deterministic; these local control stations cannot serve the purposes of wide-area network protection.

32

CONCLUSIONS 1. Wide-area network protection schemes allow economic power system operation without

impacting reliability, even considering multiple contingencies.

2. Detailed steady-state and dynamic stability contingency studies should be done to specify wide-area network protection schemes and answer questions about such field implementation issues as the features needed at each location and maximum tripping time.

3. Each contingency causes different problems and imposes different requirements on field implementation. Control devices, rather than wide-area network protection, can solve some of these problems. Only contingencies causing a system to go into the �“extreme�” state should trigger action from a wide-area network protection scheme.

4. Dynamic event records, with different variables, acquisition rates, and event report lengths for relay operation evaluation or for SCADA, provide validation of dynamic studies. The dynamic event records are also useful for evaluating the operation of wide-area network protection schemes.

5. Typical maximum tripping times for stability problems are between 200 ms and 2 seconds. Normal SCADA channels and control signals are not fast enough. Trip signals must travel by pilot protection channels directly from a contingency detection to the power plants.

6. Because of oscillations that occur with the present contingency criteria, real power after a disturbance cannot be used as a variable for determining whether to trip. Power limits must be continuously monitored, and pre-selection of generators for tripping each kind of contingency must occur before a disturbance.

7. To improve reliability, total link power limit detection should occur locally.

8. Open line detection is critical for scheme performance; current-based methods are more simple and reliable than auxiliary relay methods.

9. Direct relay-to-relay communications technology improves scheme reliability in two ways: such technology avoids unsupervised dc control circuit wiring between devices, and the technology multiplies pilot protection channels to allow channel redundancy at a reasonable cost.

10. The fault tree method is a valuable tool for analyzing and quantifying reliability improvements on wide-area network protection schemes.

11. Field performance statistics for individual system components provide excellent input data to obtain more accurate unavailability estimates.

12. The proposed scheme without redundancy provides more than four times the availability of the present scheme with partial redundancy and cuts communications channel costs in half.

13. The proposed scheme with full redundancy provides more than 71 times the availability of the present scheme without increasing costs beyond those for the present scheme with partial redundancy.

33

BIOGRAPHIES Elizabeth Godoy Alcantar has been with Comision Federal de Electricidad (the national utility of Mexico), working at Power Systems Studies in the area of electromagnetic transients in the protection office, since 1998. From 1995 to 1998, she was a systems operator at the National Control Center. She received her BSEE degree from the University Autonomous of Morelos (UAEM), Mexico in 1991. She received her MSEE degree in power systems from the National University of Mexico in 1994. Her MSEE degree thesis won the National Prize for Power Master Degree Thesis from the Electric Research Institute (Instituto de Investigaciones Electricas IIE) and CFE.

Antulio Jarquin Hatadiz has been Chief of Comisión Federal de Electricidad Power Network Analysis Department in Protection and Control national management since 1992. At CFE, he tests and approves new transmission relays for corporate use. Prior to assuming his present office, he worked in the CFE Specialized Engineering Unit, where he was in charge of transient network analyzer studies (TNA) and commissioning of static var compensators (SVC) as well as other jobs related to supervisory control and field service. He received his BSEE degree from Instituto Tecnologico de Veracruz.

Jean Leon Eternod is a field application engineer for Schweitzer Engineering Laboratories at Mexico City. Prior to joining SEL in 1998, he worked for the Comisión Federal de Electricidad Power Systems Studies Office in protection and control corporate management. While with CFE from 1991 to 1998, Leon worked with wide-area network protection schemes, single-pole trip and reclose studies, and database validation for short circuit, load flow and dynamic simulation. He received his BSEE from the National Autonomous University of Mexico (UNAM), where he also completed postgraduate coursework in power systems. He received training in power system simulation from Power Technologies Inc. He has delivered technical papers for the summer meeting of the Mexican chapter of IEEE, Monterrey�’s Iberoamerican Protections Symposium, and the AMIME Rotating Machinery Conference in the fields of power systems simulation, synchronized phasor measurement applications, and generation and distribution protection.

Eduardo Mora Alcaraz has been a regional supervisor protection engineer for Comisión Federal de Electricidad at Veracruz, México since 1988. He has worked on several projects for extra high-voltage protection and control equipment, single-pole trip and reclose applications, wide-area load and generation shedding schemes, and new device evaluation. He received his BSEE degree from Metropolitan Autonomous University UAM in Mexico City.

REFERENCES [1] L.H. Fink and K. Carlsen, �“Operating Under Stress and Strain,�” IEEE Spectrum, Vol. 15,

pp. 48�–53, March 1978.

[2] J.L. Eternod, �“Pruebas de Registro Dinámico Utilizando la Medición Sincronizada de Fasores En la Red de CFE,�” Memoria Técnica III Simposio Iberoamericano de Proteccion de Sistemas Electricos de Potencia, Monterrey, Nuevo León, Mexico, Noviembre de 1996.

[3] R. Burnett, M.M. Butts, T.W. Cease, V.Centeno, G. Michel, R.J. Murphy, A.G. Phadke, �“Synchronized Phasor Measurement of a Power System Event,�” IEEE Transactions on Power Systems, vol.9, no.3, pp. 1643�–1649, 1994.

34

[4] K. C. Behrendt, P.E., �“Relay-To-Relay Digital Logic Communication for Line Protection, Monitoring, and Control,�” 51st Annual Georgia Tech Protective Relaying Conference, Atlanta, Georgia, April 30�–May 2, 1997.

[5] E.O. Schweitzer, III, B. Fleming, T.J. Lee, and P.M. Anderson, �“Reliability Analysis of Transmission Protection Using Fault Tree Methods,�” 24th Annual Western Protective Relay Conference, Spokane, Washington, October 21�–23, 1997.

[6] G.W. Scheer, �“Answering Substation Automation Questions Through Fault Tree Analysis,�” Proceedings of the 4th Annual Texas A&M Substation Automation Conference, College Station, Texas, April 8�–9, 1998.

[7] Workshop on Synchronized Phasor Measurement Theory and Applications, Center of Power Engineering of Virginia Tech, Arlington Virginia 1996.

[8] J.L. Eternod, �“Propuestas para un Relevador de Energía Transitoria,�” Primera Parte-Algoritmos de Decision, Memoria Tecnica III Simposio Iberoamericano Sobre de Proteccion de Sistemas Electricos de Potencia, Monterrey, Nuevo León, Mexico, Noviembre de 1996.

Copyright © SEL 2002 (All rights reserved)

Printed in USA 20021022

RELEVADOR ADAPTIVO DE SOBRECORRIENTE DE TIEMPO INVERSO: DISEÑO Y PRUEBAS

Arturo Conde Enríquez Ernesto Vázquez Martínez Héctor J. Altuve Ferrer*

Universidad Autónoma de Nuevo León

Facultad de Ingeniería Mecánica y Eléctrica Apdo. Postal 114-F, Ciudad Universitaria, CP 66450

San Nicolás de los Garza, Nuevo León, México [email protected] [email protected]

*SEL, S.A. de C.V. CINTERMEX, Local 31, Planta Baja

Av. Parque Fundidora No. 501, Col. Obrera Monterrey, Nuevo León, México 64000

[email protected]

Resumen: En este artículo se propone un relevador adaptivo de sobrecorriente de tiempo inverso. Este relevador adaptivo tiene una mayor sensibilidad y tiempos de operación más reducidos, y utiliza solo la información disponible en la ubicación del relevador. Se describe la estructura funcional del relevador adaptivo de sobrecorriente, su implementación y los resultados obtenidos en pruebas de laboratorio con señales adquiridas en tiempo real. Palabras clave: relevador adaptivo de sobrecorriente, corriente de arranque adaptiva, tiempo de operación adaptivo.

I. INTRODUCCIÓN

Las protecciones convencionales tienen parámetros de ajuste fijos, lo que hace difícil cumplir con los requerimientos de protección en condiciones de régimen de operación variable en el sistema de potencia. Una respuesta a este problema es la protección adaptiva, que puede variar sus parámetros de ajuste o sus características de operación en respuesta a cambios en el sistema de potencia. En [1,2,3] se sugiere una lógica adaptiva a nivel de subestación, en que los parámetros de ajuste del relevador son modificados desde una computadora central de subestación. En redes aisladas o rurales en que no es costeable implementar una estructura de comunicación adecuada, es posible realizar la adaptación del relevador con información de corriente local y con información del ajuste del dispositivo respaldado. En este trabajo se propone un relevador adaptivo de sobrecorriente que no requiere canales de comunicación para modificar sus parámetros de ajuste. El relevador adaptivo de sobrecorriente de tiempo inverso propuesto en este trabajo tiene la capacidad de variar su corriente de arranque en función de la corriente de carga de la línea protegida, y ajustar su tiempo de operación en función de la curva de tiempo del dispositivo respaldado.

El relevador adaptivo se implementó en una computadora personal equipada con una tarjeta de adquisición de datos y se realizaron pruebas de laboratorio con señales adquiridas en tiempo real de un modelo físico del sistema eléctrico de potencia.

II. RELEVADOR DIGITAL DE SOBRECORRIENTE DE TIEMPO INVERSO

En la Fig. 1 se presenta el diagrama funcional de un relevador digital de sobrecorriente de tiempo inverso [4], en el que por simplicidad no se muestra el procesamiento analógico y la conversión análogo-digital de la señal de corriente. El filtro recibe como entrada las muestras digitalizadas ( )kri de la corriente, y entrega a su salida,

para cada instante de muestreo, el módulo ( )krI del fasor

que representa la componente fundamental de ( )kri . El generador de funciones recibe como entradas la corriente ( )krI y la corriente de arranque Ia, y forma la señal de salida H(Ik), donde Ik =(Ir)k/Ia es la magnitud del fasor de componente fundamental de la corriente de entrada al relevador normalizada con respecto a la corriente de arranque. El integrador es el elemento que introduce la variable tiempo en el proceso; su señal de salida es:

( ) ( )k

k

1kk

k

1kk IHttIHG ∑∑

==

== ΔΔ (1)

donde Gk representa el valor acumulado del integrador en el instante de procesar la muestra k, y Δt es el período de muestreo. En el comparador (Fig. 1) se determina la condición de operación:

Integrador Comparador

H(Ik) Gk=G{H(Ik),t} Generador

de funciones

Ia

( )krI

J(Ik)

Filtro digital ( )kri

Fig. 1. Diagrama funcional simplificado de un relevador digital de sobrecorriente de tiempo inverso.

( ) ( )kk

k

kk IJIHtG

op

=Δ= ∑=1

(2)

El relevador opera en el instante en que k alcanza un valor igual a kop y se cumple (2). El tiempo de operación T está dado por:

tkT opΔ= (3)

Despejando Δt en (2) y sustituyendo en (3), se obtiene la ecuación de la característica tiempo-corriente T=F(Ik) del relevador:

( ) ( )

( )∑=

==opk

kk

kopk

IH

IJkIFT

1

(4)

Si para fines de análisis se considera constante la corriente durante la falla ( Ik=I ), (4) toma la forma:

( ) ( )( )IHIJIFT == (5)

El tiempo de operación del relevador para corrientes de falla constantes está definido por la relación de las funciones J(I) y H(I).

Dentro de cada período de muestreo la corriente permanece constante en el valor calculado en la muestra anterior. Por tanto, la ecuación (5) también puede escribirse para I=Ik (período de muestreo correspondiente a la muestra k):

( ) ( )( )k

kk IH

IJIF = (6)

Despejando H(Ik) en (6) y sustituyendo en (4), se tiene:

( )( )( )

1

1

=

⎥⎦

⎤⎢⎣

⎡Δ ∑

=

op

op

k

k k

k

k

IFIJ

t

IJ (7)

La ecuación (7) es la ecuación generalizada del relevador digital de sobrecorriente de tiempo inverso [4]. Como en los relevadores analógicos, en los relevadores digitales de sobrecorriente por lo general se hace la simplificación de ( ) ( ) KIJIJ

opkk == .

Sustituyendo esta condición en (7), se obtiene la forma digital de la expresión dinámica (8), propuesta en [5].

( )∑=

=Δ⎟⎟⎠

⎞⎜⎜⎝

⎛opk

k k

tIF1

11 (8)

III. LIMITACIONES DE LA PROTECCIÓN DE SOBRECORRIENTE

La protección de sobrecorriente utiliza la corriente como parámetro indicador del lugar del cortocircuito, teniendo en cuenta que su magnitud depende de la distancia eléctrica hasta el punto de falla. Sin embargo, hay dos problemas fundamentales: a) la corriente de cortocircuito depende también del tipo de falla y del régimen de operación del sistema; b) la corriente de carga o de pre-falla puede ser comparable con la de cortocircuito, lo que dificulta la discriminación entre el régimen normal y el de falla. De estos problemas se derivan las limitaciones de la protección de sobrecorriente.

Una primera limitación de la protección de sobrecorriente

es que su alcance (longitud de la zona de protección) depende del tipo de cortocircuito y del régimen de operación del sistema. El alcance del elemento instantáneo de la protección puede reducirse e inclusive desaparecer completamente en condiciones de generación mínima. El alcance del elemento de tiempo inverso también se reduce, dejando en ocasiones de respaldar debidamente a la línea adyacente para generación mínima. Esto es particularmente probable en la protección de fase, en que la corriente de carga máxima establece un límite inferior al valor de la corriente de arranque del relevador, limitando así su sensibilidad para la detección de fallas en condiciones de generación mínima. La segunda limitación de la protección de sobrecorriente son los elevados tiempos de operación para fallas no máximas. El criterio de coordinación se determina para los valores máximos de corriente de falla. Para las fallas restantes, que son las más frecuentes, el tiempo de operación del relevador es mayor. La limitación de tiempo es acentuada cuando el dispositivo respaldado tiene una característica de operación con diferente grado de inversión que la del relevador.

IV. CRITERIOS ADAPTIVOS PARA EL ELEMENTO DE TIEMPO

INVERSO Una alternativa de solución a las limitaciones de la protección de sobrecorriente es dotarla de la capacidad de variar automáticamente sus parámetros de ajuste dependiendo del régimen de operación del sistema, es decir, convertirla en adaptiva. Un relevador de sobrecorriente adaptivo de tiempo inverso debe tener la capacidad de variar su corriente de arranque y su curva de tiempo en función de la corriente de carga de la línea protegida. La adaptación del relevador requiere dos lógicas adaptivas: a) adaptación de la corriente de arranque, para mejorar la sensibilidad de la protección; b) adaptación del tiempo de operación, para garantizar que la ecuación de coordinación se cumpla para cualquier nivel de corriente de falla.

A. Adaptación de la corriente de arranque Se propone hacer variable la corriente de arranque Ia del elemento de tiempo inverso en función de la corriente de carga Ic:

ca III +Δ= (9)

donde ΔI representa un margen de seguridad, con valor propuesto del 15% de la corriente máxima. Con (9) se obtiene un valor dinámico en la corriente de arranque del relevador en función de la corriente de carga del sistema. Al variar este ajuste se varía la sensibilidad del relevador. Este proceso es ventajoso sobre todo durante regímenes de carga mínimos; mientras que el relevador convencional presenta una sensibilidad reducida, el relevador adaptivo tendrá mayor sensibilidad. En la Fig. 2 se representa el comportamiento del ajuste adaptivo ante variaciones en estado estable de la corriente de carga; para efectos de comparación se incluye el ajuste de un relevador convencional. Esta propiedad de adaptabilidad puede representarse mediante una característica de operación representada en un plano tiempo-impedancia T=F(Z,Ia); esta característica, a diferencia del caso del relevador convencional, posee una tercera variable, que es la corriente de arranque Ia, por lo que debe representarse en forma tridimensional. En la Fig. 3 se representan las características T=F(Z) para el relevador convencional; la representación es tridimensional, como base para la comparación con el caso adaptivo. La característica está en realidad en el plano correspondiente a Ia=50 A para este ejemplo. La región sombreada representa todas las posibles características T=F(Z) en diferentes estados del sistema.

En la Fig. 4 se muestran las características T=F(Z, Ia) del relevador adaptivo. Se observa que la zona de operación del relevador se convierte en una superficie con el concepto adaptivo. Se generan dos superficies; la superficie 1 corresponde a fallas bifásicas y la 2 a fallas trifásicas; estas superficies están delimitadas por el intervalo de variación de la corriente de arranque Ia del relevador (Ia máx= 50A, Ia mín= 25A). Se observa en la Fig. 4 que en generación mínima hay mayor sensibilidad y menor tiempo de operación del relevador para los diferentes estados operativos del sistema. Se ha sugerido [6] la posibilidad de calcular Ic como una demanda de corriente. De (9) se obtiene:

IIN

IN

jja Δ+= ∑

=1

1 (10)

donde N debe seleccionarse de modo que el intervalo N(Δt) tenga una duración del orden de uno o varios minutos.

Durante todo el intervalo de cálculo de demanda está vigente en el relevador el valor de Ia calculado al final del intervalo anterior. La acción de filtrado pasabajos inherente en el concepto de demanda de corriente simplifica la lógica del relevador adaptivo.

Es necesario mantener Ia constante durante las fallas para asegurar la correcta operación del elemento de tiempo

0

10

20

30

49

49.5

50

50.5

51

0

0.1

0.2

0.3

0.4T

Ia

Z

Fig. 3. Característica T=f(Z) de un relevador convencional.

0

10

20

30

20

30

40

50

0

0.1

0.2

0.3

0.4T

Ia

Z

2

1

Fig. 4. Característica T=f(Z,Ia) de un relevador adaptivo.

2 4 6 8 10 12 14 16 0 1 2 3 4 5 6 7 8 9

10 I

T

Ic

Ia conv= k Icmáx

Ia

Fig. 2. Corriente de arranque adaptiva y convencional.

inverso. La lógica de control de la corriente de arranque evita la variación del valor de Ia, aún cuando el fin del intervalo de demanda ocurra durante el período de falla. Si la línea está desenergizada, la lógica de control asigna un valor máximo de Ia; este valor debe ser igual al ajuste de un relevador convencional. Se utilizaron señales de prueba para evaluar la operación del relevador adaptivo ante diferentes estados operativos. En la Fig. 5 se muestra el comportamiento de la corriente de arranque adaptiva para un caso de prueba. Puede observarse que la corriente de arranque Ia calculada según (10) sigue las variaciones de la corriente de carga en forma escalonada. Al ocurrir una falla, la corriente de arranque permanece constante. Esta condición se cumple, debido a que durante todo el intervalo de cálculo de demanda está vigente en el relevador el valor de Ia calculado al final del intervalo anterior. La acción de filtrado pasabajos inherente en el concepto de demanda de corriente simplifica la lógica del relevador adaptivo.

Los criterios adaptivos propuestos para la adaptación de la corriente de arranque Ia mejoran la sensibilidad del elemento de tiempo inverso y reducen el tiempo de operación del relevador, como se demuestra en las pruebas de laboratorio. B. Adaptación de la curva de tiempo La idea básica de la adaptación de la curva de tiempo es que para cualquier valor de corriente se cumpla:

( ) ( ) TITIT nknkn Δ+= −

−1

1 (11)

La adaptación consiste en determinar la función Tn(Ik) del elemento de tiempo inverso adaptivo que haga que opere con un retardo de tiempo ΔT con respecto al relevador respaldado, para todo valor de corriente. Esto significa modificar la forma de la curva de tiempo del relevador, y su equivalente en relevadores analógicos sería modificar el diseño del relevador.

La ecuación de operación del relevador resulta en este caso de sustituir la versión de (11) para una muestra en (8). El múltiplo de corriente es determinado con el valor de corriente de arranque del dispositivo respaldado ( ) 11 / −− = n

akrnk III :

( )11

11

1

=Δ+

Δ ∑=

−−

opk

knkn TIT

t (12)

Cuando la curva del relevador respaldado es conocida analíticamente, el ajuste de curvas se realiza sustituyendo

( )11

−−

nkn IT por su ecuación. Cuando no se conoce la ecuación

de la característica tiempo-corriente del relevador respaldado, pero se dispone de su gráfica, es posible obtener parejas de valores (I, T) de esa característica y utilizar un procedimiento fuera de línea de ajuste de curvas para obtener ( )1

1−

−nkn IT .

V. PROGRAMA DE AJUSTE DE CURVAS

En la Fig. 6 se muestra el diagrama de flujo del programa

de ajuste de curvas. El programa cuenta con subrutinas de regresión no lineal y regresión polinomial, que incluyen las ecuaciones propuestas en [7,8,9,10]. En estas subrutinas el programa selecciona automáticamente el orden del polinomio que da el mejor ajuste. El programa permite obtener los resultados del ajuste para una determinada ecuación, y determinar automáticamente el mejor ajuste resultante de todas las ecuaciones disponibles. Esta segunda opción es de gran valor, pues por lo general no se conoce de antemano la ecuación que da el mejor ajuste de la característica de un determinado relevador. También es importante esta opción para el ajuste de características de otros dispositivos de protección de sobrecorriente, como fusibles o restauradores automáticos. Se acepta como un buen ajuste aquel en que, para cualquier valor de corriente, el tiempo calculado por la

0 1 2 3 4 5 6 7 0

0.5

1

1.5

2

I (A )

T (s)

Icc

Ic

Ia

Energización

Variación de corr iente de carga en estado estable

Falla y disparo

1 2 3 2

1

2

3

1

Fig. 5. Comportamiento de la corriente de arranque adaptiva.

Datos

Subrutina de regresión no

lineal

Selección de la mejor ecuación

Selección del mejor polinomio

Determinación del mejor ajuste

Resultados: • Ecuación • Coeficientes • Coeficientes • Indicadores

estadísticos

Algoritmo Regresión No lineal

Algoritmo Regresión Polinomial Subrutina de

regresión polinomial

Fig. 6. Diagrama de flujo del programa de ajuste de curvas.

ecuación tenga un error no mayor de tres ciclos con respecto al valor exacto [8,9,10]. Estos requerimientos de exactitud se aplican por lo general a partir del múltiplo 2. En este trabajo se evaluaron las curvas en todo el intervalo de valores de corriente dado por los fabricantes. Para cualquier dispositivo seleccionado siempre se obtuvo un ajuste menor que 2 ciclos de error. La obtención de una expresión analítica para dispositivos tales como relevadores, fusibles y restauradores hace válida la expresión de adaptabilidad de tiempo propuesta en (12). Con la adaptación del tiempo de operación se obtiene una coordinación automática, conservando un intervalo de coordinación constante para cualquier valor de corriente de falla. Este criterio evita la necesidad del proceso de coordinación por parte del usuario.

VI. EJEMPLOS DE COORDINACIÓN

En el sistema radial de 13.8 kV mostrado en la Fig. 7 se muestra un ejemplo de coordinación entre dos relevadores convencionales (n-2 y n) y el relevador (n-1), que para fines de comparación se define como convencional (n-1)1 y adaptivo (n-1)2. Se ilustran los niveles máximos de corriente de cortocircuito para cada punto de coordinación entre protecciones.

En la Fig. 7 se muestra el resultado de la

coordinación de los relevadores. El proceso de coordinación fue efectuado con un “software” comercial de análisis de sistemas eléctricos [11]. El relevador n-2 tiene incorporados los elementos de tiempo inverso e instantáneo. La curva del relevador adaptivo se obtuvo editando la característica de operación del relevador n-2 con el editor de relevadores del mismo programa comercial. El proceso es el siguiente: se copia la característica de operación del relevador respaldado (n-2), se edita la característica sumando el intervalo ΔT (0.3 a 0.4 seg) a los valores de tiempo (10 puntos) que describen esta característica, para los mismos valores de corriente; finalmente, la curva resultante es almacenada con otro nombre.

Se observa la reducción en el tiempo de operación

de la protección (n-1)2 en la zona de respaldo de n-2, con una disminución de tiempo de hasta un segundo con respecto a la curva del relevador convencional. También se observa que para el punto de coordinación con n-2 (ajuste del elemento instantáneo = 200A) ambas protecciones conservan el intervalo de coordinación de 0.3 segundos.

El tiempo de operación del relevador adaptivo para

fallas en su línea protegida (sección a-b de la Fig. 7) es incrementado (así como el del relevador n) con respecto al tiempo del relevador convencional (puntos 1 y 2 de la Fig. 7), debido a la falta de convergencia de

sus curvas de tiempo. La adición del elemento instantáneo en n-1 elimina este incremento en tiempo. Donde no es posible aplicar el elemento instantáneo, el incremento de tiempo puede ser permitido porque la reducción de tiempo de n-1 con respecto al incremento de tiempo de n (0.1 seg) es mucho mayor; para el ejemplo mostrado en la Fig. 7 es de 10 veces.

En la Fig. 8 se presenta un ejemplo de coordinación entre

un fusible (n-2), un relevador adaptivo (n-1) y un relevador convencional (n). En el sistema radial se indican los niveles máximos de corriente de cortocircuito; se considera que el cable del alimentador entre las barras A y B es de poca longitud, por lo que los niveles de cortocircuito son comparables. El relevador adaptivo es la réplica de la curva de liberación máxima del fusible, más el ΔT de coordinación; la curva del relevador adaptivo fue obtenida mediante el “software” de análisis [11], siguiendo el mismo proceso que se utilizó para la coordinación mostrada en la Fig. 7.

El relevador convencional n se puede coordinar con el

fusible como si el relevador adaptivo no existiese, pero dejando un intervalo de coordinación igual a 2ΔT entre ambas curvas. Si se tiene la representación gráfica del relevador adaptivo, obtenida mediante el proceso indicado en el segundo párrafo de la presente sección, la coordinación entre n y n-1 se realiza directamente con la curva del relevador adaptivo.

En el diagrama de la Fig. 9 se muestra una red radial de

13.8 kV donde se tiene un restaurador (n-2), un relevador adaptivo (n-1) y un relevador convencional (n). En la misma figura se presenta el resultado de la coordinación; se considera la curva acumulada del restaurador para efectos de coordinación, de modo que no es necesario considerar el tiempo de restauración y el efecto de sobreviaje del disco. El relevador adaptivo duplica la característica de operación de la curva acumulada del restaurador, más el ΔT de coordinación. La curva del relevador adaptivo es obtenida mediante el

10 100 1000 Corriente (A)

T (s)

10

1

0.1

(n-2)

(n-1)2

(n-1)1 (n)

A

B

n

n-1

n-2

200 A

500 A

ΔT

ΔT

a b

1

2

Fig. 7. Ejemplo de coordinación con un relevador convencional.

“software” de análisis [11] con el mismo proceso que se utilizó en los ejemplos anteriores. La coordinación del relevador convencional puede realizarse en forma directa con la curva del relevador adaptivo, o puede realizarse directamente con el restaurador n-2.

100

10

1

0.1

1 10 100 1000

Tiempo (seg)

Corriente (A) 430

A

B

n

n-1

n-2

50 A

50 A

(n-2)

(n-1)

(n)

50 AMP

50 AMP

Fig. 8. Ejemplo de coordinación relevador convencional - relevador adaptivo - fusible.

10 100 1000 Corriente (A)

100

Tiempo (seg)

10

1

0.1

A

B

n

n-1

n-2 R

500 A

500 A

(n-2)

(n-1)

(n)

500 AMP

500 AMP

Fig. 9. Ejemplo de coordinación relevador convencional - relevador

adaptivo - restaurador.

VII. ESTRUCTURA FUNCIONAL El diagrama funcional del relevador adaptivo de sobrecorriente de tiempo inverso se muestra en la Fig. 10. El controlador de tiempo de operación recibe como entradas el valor de ajuste de la corriente de arranque Ia(n-1) y la expresión analítica de la característica ( ) ( )kn IFT =−1 del dispositivo respaldado. También recibe los valores de la impedancia de línea ZL y el voltaje del sistema VLL para el cálculo de la corriente de cortocircuito en la ubicación del dispositivo respaldado. El controlador de tiempo de operación calcula el valor de la función ( )kn IH que será integrada por el elemento de tiempo inverso. Este proceso de integración establece el comportamiento dinámico del relevador ante corrientes de falla variables. El bloque controlador de corriente de arranque recibe como entradas el valor máximo admisible de la corriente de arranque, Iamáx(n) y la magnitud del fasor correspondiente (Ir)k,. Este bloque emite a su salida una señal lógica 1/0, como señal permisiva de operación del relevador. Si se ha detectado una falla (Icc > Ia) el bloque controlador de corriente de arranque emite a su salida la señal lógica 1 para establecer la condición de arranque y dar inicio al proceso de integración por parte del bloque del elemento de tiempo inverso. La señal 0 es una orden de bloqueo, que impide o interrumpe el arranque del relevador, esta señal se emite cuando se cumple la condición (Icc < Ia). El bloque controlador de corriente de arranque también verifica que no se viole el límite establecido para la corriente de arranque (Iamáx). Si se cumple la condición (Icc < Ia(n-1)) el elemento de tiempo inverso utiliza la Ia calculada en el propio relevador adaptivo; con esto se permite en forma conveniente la operación del relevador adaptivo cuando el dispositivo respaldado no tenga la sensibilidad suficiente para operar.

Por último, en el elemento de salida se verifica la condición de operación y se emite la señal de disparo al interruptor.

Elemento de salida

Filtro digital

Controlador de corriente de arranque

(Ir)k (ir)k

Elemento de tiempo inverso

Controlador de tiempo

de operación ( )kn IH ( )1−naI

( ) ( )kn IFT =−1

( )nmáxaI 1/0

aI

LLL VZ −,

Fig. 10. Diagrama funcional simplificado de un relevador adaptivo de

sobrecorriente.

Los bloques de control para la adaptación de la corriente de arranque y el tiempo de operación pueden ser implementados de forma independiente; esto se aprecia en la Fig. 10, donde la información de entrada y salida de cada bloque es independiente. Un mayor nivel de adaptación puede obtenerse con la utilización combinada de estos dos criterios adaptivos. La información requerida por los bloques de control está disponible en forma local, por lo que el relevador adaptivo no requiere enlaces de comunicación.

VIII. IMPLEMENTACIÓN La estructura general del relevador adaptivo

(Fig. 11) consta de un módulo de conexión, para concentrar las señales de entrada y salida, una tarjeta de adquisición de datos, y una computadora personal, donde residen los programas de operación del relevador adaptivo. El diagrama de bloques del programa del relevador adaptivo se muestra en la Fig. 12, donde se indican las funciones principales de cada subrutina.

La subrutina de adquisición de señales de entrada (ver Fig. 12) incluye un módulo de pruebas, con dos alternativas: a) Adquisición por medio de lectura de señales de archivos; esta opción permite extraer datos de archivos externos en formato ASCII generados en programas de simulación (como el EMTP), o de archivos de datos que contienen registros de fallas reales; b) Generación de señales internas; esta variante ofrece gran versatilidad para la simulación de diferentes estados operativos, variación de parámetros, procesamiento digital y contaminación de la señal con ruidos diversos.

En la subrutina de acondicionamiento y filtrado

digital se realiza la formación de la ventana rectangular de datos de un ciclo de longitud, con 16 muestras por ciclo. Se realiza un filtrado digital de la señal para obtener el fasor correspondiente de la componente fundamental; este fasor representa el valor eficaz fundamental de la corriente que es suministrado al relevador digital. Por último en esta

subrutina se realiza el cálculo de la corriente de arranque (demanda) del relevador.

La subrutina de protección es la encargada de realizar el

control de la corriente de arranque y del tiempo de operación de la protección adaptiva. Se realiza el proceso de actualización del ajuste conforme al valor calculado de corriente de arranque y el proceso de integración de la función H(Ik) para determinar el tiempo de operación del relevador. La subrutina de tolerancia a la carga fría realiza un cambio temporal en el ajuste del relevador, se utilizan dos pasos de ajuste tanto en la magnitud de la corriente de arranque como en el tiempo de duración del ajuste.

La subrutina de salida realiza la emisión de la señal de

disparo del relevador a través de un canal analógico de la tarjeta de adquisición. Se representa en forma gráfica el estado del integrador. Esta función es de utilidad en pruebas de coordinación con restauradores. Se cuenta con dos salidas gráficas: la señal de entrada al relevador, y el ajuste del relevador junto con la corriente de carga del sistema. También es posible almacenar la corriente de arranque del relevador y la corriente de carga del sistema en un archivo de datos (EXCEL).

IX. PRUEBAS Y RESULTADOS

El desempeño del relevador adaptivo ha sido evaluado

exhaustivamente con pruebas de laboratorio y simulación digital. En las pruebas se utilizaron las tres alternativas de adquisición de señales: archivo de señales, generación interna y adquisición en tiempo real. A continuación se presentan los resultados obtenidos con la adquisición de

Modelo físico del sistema eléctrico de

potencia

Transformador de corriente Interruptor

Módulo de conexión

Tarjeta de adquisición

de datos

Computadora personal

Señal de corriente

Señal de disparo

Relevador adaptivo

Fig. 11. Estructura general del relevador adaptivo.

Adquisición de señales de entrada

Acondicionamiento y

filtrado digital

Subrutina de protección

Subrutina de salida

Adquisición de señales en tiempo real Filtrado anti-aliasing Conversión A/D Lectura de señales de un archivo Generación interna de señales

Formación de ventana de datos Cálculo de fasores (filtrado digital) Cálculo de demanda

Control adaptativo de Ia Control adaptativo de T Algoritmo de protección de

sobrecorriente

Disparo de interruptor Estado del integrador (Posición

del disco) Salidas gráficas Registro de eventos

Fig. 12. Diagrama de bloques del programa.

señales en tiempo real. En estas pruebas se simulan diferentes estados operativos del sistema para evaluar la lógica de operación del relevador adaptivo; se utilizó el sistema mostrado en la Fig. 13.

A. Adaptación de la corriente de arranque

En la Fig. 14 se presentan los resultados de una prueba. Se observa la variación escalonada en intervalos de demanda de la corriente de arranque Ia. En cualquier instante de tiempo está vigente el valor de Ia calculado al final del intervalo anterior. Al ocurrir la falla el relevador mantiene constante el valor de Ia, asegurando la sensibilidad necesaria para la detección de la falla.

Se realizaron pruebas de adquisición de señales con

secuencias de operación de energización, falla y disparo. El cortocircuito de 6 A no es detectado por la protección primaria (relevador n-1), ni por el relevador convencional de respaldo n. Este cortocircuito sí es detectado y eliminado por el relevador adaptivo n. El valor de Ia es el correspondiente al período anterior de cálculo de demanda, y permanece constante durante la falla; posteriormente, al eliminarse la corriente de cortocircuito y tenerse la condición de línea abierta, el valor de Ia es ajustado a su valor máximo.

El relevador adaptivo para este evento tiene una

corriente de arranque más sensible que el relevador convencional n-1, permitiendo la pérdida conveniente de selectividad para que el relevador adaptivo opere para la falla en el bus C. El relevador adaptivo operará con una característica de operación igual a la asignada al relevador n-1, pero con la Ia calculada por el propio relevador adaptivo.

B. Adaptación del tiempo de operación Las pruebas con corriente de falla constante

consisten en la medición del tiempo de operación del relevador adaptivo respaldando al relevador electromecánico para un valor constante de corriente de falla. El relevador adaptivo cumple funciones de respaldo cuando tiene un tiempo de operación ΔT (0.3 segundos) mayor que el relevador respaldado para cualquier valor de corriente de falla; esto garantiza una coordinación automática entre estos

relevadores para cada valor constante de corriente de falla. Se tomaron 10 parejas de valores tiempo-corriente de la

curva de un relevador electromecánico, publicada por el fabricante. La palanca de tiempo es 5. Los valores de corriente se definieron con espaciamiento uniforme según se propone en [9].

La expresión analítica obtenida del programa con menor

error fue la ecuación polinomial de la forma:

( ) ( ) ( ) ( ) ( )5432 log278.710

log374.2098

log571.2418

log038.1342

log416.362608.37

IIIIIT +−+−+−=

Esta expresión, introducida al relevador adaptivo, define

su característica de operación. En la Tabla 1 se muestran los tiempos obtenidos para el relevador electromecánico y los tiempos calculados por el relevador adaptivo. Se observa que el intervalo entre los tiempos de operación del relevador electromecánico y el adaptivo es muy cercano al intervalo de coordinación previamente definido (0.3 segundos). Las variaciones observadas son debidas al ruido presente durante la prueba.

En la Fig. 15 se representan en forma gráfica los

resultados mostrados en la Tabla 1. Se presentan los valores tiempo-corriente obtenidos en un plano bilogarítmico. Se grafican los valores tiempo - corriente del relevador electromecánico y del relevador adaptivo. Se observa que el intervalo de coordinación es conservado, garantizando así una coordinación automática entre estos dos relevadores.

X. CONCLUSIONES

• El relevador adaptivo propuesto en este trabajo modifica

su corriente de arranque en función de la corriente de carga, y ajusta su tiempo de operación en función de la respuesta de operación del dispositivo respaldado.

• La introducción del concepto de demanda de corriente en el

cálculo de la corriente de arranque simplifica la lógica de control de la corriente de arranque del relevador.

Ia

Ic

Icc

t (s)

I (A)

Iamáx

Fig. 14. Prueba con adquisición de la señal en tiempo real.

A B C

100 V Carga: 2-3.5 A Falla: 6 A

Icarga máx= 8.33 A n-1

IcB=3.33 A

Icarga máx= 5 A n

AIacn 5.12= AIa c

n 5.71 =−

Fig. 13. Modelo físico del sistema eléctrico de potencia.

• El relevador adaptivo no requiere ningún valor de ajuste y no necesita coordinarse; estos procesos se realizan de forma automática por el relevador.

• Los criterios adaptivos propuestos, la adaptación de

la corriente de arranque y adaptación del tiempo de operación, pueden ser implementados de forma independiente.

• Las pruebas realizadas al relevador adaptivo con

señales reales adquiridas en tiempo real dieron resultados que cumplen con los criterios de coordinación de tiempo y corriente.

XI. REFERENCIAS

[1] K.R. Shah, E.D. Detjen and A.G. Phadke, “Feasibility of adaptive distribution protection system using computer overcurrent relaying concept,” IEEE Transactions on Industry Applications, vol. 24, No. 5, September/October 1988, pp. 792-797.

[2] M.S. Sachdev, T.S. Sidhu, B. Chattopadhyay, et. al., “Design and evaluation of an adaptive proteccion system for a distribution network,” Cigré Paper 34-202, París, 1995.

[3] J. Eisman, G. Gómez and J. Torres, “Applied adaptive protection practices based on data transmission between relays,” Cigré Paper 34-207, París, 1995.

[4] H.J. Altuve, y G.P. Kasianov, “Simulación matemática de un relevador estático de sobrecorriente,” Ingeniería Energética (Cuba), Vol. VI, no. 4, 1985, pp. 347-353.

[5] IEEE Std C37.112-1996, IEEE Standard Inverse-Time Characteristic Equations for Overcurrent Relays, September 1996.

[6] A. Guzmán, Schweitzer Engineering Laboratories Inc. Comunicación personal.

[7] J.E. Hieber, “Empirical equations of overcurrent relay curves for computer application,” IEEE Winter Power Meeting, New York, N.Y., January/February 1965, Paper No. 31 CP 65-91, pp. 1-11.

[8] G.E. Radke, “A method for calculating time-overcurrent relay settings by digital computer,” IEEE Transactions on Power Apparatus and Systems, Vol. 82, Special Supplement, 1963, pp. 189-205.

[9] IEEE Committee Report, “Computer representation of overcurrent relay characteristics,” IEEE Transactions on Power Delivery, Vol. 4, No. 3, July 1989, pp. 1659-1667.

[10] M.S. Sachdev, J. Singh, and R.J. Fleming, “Mathematical models representing time-current characteristics of overcurrent relays for computer application,” IEEE Paper A78 131-5, January 1978.

[11] Professional Electrical Power System Design and Simulation Software EDSA, ver. 2.95, EDSA, San Diego, CA, USA.

XII. BIOGRAFÍAS

Arturo Conde Enríquez nació en Naucálpan de Juárez, Edo. de México, México, en 1971. Se graduó de Ingeniero Mecánico Electricista en la Universidad Veracruzana en 1993. Obtuvo la Maestría en Ciencias de la Ingeniería Eléctrica y el grado de doctor en Ingeniería Eléctrica en la Universidad Autónoma de Nuevo León en 1996 y 2002 respectivamente. Actualmente es catedrático de la Facultad de Ingeniería Mecánica y Eléctrica de la Universidad Autónoma de Nuevo León. Ernesto Vázquez Martínez e graduó de Ingeniero en Electrónica y Comunicaciones en 1988, y obtuvo su Maestría y Doctorado en Ingeniería Eléctrica en 1991 y 1994 respectivamente, en la Universidad Autónoma de Nuevo León, México. Desde 1996 es Profesor Investigador del Programa Doctoral en Ingeniería Eléctrica de la Universidad Autónoma de Nuevo León, México, y de 2000 a 2001 realizó una estancia posdoctoral en la Universidad de Manitoba, en Canadá. Actualmente es el Coordinador del Postgrado en Ingeniería Eléctrica de la misma universidad. Es miembro del Instituto de Ingenieros en Electricidad y Electrónica (IEEE) de Estados Unidos y miembro del Sistema Nacional de Investigadores de México, Nivel I. Sus áreas de investigación son la protección de sistemas eléctricos de potencia y la aplicación de técnicas de inteligencia artificial en sistemas eléctricos de potencia. Héctor J. Altuve Ferrer se graduó de ingeniero electricista en la Universidad Central de Las Villas (UCLV), Cuba, en 1969, y obtuvo su grado de Doctor en Ingeniería Eléctrica en el Instituto Politécnico de Kiev, URSS, en 1981. Fue profesor de la Facultad de Ingeniería Eléctrica de la UCLV entre 1969 y 1993. Fue profesor del Programa Doctoral de la Facultad de Ingeniería Mecánica y Eléctrica de la Universidad Autónoma de Nuevo león, en Monterrey, México, entre 1993 y 2001. Fue miembro del Sistema Nacional de Investigadores de México. Fue Profesor visitante de Washington State University, Pullman, WA, U.S.A., en el curso académico 1999-2000. En enero del 2001 comenzó a trabajar como Senior Research Engineer de Schweitzer Engineering Laboratories (SEL), con oficina en Monterrey, N.L., México. En noviembre del 2001 pasó a ser el Director General de SEL en México. Es Senior Member del Instituto de Ingenieros en Electricidad y Electrónica (IEEE) y Conferencista Distinguido de la Sociedad de Potencia del IEEE. Es autor o coautor de publicaciones técnicas y patentes de invención.

10

Corriente (A)

Tiempo (sec)

10

1

0.5

20

Relevador adaptivo

Relevador convencional

Fig. 15. Características de operación de un relevador electromecánico

y el adaptivo obtenidas en pruebas de laboratorio.

Tabla 1. Tiempos de operación promedio e intervalos de coordinación.

Tiempo de operación

Icc Relevador adaptivo

Relevador electromecánico

? Tmed

3.18 6.63 6.26 0.37 4.12 3.99 3.69 0.30 5.10 2.93 2.56 0.37 6.44 2.13 1.75 0.38 8.23 1.61 1.31 0.30 10.0 1.41 1.08 0.33 11.9 1.31 1.00 0.31 13.9 1.22 0.9 0.32 14.7 1.18 0.88 0.30 15.7 1.14 0.81 0.33

1

This paper reviews ground fault protection and detection methods for distribution systems. First, we review and compare medium-voltage distribution-system grounding methods. Next, we describe directional elements suitable to provide ground fault protection in solidly- and low-impedance grounded distribution systems. We then analyze the behavior of ungrounded systems under ground fault conditions and introduce a new ground directional element for these systems. Then we examine the behavior of compensated systems during ground faults and describe traditional fault detection methods. We conclude by introducing new ground fault detection methods for compensated systems.

I. INTRODUCTION Ground fault current magnitudes depend on the system

grounding method. Solidly- and low-impedance grounded systems may have high levels of ground fault currents. These high levels typically require line tripping to remove the fault from the system. Ground overcurrent and directional overcurrent relays are the typical ground fault protection solution for such systems. However, high-impedance ground fault detection is difficult in multigrounded four-wire systems, in which the relay measures the ground fault current combined with the unbalance current generated by line phasing and configuration and load unbalance.

Ungrounded systems have no intentional ground. For a single-line-to-ground fault on these systems, the only path for ground current to flow is through the distributed line-to-ground capacitance of the surrounding system and of the two remaining unfaulted phases of the faulted circuit.

In resonant-grounded or compensated distribution networks the system is grounded through a variable impedance reactor connected to the power transformer secondary neutral or the neutral of a grounding bank. This reactor compensates the system phase-to-ground capacitance such that the zero-sequence network becomes a very high impedance path. The reactor, known as the Petersen coil, permits adjustment of the inductance value to preserve the tuning condition of the system for different network topologies.

Resonant grounding provides self-extinction of the fault arc in overhead lines for about 80 percent of temporary ground faults [1]. Considering that about 80 percent of ground faults are temporary, we conclude that more than 60 percent of overhead line ground faults clear without breaker tripping. High-impedance grounded systems are grounded through a high-impedance resistor or reactor with an impedance equal to

or slightly less than the total system capacitive reactance to ground. The neutral resistor is of such a high value that ground faults on such systems have very similar characteristics to those of resonant-grounded systems.

Because ground faults in ungrounded, high-impedance grounded, and compensated systems do not affect the phase-to-phase voltage triangle, it is possible to continue operating either system in the faulted condition. However, the system must have a phase-to-phase insulation level and all loads must be connected phase-to-phase.

Ground relays for these systems require high relay sensitivity because the fault current is very low compared to solidly grounded systems. Most ground-fault detection methods use fundamental-frequency voltage and current components. The varmetric method [2] is the traditional ground fault detection solution in ungrounded systems. We may also use this method in high-impedance grounded systems. The wattmetric method [2][3] is a common directional element solution for compensated systems, but its sensitivity is limited to fault resistances no higher than a few kilohms. We may also use the wattmetric method in high-impedance grounded systems and isolated neutral systems. Other fundamental-frequency methods for compensated systems (such as the admittance method [4][5]), provide increased sensitivity but require information about all feeders, about the possibility of making control actions on the Petersen coil, or both. There are also methods that use the steady-state harmonic content of current and voltage to detect ground faults [6][7]. Another group of methods detects the fault-generated transient components of voltage and current [6][8]. These methods have limited sensitivity, because high-resistance faults reduce the level of the steady-state harmonics and damp the transient components of voltage and current.

II. GROUNDING METHODS OF MEDIUM-VOLTAGE DISTRIBUTION NETWORKS

The main goals of system grounding are to minimize voltage and thermal stresses on equipment, provide personnel safety, reduce communications system interference, and give assistance in rapid detection and elimination of ground faults.

With the exception of voltage stress, operating a system as ungrounded, high-impedance grounded, or resonant grounded restricts ground fault current magnitudes and achieves most of the goals listed above. he drawback of these grounding methods is that they also create fault detection (protection)

Review of Ground Fault Protection Methods for Grounded, Ungrounded, and Compensated

Distribution Systems Jeff Roberts, Dr. Hector J. Altuve, and Dr. Daqing Hou

Schweitzer Engineering Laboratories, Inc. Pullman, WA USA

2

sensitivity problems. We can create a system grounding that reduces voltage stress at the cost of large fault current magnitudes. However, in such a system the faulted circuit must be deenergized immediately to avoid thermal stress, communications channel interference, and human safety hazards. The disadvantage of this system is that service must be interrupted even for temporary faults.

The following is a brief description of the grounding methods typically used in medium-voltage distribution circuits. Table I summarizes the main characteristics of these grounding methods.

A. Ungrounded or Isolated Neutral In an isolated neutral system (see Fig. 1), the neutral has no

intentional connection to ground: the system is connected to ground through the line-to-ground capacitances. Single line-to-ground faults shift the system-neutral voltage but leave the phase-to-phase voltage triangle intact.

A

B

CGround

faultG

CA CB CC

CAB CAC

CBC

N

Fig. 1. Isolated Neutral System

For these systems, two major ground fault current magnitude-limiting factors are the zero-sequence line-to- ground capacitance and fault resistance. Because the voltage

triangle is relatively undisturbed, these systems can remain operational during sustained, low-magnitude faults.

Self-extinction of ground faults in overhead-ungrounded lines is possible for low values of ground fault current. At higher magnitudes of fault current, faults are less likely to self-extinguish at the fault current natural zero-crossing because of the high transient recovery voltage. Later, we discuss how a resonant-grounded system damps this recovery voltage rise, thereby increasing the likelihood of causing the ground fault to self-extinguish.

Zero-sequence [9], or three-phase voltage relays can detect ground faults in ungrounded systems. This method of fault detection is not selective and requires sequential disconnection or isolation of the feeders to determine the faulted feeder. A sensitive, directional ground varmetric element is the typical alternative to sequential disconnection [2]. These elements respond to the quadrature component of the zero-sequence current with respect to the zero-sequence voltage. Later we introduce a new directional element that uses the measured impedance as the measurand for differentiating forward and reverse ground faults.

B. Effective or Solid Grounding Effective, or solid, grounding is popular in the United

States. To be classified as solidly grounded, the system must have (X0 / X1) 3 and (R0 / X1) 1, where X0 and R0 are the zero-sequence reactance and resistance, and X1 is the positive-sequence reactance of the power system [10]. In practice, solidly grounded systems have all power system neutrals connected to earth (or ground) without any intentional impedance between the neutral and earth.

TABLE I COMPARISON OF GROUNDING METHODS FOR MEDIUM-VOLTAGE DISTRIBUTION NETWORKS

Grounding Method

Issues Isolated Neutral

Solid Grounding (Unigrounding)

Solid Grounding (Multigrounding)

Low-Impedance Grounding

High-Impedance Grounding

Resonant Grounding

Some Countries of Application

Italy, Japan, Ireland,

Russia, Peru, Spain

Great Britain USA, Canada,

Australia, Latin America

France, Spain Northern and

Eastern Europe, China, Israel

Permissible Load Connection Phase-phase

Phase-phase (3 wires) and phase-neutral (4 wires)

Phase-phase and phase-ground Phase-phase Phase-phase Phase-phase

Required Insulation Level Phase-phase Phase-neutral Phase-neutral Phase-neutral Phase-phase Phase-phase

Limitation of Transient Overvoltages Bad Good Good Good

Good (R-grounding), Average (L-grounding)

Average

Possible Operation With a Ground Fault Not always No No No Not always Almost always

Self-Extinguishing of Ground Faults Not always No No No Not always Almost always

Human Safety Average Good Bad Good Average Good Equipment Thermal Stress Low High High High Low Lowest Interference With Communication Lines Average High High High Low Lowest

Ground Fault Protection Sensitivity Average Good Bad Good Average Average

3

There are two different practical implementations of solid grounding in medium-voltage distribution systems: unigrounded and multigrounded. In unigrounded systems there may only be three wires with all loads connected phase-to-phase (see Fig. 2(a)), or there may be four wires with an isolated neutral and all loads connected phase-to-neutral (see Fig. 2(b)). In the latter application the load unbalance current returns through the neutral while the ground fault current returns through the earth to the substation neutral. In multigrounded systems with four wires and phase-to-neutral loads (see Fig. 2(c)), the system is grounded at the substation and at every transformer location along the circuit. In some instances some single-phase branch loads are connected to a line and earth without running a neutral conductor. In these systems both load unbalance and ground fault currents divide between the neutral conductor and earth. Detecting high-resistance ground faults on these systems is difficult because the protective relay measures the high-resistance ground fault current combined with the unbalance current.

Ground faults on these systems may produce high-magnitude currents that require tripping the entire circuit and interrupting load to many customers. About 80 percent of ground faults occurring on overhead distribution lines are transient. For these systems automatic multishot reclosing is widely used. The resulting interruption/restoration cycle can represent a problem to customers with large rotating loads or those with loads intolerant of voltage sags.

Solid grounding reduces the risk of overvoltages during ground faults. These faults do not shift the system neutral (see Fig. 2(e)). Thus, the system does not require as high a voltage insulation level as does an isolated neutral system. Transmission systems are typically solidly grounded throughout the world. Distribution systems are commonly unigrounded in Great Britain and multigrounded in North America, Australia, and some Latin American countries.

The typical ground fault protection for solidly grounded systems consists of residually connected (or equivalent mathematical summation) nondirectional and directional overcurrent relays. Detecting high-impedance ground faults is difficult for the nondirectional relay applications on four-wire installations with phase-to-neutral loads because you must set the minimum relay sensitivity greater than the normal load unbalance. Coordination with lateral circuit fuses, primarily sized to carry load, is yet another limiting factor for ground protection sensitivity in these systems [10]. As a result, many downed conductors have remained undetected and energized for a significant time. Recent directional relay developments take into account the normal load and line unbalances and do not require significant degradation of ground relay sensitivity. This latest directional element technology greatly improves sensitivity during low load conditions as compared to nondirectional protection, but it still must limit relay sensitivity during periods of very high load flow.

C. Low-Impedance Grounding In this type of grounding the system is grounded through a

low-impedance resistor or reactor with the objective of

limiting the ground fault current. By limiting the ground fault current magnitudes to tens or hundreds of amperes, you reduce equipment thermal stress, which allows you to purchase less expensive switchgear. This method is equivalent to solid grounding in many other ways, including ground fault protection methods.

Many of the distributed networks in France are low-resistance grounded. In rural distribution networks the ground fault current is limited to 150�–300 A primary, and in the urban networks, which have higher capacitive currents, the resistor is selected to limit the ground fault current to a maximum of 1000 A [8]. Industrial plant engineers also use low-impedance grounding in their plant and distribution circuits. D. High-Impedance Grounding

In this method the system is grounded through a high-impedance resistor or reactor with an impedance equal to or slightly less than the total system capacitive reactance to ground. The high-impedance grounding method limits ground fault current to 25 A or less. High-resistance grounding limits transient overvoltages to safe values during ground faults. The grounding resistor may be connected in the neutral of a power or grounding transformer, generator or generator-grounding bus, or across a broken delta connection of distribution transformers [10].

A

C B C B

A, N, GVBN=VBGVCN=VCG

VAN

N, G

VBN

VCN

(d) (e)

A

B

C

N

G

G

Groundfault

(a)

Loads

A

B

C

N

G

GGround fault

(b)Isolated neutral conductor

(c)

A

B

C

N

G

GGround fault

Neutral conductor

Fig. 2. Solidly Grounded Systems: (a) Three-Wire Unigrounded System, (b) Four-Wire Unigrounded System, (c) Four-Wire Multigrounded System, (d) Phasor Diagram for Normal Operation, (e) Phasor Diagram for a Ground

Fault

4

As with isolated neutral systems, ground faults on these systems shift the system neutral voltage without modifying the phase-to-phase voltage triangle. Again, this grounding method permits the utility to continue operating the system during sustained ground faults.

Nonselective ground fault detection is possible by sensing system zero-sequence voltage magnitude and comparing it with an overvoltage threshold, or by measuring all three phase-to-ground voltages and comparing each voltage magnitude against an undervoltage threshold. To find the faulted feeder, you must use sensitive zero-sequence directional elements or disconnect feeders to determine when the zero-sequence voltage drops to a normal level. The traditional directional element is the wattmetric type [2][3], which responds to the in-phase, or active, component of the zero-sequence current with respect to the zero-sequence voltage. For reactance-grounded systems, you can also use a varmetric directional element that responds to the reactive or quadrature component of the zero-sequence current [2].

Typical fields of application for high-resistance grounding include generators connected in a generator-transformer unit [10] and medium-voltage industrial plant distribution networks [11].

E. Resonant Grounding In this method of grounding, the system is grounded

through a high-impedance reactor, ideally tuned to the overall system phase-to-ground capacitance (see Fig. 3). The variable impedance reactor is called a Petersen coil after its inventor, who introduced the concept in 1917. It is also known as an arc-suppression coil or ground-fault neutralizer. The coil is typically connected to the neutral of the distribution transformer or a zigzag grounding transformer. Systems with this type of grounding are often referred to as resonant-grounded or compensated systems. When the system capacitance is matched by the inductance of the coil, the system is fully compensated, or at 100 percent tuning. If the reactor inductance does not match the system capacitance, the system is off tuned. It can be over- or undercompensated, depending on the relationship between inductance and capacitance.

A

B

C

N

G

Ground faultPetersen

coil

CABCAC

CBC

CA CB CC

Fig. 3. Compensated System

Older installations use a low-cost, fixed value reactor. In these systems the tuning condition, whether under- or overcompensated, changes with the configuration of the distribution network. Tap-changing reactors permit manual or automatic control of the tuning conditions. Modern

installations include a moving-core (plunger) reactor equipped with a control system to provide almost 100 percent tuning for all system-operating conditions. These plunger systems also provide a smooth means of system tuning.

Resonant grounding a system can reduce the ground fault current to about 3 to 10 percent of that for an ungrounded system. For 100 percent tuning, the active coil losses, system harmonics, and system active leakage current determine the fault current magnitude [1]. Residual current compensation methods inject a current through the reactor to the system during the fault, reducing the fault current almost to zero [12].

The arc self-extinction action depends not only on the fault current magnitude, but also on the transient recovery voltage rate after successful arc extinction at the current zero crossing. In compensated systems this voltage recovery time is much slower than in ungrounded systems.

Detecting high-impedance faults in compensated distribution circuits requires a device with a very sensitive residual current input. The wattmetric directional method, described later in this paper, is the most commonly used type of directional element. However, the ground fault resistive coverage of this element is limited by the supervisory zero-sequence overvoltage element setting.

Utilities that trip the lines for permanent faults frequently detect the fault by measuring the zero-sequence voltage and then automatically change the system grounding condition. This switching operation is intended to enhance the sensitivity of the selective ground fault protection. The switching operation has an intentional delay of several seconds to allow the compensated system to extinguish the arc for temporary faults. The thermal rating of the Petersen coil sets this delay time. In some systems, [13] the practice is to by-pass the reactor with a single-pole breaker. Another alternative is to connect a resistor in parallel with the Petersen coil or to an auxiliary winding of the Petersen coil. A problem with these methods is that the connection of the resistor reduces the zero-sequence voltage without necessarily significantly increasing the zero-sequence fault current magnitude. The result is that the action taken to increase fault detection sensitivity can decrease the wattmetric ground directional element sensitivity.

III. DIRECTIONAL GROUND-FAULT ELEMENTS FOR SOLIDLY- AND LOW-IMPEDANCE GROUNDED SYSTEMS

Modern distribution systems are becoming looped or multifeed systems because dispersed generation has increased and loops are needed at the distribution level to improve supply reliability to critical loads.

Ground faults in solidly- and low-impedance grounded systems may produce high current levels that require circuit tripping. Use directional elements to provide ground-fault protection in these systems.

Classical ground directional relays (GDRs) respond to either negative- or zero-sequence quantities. For these classical GDRs, we must select which sequence quantities to use for each application and particular system operating conditions. Once you select a GDR model type, with fixed polarizing and operating quantities, the traditional GDRs use

5

these quantities at all times. This restriction may result in directional element misoperation for changing system configurations.

A new GDR, however, selects the best sequence quantities to use for ground faults according to system conditions. It is possible for this new GDR to use a negative-sequence directional element for one fault and a zero-sequence voltage-polarized directional element for the next ground fault.

This new ground directional relay consists of a combination of three directional elements: Zero-Sequence Current-Polarized (32I), Negative-Sequence Voltage-Polarized (32Q), and Zero-Sequence Voltage-Polarized (32V). The new relay uses negative- and zero-sequence voltage-polarized directional elements that overcome the dependability and security problems of traditional voltage-polarized elements.

A. Current-Polarized Directional Element (32I) The 32I element is the traditional current-polarized

directional element. The analog input quantities to this element are the operating quantity, 3I0, and the polarizing quantity, IPOL [14]. The 32I element calculates a torque-like product based on the magnitudes and the relative angle of the analog input quantities (Equation (1)). The 32I element compares the result of the torque calculation, T, against present thresholds. If T is positive and above the positive threshold, the element asserts to declare a forward ground fault. If T is negative and below the negative threshold, the element asserts to declare a reverse ground fault.

03cos�•03�• IPOLIIPOLIT (1)

Where IPOL is the polarizing quantity. 3I0 is the operating quantity: 3I0 = IA+IB+IC All reliable directional elements require supervision. We

enable the 32I element (32IE Enable bit asserts) when all of the following conditions are true:

1) The zero-sequence current, I0, is greater than the positive-sequence current, I1, times the a0 factor (I0 > a0�•I1). The a0 factor increases the 32I element security for zero-sequence currents, which circulate because of line asymmetries, slight CT saturation, etc. [15].

2) The operating quantity, 3I0, is greater than the 50G sensitivity threshold.

3) The polarizing quantity, IPOL, is greater than the preset sensitivity threshold.

4) The E32IV programmable variable asserts (logical 1). The E32IV variable deasserts to identify zero-sequence source isolation [16]. The programmable variable can be set locally or remotely via command or contact input. With this control capability, events that occur locally or in remote parts of the system may control the relay to accommodate new system conditions.

B. Negative-Sequence Voltage-Polarized Directional Element (32Q)

The analog input quantities to this element [17] are the negative-sequence voltage, V2, and the negative-sequence

current, I2. The 32Q element calculates the negative-sequence impedance, Z2, presented to the relay using . If Z2 is below the Z2F threshold the 32Q element declares a forward fault. If Z2 is above the Z2R threshold the 32Q element declares a reverse fault.

22

2�•21�•2Re2

I

ILVZ (2)

Where V2 is the negative-sequence voltage: 3/�•�•2

2 CVaBVaAVV

I2 is the negative-sequence current:

3/�•�•22 CIaBIaAII

L2 is the line negative-sequence impedance angle.

We enable the 32Q element (32QE Enable bit asserts) when all of the following conditions are true:

1) The negative-sequence current, I2, is greater than the zero-sequence current, I0, times the k factor (I2 > k �• I0). In the event that the 32Q and 32V elements have sequence currents above their minimum current sensitivity thresholds, the relay selects the 32Q element if I2 > kI0. This check ensures that the relay uses the most robust analog quantities even if the relay sensitivity settings are not optimized.

2) The negative-sequence current, I2, is greater than the positive-sequence current, I1, times the a2 factor (I2 > a2�•I1). The a2 factor increases the 32Q element security in the same way the a0 factor increases the 32I element security.

3) The negative-sequence current, 3I2, is greater than the 50F or 50R sensitivity threshold. The relay avoids making erroneous directional decisions for low input values of 3I2 by requiring that 3I2 be greater than the 50F or 50R threshold.

C. Zero-Sequence Voltage-Polarized Directional Element (32V)

The 32V element is the zero-sequence analogy of the 32Q element. Equation (3) shows the algorithm used to calculate Z0. The 32V element makes directional decisions in the same way as the 32Q element. The element compares Z0 against the Z0F and Z0R thresholds to determine the direction of the ground fault.

203

03�•01�•03Re0

I

ILVZ (3)

Where V0 is the zero-sequence voltage:

3/0 CBA VVVV

I0 is the zero-sequence current:

6

3/0 CBA IIII

L0 is the line zero-sequence impedance angle.

We enable the 32V element (32VE Enable bit asserts) when all of the following conditions are true:

1) The zero-sequence current, I0, is greater than the positive-sequence current, I1, times the a0 factor (I0 > a0�•I1). The circulating zero-sequence currents, which are due to line asymmetries, are typically less than the circulating negative-sequence currents for most phase conductor configurations [15]. Thus, the a0 factor is usually smaller than the a2 factor. Because of this fact, we can set the 32V element more sensitively than the 32Q element in nontransposed line applications where severe CT saturation is not a possibility.

2) The residual current, 3I0, is greater than the 50F or 50R sensitivity threshold.

3) The E32IV programmable variable asserts (logical 1). The GDR uses the status of the 32IE, 32QE, and 32VE

Enable bits in the relay priority logic to select the optimal directional element to run. This priority logic is explained later in this paper.

D. 32Q and 32V Element Operation for Ground Faults The way that the 32Q and 32V elements declare forward

and reverse ground faults is similar. Let us examine the 32Q element operation for forward and reverse ground faults in a two-source system. Fig. 4 shows the relay negative-sequence voltage, V2, and negative-sequence current, I2, for a ground fault at the remote terminal. I2 is the current contribution from the local end. Notice that the primary current I2 is flowing in at the CT polarity mark. At the relay location, V2 = �–I2·ZS2. If the negative-sequence impedance angles, ZS2 and L2, are the same, the calculated Z2 quantity is Z2 = �–|ZS2 .

Fig. 5 shows the relay quantities, V2 and I2, for a reverse ground fault. The polarity of V2 is the same as for forward ground faults. For reverse faults, the relay current I2 is the contribution from the remote end. The primary current I2 is flowing out at the CT polarity mark. At the relay location, V2 = I2�•(ZL2 + ZR2). If the angles, (ZR2 + ZL2) and L2, are the same, the calculated Z2 quantity is Z2 = ZL2+ZR2 .

ZR2ZL2ZS2

I2

V2

NetworkConnection

NetworkConnection

R Relay Fig. 4. Relay Negative-Sequence Voltage, V2, and Negative-Sequence

Current, I2, for a Ground Fault at the End-Of-Line

ZR2ZL2ZS2

I2

V2

NetworkConnection

NetworkConnection

R Relay

Fig. 5. Relay Negative-Sequence Voltage, V2, and Negative-Sequence Current, I2, for a Ground Fault Behind the Relay

After the Z2 calculation, the relay compares Z2 against the forward and reverse thresholds (Z2F and Z2R, respectively) to make the fault direction declaration. The Z2F threshold must be greater than the maximum Z2 result for forward faults (Z2F > Z2). The Z2R threshold must be less than the minimum Z2 result for reverse faults (Z2R < Z2). The 32V element does identical comparison of Z0 with thresholds Z0F and Z0R. Fig. 6 shows the operating characteristics of the 32Q (Fig. 6 (a)) and 32V (Fig. 6 (b)) elements, and also the measured impedance for forward and reverse faults. For the 32Q element we use ZL1 instead of ZL2 in Fig. 6 (a). Recall that ZL1 = ZL2 for lines.

One of the advantages of the 32Q and 32V elements is that the element sensitivity does not depend on the voltage magnitude at the relay location. For this reason, the elements can be applied in very strong systems where the magnitudes of V2 and V0 are very small.

E. Selecting the Optimal Directional Element (Best Choice Directional Element)

Each of the three directional elements, 32I, 32Q, and 32V, has advantages and disadvantages for various system conditions [18]. The relay selects the optimal directional element for a particular system configuration according to the selected processing sequence and the enable variables, 32IE, 32QE, and 32VE.

Z2 Plane

Z2R = ZL1 /2 + 0.1Z2F = ZL1 /2

R2

X2

ZL1

ZL1+ZR2 (Reverse Faults)

-ZS2 (Forward Faults)

Z0 Plane

(a) 32Q Element (b) 32V Element

Z0R = ZL0 /2 + 0.1Z0F = ZL0 /2

R0

X0

ZL0

ZL0+ZR0 (Reverse Faults)

-ZS0 (Forward Faults)

Fig. 6. Operating Characteristics of the 32Q and 32V Directional

Elements

A setting called ORDER uses the enables described earlier to determine the directional element processing sequence. Assign 32I, 32V, or 32Q enable elements to these variables in the desired sequence. For example, assigning ORDER=32IE, 32QE, 32VE sets the element processing sequence as 32I first, 32Q next, and 32V last.

7

With this processing sequence, the relay uses 32I when IPOL and 3I0 are above the sensitivity thresholds. If the currents do not exceed these thresholds, the 32IE variable does not assert. The relay then proceeds to the 32Q element and checks the status of 32QE. If 32QE does not assert, the relay next checks the status of 32VE. The relay uses the k factor in the 32QE variable to select the most reliable sequence current, I2 or I0, in making the directional decision.

In the automatic setting mode, the GDR provides settings for the 32Q and 32V directional elements. This feature simplifies the relay setting procedure.

The relay selects the most suitable settings from given system parameters. For example, the relay uses the line impedance parameters to set the Z2F, Z2R, Z0F, and Z0R thresholds used in the 32Q and 32V elements.

Fig. 6 shows the calculated sequence impedances, Z2 and Z0, for forward and reverse faults. If we assume infinite sources at both line ends, the line impedance separates the calculated impedances for forward and reverse faults. To safely discriminate between forward and reverse faults, we can set the forward and reverse thresholds at one-half of the corresponding sequence line impedance. The relay sets Z2F to ZL1/2 , and Z2R to ZL1/2+0.1 . Z0F and Z0R are set the same as Z2F and Z2R except the relay uses the zero-sequence line impedance, ZL0, instead of the positive-sequence line impedance, ZL1. These voltage directional element threshold settings guarantee that the relay makes the correct directional decisions for any source-switching conditions.

The relay automatic mode sets the a0 and a2 factors to 0.1. These conservative settings make the directional elements secure under almost all transmission line configurations while allowing sensitive settings for 50F and 50R. The relay automatic selection mode sets 50F to 0.5 A and 50R to 0.25 A. The 50R setting is more sensitive for reverse faults than for forward faults. For example, Directional Comparison Blocking (DCB) schemes require more sensitive reverse elements than forward elements.

IV. UNGROUNDED DISTRIBUTION SYSTEM ANALYSIS In this section we analyze the steady-state behavior of

ungrounded systems in both the phase and the symmetrical component domains.

A. Three-Phase Analysis Fig. 7 shows a simplified representation of a three-phase

ungrounded distribution system. The relay location defines the protected line. All the other distribution lines are lumped in an equivalent line representing the remainder of the distribution system. For simplification in our steady-state analysis, we assume ideal sources operating at nominal frequency and no load, and disregard line series impedances, resistance and reactance. We justify disregarding load on the basis that all loads for these systems must be connected phase-to-phase and thereby do not generate any zero-sequence unbalance. These assumptions introduce no significant error in the results but greatly simplify the calculations.

In Fig. 7 CAL, CBL, and CCL represent the phase-to-ground capacitances of the protected line, and CAS, CBS, and CCS are the phase-to-ground capacitances of the remaining network. We do not represent the phase-to-phase capacitances of the system in Fig. 7 because they do not contribute to the residual current and so are irrelevant to this analysis.

Using the circuit of Fig. 7, we may write:

0CSBSASCLBLAL IIIIII (4)

The relay element measures the residual current 3I0L of the protected line. From (4):

)(3 0 CSBSASCLBLALL IIIIIII (5)

We may represent the occurrence of a forward solid A-phase fault in the system of Fig. 7 by closing Switch SF. In this case, the fault current IF equals IAL:

)( CSBSASCLBLALF IIIIIII (6)

From (5), we see that the residual current measured by the relay is actually the residual current supplied by the remainder of the system. This also shows us that if the protected line were the only feeder connected to the bus, the residual current measured by the relay would equal zero (i.e., 3I0 = IAL �– (IBL + ICL) = 0). For this switching configuration, detecting a ground fault is easily accomplished with a simple zero-sequence overvoltage element.

A

B

C

G

CBL

NN

A

B

CR

R

R

A

B

C

ProtectedLine

RemainingSystem

IBL

CCL

ICL

CAS

IAS

CBS

IBS

CCS

ICS

SFCAL

IAL

Fig. 7. Three-Phase Simplified Representation of an Ungrounded

Network

In a symmetrical unfaulted system the residual current for the protected line is zero, 3I0L = 0, and the system neutral N is at ground potential, VNG = 0, (see Fig. 8(a)). Natural system asymmetry produces some neutral current and shifts the system neutral from the ideal ground potential of VNG = 0.

A

C B

C B

N

A, G

VAN

VBNVCN

VAN

N, G

VBN

VCN

(a) (b) Fig. 8. Voltage Phasor Diagrams for the System of Fig. 7: (a) Unfaulted

System, (b) Faulted System (Solid A-Phase Fault, RF=0)

8

For a solid A-phase-to-ground fault, RF = 0, in the ideal lossless system, the faulted phase and ground potential are equal (see Fig. 8(b)). The phase-to-ground voltage of the two remaining unfaulted phases equals the phase-to-phase voltage (VBG = VBA, VCG = VCA) and the neutral-to-ground voltage equals the negative of the source phase-to-neutral voltage corresponding to the faulted phase (VNG = �–VAN).

B. Symmetrical Component Analysis The phase-domain analysis provides an exact

representation of the ungrounded system, which is valid even for asymmetrical systems. However, ground fault detection methods are typically based on zero-sequence quantities. It is then also important to outline a symmetrical-component-domain analysis of ungrounded systems operating in steady-state. If we consider that a ground directional relay relying on phase quantities would be supplied by high ratio phase current transformers (CTs), we immediately see that the need to size the phase CT ratio to sustain full load current automatically makes such a design less sensitive than a design that can use a lower ratio core-flux summing CT.

The zero-sequence impedance of an ungrounded system has a very high magnitude. This high value permits us to ignore the positive- and negative-sequence impedances without significant loss of accuracy when evaluating single line-to-ground faults. Fig. 9 shows an approximate zero-sequence representation of the forward ground fault in the system depicted in Fig. 7 (Switch SF closed). We assume that the system is symmetrical (CAL = CBL = CCL = CL, CAS = CBS = CCS = CS), and consider that the Thevenin voltage, the prefault voltage at the fault point, is equal to the nominal, phase-to-neutral, system voltage, Vnom.

Z0L

I0

V0 XC0LXC0S

R

Fig. 9. Zero-Sequence Network for the Forward Ground Fault in Fig. 7 System

Note that in Fig. 9 the relay measures V0 across XC0S and the current I0 through XC0S, where XC0S is the zero-sequence capacitive reactance of the remaining system in Fig. 7. The primary current I0 is flowing in at the CT polarity mark. At the relay location, V0 = �–I0�•(�–jXC0S) = jXC0SI0. If the relay calculates Z0 according to (14), with L0 = 90°, for this forward fault the resulting Z0 value is + XC0S.

Fig. 10 shows the zero-sequence network corresponding to a reverse fault in the system depicted in Fig. 7. The relay measures V0 across the series combination Z0L �– jXC0L, and the current I0 through the same series combination, where Z0L is the zero-sequence line impedance and XC0L is the zero-sequence capacitive reactance of the protected line. The primary current I0 flows out of the CT polarity mark for this

reverse fault. At the relay location, V0 = I0�•(Z0L �– jXC0L). Typically XC0L Z0L, so a good approximation is V0 = �–jXC0LI0. The resulting Z0 value for the reverse fault is �–XC0L.

Z0L

I0

V0 XC0LXC0S

R

Fig. 10. Zero-Sequence Network for the Reverse Ground Fault in Fig. 7 System

V. NEW GROUND FAULT DETECTION ELEMENT FOR UNGROUNDED SYSTEMS

A. Element Description Fig. 11(a) shows the phasor diagram for forward and

reverse faults in the system shown in Fig. 7. Fig. 11(b) shows a patent-pending directional element characteristic for ungrounded systems (32U). The function of a directional element is to determine forward and reverse conditions: i.e., differentiate +XC0S from �–XC0L. This new element does this with two thresholds set between these two impedance values. If the measured impedance is above the forward threshold (and all of the supervisory conditionals are met), the fault is declared forward.

3I0 (Forward fault)

3V0

Relay operatingcharacteristic

Reverse faultthreshold

Zero-Sequence Impedance Plane

R0

+XCOS

�–XCOL

Forward faultthreshold

3I0 (Reverse fault)

X0

(a) Zero-SequencePhasors

(b) New Impedance-Plane Directional ElementCharacteristic

Fig. 11. New Ungrounded System Ground Directional Element (32U) Characteristic

B. System Unbalance Affects Sensitivity CT inaccuracies could adversely affect directional element

sensitivity. Similarly, if the line-ground capacitances are not equal, the system produces standing or unfaulted zero-sequence quantities. Typically these quantities are small, but in a very large system the cumulative effect of unequal capacitances can generate appreciable zero-sequence voltage. To preserve fault resistance sensitivity, do not use a zero-sequence overvoltage element to supervise the directional element.

Let us review the effect of zero-sequence voltage supervision on ground relaying sensitivity. For this example, assume the end-of-line ground fault shown in Fig. 12 delivers 5 mA of secondary current to the relay on a system where the nominal secondary line-neutral voltage is 66.4 V.

9

3RF

V0C0 = C

VNOM52-1

52-2

52-3

R

SourceI0

(a) System single-Line Diagram (b) Zero-Sequence NetworkRepresentation

Fig. 12. Zero-Sequence Overvoltage Sensitivity Example

From Fig. 12(b):

CjIV 1�•00 (7)

0

0

�•VjIC (8)

Next, set the minimum V0 at 2 V for a starting place to calculate C in (8), given a minimum I0 of 5 mA. If we do this for a 60 Hz system, C = 6.63 F. Then, let us evaluate another similar system, but with Breaker 3 (52�–3) closed to increase C. If this new system only produced 5 mA secondary and the capacitance equaled 13.26 F, then |V0 | = 1 V secondary. Given a 3V0 threshold of 6 V secondary, the relay using supervisory zero-sequence overvoltage would not operate because of an incorrect supervisory setting.

Looking again at Fig. 12, we can calculate RF as:

0

0

�•3 IVVR NOM

F (9)

As shown in (9), raising the V0 threshold decreases the numerator and thereby decreases the available fault resistance coverage, or sensitivity, for a given minimum magnitude of I0. An alternative to 3V0 security supervision is to require the ratio of residual current to positive-sequence current to exceed a minimum scalar threshold value. The benefit of this supervision is that the minimum sensitivity of each feeder relay is not dependent upon the total system unbalance.

C. New Ungrounded Directional Element Performance To demonstrate the performance of this new ground

directional element for ungrounded systems we modeled a distribution power system using EMTP (Electromagnetic Transients Program). Fig. 13 shows the simulated system and the placement of an A-phase-to-ground fault with RF = 10 k . All of the seven feeders are overhead lines with differing lengths.

35 kV

110 kV

11 kV

L100

L400

L1000

L200

L300

L700

L1200

AG

Fig. 13. Ungrounded System Single-Line Diagram

Fig. 14 shows the zero-sequence voltage and current presented to the relay on Feeder L400. Fault inception is at cycle 18. From the upper plot (V0 = 1.5 V), notice that the high RF restricted the available 3V0 to only 4.5 V secondary.

10 20 30 40 50 60 700

0.02

0.04

Seco

ndar

y Am

ps

cycle

0

1

2

Seco

ndar

y Vo

lts

Zero-Sequence Volts: V0

Zero-Sequence Amperes: I0

Fig. 14. Line L400 Zero-Sequential Voltage and Current for End-of-Line

AG Fault, RF = 10 k

Fig. 15 shows the new directional element calculation results for this fault (the results prior to fault inception are erroneous because there is no prefault zero-sequence current). Note that the forward directional decision is very stable after the current magnitude rises above the minimum threshold of 5 mA (indicated by the thicker line in the plot).

0 10 20 30 40 50 60 70-500

0

500

Cycle

Cal

cula

ted

Z0 (O

hms) Fault Inception

ForwardRegion

ReverseRegion

Fig. 15. Line L400 Relay New Ungrounded System Directional Element

(32U) Calculates Forward Direction Independent of Fault Resistance

VI. COMPENSATED DISTRIBUTION SYSTEM ANALYSIS Operation analysis of compensated distribution networks is

important for identifying the available alternatives for ground fault detection in these systems. In this section we analyze the steady-state behavior of compensated systems both in the phase and in the symmetrical component domains. We also

10

summarize the harmonic content and the transient behavior of these systems.

A. Three-Phase Analysis Fig. 16 shows a simplified representation of a three-phase

compensated distribution system in which all the feeders appear in a single circuit. Again, for simplification in our steady-state analysis, we assume ideal sources operating at nominal frequency and no load, and disregard line series impedances, resistance and reactance.

We represent the Petersen coil in Fig. 16 as a parallel combination of an inductance (LN) and a resistance (RN). This configuration is the equivalent parallel circuit of the coil (the coil inductance and resistance are actually in series). This same combination may also represent the equivalent circuit for the case in which we connect a resistor in parallel with the Petersen coil or to an auxiliary winding of the coil. In Fig. 16, CA, CB, and CC represent the phase-to-ground capacitances of the network. Resistances RA, RB, and RC represent the phase-to-ground leakage resistances. Both the capacitance and leakage resistance values could be different for the different phases, especially for overhead lines, which means that the system may be asymmetrical. Typical values of the phase-to-ground leakage resistances are approximately ten to 20 times the phase-to-ground capacitive reactances [3]. We do not represent the phase-to-phase capacitances of the system in Fig. 16 because they do not contribute to the residual current and so are irrelevant to this analysis.

We may represent the occurrence of an A-phase fault in the system of Fig. 16 by closing Switch SF. RF represents the fault resistance. For solid faults RF = 0.

To discuss the basic principles of resonant grounding [19], we can disregard all active losses in the equivalent circuit of Fig. 16 (RN = RA = RB = RC ). Using the circuit of Fig. 16, we can calculate the current through the system grounding ING using the phase currents as:

0CGBGAGNG IIII (10)

)( CGBGAGNG IIII (11)

A

B

C

N

ING ICG IBG IAG

VAN IA

VBN IB

VCN IC

LN RN CC RC CB CARB

RF

RASF

IF

G Fig. 16. Three-Phase Simplified Representation of a Resonant Grounded

Network

In a symmetrical unfaulted system the neutral current is zero (ING = 0) and the system neutral N is at ground potential (VNG = 0), similar to that of the ungrounded system voltage

phasor diagram shown in Fig. 8(a). Natural system asymmetry produces some neutral current and shifts the system neutral from the ideal ground potential of VNG = 0. For a solid A-phase-to-ground fault (RF = 0) in the ideal lossless system, the faulted phase and ground potential are equal. The phase-to-ground voltage of the two remaining unfaulted phases equals the phase-to-phase voltage (VBG = VBA, VCG = VCA) and the neutral-ground voltage equals the negative of the source phase-to-neutral voltage corresponding to the faulted phase (VNG = �–VAN).

For the solid fault in Fig. 16, calculate the total fault current, IF:

)( CGBGNGAGF IIIII (12)

Note that ING is 180 degrees out-of-phase with respect to (IBG + ICG) in the ideal lossless system. By selecting the appropriate Petersen coil inductance, we can theoretically reduce the fault current to zero. This is the tuning, or compensation, condition, in which the system is in parallel resonance:

CGBGNG III (13)

For an off-tuned system, in other words a system not 100 percent tuned, the two possible operating conditions are:

CGBGNG III (overcompensated system) (14)

CGBGNG III (undercompensated system) (15)

In a system with losses, the 100 percent tuning condition does not result in a zero-ampere fault condition. For these systems the coil and system active losses, and the amount of RF, determine the ground fault current magnitude.

If RF 0, the current through the fault is only a portion of the faulted phase current (i.e., IF IAG for an A-phase fault). We can represent the fault inception as a change in the faulted phase admittance [19][20]. In this case the magnitude of the neutral voltage shift as a result of the fault is less than the source phase-to-neutral voltage (VNG < VAN).

The normalized neutral-to-ground voltage [12][19] for the system of Fig. 16 is given by:

CBAN

CBA

nom

NG

YYYYYaYaY

VV 2

(16)

Where Vnom is the nominal system voltage, a = 1 120 , a2 = 1 �–120 , and

AA

AAA CjR

jBGY 1

(A-phase-to-ground admittance) (17)

NNNNN LjR

jBGY 11

(Neutral admittance) (18)

11

RLRRN

222

(19)

LLRLN 2

222

(20)

where R and L are the Petersen coil resistance and inductance, respectively.

For the faulted system (Switch SF of Fig. 16 closed), all parameters remain unchanged except the A-phase admittance (YA). We represent YA for the faulted system as:

AAF

A CjRR

Y 11 (21)

As a result, the fault changes both the magnitude and phase of the neutral-ground voltage VNG. However, the phase changes are not a clear indicator of the fault, and the magnitude changes could be very small or even negative for high-resistance faults. As a result, faults can enhance system balance instead of perturbing it. This represents a limit to the sensitivity of the fault detection based on the magnitude of the neutral-to-ground voltage. The variation of this voltage, incremental neutral-to-ground voltage, is a better indicator of the fault. The value of this voltage is free from the prefault normal system unbalance; however, the voltage could be affected by system switching operations such as connection or disconnection of distribution lines, by tap-changing operations, or by resistor insertion in the Petersen coil.

B. Symmetrical Component Analysis Ground fault detection methods are typically based on

zero-sequence quantities. So, it is also important to outline a symmetrical-component-domain analysis of compensated systems operating in steady-state.

The zero-sequence impedance of a compensated system has a very high magnitude. This high value permits us to ignore the positive- and negative-sequence impedances without significant loss of accuracy when evaluating single line-to-ground faults. Thus, we represent the ground fault by connecting an equivalent Thevenin source in series with a resistance at the point of fault in the zero-sequence network. Fig. 17 shows an approximate zero-sequence representation of a ground fault in the system depicted in Fig. 16. We assume that the system is symmetrical (CA= CB = CC = C), disregard the leakage resistances (RA = RB = RC ), and consider that the Thevenin voltage, prefault voltage at the fault point, is equal to the nominal, phase-to-neutral system voltage, Vnom.

V0Vnom

3LN 3RN C0=C

3RF

+

Fig. 17. Zero-Sequence Representation of a Ground Fault in Fig. 16

System

From Fig. 17 we obtain:

)3

1(31

10

NF

N

Fnom

LCRj

RRV

V (22)

For ground faults the system zero-sequence voltage equals the neutral-to-ground voltage [10]. Then, (22) is the zero-sequence version of (16) for a symmetrical system if we disregard leakage resistances.

Fig. 18(a) shows a single-line diagram of a compensated radial distribution system in which the relay location defines the protected line. All of the other distribution lines are lumped in an equivalent impedance representing the remainder of the distribution system.

Fig. 18(b) is an approximate zero-sequence representation of the system. C0L and R0L are the protected line zero-sequence capacitance and leakage resistance, respectively. C0S and R0S are the corresponding values for the remainder of the system. In this case C0 = C0L + C0S and 1 / R0 = 1 / R0L + 1 / R0S, where C0 and R0 are the system zero-sequence capacitance and leakage resistance, respectively.

Again, we represent the fault by connecting an equivalent Thevenin source in series with a resistance. In Fig. 18(b) we close Switch SF to represent a ground fault on the protected line (forward fault direction from the perspective of the relay) and close Switch SR to represent a ground fault elsewhere in the system (reverse direction fault).

We may also approximately represent system unbalance by connecting an equivalent Thevenin source in series with an impedance in the zero-sequence network. In Fig. 19(a), ZLU represents the unbalance of the protected line and ZSU is the unbalance impedance corresponding to the rest of the system. At the relay side the unbalance includes the primary system neutral asymmetry and the additional unbalance introduced by the current transformers.

In the typical residual connection of the relay to three phase current transformers (CTs) the measurable zero-sequence current is corrupted by CT errors; the CT phase-angle error is particularly influential in this case. The Holmgreen connection of the CTs is a residual connection of the relay to phase CTs that are specially matched to reduce the residual current measurement error.

12

(b)

RelayProtected line

Remaining system

(a)

3LN 3RN C0L 3RFR0L

SF

C0S 3RFR0S

SR

RelayI0

V0

+

Fig. 18. Compensated Distribution System: (a) Single-Line Diagram,

(b) System Zero-Sequence Network

By far the best solution is the flux-summation or window-type CT, because it transforms the zero-sequence current directly and therefore does not include the cumulative error of the phase CTs. In addition, the CT ratio may be as low as 10:1, thus providing a significant increase in secondary current delivered to the protective relay. Hence, increasing the zero-sequence current means more sensitivity for ground faults.

It is possible to determine the unbalance impedances ZLU and ZSU or their corresponding admittances YLU and YSU. The admittance method for sensitive ground fault detection in compensated distribution circuits [4] requires the protective instruments to calculate these admittance values for the prefault system for use as a reference. The same calculation is also the reference for a residual current compensation method in resonant systems [12]. To calculate the unbalance admittances, you must have system information for two different tuning conditions. Typically, this means the protective instrument must control the Petersen coil. An obvious restriction to this requirement is that the protective instrument must be located within the substation, unless we include fast and secure external communication in the protection scheme.

A new approach to this problem is to exclude the effect of the unbalance from the zero-sequence network by using incremental, or delta, quantities [21]. The advantages of this method are that it does not require Petersen coil control and that it uses information from the protected line. Thus, the new type of protective relay can be located elsewhere in the system without the need for dedicated communications channels. The incremental zero-sequence voltage ( V0) and current ( I0) are:

FAULTPREFAULT VVV ,, 000 (23)

FAULTPREFAULT III ,, 000 (24)

(a)

3LN 3RN C0L R0L

SF

I0

V0

+ Relay

3RFZLU C0S R0S

SR

3RFZSU

(b)

3LN 3RN C0L 3RFR0L

SF

C0S 3RFR0S

SR

RelayI0

V0

+

Fig. 19. Zero-Sequence Network for the System of Fig. 18(a) Considering

System Unbalance: (a) Total Quantities, (b) Incremental Quantities

Fig. 19(b) depicts the zero-sequence system network for incremental quantities. Note that it only shows those components necessary for calculating fault quantities. The results are independent from the natural system asymmetry and from CT errors. The latter feature permits a relay with this new means of ground fault detection to be used with conventional CTs. While the incremental quantities do reduce the zero-sequence current measurement error, this method cannot overcome the problem of extremely low zero-sequence current being presented to the relay because of the large ratios used for the phase CTs.

C. Ground Faults in Compensated Distribution Networks The self-extinction mechanism for arcing ground faults in

cable lines is not as effective as in overhead lines because cable insulation breakdown is generally definitive. Even if the arc self-extinguishes during the current zero-crossings, the damaged insulation fails again when the instantaneous phase-to-ground voltage reaches a level higher than the damaged dielectric withstand level. The result is a restriking fault. Restriking ground faults produce repetitive overvoltages in the unfaulted phases that can eventually lead to these phases faulting to create a cross-country fault. Given the low probability of self-clearing, many utilities trip, without reclosing, cable distribution lines when the protection detects a ground fault. Another solution is to introduce residual current compensation in the Petersen coil control system [12]. This system reduces the ground fault current to zero, thus reducing the restriking mechanism. It is then possible to operate the system with a faulted cable.

In summary, the basic types of ground faults in compensated distribution networks are self-clearing faults, restriking faults, and sustained faults. Restriking faults are self-clearing faults that become repetitive as a result of permanent insulation breakdown. Sustained faults include all permanent faults and some temporary faults not cleared by the arc-extinction mechanism of resonant grounding.

13

VII. GROUND FAULT DETECTION METHODS FOR COMPENSATED DISTRIBUTION NETWORKS

Ground fault detection methods for compensated distribution networks may be classified according to the components of the relay input signals that they use to detect the fault. For this discussion, we grouped the methods into the following four classifications:

1) Fundamental frequency 2) Harmonic based 3) Transient-components based 4) Other

Methods included in the first two groups use information corresponding to the steady-state of the faulted distribution network; some of these methods also require steady-state prefault information. Group 3 methods use information on the transient process generated by the fault. Group 4 includes methods that basically use steady-state information, but require control actions on the Petersen coil, either current injection or temporary detuning.

A. Voltage Detection Ground faults in compensated systems reduce the line-to-

ground voltage of the faulted phase and shift the system neutral, increasing the system zero-sequence voltage in most cases. Both the zero-sequence voltage and the phase-to-ground voltages have been used as indicators of ground faults [3][9]. However, for high-resistance faults the voltage change could be very small. Some researchers have proposed using the incremental zero-sequence voltage in order to increase the detection sensitivity [19].

Another problem is that the zero-sequence voltage has almost the same value in the entire distribution network. The voltage drops caused by the zero-sequence currents circulating through the zero-sequence impedances of the lines and transformers are very small as compared to the voltage drops across the phase-to-ground system impedances. In other words, voltage-based ground fault detection methods are not selective. They detect the ground fault, but do not determine the faulted element. It is then necessary to manually disconnect and reclose every feeder of the substation to locate the fault. These short service interruptions represent a power quality problem. For this reason, voltage detection is typically used as a starting function of selective ground fault detection methods in compensated distribution networks.

B. Wattmetric Method Selective ground-fault detection methods require current

information. Zero-sequence directional relays are a classical solution [2][3][9]. For ungrounded networks the varmetric relay responds to the quadrature (imaginary) component of the zero-sequence current with respect to the zero-sequence voltage [2][9]. For compensated networks the wattmetric relay uses the in-phase (real) component of the zero-sequence current [2][3].

We may analyze these fault-detection methods, referring to the system depicted in Fig. 18(a). Fig. 18(b) shows the approximate zero-sequence representation for the balanced version of the system.

For a forward fault, such as Switch SF closed and Switch SR open in Fig. 18(b), the relay zero-sequence current, I0, is:

NS

NS LCj

RRVI

31

311

00

00

(Forward fault) (25)

For a reverse fault, such as Switch SF open and Switch SR closed in Fig. 18(b), the relay current is:

LL

CjR

VI 00

001 (Reverse fault) (26)

In ungrounded systems (RN = LN ), (26) does not change. For ungrounded systems (25) takes the form:

SS

CjR

VI 00

001 (Forward fault) (27)

In compensated networks the direction of the quadrature component of I0 may change for forward faults depending on the system tuning conditions (see (25)). This direction depends on the values of the equivalent zero-sequence capacitance of the unfaulted lines, C0S, and the zero-sequence parallel-equivalent inductance of the Petersen coil, 3LN. Fig. 20 shows the phasor diagram for ground faults in compensated networks. For reverse faults I0 exhibits the same behavior as in ungrounded systems. For forward faults the angular position of I0 with respect to V0 may vary widely. The quadrature component of I0 is negative, as in ungrounded systems in an undercompensated (undertuned) network, and positive in an overcompensated network. On the other hand, the sign of the in-phase, active, component of I0 is always positive for reverse faults and negative for forward faults. We may use a wattmetric directional relay having the following output quantity (* = complex conjugate):

00000 cos*�•Re IVIVW (28)

We may compare W with positive and negative thresholds (+ and �– ). Then W < �– indicates a forward fault and W > indicates a reverse fault. The wattmetric relay operating characteristic is also represented in Fig. 20.

Fig. 21 depicts a simplified logic diagram for a wattmetric element. The active component of I0 is very low during ground faults, so the relay should be very sensitive ( should be very small). To avoid relay misoperations during normal system conditions, add a starting function responding to the magnitude of V0. Then wattmetric relay sensitivity is determined by the V0 element sensitivity. The threshold V0 value should be greater than the value of V0 for normal system unbalances. A typical setting is 20 percent of the nominal system voltage.

The wattmetric method has been in use for many years in compensated systems. It is simple, secure, and dependable for low-resistance ground faults. However, the requirement of V0 detection limits the sensitivity of the wattmetric method for high-resistance faults. Another drawback is that the method is

14

very sensitive to CT accuracy problems. In the typical residual connection of the relay to three CTs, CT angle errors may produce a change in the sign of the element output, W. Careful calibration of the CTs is a possible solution, but flux-summation CTs are strongly recommended for wattmetric relays.

Some utilities use zero-sequence voltage relays to detect ground faults in compensated systems and automatically connect a resistor in parallel with the Petersen coil or to an auxiliary winding of the coil. For high-resistance faults, connecting the resistor reduces the zero-sequence voltage without necessarily increasing the zero-sequence fault current; the output of the wattmetric element actually decreases. The other limitation of this method is reduced sensitivity resulting from the use of the zero-sequence voltage for fault detection.

We may apply the wattmetric method for ground fault detection in all types of distribution systems having low ground-fault-current values. This includes isolated-neutral, high-impedance-grounded and compensated systems. However, for isolated neutral systems the varmetric method provides higher sensitivity than the wattmetric method, because the quadrature component of the zero-sequence current is always greater than the in-phase component for ground faults in these systems.

I0 forwardfault region

I0Overcompensated

system

I0Undercompensatedsystem

I0 (Reverse fault)

v0

Forward fault Reverse fault

Relay operatingcharacteristic

Fig. 20. Typical Phasor Diagram for Ground Faults and Wattmetric Relay

(32W) Operating Characteristics in a Compensated Network

_+

_+

_+

W = Re[V0 · I0*]I0

V0

ABS

THR

Forward fault

Reverse fault

Fig. 21. Simplified Logic Diagram of a Wattmetric Relay Element (32W)

VIII. NEW FAULT DETECTION METHODS FOR COMPENSATED DISTRIBUTION NETWORKS

Fundamental-frequency components of current and voltage provide the most reliable and significant information for detecting high-resistance faults in compensated distribution

networks. The harmonic content of ground-fault current is small and variable, especially for high-resistance faults.

The transient components of current and voltage are severely damped for high-resistance faults. Methods that need some type of control on the Petersen coil or current injection provide high sensitivity, but require costly equipment.

All known fundamental-frequency methods that provide high sensitivity require information on all feeders and/or some sort of control on the Petersen coil or current injection. For example, the admittance method requires information from all feeders. The admittance method also needs control on the Petersen coil or current injection. The wattmetric method is a good solution for low-resistance faults, but has sensitivity limitations in detecting high-resistance faults.

We may provide high-resistance coverage by measuring the zero-sequence conductance or the zero-sequence resistance in the protected feeder. The new methods we introduce in this paper use only information from the feeder and do not require control on the Petersen coil or current injection. The methods are suitable for stand-alone, high-sensitivity devices that can be located at any point of a distribution network. A typical application is in the control unit of an automatic recloser.

A. Conductance Method We use the distribution network depicted in Fig. 18(a) to

outline the conductance method. First, we disregard the system unbalance. Fig. 18(b) shows the zero-sequence network for the balanced system. Equations (25) and (26) describe the zero-sequence relay current, I0, for forward and reverse faults. We may then also use Equations (25) and (26) to calculate the apparent zero-sequence admittance, Y0, measured by the relay for forward and reverse faults:

NS

NS LCj

RRVI

Y3

13

110

00

00


LL

CjRV

IY 000

00


Taking the real part from Equations (29) and (30), we can determine the conductance, G0, measured by the relay for both faults:

NSNS

GGRRV

IG 0000

00 3

11Re


LL

GRV

IG 000

00

1Re (Reverse fault) (32)

where G0L=1/R0L is the zero-sequence leakage conductance of the protected feeder, G0S=1/R0S is the equivalent zero-sequence leakage resistance of the remaining feeders, and G0N=1/3RN is the zero-sequence conductance corresponding to the parallel equivalent of the Petersen coil.

15

We may use a conductance element responding to the real part of the I0/V0 ratio to detect ground faults. The element compares the measured conductance, G0, with positive, , and negative, �– , thresholds. Then G0 < �– indicates a forward fault, and G0 > indicates a reverse fault. For forward faults (see (31)) the conductance element measures the equivalent conductance behind the relay. This includes the conductance of the remaining feeders and that of the parallel equivalent of the Petersen coil. For reverse faults (see (32)) the conductance element measures the conductance of the protected feeder.

The conductance method is inherently directional. It responds to the sign of the real part of the measured admittance. This is an advantage compared to the admittance method, which responds to the magnitude of the admittance and does not use valuable phase information.

The conductance method responds to the current/voltage ratio. Its output is relatively independent of the magnitudes of the zero-sequence current and voltage. This is an advantage compared to the wattmetric method, which fails to detect high-resistance faults caused by low values of both V0 and I0. Recall that the wattmetric element is a product-type element and the conductance element is a ratio-type element.

B. Incremental Conductance Method The traditional conductance method described above works

well in balanced systems and for low CT errors, such as using flux-summation CTs, for example. System and CT unbalances introduce errors in the measured conductance and thereby limit sensitivity. A solution to this problem is the incremental conductance method. We use the incremental zero-sequence current (Equation (24), and voltage (Equation (23), to calculate an incremental zero-sequence conductance, G0:

0

00 Re

VIG (33)

Fig. 19(b) shows the zero-sequence network of the Fig. 18(a) system with incremental quantities. Note that this circuit is equal to that of Fig. 18(b) for a balanced system. Then, the measured values of G0 coincide with those given for G0 by Equations (31) and (32):

NSNS

GGRR

G 000

0 311 (Forward fault) (34)

LL

GR

G 00


The incremental conductance element (32C) compares the measured incremental conductance, the real part of the incremental current/incremental voltage ratio, with positive and negative thresholds to discriminate forward faults from reverse faults. It is a directional, ratio-type method that is not affected by system unbalance and that can be used with conventional CTs. Fig. 22 depicts the measured incremental conductances for forward and reverse faults and the incremental conductance relay operating characteristic (two parallel straight lines).

Reverse faultForward fault

G0G0L-(G0S+G0N)

Relay operating characteristic

B0

Fig. 22. Incremental Conductance Element (32C) Operating Characteristic

Fig. 23 depicts the simplified logic diagram for an incremental conductance element. It retains the advantages of the incremental conductance methods and does not require incremental zero-sequence voltage and current as input information.

0V

0I

+

-Forward Fault

+-

Reverse Fault

0

00 V

IReG

-

Fig. 23. Simplified Logic Diagram of an Incremental Conductance Element

(32C)

C. Compensated System Directional Element Performance To demonstrate the performance of this new ground

directional element, we modeled a distribution power system using EMTP (Electromagnetic Transients Program) and played the results through a relay simulation program and a test system connected to the relay. The simulated system includes a 138 kV three-phase balanced source, a delta-wye 138-13 kV step-down transformer, a Petersen coil, and three feeders. All three overhead line feeders have the same tower structure (see Fig. 24) but different line lengths as shown in Fig. 25.

At 100 percent tuning, the total zero-sequence capacitive current from all feeders equals the inductive current provided by the Petersen coil in the normal unfaulted system condition. The Petersen coil has an inductance of 5.15 Henries and an X/R = 30.

Because of the high zero-sequence impedance presented by the Petersen coil resonant system, the fault voltage and current profiles are almost identical for both close-in and remote ground faults.

Each of three feeders has a leakage conductance of five micro-Siemens primary. The equivalent shunt conductance of the Petersen coil is 5.72 micro-Siemens primary. From our previous analysis, for a ground fault anywhere on this distribution system, the relay at the faulted feeder calculates an incremental conductance that equals the negative sum of the conductances of all remaining feeders plus that of the Petersen coil. This negative sum of conductance is -15.72 micro-Siemens primary. Relays that see the fault as a

16

reverse fault calculate an incremental conductance that equals its own feeder leakage conductance of five micro-Siemens primary.

0.5 m 0.5 m

0.9 m8.32 m

A

B

C

Fig. 24. Tower Configuration of Simulated System

52-1

138 kV

20 km

40 km

13 kV

40 km

Feeder 1

Feeder 2

Feeder 3

X/R=30L=5.15H @ 100% tuning

Reverse52-2

52-2

ForwardR

Fig. 25. Simulated Power System Single-Line

The first example simulates an A-phase-to-ground fault on Feeder 3. The system is initially at the 100 percent tuning condition. The fault resistance is 80 k primary. The fault is a self-clearing fault that lasts one second. Fig. 26 shows the zero-sequence voltage, 3V0, and current, IR, of Feeder 3. The A-phase-to-ground fault occurs at cycle 60 and self-extinguishes at cycle 120. Notice that in Fig. 26, the standing zero-sequence voltage, 3V0, is 26.1 V secondary. Recall that the traditional wattmetric directional element requires that |3V0| exceeds 0.2 Vnom, or 13.9 V in this application, to be enabled. This standing voltage is already 62 percent of the enabling threshold, 59RES. Note, too, that the |3V0| during the A-phase fault is even lower than the system standing zero-sequence voltage. In this situation, the traditional wattmetric method is not sensitive enough to detect this 80 k fault.

Fig. 27 shows the wattmetric and conductance directional calculations and the forward (FWD) and reverse (REV) threshold settings for a relay installed on Feeder 3. Note that the incremental conductance directional element correctly detects the high-impedance fault: In other words, it produced a negative result that overcomes the forward fault threshold for the fault. Also, note that the incremental conductance value during the fault is about 157 micro-Siemens secondary, a correct value given a PTR to CTR ratio of ten. The wattmetric element calculated a zero-sequence active power value that indicated a trend for a forward fault (a negative value). However, this power value is far short from overcoming the forward fault detection threshold. Remember that the zero-sequence voltage magnitude during the fault is even lower than that of the system standing zero-sequence voltage, therefore the wattmetric element is not enabled for this fault.

0 20 40 60 80 100 120 140 160 1800

20

40

60

80

sec.

vol

t

0 20 40 60 80 100 120 140 160 1800

0.005

0.01

0.015

0.02

sec.

am

p

cycle

59RES

3V0

IR

Fig. 26. Zero-Sequence V and I Plots for an 80 k AG Fault on Feeder 3

0 20 40 60 80 100 120 140 160 180-0.6

-0.4

-0.2

0

0.2

0.4

sec.

wat

t

0 20 40 60 80 100 120 140 160 180-4

-3

-2

-1

0

1x 10

-4

sec.

sie

men

s

cycle

REV

FWD

Re[3V0*IR]

REV

FWD

Delta G

Fig. 27. Feeder 3 Wattmetric and Conductance Element Calculations for

Case 1

The second case simulates a B-phase-to-ground fault on Feeder 1, which is a reverse fault for the relay at Feeder 3. The B-phase-to-ground fault occurs at cycle 60 and self-extinguishes at cycle 120. The fault resistance is 30 k primary. The system is 20 percent over-compensated initially. Fig. 28 shows the zero-sequence voltage and current of Feeder 3. From the figure, notice that the standing zero-sequence voltage, 3V0, is 15.3 V secondary, a much lower value than that of Case 1 with a 100 percent tuned system. The zero-sequence voltage magnitude, |3V0|, during the fault is much higher than the standing voltage, but it is still lower than the enable threshold of the wattmetric element.

Fig. 29 shows the wattmetric and conductance directional calculations and thresholds for the relay installed on Feeder 3. Note that the incremental conductance directional element calculates a correct positive conductance result, about 50 micro-Siemens secondary, which is above its reverse fault detection threshold. The relay therefore successfully detected the high impedance reverse fault. The wattmetric element is not enabled by the zero-sequence overvoltage element; its zero-sequence power value also fell below the reverse fault

17

threshold, and therefore the element failed to detect the reverse fault.

0 20 40 60 80 100 120 140 160 1800

20

40

60

80

sec.

vol

t

0 20 40 60 80 100 120 140 160 1800

0.01

0.02

0.03

0.04

0.05

sec.

am

p

cycle

59RES

3V0

IR

Fig. 28. Zero-Sequence V and I Plots for a 30 k BG Fault on Feeder 1

0 20 40 60 80 100 120 140 160 180-0.6

-0.4

-0.2

0

0.2

0.4

sec.

wat

t

0 20 40 60 80 100 120 140 160 180-4

-3

-2

-1

0

1x 10

-4

sec.

sie

men

s

cycle

Re[3V0*IR']

REV

FWD

REV

FWD Delta G

Fig. 29. Feeder 3 Wattmetric and Conductance Element Calculations for

Case 2

IX. IMPROVED BROKEN DELTA VOLTAGE TRANSFORMER CONNECTION

Modern relays accept many different configurations for voltage transformers (VTs): three-phase four-wire, open-delta, and broken delta. The latter VT connection is suitable for ground directional element applications and is very common in most of the existing ungrounded and resonant-grounded systems. However, this standard broken-delta connection can present a unique set of problems outside of ferroresonance. Fig. 30 shows a classical broken-delta VT connection.

While the broken-delta VT connection does provide zero-sequence voltage for measurement during ground faults, the nominal output voltage for a bolted ground fault on the ungrounded or Petersen system can be over 360 VAC if the nominal voltage for each phase transformer of the broken delta is 120 VLN. Such a high voltage is higher than most relay input ratings. Because the ideal prefault zero-sequence voltage magnitude is zero, it is difficult to detect a blown VT fuse. Because many ground directional elements require a minimum

zero-sequence voltage, a blown VT fuse could defeat ground protection.

Many installations require maintaining the broken-delta VT connection for existing control devices. Thus, the challenge is to extract three-phase four-wire voltage signals from an existing broken-delta connection. It is also very desirable to have a relaying system applicable to a standard three-phase, four-wire VT application. Fig. 31 shows a simple, patent-pending solution. This solution requires connecting the relay input transformers as shown. With this connection, the relay can then measure each phase voltage and calculate the necessary 3V0 for the zero-sequence directional element described earlier.

VA

VB

VC

ABC

VTs SecondariesConnected inBroken Delta

Power System

3V0 = VA + VB + VC

Fig. 30. Traditional Broken Delta VT Connection Diagram

VA

VB

VC

VN

VaRELAY

ABC

VTs SecondariesConnected inBroken Delta

Relay Internal VTPrimaries ConnectedBroken Delta

VbRELAY

VcRELAY

Power System

Fig. 31. Single-Line Diagram and New Broken-Delta VT Connection

Diagram

The benefits of this new VT connection approach are: No relay input transformer has to be rated for 360 VAC. The relay system can now check for blown potential fuses.

In a relay using the traditional broken-delta connection on a system with little or no unbalance, the 3V0 measurable before and after a blown secondary fuse is the same, zero volts.

The relay can measure each individual phase voltage and calculate the necessary sequence components. This allows the relay to use the same VTs for phase and ground directional control elements.

It does not require disturbing existing wiring for devices using the broken-delta voltage output. Simply add wires from

18

the B- and C-phase polarity marks of the VT secondaries to the respective inputs on the relay.

It allows dual phase directionality from differing VTs: Main 1 could use this new connection from the broken-delta system while Main 2 could use the existing open-delta VTs for polarizing.

X. SUMMARY OF GROUND FAULT PROTECTION METHODS TABLE II summarizes available fundamental-frequency

methods for stand-alone relays in distribution systems. For solidly- and low-impedance grounded systems we recommend using the best choice directional element. This element automatically selects the optimal directional element, 32I, 32Q, or 32V, for each particular system configuration and ground fault condition. For isolated neutral systems we recommend using the zero-sequence impedance element, 32U. Finally, for high-resistance-grounded systems, high-reactance grounded systems and compensated systems we recommend using a combination of the wattmetric, 32W, and the incremental conductance, 32C, elements. The 32W element provides reliable detection of low-resistance faults, up to about 10 k . The 32C element adds the sensitivity required to detect very high-resistance faults.

TABLE II SUMMARY OF GROUND FAULT PROTECTION METHODS FOR STAND-ALONE

RELAYS IN MEDIUM-VOLTAGE DISTRIBUTION SYSTEMS

Grounding Method Available Fault Detection Method

Recommended Method

Solidly and Low-Impedance Grounding

32I 32Q 32V

Best Choice Directional

Isolated Neutral 32U 32VAR 32W 32C

32U

High-Resistance Grounding 32W 32C 32VAR

Combination of 32W and 32C

High- Reactance Grounding 32W 32C 32VAR


Resonant Grounding 32W 32C


Legend: 32I: Current-Polarized Directional Element 32Q: Negative-Sequence Voltage-Polarized Directional Element 32V: Zero-Sequence Voltage-Polarized Directional Element 32U: Zero-Sequence Impedance Element for Ungrounded Systems 32VAR: Varmetric Element 32W: Wattmetric Element 32C: Incremental Conductance Element

TABLE III depicts the main sensitivity limiting factors of the ground directional elements. Line asymmetry generates negative- and zero-sequence currents for three-phase faults [14]. We need to use a2 and/or a0 settings to avoid misoperation of 32I, 32Q, 32V and 32U elements. CT and VT magnitude and angle errors produce standing negative and zero-sequence voltages and currents [15]. The negative- and zero-sequence quantities produced by ground faults should overcome these standing voltages and currents. Total system unbalance is the main sensitivity limiting factor in wattmetric,

32W, and varmetric elements, 32VAR. The starting function responding to V0 in these elements needs to be set above the standing V0 value because of system unbalance. The 32C element is highly sensitive. The inherent sensitivity of a particular relay design determines the 32C element sensitivity.

TABLE III also shows the basic CT and VT requirements for ground directional elements in distribution systems. We strongly recommend minimizing phase CT saturation problems to avoid false negative- and zero-sequence currents for three-phase faults [15] in the application of 32I, 32Q, and 32V elements. We also recommend using flux-summing CTs to provide zero-sequence current information to 32U, 32W, and 32C elements. We recommend using Class 2 VTs or better in all ground directional element applications [15].

TABLE III FACTORS AFFECTING SENSITIVITY OF GROUND DIRECTIONAL ELEMENTS FOR

DISTRIBUTION SYSTEMS

Element Sensitivity Limiting Factors

CT Requirements VT Requirements

32I 32Q 32V

-Line asymmetry -CT and VT inaccuracies -CT Saturation (32V only)

-Select phase CTs to minimize saturation

-Class 2 VTs or better recommended

32U

-Line asymmetry -CT and VT inaccuracies

-Flux-summing CTs recommended


32W 32VAR

-Total system unbalance -Flux-summing CTs recommended


32C -Relay sensitivity limits

-Flux-summing CTs recommended


XI. CONCLUSIONS Ground faults in solidly- and low-impedance-grounded

systems may produce high current levels that require line tripping. Use directional elements to provide ground-fault protection in these systems. In multigrounded four-wire distribution systems, the relay measures the ground fault current combined with the unbalance current. As a result, detecting high-resistance faults in these systems is very difficult.

Selecting a fixed ground directional element for all system conditions may sacrifice protection reliability and sensitivity in solidly and low-impedance grounded systems. The ground directional relay described in this paper includes two zero-sequence elements and a negative-sequence element. The relay selects automatically the best directional element to use for each particular fault and system condition (patented).

Ungrounded systems are connected to ground through the line-to-ground capacitances. Single line-to-ground faults shift the system neutral but leave the phase-to-phase voltage triangle intact. Self-extinction of ground faults in overhead-ungrounded lines is only possible for low values of ground fault current.

Zero-sequence or three-phase voltage relays can detect ground faults in ungrounded systems. However, this method is not selective. A sensitive, directional ground varmetric element is the classic solution to ground fault detection in ungrounded systems.

19

A new ground directional element for ungrounded systems (patent pending) measures the zero-sequence reactance and compares its value with two settable thresholds. For a forward ground fault, the element measures the zero-sequence capacitive reactance of the equivalent system behind the relay. For reverse faults, the new element measures the series combination of the protected line zero-sequence series impedance and the line capacitive reactance. The new directional element includes a security supervision logic that requires the ratio of residual current to positive-sequence current to exceed a minimum scalar threshold value. The benefit of this supervision as compared to the traditional 3V0 security supervision is that the minimum sensitivity of each feeder relay is not dependent upon the total system unbalance.

Compensated systems are grounded through a variable impedance reactor, which compensates the system phase-to-ground capacitance. The system remains operational during ground faults. Resonant grounding provides self-extinction of the arc in overhead lines for about 80 percent of temporary ground faults. Then, ground faults that clear without breaker tripping represent more than 50 percent of all faults in overhead lines.

The wattmetric method is the most widely used solution to ground fault detection in compensated systems. The relay element responds to the in-phase (real) component of the zero-sequence current with respect to the zero-sequence voltage. The requirement of V0 detection limits the sensitivity of the wattmetric method for high-resistance faults. All known fundamental-frequency methods that provide high sensitivity, for example, the admittance method, require information on all feeders and/or some sort of control on the Petersen coil or current injection.

We may provide high-resistance coverage by measuring the zero-sequence conductance in the protected feeder. This method works well in balanced systems and for low CT errors, such as using flux-summation CTs. An enhancement to this method is to calculate the incremental conductance as the real part of the ratio of the incremental zero-sequence current to the incremental zero-sequence voltage. The incremental conductance method is inherently directional, exhibits high fault resistance coverage, and is applicable in a stand-alone feeder relay.

We conducted extensive digital simulation testing of a new incremental conductance ground directional element for compensated systems using EMTP. The relay element performed well even for high-resistance faults. The new element detected faults with resistance values over 60 k .

XII. REFERENCES [1] M. Pühringer, Resonant Grounding as Approach to System Neutral

Grounding, Haefely Trench, February 1998. [2] AIEE Committee Report, �“Sensitive Ground Protection,�” AIEE

Transactions, Vol. 69, 1950, pp. 473�–476. [3] E. T. B. Gross, �“Sensitive Fault Protection for Transmission Lines and

Distribution Feeders,�” AIEE Transactions, Vol. 60, Nov. 1941, pp. 968�–972.

[4] G. Druml, �“Detecting High-Ohmic Earth Faults in Compensated Networks,�” Proceedings of the International Symposium NMT, 1995.

[5] Haefely Trench, EPSY-Earthfault Protection System: Summary of the Results of Earthfault Field Tests.

[6] Ya. S. Guelfand, Protection of Distribution Networks, Moscow: Energoatomizdat, 1987 (Russian).

[7] A. M. Fedoseev, Protective Relaying of Electrical Systems, Moscow: Energia, 1976 (Russian).

[8] D. Griffel, Y. Harmand, and J. Bergeal, �“New Neutral Earthing Technologies on MV Networks,�” Revue Generale D�’Electricite, No. 11, December 1994, pp. 35�–44 (French).

[9] L. F. Hunt and J. H. Vivian, �“Sensitive Ground Protection for Radial Distribution Feeders,�” AIEE Transactions, Vol. 59, February 1940, pp. 84�–90.

[10] J. L. Blackburn, Protective Relaying: Principles and Applications, Second Edition, New York: Marcel Dekker, Inc., 1998.

[11] B. Bridger, �“High-resistance grounding,�” IEEE Transactions on Industry Applications, Vol. IA-19, No. 1, January/February 1983, pp. 15�–21.

[12] K. M. Winter, �“Swedish Distribution Networks�—a New Method for Earthfault Protection in Cable and Overhead Systems,�” Proceedings of the Fifth International Conference on Developments in Power System Protection�—DPSP �’93, United Kingdom, 1993, IEE Conference Publication No. 368, pp. 268�–270.

[13] A. Haman, �“Petersen Coils as Applied to 41.6 kV Systems,�” Proceedings of the 34th Annual Minnesota Power Systems Conference, St. Paul, Minnesota, October 6�–8, 1998.

[14] J. Roberts and A. Guzman, �“Directional Element Design and Evaluation,�” Proceedings of the 21st Annual Western Protective Relay Conference, Spokane, WA, October 1994.

[15] J. Roberts, E.O. Schweitzer III, R. Arora, and E. Poggi, �“Limits to the Sensitivity of Ground Directional and Distance Protection,�” Proceedings of the 22nd Annual Western Protective Relay Conference, Spokane, WA, October 1995.

[16] L. Blackburn, �“Negative Sequence Relaying for Mutually Coupled Lines,�” Proceedings of the 1972 Conference for Protective Relay Engineers, Texas A&M University, College Station, TX, April 1972.

[17] E.O. Schweitzer III and J. Roberts, �“Distance Relay Element Design,�” Proceedings of the 19th Annual Western Protective Relay Conference, Spokane, WA, October 1992.

[18] A. Guzman, J. Roberts, and D. Hou, �“New Ground Directional Elements Operate Reliably for Changing System Conditions,�” Proceedings of the 23rd Annual Western Protective Relay Conference, Spokane, WA, October 15�–17, 1996.

[19] V. Leitloff, L. Pierrat, and R. Feuillet, �“Study of the Neutral-to-Ground Voltage in a Compensated Power System,�” European Transactions on Electrical Power Engineering, Vol. 4, No. 2, March/April 1994, pp. 145�–153.

[20] V. Leitloff, R. Feuillet, and L. Pierrat, �“Determination of the Phase-to-Ground Admittance in a Compensated MV Power Distribution System,�” Proceedings of the 28th Universities Power Engineering Conference�—UPEC �’93, Vol. 1, Stafford, United Kingdom, September 21�–23, 1993, pp. 73�–76.

[21] G. Benmouyal and J. Roberts, �“Superimposed Quantities: Their True Nature and Their Application in Relays,�” Proceedings of the 26th Annual Western Protective Relay Conference, Spokane, WA, October 26�–28, 1999.

20

XIII. BIOGRAPHIES

Jeff Roberts received his BSEE from Washington State University in 1985. He worked for Pacific Gas and Electric as a System Protection Engineer. In 1988, he joined Schweitzer Engineering Laboratories, Inc. as the first Application Engineer. Later he became Application Engineering Manager.

He now serves as the Research Engineering Manager. He has written many papers in the areas of distance element design, sensitivity of distance and directional elements, line differential and directional element design, protection scheme design and analysis of event report data.

Mr. Roberts holds over twenty patents granted or pending. He is also a senior member of IEEE.

Hector J. Altuve received his BSEE from Central University of Las Villas (UCLV), Cuba, in 1969 and his PhD from Kiev Polytechnic Institute, USSR, in 1981. He served as a professor in the School of Electrical Engineering at UCLV from 1969 to 1993.

From 1993 to 2001 he served as a professor in the PhD program of the Mechanical and Electrical Engineering School at Autonomous University of

Nuevo Leon, in Monterrey, Mexico. He was the 1999-2000 Schweitzer Visiting Professor at Washington State University. In 2001 he joined Schweitzer Engineering Laboratories, Inc. as a Research Engineer. Later he became the General Director for Mexico operations.

He is a Senior Member of IEEE, and a PES Distinguished Lecturer. He has authored and coauthored many technical papers and holds a patent.

Daqing Hou received BS and MS degrees in Electrical Engineering at the Northeast University, China, 1981 and 1984, respectively. He received his Ph.D. in Electrical and Computer Engineering at Washington State University in 1991.

Since 1990, he has been with Schweitzer Engineering Laboratories, Inc., Pullman, Washington, USA, where he is currently a Development Engineering Manager. His work includes system modeling, simulation and signal processing for power system digital protective relays. His research interests include multivariable linear systems, system identification, and signal processing.

He holds several patents and multiple patents pending. He has authored or co-authored several technical papers, and is a senior member of IEEE.

An Adaptive Protection Scheme for Power Distribution Systems

J. De La Ree* David C. Elizondo* Juancarlo Depablos* James Stoupis **

*Center for Power Engineering **ABB ETI The Bradley Department of Electrical Engineering Raleigh, North Carolina Virginia Tech, Blacksburg, VA 24061 Introduction This paper describes the application of microprocessor based relay with internet communication capabilities in distribution protection systems. The first part of the paper presents an analogue simulator which was built to asses the benefits that can be obtained when power systems are combined with communication and computer systems. The second part of the paper details one of the several possible applications of internet based peer to peer communication in distribution power systems. The traditional distribution protection system (recloser, sectionalizers) was configured to automatically isolate faulted circuits as well as to reenergize unfaulted loads after a certain number of reclosing operations. According to the results, adding peer-to-peer communication to a traditional distribution protection system significantly enhances its performance. They can be summarized as follow: Distribution Protection System: Background It is well understood that microprocessor based relays have opened a new era in protection systems. These new devices provide us with the possibilities of storage, data analysis and communications. Furthermore they enable us to develop new protection schemes capable of rendering decisions based upon, not only local signals, but remote inputs as well. This paper presents an analogue simulator which was built to asses the benefits that can be obtained when power systems are combined with communication and computer systems. The analogue simulator models a typical power distribution feeder. The elements included in the model are: the distribution line, local loads, recloser and sectionalizer with its microprocessor based relay, and communication equipment. The system conditions that may be modeled by the simulator are classified as normal and contingency. During normal conditions, load switches connect and disconnect system load at a particular location, and pilot lights

indicate the load status. During contingency state, three-phase faults as well as single-phase to ground faults may be applied by switches with appropriate connections. The faults may be permanent faults and transient faults. The faults can be applied in selected locations along the feeder and the recloser and sectionalizers are coordinated in order to properly isolate the faulted segment depending on the fault location and the fault characteristic. The coordination of the recloser and sectionalizers obtained by using the protection functions of the microprocessor based relays and applying the traditional protection practices reveled that there is an area of opportunity to improve the distribution feeder reliability. The use of the communication capabilities of the microprocessor based relays is a key factor. Enabling microprocessor based relays to share information among each other in a peer-to-peer communication scheme makes possible to significantly enhance the distribution protection system performance. In this particular application, microprocessor based relays communicate with each other using a standard TCP/IP protocol. Therefore, devices can access both local and remote data-bases providing a complete on-time picture of the system status. Also, it is possible to program every device to respond to the different contingency scenarios and fault isolation schemes. Furthermore, the system was programmed to automatically reenergize unfaulted loads by closing tie-breakers connecting to nearby feeders. The general results are shown at the end of the paper.

Distribution feeder analog simulator: Panel description

This section describes the analog simulator that models the distribution feeder. Figure 1 shows a schematic of the feeder as it would be deployed in the field, along with its single line diagram.

Figure 1: Distribution Feeder Model: Virtual Picture and Single Line

Diagram

The model of a distribution feeder The model of the distribution feeder was built in a 19 inch panel. A picture of the actual panel is included in the Appendix. The main components of the panel are: a source section, four line sections, one recloser section, and three sectionalizer sections. The source section is composed by a variable three-phase voltage source and a thermo-magnetic circuit breaker. The three-phase voltage source represents the low voltage winding of the substation transformer, and the thermo-magnetic circuit breaker represents the feeder circuit breaker. The line section models the distribution feeder lines. Each unit includes: three reactors (one per phase), one neutral reactor, one single-pole toggle switch, two three-pole toggle switches, and three resistors. Figure 2 shows the arrangement of the elements in the line section. The phase reactors (RL) represent the impedance of the distribution line. A neutral reactor (RN) was added in order to obtain the proper line to neutral fault currents. A three-pole toggle switch (TFF) applies a solid three-phase fault in the particular line section of the distribution feeder. In a similar fashion, a single pole toggle switch (SFF) is used to apply single-phase to ground faults. The resistors (Ω) represent a three phase local load, which is energized by a three-pole toggle switch.

Figure 2: Elements arrangement inside the Line Section

The recloser section includes a low voltage contactor which models the recloser, as well as the actual relay which is normally installed in the distribution line pole. The microprocessor based relay controls the low voltage contactor for manual commands as well as for automatic commands related to the contingency state of the feeder. The low voltage contactor in this section has the capability of interrupting short circuit currents. The opening of this device isolates the source and de-energizes the distribution feeder. The sectionalizer section was modeled similarly to the recloser section. A fundamental difference, however, is the fact that the low voltage contactors which model the sectionalizers do not have the capability of interrupting short circuit currents. These devices are allowed to open only during the period of time the recloser is opened, i.e., when the distribution feeder is de-energized. A microprocessor based relay controls the low voltage contactor for manual commands as well as for automatic commands related to the contingency state of the feeder.

Traditional protection scheme applied to the model As it was mentioned earlier, the analogue simulator may operate under normal and contingency conditions. When the analogue simulator is operating under the normal state it feeds all loads in the distribution feeder. The contingency state of the analogue simulator presents different scenarios according to the fault location and the fault characteristics. As far as fault location is concern, faults may be applied on four different physical locations along the distribution feeder, as shown in Figure 3. Regarding to the fault characteristics, two sub classifications apply: fault nature and fault type. The nature of faults that may be applied in the analogue simulator are permanent and transient faults. According to fault type, three-phase and single phase to ground faults are possible to implement. Clearly the response of the recloser and sectionalizers in the simulator depends on the fault location and characteristic that is applied.

Figure 3: Possible Fault Location to be applied in the Analogue Simulator

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer

SectionalizerSubstation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3T

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR S

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer

Sectionalizer

S1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3T

Recloser

Normally

Open

Sectionalizer

Sectionalizer

R L

R N

S F F

T F FΩ

Ω

Ω

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3S3

Recloser

Normally

Open

Sectionalizer

Sectionalizer

F1 F3F2

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

T

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer

Sectionalizer

SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3S3

Recloser

Normally

Open

Sectionalizer

Sectionalizer

F1 F3F2

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3S3

Recloser

Normally

Open

Sectionalizer

Sectionalizer

F1 F3F2

Substation

CB

Sectionalizer

Breaker

CB RR S

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

T

Recloser

Normally

Open

Sectionalizer

Sectionalizer

S1 SS2

T

Recloser

Normally

Open

Sectionalizer

Sectionalizer

The microprocessor based relays provide measurement, protection and control functions for the distribution feeder. The protection functions were used in order to coordinate the recloser with the sectionalizers using the traditional protection scheme for overhead distribution feeders [2]. The recloser protective functions for phase overcurrent, ground overcurrent, and reclosing sequence were utilized. The recloser sequence of operations is shown in Figure 4. For the sectionalizers, the number of counts and current threshold were set.

Figure 4: Sequence of operations for the Recloser

After the recloser – sectionalizers coordination was performed, the analogue simulator is able to correctly respond to the contingency scenario that was applied. In other words, the analogue simulator isolates the faults according to the fault location and the fault characteristics. Table 1 shows the element responsible of opening the distribution feeder for each fault location as depicted in Figure 3. We can see that there is no operation of sectionalizer T, since this sectionalizer is normally open. Considering permanent faults, the last column in the table shows the recloser sequence of operations, according to Table 1.

Table 1: Coordination of the protection elements in the Distribution Feeder

Fault Location Distribution Equipment

Number of Reclosing

Operations F1 Recloser 3 F2 Sectionalizer 1 2 F3 Sectionalizer 2 1

For permanent faults, Table 1 shows the number of reclosing operations that the recloser has to perform in order to isolate the fault as a function of the fault location under the traditional protection practices [1]. This table reflects the coordination that is possible to obtain using the power system and the protection functions of the

microprocessor based relays; it also revels an area of opportunity to improve the distribution feeder reliability.

Internet Peer to Peer communication based protection system

In this particular application, peer-to-peer communication system enables distribution relays to share information with others connected to the communication network without having a master device. Every relay is able to ask from, and send to, the network un-requested information. Furthermore, and even more important, any relay can master the re-configuration of the distribution system itself after a contingency occurs. Thus, the system can be programmed to isolate every possible fault after a certain number of reclosing operations as well as to reenergize unfaulted loads. As a result, the traditional protection system is transformed into an adaptive protection system able to reconfigure itself to successfully face contingency conditions.

TCP/IP communication based A TCP/IP protocol has been chosen as a communication platform. It enables devices to have access to either Internet or a private Intranet. Therefore, communication availability constrictions are curbed by using and widely spread protocol. Figure 5 shows the architecture of the proposed system. The scheme must be implemented with a fall-back operation procedure for those times when communication is not available.

Figure 5 Web-Based Communication Loop Control

The access to the TCP/IP protocol was accomplished by using low cost Java Application Control Engines built

Time

0.5 seconds, 1ST Reclose

15 seconds, 2 nd Reclose

45 seconds, 3rd Reclose

F1

F3

F2

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3T

Recloser

Normally

Open

Sectionalizer

PCD SCD1 SCD2

T

TCP/IP

Web

Server

F1

F1

F3

F3

F2

F2

Substation

CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3

Recloser

Normally

Open

Sectionalizer


CB

Sectionalizer

Breaker

CB RR SS1 SS2

S3T

Recloser

Normally

Open

Sectionalizer

PCD SCD1 SCD2

T

TCP/IP

Web

Server

upon a compact embedded PowerPC™ platform with Flash Memory for backup. Furthermore, they are capable to configure a Modbus serial communication network with one o more Modbus devices. Communication can be established by using either RS-232 or RS-485 ports. A complete set of Java™ based control, application, logging, and user interface "objects" are included in a library for the Systems Integrator to create a robust control system for any size system. The Java based control can also be configured with optional Web User Interface Services. In this configuration, the system's graphical views can be accessed using any standard Web browser such as Netscape Navigator™ or Internet Explorer™.

Operating philosophy A typical radial overhead distribution system as the one we have shown in Figure 5 was used during the implementation of the Web-Based Peer-to-Peer Automation Distribution System in the Analogue Simulator. The firmware of the PCD (Power Control Device) and SCDs (Switch Control Device) of the system included a ModBus register (40819) which provided us with a count within the sequence of events during a fault. This counter allowed us to determine the precise moment of inception of a fault and helped us to determine which protective devices (PCD or SCDs) registered its presence. The operating philosophy of this system is as follows: The recloser is the only device capable of opening short circuit currents. Sectionalizers that see the fault ahead should be coordinated in such a way that only the device closest to the fault should operate after the second reclosing operation and while the recloser is open. Furthermore, it has to send an open/lock command to the device located at the end of the faulted segment. Every device that remains closed down-stream of the faulted line switches to alternating settings. Once the fault was completely clear (both edge devices open) the tie device close and un-faulted load is reenergized. The communication system works as follow: Devices are constantly posting in the network their event counter (40819) and status registers. Furthermore, they are constantly reading from the network event counter and status of all devices in the power distribution network. Having this information enables them to evaluate the system condition. When a fault occurs anywhere in the distribution system, the recloser starts its sequence of operations. Thus, event counter registers increase with every reclosing operation only in devices that see the fault current. The fault location is accomplished by comparing 40819 register of successive devices. If a device sees the current but its down stream fellow does not, the fault is between them. Once the fault was located, the closest

device master the reconfiguration of the network. It generates and sends open/close/lock commands to other devices according to the operating philosophy described above. In order to better understand what is stated in the previous paragraphs, let’s consider Figure 5. The logic schemes for permanent faults in F1, F2 and F3 are shown in Tables 2, 3 and 4 respectively.

Table 2 Logic Scheme for a Permanent Fault at F1

Time 40819 Status Response 0 PCD=0

SCD1=1 SCD2=1

PCD=C SCD1=C SCD2=C T=O

0+ PCD=0 SCD1=1 SCD2=1

PCD=O SCD1=C SCD2=C T=O

PCD opens and momentarily clears the fault current.

0.5 Sec

PCD=1 SCD1=1 SCD2=1


PCD recloses. First reclosing operation. Event counter register increase only in PCD. Neither SCD1 nor SCD2 senses current fault. Therefore they do not increase their event counter and remain close.

PCD=1 SCD1=1 SCD2=1


PCD senses the fault and opens clearing the fault current momentarily. All registers remain unchanged.

PCD=1 SCD1=1 SCD2=1


SCD2 reads from the network SCD1-40819 register. Since SCD1-40819=1 and SCD2-40819 = 1 , the fault is not between SCD1 and SCD2. Therefore they do not have to take any action.

PCD=1 SCD1=1 SCD2=1


SCD1 reads from the network PCD-40819 register. Since PCD-40819=1 and SCD-40819 = 1 , the fault is between PCD and SCD.

PCD=1 SCD1=1 SCD2=1

PCD=O/L SCD1=O SCD2=C T=O

SCD1 masters the system reconfiguration. SCD1 opens itself. It also sends a lockout command to PCD.

PCD=1 SCD1=1 SCD2=1

PCD=O/L SCD1=O SCD2=C T=O

Tie breaker read from the network PCD counter, PCD status, SCD1 counter and SCD1 status. Since PCD-40819 = 1 and PCD status is locked and SCD1 status is open, Tie breaker send a changing settings command to SCD2.

PCD=1 SCD1=1 SCD2=1

PCD=0/L SCD1=0 SCD2=C/AS T=O

SCD2 changes to alternating settings

PCD=1 SCD1=1 SCD2=1

PCD=O/L SCD1=O SCD2=C/AS T=C

SCD2 reads from the network confirmation that T alternating settings have been set up. Thereafter, it close itself

10 Sec

PCD=1 SCD1=1 SCD2=1

PCD=O/L SCD1=O SCD2=C/AS T=C

Unfaulted load are fed by the backup source.

Table 3 Logic Scheme for a Permanent Fault F2


SCD1=1 SCD2=1





0.5 Sec

PCD=1 SCD1=3 SCD2=1


PCD recloses. First reclosing operation. Event counter register increase in PCD and SCD1 because they see fault current. SCD2’s event counter remains unchanged.

PCD=1 SCD1=3 SCD2=1



PCD=1 SCD1=3 SCD2=1


SCD1 reads from the network PCD-40819 register. Since PCD-40819=1 and SCD1-40819 = 3 , the fault is not between PCD and SCD. Therefore they do not have to take any action

PCD=1 SCD1=3 SCD2=1


SCD2 reads from the network SCD1-40819 register. Since SCD1-40819=3 and SCD2-40819 = 1 , the fault is between SCD1 and SCD2.

PCD=1 SCD1=3 SCD2=1

PCD=O SCD1=O SCD2=O T=O

SCD2 masters the system reconfiguration. SCD2 opens itself. It also sends an opening command to SCD1.

PCD=1 SCD1=3 SCD2=1

PCD=C SCD1=O SCD2=O T=O

PCD read from the network SCD1 counter, SCD1 status, SCD2 counter and SCD1 status. Since PCD-40819 = 1 and SCD1 status is open and SCD2 status is open, PCD closes itself .

PCD=1 SCD1=3 SCD2=1

PCD=C SCD1=O SCD2=O T=C

Tie breaker read from the network PCD counter, PCD status, SCD1 counter and SCD1 status. Since PCD-40819 = 1 and SCD1 status is open and SCD2 status is open, Tie breaker (T) closes itself .

15 sec

PCD=1 SCD1=3 SCD2=1

PCD=C SCD1=O SCD2=O T=C

Unfaulted load are fed by the backup source.

Table 4 Logic Scheme for a Permanent Fault at F3


SCD1=1 SCD2=1





0.5 Sec

PCD=1 SCD1=3 SCD2=3


PCD recloses. First reclosing operation. Event counter register increase in PCD, SCD 2 and SCD1 because they all see fault current.

PCD=1 SCD1=3 SCD2=3



PCD=1 SCD1=3 SCD2=3

PCD=O SCD1=C SCD2=O T=O

SCD2 opens itself by internal settings.

PCD=1 SCD1=3 SCD2=3

PCD=C SCD1=C SCD2=O T=O

PCD reads from the network SCD1-40819, SCD2-40819 and SCD2 status registers. Since SCD1-40819=3 and SCD2-40819 = 3 and SCD status is open, PCD close itself.

Results and Conclusions

Table 5 summarizes the above results. It shows a comparison in terms of the number of reclosing operations, unfaulted load losses, and recovery time for several contingences between the traditional protection system and Peer-to-Peer Communication-Based protection system applied to distribution networks. Table 5: Comparison of the Traditional Protection Scheme and the Peer-

to-Peer Communication Based Protection System.

Traditional Protection System Fault Location Reclosing

Operations Unfaulted

Losses Recovery

Time F1 3 100% N/A F2 2 50% 180 Sec F3 1 0% 45 Sec

Peer-Peer Communication Based Protection System Fault Location Reclosing

Operations Unfaulted

Losses Recovery

Time F1 1 0% 15 Sec F2 1 0% 15 Sec F3 1 0% 15 Sec

According to Table 5, adding peer-to-peer communication to a traditional distribution protection system significantly enhances its performance. Assessing quality of service by means of measuring the percentage of undesired load losses, peer-to-peer communication based protection gets the highest score. In other words, for any contingence that may happen in the system no unfaulted load is lost. Indeed, only faulted load is permanently shut down. Although the peer-to-peer communication based protective system was implemented and tested for a radial overhead distribution system, it is only one of many possible applications of an adaptive protection systems. Among these applications may be:

- Underground Distribution System. - Fault detection - Power quality - Remote supervision

As we have shown in this work, the adaptive protection distribution system is the result of bringing all together the latest advances in computer, communication and

protection. Thus, a significant improvement in distribution system reliability has been achieved by blending all these new concepts in one specific application.

Acknowledgments

The authors of this paper would like to express their gratitude to ABB/ETI and Tridium Corporation for their support during the development of this study.

References

1. Electric Utility Engineering Reference Book, Volume

3, Distribution Systems. Westinghouse Electric Corporation. R.R. Donnelley and Sons Company, Chicago and Crawfordsville, Indiana. 1965.

Appendix

Distribution feeder analog simulator

2

High Impedance Fault DetectionOn Distribution Feeders

by

Mark Adamiak Craig Wester Manish Thakur Charles JensenGE Power Management JEA

AbstractThe ability to detect High Impedance (HiZ) faults hasbeen a topic of research and development for over 30years. About seven years ago, products began toappear on the market that could securely perform thisfunction. Over this time period, several hundred HiZdetection devices have been placed in service andhave performed to expectations. This paper reviewsthe operating principle of HiZ fault detection, looksat the application issues encountered over this time,highlights some of the actual detections, and looks atpossible future directions of the technology.

I. IntroductionFrom the beginning of power distribution, the powersystem protection engineer has been challenged withthe detection of HiZ faults. The IEEE Power SystemRelay Committee working group on High ImpedanceFault Detection Technology [1] defines HiZ faults asthose that “do not produce enough fault current to bedetectable by conventional overcurrent relays orfuses”. As such, it should be noted that whereastraditional protection is designed to protect the powersystem, HiZ protection is primarily focused on theprotection of people and property.

The typical HiZ fault is when a conductor physicallybreaks and falls to the ground. The break in theconductor will usually result in either a drop in loadon the affected feeder or possibly a momentaryovercurrent condition as the falling conductor brieflycomes in contact with a solidly grounded object.Once on the ground, the resulting electrical signatureis very much a function of the contacted surface.Surfaces such as concrete, grass, dirt, and wetsurfaces in general will result in an “arcing fault”with RMS fault currents in the range of 10 to 50amps whereas surfaces such as dry sand and asphaltwill result in a constant low level of current flow.Arcing faults result in a very definable and detectablepattern whereas the signatures presented by the lattersurfaces present a challenge to secure and reliabledetection.

A related type of HiZ fault is when the conductordoes not break, but comes into contact with groundedobjects either through a failure of the conductormounting system, insulation failure, or inadvertentcontact with some external element such as a treelimb. These faults will usually exhibit the same“arcing” signature as a broken conductor lying on theground, however, the event will not be preceeded byany change in fundamental current.

A third type of event is a sagging conductor.Although not technically a “fault”, it does present aconsiderable public safety hazard. In thiscircumstance, a conductor hangs low enough toenable human or other contact. Note that this type ofevent offers no electrical signature for detection.

The frequency of downed conductors is a topic fordiscussion as most occurrences are not logged byfield crews. Best estimates are that between 5% to10% of all distribution system fault events aredowned conductors. See below photo of downedconductor.

II. Detection TechniquesDetection of HiZ faults fall into two categories:mechanical detection and electrical detection. The

3

following sections offer a brief review of the varioustechniques that have been developed in these areas.

a. Mechanical DetectionMechanical detection usually involves some way offorcing contact with a solid ground in order to allowconventional overcurrent protection to operate.

The first type of mechanical HiZ detection methodconsists of a device(s) mounted to a cross arm orpole. The device is mounted under each phase wirein order to catch the conductor as it falls to theground. The force of the falling conductor releasesan internal spring that ejects a bus bar to makecontact with the fallen wire and create a lowimpedance ground fault. The ground fault createdwill cause conventional overcurrent protection tooperate. Sagging conductors that do not come incontact with earth or a grounded object could bedetected by this mechanical method. The installationand maintenance costs are high. For bi-directionalcoverage, six units would have to be mounted oneach pole. Even though the cost may be high toallow usage on every pole, utilities may install incertain areas, such as churches, schools, or hospitals.

A second type of mechanical HiZ detection methoduses a pendulum mounted aluminum rod with hookedends. It is suspended from an under-built neutralconductor. The falling conductor is caught andproduces a low impedance ground fault, whichoperates conventional overcurrent protection.Typically, two units are mounted per span. Saggingconductors that do not come in contact with earth or agrounded object could be detected by this mechanicalmethod. Ice, wind, and tree growth could cause afalse detection.

b. Electrical DetectionThere are three primary “algorithmic” techniques thathave been developed and field tested to date. Asummary of these three systems follows:

High Impedance Fault Analysis SystemThis electrical HiZ detection method measures thethird harmonic current phase angle with respect to thefundamental voltage. There is a distinct phasorrelationship between the third harmonic current andthe faulted phase voltage. The device calculates andstores the average ambient third harmonic currentphasor. When a fault occurs, the new third harmoniccurrent phasor is vectorially subtracted from thestored value. A high impedance fault is issued if themagnitude is above setting and angle matches apredetermined value for a downed conductor. Thedevice acquires current and voltage values from the

relaying current and voltage transformers. Typically,one unit is installed in each distribution breaker.Units have been in service since the early 1990’s.

Open Conductor DetectionThis electrical HiZ detection method detects loss ofvoltage to determine a broken conductor. The systemmeasures the voltage at each end of a single phaselateral. When the voltage of any phase drops belowthe specified threshold, a transmitter sends a signalon the neutral conductor to a receiver at the upstreamdevice. The upstream device opens if voltage ispresent at the upstream device. Systems have beenunder test since 1992.

Signature Based HiZ DetectionThe signature based HiZ IED performs expert systempattern recognition on the harmonic energy levels onthe currents in the arcing fault. This technique isbased on the technology developed at Texas A&MUniversity after more than two decades of research,funded in part by the Electric Power ResearchInstitute. The HiZ IED uses a high waveformsampling rate (64 samples/cycle) on the ac currentinputs to create the spectral information used in thesignature analysis. Expert system techniques areemployed to assure security while maintainingdependability.

The overall process incorporates nine algorithms,each performing a specific detection or classificationfunction. High impedance fault detection requiresinputs from the three phase and ground currents viarelaying current transformers. Voltage inputs are usedto enhance security and to provide supplementalphase identification and are not required for arcingdetection.

The primary detection algorithms are the Energy andRandomness algorithms. The Energy algorithmfocuses on the fact that arcing causes bursts of energythat register throughout the frequency spectrum. Theenergy values – computed as the square of theharmonic and non-harmonic spectral components(excepting the fundamental) – are integrated into odd,even, and non-integer harmonics values. Sampling at64 samples per cycle allows computation offrequency components up to the 25th harmonic. TheEnergy algorithm monitors these computedharmonics on all phase and ground currents. Afterestablishing an average energy value for a givensignal, the algorithm indicates “arcing” if it detects asudden, sustained increase in the value of thatcomponent. Figure 1 shows “normal” energy levelsas measured on an actual feeder. Indications ofenergy increase are reported to the Expert Arc

4

Detector (EAD), which performs a probabilisticintegration of the arcing inputs from all phases andall harmonic components.

The second detector in the algorithm suite is theRandomness algorithm. This algorithm keys on asecond characteristic of an arcing fault, which is thefact that the energy magnitudes tend to varysignificantly on a cycle-to-cycle basis. Figure 2shows the energy values during an arcing fault. Thehigh level of energy as well as the variance in theenergy can clearly be seen. The Randomness

measures these magnitude variations and reportdetection of magnitude variation to the Expert ArcDetector.

The purpose of the Expert Arc Detector algorithm isto assimilate the outputs of the basic arc detectionalgorithms into one cumulative arc confidence levelper phase. There are actually 24 independent basicarc detection algorithms, since both the Energy andRandomness algorithms are run for the odd, even andnon-integer harmonics for each phase current and forthe neutral/ground current. An arc confidence level

is determined for each phaseand neutral/ground. Theexpert arc detector algorithmcompares the cumulative arcconfidence level values or highEAD counts to the user’s arcsensitivity setting. Figure 3shows the block diagram ofhow the Energy, Randomness,and Expert Arc Detectoralgorithms function together.

For the device to be secure anddependable, the Expert ArcDetector integrates the outputsfrom the Energy andRandomness algorithms. Thenumber of times that theintegration is performed is, aswell as the integration level.depends on the arc sensitivitysetting. The more sensitive thesetting, the lower theintegration level and the fewerintegrations required.

An “arcing detected” output isissued once all the EADrequirements are satisfied. Ifeither a loss of load or amomentary overcurrentcondition is detectedimmediately before an “arcingdetected” output is registered,the “downed conductor” outputis set to indicate that there isactually a conductor on theground.

If the device determines that adowned conductor or arcingexists, it attempts to determinethe phase on which the highimpedance fault condition

Figure 2Arcing Fault Energy Levels & Randomness Signature

Figure 1Normal Energy Levels

5

exists in a hierarchical manner. First, if a significantloss of load triggered the arc detection algorithms,and if there was a significant loss on only one phase,that phase is identified. If there was not a singlephase loss of load, and if an overcurrent condition ononly one phase triggered the algorithm, that phase isidentified. If both of these tests fail to identify thephase, the phase with a significantly higherconfidence level (e.g. higher than the other twophases by at least 25%) is identified. Finally, if noneof these tests provides phase identification, the deviceanalyzes the correlation between the peak portion ofthe voltage waveform with the neutral/ground arcbursts. If there is correlation with a particular phasevoltage, that phase is identified. If that test fails, thephase is not identified.

Conductors that do not continuously arc, but havetime periods between arcs can be detected by thearcing suspected identifier algorithm. For example, ifarcing is caused by tree limb contact or insulator

degradation, arcing will typically be presentintermittently with relatively long periods ofinactivity. In such cases, arcing may be affected bysuch factors as the motion of a tree limb or themoisture and contamination on an insulator. Thepurpose of the arcing suspected identifier algorithm isto detect multiple, sporadic arcing events. If takenindividually, such events are not sufficient to warrantan arcing alarm. When taken cumulatively, however,these events do warrant an alarm to system operators,so that the cause of the arcing can be investigated.The user can select the number of maximum numberof arcs and an acceptable period of time. Due to thepossible long periods of arcing inactivity, a HiZdecision could be reached in up to 5 minutes.

III. Signature Based HiZApplication Issues

The following sections highlight a number ofapplication guidelines developed over the last severalyears of HiZ detection device installations.

OptionalVoltageInputs

60 Hz Block& Amplifier

Convert ToOdd, Even &

Non-Harmonics

IdentifyLarge

ChangesIn Energy

RMSCurrentDetector

PhaseIdenti-fication

DownedConductor

Ia,b,c,n

Ia,b,c,

RandomnessAlgorithm

Load EventDetector

Load AnalysisAlgorithm

Load Extraction AlgorithmArc Burst PatternSpectral Analysis

Arcing SuspectedIdentifier

ExpertArc

Detector

ArcingSuspectedIdentifier

PatternAnalysis

ConfirmationChecks

ArcingSuspected

Alarm

Expert ArcDetector

IdentifyRandom

Changes InEnergy

Figure 3Signature Based HiZ Detection Block Diagram

In

Energy Algorithm

6

a. Arcing Fault Response ProceduresAs previously described, signature based HiZalgorithms can provide three different outputdesignations, namely: arcing suspected (orintermittent arcing), arcing detected, and downedconductor. Each utility needs to establish standardresponses to each of these outputs. At this stage inthe implementation cycle, typical responses havebeen as outlined in Table 1.

If tripping of the feeder is chosen as a course ofaction, one of the ensuing challenges is locating theHiZ fault. While energized, the arcing fault / downedconductor can often be located via sight, sound, radiofrequency interference (RFI), or loss of power in anarea. Once the feeder is de-energized, all the abovebecome non-functional. As such, the decision to de-energize or not to de-energize must be based on therelative consequences of each action. For example, ifthe region is around a school or residential area, thereis a strong bias to de-energize. On the other hand, ifthe arcing line is feeding a hospital or an industrialregion, the decision might be to alarm.

Condition PrimaryResponse

SecondaryResponse

ArcingSuspected Alarm -

ArcingDetected Alarm Trip

DownedConductor Trip Alarm

It is strongly recommended that any utility installingHiZ detection devices develop a written responseprocedure to each of the above HiZ conditions.

b. Line GroundingThe HiZ element was primarily designed for solidlygrounded systems. The same algorithm has beentested with some degree of success on impedancegrounded systems as well a few tests on ungroundedsystems. The algorithm did pick up, however,consistency of operation was an issue. One other testperformed involved a downed conductor opposite thesource side of the line (see Figure 4). In thisconfiguration, there was a down-stream transformer.When the transformer was loaded, detection of thedowned conductor back in the substation wasachieved.

c. CT RatioThe ground current on a downed conductor may beonly a few amperes on a feeder with several hundredamperes of load. Choosing as small a CT ratio aspossible maximizes the arcing component in thewaveform and optimizes the ability of the HiZalgorithms to detect the HiZ fault. The algorithm hasbeen successfully tested with CT ratios on the orderof 1200:5. The HiZ algorithm use standard relayaccuracy CTs.

d. Sensitivity Vs. SecurityThe major setting in a HiZ device is ArcingSensitivity. HiZ detection is no different from anyother protection scheme in that there is a trade-offbetween sensitivity and security. An algorithm canbe designed to pick-up on almost any disturbance onthe feeder. The challenge is being able todiscriminate between events. The sensitivity settingrepresents a balance control between sensitivity andsecurity. Security can be enhanced by requiringmultiple detections of the arcing condition before aHiZ condition is declared.

Typical recommendations are for a balance ofsensitivity and security in the normal operating mode.Under conditions such as an impending storm, it maybe desirable to actually de-sensitize the algorithm, aswith everything wet, there is usually much arcingleakage around the system. On the opposite extreme,if a region has been experiencing a dry spell, it maybe desirable to set the sensitivity to maximum. Inany event, remote control of setting groups to allowsuch changes is desirable.

e. Overcurrent CoordinationThe general consensus for feeder fault protection isthat, given there is sufficient current, to have theovercurrent element(s) operate and trip out the feederbefore the HiZ element operates. This dictates theneed for an overcurrent coordination timeout period.Setting of this coordination time should be based onthe operating time of the time overcurrent (TOC)element for a fault located at the end of the feeder.The HiZ algorithm can operate in as little as 20seconds whereas a TOC relay may take much longerto operate. Too long a coordination time (> 1minute) is not recommended as HiZ faults tend todecrease in magnitude over time as the conductor“glasses over” and/or breaks – resulting in a smallerground contact area.

A related application note on TOC relays is the needto coordinate not only the operate time, but also thereset time. On HiZ faults, a TOC element will“ratchet”, that is, move forward for a period of time

Table 1Typical Arcing Condition Responses

7

and then begin to reset as the fault current dropsbelow the pickup level of the relay. If a TOC relaywith instantaneous reset is placed downstream of aTOC relay with timed reset, the relays may mis-coordinate resulting in the disconnection of more ofthe feeder than desired.

IV. Experience to DateTo date, utilities around the world have installed overseveral hundred HiZ detection devices. Dozens ofreal arcing suspected, arcing detected, and downed-conductor events have been recorded with a numberof the installations connected to trip. Almost allreport into SCADA. The ratio of “detected” downedconductors to the total population of downedconductors has been about 80%. The following are afew highlights from the accumulated experiencebase:

• On the JEA Jacksonville, FL system, reports of“arcing suspected” were being received from a HiZdevice at the same time every day for a period oftime. Figure 5 shows “arc confidence”, integratedarcing information from the reporting IED. Note thatthe arc confidence rose quickly for two integrationperiods then settled out. As a result, the detection onthis event was reported as “arcing suspected” initiallyand shortly after, “arcing detected” was declared.Inspection of the line uncovered no obvious arcingsites. Following this result, an analysis of thecustomer base connected to the suspect feeder was

performed and one customer with a heavy-dutyprocess was identified. A phone call to the identifiedcustomer was made to inquire if any of his processesincluded arc furnaces or other arcing loads to whichthe customer responded “no”. On the day followingthe inquiry by JEA, the customer phoned back andstated that a large motor in their facility had justfailed. The HiZ device was able to see through thedistribution transformer into the customer site and thecustomer motor.

• The connection to the high voltage bushing ofa distribution transformer had become loose andbegan to arc. The resulting signature was detected bya HiZ device (as well as the customer, when his lightswent out once the connection burned through.)

• After a long dry spell, a rainstorm came intothe area. Many of the insulators on the feeders,which had become quite contaminated, began toconduct in an arcing manner. In conjunction with thestorm was lightning, which produced a transient faulton one of the feeders. As a result, the HiZ IED sawfault current followed by arcing and declared a“downed conductor”.

• Many utilities have performed staged fault testson their systems in order to test the effectiveness ofHiZ detection. In most cases, the utility wouldinclude a “challenge” test case – typically aconductor dropped on asphalt or sand. In this onetest, the conductor was dropped on asphalt with the

To Substation

3-PhXfmr

IndustrialLoad

Downed Conductor

Figure 4Load Side Downed Conductor

8

expectation of no detection. What occurred,however, was that the arc found paths through cracksin the asphalt that permitted arcing and subsequentdetection by the HiZ device.

• One question often asked is how “directional”is the HiZ algorithm? To find the answer to thisquestion, one utility ran staged fault tests with HiZIEDs installed on two parallel feeders (see Figure 6).HiZ faults were placed on one feeder while theperformance of the parallel feeder was observed. Inall cases, the HiZ IED on the non-faulted feeder didnot detect any arcing, while the HiZ IED on thefaulted line detected about 80% of the HiZ stagedfaults.

• As utilities expand their usage of HiZ devices,they are surprised by the number of arcing conditionsexisting on their distribution system. As JEA addedHiZ signature devices to 27 feeders, it came as a total

surprise that 50% of these feeders began to report“arcing suspected” conditions. Now that JEA knowsthat something is happening, they plan to use otherdevices to help locate/determine the root of the arcingconditions

• Finally, in the challenge arena, several HiZfaults were staged on dry sandy soil. In most cases,the HiZ IED did not detect arcing. Analysis of thewaveforms from these faults does show a change inenergy; however, sand does not exhibit the“randomness” of other material types.

V. Lessons LearnedAs a result of the knowledge base garnered fromseveral years of field experience, a number ofenhancements to the HiZ algorithms have beenidentified.

Figure 5Arcing Signature for Failing Motor

9

Loads Arcing Fault

X

HiZIED1

HiZIED2

a. Downed Conductor MisclassificationFirst and foremost has been the issue of mis-classification of an arcing fault as a downedconductor for a fault on a parallel feeder followed byarcing. The parallel feeder fault drops the voltage onthe substation bus and assuming near unity powerfactor operation of the feeders, all feeders connectedto the bus subsequently see a loss of load. If this lossof load is followed by arcing (as was the case withthe contaminated insulators previously mentioned),the HiZ IED will declare a “downed conductor”. Asimple fix in the form of an under-voltage restraintwas added to the loss of load logic. Now, if a loss ofload occurs in conjunction with an undervoltage, theloss of load logic flag is not set.

b. Transformer Inrush RestraintAlthough no reported cases exist, the inrushwaveform resulting from the energization of atransformer can look like arcing. The inrushwaveform, however, is very distinguishable ascompared to arcing. In particular, inrush has a veryhigh second harmonic component – much higher thatthat seen in arcing. Given this simple differentiator,

an arcing restraint was added to the algorithm suchthat if the 2nd harmonic component of the waveformis greater that a percentage of the fundamental (a usersetting – typically about 15%), the arcing detectedalgorithm is reset and block from operation.

c. Dynamic Energy Level AdjustmentIn the course of field experience, it was clearlyobserved that not all feeders were created equal withregards to the steady state harmonic energy levelsthat existed. This variance required the setting of aminimum energy threshold significantly above whatthe energy levels on a typical feeder would be. Inorder to optimize the sensitivity for each individualinstallation, a “dynamic” energy threshold was addedto the algorithm. In this mode of operation, theaverage harmonic energy level on a feeder ismeasured over a 3-day period. The harmonic energythresholds are then set at a value of 3-sigma abovethe average energy value thereby allowing eachfeeder to operate at maximum sensitivity.

d. Oscillography and Sequence Of Events(SOE) Overrun

Arcing events tend to be bursty, that is, an event maypick-up for awhile, settle out, then pick up again. AsIEDs try to log the activity, SOE and oscillographylogs tend to overrun. Solutions to this problem aretwofold: With regard to oscillography, the concept ofpriority was developed. All file types were assigneda priority and depending on the priority, it couldover-write a file of lower priority. For example, afile created by a “downed conductor” event (highestpriority) would be allowed to over-write either an“arcing detected” waveform file (medium priority) oran “arcing suspected” waveform file (low priority).With regard to SOE overrun, arcing events werelatched for up to 10 minutes thereby allowing onlyone arcing event entry every 10 minutes – asignificant reduction in the possible number of eventsthat could be entered in the SOE log.

VI. Future DirectionsGiven the HiZ detection experience to date, there area number of areas where further investigation andresearch are desirable. This section highlight a fewof these areas:

a. HiZ Fault LocationAs mentioned earlier, once an arcing fault is detected,there is the challenge of locating the faulted circuit.A distance to fault calculation has often been talkedabout, but at this juncture, it is still some ways away.JEA is investigating using a corona camera (Figure 7)to aid in the fault location process.

Figure 6Paralleled Feeder Selectivity

10

The camera spectrally images the corona energy fromthe conductor and then superimposes the spectralenergy onto the background object. The benefit ofthis technology is that it can be operated in directsunlight.

b. HiZ DirectionalityAs HiZ devices become more common in thedistribution system, the need to coordinate arcdirection on the same feeder becomes desirable. Inparticular, in the scenario of a main breaker andseveral down-stream reclosers, it would be desirableto be able to sectionalize the HiZ faulted section as ispresently done for low impedance faults.Sectionalization could be optimized with the additionof recloser-to-recloser communication. Radiocommunication systems are readily available todaythat can provide the communication channel andUCA based relays already incorporate the ability tomessage among themselves.

c. Sand SettingsAs noted in the paper, sandy soils do produce arcingenergy, however, they do lack the randomnesscomponent. Future developments need to explore thepossibility of creating a sand setting that focuses onthe energy aspect of an arcing fault and de-emphasizes the randomness component.

d. HiZ Fault Type DeterminationIt is desirable to be able to determine the type of HiZfault based on the signature of the energy waveforms.Ideally, the signature analysis would be able toidentify not only an arcing conductor but alsoequipment trouble such as a contaminated insulator, afailing transformer or an arcing motor. Effort isneeded to build the database of these disturbances toallow such discrimination.

ConclusionsHiZ Detection technology has taken major strides inthe last several years and the knowledge andexperience base surrounding it has growndramatically. It is clear that as technology advances,so will our ability to do more with arcing waveformsincluding advanced sensitivity and detailed eventtype analysis. It has also become clear that utilitiesneed to take a “system” approach to HiZ detection ontheir distribution system by taking advantage of allthe mechanical and electrical HiZ detection devicesoffered by the industry.

Figure 7Corona Camera

11

Bibliography

1. High Impedance Fault DetectionTechnology. Report of PSRC WorkingGroup D15; http://www.pes-psrc.org/d/D15MSW60.html

2. B.M. Aucoin, R.H. Jones, “HighImpedance Fault Implementation Issues”,IEEE Transactions on Power Delivery,January 1996, Volume 11, Number 1, pp139-148.

3. R. Patterson, W. Tyska, “AMicroprocessor-Based Digital FeederMonitor with High-Impedance FaultDetection”, presented at 47th AnnualConference for Protective Relay Engineers– Texas A&M University, March 1994.

Authors

Mark Adamiak received his Bachelor of Scienceand Master of Engineering degrees from CornellUniversity in Electrical Engineering and an MS-EEdegree from the Polytechnic Institute of New York.Mark started his career with American Electric Power(AEP) in the System Protection and Control sectionwhere his assignments included R&D in DigitalProtection, relay and fault analysis, Power LineCarrier and Fault Recorders. In 1990, Mark joinedGeneral Electric where his activities have rangedfrom development, product planning, and systemintegration. In addition, Mr. Adamiak has beenactively involved in developing the framework forthe implementation of the MMS/Ethernet peer-to-peer communication solutions for next generationrelay communications. Mark is a Senior Member ofIEEE, past Chairman of the IEEE RelayCommunications Sub Committee, and a member ofthe US team on IEC TC57 - Working Group 11 onSubstation Communication.

Craig Wester was born in Belgium, Wisconsin, andreceived a B.S. in Electrical Engineering with astrong emphasis on power systems from theUniversity of Wisconsin-Madison in 1989. Craigjoined General Electric in 1989 as a utilitytransmission & distribution application engineer. Mr.Wester is currently an employee of GE PowerManagement as Regional Sales Manager. Mr.Wester’s role consists of providing salesmanagement, power system protection applicationand support to the investor-owned utilities, ruralelectric cooperatives, electric municipals, consultants,and OEMs throughout the southern US for GErelaying equipment. Mr. Wester is a member of theIEEE.

Charles Jensen is Manager, System Protection andControl for JEA, Jacksonville, FL. In this role, Mr.Jensen, oversees and manages all protective relayingand control from generation through the TS&Dsystems up to the ultimate utilization customer. Priorto this position, Mr. Jensen was Director, SystemInformation and Distribution Automation (DA) forthe Jacksonville Electric Authority. As Director ofDA, Mr. Jensen's responsibilities includedmanagement of a multi-million, multi-year, multi-phase Distribution Automation Project. Mr. Jensenhas held progressively responsible positions withinthe utility industry leading to his present position. AsDivision Chief of Technical Support Engineering,Mr. Jensen was responsible for distribution systemstandards, AM/FM/GIS systems, distribution systemconnectivity databases, power quality, distributionsystem analysis and developing of engineering designsystems. Other areas of specialty include, troublecall management systems, distribution transformertest system design and implementation, protectionand control systems, substation designs, preventativemaintenance program development andimplementation and fleet and equipmentmanagement. Mr. Jensen received his M.S. degree inElectrical Engineering/Systems Science and a B.S.degree in Electrical Engineering from Michigan StateUniversity. Mr. Jensen is a dually registeredProfessional Electrical Engineer and MasterElectrician in both Florida and Michigan. Mr.Jensen is a member of the IEEE and otherprofessional organizations.

Manish Thakur completed his B.S. in ElectricalEngineering from REC-NAGPUR of India. From1996-1999, Mr. Thakur worked for ABB NetworkControl & Protection Business Area. From 1999-2001, Mr. Thakur completed his M.S. in ElectricalEngineering from the University of Manitoba. He iscurrently working for GE Power Management as anapplication engineer. He is a member of IEEE.

Authors: Eng. Eloi Rufato Junior –DISED/DES COPEL Eng. José Molinari Pinto–DISED/DES-COPEL Eng. Volney Pedroni – LACTEC Eng. Mário Klimkowski - LACTEC. Companies: COPEL-DIS- Compania Paranaense de Energia (Energy Company of Paraná). LACTEC. Country of Origin – BRAZIL Key Words: High Impedance, Falt, Short Circuit, Overcurrent

SUMMARY OF THE PROJECT: This project presents a summary of the tests realized with equipment, with the purpose of detecting the high impedance errors that occur in the electrical systems, and also the researches done to evaluate the status of the art of equipment and/or processes used to detect the high impedance errors of the electrical energy distribution system. During the tests, short circuit tests were made, with normal, aluminum and aluminum with steel soul cables, provoking with same fallen cables over several different kinds of soil, with a great variety of contact resistivity with the soil, also the cases of covered cables were tested, used in compact network, in both 13.8 kV and 34.5 kV,

proving that these currents of short circuit of cables fallen to the soil are very small. High impedance defects in distribution networks are those that occur with low values of short circuit currents, lower than the ones that are possible to be detected by the conventional protection equipment of overcurrent. They are usually caused by contact with branches or broken and fallen cables. The fallen cables that remain energized risk the lives of people that walk on the thoroughfares. They may generate electrical arches in the contact, causing fire if the material that they are resting on is inflammable. Besides the risk to others, these defects usually cause small harm located at the circuit, and the

Author: Name: José Molinari Pinto Address: Compania Paranaense de Energia- COPEL Rua: José Izidoro Biazzeto-158 Bloco C - Mossunguê Curitiba - Paraná - Brasil Fone: (00-55)(41) 331-2736 Fax: (00-55)(41) 331-3266 email: [email protected]

TECHNOLOGICAL SOLUTIONS TO DETECT HIGH IMPEDANCE FAULTS

José Molinari Pinto - Eloi Rufato Junior

COPEL - Companhia Paranaense de Energia.

companies have always neglected to these facts due to the low cost to repair them. That’s why investing in solutions for high impedance errors have never been justified by a conventional cost/benefit analysis. Even though they don’t cause harm to the circuit, they may generate expenses with reparations to others because of their consequences. That’s why all the searches for a solution to this problem have been motivated by the wish of assuring safety to every one and the prevention of fire.

1 Introduction

At a meeting in March/2000, there was consent about the importance of this research, because evaluations and tests that were done in commercial equipment of which it was expected to get as a product from the detection of high impedance errors not to present satisfactory results. In these tests there were analysis of short circuit involving also compact network cables, both in 13.8 kV and 34.5 kV, proving that these currents of short circuit of cables fallen to the soil are very small.

2 Existing Protection Practices

2.1 Existing Protection Against Short Circuits

The system used to identify short circuit events in the distribution networks by electrical energy companies, in general, is the one that uses relays that are sensitive to overcurrents. This kind of protection is efficient to defects such as network cables contact or internal short circuits in equipment that generate currents of high short circuit in the phases, that is, currents much higher than the maximum currents of charges accepted as normal to each circuit. The phase relays are adjusted to operate above the maximum charge value, turning off the circuit. Other causes of defects such as contact with branches, loss of current in isolators or equipment, for example, may generate defects with values smaller than the value of the maximum current of charge of the circuit. During some periods of the

day the charges may be small and it may occur that even adding the charge current with the defective current, the result won’t be higher than the maximum charge value accepted for the relay, making the identification by the relays of this defective current as phase overcurrent difficult. To detect this kind of event, another kind of relay is used; one that overlooks the resulting value of the vectorial totality of the phase currents that, in a circuit with no defects and with the equal values of three phase currents, would result in a null value. Unbalance in the phase charges due to the loss of current to the ground when a fall of a cable to the soil occurs, generates a result of the addition of the phase currents different from zero. The overlooking relay of this addition is called neutral or ground current relay, and it is adjusted to operate starting at a certain current value. That’s why the use of the term neutral overcurrent, associated to low defective currents, always refers to the values resultant from the addition of the phase currents that are above the value adjusted for action of the neutral relay, predetermined for each circuit. However, because of the kind of connection of network used, technical problems in the use of individual current transformers, and of normal unbalance of the phase charges, there will always be a result different from zero for the vectorial addition of the ground currents. The overlooking relay of this current should be adjusted to act above certain values, to guarantee the presence of a band of ground current below its adjustment as a normal condition for operation. Putting on a chart, (picture 1) we are able to see the field of action of these relays: The protection technology for 60-hertz overcurrents is the kind of protection that is predominant in the distribution systems. Cables that are fallen to the soil and that do not generate 60 hertz current enough to activate this kind of protection remain energized.

2.2 Limits to minimal protection adjustments First, we must highlight two important factors that contribute to the appearance of values still found in general companies nowadays for the insensitivity protection band based in relays of neutral for 60 Hz currents. These factors are:

2.2.1 System Architecture Connections of the triangle type at the exit of transformers of substations and of the same kind at the primary side of transformers of the network contribute to the unbalance of the charges at the primary phases of the transformer of the Substation, allowing lower values resulting between the phases and, therefore allowing smaller adjustment values on the sensitivity of the neutral relays, with better results on the monitoring of neutral overcurrent to identify errors in the network through sensors. Star Connections grounded at the exit of the substation and star grounded at the network transformers, are too dependent of the balance of charge because it is transferred to the primary of the transformer. At these networks, the balance that comes from charges in consumers with three-phased connection is worse because of the use of mono-phased patterns of network transformers and “MRT” (monofilar with underground way back) networks. These conditions require the use of higher values of relays shots, in a way that avoids interruptions due to the abnormal charge unbalances. The maximum adjustment values of the protections of neutral overcurrent recommended normally are:

• Until the maximum of 27A for RA’s 34.5kV

• Until the maximum of 25A for RA’s 13.8kV

This criterion is not always possible to be followed because of characteristics of the existing network and values of charges of the supplied consumers. In this case if it is necessary to use higher values than the ones listed above, they may be used as long as they obey the sensitivity and security criteria, defined by the relation: Equation 1 - Neutral adjustment current

5,11,0 TmínIccIneutroIc φ

≤≤×

Ic = charge current Ineutral = Neutral adjustment current Iccφ Tmin = phase-ground short circuit current minimum at the end of the circuit.

2.2.2 Limits of Equipment used Another factor is linked to the limits of the protection equipment that is used in the electrical systems. Working at the distribution systems of several companies we find a variety of models, from equipment controlled by electromechanical, electromechanical re-closers, static, and even the most modern relays, with micro processed controls. At the chart 1, there are some examples of equipment and their minimum current of neutral shots adjustment. Table 1 – Neutral adjustment current

Equipment Made by Type Adjustment Neutral (A)

Recloser Weco ESM 800 50 Recloser Brush PMR38 20 Recloser McGraw CXE 12,5 Recloser McGraw KFE 10 Recloser Cooper Nova 1 Recloser Whipp Panacea 1 Recloser Brush Microtrip

II 5

More recent equipment have the function SGF – Sensitive Ground Fault, which allows minor adjustments for the ground current with defined adjustment times. It is a function that contributes to reduce the currents of detection of the ground-phase short circuits of lower values, it is not considered as a final solution to this problem. Fig 1 – SEF Unit

SEF Ippn Ippf

t(s)

Amp

2.3 Operational Procedures in occurrences of high impedance defects

No alarm is activated at the substations or at the systems of automatic supervision when an event with fallen and energized cables that are in conditions that result in a high impedance defect incapable of activating the overcurrent protection occurs. The Distribution Operational Center is only aware of the occurrence when people who are near the locality of the defect notice this fact and reports it via telephone. The operator tries to identify the nature of the fallen cable through questions to the caller, but not always the answers bring certainty. If there are doubts, due to the risks, the solution that is used is the turning off of the feeder circuit or a switch near the area. This problem occurs due to the pattern of aerial networks. A fallen and energized cable with high impedance contact may be on for minutes or hours.

Copel Procedures to reduce the risks of falling cables The engineering areas of Copel are up-to-date with the technological evolution that occur in the area of the electrical sector and count with the support of Lactec for studies of research and tests of new protection equipment or new ways that present proposals to improve the sensitiveness of protection of the distribution network. The areas of equipment maintenance of Copel are constantly doing the preventive maintenance of the protection equipment installed in substations and feeder. The areas of network maintenance are constantly doing the preventive inspection and maintenance of the distribution networks aiming the unchanged conservation of the physical conditions of the distribution system.

2.4 Risks of new Network Patterns One of the patterns that eliminate the problem of cables to the soil, is the pattern of network called underground, where the energized cables are isolated and settled in ducts underground or are buried directly. However, due to the higher cost compared to the cost of aerial networks, they have an insignificant participation in the Brazilian distribution networks.

The Brazilian companies have been using an intermediary pattern named compact network, which is still an aerial network, however it presents characteristics that aim the reduction of interruptions because of contact with trees. It uses cables covered with full isolation to avoid turning off the circuit by contact with tree branches. It doesn’t have metallic external shield. The energy cables are hung by isolated fixing spacers that are hung permanently at a steel cable distended between the posts. This technique reduces the aerial space taken by the cables for better coexistence with the arborization and reduction of the trimming area for maintenance, besides allowing the energized cables to be suspended with less pulling traction due to the reduction of the space between the sustentation structure, that is, an average of 3 sustentations added between posts. Even though the probability of occurrence of cable fallen to the soil is smaller for this kind of network, cases that occur may behave similar to the one with aerial network with uncovered cable (without isolation) where the contact resistance is so high that it works as isolation, not generating current of loss able to be detected.

3 Position of the problem in Brazil

3.1 Concessionary Companies An exchange of experiences realized at the “Subcomite de Operacao e Manutencão do Comite de Distribuicao” – “CODI” (Subcommittee of Operation and Maintenance of the Distribution Committee), in October 1996, nowadays included in the new association of electrical energy companies ABRADEE, evaluated the craft condition about Detection of high impedance errors and of broken cables, with the 16 concessionaries that belonged to the committee on that date. (Ref. 6) Some highlighted points:

• Adoption of minimum values of

protection adjustment relays of ground (neutral), used by the companies, ranged according the following table:

Table 2 – Values of Adjustments of the Ground Current Values of Adjustment of

Neutral Relay Number of companies

1 Lower than 20A 3 2 > than 20A and <30A 2 3 > than 30A and <40A 1 4 > than 40A and <50A 1 5 Higher than 50A 1 6 Case with up to 80A 1 7 Between 20A and 50A 1 8 >10% of the maximum charge 1 9 20% of the liberation of charge

and 66% of the lower level of short circuit phase/ground minimum calculated with Zt=100 Ohms

1

10 Maximum 60A for the ground relay circuit breaker / Max 25 Amp for SE re-closer

1

11 Didn’t answer / didn’t clarify 3 This result shows that the problem does not happen only at Copel. The problem is due to the aerial pattern adopted all over Brazil. The factors that contribute to the differences of criteria among the companies are the same approached at the item 2.2, as follows:

• There is too much old equipment at the networks which minimal sensitivity adjustments for ground current are high;

• Limitations imposed by the architecture of the electrical system used.

3.2 Researches in Brazilian Institutions The research of new techniques or equipment aiming the detection of high impedance faults in Brazil has not been a priority, because big manufacturers and foreign institutions have been putting themselves into this subject during decades without a result 100% trustworthy and with low cost, so far. Among more recent researches in Brazil, the proposal of the "Sistema de Protecao de Faltas de Alta Impedancia" (Protection of High Impedance Faults System) – (Ref. 4) stands out, published at the V Seminario Tecnico de Protecao e Controle (5th Technical Seminar of Protection and Control) – Curitiba – Paraná, in 1995. It is based on the captation of

values of the electrical field by sensors installed bellow the high-tension network, overlooking the state of balance of the sequence tensions at the high-tension circuit. This device works with tension parameters instead of the current as it has traditionally happened. It is a solution that is bond to the necessity of a very big number of installed units to cover the entire network and it depends on a communication system from each location to the substation so that the protection device can work. Technologically it is a feasible solution, but it depends on better evaluation of the costs and precision of results.

4 Status of the Problem in other Countries

4.1 Attempts of solution with parameters of 60 Hz (ref.1)

The search for the solution to the problem of high impedance defects has passed through many attempts throughout decades of research, and among them we can mention the following:

• Relays with high sensitivity adjustment that fits the charge characteristics so that at the times of light charge they can present higher probability of identification of faults with cables fallen to the soil.

• Relays with the objective of

identification of the speed of change of the current values, assuming that the charges increase their values slower and that the increases due to short circuit increase their values faster.

• Relays with the objective of determining

the relations at the Fault Current (increased current) with the relations at the preceding charge current. It considers that these relations are uneven and that they are more balanced at normal currents.

• Other studies not listed may be found in

long bibliography of the international electrical sector.

Besides the techniques listed above, many other have been tried. However, the path aiming the use of 60 Hz parameters did not present solution perspectives up to the moment.

4.2 Attempts of solution with characteristics different from the 60 Hz parameters

Other techniques were researched, always trying to recognize a high impedance error “signature” at the network and through it decide when to interrupt the circuit. This signature would be offered by characteristics that are not associated to the 60 Hz characteristics, according to the fact that these parameters are less affected by the changes in charges and greatly affected by the changes in fault conditions. In this area, ref. 1 mentions some contracts done by EPRI with highly visible American companies in the scientific area for the development of researches, such as:

• With Power Technologies Inc. – Research of statistics technique to detect faults using sequence components of 60 Hz and its harmonics.

• With Hughes Aircraft – Research of detector based on the changes of the third harmonic caused by the fault.

• With Texas A&M University – Research of High Frequency Fault Detector at the range of 2 – 20 kHz basing on the fact that this high frequency signal increases due to the presence of electrical arch at the point of defect.

Even though these techniques may be useful to increase the number of cases detected they may also end up working in case of lack of faults because the parameter used as a reference occurred in the line for another reason, nullifying their results as a definitive solution.

4.3 More recent researches More recent jobs indicate that there isn’t a technology solidified as a definitive solution yet and the research focuses at the search for non-conventional options to identify the high impedance errors, such as the example mentioned at the item 3.2, in Brazil. Abroad, we highlight two examples: 4.3.1 High Impedance Fault Detection Using a

Morlet Transform Approach – (Ref.5)

Ref. 5, published by IEEE, describes researches with the use of algorithms using resources of the theory of the Wavelet Transformed, with discussion with foreign experts about the results that are considered promising. It is a new option to be considered. 4.3.2 Texas A&M University, EPRI, and

NASA Project Many research groups around the world have been trying to find a solution to the problem. One of these groups including Texas A&M University, EPRI, and NASA, researched the subject for more than a decade involving 40 concessionaries, more than 50 consultants with the objective of defining the nature of the high impedance defects for the electrical energy distribution systems and the characteristics of the algorithms necessary and products possible to be obtained at the detection of fallen cables. As a product of this project it is important to highlight a conclusion that is not homogenously considered in Brazil yet, that is that within the technologies available, not all the high impedance defects may be detected via electrical system, from the distance. Through this research, the characteristics of the equipment used to detect high impedance defects end up deviating from the traditional fast action protection equipment. Digital Feeder Monitors are constantly analyzing the presence of all the detectable disturbances and have artificial intelligence to differentiate between conductor arch on the soil and intact conductor arch and also between faults of similar phenomenon generated at the charges themselves or that occur because of opening and closing of switches. As an example of the recommendations of this research, a version of the GE's DFM (Digital Feeder Monitor) includes 9 sophisticated algorithms to detect high impedance associated to an analysis system of arch detection and another of charge monitoring. Its algorithms correlate events order, conditions of the pre-fault and pos-fault charge, presence of electrical arch, presence and duration of overcurrent, energy arches blocks, presence of harmonics, sudden or sustained energy changes, and the like (pic.2). In short it is a sophisticated device for monitoring of the presence of high impedance defects and surveillance of the energy quality. As an option it can do traditional protection functions.

Fig 2. Algorithm of Random Document information from the manufacturers estimates, based on the tests realized during the project, that with this device it is possible to detect 80 to 85% of the high impedance defects. At the others, it is not an error of the existing devices, because in these cases fault impedance has such high values that may be considered isolating contacts.

5 Evaluation of the Digital Feeder Monitors Through the Distribution Engineering, Copel has done short circuit tests in distribution networks simulating high impedance defects provoked and observed with GE’s DFM equipment – Digital Feeder Monitor, in Dec/95 at the didactic SE (CDTH) and in 04/28/96 at Santa Fé feeder at Fazenda Rio Grande SE. The results are at reports of the experiments. These tests subsidize some statements that are in this text, that not all the defects will be detected. In these tests, there were indications of parameters of the occurrence when the cable on the soil was of the source side of the network, but not when the cable on the soil was of the charge side. It was supposed that the error was due to the low potency of the didactic SS, which did not cause significant change of charge to shot triggers of algorithms. However this condition does not suit with high impedance detection equipment, because cables on the soil in distant or low charge ramifications will also be similar to this condition, and will not generate significant change of charge.

There were also cases when the detection took several minutes to identify the event, characteristic that is not pleasing to the purpose of safety of others.

6 Conclusions

• The detection of high impedance defects is one of the problems of the electrical system that does not have a solution yet.

• The current belief is that there is no technical solution to eliminate 100% of the risks of fallen and energized cables on the soil of public ways.

• Energized cables on the soil are a risk of all the aerial distribution systems.

• The companies are responsible for actions to control this risk.

7 References [1]IEEE: “Status of High Impedance Fault Detection N. 3” - IEEE, March 1985. [2]Perform Evaluation of High Impedance

Fault Detetion Algorithms basead on Stagged Fault Tests.

[3]Engro/GEC/Alsthom,:”Proteção de Circuitos para Faltas Intermitentes de Alta Impedância” – UERJ/Marte Eng. Ltda – V Seminário Técnico de Proteção e Controle, Brasil, 1995.

[4]CED/USP:” Sistema de Proteção de Faltas de Alta Impedância ” – V Seminário Técnico de Proteção e Controle, Brasil, 1995.

[5]Huang,Shyh Jier; Hsieh, Cheng-Tao: “High Impedance Fault Detetion Utilizing a Morlet Transform Approach”

[6]ABRADEE: “Troca de Experiências para avaliar o estado da arte sobre Detecção de Faltas de Alta Impedância e de Cabos Rompidos”, 1996.

[7]GE,”Manual do Equipamento DFM” – GEK 1000693.

[8]Rufato, Eloi Junior; “Relatório de Ensaios de Curto Circuito na SE Fazenda Rio Grande, Francisco Beltrão e Marmeleiro”. Copel - CNDS - Relatório 01/98, Curitiba, Brasil, 02/02/1998.

[9]Arroyo, Carlos A.: “Selecion de La Proteccion de Fallas a Tierra Con Altas Resistências en Las Redes de Media Tension”. I Seminário sobre Sistemas de Distribucion Y Utilizacion de Energia Electrica, Ica, Peru, octubre/1996.

(2-Cycle Multiples)

NormalEvent

SeriousEvent

True Output

Normal

(30 of 150)(3 of 5)

Threshold (125% of Average Value )

Average Value

150 Points30 Pts

Non-Consecutive Points

ANEXO 1

rr

ARC

138KV

FRG

ARC

34,5KV

3 2

4

FRG

10 5H

ARC

13,8

Agudos do Sul

CXE

ESV

ESV

ESV

ESV

75 kVA

5,0MVA

41 MVA

13,8kV

TA

Faz.Rio Grande

ESV

CO-11

X9

X4

X1

34,5KV

13,8KV

80ES

80ES

1,0 MVA

5,0 MVA

2,5 MVA

13,8kV

13,8kV

9

127/220 V

Fig 2- Diagram System

DISTRIBUTION FUSE LINKS: RESPONSIBLE FOR THE DAMAGE

OF EQUIPMENT AT SUBSTATIONS !!!

Eloi Rufato Junior - José Molinari Pinto

COPEL - Companhia Paranaense de Energia. Authors ELOI RUFATO JUNIOR – COPEL JOSÉ MOLINARI PINTO – COPEL Companies: COPEL – DIS – Companhia Paranaense de Energia (Energy Company of Paraná) COUNTRY OF ORIGIN – BRAZIL Key Words: Fuse links, Distribution, Short circuit, Substations Protection

SUMMARY: The main objective of this project is to present the result of studies and tests done on field and in laboratories, with real short circuits, with the purpose of improving the quality of distribution fuse links, specially the ones used to protect distribution transformers (15 kVA, 45 kVA, 75kVA, 225 kVA, etc), “H” types, in the classes of tensions of 13.8 kV and 34.5 kV. These protection devices, due to their low quality, were causing several kinds of problems at the distribution circuits, and also at the equipment installed at the Substations. At the distribution circuits, the main problems that occur are that, defects caused at the low tension circuits (127/220 Volts), such as colliding cables, tree branches at the network, collisions, storms,

etc, were causing bi-phased or three-phased short circuits at the feeder circuits of 13.8 kV and 34.5 kV that come from the Substations. These defects cause, many times, the touching of conductors at the side of High Tension, which is detected by the protection equipment installed at the Substations, work done by Re-closer or Circuit Breakers with overcurrent relays. With the occurrence of collisions at the side of high tension, the re-closers operate until they get to the blockage, letting the feeders off for a long time, until the defect that was generated at the low-tension side is found, in distribution transformers. Because of this, we have a considerable increase of the DEC of the feeders, and as a consequence, we are decreasing the quality of the energy supplied to the consumers. Another problem that the low quality of these breakers causes is the damage of the equipment installed at the Substations, mainly the Power Transformer of these Substations which are submitted to often and high levels of short circuits, which cause defects at the

Author: Name: Eloi Rufato Junior Address: Companhia Paranaense de Energia - COPEL Rua: José Izidoro Biazzeto – 158 Bloco C -Mossunguê Curitiba - Paraná - Brasil Fone: (00 55)(41) 331-2736 Fax: (00 55)(41) 331-3266 email: [email protected]

isolation and loosening of their magnetic part due to the electromechanical efforts to which they are submitted. With these studies and tests, it was possible to develop, with the manufacturers of fuse links, new material with the purpose of improving the quality of this equipment, decreasing the problems caused by them, at the Electrical Systems. 1 Introduction The distribution system of COPEL was conceived to meet the premises of low initial investment and reduced operational cost. Within this philosophy two distribution tensions were adopted, 13.8 and 34.5 kV the last also being the sub transmission one. The sub transmission lines of 34.5 kV that come from 138/34.5/13.8 kV transmission substations feed 34.5/13.8 kV substations (up to four substations) that attend small localities. The arrangement of these substations allows the installation of up to four 5 to 7 MVA transformers and four 13.8 kV feeders, with the entrance of two sub transmission circuits in 34.5 kV, besides the possibility of installation of capacitor and tension regulator banks. The kind of connection of the power transformers is grounded star at the 34.5 kV side and delta with to-and-fro ground connection at the 13.8 kV side. The powers of the transformers are 3.75 MVA, 4.2, 5 and 7 MVAs. There are cases of the past of 1.0 MVA, 1.5 MVA, 2.0 MVA and 2.5 MVA, and all of them have taps (34.5/33.75/33.0/32.25/31.5 kV) at the primary side.

1.1 Protection at Substations The 13.8 kV and 34.5 kV exits of the substations are protected by automatic re-closers, with hydraulic, electronic, electro-mechanic and micro processed control, having a tendency to the last ones, that make more precise and small adjustments possible, making the coordination and sensitiveness easier. These re-closers use vacuum, SF6 or isolating oil for interruption, but the last ones are not being used. PROTECTION AT DISTRIBUTION LINES With the purpose of improving the reliability and continuation of the supply of electrical energy,

automatic re-closers are being used in ramifications that come from the distribution lines in 34.5 kV and feeders of 13.8 kV, coordinated with fuse links and re-closers switches. 2 History Through occurrences survey, we verified the existence of problems related to the protections at the Francisco Beltrao SS and at other substations of the distribution system of Copel and problems related to collision of conducers at the Distribution Lines of 13.8 kV and 34.5 kV, caused by phase-ground defects throughout these lines and that would be developing to defects between phases, at the spans with problems, and this would cause the stress of the electrical system equipment involved, mainly the transformers of the Substations (ref.2) Defects at the Low Tension circuits 127/220 Volts, at the secondary of distribution transformers would be causing the burning of the fuse links of the distribution transformers and causing the action of re-closers and circuit breakers at the source substations. These new tests have the purpose of checking some new solutions proposed to solve the problems that are taking place, among them are: Tests of new fuse links used to protect distribution transformers, from 4 different manufacturers, three of them are imported and one is national, normal type. CABRAL test, developed by Copel, device to be installed at the ESV re-closers from Cooper with Resco’s control, with the purpose of making the blockage of the re-closer for short circuits of high currents possible.

2.1 Feeders where the tests took place (ref.1): • Jacaré feeder – 34.5 kV; • Enéas Marques circuit – 34.5 kV; • Pinheirinho Circuit – 13.8 kV, test at LT

127/220 V; • Ovetril Circuit – 13.8 kV, test at LT 127/220

V;

2.2 Locales where the tests took place: With the purpose of simulating the short circuits closer to reality, we used the locales where once there were defects at the system.

2.3 Preliminary script to realize the short circuits:

During the tests, the FBL SS and LD signals were monitored. 3 Planned Tests Four ( 4 ) kinds of fuse links of different manufacturers were tested, three of them special breakers (imported) and one national, normal type, named (normal, TYPE A, TYPE B, TYPE C) (ref.1)

3.1 Location 9 – Pinheirinho circuit 13.8 kV test BT transformer of 112.5 kV (see photos annex 3)

17. Test 17 Short-circuit phase-ground at phase A, breaker switches of the transformer with normal link of 5H. 18. Test 18 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE A” link of 5H. 19. Test 19 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE B” link of 5H. 20. Test 20 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE C” link of 5H. 21. Test 21 Short-circuit phase-phase between the phases A and B, breaker switches of the transformer with normal links.

3.2 Location 10 – Pinheirinho circuit 13.8 kV test BT transformer of 75 kVA.

26. Test 26 Short-circuit phase-ground at phase A, breaker switches of the transformer with normal link of 5H. 27. Test 27 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE A” link of 5H. 28. Test 28 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE B” link of 5H.

29. Test 29 Short-circuit phase-ground at phase A, cable on ground, breaker switches of the transformer with “TYPE C” link of 5H.

Picture 1 – Short-circuit at low-tension test – link 5H – normal. 4 Detailed studies of the incidents The Studies of Protection, in theory, were redone, involving the protection equipment of Substations, the Fuse links installed at the derivations and at the protection of Distribution Transformers, concluding that there is no lack of coordination of the protection equipment involved. Because of this, we decided to check if the problem mentioned above might also occur in other distribution lines and substations, which was verified, according to the answers received from the regional units.

4.1 Studies of the Problems Through the analysis of all the tests of short-circuit realized in field, it was possible to verify that there is the appearance of electrical arch between the breaker switches installed as a protection at the primary of the distribution transformers (13.8 kV and 34.5 kV), when short-circuits happen at the low tension side of these transformers (127/220 V), and when a normal fuse link is used, causing a bi-phase or three-phase short-circuit at the high tension side, which makes the protection equipment installed at the Substations work. Depending on the kind of circuit, this will occur until the circuit is blocked. This phenomenon is closely related to the Transitory Restoration Tension (TRT). At the breaker switch, there is the appearance of electrical arch inside its cartridge when the part fuses, in these conditions the breaker switch should extinguish the arch and keep the circuit open. When the short-circuit current is due to a high impedance fault, the heat emanated by the fusion of the link should decompose the inner part of the cartridge of the link, and then, a gas is formed and it de-ionizes the inner arch. The pressure developed inside the cartridge helps keep the characteristics of open circuit after the arch is extinguished. The pressure and the recomposing of the dielectrical are related to the size and duration of the defect current. If during the process of elimination of the short-circuit the tension of the arch breaks the dielectric, it will be re-started, causing the TRT (see annex – pictures). If the dielectric is enough to support the TRT, the re-start of the arch occurs and the short-circuit current will still pass through the breaker switch, even after the cartridge has fallen, until the current passes through the reference again. (see pic.2) At the picture 2, the “0AC” covering is the characteristic curve of the breaking tension of the dielectric during the restoration and the curve represents the resulting TRT. 5 Transitory Restoration Tension The distribution systems consist of inductances, capacitances and resistances, that’s why, through

the analysis of electrical circuits, we have: (ref 6,7 and 8)

• The current at a inductor doesn’t change instantly because the tension will be endless, according to the expression bellow:

Equation 1

dtdiLtV

T

×= ∫1

0

)(

If dt1 ⇒ zero ∴ V (t1) ⇒ ∞

• The tension at a capacitor doesn’t change instantly, because the current will be endless, according to the expression bellow:

Equation 2

∫ ××=1

0

)()(t

dttvCti

Analyzing the TRT phenomenon that appears during the interruption of faults, when the breaker switch opens, the short-circuit current will be interrupted and the transitory tension appears between the two terminals of the breaker switches. The equation for the fault current will be given by the equation 3:

Equation 3

( )[ ]θθ sensen −+×= wtwLEicc

Picture 2 – Example of the testing TRT with covering at two parameters that meet the demands for the test. The short-circuit current has a continuous component that is proportional to the sine θ, as shown at picture 3. The transitory restoration tension can be written according to equation 4:

Equation 4

⎟⎠

⎞⎜⎝

⎛⎟⎠

⎞⎜⎝

⎛=⎟

⎠

⎞⎜⎝

⎛dtdxZ

dtdicc

dtdVtrt

Through mathematic solution, we get to the transitory restoration tension:

Equation 5

( )⎥⎦

⎤⎢⎣

⎡−−××= θθ wt

LCtEVtrt coscoscos

The effect of the resistance at the circuit is absorbing the transitory component of the tension, diminishing the TRT during the opening of the breaker switch when a short-circuit at the low tension occurs. Also at equation 5, the values of L and C are from the distribution transformers. So, to determine the TRT value at the breaker switches, when there are short-circuits at the low tension side, we should consider the transformer parameters (L and C) and the resistance value of the low tension conductors. Besides the characteristics of the circuits where the defects occur, we also have to consider the quality of the breakers so that there is a complete extinction of the electrical arch during a short-circuit at the low tension. A bad quality fuse link will take too long to extinguish the arch, allowing the electrical arch to re-start, not interrupting the defects that occur at low tension.

Picture 3 – Icc Components 6 Analysis of the results We have done more than 60 real short-circuit tests, in field, at different locations of the system and using 45 kVA, 75 kVA and 112.5 kVA transformers and using 4 kinds of fuse links from different manufacturers, according to item 3. Through filming and observance of the short-circuits, and also analysis of all the oscillographic values at digital ocillograph, Hioki type, (see oscillographies – annex), we could note and define what kind of link does not generate TRT during a short-circuit at the low-tension network. Consequently it doesn’t cause the electrical arch at the breaker switch for a long time, not ionizing the air around it and not causing short-circuits at the High Tension side, avoiding the repetitive action of the protections at Substations, and also reducing the stress and the damage of the equipment installed at Substations (specially the power transformers that suffer more efforts).

Picture 4 – Oscillography of the time of link arch of 5H. Comparing the oscillographic values to the observance and analysis of the films, during all the tests, with conventional links and special fuse links (ref.9), we were able to define a maximum time of arch so that its re-start won’t occur at the breaker switch. This maximum time of arch was only possible with fuse links of imported technology. At that time this technology was not available in the country, consequently the costs are much higher than the ones of a conventional link. These fuse links use vulcanized fiber tubes, which do not present the carbonization that the common paper and fenolite resine links present as they burn during the short-circuit. The vulcanized fiber contributes to the formation of gas and water vapor during the short-circuit, which is fundamental to avoid the re-start of the arch during the action of the fuse link for low value circuits.

Picture 5 – Oscillography of the time of fuse link arch 3H. 7 Conclusions

• Through the analysis and verifications of

all the oscillographies and films of all short circuit tests, we could rate the parameter found at the distribution systems better;

• With the purpose of avoiding the

increase of stress and damage of the power transformers installed at substations, we reduced the maximum number of operations for blockage of the protection equipment installed at substations;

• We have been doing a Pilot Project at the 5 COPEL’s regional units, in 13.8 kV and 34.5 kV circuits, with special fuse links, for the protection of distribution transformers, from a national manufacturer that fulfilled the requisites specified after the short-circuit tests in field;

• This Pilot Project has the purpose of verifying if the solution proposed is

economically viable once these breakers cost 6 to 10 times more than the conventional ones;

• Through these tests we could change the Technical Specification of Fuse links of Copel, in reference to the tests of type and receipt.

8 References [1] Rufato, Eloi Junior; “Relatórios de Ensaios de Curto Circuito na SE de Francisco Beltrao e Marmeleiro”. Copel – CNDS – Relatório 01/98, Curitiba, Brasil, 02/02/ [2] Rufato, Eloi Junior; “Encontro Técnico de Transformadores de Potencia”. Instituto de Engenharia do Paraná, Curitiba, Brasil, 1997. [3] ABNT, “NBR 7282/89 – Dispositivos Fusíveis tipo Expulsão” – Brasil, 1989. [4] COPEL, “Norma Técnica da Copel – Elo Fusível de Distribuição – 810032”, 12/98. [5] COPEL, “Norma Técnica da Copel – Elo Fusível – NTC 813810/29”, 12/98. [6] Hermeto, Eric e outros “Desligamentos Instantâneos em Alimentadores e sua relação com a Interrupção de Faltas Secundarias”, Brasil, 1976. [7] Harner, Robert “Distribution Systems Recovery Voltage Characteristics: I – Transformers Secundary – Fault Recovery Voltage Investigation”, IEEE Trans., 02/1968. [8] Peterson, Harold A, “Transients in Power Systems” – John Wile and Sons, New York, 1951. [9] Costa, Eleilson S e outros, “Elos Fusíveis de distribuição: Especificação, desenvolvimento e ensaios”, Eletricidade Moderna, 01/2000. DADOS DA EMPRESA: Destinatário: Companhia Paranaense de Energia - COPEL Endereço: Rua José Izidoro Biazetto, 158 – Bloco C – Mossungue – Curitiba-Paraná-Brasil Código Postal: 81200-240. Telefone: (55-41) 331-2736. Telefax: (55-41) 331-3266. Email: - [email protected] - [email protected]

Annex 1

Picture 6 – Oscillographies of the Arch Time at Fuse links

Picture 7 – Short-Circuit at Low-Tension with Normal 5H Fuse link

Picture 8 – Short-Circuit at Low-Tension with Special Fuse link

Picture 9 – Short-Circuit at Low-Tension Normal Fuse link

Picture 10 – Short-Circuit at Low-Tension with Special Fuse link.

Analysis of Hidden Failures of Protection Schemes in Large Interconnected Power Systems

J. De La Ree David C. Elizondo (Senior Member) (Student Member)

Center for Power Engineering

The Bradley Department of Electrical Engineering Virginia Tech, Blacksburg, VA 24061

Abstract: Hidden failures in protection systems have been identified as key contributors in the cascading of power system wide-area disturbances. This paper describes a methodology to evaluate the hidden failure’s effects based on regions of vulnerability and areas of consequence in protection systems. The mechanisms of the hidden failures as well as the development of the regions of vulnerability from the hidden failures are shown. A number of the regions of vulnerability in a sample 179 test system are represented as physical areas. Two scenarios evaluate the consequences of the unwanted disconnections of the transmission lines caused by hidden failures. The development of the index of severity, which combines the magnitude of the regions of vulnerability and the consequences of unwanted disconnections caused by hidden failures is proposed. This index would identify the critical protection systems, whose unwanted operations would result in a significant loss of the power system integrity and/or a large cascading event. Keywords: Hidden Failures, Blackouts, Wide-Area Disturbances, Protection Systems, Regions of Vulnerability, Index of Severity, Cascading Events.

I. INTRODUCTION

It is well known that protection systems play an important role in the difficult task of delivering electric power under all power system scenarios. Failures of these protection systems may affect considerably the power system operation and its capacity to deliver power. The purpose of this paper is to analyze the effect of a specific type of failure within the power system protection, the hidden failures. Hidden failures are defined as “a permanent defect that will cause a relay or a relay system to incorrectly and inappropriately remove a circuit element(s) as a direct consequence of another switching event” [1]. From the definition, it can be emphasized that hidden failures bring as a result the disconnection of a circuit element. Then, a "failure to operate" will not be considered a hidden failure due to the fact that some other protection systems will react and finally eliminate the abnormal condition. Power Systems are biased towards dependability, and, sooner or later, all

"events" should be cleared by the existing protection systems. It will be convenient at this time to include a list of elements necessary to have a hidden failure: • A Protection Element Functionality Defect, PEFD. In

general, a PEFD takes place when the protection devices are unable to perform their designed and expected task. They may take the form of hardware failures, outdated settings, human errors and/or negligence.

• The second element is the logic involved around the PEFD. This logic will determine if the failure will or will not remain hidden within the protection scheme until another event uncovers the defect.

Figure 1 shows the schematic of the control circuit of a distance relay. It is easy to identify the three-zones of protection and the corresponding coordination timers for zone 2 and zone 3. For this device, a PEFD in the form of continuously closed contacts T2 or T3 will result in a hidden failure of this protection system. Under this condition, it is only necessary to have a fault with an impedance smaller than zone 2 or zone 3 setting to render a trip decision. The device will loose coordination with down-stream protective devices and the transmission line will be unnecessarily disconnected for an external fault. A PEFD of Z1 contact will result in no hidden failure. The logic is such that this PEFD will manifest itself as soon as the system is place in service. This indeed is a relay miss-operation, but not a hidden failure.

Z2

T2

Trip Coil52a

Z 3

T 3

Timer 3Timer 2

Z 1

NHF

HF

Figure 1: Logic Schematic for Z2 and Z3 Distance Relays

Prior research has resulted in a catalogue of hidden failure Modes of the most common protection schemes used in transmission lines, transformers, buses and generators [2]. For each of these failure modes it is possible to define a region of vulnerability as the physical area in which the occurrence of an event will trigger the unwanted disconnection due to the hidden failure. This region will be defined by a particular location, dimension, and magnitude depending on the hidden failure mo de of the protection system being analyzed and the protection system location within the power system. The size of the region of vulnerability and its corresponding consequence will provide us with the necessary elements to create an index of severity to a given hidden failure within the power system. II. REGION OF VULNERABILITY DEVELOPMENT – QUALITATIVE APPROACH In order to explain how the regions of vulnerability are developed based on the hidden failure modes, an example related to a protection system for a transmission line is presented next. Figure 2 shows the one-line diagram of the directional comparison blocking scheme, in which the region of vulnerability will be developed. Figure 3 shows the logic schematic for the directional comparison blocking with the hidden failure mode “FDA - cannot pick up.” The nomenclature for Figure 2 and Figure 3 is included after Figure 3.

CBA CBB

DA

DB

FDA

FDB

FA FBFAB

Figure 2: One-line Diagram for the Directional Comparison Blocking.

DA

RA

Trip CoilA

52aA

Start BlockingSignal, ActivatesTransmitter

FDAStopsBlockingSignal

TA

DB

RB

Trip CoilB

52aB

Start BlockingSignal, ActivatesTransmitter

FDBStopsBlockingSignal

TB

1

Figure 3: Logic Schematic for the Directional Comparison Blocking, A and

B line sides.

Nomenclature and Protection Scheme Elements for Figure 3 and Figure 4*.

Element Description Function FDA Directional fault

detector Detect external faults and start transmitter

DA Directional tripping relay

Detects faults and stops transmitter

RA Receiver Receives remote/local signal

TA Transmitter

Transmits communication signal

CBA Circuit Breaker Disconnect one line side 52aA CB auxiliary contacts Monitor breaker status FA Fault behind bus A FB Fault behind bus B FAB Fault between A and B * The same elements apply also for line side B.

Table 1 shows the sequence of events for an unwanted disconnection caused by a hidden failure. The circuit breaker that is opened incorrectly is CBB. The sequence of events described in Table 1 is related to the PEFD in the logic schematic shown in Figure 3, “1”, and the electrical fault located at “FA. ” See Figure 2. Table 1: Sequence of Events for an Unwanted Disconnection Caused by a

Hidden Failure.

Case for FA

PEFD:FDA can not pick up Operated Relays

Consequences at side A logic schematic

Consequences at side B logic schematic

Final Result

FDA FDA cannot close, T A is not activated, no BSA. RA remains closed.

No BSA, so RB remains closed.

DB DB closes its contacts.

Unwanted trip at B side, since BSA was not received, RB remained closed and DB was activated.

Nomenclature: BSA: Blocking signal of side A

The first event in the unwanted disconnection caused by a hidden failure is the PEFD that was shown in Figure 3, “1”. The FDA relay has its contacts normally open, and the PEFD will prevent of any change in contacts status. Therefore, the relay contacts will remain open regardless of the power system conditions. Since the FDA is a distance relay, the relay reacts to the apparent impedance seen from its terminals. The power system condition required to operate this relay is an electrical fault within the relay protective zone. In the directional comparison blocking scheme, the receiver relays RA and RB take input signals from both fault locator relays, FDA and FDB. This means that the receiver relays RA and RB change its contacts status, from normally-closed to open, if the blocking signal coming from either FDA or FDB is received. The PEFD on FDA makes the relay unable to react to the power system conditions for which it was designed. In other words, the relay cannot close its contacts

even though there is a power system condition that is observable to it. This FDA inability to close its contacts has the consequence of not sending a blocking signal to the receiver relays. Since the receiver relays (RA and RB) react to the blocking signal coming from either FDA or FDB, there is a case in which the inability of a fault detector to close its contacts leaves a flaw in the logic—a vulnerable logic. The vulnerable logic that results from the PEFD in the FDA relay does not reveal itself until another condition occurs in the power system. This other condition has particular characteristics and becomes observable to one of the relays in the directional comparison blocking scheme. This other condition produces a relay reaction and because of this relay operation only, an unwanted disconnection caused by a hidden failure may take place. Under this vulnerable logic scenario, in which the receiver RB can only receive input signals from FDB (FDA has a PEFD), the occurrence of “FA” makes the DB relay operate resulting in an unwanted disconnection caused by a hidden failure; CBB is opened incorrectly. The second event in the unwanted disconnection caused by a hidden failure is the electrical fault “FA.” See Figure 2 and Figure 3. Figure 4 shows the representation of the region of vulnerability for the hidden failure mode “FDA - cannot pick up.” The relay DB defines the region of vulnerability location, dimension, and magnitude. The region of vulnerability is located in the reverse part with respect to the protection system element with the hidden failure mode, FDA. Since DB is a distance relay, the region of vulnerability will be expressed in ohms. The magnitude of the region of vulnerability is calculated based on the DB relay settings, which is the relay that reacts to the power system “other condition FA,” as well as other factors related to the network topology, which will be presented in the next section.

CBA CBB

DA

DB

FDA

FDB

FA FBFABCBA CBB

DA

DB

FDA

FDB

FAFA FBFABFAB

Figure 4: Region of Vulnerability for the Hidden Failure Mode “FDA

cannot pick up.”

III. REGIONS OF VULNERABILITY DEVELOPMENT: A QUANTITATIVE APPROACH ON A TEST SYSTEM

This section presents the quantitative evaluation of a number of the regions of vulnerability for two protection schemes for transmission lines. The representation of the regions of vulnerability in the power system is shown in a sample 179

test system, which serves as framework for this development. The pilot protection scheme is the directional comparison blocking, described in a previous section. The non pilot protection scheme is a Z2 distance relay. This analysis applies for both the Directional Comparison Blocking and the Zone 2 distance relay due the similarities that these scheme present between each other. Previous research presented the factors affecting the evaluation of the region of vulnerability as: (1) the relay’s settings, DB and DA, (2) the number of transmission lines or transformers connected to each bus of the protected line, which we will call bus density, and (3) the ratio of the impedance of the protected line with respect to the impedance of the transmission lines or transformers connected to each bus of the protected line. For the quantitative evaluation of the regions of vulnerability of the line A-B shown in Figure 4, the calculation of the settings of relays DB and DA is required. For this development, the directional relays DB and DA are set to overreach the protected line by a factor of kD equals to 0.2 [3]. Then, for each of the transmission lines in the sample system, the setting of the relay DB and DA is 1.2 times the impedance of the respective line. INFEED EFFECT ON THE RELAY SETTINGS AND THE REGION OF VULNERABILITY EVALUATION. Distance relays react to the apparent impedance resulting from the voltage and current inputs. The apparent impedance, however, may be different from the actual impedance between the relay location to the fault because of current infeed (or outfeed) at some point between the relay and the fault. Current infeed has the effect of causing a distance relay to underreach for all faults past the point where the infeed of current occurs. [4]. Since the current infeed causes to a distance relay to underreach, then the security margin of the relay may be reduced. The objective of the Directional relay (DB and DA) is to detect a fault anywhere on the transmission line between the buses A and B (see Figure 4). In order to meet this objective, the relay is set at 120% of the impedance of the line, and the 20% is used as a security margin. This margin may be reduced due to the infeed current and that reduction will depend on the infeed/relay current ratio and the fault location. In cases in which the infeed current reduces considerably the security margin, the setting of the DB relay may be increased to compensate the loss of security margin. In order to evaluate the regions of vulnerability of this specific test system, and to obtain comparable results, the setting of the DB relay has been chosen as 120 % of the impedance of the line. We have neglected the possible loss of security in the protection system design and we have

obtained a consistent setting method for the relays. This enables us to obtain a fair comparison of the regions of vulnerability in the system. Infeed currents are not always present in the power system. The worst case scenario is the absence of infeed currents, since the relay will not underreach. Then, we have ignored the effect of current infeed in the evaluation of the regions of vulnerability. REPRESENTATION OF THE REGIONS OF VULNERABILITY ON THE POWER SYSTEM: THE PHYSICAL AREAS Figure 5 shows a representation of the region of vulnerability of the directional comparison blocking scheme applied to the transmission line number 73. The single line diagram of the sample power system is included in the Appendix and this portion of the system is located in the B – 1 coordinates (see appendix). The DB relay is shown in the figure with the square and it is set to 120 % the impedance of the line. The power system elements in which the region of vulnerability is represented are marked with a circle, as shown in the figure.

23

21

132

133

134 69

93

94

77

78

7980

92

73

90

DB

orZ2

Rel

ay

Figure 5: Representation of the Region of Vulnerability, Case Line 73

Considering the factors previously mentioned the region of vulnerability is calculated in Table 2. In order to calculate the physical areas, i.e., the regions of vulnerability in Km, we have used typical values of ohms per unit of length for the transmission line voltage [5]. Furthermore, we have assumed that all lines connected to the bus have the same ohms per unit of length.

Table 2: Magnitude of the Region of Vulnerability, Case Line 73.

Relay Setting Z (73)*1.2 = 0.0135

Bus Density 2

Region of Vulnerability Per Unit Z(73)*.2 = 0.0022

Region of Vulnerability Ohms 5.60 Ohms

Region of Vulnerability Km 17.254 Km

Figure 6 shows the representation of the region of vulnerability of the directional comparison blocking scheme applied to the transmission line number 81. This portion of the system is located in the B – 1 coordinates (see appendix). The DB relay is shown in the figure with the square and is set to 120 % the impedance of the line. The power system elements in which the region of vulnerability is represented are marked with a circle, as shown in the figure. As we can see, in this case the power system elements include transmission lines as well as transformers.

93

95

17881 82 83

177

G

96

97

DB

or

Z2 R

e lay

Figure 6 Representation of the Region of Vulnerability, Case Line 81

The region of vulnerability is calculated in Table 3.


Relay Setting Z (81)*1.2 = 0.0812

Bus Density 5

Region of Vulnerability Per Unit Z(81))*.2 *4 = 0.0542

Region of Vulnerability Ohms

21.66

Region of Vulnerability Km *

22.19 Km

* In this case a problem exists with the representation of the physical area when transformers are present. The region of vulnerability expressed in Km has been calculated for the transmission lines only. This will be explained later in the section. Figure 7 shows the representation of the region of vulnerability of the directional comparison blocking scheme applied to the transmission line number 65. This portion of the system is located in the A – 1 coordinates (see appendix). The DB relay is shown in the figure with the square and is set to 120 % the impedance of the line. The power system elements in which the region of vulnerability is represented are marked with a circle, as shown in the figure. Once again, the power system elements include transmission lines as well as transformers.

9

55 5611

51

10

54

126566 67 175

G

52

13

53

63 64

1470 71 68 173 17461 62

15

25DB or

Z2 Relay

Figure 7 Representation of the Region of Vulnerability, Case Line 65

The region of vulnerability is calculated in Table 4.


Relay Setting Z (65)*1.2 = 0.0170

Bus Density 9

Region of Vulnerability Per Unit

Z(65))*.2 *8 = 0.0226

Region of Vulnerability Ohms

56.53

Region of Vulnerability Km *

152.21 Km

* A similar problem exists with the presence of transformers. The region of vulnerability expressed in Km has been calculated for the transmission lines only. Section I mentioned that is possible to define a region of vulnerability as the physical area in which the occurrence of an event will trigger the unwanted disconnection due to the hidden failure. Some clarifications regarding the physical area are required when these theories are applied in a sample power system. For Figure 5, the evaluation of the region of vulnerability as a physical area is a straight forward procedure, since we have one transmission line and we know its nominal voltage, the ohms per unit of length, etc. To obtain the physical area of the region of vulnerability of the Figure 7 is a more complex task. First, if we consider the transmission lines only, there are a number of different estimates of the ohms per unit of length derived from the independent designs of each of the transmission lines. The second point is related to the transformers. The inclusion of the transformers in the calculation of the region of vulnerability is conceptually correct. If there is a fault inside the transformer within the reach of the DB or Z2, the transformer breakers should be opened by its differential protection and the line with the hidden failure incorrectly would be disconnected. The problem lays, however, in the representation of the physical area. Unlike transmission lines, the representation of the region of vulnerability on transformers will be confined in

the substation switchyard. Practically, the physical area may be the whole transformer bank or a part of the transformer windings within the transformer bank. That is why in the previous calculations we have excluded the transformers in the representation of the region of vulnerability expressed in kilometers. We have, however, left the transformers in the calculation of the region of vulnerability expressed in ohms. It is important to recall that this analysis is based on per unit values of the transmission line and transformer impedances. The ohms per unit of length of a given transmission line depends of many factors which are related to the transmission line design [6]. Generally, the ohms per unit of length of a transmission line tend to decrease with increments in the nominal voltage of the transmission line. However, the range of this decrement is quite narrow [5]. In power system analysis a normal practice is to set SBASE to a constant value of 100 MVA. Given this SBASE of 100 MVA, the ZBASE varies with the square of the base voltage. Since the per unit impedances are obtained by dividing the ohmic impedance by ZBASE , then there is a tendency to obtain smaller per unit impedance values for bigger transmission line nominal voltages. [4] states that for a given line length, the per unit impedance varies much more with the nominal voltage of the line than the ohmic impedance. In order to appreciate the results presented in the previous tables in the appropriate perspective we should analyze the regions of vulnerability expressed in Ohms . We can see that the magnitude of the region of vulnerability for the 500 kV line in Figure 7 is 56.53 Ohms and the magnitude of the region of vulnerability for the 200 kV line shown in Figure 6 is 21.66 Ohms. However, these results are not consistent when we analyze the regions of vulnerability expressed in per unit values. We can see that the magnitude of the region of vulnerability for the 500 kV line in Figure 7 is 0.0226 pu and the magnitude of the region of vulnerability for the 200 kV lines shown in Figure 6 is .0542 m per unit. This is an obvious effect of the different voltage base values used to calculate the ZBASE of the transmission lines.

IV. REGIONS OF VULNERABILITY: CONSEQUENCE EVALUATION

The impact or consequences of the unwanted disconnection of power system elements caused by hidden failures in the power system is critical and must be considered in the methodology. The magnitude of the regions of vulnerability and the consequences of the unwanted disconnections caused by hidden failures are combined to determine an index of severity. If the magnitude of a region of vulnerability of a transmission line end is bigger than the other end, then the exposure to the occurrence of other conditions—such as electrical faults—is greater, and then the region of vulnerability magnitude must be taken into account. The consequences of unwanted disconnections caused by hidden

failures in the power system must also be considered, since the unwanted trip of a key transmission line in the backbone of the system transmission network would be more important than the unwanted trip of a radial transmission line. Applying these ideas to the portion of the system depicted in Figure 6 and Figure 7 we have simulated a double contingency in the sample 179 test system. Two transmission lines are taken out of the power system and an Electric Transient Stability program was used to simulate the effects in the system. SCENARIO I The contingency simulates a permanent three-phase fault on the transmission line 82. We have assumed that there is a hidden failure in the relay shown in Figure 8 and that the three phase fault takes place inside the region of vulnerability. For the transmission line 82, the relays will detect that there is an internal fault and will open both ends; this is a correct operation. Because of the hidden failure, and the fact that the fault is located inside the region of vulnerability, the circuit breaker associated with the relay shown will also be opened. This is an unwanted disconnection caused by the hidden failure.

93

95

17881 82 83

177

G

96

97

DB

orZ2

Rel

ay

Figure 8: Contingency Description: A Fault on Transmission Line 82, Lines

81 and 82 are Disconnected of the Power System.

The relative angles of the nearby generators are plotted in Figure 9. As we can see in the figure, the angles of the machines present some oscillations and they tend to return to a new steady state. We could say that the consequence of this double contingency in the power system is not severe. However, further analysis must be performed to evaluate possible overloads, undervoltages or other undesirable conditions.

0 1 2 3 4 5 6 7 8 9 10Time in Seconds

100

50

0

-50Gen

erat

or R

elat

ive

Angl

e, D

egre

es

0 1 2 3 4 5 6 7 8 9 10Time in Seconds

100

50

0

-50Gen

erat

or R

elat

ive

Angl

e, D

egre

es

Figure 9 Generator’s Relative Angles for the Double contingency:

Transmission Lines 81 and 82 are taken out of service.

SCENARIO II The contingency simulates a permanent three-phase fault on the transmission line 66. The same assumptions considered in Scenario I apply in this case. We have disconnected two lines out of the power system, in this case line 66, and the circuit breaker associated to the relay shown in Figure 10.

9

55 5611

51

10

54

126566 67 175

G

52

13

53

63 64

1470 71 68 173 17461 62

15

25DB or

Z2 Relay

Figure 10: Contingency Description: A Fault on Transmission Line 65,

Lines 65 and 66 are Disconnected of the Power System.

The relative angles of the nearby generators are plotted in Figure 11. As we can see in the figure, the angles of a number of machines present severe accelerations, and a system separation is evident. We could say that the consequence of this double contingency in the power system is quite severe.

0 1 2 3 4 5 6 7Time in Seconds

500

0

-500

-1000

-1500

-2000Gen

erat

or R

elat

ive

Angl

e, D

egre

es

0 1 2 3 4 5 6 7Time in Seconds

500

0

-500

-1000

-1500

-2000Gen

erat

or R

elat

ive

Angl

e, D

egre

es

Figure 11 Generator’s Relative Angles for the Double Contingency:

Transmission Lines 81 and 82 are taken out of service.

V. CONCLUSIONS

This paper has presented a methodology to evaluate the hidden failure’s effects based on regions of vulnerability and areas of consequence in protection systems. The mechanisms of the hidden failures were presented and applied to a three zone distance relay. The development of the regions of vulnerability from the hidden failures was derived for the directional comparison blocking scheme. The region of vulnerability was represented as a physical area in the transmission line used in the example. The directional comparison blocking scheme and the zone 2 relay were applied to a number of transmission lines in the sample 179 test system. The regions of vulnerability were calculated using the values of the power system elements in per unit and in ohms, and they were represented in the power system as physical areas. For the representation of the physical areas we used typical values of the ohms per unit of length for the transmission lines. The role of the transmission lines and transformers in the representation of the region of vulnerability was highlighted. We realized that in order to appreciate the physical areas, the regions of vulnerability should be expressed in ohms. Two scenarios were presented in order to evaluate the consequences of the unwanted dis connections of the transmission lines caused by hidden failures. Two transmission lines were taken out of service as a consequence of a permanent three phase fault inside the region of vulnerability. The consequences of the unwanted disconnections in the power system were examined, and together with the magnitude of the regions of vulnerability (represented as physical areas) will be used to create an index of severity. This index would identify the critical protection systems, whose unwanted operations would disconnect power system elements affecting considerably the power system integrity and its capacity to deliver energy.

VI. ACKNOWLEDGMENTS

The second author of this paper appreciates the support of the National Council for Science and Technology, CONACYT , Mexico.

VII. REFERENCES

1. Surachet Tamronglak, “Analysis of Power System Disturbances due to Relay Hidden Failures”, Ph.D. Dissertation, Virginia Polytechnic and State University, Blacksburg, Virginia, March 1994.

2. David C. Elizondo, “Hidden Failures in Protection Systems and its impact on wide-area disturbances", MSEE Thesis, Virginia Polytechnic and State University, Blacksburg, Virginia, April 200.

3. Stanley H. Horowitz and Arun G. Phadke “Power System Relaying,” second edition. Research Studies Press Ltd., England, and John Wiley and Sons Inc. New York, 1995.

4. IEEE Guide for Protective Relay Applications to Transmission Lines. IEEE Std C37.113-1999. Power System Relaying Committee. September 1999

5. Prabha Kundur, “Power System Stability and Control,” Electric Power Research Institute, Power System Engineering Series. McGraw-Hill Inc. 1994

6. William D. Stevenson Jr. Elements of Power System Analysis, Fourth Edition, McGraw-Hill, 1982.

VIII. BIOGRAPHIES

Jaime De La Ree (Senior Member) received the BSEE degree with distinction from the Instituto Tecnologico y de Estudios Superiores de Monterrey in 1980, and the MS and The Ph.D. degrees from the university of Pittsburgh, Pittsburgh, Pennsylvania in 1981 and 1984 respectively. In 1984, he joined the faculty of Virginia Polytechnic Institute and State University, Blacksburg, Virginia, Where is an associate professor. His research interest in the areas of power systems and rotating machinery. He is a member of the IEEE, PES, IAS, as well as Tau Beta Pi and Eta Kappa Nu Honorary Societies. David C. Elizondo (Student Member) was born in Monterrey, Nuevo Leon Mexico, in 1972. He received the BSEE degree from the Instituto Tecnologico y de Estudios Superiores de Monterrey in 1994. His professional experience took place from 1994 to 1998. His job was related to the electrical business in some Mexican cities, being involved in the Industry as well as Utilities. Mr. Elizondo completed his MSEE in April 2000 and he is currently enrolled in the Ph.D. program at Virginia Tech.

IX. APPENDIX*

* The sample 179 test system was modified to 127 buses.

G 126

203125

127

42

39

43

4849

50

40

47

12441

44 45 46202

118

119

G

37 38

3231201

123

117

116

197 198

23

120

27

17 1816

11525 114

36 200

113G

112 22

33 3419

30

110

111

35

196195

20 21109

108

194193

107 106

10514 15

166

104

199

24

122

26

121

28 29

103

20482

141142

143

139

138

14078

80205

81G

109

108

111 79

110

190

76

G

75

11274

113

185186 187

77

135

136

137

192

73114

191

72G

71

98

97107106

68

69

103

104

105

70

64

63

189

G67

188

G65

100 66

101

99

57

170

8G

7

9

58

55 5611

51

10

54

1265 66 67175

G

52

13

53

63 64

14

70 71

23

21

6816

17317461 62

171176

172

1759 60

18G19 20

15

132

133

134 69

93

95178

81 82 83

177

G

96

9785 87

8984100 99

9886 8890

101

94

77

78

7980

92

73

90

181180

G102 179

89G 91

92

94

87

84

184

88G

9695 93 85

74

75

86

76

182 183

83

130

G

1

165

2

3

13

163

G

5164

6

4

12

G

26168

24

131

25

162

G62

577

144

61

5

8

58

6

59

9 10

154

G60

145146

11

51

157158

52

153

16053

159

G

50

254

161

55 155

56

G

147

22

208

41

40

12037

3938

207

G

125124

122

12335

36129121

42

127126

118119

211

169

G

28

72

27

29

117

30

34

115

116

210

209

G

32

31

151

348

15649

G

102

45

447

152

G46

206

G

144

167

G

43

150149

128

33

91

148

1 2

A

B

C

D

!"#$

!""

%#

%"&

! "dt

)t(idL)t(i•Rt•cos•V !"#!$

! "!"#$ t•cos•I)t(i

!"!!

jj e•II and e•VV!!

! " !"!"#

jj e•I L• jRe•V

L•jR

e•Ve•I

jj

!"

#

!"

!"#$#%

#&#

#&#

$#'

$#'

"#$

%"#$

%

%"#"&#$

&###

%"#"&#$

'"&#$

&#'$''&#

!"

'$&'$!

!

!"##"#$

ema T–TT !

!" •TP

(

%(

dtd!

"#

)

!" •JT

!! ! )! ! (

!" •JM

)%

2

2

aa dtd

•M•M•TP!

"#"$"

(%(*

!"#$% t•)t( syn

(

dtd

dtd

syn!

"#$%

$#

*

2

2

a dtd

•MP!

"

dtd

•H•2

dt

d•

H•2PPP

syn2

2

synema

!

!"

#

!"$"

!

*

##

!

"

#

!"# !"$

! " ! "ijijijijjiijsh2

iij sin )Y( imag cos )Y( realV•VYY real• VP !"!#"$!!!!

! " ! "ijijijijjiijsh2

iij cos )Y( imag sin )Y( real V•VYY imag• V –Q !"!"#$!!!!

++

-1 0 -5 0 5 10-1 0

-8

-6

-4

-2

0

2

4

6

8

1 0

!

!

!

!

!

!"#"$%&

!

!"

#$

%&

!

! "

!!"#$!

! " !"#"!"#!

!"

!

! "!"#!!"

!

!"

!

!

!

!"!

!"

!"#$"

%"

%

!"#

"$%$

!!"

#

!"#!

$sin•PP

dtdH2

maxm2

2

syn

13

31max X

E•EP

!"

!

!

SLTd13 XXXXX !!!"#

!

pu 536.002.0066.0152.0296.0X13 !"""!

!"#$

%23.70 sin•866.175.0

dtdH 2

2

2

syn

# Hz 597.13.2 •21.707 •377

21

H •2

S•

21 ps

!

"

!

#

"

!

pu 1.707 23.70 cos •1.866 cos •Pdt

dPS 0max

ep

0

!"!#!!

!"!

!"# !"$

1212

2112 sin

XV•V

P !"

1212

21

12

21

12 cos X

VVXV

Q !"#

12

21

12122

1

1

2

cos•V

Q•XVVV

#

$!

!

$%

12

12122 cos

Q•X1V

!

"#

)t( sin•e •A)t( t0 !"#"$%$ !

&'(

!!!

"

#

$$$

%

&'

(

!!!

"

#

$$$

%

&

(

2

1

12

3

2

1

V

V

x

x

x

x

&!")"

!!!!!!

"

#

$$$$$$

%

&

'

'

'

'

'

(

!!!!!!

"

#

$$$$$$

%

&)

*

5

4

3

2

1

12

12

2

1

12

Q

P

V

V

z

*

!!!!!!

"

#

$$$$$$

%

&

'

'

'

'

'

(

!!!!!!!!!

"

#

$$$$$$$$$

%

&

)

*'(*

5

4

3

2

1

12

13222

12

132

3

2

1

Xxcos•x•xx

Xxsin•x•x

x

x

x

)x(hz

*+"

! " ! ")x(hzR •)(xH )x(H•R•)x(Hxx k1k1k1Tkk1k!"#

!!!"

!!!!

, ,

, ,

, ,

! "!"#$ t• cos• I)t(i

! " ! "

!"

#$%

& '(

!"#$!"#

2ee

V)t(vtjtj

! "! " ! "

!"

#$%

& '(

!"#$!"#

2ee

Ititjtj

! " ! "

! "

dt

2ee

I d

L2

ee I•R

2ee

V

tj)t(j

tj)t(jtj)t(j )*

)+,

)-

)./

!"

#$%

& '

'!"

#$%

& '(!

"

#$%

& '

%"#$%"#

%"#$%"#!"#$!"#

! " ! "! "! "

dte•Id

Le•I•Re•Vtj

tjtj

!

!"!#!"

! "! "! "

dte •I d

Le •I •Re•Vtj

)t(jtj!"#$

!"#$%"#$!"

! "!!"#"!"#

jjtjjt•j e• I •L• je• I •R ee•Ve

! "!!#!"#

jjj e•I •L• je •I •Re•V

1

Resumen: Una completa descripción del proceso de medición fasorial conduce a los criterios ideales que un filtro digital debe de cumplir para ofrecer estimaciones fasoriales más precisas bajo oscilaciones de sistemas de potencia. El filtro Coseno Elevado, ampliamente utilizado en transmisión digital, se describe y propone para aproximar dichos criterios. Los resultados de simulación corroboran una sustancial mejora en el rendimiento (exactitud y velocidad) de los estimados fasoriales con respecto a los obtenidos con el filtro combinado Fourier-boxcar, utilizado comúnmente en las aplicaciones comerciales. Palabras Clave: medición fasorial, oscilaciones de sistemas eléctricos de potencia, control retroalimentado, resonancias de la red, FACTS, demodulación en cuadratura, demodulación de amplitud, ventanas, filtros digitales.

I. INTRODUCCIÓN La representación fasorial de una señal de corriente alterna

(CA) es una técnica estandarizada utilizada en muchas áreas de la ingeniería eléctrica. En ella se reduce una señal de CA a un vector representando la magnitud y fase de una senoidal. La transmisión de energía eléctrica generalmente usa CA, de manera que la señal es comúnmente representada por un fasor. Un microprocesador barato puede manejar fasores directamente de las muestras de la forma de onda en tiempo real. La disponibilidad de tiempo preciso sobre una vasta área geográfica a través del Global Positioning System (GPS) [1] ofrece una referencia común para la medición fasorial. Estos avances han permitido el desarrollo de sistemas que miden en tiempo real los fasores del sistema de potencia, en las subestaciones distribuidas por todo el sistema de potencia.

El propósito de la medición fasorial es producir una representación reducida y simplificada de las señales del sistema eléctrico de potencia, que describen con precisión el sistema real. En general, las mediciones son únicamente representaciones de los sistemas reales, y las fasoriales no son la excepción. Al evaluar el rendimiento óptimo tiene que tomarse en cuenta, de manera integral, no sólo la tasa de medición, sino también la exactitud, la técnica de conversión empleada, y la tasa de datos final. Esos criterios se derivan de

Este trabajo fue financiado por la Comisión Federal de Electricidad

(CFE), bajo el proyecto “Medición Sincronizada de Fasores en Sistemas Eléctricos de Potencia”.

J. A. de la O trabaja en la Universidad de Nuevo Léon, PO Box 113-F, San Nicolás de los Garza, N.L., 66450 Mexico (e-mail: [email protected]).

K.E. Martin trabaja en Bonneville Power Administration, PO Box 491, Vancouver, WA 98666, USA(email: [email protected]).

las características de las señales y anchos de banda requeridos por las aplicaciones que utilizarán las mediciones. Este artículo discute dichas características y demuestra un mejoramiento en el rendimiento al utilizar un algoritmo de conversión fasorial alternativo. Por simplicidad, el análisis y los ejemplos en este artículo serán dados para un sistema de 60 Hz, pero todos los principios son aplicables igualmente a sistemas de 50 Hz haciendo los ajustes apropiados.

II. REQUISITOS DE LA MEDICIÓN El proceso de medición fasorial puede separarse en dos

subprocesos. El primero consiste en muestrear las formas de onda de voltaje o corriente con un convertidor análogo/digital (A/D). Este proceso requiere prefiltrar las señales analógicas para evitar el aliasing, i.e., que las señales de alta frecuencia, arriba de la tasa de Nyquist, sean percibidas en la señal numérica como de baja frecuencia. El sistema de muestreo debe también ofrecer suficiente aislamiento eléctrico para prevenir las interferencias y daños de los transitorios.

El Segundo es el proceso de cálculo fasorial. Este proyecta la señal de potencia sobre las funciones Coseno y Seno, para ofrecer la componente real e imaginaria de la representación fasorial. Este proceso en efecto filtra todas las componentes de señal exceptuada la de 60 Hz. Estas componentes serán en general funciones del tiempo ya que la magnitud y frecuencia de la señal de potencia cambian constantemente. La exactitud con la cual estas funciones representan al sistema de potencia real, depende principalmente del algoritmo utilizado para la conversión fasorial. No solamente habrá dinámicas lentas representadas en las componentes, sino también muchos efectos de más alta frecuencia, causados por resonancia, oscilaciones torsionales de la flecha del generador, y controladores como los PSS y FACTS. Capturar las variadas y complejas dinámicas en la generación fasorial implica resolver compromisos, de manera que esta sección revisa las características de la conversión fasorial con relación a los datos que serán obtenidos.

Existen cuatro consideraciones principales al diseñar el algoritmo de cálculo fasorial: la tasa de muestras fasoriales, exactitud de la representación de la información incluida (en la banda de paso), la capacidad de rechazar la información excluida (en la banda de paro), y el retraso temporal de la medición.

A. Tasa de Muestreo Existen dos tasas de muestreo a considerar: la de la

numerización de las señales analógicas, y la de los fasores calculados. La tasa de muestreo de las formas de onda debe de

Refinando las Mediciones Fasoriales bajo Oscilaciones de Sistemas de Potencia

José A. de la O, Miembro, IEEE, y Kenneth E. Martin, Senior, IEEE

2

ser lo suficientemente alta como para hacer una buena reproducción. Ésta requiere filtrado analógico antiréplicas (antialiasing) antes del muestreo. Generalmente esta tasa de muestreo es mucho más alta que la frecuencia del sistema y a menudo dictada por los algoritmos de derivación fasorial.

La ‘tasa de muestras’ fasoriales es la tasa a la cual los fasores son producidos por el algoritmo fasorial. Esta puede variar desde muchos fasores por ciclo, hasta un fasor cada varios ciclos. Sin embargo, el criterio de Nyquist limita el ancho de banda de la información a la mitad de la tasa de muestras fasoriales, de manera que puede requerirse de filtrado adicional para evitar el solapamiento de réplicas frecuenciales en los estimados fasoriales. Este filtrado puede ser antes, durante o después del proceso de estimación. Ya que los fasores son estimados para la frecuencia fundamental del sistema, el filtrado antes de la estimación fasorial queda restringido a frecuencias superiores a la fundamental.

Un fasor puede ser calculado sobre cualquier cantidad de muestras de la forma de onda, pero la mayoría de los algoritmos usan bloques de muestras que abarcan un número entero de ciclos. Este lapso de tiempo sobre el cual los fasores son calculados es llamado ‘ventana’. La onda de 60 Hz está cambiando constantemente de manera que el cálculo fasorial representará un valor promedio sobre la ventana de cálculo. Las ventanas más angostas (que contienen menos ciclos) proveen una representación instantánea más rápida del sistema de potencia, pero también tienen un espectro frecuencial más amplio. Un espectro más amplio requiere una tasa de muestreo más alta o filtrado adicional para una representación exacta. Inversamente, las ventanas más amplias tienen espectros frecuenciales más angostos por lo que requieren tasas de muestreo más bajas y menos filtrado adicional. El ancho de banda de la información y la tasa de muestreo fasorial constituyen las consideraciones básicas que serán dictadas por la capacidad de manejo de datos de las supuestas aplicaciones. Estas determinarán algunas de las otras alternativas en el diseño del algoritmo de estimación fasorial. Los sistemas actuales usan o calculan fasores a 12, 30, 720, y 2880 muestras/segundo [2]. En nuestras simulaciones numéricas, 3840 muestras fasoriales/segundo (64 por ciclo) serán generadas.

B. Exactitud La señal de potencia generalmente se encuentra cerca de

60 Hz y está cambiando constantemente. La técnica de derivación fasorial requiere producir una medición exacta en ambas, amplitud y fase, no sólo para 60 Hz, sino para un rango de frecuencia alrededor de 60 Hz. La cantidad de desviación típica de 60 Hz y la razón de cambio dependen del sistema de potencia. Además hay muchas otras señales inyectadas en el sistema de potencia por gobernadores, dinámicas de flecha, estabilizadores, y otros elementos del sistema. Estos introducen en la señal de potencia información con frecuencias que pueden ir desde las bajas frecuencias (2 Hz o menos) hasta las altas, digamos 25 Hz, y aun acercándose a la fundamental de 60 Hz.

La banda de paso del algoritmo fasorial, alrededor de 60 Hz, necesita ser lo suficientemente amplia para representar las señales de interés. Debido a la transición del filtro, la exactitud

necesariamente decrecerá en la medida en que la frecuencia se aleja de 60 Hz (por arriba o por abajo). La transición es función del algoritmo, y es restringida por la tasa de muestras fasoriales, el rechazo fuera de banda, y los requisitos de retraso. Una tasa alta de muestras fasoriales permite usar respuestas en frecuencia más anchas de manera que el filtro puede ser más plano en la banda de paso. La tasa de datos es generalmente limitada por la capacidad de comunicación y de procesamiento de datos. El usuario tiene que decidir cuál es el ancho de banda de interés y los requisitos dentro de esa banda. Al final habrá que resolver un compromiso entre estos requisitos y las facilidades de comunicación y procesamiento de datos.

C. Rechazo Cuando los fasores son calculados, el criterio de Nyquist

requiere eliminar las frecuencias arriba de la mitad de la tasa de muestras para evitar el empalme de las réplicas en la señal muestreada. El nivel de filtrado requerido depende del contenido frecuencial en la señal original y la sensibilidad de la aplicación a señales extrañas. El mejor filtrado generalmente requiere más procesamiento de datos y un retraso de tiempo mayor. También puede introducir cambios de fase y atenuación de amplitud. Menos filtrado es más rápido pero puede introducir información errónea. Ultimadamente, el usuario y los requisitos de la aplicación determinan los límites.

D. Retraso en la Medición El retraso de la medición es el tiempo desde que un cambio

ocurre en la señal de potencia hasta que se manifiesta en los datos de salida. Un cambio repentino en la señal de potencia será promediado con valores en la ventana de medición anteriores y posteriores al cambio. Si las ventanas de cálculo se solapan, el cambio podría ser parcialmente reflejado en más de una ventana. En cualquier caso, el retraso entre el cambio y el tiempo en que el estado del nuevo sistema es representado completamente en los valores fasoriales depende de la longitud de la ventana de cálculo y de los filtros adicionales. Una ventana más ancha integrará cambios sobre un tiempo más largo y tomará más tiempo en reflejar los cambios del sistema de potencia. Por ejemplo, una ventana de cuatro ciclos tomará cuatro ciclos para reflejar el Nuevo valor después de un cambio escalón, mientras que el máximo retraso de una ventana de un ciclo es de un ciclo. Pero el ancho de banda de una ventana de un ciclo será mucho más ancho, requiriendo una tasa fasorial más alta o filtrado adicional. El filtrado adicional incrementará el retraso, de manera que podría no haber mejoría con una ventana más angosta. El usuario tendrá que determinar los requisitos para el retraso tolerado por la aplicación para optimizar esta característica.

E. Requisitos de la Aplicación Ningún algoritmo es ideal para todas las aplicaciones.

Todos ellos exhiben compromisos para acomodar el hardware y el software disponible a las necesidades de la aplicación. La tasa de muestras fasoriales limita el ancho de banda de la información, el cual es una consideración de todas las aplicaciones. También afecta el manejo de los datos

3

incluyendo su almacenamiento y transmisión. La exactitud en la banda de paso es también fundamental, pero fácil de lograr. Ella resulta al resolver un compromiso con el rechazo fuera de la banda. Si es muy probable que exista mucho contenido frecuencial fuera de la banda de paso permisible, es probablemente mejor sacrificar algo de planicie en la banda de paso por mejor rechazo en la banda de paro. Y una más alta tasa de muestreo permite tanto más rechazo como más planicie. Finalmente los efectos del retraso deben de ser evaluados. Éste es únicamente importante en aplicaciones de control ya que las diferencias entre los diversos algoritmos constan únicamente de unos cuantos ciclos.

Para aplicaciones de análisis, la exactitud de la banda de paso y el rechazo pueden ser muy importantes, sin embargo el retraso no lo es tanto. Una tasa de 10-muestras/seg es generalmente adecuada para dibujos de energía, voltaje, ángulo de fase y frecuencia utilizados para comparaciones visuales. Para comparaciones de modelos de sistemas, de 15 a 30 muestras/seg son probablemente necesarias. Para análisis de respuestas dinámicas, 20 muestras/seg es el mínimo que debería usarse. El usuario tendrá que determinar el mínimo requerido basado en las frecuencias de interés.

Tasas de datos de 10 muestras/seg son adecuadas para el monitoreo y respuesta de operador en tiempo real. La exactitud en la banda de paso y el rechazo no son críticos pues esos efectos caen generalmente bajo la visibilidad gráfica. Las diferencias de retraso no son visibles tampoco.

Las aplicaciones de control son las más exigentes. Ellas requieren tanto retrasos cortos como buen procesamiento de señales. Se requiere un buen rechazo fuera de banda, lo que requiere un buen enventaneo y filtrado. La ganancia en la banda de paso deberá ser constante si no se desea aplicar un postprocesamiento de compensación. Una alta tasa de datos puede ser requerida para minimizar el retraso y optimizar la exactitud en la banda de paso y el rechazo fuera de ella.

III. ALGORITMOS DE MEDICIÓN Hoy en día, uno de los algoritmos más utilizados para estimación fasorial es el filtro de Fourier. Su rechazo armónico y rapidez de estimación, tanto como su propiedad de recursividad, lo hacen el estimador favorito. Sin embargo, su rendimiento (exactitud y velocidad) en estado estable deberá de ser evaluado junto con sus deficiencias transitorias y las limitaciones impuestas por las tasas de datos fasoriales del usuario. El filtro de Fourier de un ciclo generalmente ha sido implementado con ventanas solapadas produciendo varios estimados fasoriales por ciclo. Pero debido a las limitaciones de la transmisión y el registro de datos, los fasores han sido utilizados a submúltiplos de 60 Hz, utilizando 1 fasor/N ciclos, donde N es un entero entre 2 y 60. El conjunto de datos ha sido generalmente reducido a 1/N ciclos usando únicamente un fasor cada N-ésimo ciclo (por submuestreo) o promediando sobre 2N ciclos cada 1/N ciclos (promediado uniforme o filtrado “Boxcar”). Desgraciadamente ambos procesos son muy susceptibles a la infiltración de la frecuencia negativa de la componente fundamental, y al solapamiento de réplicas (aliasing) de las componentes de frecuencias de alto orden en la banda de paso. En efecto, el submuestreo sin un buen filtro

pasabajas reproduce las componentes de alta frecuencia dentro de la banda de paso. El problema de la infiltración de frecuencia negativa (o frecuencia imagen) ha sido reportado en [3], en donde para rechazarla se propone un postfiltro ranura adaptivo. La vulnerabilidad de los PMU (Phasor Measurement Unit) a las señales fuera de banda bajo condiciones de campo ha sido reportada en [4] pero su autor reconoce que el PMU estudiado “ofrece la base para un excelente transductor digital” y que las razones de los detalles reportados “no son conocidas completamente” El problema es que el filtro Boxcar afila abruptamente el pico de la respuesta en la fundamental, de manera que las demás frecuencias son más severamente atenuadas que cuando se usa exclusivamente el filtro básico de Fourier. Los lóbulos laterales del Fourier son tan altos, que permiten que las frecuencias fuera de la banda de paso sean reproducidas dentro de ella. El efecto global de filtrado (Fourier-Boxcar) se muestra en la Fig. 1, donde la frecuencia ha sido normalizada con respecto a la fundamental. Se puede apreciar que dicho sistema atenúa fuertemente las frecuencias de la oscilación de potencia que se encuentran entre cero y fo/4 Hz. En la próxima sección se consideran los criterios de diseño que deben aplicarse al filtro para medir las oscilaciones de entrada con mejor fidelidad, considerándolas como un caso especial de modulación en amplitud.

A. Modulación de Amplitud Si la señal básica de potencia cos(2πf0 t+ θ) es modulada en amplitud por la señal a(t), el modelo de señal es el siguiente:

)2cos()()( θπ += tftats o , (1)

donde a(t) es la señal de modulación de amplitud, fo la frecuencia fundamental, y θ la fase; las cuáles, por simplicidad, se asumen constantes. Su transformada de Fourier será:

[ ])()(21

)( 00 ffAeffAefS jjv ++−= − θθ (2)

donde A(f) es el espectro de la función envolvente a(t). La transformada de Fourier de la señal proyectada en el plano fasorial [5] obtenida a la salida del filtro con respuesta en frecuencia Φ(f) será:

( ))()(2

)()( 00 ffAeffAef

f jj ++−Φ

=Σ − θθ (3)

Fig. 1 Respuesta en frecuencia de la combinación en cascada de un filtro de Fourier de un ciclo y un filtro (antialiasing) Boxcar de cuatro ciclos.

4

Y dado que la función de proyección requiere ser inversamente rotada a la frecuencia fundamental para obtener la función fasorial, la transformada de Fourier de esta última es una simple versión de Σ(f), trasladada hacia la izquierda en frecuencia:

)()( 0fffP +Σ= (4) Cuando el filtro pasabanda usado en el proceso de demodulación o proyección es de la forma:

)(2

)( 0ffWA

fw

−=Φ , (5)

donde W(f-fo) es la translación a la frecuencia fundamental del espectro de la ventana w(t), y Aw =W(0). Entonces la transformada de Fourier de la función fasorial será finalmente:

[ ])2()()(

)( 0ffAefAeA

fWfP jj

w

++= − θθ (6)

El primer término corresponde a la salida deseada del filtro, mientras que el segundo representa a la indeseable infiltración de frecuencia negativa. Así, para asegurar el rechazo completo de la frecuencia negativa, el espectro de la ventana deberá de cumplir con la siguiente condición:

0)2()( =+ offAfW . (7)

El Segundo factor, A(f+2fo), es una copia del espectro de la amplitud, trasladada a f=-2fo. Cuando la amplitud es constante (en estado estable), este espectro está concentrado en esa frecuencia, pero cuando la amplitud oscila, su espectro se esparce sobre una banda centrada en dicha frecuencia. A mayor variación dinámica en la amplitud, mayor separación de las bandas laterales de la oscilación. Entonces, de acuerdo con (7), se puede constatar que la condición W(-2fo)=0 , que se cumple en el caso del filtro de Fourier, es necesaria pero no suficiente para evitar la infiltración de la frecuencia negativa. Para excluirla completamente, W(f) deberá de ser nula sobre todo el intervalo de frecuencia en el cual A(f+2fo)≠0.

Por otra parte, la fidelidad a la amplitud de entrada requiere, del término de la izquierda en (6), que:

Aw

Ff1A

fW≤= para ,

)(.

(8)

donde FA es la máxima frecuencia contenida en la función de amplitud a(t). Al cumplir con las dos condiciones anteriores, a la salida del filtro se obtendría:

θjetatp )()( = (9) que constituye la salida deseada.

En la próxima sección, se presenta un filtro que ayuda a lograr en la práctica las condiciones teóricas anteriores. Note que en este artículo únicamente se consideran filtros de respuesta impulsional finita (FIR), o formados por una ventana. Aunque los filtros de respuesta impulsional infinita (IIR) sean más rápidos que los FIRs, éstos no se consideran aquí debido a que su respuesta en fase no lineal introduciría sistemáticamente distorsión en fase a la salida.

B. Filtro Coseno Elevado Un filtro particular de Nyquist [6,7] con amplias

aplicaciones prácticas es el filtro Coseno Elevado (CE) cuya respuesta en frecuencia es definida por:

)()( fcossTfW 2Ω= , (10) donde f es la frecuencia en Hz, Ω(f) es la siguiente función a trozos lineales de simetría impar, expresada aquí abajo para f≥0:

⎪⎪⎪

⎩

⎪⎪⎪

⎨

⎧

⎟⎠

⎞⎜⎝

⎛

+>−

+≤≤

−+−

<

=Ω

s

sss

s

s

2Ta1

f

2Ta1

f 2T1-a

Tf

T

2T1-a

f

f

en 2

en ,2

12

en ,0

)(

π

α

α

π ,

(11)

en la cual Ts es conocido como período simbólico y el parámetro α (0≤α≤1) es llamado factor de redondeado. La Fig. 2 muestra la función de transferencia W(u) para α=0, 0.7 y 1, en función de la frecuencia normalizada u=fTs. El factor de redondeado representa el exceso normalizado del ancho de banda espectral más allá de la frecuencia de Nyquist 1/2Ts. Como puede verse, éste define, en porcentaje, el ancho de la banda de transición, e indirectamente el de la banda de paso con ganancia constante.

La respuesta impulsional del filtro CE, también conocida como pulso de conformación de amplitud, es dado por la siguiente expresión en forma cerrada:

⎟⎟⎠

⎞⎜⎜⎝

⎛++−= )()()(

4)(

2

1

T

tsinc

2

1

T

tsinc

T

tsinctw

sss

ααπ (12)

donde sinc(x) es el seno cardenal, definido como sinc(x)=sen(πx)/πx. La Fig. 3 muestra la mitad derecha de los pulsos CE para α=0, 0.7 y 1, como funciones del tiempo normalizado (u=t/Ts). Note que el período simbólico Ts corresponde a la separación entre dos sucesivos cruces por cero en todos los pulsos. En transmisión digital, los pulsos portadores del código de cada símbolo son separados por la misma cantidad de tiempo con el fin de evitar la interferencia intersimbólica (Intersymbol Interference, ISI).

En nuestras simulaciones digitales, un filtro con frecuencia de corte (-6 dB) en 30 Hz será utilizado, de él resulta un pulso con Ts igual al período fundamental (Ts=To=1/fo). La compactación temporal lograda por el parámetro α puede ser explotada para reducir los lóbulos laterales y el rizado que aparece cuando el pulso ideal es truncado para implementarse en un filtro FIR. Para un truncamiento en ±2Ts, se encontró

u=fTs

1

0

W 1 u,( )

W 0.7 u,( )

W zero u,( )

11− u

Fig. 2 Función de transferencia del filtro CE α=0, 0.7 y 1

5

Fig. 3 Respuesta impulsional del filtro CE, α=0, 0.7 y 1. que α=0.7 es la mejor alternativa.

Para implementar el filtro digital, la función truncada de w(t) deberá ser trasladada y muestreada para derivar los coeficientes de ponderación del filtro de Fourier. Si N es la cantidad de muestras por ciclo, los coeficientes de ponderación para el filtro CE de cuatro ciclos son dados por:

11,2,...,4NkTkNT

wkw oo −=−= ,0 ),2(][ (13)

Y si s[n] es la señal de entrada al filtro, la salida p[n] del filtro CE será dada por:

∑−

==

1 2

]e[][2

][4N

0k

Nj

w

kkwn-ks

Anp

π

(14)

donde Aw es una constante, suma de todos los coeficientes:

∑−

==

1][

4N

0kw kwA (15)

Como se puede apreciar, los coeficientes del filtro CE en (13) fueron extraídos de uno de los pulsos de configuración de amplitud definido en (12) y truncado y muestreado en (13), y no de la ventana de Hanning, también conocida en la literatura como ventana Coseno Elevado (Raised Cosine Window).

Finalmente, si los fasores deben calcularse cada L muestras, una rotación inversa es necesaria para compensar la rotación de la señal deslizante, por lo que la secuencia de fasores estimados será dada por:

0,1,2,...nnLpnLn

NLj

e =−

= ][][ ,2π

ρ (16) La mínima separación es L=1, la cual corresponde a un

fasor por muestra de señal, con una tasa de muestras fasoriales de Nfo. La separación máxima está limitada por la tasa de Nyquist, la cual, a la salida de un filtro con un ancho de banda a (-6 dB) de 30 Hz, corresponde a 60 muestras fasoriales por segundo, i.e., a una separación máxima de un ciclo (Lmax=N). En las siguientes simulaciones, se tomó L=1, de manera que cualquier otra tasa inferior podrá obtenerse simplemente submuestreándolos.

IV. RESULTADOS DE SIMULACIÓN Como ejemplo de una oscilación de potencia, consideramos la siguiente función de modulación de amplitud:

)2cos()( mmtf21

1.5ta θπ ++= (17)

donde fm=6Hz, θm=π. Ésta es aplicada en (1) con los parámetros fo=60 Hz y θ=0.

La Fig. 4 muestra los primeros dos ciclos de la amplitud a la entrad (línea lisa) y salida (línea granular) del filtro de Fourier de un ciclo. La granularidad en la señal de salida se debe a la significativa infiltración de la componente de frecuencia negativa por el filtro de Fourier, como puede comprobarse en su espectro, mostrado en la Fig. 5, en función de la frecuencia normalizada fTo. Junto con las componentes de la oscilación en ±6 Hz, se perciben dos componentes en –2fo±6 Hz que corresponden a la infiltración de la componente de frecuencia negativa de la portadora. Antes de obtener el espectro, la señal de salida fue centrada, con el fin de excluir la componente de frecuencia nula. Una explicación detallada de la manifestación de este fenómeno en el plano fasorial aparece en el Apéndice. Con el fin de filtrar la granularidad de la señal de salida de la Fig. 4, un postfiltro Boxcar de cuatro ciclos ha sido empleado (en cada instante, la amplitud es promediada sobre sus anteriores cuatro ciclos). Este proceso atenúa fuertemente la amplitud e introduce un retraso de dos ciclos. Además, la compensación de amplitud es solamente posible cuando la frecuencia de la oscilación es única y conocida. La Fig 6 compara la amplitud de salida del postfiltro de cuatro ciclos con la original de entrada. Aún cuando la granularidad desaparece, la amplitud es fuertemente atenuada y retrasada hasta dos ciclos y medio de la fundamental. La comparación con un filtro de Fourier de cuatro ciclos se muestra en la Fig. 7. Ésta claramente muestra tanto una persistente infiltración de frecuencia negativa, como la reducción de amplitud y el retraso en la señal de salida. Por supuesto que si un filtro Boxcar de un ciclo fuese utilizado para alisar la señal, el resultado final sería el mismo mostrado en la Fig. 6. Este ejemplo demuestra claramente que la infiltración de frecuencia negativa constituye el mayor inconveniente del filtro de Fourier. Finalmente, como solución a los inconvenientes anteriores, proponemos el uso del filtro CE. La amplitud de la oscilación obtenida con el filtro CE de cuatro ciclos (α=0.7) es mostrada en la Fig. 8. Dado que la latencia de este filtro es de dos ciclos, la amplitud aparece retrazada a dos ciclos fundamentales de la entrada. Como puede apreciarse una curva de amplitud lisa y sin atenuación es obtenida directamente con este filtro. De hecho, su amplitud de salida es (casi) una réplica retrasada de la entrada, debido a la respuesta en frecuencia plana de este filtro. Además, la granularidad, producida por la infiltración de la componente de frecuencia negativa de la portadora, es suprimida casi completamente, debido al fuerte rechazo en su banda de paro. Con esta nueva solución para estimar fasores, no se requiere ningún filtro adicional ni compensación de amplitud. Además, la salida de amplitud no es sensible a las persistentes desviaciones de la frecuencia fundamental del sistema, debido a la planicie de la respuesta en frecuencia alrededor de la frecuencia fundamental. Finalmente, su señal fasorial puede ser submuestreada hasta 60 Hz sin problemas de aliasing ya que tiene un ancho de banda (a -6 dB) en 30 Hz.

6

Fig. 4 Entrada y salida de amplitud obtenida con el filtro de Fourier de un ciclo.

Fig. 5 Espectro de la amplitud de salida en la Fig. 4

Fig. 6 Amplitud de entrada y salida después del posfiltro Boxcar de cuatro ciclos.

Fig. 7 Entrada y salida de amplitud del filtro de Fourier de cuatro ciclos.

Fig. 8 Entrada y salida (retrasada) del filtro CE de cuatro ciclos. Cualquier oscilación con contenido frecuencial debajo de 9 Hz, límite de la respuesta plana considerada en este ejemplo, pasará a través del filtro sin distorsión significativa alguna, obteniendo a la salida una copia fiel (sólo retrasada) de la oscilación de amplitud.

V. CONCLUSIONES Los requisitos de datos para mediciones fasoriales difieren de aplicación en aplicación. Los análisis de datos registrados no necesitan datos transmitidos en milisegundos, pero una aplicación de control podría requerirlos. Mediciones hechas en un sitio cercano a un generador particular podrían necesitar mejor rechazo subarmónico que en otros lugares. La tasa de muestras fasoriales, la exactitud, el rechazo fuera de la banda de paso, y el retraso de la medición constituyen los parámetros interactivos en el cálculo fasorial. La aplicación de interés dictará los requisitos mínimos y los compromisos preferidos en la derivación. La técnica de estimación fasorial debería ser confeccionada para optimizar los parámetros elegidos. Un filtro Coseno Elevado permite capturar con precisión el comportamiento dinámico de una oscilación bajo la banda de frecuencia definida por su respuesta en frecuencia plana. Este rechaza casi perfectamente la componente de frecuencia negativa de la señal senoidal predominante. Dicho filtro ofrece un excelente rechazo fuera de banda, de manera que no requiere filtrado antialiasing adicional para el submuestreo fasorial. Ofrece una salida más lisa, menos distorsionada y con menos retraso que la solución Fourier-Boxcar, y constituye una buena alternativa para su remplazo.

VI. APÉNDICE La infiltración de la componente de frecuencia negativa se

manifiesta en el plano complejo por la generación de una trayectoria de cicloides concatenadas. La Fig. 9 ilustra la trayectoria (amplitud y fase) seguida en el plano complejo por los fasores estimados a la salida del filtro de Fourier de un ciclo. Dado que en este ejemplo, la amplitud a(t) es real, y θ=0, la trayectoria ideal debería de ser una línea recta sobre el eje real, oscilando entre los puntos (1,0) y (2,0). Pero como puede apreciarse en la Fig. 9, la trayectoria resultante se encuentra lejos de la ideal, siendo una concatenación de cicloides fuera del eje real. Estas cicloides son principalmente

7

Fig. 9 Trayectoria seguida en el plano complejo por los fasores obtenidos con el filtro de Fourier de un ciclo. debidas a la permeabilidad de frecuencia negativa del filtro de Fourier, y explican la granularidad de la amplitud de salida en las Figs. 4 y 7. Cicloides hacia arriba corresponden a amplitudes crecientes, mientras que las amplitudes decrecientes producen cicloides hacia abajo, con tamaño máximo en el punto de inflexión (1.5,0), donde la tasa de cambio de la amplitud es máxima. De manera que el filtro de Fourier siempre produce cicloides cuando el fasor se mueva de un estado estable a otro (amplitud, frecuencia y ángulo constantes). En la salida del filtro de Fourier de cuatro ciclos, la parte imaginaria máxima, alcanzada por la cicloide mayor del centro, se reduce apenas 80% (a 0.02) de la correspondiente al filtro de Fourier de un ciclo (0.025) en la Fig. 9. Mientras que el filtro Coseno Elevado de cuatro ciclos, la reduce a una centésima de dicho valor.

VII. RECONOCIMIENTO Este trabajo fue financiado por la Comisión Federal de Electricidad

(CFE), bajo el Proyecto “Medición Sincronizada de Fasores en Sistemas Eléctricos de Potencia”.

VIII. REFERENCIAS [1] R.J. Murphy, “Power System Disturbance Monitoring”, 21st Annual

Western Protective Relay Conference, Spokane, Washington, Oct. 18-20, 1994.

[2] K.E. Martin, R. Kwee, “Phasor Measurement Unit Performance Tests,” Precise Measurements in Power Systems Conference, Arlington, VA, November 1995.

[3] C.P. Denys, C.Counan, L. Hossenlopp and C. Holweck, “Measurement of Voltage Phase for the French Future Defence Plan Against Losses of Synchronism,” IEEE Transactions on Power Delivery, vol 7., no.1, pp. 62-69, January 1992.

[4] B.J.F. Hauer, “Validation of Phasor Calculation in the Macrodyne PMU for California-Oregon Transmission Project Tests of March 1993,” IEEE Trans. Power Delivery, Vol. 11, no.3, pp 1224-1231,July 1996.

[5] De la O J. A., “New Theory for Phasor Measurement,” Proceedings of the 19th IEEE Instrumentation and Measurement Technology Conference (IMTC), Anchorage, AK, May 2002, pp 1405-1410.

[6] De la O J. A., “On the Use of Amplitude Shaping Pulses as Windows for Harmonic Analysis,” IEEE Trans. on Instrumentation and Measurement, Vol. 50, No. 6, Dec. 2001.

[7] J.B. Anderson, Digital Transmission Engineering, New York: IEEE Press, 1999, pp 14-28 .

IX. BIOGRAFÍAS José Antonio de la O Serna nació en San Pedro, Coahuila, México in 1953. Recibió su grado de Doctor Ingeniero en Telecom, Paris, Francia, en 1982. De 1982 a 1986 fue profesor en el Tecnológico de Monterrey, en Monterrey, México. En 1987, ingresó al Programa Doctoral en Ingeniería Eléctrica en la Universidad Autónoma de Nuevo León, donde fue miembro del Comité Doctoral. De 1988 a 1993 se incorporó al Departamento de Eléctrica del Politécnico de Yaoundé, Camerún. En 1994 regresó al Programa Doctoral de la UANL, en Monterrey, donde es

ahora Profesor Investigador. Es Presidente del IEEE Sección Monterrey y miembro del Sistema Nacional de Investigadores.

Kenneth E. Martin es ingeniero de la BPA (Bonneville Power Administration) desde 1975 donde ha trabajado con sistemas de protección, sistemas de control y telecomunicaciones e instrumentación. Desde 1988 su área de interés ha sido los sistemas de temporización basados en GPS y las mediciones de área amplia y en tiempo real de sistemas de potencia. Entre sus proyectos recientes se incluyen un sistema de temporización basado en tecnología GPS y el

desarrollo e implantación de sistemas de medición fasorial en el WSCC. Obtuvo su título de ingeniero en la Universidad de Colorado y una maestría en Matemáticas en la Universidad de Washington. Es Senior del IEEE, e ingeniero registrado en el estado de Washington.

1

Resumen: Se desarrolla la formulación matemática para estimación de fasores a partir de fasoretas de segmentos de señal infracíclicos y se cuantifican las respuestas en frecuencia de los filtros que las generan. Para la evaluación del ruido en las fasoretas se consideran dos señales de ejemplo. De la teoría y las simulaciones, se concluye que, cuando el segmento de señal involucrado en el cálculo de una fasoreta corresponde a una senoidal pura en estado estable, ésta ofrece la más exacta y rápida estimación fasorial. Sin embargo, ante señales transitorias o senoidales impuras, la exactitud y comportamiento dinámico de éste método de estimación fasorial dependerá en gran medida de la técnica de extracción de ruido aplicada a la salida, dada la gran sensibilidad de las fasoretas al ruido. Palabras Clave: estimación fasorial, fasoreta, filtro digital, respuesta en frecuencia, proyección de señales, bases no ortogonales, plano oblicuo.

I. INTRODUCCIÓN La medición fasorial es muy importante para monitorear y

diagnosticar sistemas eléctricos de potencia [1]. Los fasores portan la amplitud y la fase de una señal senoidal en estado estable (frecuencia, amplitud y fase constantes). Para calcularlos, los algoritmos actuales procesan segmentos de señal (de voltaje o corriente) de duración igual a un múltiplo de ciclo fundamental. Esto se debe a que las señales Coseno y Seno de frecuencia fundamental, referencias para obtener la componente real e imaginaria del fasor, son ortogonales entre sí solamente con esas duraciones. Bajo esa ortogonalidad, las componentes horizontal y vertical del fasor se obtienen directamente, con la ventaja adicional de que las armónicas, siendo ortogonales al plano fasorial, se eliminan automáticamente en el proceso de estimación.

Para señales de duración inferior al ciclo, las referencias Coseno y Seno de frecuencia fundamental, truncadas en segmentos de duración fraccionaria, forman bases oblicuas que generan planos no ortogonales al plano fasorial. En este caso es necesario referir al plano fasorial la proyección ortogonal de la señal sobre el plano oblicuo, mediante una transformación adicional. Desgraciadamente, en esta composición de transformaciones se infiltran armónicas y otras componentes no ortogonales a los planos oblicuos.

Recientemente se ha propuesto un algoritmo para estimar fasores a partir de estimaciones obtenidas sobre segmentos sucesivos de señal de duración inferior al ciclo, ahí llamadas

Este trabajo recibió un apoyo complementario por la Universidad de

Nuevo León (UANL), bajo el proyecto PAICYT CA561-01 “Medición Fasorial sobre Bases Oblicuas”.

J. A. de la O trabaja en la Universidad de Nuevo Léon, PO Box 113-F, San Nicolás de los Garza, N.L., 66450 Mexico (e-mail: [email protected]).

fasículas (phaselets) [2]. Y con secuencias de ellas se va formando un estimado fasorial de duración variable, hasta llegar a un ciclo.

El presente artículo se propone desarrollar matemáticamente el proceso de la estimación fasorial a partir de señales de duración infracíclica, analizar su respuesta en frecuencia y evaluar tanto los errores por infiltración como su comportamiento dinámico ante señales transitorias. Lo anterior con la idea de llegar a plantear nuevas y mejores estrategias para obtener estimaciones fasoriales infracíclicas más precisas y con mejor comportamiento transitorio.

II. ESTIMACIÓN FASORIAL Las partes real e imaginaria de la secuencia:

10,1,...,N-nnn

Ni

N e == ,2

2)(π

ϕ (1)

que resulta al tomar N muestras por ciclo de una exponencial compleja continua, forman una base ortogonal que genera un plano en el hiperespacio de dimensión N. En cualquier instante k, la señal discreta definida por la secuencia {s(n), n=0,1,2,..., N-1}k se proyecta ortogonalmente en el punto σ(k) de dicho plano, mediante la suma de convolución:

)(1

0)()( nks

N

nnk −∑

−

== ϕσ (2)

En el caso particular de una señal exponencial compleja de la forma:

πωθω 20)( ,)( <≤= +nieAns , (3) la secuencia de proyecciones es dada por:

)(1

0

)( )()()( θωωθω ωϕσ +−−

=

+ Φ== ∑ kiniN

n

ki AeenAek , (4)

donde Φ(ω) es la función de transferencia de ϕ(n). Así, el proceso de proyección se puede considerar también como un proceso de filtrado. Siendo el filtro un sistema lineal con respuesta impulsional ϕ(n). Note que para obtener el fasor de una cosenoidal de frecuencia fundamental ω1=2π/N, se requiere que Φ(-ω1)=0 y Φ(ω1)=2 y compensar negativamente la rotación en (4), de manera que la secuencia fasorial obtenida es dada por:

θπ

σρ ikN

iAekk e ==

−)()(

2

(5)

Al calcular la proyección en (2), es posible separar la suma de convolución en Pj=2j sumas parciales, j=0,1,2...,log2(N), con Wj=N/Pj términos cada una:

∑∑==

−

−=∑ =−=

jj j

j

P

p

jp

P

p

pW

Wpnknksnk

11

1

)1()()()()( σϕσ (6)

Para cada partición j, los términos σ jp (k), p=1,2,..,2j serán

llamados fasoretas. Con más precisión diremos la fasoreta p de

Estimación Fasorial sobre Bases Oblicuas José A. de la O, Miembro, IEEE

2

orden j. La Ec. (6) se puede interpretar como la descomposición de la proyección (o del fasor) en fasoretas de orden j.

También es posible obtener las fasoretas mediante una sola convolución, definiendo 2j secuencias )(nj

pϕ , en las cuales se

anulan los elementos de ϕ(n) que no intervienen en el cálculo:

⎭⎬⎫

⎩⎨⎧

=−≤≤−

resto elen 0

)()(

1)1( jpWnjWpjp

nn

ϕϕ (7)

Así, para una partición j, la fasoreta p se obtiene mediante la siguiente ecuación de convolución:

∑=−

==−

1

0 ,)()()(

N

n

jjp

jp 1,2,...,2pnksnk ϕσ (8)

Note que las fasoretas de un nivel de resolución inferior se forman agregando de dos en dos fasoretas vecinas de nivel superior, es decir:

j-1jp

jp

jp 1,2,...,Ppkkk =+= −− )()()( 2121 σσσ (9)

La Fig. 1 ilustra con cruces en el plano fasorial las fasoretas más pequeñas, de sexto orden, del fasor e iπ/4. Los puntos corresponden a las fasoretas de 64 puntos equiespaciados en la circunferencia unitaria del plano fasorial: σn= e i2πn/64 , n=0,1,2, ...63. Las fasoretas de cualesquiera de los puntos se obtiene girando el haz de cruces en la Fig. 1 a la fase correspondiente. Como se puede apreciar, las fasoretas son la perspectiva en el plano fasorial de las proyecciones de un fasor en los planos oblicuos, una especie de vista caleidoscópica del fasor desde el plano fasorial. Por ello, al variar la fase del fasor, se generan en el plano fasorial las perspectivas de los distintos planos oblicuos en el hiperespacio. Dado que se trata del nivel de resolución más fino, la separación entre planos adyacentes es la más pequeña. Al ir reduciendo el orden j, los planos se van separando ya que están formados por la suma de dos fasoretas adyacentes de orden superior, las cuales a su vez se encuentran en planos adyacentes. Este proceso termina hasta formar el plano fasorial en el que las fasoretas coinciden con los 64 fasores unitarios.

Fig. 1. Fasoretas (+) de señal cosenoidal de amplitud unitaria y π/4 rad de fase en el plano fasorial. Y fasoretas (.) de cosenoidales con 64 fases distintas formando planos oblicuos.

Fig. 2 Descomposición de 64 puntos equiespaciados en la circunferencia en fasoretas de orden 6. Con cruces la trayectoria del punto σ= e iπ/4 .

La Fig. 2 muestra la suma de las fasoretas de resolución más fina correspondientes a los 64 puntos equiespaciados sobre la circunferencia unitaria. Note que para llegar a cualquier punto de la circunferencia, la agregación de fasoretas siempre forma dos trayectorias cicloides desde el origen. Como ejemplo se ilustra con cruces la composición de e iπ/4.

Dada la Ec. (9) y la naturaleza aditiva de la fig.2, la composición de los fasores, con fasoretas de un nivel inferior se obtiene simplemente diezmando por dos los puntos de la trayectoria correspondiente. Desde las fasoretas de más fina resolución, el proceso puede continuar, con diezmados por factores de 4, 8, 16, 32 y 64, con el que se llega al nivel de resolución más bajo, en el cual la única fasoreta corresponde al punto terminal. Note que las trayectorias repiten dos veces un mismo patrón. Es sencillo demostrar que cuando la señal de entrada es una senoidal pura, de la forma s(n)= A cos(2πn/N+θ), las fasoretas son periódicas de periodo Pj/2. Para esas señales, que corresponden exactamente con puntos del plano fasorial, se tiene por tanto

2/ )()( 2/ jj

jPpjp 1,2,...,Ppkk == +σσ (10)

y basta con calcular la mitad. Es por ello que en la Fig. 1 solamente se ven 32 cruces en vez de 64.

III. FUNCIONES DE TRANSFERENCIA De acuerdo con (7), las fasoretas son generadas por filtros

cuya respuesta en frecuencia [3] es dada por su función de transferencia:

jniN

n

jp

jp 1,2,...,2pen =Φ −

−

=∑= )()(

1

0

ωϕω (11)

Y es evidente que la función de transferencia total es:

∑Φ=Φ=

j

p

jp

2

1)()( ωω (12)

Aplicando series geométricas se obtiene:

⎪⎪

⎩

⎪⎪

⎨

⎧

=−−−−

−

−

=

Φ]

21

2)2)[(2(

)]21)([(

)])([(

(13) N

2 cuando

22

2

2 ,

)(jW

1pN

i

jp

eN

jW

N

NjW

sen

senN2

πω

ω

ω

πω

π

πω

3

La Fig. 3 muestra la magnitud de las respuestas en frecuencia Φj

p(ω) para j=0,1,2. Se puede observar que en la medida en que aumenta el nivel de la partición, aumenta el ancho de banda de los filtros. En la Fig. 3 se puede constatar que las respuestas en frecuencia de los filtros son asimétricas, ponderando diferentemente las componentes de frecuencia positiva y negativa de cada armónica. Por ejemplo, el filtro Φ2

1(ω) permite el paso de la frecuencia fundamental positiva, pero no impide el paso de la negativa, por otra parte deja pasar una fracción de la componente de frecuencia positiva de la 3ª armónica, pero elimina completamente la componente de frecuencia negativa.

Note en (13) que la respuesta en magnitud es igual para todos los filtros de una misma partición, ya que sólo depende de la cantidad de coeficientes del filtro Wj y no del índice p. Este último sólo afecta la respuesta en fase, con una diferencia de fase entre un filtro y su vecino de Wj(ω-2π/N), lo que corresponde al retraso por su posición relativa en el intervalo de tiempo. Por tanto, ante una misma señal de entrada, las salidas de los filtros de una misma partición serán idénticas, a excepción de sus diferentes retrasos. Esto hace que por cada partición, sólo un filtro sea de interés, los demás son redundantes.

IV. CÁLCULO FASORIAL COMO TRANSFORMACIÓN LINEAL Es interesante notar que la formulación anterior puede expresarse en términos matriciales, para ello note que la fasoreta p relativa al instante k se puede definir en términos matriciales como:

kjp

jp k sFs =)( (14)

donde el vector sk es un vector columna (N×1) conteniendo las muestras de la señal anteriores al instante k:

Tk Nksksks )](),.....,1(),([ −−=s (15)

y la matriz Φjp es la matriz rectangular 2×N conteniendo la

parte real e imaginaria de los coeficientes del filtro complejo correspondientes a fasoreta p. Por lo que las fasoretas son las proyecciones del vector sk sobre los planos generados por los vectores renglón de Φj

p. Dado que el vector de la secuencia cosenoidal

sc(n)=Acos(2πn/N+θ) se puede generar con los coeficientes del filtro de Fourier:

kk Fss = (16)

Fig.3 Respuesta en frecuencia de los filtros Φ jp(ω), j=0, 1 y 2.

donde sk es el fasor de la señal en k

⎥⎦

⎤⎢⎣

⎡=

)(

)(

θ

θ

sen

cosAks (17)

y F es la matriz N×2 formada por las partes real e imaginaria de los coeficientes del filtro de Fourier:

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

−−

=

)( )(

)( )(

)( )(

)1(2

)1(2

22

00

NNNN

NN

sencos

sencos

sencos

ππ

ππ

LLF

(18)

Al sustituir (16) en (14) y resolver para sk se obtiene:

( ) jp

jpk sFFs

1−= (19)

Por tanto, cuando el segmento de señal considerado en el cálculo de la fasoreta p corresponde a una señal senoidal pura en estado estable, su fasor puede estimarse sin error alguno a partir de ella mediante (19). El problema del error aparece cuando la señal de entrada no es una senoidal pura en estado estable como la expresada en (16). Esto incluye los casos de señales transitorias, señales con armónicas (periódicas), o señales con ruido aditivo, tal como la exponencial atenuada que aparece comúnmente en las corrientes de falla. En estos casos habrá que eliminar el error por otros métodos. En la siguiente sección presentamos simulaciones numéricas de estimación fasorial a partir de fasoretas de orden 4.

V. RESULTADOS DE SIMULACIÓN Con el fin de evaluar los errores de estimación fasorial a

partir de fasoretas, consideramos a continuación los fasores estimados de fasoretas de diversos órdenes, comparándolos con los de la de orden 0. Ya que los de esta última corresponde al fasor estimado directamente con un filtro de Fourier de un ciclo. Para ello compararemos los estimados de amplitud y fase de una señal de corriente y otra de voltaje, registradas durante una falla. Dichas señales se encuentran disponibles en [4], en un archivo llamado reclosed en donde vienen muestreadas a N=64 muestras por ciclo.

La Fig. 4 muestra los estimados de amplitud de la señal de corriente, la cual se muestra con línea punteada. Como se puede apreciar dicha señal contiene al inicio una importante exponencial amortiguada, seguida por un segmento de estado estable, una súbita interrupción y un corto intervalo de

Fig. 4 Estimaciones de amplitud con fasoretas σ0

1 (rayas) y σ41 (continua)

de una señal de corriente (línea punteada)

4

restablecimiento. Se puede observar que los estimados de amplitud obtenidos a partir de la fasoreta σ 41 (línea continua) contienen una fuerte infiltración de la exponencial atenuada, la cual se manifiesta por oscilaciones de gran amplitud y frecuencia fundamental. Mientras que en el segmento de estado estable, las estimaciones de amplitud contienen ruido de alta frecuencia. Este ruido se detectó en las estimaciones de fasoretas de alto orden, superior a tres. Y se debe a que, en la medida en que los filtros de estimación fasoretal tienen menos coeficientes, su respuesta en frecuencia es más ancha por lo que son más susceptibles al aliasing. Por otra parte, con menos coeficientes, la matriz inversa en (19) es numéricamente más inestable. De hecho la estimación de fasores a partir de fasoretas de orden 6 fue imposible, debido a que, para ese orden, la inversa de la matriz en (19) no se pudo calcular debido a su proximidad a la singularidad.

La Fig. 5 muestra los estimados de amplitud de la señal de voltaje. Esta señal se apega más a una senoidal pura en sus respectivos segmentos de estado estable. Como se puede apreciar, los estimados fasoriales de amplitud contienen nuevamente ruido de alta frecuencia con algunos picos. Este ruido desaparece cuando los estimados se obtienen con fasoretas de más bajo orden o menos resolución (inferior o igual a 3).

Cuando se comparan las estimaciones de amplitud obtenidas con fasoretas de diferentes órdenes, las señales de salida poseen dinámica muy semejante (exceptuado el ruido de alta frecuencia) a las de la Fig. 4 y 5, teniendo como límite la estimación de la fasoreta de orden cero, que ofrece el estimado con menos ruido. La Fig. 6 muestra las estimaciones de amplitud cuando a la corriente de la Fig.4 se le aplica un extractor de componente exponencial [5]. Se muestran los estimados a partir de las fasoretas de orden cero (línea a rayas) y tres (línea continua). Como se puede constatar, las oscilaciones de la Fig. 4 fueron eliminadas, sin embargo las transiciones continúan teniendo picos abruptos, lo que evidencia la necesidad imperiosa de posfiltrado sobre los estimados provenientes de fasoretas. Finalmente la Fig. 7 muestra los estimados de amplitud de voltaje a partir de las fasoretas de orden cero (línea a rayas) y dos (línea continua). Se observa que el ruido de alta frecuencia es eliminado al reducir el orden de la fasoreta, sin embargo los picos persisten.

En cuanto a los estimados de fase, se cumplen las mismas observaciones hechas para las estimaciones de amplitud, tanto en su alta sensibilidad a señales distintas a una senoidal pura,

Fig. 5 Estimaciones de amplitud de señal de voltaje (línea punteada) con fasoretas σ0

1 (rayas) y σ41 (continua)

Fig. 6 Estimados de amplitud de corriente sin aperiódica exponencial (línea punteada) con fasoretas σ 01 (a rayas) y σ 31 (continua).

Fig. 7 Estimados de amplitud de voltaje (línea punteada) con fasoretas σ 01 (a rayas) y σ 21 (continua)

Fig. 8 Estimaciones de fase con fasoretas σ 01 (a rayas) y σ 41 (continua) en la señal de corriente (punteada) de la Fig. 4.

Fig. 9 Estimaciones de fase con fasoretas σ 01 (a rayas) y σ 21 (continua) en la señal de voltaje (punteada) de la Fig. 7. como al ruido de alta frecuencia para estimados provenientes de fasoretas de alto orden. Como ejemplo se presenta en la Fig. 8 los estimados de fase correspondientes a la Fig. 4 y en la Fig. 9 los de la señal de voltaje de la Fig. 7. Se puede observar nuevamente las oscilaciones (de fase) provocadas por la componente exponencial en la fase de la corriente y los picos en los estimados de fase de la señal de voltaje.

5

De la consideración de la formulación matemática y los ejemplos anteriores, se desprende que es necesario que los estimados fasoriales a partir de fasoretas vayan acompañados de un estricto proceso de filtrado de ruido. Éste puede darse a priori (prefiltrado) para depurar la componente de frecuencia fundamental antes de entrar al algoritmo de estimación, o bien a posteriori (postfiltrado) a fin de depurar los estimados de los picos, o de los ruidos, ya sea de alta frecuencia (en fasoretas de orden elevado) o de las componentes distintas a la senoidal pura. Dado que los filtros generadores de fasoretas de un mismo orden poseen la misma respuesta en magnitud (sólo difieren en fase), el proceso de conformación del fasor deberá tener en cuenta la secuencia de fasoretas de orden decreciente σ4

1, σ31, σ2

1, σ11 para aproximarse a la estimación de un ciclo

σ01, el cual constituye el límite de extracción intrínseca de

ruido y correspnde al estimado con el filtro de Fourier de un ciclo. Como es ya bien sabido éste filtro tiene un pobre rendimiento transitorio [6], pues sólo ofrece estimados fasoriales exactos cuando la señal es senoidal pura, como cualesquiera de los estimados a partir de fasoretas, con la ventaja de que éstos últimos lo hacen más rápido.

VI. DISCUSIÓN La justificación del uso de algoritmos de estimación fasorial mediante fasoretas ha sido basada en la eliminación del ruido mediante técnicas aproximación por mínimos cuadrados. Es necesario hacer una demostración matemática de que dicho procedimiento ofrece estimaciones suficientemente exactas. De manera que este trabajo amerita ser continuado para estudiar el comportamiento de las fasoretas ante diferentes tipos de componentes comúnmente presentes en las diferentes aplicaciones de medición fasorial.

VII. CONCLUSIONES Y RECOMENDACIONES La estimación de fasores a partir de fasoretas funciona

perfectamente cuando la señal de entrada es un segmento de senoidal pura. Si no existen otros componentes de señal bajo la duración de los filtros que las generan, sus estimados son correctos sin necesidad de extraer ruidos, para fasoretas e orden inferior a tres. Sin embargo, cuando la señal no es una senoidal pura, habrá que extraer las infiltraciones que disminuyen la exactitud de la estimación. El comportamiento dinámico y la exactitud de los estimados dependerá fuertemente de la técnica de extracción de ruido utilizada.

Se recomienda no utilizar fasoretas de alto orden ya que son susceptibles de infiltración a ruidos de alta frecuencia.

VIII. REFERENCIAS [1] Advancements in Microprocessor Based Protection and

Communication, IEEE Tutorial Course, Power System Relaying Committee of the IEEE Power Engineering Society. IEEE Catalog Number 97TP120-0, N.J., 1997.

[2] M. G. Adamiak, G.E. Alexander, W. Premerlani, “Advancements in Adaptive Algorithms for Secure High Speed Distant Protection”, GE Power Management, Malvern, PA.

[3] L. Wang, “Frequency Responses of phasor-based Microprocessor Relaying Algorithms,” IEEE Trans. Power Delivery, Vol. 14, No. 1, pp 98-105, January 1999.

[4] URPCTM Program, available in software of UR Family of Protection Relays, at the General Electric Industrial Systems website: http://www.geindustrial.com/cwc/products?famid=31.

[5] J.A. de la O, “Complementary Filters for Fault Detection and Phasor Measurement”, Proceedings of the International Conference on Signal Processing, Orlando, FL, USA, Nov 99, disponible en http://www.icspat.com.

[6] J. A. de la O, “Improving the Transient Shortcomings of the Fourier Filter,” Proceedings of the Instrumentation, Systems and Automation (ISA) Conference and Exhibition, Chicago, IL, USA, Oct 2002.

IX. BIOGRAFÍA José Antonio de la O Serna nació en San Pedro, Coahuila, México in 1953. Recibió su grado de Doctor Ingeniero en Telecom, Paris, Francia, en 1982. De 1982 a 1986 fue profesor en el Tecnológico de Monterrey, en Monterrey, México. En 1987, ingresó al Programa Doctoral en Ingeniería Eléctrica en la Universidad Autónoma de Nuevo León, donde fue miembro del Comité Doctoral. De 1988 a 1993 se incorporó al Departamento de Eléctrica del

Politécnico de Yaoundé, Camerún. En 1994 regresó al Programa Doctoral de la UANL, en Monterrey, donde es ahora Profesor Investigador. Es Presidente del IEEE Sección Monterrey y miembro del Sistema Nacional de Investigadores.

of 1

RECENT ADVANCES IN DIGITAL RELAY TESTING

Miguel Gutierrez, Sales & Application Engineer Benton Vandiver III, Technical Director

OMICRON electronics Corp. USA Houston, Texas, USA

Presented to the

6th Symposium Iberoamericano

Monterrey, NL

November 17-20, 2002

INTRODUCTION: Microprocessor based relays are quickly evolving into multifunctional Intelligent Electronic Devices (IED) with built-in high-speed communication capabilities that allow them to perform distributed protection and control functions over a substation Local Area Network (LAN). At the same time they run complex protection and control algorithms designed to provide fast and adaptive fault clearing. An interim substation communication standard has been implemented in many new relays based on the UCA 2.0 [1] specification. This presents new challenges in the area of testing for both commissioning and routine testing. To adequately test such IEDs it is necessary to use specialized testing equipment with similar communication capabilities. Intelligent Electronic Devices (IED) provides data acquisition, protection, metering, and control functions in a cost effective single box solution. Substation protection and controls systems can benefit from vertical optimization using a substation local area network (SLAN) to streamline complex control logic and eliminate wiring which allows significant improvement in the functionality of the protection without any increase in the cost. These devices are beginning to gain widespread acceptance and are recognized as essential to the efficient operation and management of modern substations. Interoperability between IEDs (protective relays) from different manufacturers becomes an overriding requirement for these applications to succeed commercially. This is because it is necessary to achieve substation level interlocking, protection and control function compatibility and application efficiency between the various manufacturers’ IED products. However, microprocessor-based relays (IEDs) with high-speed peer-to-peer communication capabilities can be used in conventional, stand-alone applications, as well as in distributed, communications-based protection systems. Relays using this state-of-the-art technology require a state-of-the-art technology for their testing, since they cannot be tested by conventional test equipment designed for relays with opto-inputs and relay outputs.

of 2

HIGH SPEED PEER-TO-PEER COMMUNICATIONS: A definition of the peer-to-peer communications based on the UCA 2.0 GOOSE is necessary in order to appreciate the schemes described. A GOOSE is a Generic Object Oriented Substation Event and contains the asynchronous reporting of an IED’s protection & logic elements status to other peer devices. Only those IEDs enrolled to receive a GOOSE message will act on the status it contains. This replaces the hard wired control signal exchange between IED’s for interlocking and protection purposes. Consequently, it is mission sensitive, time critical and must be highly reliable. By definition, the GOOSE message is multicast on the substation LAN (or section of the LAN) (Fig. 1) and received by the enrolled IEDs. The associated IEDs receiving the message use the contained information to determine what the appropriate protection response is for the given state. (See Table 1 for the common components of the GOOSE Class Object.) The decision of the appropriate action to GOOSE messages and the action to take should a message time out due to a communication failure is determined by local intelligence in the IED receiving the GOOSE message. For the purposes of this paper we will describe a common Breaker Failure Protection scheme used to trip the adjacent breakers of a distribution bus protection based on GOOSE messages from a feeder protection IED. Considering the importance of the functions performed using GOOSE messages, UCA 2.0 defines very strict performance requirements. The idea is that the implementation of high-speed peer-to-peer communications should be equal to or better than what is achievable by existing copper technology. Thus the total peer-to-peer time should not exceed 4mS. A further important requirement for the GOOSE messages is very high reliability. Since the messages are not confirmed, but multicast, and considering the importance of a message such as Initiate Breaker Failure Protection, there has to be a mechanism to ensure that the receiving IED’s will receive the message and operate as expected. To achieve a high level of reliability, messages will be repeated as long as the change of status persists. To maximize dependability and security, a message will have a time to “live” which will be known as “hold time”. After the hold time expires the message (status) will expire unless the same status message is repeated or a new message is received prior to the expiration of the hold time. The repeat time for the initial GOOSE message will be short and subsequent messages have an increase in repeat and hold times until a maximum is reached. The GOOSE message contains information that will allow the receiving IED to know that a message has been missed, a status has changed and the time since the last status change. In order to achieve high-speed performance and at the same time reduce the network traffic during severe fault conditions, the GOOSE message has been designed based on the idea to have a single message that conveys all required protection scheme information regarding an individual protection IED. It represents a state

Send ing IED

Receiv ing IED

Receiv ing IED

Receiv ing IED

Receiv ing IED

E t h e rn e t

G O O SE

Fig.1 Broadcast GOOSE message

of 3

machine that reports the status of the devices in the IED to its peers. (See Table 2 for the defined use of the standardized IED DNA for GOOSE protection messages.) To allow further customization of the GOOSE messages, individual applications can map other status points (internal control logic) to the User Defined bit pairs (UserSt). A protection IED will typically subscribe to a GOOSE message or messages in order to get indication of the status of the breaker (or breakers in a ring bus or a breaker-and-a-half configuration), as well as any protection signals required by a distributed protection scheme such as a communications based line protection, distribution bus protection or breaker-failure protection in our example. In the high-speed peer-to-peer communications based schemes, the GOOSE messages will replace the hard wired signals of the traditional relay’s opto inputs. At the same time instead of operating one or more relay outputs, the protection IED will send a GOOSE message with its bit-pair status change to trip the breaker or breakers via their associated protection IEDs. BUS BREAKER FAILURE SCHEME FOR FEEDER PROTECTION IED: Figure 2 shows a conventional Feeder Scheme one-line with breaker failure that trips the feeder bus and transformer high side. The individual feeder protection starts a breaker failure timer that checks for the trip and breaker status. If both are still present after the timer expires, it trips the bus through an auxiliary tripping relay. This scheme is repeated in each feeder protection using the same breaker failure elements.

All of the interlocking in this scheme requires point to point wiring, and this is only one component in the overall control scheme. The equivalent scheme implemented with a SLAN and the UCA 2.0 GOOSE would be simplified to the one line representation in Figure 3.

of 4

In order to test the described scheme, specialized test equipment is necessary that not only outputs the required fault simulation of voltage and current waveforms, but also can monitor the GOOSE message from the IED and time-stamp the status bit pairs for analysis. Additionally, the ability to generate simulated GOOSE messages for testing the receiving IEDs becomes a necessity. TESTING PEER-TO-PEER COMMUNICATIONS BASED DEVICES AND SYSTEM: Testing of high-speed peer-to-peer based protection and control systems presents new challenges to protection engineers. For the testing of a conventional microprocessor relay with opto-inputs and relay outputs the test device has to simulate a sequence of dynamically changing conditions through its analog voltage and current outputs. At the same time the test device has to simulate the auxiliary contacts of the breakers and any other devices that may affect the performance of the relay under test. The test device has to monitor the outputs of the relay under test in order to change the states of the simulation and to evaluate the performance of the relay under test. As described earlier, the GOOSE messages replace the hard-wired signals between breakers and other devices in the substation and the tested device. At the same time, it is possible in some cases that the relay uses a combination of hard-wired signals with GOOSE messages providing a subset of the required status indication. A test system for present day peer-to-peer based protection should have the following features in order to allow appropriate testing by simulating the substation and the integrated substation protection and control system environment: • Analog signal simulators that provide currents and voltages to the tested IEDs • Digital signal simulators that represent the changes of breaker status or other remote control

signals simulator as traditional inputs to the IED(s) • Communication simulators that generate GOOSE messages in order to simulate the operation

of other IEDs connected to the substation local area network

of 5

• GOOSE message analyzer that monitors and time-stamps messages received from the IEDs under test in order to evaluate the relay performance/response

• Configuration tools that allow the user to configure the test device to subscribe to the tested IED’s GOOSE and send simulated GOOSE messages of multiple IEDs included in a distributed high speed peer-to-peer communications based protection system

• Test software that allows flexible configuration of the required test sequences and simulations that utilize the above features.

An example of a test setup for high-speed peer-to-peer communications based IED is shown on Figure 4 below.

Figure 4 - High-speed peer-to-peer communications based IED test setup

Since distributed high-speed peer-to-peer communications schemes for more complex system testing (for example a bus protection scheme) include multiple IEDs, this will require multiple test devices for the proper simulation of internal or external fault conditions. The analog outputs of all test devices have to be synchronized in order to apply the fault currents and voltages to the tested IEDs at the same time to realistically simulate the fault condition system wide in the substation. These test devices require appropriate configuration tools that will allow multiple configurations it in such a way that it can simulate GOOSE messages from multiple substation IEDs. At the same time it has to subscribe to specific GOOSE messages from multiple IEDs being tested in order to time-stamp the receiving of a monitored GOOSE message from a tested IED for further analysis and evaluation of the IED performance. Figure 5 shows an example test with the waveforms of the currents and voltages applied to the IED, as well as the operation of conventional relay outputs and a BF trip status from a received GOOSE message.

Test Device

V I Opto-In Relay-Out

Test Computer

V I Relay-Out Opto-In

Tested IED

Ethernet Hub

G O O SE

G O O SE

of 6

Figure 5 – Results From a UCA-Based IED Test CONCLUSIONS: Interoperability between protective IEDs from different manufacturers in the substation becomes a necessity in order to achieve substation level interlocking, protection and control functions and improve the efficiency of microprocessor based relays applications. High-speed peer-to-peer communications between IEDs connected to the substation LAN based on exchange of GOOSE messages can successfully be used for different protection and control applications such as the protection of distribution buses, adaptive protection or load-shedding in substations with varying configurations. Testing of high-speed peer-to-peer based protection and control systems presents new challenges to protection engineers and technicians and will require a strict methodology and new tools to be successful in the future. The UCA 2.0 is not an official standard, but a detailed specification that was evolving towards that goal. Several manufacturers have implemented it in their products to provide proof of concept and to establish a leadership role in the future of system protection and automation.

Wired Trip GOOSE BF Trip Wired BF Trip

of 7

Through cooperation between IEEE and IEC there is an emerging standard that will incorporate the best concepts and even many details of the UCA 2.0 specification. It is the IEC 61850 standard and will be published in its respective parts over the next few years. The part that incorporates the GOOSE specification of UCA 2.0 is 61850-8, which is in final approval at the time of this paper. It will be published in 2003 and available for manufacturers to implement in their products. The main difference between the UCA GOOSE and the IEC version (called a GSSE) is that all status bits are treated as user bits. (No DNA) And analog values are allowed in the message. This allows more flexibility in the possible uses and configurations of the IED. Since all data is treated as objects, the structure and use will be easily determined from the IED’s definition. The general document structure of the IEC 61850 standard is outlined below, a general association of the related parts of UCA to IEC 61850 follows.

IEC 61850:

Document Structure

SCSM A

Stack A

SCSM A

Stack B

SCSM X

Stack X

SCSM Y

Stack Y

WG 11

Part 8

Basic principles

Glossary

General Requirements

System and project management

Communication requirements

SAS Configuration

Basic Communication Structure

WG 10

Part 1

Part 2

Part 3

Part 4

Part 5

Part 6

Part 7

WG 12

Part 9

TC57 WG10,11,12

IEC 61850:

Document Structure

SCSM A

Stack A

SCSM A

Stack B

SCSM X

Stack X

SCSM Y

Stack Y

WG 11

Part 8

Basic principles

Glossary

General Requirements

System and project management

Communication requirements

SAS Configuration

Basic Communication Structure

WG 10

Part 1

Part 2

Part 3

Part 4

Part 5

Part 6

Part 7

WG 12

Part 9

TC57 WG10,11,12

of 8

The benefits and cost advantages of this new standard over previous hard-wiring practices will mean orders of magnitude to the bottom line of the utility or electricity service provider implementing it. Testing will be an important factor in these benefits, as automated testing based on the IEC 61850 standard will become the only way to perform testing, but more important, the only future of protection IED testing. Footnotes: [1] UCA 2.0 is a registered trademark of EPRI. References: 1. Testing of UCA2.0/IEC-61850 Based Protective Relays, B. Vandiver, OMICRON

electronics, Presented at the IEEE PES Summer Meeting, Chicago, IL. July 22, 2002

61850-x-y and UCA 2

UCA2

TC57 WG10,11,12

7-2

7-3

7-4 Compatible data objects

Data Templates for Substations

Abstract Communication

Service Interface (ACSI)

Mapping to MMS8-1 Common Application

Service Model (CASM)

Standard Data Types and

Common Components

Common Class Definitions

GOMSFE

Device Models

Device Models

Device Models

61850-x-y

GOOSE Communications

61850-x-y and UCA 2

UCA2

TC57 WG10,11,12

7-2

7-3

7-4 Compatible data objects

Data Templates for Substations

Abstract Communication

Service Interface (ACSI)

Mapping to MMS8-1 Common Application

Service Model (CASM)

Standard Data Types and

Common Components

Common Class Definitions

GOMSFE

Device Models

Device Models

Device Models

61850-x-y

GOOSE Communications

of 9

APPENDIX Table 1. UCA 2.0 GOOSE Message Common Components of the GOOSE Class Object

Table 2. DNA details of the UCA 2.0 GOOSE Message

Bit Order 00 01 10 11 Bit # Bit Value 0 1 2 3

Pair Definition State State State State 0, 1 1 OperDev Normal Trip Close Invalid 2, 3 2 Lock Out Invalid Normal LO Invalid 4, 5 3 Initiate Reclosing Normal Cancel Auto Reclosing Invalid 6, 7 4 Block Reclosing Normal Cancel Block Invalid 8, 9 5 Breaker Failure Initiate Normal Cancel Initiate Invalid

10, 11 6 Send Transfer Trip Normal Cancel Set Invalid 12, 13 7 Receive Transfer Trip Normal Cancel Set Invalid 14, 15 8 Send Perm Normal Cancel Send Perm Invalid 16, 17 9 Receive Perm Normal Cancel Receive Perm Invalid 18, 19 10 Stop Perm Normal Cancel Stop Perm Invalid 20, 21 11 Send Block Normal Cancel Send Block Invalid 22, 23 12 Receive Block Normal Cancel Receive Block Invalid 24, 25 13 Stop Block Normal Cancel Stop Block Invalid 26, 27 14 BkrDS Between Open Closed Invalid 28, 29 15 BkrPhsADS Between Open Closed Invalid 30, 31 16 BkrPhsBDS Between Open Closed Invalid 32, 33 17 BkrPhsCDS Between Open Closed Invalid 34, 35 18 DiscSwDS Between Open Closed Invalid 36, 37 19 Interlock DS Invalid Non Interlock Interlock Invalid 38, 39 20 LineEndOpen Between Open Closed Invalid 40, 41 21 Mode Test Offline Available Unhealthy 42, 43 22 Event Invalid Normal Event Invalid 44, 45 23 Fault Present Invalid Clear Present Invalid 46, 47 24 Sustained Arc Invalid Normal Present Invalid 48, 49 25 Downed Conductor Invalid Normal Downed Invalid 50, 51 26 Sync Closing Normal Cancel Initiate Invalid 52-63 27-32 Reserved

Common Components Required for GOOSE Name Description Data Type

Sending IED

Sending IED IDENT IDENT

t GOOSE Timestamp BTIME6 SqNum Message Sequence Number INT16U StNum Event Sequence Number INT16U

HoldTim Time to Wait before RS INT16U BackTim Time since Event INT16U

PhsID Identifies Faulted Phases INT16U DNA Protection DNA BSTR64

UserSt User defined Bitstring, used in bit pairs BSTR256

of 10

BIOGRAPHY Miguel Gutierrez Received his Bachelor in electronic and Licenciatura in Protections System from the University of Costa Rica in 1985 and 1988 respectively. He worked as a field protection engineer (later chief engineer) since 1985 at the Costarican Institute of Electricity (ICE), in San Jose, Costa Rica. He developed extensive experience in commissioning, testing, setting of protective relays, transient recorders, control and measurements system in power plants and substations. In 1993 he worked partial time with the University of Costa Rica teaching Protection Systems. In 1994 he worked as a chief engineer of the substation department at the Costarican Institute of Electricity (ICE), in San Jose, Costa Rica. In 1997 he worked for Rochester Instruments Systems (RIS) as a testing engineer in the meter department and in 1999 he joined OMICRON electronics as a Sales & Application Engineer for Latin America with primary responsibilities of sales, training, technical assistance. He is member of the IEEE . Benton Vandiver III received BSEE from the University of Houston in 1979. He began his career with the Substation Division of Houston Lighting & Power, in 1978 engineering relay protection systems for all levels of transmission, distribution, and generation. His main interests were in computer design automation of protection schemes and substation projects. He developed extensive knowledge in the application, setting, testing, modeling, and design of traditional and digital relaying systems used in all types power system protection, control, and monitoring. In 1991 he joined Multilin Corp. as a Project Manager on a team responsible for designing and developing the hardware and software for a new family of utility grade digital relays. In 1995 he joined OMICRON electronics as a Sales & Application Engineer with primary responsibilities of sales, training, and promotion of the revolutionary CMC Universal Test Set to North & South America. He is currently Technical Director for OMICRON electronics Corp. USA in Houston, TX. He is a long time member of IEEE and authored or co-authored many technical papers for various conferences in North America.

Automated testing of protective relays using Advanced Visual Test Software

Cristian Paduraru , Member IEEE

Abstract-The article describes the concept of automatic testing of protective relays in power industry. First ,the increased need for automatic test modules is explained. Next, a detailed example is provided that is “Load Encroachment Characteristic”. Relay settings and testing approach are explained step by step. Finally, the capability of the software is emphasized and conclusions are drawn.

Keywords-Automatic testing, impedance, relays ,

transmission lines

I. INTRODUCTION –THE NEED FOR AUTOMATION

Power Engineers are facing nowadays tremendous challenges because of deregulation of power market. Competition in the market increases sharply and so is the complexity of new equipment used in power industry. Among them, setting, testing and maintaining of microprocessor based relays seem to be the biggest challenges the protection engineers and technicians have to overcome.

In order to face the new challenges protective relays manufacturers are developing new microprocessor based relays each of them with newer functions built-in. Consequently the new numerical multifunction relays require hundreds of settings, protection logics, control and metering functions in order to cover all the specific applications they were designed for. It is a fact that the offer of reliable products in relay market is keeping up with the increased demand. The relay manufacturers being in a rough competition increased year by year the number of new products on the market. What about the USER, the protection technician or engineer? Is he able to cope with this overwhelming number of tests required to be executed in the same timeframe as before? Because the new relays are supposed to make protection engineer’s life easier. How are they going to perform a comprehensive test procedures within a very short period of time mostly a couple of hours? How are they supported by the relay test equipment manufacturers?

To answer to these questions a new concept was defined, developed and deployed in the last couple of years. This is Automatic Test Modules for Protective Relays. The modules run under the state of the art software Advanced Visual Test Software - shortly AVTS - a Windows based software with a very friendly interface and easy to use. 1

Cristian Paduraru is with AVO International in Dallas, TX, 75237, USA

Email: [email protected]

The automatic tests tailored to any specific relay reduces the time to perform a comprehensive set of tests from hours (and sometimes days) to just a couple of minutes. A powerful Test Connections Editor allows the user to make the testing preparation in just minutes. A picture is worth a thousand words [1] .The relay testing procedure is visualized in real time and it takes a couple of minutes (even seconds) instead of hours or days.

The end user is allowed to customize its own modules using test wizards. The tests range from less complex applications such as over/under current/voltage, timing or slope tests to End-to-End testing using Global Positioning System satellite synchronization signals. Moreover , any Digital Fault Recordings or EMTP/ATP simulations are now able to be read and played back [1].

For a better understanding of new testing concept a detailed application is described below.

II. LOAD ENCROACHMENT-CONCEPT AND NEW DEVELOPMENTS

As transmission lines are loaded to higher and higher

levels the relay manufactures were challenged to find out new algorithms to be implemented in order to avoid so called Load Encroachment. In order to prevent undesired trips of a line under heavy load conditions several approaches were mostly used in the past. One was employing a reduced distance element reach setting. Another one was a modified distance element characteristic such as lenticular or elliptic characteristics. The first approach limited the area a remote fault can be detected .The second has a lower sensitivity to resistive faults. Another attempt was to use an offset for the impedance characteristic. This characteristic will cover remote faults but may become insensitive for close faults [2].

To overcome the above shortcomings the new microprocessor based relays manufacturers developed a new concept named “Load Encroachment Characteristic”.

The new characteristic prevents long reaching phase distance elements from tripping under heavy load condition. Basically it blocks phase distance elements for balanced faults when measured positive sequence impedance is within the predefined load region (see fig.1). The major advantage will be a maximized coverage of distance elements so an increased sensitivity of the protection is obtained.

The following application will show the advantages of employing Load Encroachment characteristic. A four zone distance relay was employed.

Load encroachment characteristic is defined by a load impedance in forward direction setting ZLF and reverse direction respectively ZLR. Two defined angle settings will determine the load sectors forward and reverse direction also

(see fig1). These angles are calculated based on minimum acceptable power factor under normal conditions [3].

FIG. 1 Load Encroachment Area

The application is addressing to a transmission line having the following parameters [3] : Nominal line voltage (Unom) : 230 kV Minimum line voltage (Umin) : 220 kV Length : 150 km Line positive sequence impedance amplitude (primary value) (Z1MAG) : 78.03 ohm Line impedance angle (ZANG) : 83.97 o Maximum line load (Smax) : 470 MVA Minimum power factor (PF) : 0.85 Current Transformers Ratio (CTR) : 200/1 Potential Transformers Ratio(PTR) : 2000/1 The minimum Load impedance setting (forward and reverse) will be:

2

2

2

2

min0.9max

220 2000.9 9.272000470

U CTRZLF ZLRPTRS

KV ohmMVA

= = × × =

= × × =

The maximum acceptable Positive Line Angle Forward (PLAF) is:

1 1cos ( ) cos (0.85) 31.8PLAF PF− −= = = o In a similar way the maximum acceptable Negative Angle Forward (NLAF), Positive Line Angle Reverse (PLAR) and Negative Line Angle Reverse (NLAR) are calculated:

-1 -1360 - cos ( ) 360 - cos (0.85)360 - 31.8 328.2 o

NLAF PF= = =

= =

-1 -1

o

PLAR = 180 - cos (PF) = 180 - cos (0.85)== 180 - 31.8 = 148.2

-1 -1180 cos ( ) 180 cos (0.85)180 31.8 211.8o

NLAR PF= + = + =

= + =

The following zones reach (in secondary values) are shown below [3],[4]:

1 0.8 1 / 0.8 78.03 200 / 2000 6.24

Z Z MAG CTR PTRohm

= × × =

= × × =

2 1.2 1 /1.2 78.03 200 / 2000 9.36

Z Z MAG CTR PTRohm

= × × =

= × × =

3 1.25 2 1.25 9.36 11.7 Z Z ohm= × = × =

Since Load Encroachment is going to be employed we are

now able to set an increased reach margin for zone 4 element so the relay protection element will cover an extended area

in case of remote faults occurrence (In our example the additional reach is chosen greater than 50%). Thus,

4 1.54 3 1.5 11.7 18 Z Z ohm= × = × = Please note that the maximum reach can be set without using load encroachment is aproximatively 180% of the line impedance [2],[3] which would have been 14 ohms. The following considerations apply for testing the load encroachment in the forward directions only. A similar approach will be considered for reversed current flow . The Load Encroachment characteristic is now created overlapping the zone 4 characteristic and the load impedance loci characteristic defined by ZLF, PLAF and NLAF (see fig.1). Once the characteristic is known we are able to start testing the relay .It is obvious that testing manually requires a longer time since the voltage and current amount required to test the encroachment zone is difficult to be calculated. Using the software this task becomes a simple one. A predefined characteristic using a powerful theoretical editor has been calculated. This theoretical characteristic will follow precisely the relay “stored” characteristic as it can be seen in fig.2. All the other characteristics (zone 1,2 and 3) are shown for reference only.

FIG. 2 Snapshot of automatic testing of zone 4 in conjunction with Load Encroachment characteristic

The above picture is a snapshot from a real test of the distance relay. The user visualizes in real time the amount continuous calculated and applied to the relay (voltages and currents) and expected and actual (measured) values of the impedances as well. Also the theoretical characteristics and the actual tripping points are shown both with an instantaneous update of a customizable log. The test was performed automatically in less than three minutes and it covers the following acceptance tests all in one: zone 4 reach test, maximum torque angle, zone 4 characteristic test and Load Encroachment characteristic test. III.CONCLUSIONS

A similar approach can be used for any type of the test

required to be performed on any relay. Many other test modules and applications have been developed for AVTS in order to save time and resources for electrical utilities people and are available upon request.

AVTS has a solution for any relay on the market .The concept is coming from the future to face the present and is designed to help the electrical utilities people tackling the upcoming challenges with confidence. IV. ACKNOWLEDGEMENT

The author gratefully acknowledges the contributions of Gary Anderson and Dennis Moon for their support in developing the test procedure described in the paperwork. V. REFERENCES Bulletins: [1] Advanced Visual Test Software-[Online].Available:http://www.avointl.com Papers presented at conferences: [2] Barry Jackson – IEEE/PSRC “Transmission Line Protection Systems Loadability”

Presented to the 55th Annual Georgia Tech Protective Relaying Conference, Atlanta,Georgia,May2-4 ,2001 Instruction Manuals: [3] SEL321/SEL321-1 Instruction Manual – Section 5 – Applications [ Online ].Available at http:www.selinc.com/sel-321.htm Technical Reports: [4] Western Systems Coordination Council Relay Workgroup Report – Application of Zone 3 Distance Relays in Transmission Lines , September 10 ,1997 [ Online ].Available at http:www.wscc.com

VI. BIOGRAPHY Cristian Paduraru was born in Suceava , Romania on 28th February 1969. He graduated from Politechnical Institute of Iasi, Romania. He joined AVO International in 2001 as an Application Engineer in Technical Support Group being in charge with developing custom applications for intelligent protective relays using AVTS (Advanced Visual Test Software). He is also in charge with testing and deploying auto test modules, which allow the customer to evaluate and diagnose the operation of microprocessor, based smart protective relays. Prior position he held was System Engineer with General Electric Automation Services in Dallas, Texas (from 2000 to 2001) where he designed Generator Protection Panels for large combined cycle power plants. He also worked for Romanian Electricity Authority as a Protection Engineer (1993-2000). He has a Bachelor of Science degree in Power Engineering. He is a member of Power Engineering Society of IEEE.

1

CONSIDERACIONES PARA LAS PRUEBAS A LOS RELEVADORES

DIFERENCIALES DIGITALES

Ing. Meliton Ángeles Martínez

Servicios Especializados de Ingeniería de protecciones Eléctricas

[email protected] Oaxaca Oax. Mex.

RESUMEN: Se hace un análisis de las corrientes

que se presentan a una protección diferencial de

transformador, y las consideraciones para realizar

pruebas a un relevador diferencial digital de

porcentaje aplicado a un trasformador de potencia

con conexión delta/estrella, tomando en cuenta el

ángulo de desfasamiento producido en las

corrientes primarias por este tipo de conexión. Esta

protección debe ser estable para todas las fallas

externas que generalmente son monofásicas o

bifásicas y en menor grado las trifásicas.

INTRODUCCIÓN: El cambio de tecnología de los

relevadores de protección de electromecánicos a

digitales, requiere también de nuevas

consideraciones para realizar correctamente las

pruebas de funcionamiento a estos equipos.

Son ampliamente conocidos los desfasamientos

producidos por las conexiones delta/estrella de los

transformadores de potencia y en general de

cualquier transformador trifásico en los voltajes y

corrientes de ambos lados del transformador. Los

relevadores diferenciales electromecánicos de

protección eran monofásicos, por lo que era

necesario hacer algunos arreglos en las conexiones

de los transformadores de corriente para

compensar estos desfasamientos. Esta

compensación exigía además que los

transformadores de corriente fueran de

determinada relación, que permitiera igualar las

corrientes secundarias. El arreglo típico consistía

en conectar los tc´s al inverso de la conexión de

los transformadores de potencia, es decir; en el

lado de la conexión delta del transformador , los

tc´s se tenían que conectar en estrella, mientras

que en el lado de la conexión estrella del

transformador , los tc´s se conectaban en delta.

Esto representaba una ventaja adicional ya que las

fallas monofásicas, (con componentes de

secuencia cero en la corriente) en el lado de la

estrella del transformador se filtraban en la delta

de los tc´s, por lo que no había que darle algún

tratamiento especial a esta corriente en el

relevador diferencial (87T).

Los relevadores digitales ofrecen las ventajas de

que los tc´s pueden conectarse en estrella en

ambos lados del transformador además de que no

requieren relaciones determinadas de los tc´s, es

decir pueden aceptar una amplia gama de

relaciones en los tc´s en ambos lados del

transformador .

En forma general la filosofía a aplicar durante las

pruebas a los relevadores 87T es que deben

responder adecuadamente a las condiciones de

corriente tal como éstas se presentan en el sistema

de potencia ya sea durante condiciones normales

de operación o durante condiciones de fallas. Esto

nos permitirá establecer las consideraciones para

realizar las pruebas.

Otro punto básico de referencia es que se debe

poner especial atención a la estabilidad del

relevador, es decir que no opere para fallas

externas al área protegida, ya que para fallas

internas no existe restricción para su operación,

pero una incorrecta aplicación producirá

operaciones con fallas externas.

-10

-8

-6

-4

-2

0

2

4

6

8

10

-10 -8 -6 -4 -2 0 2 4 6 8 10

I2 = CORRIENTE DE SALIDA

I1 =

CO

RR

IEN

TE

DE

EN

TR

AD

A

ZONA DE

OPERACION

ZONA DE

OPERACION

FIGURA No. 1 CARACTERISTICA TIPICA DE UNA PROTECCIÓN DIFERENCIAL DE PORCENTAJE.

2

DESARROLLO: La representación gráfica de la

característica de operación de los relevadores de

porcentaje es su pendiente que se complementa

con su corriente mínima de operación.

Para comprobar el correcto funcionamiento de este

relevador y determinar gráficamente su zona de

operación se requieren las siguientes pruebas:

a). Corriente mínima de operación.

b). Pendiente.

La pendiente se puede definir como el grado de

tolerancia que el relevador permite en las corrientes

comparadas expresado en por ciento.

m = ( I1 - I2) / ( I1 + I2 ) donde.

m = pendiente.

I1 = Corriente de entrada. I2 = Corriente de salida.

( I1 - I2 ) = Corriente diferencial o de operación.

Expresado como la suma vectorial de

las corrientes. ( I1 + I2 ) = Corriente de restricción, expresado

como la suma de valores absolutos de

las corrientes

Estas características se observan en las figuras

números 1 y 2.

0

1

2

3

0 2 4 6 8 10

I1 + I2 (RESTRICCION)

I1 -

I2

(D

IFE

RE

NC

IAL

)

ZONA DE

OPERACION

ZONA DE

ESTABILIZACION

FIGURA No. 2. PENDIENTE DE UNA PROTECCIÓN DIFERENCIAL DE PORCENTAJE.

ANÁLISIS DE CORRIENTES

Para lograr la estabilización del revelador

diferencial se analizarán las corrientes bajo tres

condiciones diferentes

a) Condiciones normales de operación

b) Condiciones de falla monofásica

c) Condiciones de falla bifásica

Para establecer las condiciones de estabilización

en el relevador analizaremos los tres casos

anteriores a partir de las corrientes primarias para

lo cual tomaremos como ejemplo un transformador

de las siguientes características.

Transformador trifásico

CAPACIDAD: 10,000 KVA

RELACION: 115 / 13.8 KV

CONEXIÓN: Delta / Estrella ( Dy11 )

Inominal: 50.2 / 418.4 A

a).- Estabilización en condiciones normales.

Inominal: en 115 KV = 50.2 A

Inominal: en 13.8 KV = 418.4 A

El sistema balanceado con carga nominal del

transformador resulta con los siguientes corrientes

en el lado estrella.

Ia = 418.4 ∠120° Ib = 418.4 ∠0° Ic = 418.4 ∠-120°

Debido a la conexión Dy11 las corrientes en el lado

delta resultan ser:

IA = 50.2 ∠90° IB = 50.2 ∠-30° IC = 50.2 ∠-150° Asumiendo RTC1 = 100/5 en el lado de 115 KV

y RTC2 = 600/ 5 en el lado de 13.8 KV, en el

relevador tendremos los siguientes corrientes.

a1) De los TC´s lado de 115 KV

iA = IA / RTC1 = 50.2 ∠90° / 20 iB = IB / RTC1 = 50.2 ∠-30° / 20 iC = IC / RTC1 = 50.2 ∠-150° / 20 iA = 2.51 ∠90° A iB = 2.51 ∠-30° A iC = 2.51 ∠- 150° A a2) De los TC´s del lado de 13.8 KV tenemos

ia = Ia / RTC2 = 418.4 ∠120° / 120 ib = Ib / RTC2 = 418.4 ∠0° / 120 ic = Ic / RTC2 = 418.4 ∠-150° / 120 ia = 3.49 ∠120° ib = 3.49 ∠0° ic = 3.49 ∠-120°

Las corrientes primarias y secundarias de

condiciones normales incluyendo sus diagramas

fasoriales se muestran en la figura No. 3

b).- Estabilización en condiciones de falla

monofásica externa.

Para la falla monofásica externa en el lado de la

estrella del transformador consideremos las

siguientes condiciones.

Ia = 2000 ∠0° Ib = 0 Ic = 0

3

Esto se refleja con las siguientes condiciones en el

lado de 115 KV.

IA = 138.5 ∠0° IB = 138.5 ∠180° IC = 0 En el relevador tendremos las siguientes corrientes

b1) De los TC´S de 115 KV

iA = IA / RTC1 = 138.5 ∠0° / 20 iB = IB / RTC1 = 138.5 ∠180° / 20 iC = IC / RTC1 = 0 iA = 6.92 ∠0° iB = 6.92 ∠180° iC = 0

b2) De los TC´S de 13.8 KV

ia = Ia / RTC2 = 2000 ∠0° / 120 ib = Ib / RTC2 = 0 ic = Ic / RTC2 = 0 ia = 16.666∠0° ib = 0 ic = 0

Las corrientes primarias y secundarias en

condiciones de falla monofásica externa al área

protegida se muestran en la figura No. 4

c).- Estabilización en condiciones de falla bifásica

externa.

Consideramos a falla bifásica externa en el lado de

la estrella con los siguientes valores.

Ia = 0 Ib = 2000 ∠ 0° Ic = 2000 ∠ 180°

Lo que se refleja en 115 KV como:

IA = 138.5 ∠ 0° IB = 138.5 ∠ 0° IC = 277∠ 180° En el relevador tendremos las siguientes corrientes

c1) De los TC´S de 115 KV

iA = IA / RTC1 = 138.5 ∠0° / 20 iB = IB / RTC1 = 138.5 ∠0° / 20 iC = IC / RTC1 = 277 ∠180° / 20 iA = 6.92 ∠ 0° iB = 6.92 ∠ 0° iC = 13.8 ∠ 180°

c2) De los TC´S de 13.8 KV

ia = IA / RTC2 = 0 ib = IB / RTC2 = 2000 ∠0° / 120 ic = IC / RTC2 = 2000 ∠180° / 120

ia = 0 ib = 16.67 ∠ 0° ic = 16.67 ∠ 180° Las corrientes primarias y secundarias de la

falla bifásica externa al área protegida se muestran

en la figura No. 5

AJUSTES: Si los ajustes de la protección

corresponden a las consideraciones que se han

hecho, tales como:

Voltajes primario y secundario.

Conexiones y defasamientos.

Relaciones de TC1 y TC2.

Asumiendo una

Pendiente = 25%. y una corriente mínima de

operación = 20 % de la Inom. del Transformador.

PRUEBAS:

CORRIENTE MNIMA DE OPERACIÓN. Para esta

prueba debemos obtener los siguientes resultados.

Lado 115 KV)

Imin op = Inom ( 0.2 ) / RTC1 Imin op = 50.2 ( 0.2 ) / 20 Imin op = 0.5 A Lado 13.8 KV

Imin op = Inom ( 0.2 ) / RTC2 Imin op = 418.4 ( 0.2 ) / 120 Imin op = 0.7 A

PENDIENTE: Observando las corrientes obtenidas

en las figuras 3, 4 y 5, estas condiciones se

pueden reproducir usando equipos de pruebas

para comprobar la estabilización.

Las conexiones para las pruebas de estabilización

y pendiente son las mismas, solo se consideran los

fasores de corriente tal como se determinaron para

lograr la estabilización.

Para determinar la pendiente se requiere, a partir

de la estabilización incrementar una de las dos

corrientes dejando la otra fija, de esta manera se

obtendrá una diferencia entre ambas corrientes, la

cual, al alcanzar el porcentaje ajustado producirá la

operación de la protección.

En este caso se determinarán los valores

esperados para la pendiente de 25 %.

CALCULO DE LOS VALORES ESPERADOS

PARA LA PENDIENTE

m = 0.25

m = ( I2 – I1) / (I1 + I2)

Despejando I2 tenemos que:

I2 = I1 [ (1-m) / (1 + m)] ó

I1 = I2 [ (1+m) / (1 - m)]

4

Falla trifásica: tomando como base la corriente

nominal del transformador.

Valores para estabilización

I1 = iA = 2.51 = 1.0 pu. I2 = ia = 3.49 = 1.0 pu.

Valores de operación

I1 = 1.0 pu I2 = I1 [(1-0.25) / (1+0.25)] I2 = I1 ( 0.75 / 1.25) I2 = 0.6 I1 ó I1 = 1.666 I2

De donde para igualar a I1 se requiere que I2 se

multiplique por 1.667 de donde

I2 de operación = 1.667 (3.49) = 5.81 A

Condiciones de operación por pendiente con falla

trifásica.

I1 I2 iA = 2.51 ∠90° ia = 5.81 ∠90°

iB = 2.51 ∠-30° ib= 5.81∠0°

iC = 2.51 ∠-150° ic= 5.81∠-120°

Falla monofásica tomando como base la corriente

nominal lado delta del transformador.

IA = 50 ∠0° Ia = 722 ∠0° iA = - iB = 2.5∠0° ia = 6.0∠0° I1 = iA = 2.5 = 1.0 pu I2 = ia = 6.0 = 1.0 pu Por la pendiente

I1 =1.667 I2 = 1.667 ( 6.0∠0°) = 10 ∠0° A

Condiciones de operación por pendiente por falla

monofásica.

I1 I2

iA = 2.5 ∠0° iB = 2.5 ∠180°

ia = 10∠0°

Falla bifásica tomando como base la corriente

nominal del lado delta del transformador

IA = IB = 50 ∠0° Ia = - Ib = 722 ∠0° iA = iB = 2.5 ∠0° iC = 5 ∠180° ia = -ib = 6.0 ∠0° I1 = iA = 2.5 = 1.0 pu I2 = ia = 6.0 = 1.0 pu

Por la pendiente

I1 = 1.667 I2

∴ I2 = 1.667 ( 6<0° ) = 10 ∠0° A Condiciones de operación por pendiente por falla

bifásica.

I1 I2 iA = 2.5 ∠0° ia = 10∠0° iB =2.5∠0° iC = 5∠180°

Ib = 10∠-180°

CONCLUSIONES:

1.- Para la realización de las pruebas de pendiente

en condiciones de corrientes balanceados en

ambos lados del transformador se requieren 6

fuentes de corriente defasables, tres para I1A , I1B

e I1C y tres para I2a , I2b e I2c.

2.- Para realizar las pruebas de pendiente en

condiciones de falla,monofásica y bifásica, se

pueden realizar con solamente dos fuentes de

corriente defasadas 180° una de la otra.

3.- Las magnitudes de corriente de estabilización

se pueden determinar a partir de las conexiones y

defasamientos del transformador.

Análisis de las corrientes en la protección

diferencial aplicado a un transformador trifásico.

BIBLIOGRAFÍA:

La protección diferencial en los transformadores de

potencia. Meliton Angeles M. Tesis profesional.

Applied Protective Relaying. Westighouse Electric

Corp.

ING. MELITON ANGELES MARTINEZ.

BIOGRAFIA:

Nació en San Pedro Quiatoni

Oaxaca en 1953, Egresado del

Instituto Tecnológico de Oaxaca

en 1980 graduado como

Ingeniero Industrial Electricista.

Trabajó en la Comisión Federal

de Electricidad de Nov/1974 a

mayo/1992 desempeñando

diferentes cargos en la Región de Transmisión

Sureste. Miembro del Comité Nacional de

Protecciones. Instructor de diversos cursos sobre su

especialidad. Autor de varios artículos técnicos y

ponente en diversos eventos. En 1991 merecedor de

la Medalla ADOLFO LOPEZ MATEOS al mérito

electricista por su destacado desempeño dentro de la

C.F.E.

Actualmente se desempeña como consultor e

instructor en la especialidad de protección y medición

para México y Centroamérica a través de la empresa

Servicios especializados de Ingeniería de

Protecciones Eléctricas, en la ciudad de Oaxaca Oax.

México.

1

CONSIDERACIONES PARA LAS PRUEBAS DE PROTECCIONES

DIFERENCIALES DE BARRAS


Servicios Especializados de Ingeniería de protecciones Eléctricas

[email protected] Oaxaca Oax. Mex.

RESUMEN: Por las implicaciones que sobrelleva la

operación de una protección 87B, y los bajos índices

de operaciones correctas de estas, un alto porcentaje

de protecciones de buses se encuentran fuera de

servicio por la desconfianza en la protección en sí, o

en los equipos complementarios o en las pruebas de

puesta en operación. En este documento se presenta

el principio de operación y se proponen algunas

formas de pruebas que podrían ayudar a disminuir

esta desconfianza en este tipo de protecciones.

INTRODUCCIÓN: El uso de la Protección Diferencial

de barras (87B) es el método más recomendable para

protección de barras colectoras o buses.

Existen varios tipos de esta protección entre los que

podemos mencionar los siguientes:

1.- Esquema 87B con acopladores lineales Fig. DB-01.

2.- Esquema 87B de alta Impedancia. Figura DB-02.

3.- Esquema 87B de porcentaje. Figura DB-03.

Este último es el que se ha generalizado su aplicación

por su confiabilidad, habiendo sido en sus inicios de

tipo electromagneto-mecánico, luego estático y

finalmente digital, y el uso de transformadores

mezcladores, convertidores de 3 fases a 1 viene a

reducir el numero de elementos de detección de la

corriente diferencial a solamente una, para la

protección de un BUS con tres fases y varios

alimentadores.

Principio de operación.- La protección diferencial

compara la sumatoria de las corrientes que entran al

Bus con la sumatoria de las que salen de él. En otras

palabras, realiza la suma vectorial (ec. 01) de todas

las corrientes que se conectan a la barra, y el

resultado de la suma debe ser igual con cero en

condiciones normales. Esto se puede observar en la

figura DB-04.

Sin embargo para una falla interna, el resultado de

esta suma vectorial resultará una magnitud muy

grande ya que todas las corrientes tendrán el mismo

ángulo, como se observa en la figura DB-05.

L1 L4 L2 L3

87B

DB-02.- Esquema 87B de alta impedancia

L1 L2 L3 L4

87B

R1 R3 R4 R2

OP

Figura DB-03: Esquema 87B de porcentaje

L1 L4 L2 L3 87B

DB-01.- Esquema 87B con acopladores lineales

2

Como es sabido, la protección diferencial de

porcentaje representa grandes ventajas sobre los

otros tipos de protecciones diferenciales debido a que

se puede ajustar el grado de desbalance de corrientes

para su operación en donde se introduce el concepto

de pendiente.

Pendiente (K): El significado práctico de ésta sería,

grado de tolerancia a las desigualdades en las

corrientes, y su expresión matemáticamente puede

variar de un fabricante a otro, pero de manera general

se puede representar como:

K = Fuerza de operación / Fuerza de restricción.

Donde: La fuerza de operación está definida por la

corriente diferencial, y la fuerza de restricción es la

que impide la operación del relevador, también

llamada fuerza de estabilización, y puede variar en su

definición según el fabricante.

Fuerza de operación = I dif = 01.

(corriente diferencial)

En este documento definiremos que:

Fuerza de restricción = Irest = 02.

(Estabilización)

Siendo m el número de ramas conectadas al bus.

Haciendo m = 2, las ecuaciones 1 y 2 se reducen a:

Idif = ⏐I1⏐- ⏐I2⏐ 03

Irest = ⏐I1⏐+ ⏐I2⏐ 04

K = ( Idif / Irest ) K = (⏐I1⏐- ⏐I2⏐ ) / (⏐I1⏐+ ⏐I2⏐ ) 05

La representación grafica de la pendiente (K) se

muestra en las figuras DB-06 y DB-07.

0

1

2

3

0 2 4 6

I1 + I2 (RESTRICCION)

I 1 -

I2

(DIF

ER

EN

CIA

L) ZONA DE

OPERACION

ZONA DE

ESTABILIZACIO

N

Figura DB-06: Grafica 1 de pendiente

-10

-8

-6

-4

-2

0

2

4

6

8

10

-10 -8 -6 -4 -2 0 2 4 6 8 10

I2 = CORRIENTE DE SALIDA

I 1 =

CO

RR

IEN

TE

DE

EN

TR

AD

A

ZONA DE

OPERACION

ZONA DE

OPERACION

Figura DB-07: Grafica 2 de pendiente.

m →

∑ I n=1

m

∑ ⏐I ⏐ n=1

87B

R1 R2

Is2

Is2 Is1

Is1

IsDIF = Is1 + Is2

Ip1 Ip2

IF

Figura DB-05: Corrientes para falla interna en una protección 87B.

IF 87B

R1 R2

Is2

Is2 Is1

Is1

IsDIF =0

Ip1 Ip2

Figura DB-04: Corrientes para falla externa en una protección 87B.

1

2

3

3

Transformadores mezcladores: Estos

transformadores son auxiliares que permiten obtener

una sola salida de corriente de las tres fases de

entrada como se muestra en la figura DB-08, y para

cada tipo de falla representara una relación diferente

de acuerdo con la tabla TB-1 obtenida a partir de esta

figura.

Generalmente el transformador puede contener varios

devanados primarios, con las cuales se pueden hacer

combinaciones para obtener diferentes relaciones,

procurando conservar números de vueltas múltiplos de

1, 2, 3, para las diferentes fases.

Tabla TB-1: Numero de vueltas y relación del TC

mezclador para cada tipo de falla.

FALLA AN BN CN AB BC CA 3F

Np 10 20 30 10 10 20 17.3

Ns 300 300 300 300 300 300 300

Relación 30 15 10 30 30 15 17.3

En la figura DB-09 se presenta el diagrama

esquemático simplificado de una protección diferencial

de barras del tipo estático, usando los

transformadores mezcladores.

PRUEBAS A LAS PROTECCIONES

DIFERENCIALES DE BARRAS.

Podemos clasificarlas en dos categorías:

1.- Pruebas de operación y

2.- Pruebas complementarias.

Las pruebas de operación determinaran la habilidad

de la protección en sí para detectar cualquier tipo de

falla y discriminar si éstas representan operación o

bloqueo del 87B.

Las pruebas complementarias aseguran el correcto

funcionamiento de los elementos auxiliares que

intervienen en el esquema para garantizar la

confiabilidad del esquema completo, entre estas

pruebas se deben contemplar las siguientes.

• Pruebas de Transformadores de corriente:

Relación, Saturación y polaridad.

• Medición de las cargas de los

Transformadores de corriente.

• Verificación de los circuitos de corriente.

• Verificación de los circuitos de disparos y

bloqueos.

• Verificación de posiciones de cuchillas y sus

auxiliares para conectar las señales

secundarias a la sección correspondiente de

la 87B.

• Pruebas de faseo.

PRUEBAS DE OPERACION:

De acuerdo a los tipos de falla que se pueden

presentar en el sistema de potencia, es necesario

simular las indicadas en la tabla TB-1.

Estas simulaciones han de realizarse para comprobar

el comportamiento del esquema bajo dos condiciones.

a) Falla interna (dentro de la zona protegida).

b) Falla externa (fuera de la zona protegida).

Simulación de falla interna: Es premisa fundamental

de que la protección 87B deba operar bajo cualquier

circunstancia para falla interna, pero además, bajo

esta circunstancia la protección operará aunque la

misma no estuviera correctamente conectada. Por lo

tanto para simular fallas internas se deberá realizar la

prueba de la corriente mínima de operación para

todos los tipos de falla mencionados en la tabla TB-1,

en cada rama del bus. Los valores de esperados de

operación son los reportados en la tabla TB-2, y la

trayectoria de esta prueba se muestra con el número

1 en la figura DB-06.

Simulación de falla externa: Esta es la prueba de

mayor importancia ya que se trata de comprobar que

la protección no operará para fallas externas.

Esta es la prueba de pendiente y se realiza utilizando

dos fuentes de corriente, mediante las cuales se

inyectan corrientes a dos de las ramas conectadas al

bus. Asumiendo que las RTC de los TC´s principales

son uniformes.

1. Se selecciona el tipo de falla externa, para

identificar las terminales de conexión de las corrientes.

2. Se inyectaran las corrientes de la misma magnitud

pero una opuesta de la otra.

3. Se incrementaran simultáneamente las corrientes

hasta alcanzar dos veces la corriente mínima de

operación, de acuerdo con la tabla TB-2, sin que

opere la protección.

IA

IC

IB

87B

300

30 10 20

Figura DB-08.- Conexiones y relaciones de un transformador mezclador

4

4. Una vez alcanzado el valor del punto anterior (3),

este valor se presenta en la tabla TB-3, se deja fijo el

valor de corriente en la rama de referencia “ I1 ref. ” y

se incrementa la corriente de la segunda rama “ I2

esp. ”.

5.- La protección operará al valor de la pendiente

ajustada con los valores de corriente reportados en la

tabla TB-3. La trayectoria de esta prueba se muestra

en la figura DB-06 con el número 3.

A partir de haber alcanzado dos veces la corriente

mínima de operación, también se puede dejar fija la

corriente de referencia y disminuir la corriente de la

segunda rama hasta entrar a la zona de operación,

esta trayectoria se muestra en la figura DB-06 con el

número 2.

Se recomienda realizar esta prueba alimentando las

corrientes de prueba en las terminales de entrada al

tablero, de las corrientes provenientes de los TC´s. De

esta manera se verifica el cableado interno y las

polaridades correctas de los transformadores

mezcladores. Se deben comprobar las tres fallas a

tierra externas y las tres fallas entre fases externas,

puede omitirse la falla trifásica. En la figura DB-10 se

muestra la manera de realizar esta prueba.

Determinación de valores esperados para una

pendiente determinada, para este caso usaremos una

pendiente del 50 % ( K = 0.5 ). De la ecuación 05 si

despejamos I1 tendremos que:

I1 = I2 [( K+1 ) / (K-1) ] 06

Al hacer K = 0.5 para sustituir en 06 tendremos.

I1 = I2 [( 0.5+1 ) / (0.5-1) ]

I1 = I2 [( 1.5) / (-0.5) ]

I1 = -3 * I2

Con esta formula calculamos los valores de la tabla

TB-3, y el signo negativo de I2 Indica la dirección

invertida de esta.

Utilizando el secundario de un transformador

mezclador para operar un elemento ajustado para

activarse al valor de 0.25 A tendremos diferentes

valores de corriente mínima antes del transformador

mezclador para las fallas de la tabla TB-1, estos

valores esperados de corriente se muestran en la

tabla TB-2.

30 10 20

300

30 10 20

300

30 10 20

300

52 – L1 52 - L2 52 -L3

DB-09.- Esquema diferencial de porcentaje estático, para tres

alimentadores

87B

5


Relación 30 15 10 30 30 15 17.3

I esp.(A) 7.5 3.75 2.5 7.5 7.5 3.75 4.33

Isec (A) 0.25 0.25 0.25 0.25 0.25 0.25 0.25 Tabla TB-2: Corriente mínima de operación esperada.


Relación 30.0 15.0 10.0 30.0 30.0 15.0 17.3

I1 ref.(A) 15.0 7.5 5.0 15.0 15.0 7.5 8.67

I2esp.(A) 45.0 22.5 15.0 45.0 45.0 22.5 26.0 Tabla TB-3: Corriente de operación esperada en la

prueba de pendiente

Para probar el esquema en su totalidad, se elige una

de las ramas como referencia, la cual se compara

contra todas las demás.

PRUEBAS COMPLEMENTARIAS:

• Pruebas de Transformadores de corriente:

Relación, Saturación y polaridad. Es muy importante

contar con los resultados de estas pruebas para

evaluar las fallas más severas y comprobar que no se

producen saturaciones en los transformadores de

corriente.

• Medición de las cargas de los Transformadores

de corriente. Esta prueba permite conocer la

impedancia en ohms que representa la carga

conectada a los bornes secundarios de los

transformadores de corriente que influyen en el

fenómeno de la saturación. Esta prueba consiste en

hacer circular una corriente en cada circuito de

corriente (desde los TC´s) y medir la diferencia de

potencial producida.

• Verificación de los circuitos de corriente. Consiste

en hacer circular una corriente en el primario del TC y

verificar en los circuitos secundarios el reflejo de esta

corriente.

• Verificación de los circuitos de disparos y

bloqueos. Esta prueba se realiza preferentemente

durante la puesta en servicio de subestaciones

nuevas, sin embargo durante la operación de

subestaciones en servicio es necesario realizar estas

pruebas; pero se hace un tanto difícil cuando no se

cuenta con aditamentos que permitan bloquear los

disparos del 87B sobre los interruptores.

Se sugiere la inclusión de bloques de conmutación

que permitan los bloqueos en forma selectiva de todos

los disparos a los interruptores conectados a la barra.

• Verificación de posiciones de cuchillas y sus

auxiliares para conectar las señales secundarias a la

sección correspondiente de la 87B. Esta verificación

se hace repitiendo las pruebas de corriente mínima de

operación, aplicando las diferentes posibilidades de

posiciones de las cuchillas principales, y se

recomienda realizar esta prueba operando las

cuchillas principales.

• Pruebas de faseo. Esta es una de las pruebas

determinantes para una correcta operación de la

protección 87B.

30 10 20

300

30 10 20

300

30 10 20

300

DB-10.- Diagrama de conexiones de prueba de PENDIENTE para falla

entre fasers B-C.

87B

FUENTE 2 VARIABLE

FUENTE 1 VARIABLE

6

Dependiendo de las circunstancias puede hacerse de

dos maneras.

1.- Durante la puesta a punto. Para hacer circular una

corriente por el primario de los TC´s, se puede

aprovecha la carga que representa alguno de los

transformadores de potencia de la subestación. Se

alimenta el transformador desde alguna de las ramas

de la barra protegida con bajo voltaje, con las

terminales opuestas del transformador en corto-

circuito. La corriente que toma el transformador

deberá reflejar una corriente secundaria suficiente

para realizar las mediciones angulares.

Se toma la rama que alimenta el transformador como

la referencia para compararla contra todas las demás

ramas de la barra, esto implica mover la alimentación

a cada una de las ramas conectadas a la barra.

Se han de comparar los ángulos de las corrientes por

fase, I de referencia contra I de la rama comparada.

El ángulo entre ambas deberá ser 180°.

Si la protección permite tomar mediciones de los

valores de operación y restricción, se recomienda

registrar estos valores, el valor de operación deberá

ser mínima comparada con la de restricción para

polaridades correctas.

2.- En operación. Se requiere hacer un análisis del

comportamiento de las cargas de todas las ramas

conectadas al bus, que bajo estas condiciones su

suma vectorial deberá ser cero.

A partir de esta cargas obtener los fasores de

corrientes por fase, para determinar los ángulos

esperados de las corrientes secundarias. La figura

DB- 11 muestra un diagrama posible.

Realizar las mediciones de los ángulos para comparar

con los ángulos esperados.

Estas mediciones deberán realizarse para todas las

ramas sin que se presenten variaciones significativas

en las magnitudes y ángulos de las cargas.

Realizar, si es posible, las mediciones de cantidades

de operación y restricción. El valor de operación

deberá ser mínima comparada con la de restricción

para polaridades correctas.

Para facilitar esta prueba es deseable contar con

bloques de conmutación que permitan derivar las

señales de corrientes en cada rama.

CONCLUSIONES Y RECOMENDACIONES

1.- Evaluar estadísticamente el comportamiento de las

protecciones diferenciales 87B, para determinar las

causas de las operaciones incorrectas y las razones

por las que algunas no están en servicio.

2.- Promover la capacitación del personal sobre el

funcionamiento y pruebas de estas protecciones.

Para evitar operaciones incorrectas por maniobras.

3.- Realizar todas las pruebas durante la puesta a

punto ya que operación se hace mas difícil realizar las

mismas.

4. Incluir en los esquemas además de un conmutador

de bloqueo de la protección 87B, bloques de

conmutación para poner en corto-circuito las

corrientes de los TC´s, derivación de corrientes para

faseo e inyección de corrientes de pruebas. Además

de un bloque para desconexión de las señales de

disparo a todos los interruptores de potencia.

BIBLIOGRAFÍA:

El arte y la ciencia de la protección por relevadores. C.

Russell Masson.

Manual de operación de protección diferencial de

barras 7SS10. SIEMENS.

Applied Protective Relaying. Westighouse Electric

Corp.


BIOGRAFIA: Nació en San Pedro

Quiatoni Oaxaca en 1953, Egresado

del Instituto Tecnológico de Oaxaca

en 1980 graduado como Ingeniero

Industrial Electricista.

Trabajó en la Comisión Federal de

Electricidad de Nov/1974 a

mayo/1992 desempeñando

diferentes cargos en la Región de

Transmisión Sureste. Miembro del Comité Nacional de

Protecciones. Instructor de diversos cursos sobre su

especialidad. Autor de varios artículos técnicos y

ponente en diversos eventos. En 1991 merecedor de

la Medalla ADOLFO LOPEZ MATEOS al mérito

electricista por su destacado desempeño dentro de la

C.F.E.

Actualmente se desempeña como consultor e

instructor en la especialidad de protección y medición

para México y Centroamérica a través de la empresa

Servicios especializados de Ingeniería de

Protecciones Eléctricas, en la ciudad de Oaxaca Oax.

México.

I1 I2

I3

I5

I4

Figura DB- 11. Diagrama fasorial de las corrientes de una fase de las

diferentes ramas para una barra.

RESUMEN: Se presenta un equipo simulador de fallas eléctricas, sencillo compacto y económico para verificar en forma dinámica, las condiciones de operación de los relevadores de protección. Complementado con una fuente dual de corriente y / o voltaje para pruebas de relevadores de alto burden y transformadores de corriente. INTRODUCCION: Con el avance tecnológico de los últimos años se han modificado algunas características de los elementos de protección y de los elementos de pruebas como son los siguientes:

Alta confiabilidad en los relevadores de protección basados en P.

Alta precisión y flexibilidad en los equipos de prueba de estos relevadores requiriendo con ello software y equipo de computo.

La operación de los relevadores pasa de estar basado en 100 % hardware a un 10 % hardware y 90 % software.

El mantenimiento de los relevadores basados en

P tiende a cero.

Toma mayor relevancia la programación lógica de los esquemas mas que la precisión en el alcance de los ajustes.

En la práctica existen varios factores que introducen errores que dificultan alcanzar una alta precisión en la operación de los relevadores.

Debido a que los relevadores nuevos supervisan todos las variables del sistema de potencia, las pruebas a los mismos deben realizarse simulando condiciones reales de fallas.

La precisión de los relevadores digitales se comprueba en los laboratorios durante su evaluación inicial.

Los costos de inversión en equipos de pruebas resultan muy altos, y para realizar pruebas dinámicas se requiere un software y una PC.

Altos costos por mantenimiento de los equipos de prueba, cuando éstos tienen que ser transportados constantemente de una instalación a otra y por caminos no siempre adecuados.

PRESENTACIÓN: Por lo anterior, y para apoyar al personal que realiza las actividades de puesta en servicio, mantenimiento y atención de emergencias en la operación de los sistemas eléctricos, hemos desarrollado el XÄID-03, (SIMULADOR TRIFÁSICO DE FALLAS, VERIFICADOR DE RELEVADORES DE PROTECCIÓN), y el TIECH-01, (FUENTE DE CORRIENTE y / o VOLTAJE PARA PRUEBAS DE RELEVADORES Y TRANSFORMADORES DE CORRIENTE) CARACTERÍSTICAS GENERALES DE DISEÑO: Compacto Resistente Práctico Versátil Fácil de usar A prueba de lluvia Fácil de transportar Económico Diseñados para el trabajo de campo y considerando las condiciones reales de operación y sin dependencia de una PC.

EQUIPOS PARA VERIFICACION DE RELEVADORES DE PROTECCIÓN y TRANSFORMADORES DE CORRIENTE: SIMULADOR TRIFÁSICO DE FALLAS XÄID-03 Y TIECH-01

Ing. Melitón Ángeles Martínez, Armando Jiménez Rojas

Servicios Especializados de Ingeniería de protecciones Eléctricas [email protected] Oaxaca Oax. Mex.

FIGURA No. 1

CARACTERÍSTICAS DEL XÁID-03: Generalmente los relevadores de protección están diseñados para proteger contra cualquier tipo de fallas; sin embargo se da especial importancia a la detección de las fallas monofásicas y a las de fase-fase, ya que los otros tipos de fallas, son cubiertas por los mismos elementos ya mencionados. Por lo tanto para minimizar componentes y costos, el equipo XÄID-03 cuenta con los siguientes elementos. FUENTES DE VOLTAJE. (Figura 2), Cuenta con: - Tres salidas de voltajes de prueba VA, VB, VC, ajustables en magnitud mediante tres selectores de rango y tres elementos de ajuste fino, con un valor máximo de 70 Volts de fase a tierra. -Selectores de tipo de falla: se usa para la simulación de fallas de fase a fase o de fase a neutro.

Selector de volts: De acuerdo a su posición obtenemos la lectura de los voltajes en la pantalla Volts. Estas mediciones pueden corresponder a voltajes de fase a neutro o voltajes entre fases o voltaje de salida de la fuente I/v. CONTROLES: �“VOLT inicio�”, de color rojo, permite obtener salidas de voltajes normales, (condiciones de prefalla), esto se puede comprobar mediante el voltímetro en las posiciones VAB, VBC, o VCA. �” I/v inicio�”: (botón de inicio de inyección de corriente, de color azul) figura 2. Este botón opera el circuito de energización de la fuente de corriente/voltaje I/v. Si el botón �“VOLTinicio�” se pulsa previamente, la operación de �“I/v inicio�” permite que se obtengan los voltajes ajustados de simulación de falla en los

bornes de voltaje de salida al mismo tiempo que la señal de corriente se inyecta. �“PARO�” de color verde, permite interrumpir las salidas de las fuentes de voltaje y de corriente �“CTL�”: Selector que permite escoger entre control de magnitud o control de ángulo para la fuente I/v, �”SEQ�”: Selector de la secuencia para falla dinámica, 1F0 o 1F1. En la posición 1F1, nos permite simular fallas donde: primero se tienen voltajes nominales de prefalla, enseguida se aplican condiciones de falla para voltajes y corriente y luego de operar la protección se restablecen las condiciones normales. En la posición 1F0, nos permite simular fallas donde: primero se tienen voltajes nominales de prefalla, enseguida se aplican condiciones de falla para voltajes y corriente y luego de operar la protección se desconectan. Selector �“I/v�”: Este selector en la posición PERM nos permite mantener la señal de corriente permanentemente en los bornes de salida ignorando el botón �“I/v INICIO�” y el paro del cronómetro FUENTE DE CORRIENTE-VOLTAJE. Cuenta con: - Una salida de (corriente-voltaje, figura 3) I/v, conmutable ca/cd, con un selector para obtener valores según la tabla 1. Se tiene un control de magnitud y un ajuste fino para la corriente.

- Siete bornes de conexión para selección del ángulo de la corriente de prueba, con posibilidad de ajuste a cada 30 grados. - Salida de voltaje adicional, para interconexión con una fuente adicional de corriente, incluida en nuestro equipo DAIZ (generador de frecuencia variable), esta

Corriente A 0.8 5.0 20.0SALIDA I/v Voltaje V 140 24.0 6.0

TABLA 1

FIGURA 3

FIGURA 2

opción permite probar relevadores diferenciales usando el XÄID-03, y el DAIZ conjuntamente. CRONOMETRO: Figura 4, Con resolución de un milisegundo, para medición de tiempos de respuesta de los equipos a verificar. Arranque interno: con la corriente o arranque externo con contacto seco. Paro externo: con contacto seco o con voltaje de 90 a 300 V ca/cd. - Un Voltímetro con selector para medición de las salidas de voltaje en las diferentes posibilidades de medición, incluyendo la salida I/v. - Un Amperímetro para medición de la corriente de salida de I/v. - Una salida de 125 Vcd, 30 VA. - Un contacto tomacorriente, para conexión de algún equipo auxiliar CARACTERÍSTICAS DEL TIECH-01: Este equipo incluye las siguientes funciones y los instrumentos necesarios para realizar pruebas a los TC´s y relevadores de corriente y / o voltaje. El equipo se muestra en la figura 5 y consta de 3 fuentes con las opciones mostradas en la taba 2. ALIMENTACIÓN: Una fase, 127 Vca, 10 A FUENTE VA´s OPCION RANGO

1 50 o 100 A 1 500 2 10 A, 200, 500 Vca 3 280, 700 Vcd 1 20 A 2 120 2 10 A, 5 A, 140 Vca 3 10, 5 A c/Harmónica

4 5 Acd, 190 Vcd 3 30 1 125 Vcd

TABLA 2 -La FUENTE 1 consta de un control principal de magnitud, un control de ajuste fino y un selector de derivaciones para obtener los valores mostrados en la tabla 3..

I max 100 50 10 2.5 1.0 A V max 5 10 50 200 500 Vca V max - - 70 280 700 Vcd

TABLA 3 Incluye un selector para obtener CA o CD en cualquiera de los tap´s con excepción de las salidas de 50 A y 100 A. La selección de 50 A o 100 A se hace con puentes externos.

-La FUENTE 2 consta de un control principal de magnitud, un control de ajuste fino y un selector de derivaciones para obtener valores según la tabla 4.

I max 20 10 5 0.8 A V max 6 12 24 140 Vca V max - 16 34 190 Vcd

TABLA 4 Incluye un selector para obtener CA o CD en cualquiera de los tap´s con excepción de la salida de 20 A. Además de un selector para generar corrientes ARMÓNICAS en la salida de CA.

-La salida 3 consta de una única salida de 125 Vcd. Para usarse como fuente de alimentación de los equipos bajo prueba.

INSTRUMENTOS: -Un CRONÓMETRO para medición de tiempos de respuesta de los equipos a verificar con resolución de 1ms. El paro del cronometro puede ser mediante contacto seco o mediante aplicación de Vca o Vcd en un rango de 90 a 300 V.

FIGURA 5

FIGURA 4

FIGURA 6

-Un AMPERÍMETRO CA: asociado con la Fuente 1

. -Un AMPERÍMETRO / VOLTIMETRO asociado a la Fuente 2 para medir Vca, Vcd, Ica, con un selector para mediciones internas o externas. -Un PROBADOR DE POLARIDAD de Transformadores aplicable en cualquier transformador con RT hasta de 1000 /1

APLICACIONES: - El verificador de relevadores de protección XÄID-03 es un equipo con el cual se pueden simular: Fallas monofásicas: AN, BN y CN. Y Fallas bifásicas: AB, BC y CA. - El XÄID-03, puede simular condiciones muy cercanas a las reales de falla, lo que permite realizar pruebas dinámicas a los relevadores, en forma manual. Se pueden simular TP´s ubicadas en línea o en barras.

En la figura 9 se presenta un oscilograma de una falla monofásica en la que se aprecia la capacidad del XÄID-03 de generar fallas dinámicas. Esto también se genera para fallas entre fases. Consecuentemente se pueden verificar la operación de relevadores aplicables en plantas de generación y subestaciones de transmisión, distribución los que se muestran en la tabla 5. En la figura 12 se presenta una vista frontal del XÄID-03

1 Sobrecorriente, no direccional 50, 51 y 46 2 Sobrecorriente direccional y de

potencia. 32, 67F y 67N

3 De distancia y / o impedancia 21, 40G 4 De voltaje y verificación de

sincronismo 25, 27 y 59

5 Equipos de medición V, A, F.P. WM, WHM

6 Equipos de registro R.F. TABLA 5

- El verificador de relevadores TIECH-01 es un equipo con el cual se pueden verificar la operación de los equipos de la tabla 6: En la figura 10 se muestran los resultados de una prueba realizada a un transformador de corriente de clase C400 y la figura 11 usando el TIECH-01 durante una prueba de campo.

FIGURA 7

FIGURA 8

FIGURA 9

TABLA 6

1 Sobrecorriente no direccional 50, 51 y 46 2 De voltaje de CD y CA 27, 59 3 Diferenciales (pick up y

pendiente) 87T, 87G, 87B

4 Transformadores de corriente (Relación, Saturación, Polaridad)

5 Transformadores de voltaje (Relación, Polaridad )

FIGURA 10 CURVA DE SATURACION DE TC

1

10

100

1000

0.01 0.1 1

CORRIENTE DE EXCITACION

VOLT

AJE

APL

ICA

DO

RTC = 200-600/5Clase C800

RTC = 600/5

RTC = 200/5

FIGURA 11

FIGURA 12. TIECH-01

CONCLUSIONES: Debido a los altos costos que significa contar con un equipo para pruebas de relevadores de protección, consideramos esta como una propuesta digna de considerarse, a partir de que no siempre es requerido una alta exactitud en las operaciones de los equipos de protección, lo anterior se respalda con los trabajos realizados utilizando nuestros equipos para verificación y pruebas de relevadores de protección en muchas instalaciones entre las que podemos mencionar los siguientes esquemas: Protección de líneas de 230 y 400 KV: Tamos, Altamira, Poza Rica Dos, Temascal Dos, Oaxaca Potencia, Juchitán Dos, El Juile, Escárcega, Tikul, Cadereyta, Uruapan Potencia etc. Protección de Generadores: El Sauz, El encino, Cañaveral, Soyapango etc. Protección de Transformadores, líneas de 115 KV y alimentadores de 13.8 KV: Monclova, Cadereyta, Mérida Oriente, Playa del Carmen, Avandaro Bledales, Metropoli, Tecolote, Julián Carrillo, Lagos Galera, Zamora Potencia, Los Amates, Ocotlan, Leon Dos, Uruapan Potencia, Zamora Potencia, Los Amates, Ocotlan, Leon Dos Los Azufres e innumerables subestaciones a lo largo de la República Mexicana


BIOGRAFIA: Nació en San Pedro Quiatoni Oaxaca en 1953, Egresado del Instituto Tecnológico de Oaxaca en 1980 graduado como Ingeniero Industrial Electricista. Trabajó en la Comisión Federal de Electricidad de Nov/1974 a

mayo/1992 desempeñando diferentes cargos en la Región de Transmisión Sureste. Miembro del Comité Nacional de Protecciones. Instructor de diversos cursos sobre su especialidad. Autor de varios artículos técnicos y ponente en diversos eventos. En 1991 merecedor de la Medalla ADOLFO LOPEZ MATEOS al mérito electricista por su destacado desempeño dentro de la C.F.E. Actualmente se desempeña como consultor e instructor en la especialidad de protección y medición para México y Centroamérica a través de la empresa Servicios especializados de Ingeniería de Protecciones Eléctricas, en la ciudad de Oaxaca Oax. México.

FIGURA 13: XÁID-03

V I S I M P O S I O I B E R O A M E R I C A N O S O B R E P R O T E C C I O N D ES I S T E M A S E L E C T R I C O S D E P O T E N C I A


MODERN COST-EFFICIENTDIGITAL BUSBAR PROTECTION SOLUTIONS

Bogdan [email protected]

(905) 201 2199

Gustavo [email protected]

(905) 201 2402



Modern Cost-Efficient Digital Busbar Protection Solutions

Page 2 of 26

1. Introduction

Simple busbars with dedicated Current Transformers (CTs) could be efficiently protected bythe high-impedance principle – a fast and reliable scheme with tens of years of excellent field re-cord. However, new power generation added recently, or to be added in the near future, compli-cates historically simple busbar arrangements and exposes existing CTs to saturation due to in-creased fault current level. New substations are often designed to satisfy cost requirements ratherthan keep the protection task straightforward and easy. This results in complex busbar arrange-ments.

High-impedance busbar protection principle faces major problems when applied to complexbusbar arrangements. Quite often, the zones of protection are required to adjust their boundariesbased on changing busbar configuration. This calls for switching secondary currents – an opera-tion that is never considered safe and should be avoided whenever possible.

Digital low-impedance busbar protection schemes are ideal for complex busbars. Optimalzoning (dynamic bus replica) is achieved naturally by switching currents in software, i.e. bymaking logical assignments to multiple zones of protection while keeping physical currents un-interrupted. Other benefits include integrated breaker fail protection, communications, oscillog-raphy, sequence of events recording, multiple setting groups, and other natural benefits of thedigital generation of protective relays.

Till very recently digital busbar and breaker failure protection schemes for medium-size andlarge busbars were not attractive to users traditionally biased toward the high-impedance ap-proach. There used to be several reasons for that. Schemes available on the market were very ex-pensive, difficult to apply, considerably slower as compared with the high-impedance protection,and perceived less secure. All these factors have changed recently. Modern digital relays aremuch faster, use better algorithms for security, and became affordable after introduction – in late2001 and early 2002 – of a phase-segregated microprocessor-based busbar relay.

Major hardware, architectural and processing power challenges facing a digital protectionsystem for medium-size and large busbars are:(a) Large number of analog signals needs to be processed (tens of currents, few voltages). The

problem is how to bring all the required signals into a “box”.(b) Large number of digital inputs may be required to monitor isolator and breaker positions in

order to provide for the dynamic bus replica mechanism (dynamic adjustment of zoneboundaries based on changing busbar configuration).

(c) Large number of trip-rated output contacts may be required particularly in the case of re-configurable busbars when each breaker must be tripped separately depending on bus con-figuration at the moment of tripping.

(d) Several differential zones are required to cover individual sections of a large bus. This callsfor significant processing power of the hardware platform.Traditionally, the aforementioned problems of large number of inputs and outputs, resulting

power supply requirements, and the processing power requirements have been addressed by twodistinctively different architectures: distributed or centralized. Both the solutions require large


Page 3 of 26

quantities of specialized hardware. As a result, they are difficult to engineer, face certain depend-ability and reliability problems, do not have a chance to mature due to comparatively low volumeof installations, and are very expensive.

The new solutions that emerged recently address the above problems by targeting medium-size busbars only and using generic hardware platforms (such as multi-winding or small bus re-lays) to build a phase-segregated, cost-efficient, easy-to-engineer busbar relays.

This paper focuses on the new phase-segregated solution and is organized as follows.First, a general overview of busbar protection principles is given starting from simple inter-

locking schemes for single-incomer distribution busbars, to high-end microprocessor based pro-tection schemes.

Second, a novel, phase-segregated approach based on existing hardware platforms capable ofprocessing plurality of single-phase AC input signals, is presented. The new solution is discussedin details including architecture, reliability, dependability, speed of operation, security on exter-nal faults, ease of configuration, and cost.

Third, basic application principles for protection of complex busbars are presented. They in-clude a tie-breaker with a single CT, treatment of blind zones and over-tripping zones, dynamicbus replica, end fault protection and breaker failure protection. Both principles and examples arepresented.

2. Busbar Protection Techniques

Power system busbars vary significantly as to their size (number of circuits connected), com-plexity (number of sections, tie-breakers, isolator switches / disconnectors, etc.) and voltage level(transmission, distribution).

The above technical aspects combined with economic factors yield a number of protectionsolutions.

2.1. Interlocking SchemesA simple protection for distribution

busbars can be engineered as an interlock-ing scheme. OverCurrent (OC) relays areplaced on an incoming circuit and at alloutgoing feeders. The feeder OCs are set todetect feeder faults. The OC on the incom-ing circuit is set to trip the busbar unlessblocked by any of the feeder OC relays(Figure 1). A short coordination timer is re-quired to avoid race conditions.

When using microprocessor-basedmulti-function relays it becomes possible tointegrate all the required OC functions in

50

50 50 50 50 50

BLO

CK

Fig.1. Illustration of the interlocking scheme.


Page 4 of 26

one or few relays. This allows not only to reduce the wiring but also to shorten the coordinationtime and speed-up operation of the scheme.

Modern relays provide for fast peer-to-peer communications using protocols such as the UCAwith the GOOSE mechanism. This allows eliminating wiring and sending the blocking signalsover digital communications.

The scheme although easy to apply and economical is limited to simple distribution busbars.

2.2. Overcurrent DifferentialTypically a differential current is created externally by summation of all the circuit currents

and supplied to an overcurrent relay (Figure 2). Preferably the CTs should be of the same ratio. Ifnot, matching CTs are required. This in turn may increase the burden for the main CTs and makethe saturation problem even more significant.

Historically, means to deal with the issue of CT saturation include definite time or inverse-time overcurrent characteristics.

Although economical and applicable to distribution busbars, this solution does not match per-formance of more advanced schemes and should not be applied to transmission-level busbars.

The principle, however, may be available as a protection function in an integrated microproc-essor-based busbar relay. If this is the case, such unrestrained differential element should be setabove the maximum spurious differential current and may give a chance to speed up operationduring heavy internal faults.

2.3. Percent DifferentialPercent differential relays create a restraining signal in addition to the differential signal and

apply a percent (restrained) characteristic. The choices of the restraining signal include “sum”,“average” and “maximum” of the bus currents. The choices of the characteristic include typicallysingle-slope and double-slope characteristics.

This low-impedance approach does not require dedicated CTs, can tolerate substantial CTsaturation and provides for comparatively high-speed tripping.

Many integrated relays perform CT ratiocompensation eliminating the need formatching CTs.

This principle became attractive withthe advent of microprocessor-based relaysbecause of the following:

• Advanced algorithms supplement thepercent differential protection functionmaking the relay very secure.

• Protection of re-configurable busbarsbecomes easier as the dynamic bus rep-

51

Fig.2. Unrestrained differential protection.


Page 5 of 26

lica (bus image) can be accomplished without switching secondary currents.

• Integrated Breaker Fail (BF) function can provide for optimum tripping strategy dependingon the actual configuration of a busbar.

• Distributed architectures could be used that place Data Acquisition Units (DAU) in bays andreplace current wires by fiber optic communications.

2.4. High-Impedance ProtectionHigh-impedance protection responds to a voltage across the differential junction points. The

CTs are required to be of a low leakage (completely distributed windings or toroidal coils). Dur-ing external faults, even with severe saturation of some of the CTs, the voltage does not riseabove certain level, because the other CTs will provide a lower-impedance path as comparedwith the relay input impedance. The principle has been used for more than half a century becauseis robust, secure and fast.

The technique, however, is not free from disadvantages. The most important ones are:

• The high-impedance approach requires dedicated CTs (significant cost associated).

• It cannot be easily applied to re-configurable busbars (switching currents with bi-stable aux-iliary relays endangers the CTs, jeopardizes security and adds an extra cost).

• The scheme requires only a simple voltage level sensor. If BF, event recording, oscillography,communications, and other benefits of microprocessor-based relaying are of interest, thenextra equipment is needed (such as a Digital Fault Recorder or dedicated BF relays).

2.5. Protection using Linear CouplersA linear coupler (air core mutual reactor) produces its output voltage proportional to the de-

rivative of the input current. Because they are using air cores, linear couplers do not saturate.During internal faults the sum of the busbar currents, and thus their derivatives, is zero.

Based on that, a simple busbar protection is achieved by connecting the secondary windings ofthe linear couplers in series (in order to respond to the sum of the primary currents) and putting avoltage sensor to close the loop (Figure 3).

Disadvantages of this approach are similar tothose of the high-impedance scheme.

2.6. Microprocessor-based RelaysThe low-impedance approach used to be per-

ceived as less secure when compared with thehigh-impedance protection. This is no longer trueas microprocessor-based relays apply sophisticatedalgorithms to match the performance of the high-impedance schemes [1-6]. This is particularly rele-vant for large, extra high voltage busbars (cost ofextra CTs) and complex busbars (dynamic bus

59

Fig.3. Busbar protection with linear couplers.


Page 6 of 26

replica) that cannot be handled well byhigh-impedance schemes.

Digital low-impedance relays couldbe developed in one of the two distinc-tive architectures:

• Distributed busbar protection usesDAUs installed in each bay to sam-ple and pre-process the signals andprovide trip rated output contacts(Figure 4). It uses a separate Cen-tral Unit (CU) for gathering andprocessing all the information andfiber-optic communications be-tween the CU and DAUs to deliverthe data. Sampling synchronizationand/or time-stamping mechanismsare required. This solution bringsadvantages of reduced wiring at theprice of more complex, thus lessreliable, architecture.

• Centralized busbar protection re-quires wiring all the signals to acentral location, where a single“relay” performs all the functions(Figure 5). The wiring cannot bereduced and the calculations cannotbe distributed between plurality ofDAUs imposing more computa-tional demand for the central unit.On the other hand, this architecture

is perceived more reliable and suits better retrofit applications.

Algorithms for low-impedance relays are aimed at [8]:(a) Improving the main differential algorithm by providing better filtering, faster response,

better restraining technique, robust switch-off transient blocking, etc.(b) Incorporating a saturation detection mechanism that would recognize CT saturation on ex-

ternal faults in a fast and reliable manner.(c) Applying a second protection principle such as phase directional (phase comparison) for

better security.

Digital relays for large busbars dominating the market till recently provide for a trip time inthe range of 0.75 to 1.5 power cycles, and use either phase comparison principle or decaying re-straining current for increased security on external faults. They were designed several years agobased on technology that since then was outdated by several generations of microprocessors.

52

DAU

52

DAU

52

DAU

CU

copperfiber

Fig.4. Distributed busbar protection.

52 52 52

CU

copperFig.5. Centralized busbar protection.


Page 7 of 26

In the meantime several new busbar relays have been introduced based on modern, morepowerful hardware platforms [7,8]. These relays provide for faster tripping time and modernfeatures, but till recently their capabilities were limited to small (typically six-circuit) busbars.

This changed with the introduction of digital, phase-segregated, low-impedance protectionschemes.

3. Phase-Segregated Busbar Relays

The problem of large number of inputs and outputs required for protection of medium-sizeand large busbars, as well as computational power required to perform all the necessary opera-tions on the inputs, could be solved by using phase segregated approach to busbar protection.

From the perspective of the main differential protection, the algorithm is naturally phase-segregated. This means that no information is required regarding currents in phases B and C inorder to fully protect phase A. This bears several important consequences and advantages.

First, completely independent microprocessor-based devices could process the AC signalsthat belong to phases A, B and C. No data transfer is required between the devices.

Second, sampling synchronization is not required between the separate devices processingsignals that belong to individual phases.

The above observations facilitate phase segregated busbar protection. With reference to Fig-ure 6, three separate relays (Intelligent Electronic Devices, IEDs) could be used to set up protec-tion for a three-phase busbar. Each device is fed with AC signals belonging to the same phase,processes these signals, and arrives at the trip/no-trip decision. On solidly grounded systems, atleast one device would operate for any type of fault. For phase-to-phase faults, two relays wouldoperate.

In order to protect a medium-size busbar it is enough that each IED supports some 18-24 ACinputs. Present protection platforms support this amount of AC inputs. A modern multi-windingtransformer, or small busbar relay, could thus be converted from a three-phase device into a sin-gle-phase differential device. Instead of supporting four or more three-phase inputs, and someground inputs, the relay supports 18-24 generic AC inputs and allows for configuring them as in-puts to the same zone of differential protection.

This approach yields a number of significant benefits. The three most important ones are:First, by building on existing platforms, the vendors could develop such a solution in a short timewith a very low investment. Second, utilizing standard platforms brings extra maturity and fea-tures into the busbar applications. Third, by building on standard hardware platforms, the manu-facturing cost is also reduced. Consequently, the overall cost of the phase-segregated solution issubstantially lower as compared with traditional, “specialized” digital busbar relays. Other fea-tures, benefits and peculiarities of the phase-segregated approach are discussed in subsection 3.6.

Busbar protection is more then a plain differential function. The following subsections ad-dress several issues related to features such as breaker failure protection, undervoltage supervi-sion, dynamic bus replica, etc.


Page 8 of 26

3.1. Differential ProtectionThe main differential protection function is implemented on a per-phase basis. A given solu-

tion could be more flexible, cost-efficient, and allowing more demanding applications, if multi-ple zones of protection are available.

A full-featured digital busbar protection system incorporates dynamic bus replica function.This includes both ability to dynamically assign currents to the relay differential zones, and pro-vide for reliable monitoring of the status signals for isolator switches and breakers. The latter istypically implemented by utilizing both normally open and normally closed auxiliary switches asexplained in subsection 3.4.

Undervoltage supervision (release) of the main differential function is an often-used feature.This feature guards the system against CT trouble conditions and problems with the dynamic busreplica (false position of a switch/breaker). Typically, phase undervoltage, or neutral and/ornegative-sequence overvoltage functions are used. The phase-segregated approach treats single-phase voltage inputs in a generic way. The user could wire phase voltages or neutral (brokendelta VT) voltages for the purpose of voltage supervision to all, or selected IEDs only. As a rule,a voltage abnormality in any phase releases all three phases of differential protection. This callsfor simple inter-IED communications. This could be done via input contacts, or digital inter-IEDcommunications means.

In the phase-segregated approach each phase IED could drive an output contact for the tripcommand. This could be done on a per-breaker basis, if required. External lockout relays may beused to gather the per-phase trip commands in order to generate a single three-pole trip signal, ifrequired.

3.2. Inter-relay Digital CommunicationsEliminating the AC data traffic between the devices facilitates the digital phase-segregated

busbar protection scheme. It is very beneficial, however, to provide for fast, reliable, fully pro-grammable communications mean for sharing on/off states between the IEDs comprising thebusbar protection system. Important applications of such a communications mean are as follows:

phas

eA

Cur

rent

sph

ase

Atri

p

phas

eB

Cur

rent

sph

ase

Btri

p

phas

eC

Curre

nts

phas

eC

trip

Fig. 6. Phase-segregated digital busbar scheme.


Page 9 of 26

• A large number of I/O points could be freely distributed between the devices. This meansthat the same information that is common for all three-phases, such as position of a mo-torized switch, could be wired just to one IED, and distributed to the remaining IEDs viadigital communications.

• Applications such as cross triggering of oscillography, undervoltage supervision, check-zone could be easily accomplished by sharing on/off signals via communications chan-nels.

• Certain breaker failure architectures rely heavily on digital communications as explainedin subsection 3.3.

• Certain dynamic bus replica solutions rely on communications as well. Tens or hundredsof digital inputs required for monitoring all the switches and breakers, could be wired toan extra relay (relays), where basic filtering (contact discrepancy and alarming) is per-formed, and the final status signals for the differential protection is sent via communica-tions (subsection 3.4).

One particular solution [9] uses redundant token-ring-like communications mechanism asshown in Figure 7. Up to eight devices could be connected. Each device could share up to 96points with all the other devices. The mechanism is based on the “auto-forwarding”: each mes-sage is received, decoded if required, and forwarded to the next device. When the originating de-vice receives its own message back, it stops forwarding it. This is also a sign that the communi-cations ring is healthy. Two rings increase reliability and reduce (by half) the maximum messagedelivery time. Available physical media include fiber, RS422 and G.703. As the devices are con-nected directly, the 820 nm fiber connection is typically used. 32-bit CRC is implemented for se-curity, and the messages are repeated for extra dependability. The maximum delivery time be-tween two neighboring devices is 2-3 msec. Broken-ring alarm is incorporated for overall reli-ability of the scheme. Default states are available in order to program response of the schemeshould the communications fail.

TX1

RX2

RX1

TX2

TX2

RX1

RX2

TX1

RX1 TX2

TX1 RX2

RX2 TX1

TX2 RX1

IED1

IED3

IED4 IED2

Fig. 7. Inter-IED communications for sharing on/off signals.


Page 10 of 26

3.3. Breaker Failure ProtectionIntegrated Breaker Failure (BF) protection is quite beneficial for complex busbars. The BF

system must monitor configuration of a busbar at any given time, in order to initiate appropriatetrip action upon a failure of a given breaker. Basically, this information is identical with the in-formation required by the main differential protection. Doubling this information, i.e. using twoseparate systems for differential protection and breaker failure, is not economical as a very sig-nificant number of I/O points is required to interface the scheme with auxiliary switches of all theisolator switches and breakers. This leads towards integration of the busbar and breaker failureprotection functions.

The BF imposes a design challenge for phase-segregated approaches. Large number of cur-rents must be monitored, large number of I/O points must be monitored or driven, and largecomputational power is required to process those signals.

One particular solution monitors the currents using the phase IEDs and sending the corre-sponding pickup / dropout signal via digital communications. This means that the devices wiredto a given signal is responsible for monitoring the level of the signal and informing the BF-dedicated IED regarding the level of the current.

With reference to Figure 8, devices 1, 2 and 3 are main protection devices for phases A, Band C, respectively; while device 4 is a BF-dedicated device. Devices 1, 2, and 3 feed constantlythe fourth device with information regarding the current level. This does not create any extrabandwidth requirements as only changes of the states initiate transmission. Device 4 hosts all theBF schemes for all the breakers. Typically this device is configured to interface all the requiredI/O points. The BF functions could be initiated externally (typically via contact input), or inter-nally (typically via communications). If any of the BF functions times out and operates, it closesan output contact on device 4, or sends a signal via communications to close any dedicated con-tact located on any of the remaining devices of the scheme.

IED1 (A)

IED3 (C)IED4 (BF)

IED2 (B)

BFI, BF OC (A)

BFI, BF OC (B)

BFI, BF OC (C)

BF TRIP

BF TRIP

BF TRIP

BFI(

EXT.

)

CB

POSI

TIO

N

B FT R

IP BF TRIP

BF TRIP

BFTR

IP

Fig. 8. Sample Breaker Failure Solution.


Page 11 of 26

For simple busbars external BF relays could be used. When the busbar is re-configurable,however, it is always beneficial to have the BF function integrated with the dynamic bus replicafunction of the main differential protection.

3.4. Dynamic Bus ReplicaDynamic bus replica feature is critical for re-configurable and complex busbars. An actual

bus image must be monitor by the protection system for the following purposes.First, for circuits with a single CT point that could be routed between different sections of

the busbar, position of the isolator switch must be known in order to dynamically decide if theCT current belongs to a given zone of protection.

Second, for circuits with a single CB point that could be routed between different sections ofthe busbar, position of the isolator switch must be known in order to dynamically decide if theCB should be tripped upon operation of a given zone of protection.

Third, the BF protection should monitor all the breakers connected to a given breaker, in or-der to decide a tripping strategy should the said breaker fail.

Fourth, positions of breakers and tie-breakers should be monitored in order to avoid blindspots or over-tripping zones.

Fifth, certain complex switching strategies call for significant re-adjustments of zone bounda-ries. The re-adjustment should be programmed as a response to the changing configuration of thebusbar.

Section 4 addresses the aforementioned application considerations. Here, the aspect of dy-namic zone boundaries, and the isolator monitoring feature are discussed.

Dynamic zone boundary could be programmed using a very straightforward mechanism ofconfiguring a zone of protection as a list of pairs: current input – on/off status signal. A givencurrent becomes a part of the zone only if the corresponding status signal is asserted. Such amechanism allows for maximum flexibility as the connection status signals could be freely pro-grammed in user-programmable logic of the relay based on a number of conditions, and differentprotection philosophies.

Auxiliary switches could fail to respond correctly. This is particularly true for motorized switches.Wrong assignment of a given current to a given zone – caused by incorrect information regarding busconfiguration – could result in a false trip (typically) or a failure to trip (rarely). Therefore, an isolatormonitoring element should respond to both normally open and normally closed auxiliary contacts of anisolator or a tie-breaker in order to assert the actual position of the isolator for the dynamic bus image.Ideally, the element should assert two extra outputs for isolator alarm (contact discrepancy), and forblocking switching operations in the substation.

Traditionally, the following logic is applied.


Page 12 of 26

Table 1. Standard Isolator Monitoring Logic.Isolator OpenAuxiliary Contact

Isolator ClosedAuxiliary Contact

Isolator Position Alarm Block SwitchingOperations

Off On CLOSED No No

Off Off LAST VALID

On On CLOSED

After time delayuntil acknowledged

Until IsolatorPosition is valid

On Off OPEN No No

Typically, an alarm is set when contact discrepancy is detected. Depending on the type of dis-crepancy, either the last valid isolator position is assumed, or a “close” position is declared.

It may be beneficial to block switching operations in the substation should a problem with thebus image occur. An operator could remove such blocking signal once the nature of the problemis discovered and rationalized.

3.5. ModularitySome phase-segregated solutions offer extra modularity at the IED level [9]. This may in-

clude variable number of DSPs, and I/O cards as well as communications and redundant powersupply. This brings an advantage of shaping each IED of the busbar protection system to fit theneeds of a particular application.

With reference to Figure 9b, an IED may be configured with 2 DSPs only, allowing measur-ing 2 x 8 currents, thus protecting busbars of up to 16 breakers. Figure 9c presents a sample con-figuration with 3 DSPs (24 AC inputs), 3 I/O modules (up to 3 x 16 inputs, or 3 x 8 outputs) anddigital communications card. Figure 9d shows a configuration aimed at interfacing exclusivelyI/O points, with no DSPs, but 5 I/O modules, communications, and a redundant power supply.

3.6. Sample Protection System ConfigurationsPhase-segregated protection schemes, particularly the ones providing for extra I/O capabili-

ties, equipped with digital communications means, and supporting multiple zones of protection

POWER

CPU

(a)

POWER

CPU

DSP

I/O

DSP

I/O

(b)

POWER

CPU

DSP

I/O

DSP

I/O

DSP

I/O

COM

(c)

POWER

CPU

I/O

I/O

I/O

I/O

I/O

COM

POWER

(d)

Fig. 9. Sample configurations of a modular system: (a) power supply and CPU are mandatory, (b) 2 DSP, 2 I/Oconfiguration for up to 16 current inputs, (c) 3 DSP, 3 I/O configuration with communications for up to 24 current

inputs, (d) 5 I/O configuration with communications and dual power supply.


Page 13 of 26

could be configured to protect a variety of busbar configurations.Modular hardware platforms such as [7-9] are particularly attractive as they provide for two

levels of scalability. First, each IED could be configured to suit the needs of a given application(number of current and voltage inputs, number of digital inputs, number and type of contact out-puts, etc.). Second, the scheme could be configured from 1, 2, 3, 4, 5 or even more IEDs de-pending on complexity of a given application.

Figures 10 through 13 present sample applications.Figure 10 shows a single-IED protection for a simple eight-input busbar. The 24 current

channels available could be wired to 8 three-phase inputs; while 3 single-phase zones of protec-tion could be configured to provide differential protection for phases A, B and C.

A similar solution for busbars of up to 12 inputs could be built on 2 IEDs. The first IED uses2 zones to protect 12-input busbars in phases A and B. The second IED uses one of its zones and12 input signals to protect the remaining phase C.

Using multiple protection systems one could cover large busbars as long as each section is ofa medium size and a check zone is not required or done externally (Figure 13).

3.7. Advantages and BenefitsThe main advantages and benefits of a phase-segregated approach to digital low-impedance

busbars protection schemes are as follows.Because the phase-segregated relays are typically developed based on existing hardware plat-

forms, reusing probably some 90% of existing firmware, they are much more cost-efficient ascompared with dedicated schemes for large busbars.

Being built on existing hardware and firmware, with thousands of unit-years of field record,the new solutions are much more mature, and could reach better maturity indices fast, as com-pared with dedicated schemes installed in very low volumes. This allows reducing a risk of in-stalling a “new” busbar relay – the new relays are actually quite mature as their hardware and

CB 1 CB 2 CB 7 CB 8

ABC

...

ZONE 1ZONE 2ZONE 3

8 PHASE ACURRENTS

8 PHASE CCURRENTS

8 PHASE BCURRENTS

Fig. 10. Three-phase protection for small busbars.


Page 14 of 26

majority of firmware work already as a transformer or small busbar relay in numerous installa-tions.

1

2

3

4

21

22

23

24

ZONE 1

ZONE 2

1

ZONE 1

ZONE 2

23 24

ZONE 3

2 21

22

1 2 11

ZONE 1

12 13 22

23 24 ZONE 2

Fig. 11. Sample system configurations: breaker-and-half, two-section busbar with a tie-breaker,double busbar with tie-breaker and a transfer bus.


Page 15 of 26

The new relays have been developed from a simple solution up towards a sophisticated one.In this way the design was not biased towards ability of protecting 50+ circuit busbars. Becauseof that the configuration mechanisms, associated software and settings are simple and alreadyknown to the user from transformer and small busbar applications. The new solutions are easierto engineer as compared with dedicated large-busbar protection systems.

Being built on existing hardware and firmware platforms, the new busbar relays are membersof existing relay lines. They share common tools. They could be integrated with other membersof the relay line. The overall user learning curve could be significantly reduced.

1 2 3 23 24

ZONE 1

1 2 3 21 22

ZONE 1

ZONE 2

23 24

1

ZONE 1

ZONE 2

ZONE 3

2 2019

21 22

23 24

Fig. 12. Sample system configurations: single-, double- and triple busbars.


Page 16 of 26

1

BUS SYSTEM NO. 1: ZONE 1 25 23 BUS SYSTEM NO. 2: ZONE 1

BUS SYSTEM NO. 1: ZONE 2 26 24 BUS SYSTEM NO. 2: ZONE 2

... 23 27 ... 48

Fig. 13. Sample system configurations: a large busbar protected by two protection systems.

Certain configurations of phase-segregated solutions exhibit enhanced immunity to relay fail-ures. Consider for example, a simple system built on three devices protecting phases A, B and Cwithout any communications. If one of the devices fails, the system is still operational providingprotection for all kinds of faults except a SLG fault in the affected phase. Even more complexsystems with communications could be programmed so that the main functions remain intactwhen certain components of the scheme fail.

Oscillography and Sequence of Events recording capabilities are multiplied by using multipleIEDs. Simple software tools exist to “consolidate” SOE records, and comtrade files into singlefiles for easier analysis. Programming the IEDs for cross triggering of various records, and set-ting up a common clock reference signal (IRIG-B) allows good synchronization of records be-tween the IEDs comprising the busbar protection system.

User-programmable logic is available in each of the IEDs allowing easier engineering andmore flexible application as compared with dedicated “hard-coded” schemes that had to be set upby the vendor, rather than the user.

4. Application Considerations

This section presents application considerations with respect to protecting complex busbar ar-rangements and/or when the objective is to provide for optimum protection by avoiding blindspots or unnecessary bus outages. As a rule, this task calls for dynamic adjustments of boundariesof differential zones of protection, and can be safely accomplished when using numerical relays.


Page 17 of 26

In the following examples it is assumed that the dynamic bus replica feature is implementedby configuring a differential zone using pairs of (current signal, associated connection status sig-nal). A given current is included in the zone, only if the associated status signal is logic one. Inthis way each current input of the relay could be logically added or removed from the zone, de-pending on changing busbar configuration.

It is also assumed that the auxiliary logic variables required for bus configuration are pro-grammed in user-programmable logic of the relay.

4.1. Switchable Bus CircuitsFigure 14 presents a sample double busbar arrangement with bus sections 1 and 2, tie-breaker

CB-1, three outgoing circuits C-1, C-2 and C-3, and a number of CTs and isolator switches. It isassumed that the two sections could operate independently, with the tie-breaker opened or closed.Each section could be used as a transfer bus.

Several complex configurations are possible in the busbar of Figure 14. They are addressed inthe following subsections. However, if none of the isolators S-5 and S-6, S-7 and S-8, and S-9and S-10 could be closed simultaneously, the two sections could and should be protected inde-pendently by two zones of protection. This could be programmed as the following zone configu-rations.

CT-5

CT-6

CB-1

S-5

S-6

CT-2

CB-2S-2

S-7

S-8

CT-3

CB-3S-3

S-9

S-10

CT-4

CB-4S-4

SECTION 1

SECTION 2

C-1 C-2 C-3

Fig. 14. Sample double-bus arrangement.


Page 18 of 26

Table 2. Sample Zone Configuration for busbar of Figure 14.ZONE 1 ZONE 2

Current CT-6 Current CT-5Z1 Input 1Status On

Z2 Input 1Status On

Current CT-2 Current CT-2Z1 Input 2Status S-5

Z2 Input 2Status S-6





The setting of Table 2 are interpreted by the relay as follows: the CT-5 current belongs al-ways to Z1; the CT-2 current belongs to Z1 only if switch S-5 is closed; the CT-3 current belongsto Z1 only if switch S-7 is closed, etc.

Operation of a given zone should be routed to the breakers using bus configuration at themoment of tripping. In this example, the following trip commands should be created:

TRIP CB-1 = Z1 OR Z2 (1)TRIP CB-2 = (Z1 AND S-5) OR (Z2 AND S-6) (2)TRIP CB-3 = (Z1 AND S-7) OR (Z2 AND S-8) (3)TRIP CB-4 = (Z1 AND S-9) OR (Z2 AND S-10) (4)

Where Z1 and Z2 stand for operation of differential zones 1 and 2, respectively.Breaker Failure protection should force an associated differential zone to operate in order to

trip all the breakers currently connected to the failed breaker. For the busbar of Figure 14 thisshould be programmed as follows:

FORCE Z1 = BF-1 OR (BF-2 AND S-5) OR (BF-3 AND S-7) OR (BF-4 AND S-9) (5)FORCE Z2 = BF-1 OR (BF-2 AND S-6) OR (BF-3 AND S-8) OR (BF-4 AND S-10) (6)

Assume for example S-5, S-8 and S-9 closed; S-6, S-7 and S-10 opened; and CB-2 failing totrip for a circuit fault. BF-2 would force Z1 to operate (equation (5)). Z1 would trip breakers CB-1 (equation (1)) and CB-4 (equation (4)). This would clear the currents towards CB-2, but wouldleave section 2 and circuit C-2 in service.

The above example illustrates a classical situation when the dynamic bus replica is required.The application may get more complex if certain switching scenarios are allowed in the substa-tion of Figure 14 (subsection 4.6).

Application of the dynamic bus replica is also beneficial when protecting simple and non-switchable busbars as explained in subsections 4.2, 4.3 and 4.4.


Page 19 of 26

4.2. Line-Side CTsA differential zone must be terminated by points where a CT is present in order to locate a

fault (selectivity) and a CB is present in order to isolate the fault (trip action). Although posi-tioned very closely, the two devices have a finite space between them. This space is potentiallyexposed to faults. Depending on the mutual position of the CT and CB, the said space may be-come either a blind spot or may cause unnecessary trip of the busbar.

Consider an arrangement shown in Figure 15. When the CB is opened, the space between theline-side CT and the CB is an over-tripping zone: a fault between the CB and CT is not a busfault as the breaker is already opened and no current is supplied from the bus towards the fault.From the metering perspective, however, the fault is within the zone. If the current is staticallyassigned to the zone of protection, an unnecessary trip takes place.

Using breaker position as a connection status for the associated current could easily solve theproblem. When the breaker is opened, the current is removed from the zone. As a result the zoneboundary moves from the CT point to the bus-side pole of the opened CB. The zone contractsand the unnecessary trip of the busbar is prevented.

The drop out delay applied to the breaker position signal is necessary to allow measuring al-gorithms of the relay to ramp down after the current is interrupted by the breaker. Otherwise afalse trip would take place when the breaker opens.

Refining example of Figure 1, one should take care of the spots between CT-5 and CB-1 (forzone 2) and CT-6 and CB-1 (for zone 1). Consequently, the status signals for CT-5 and CT-6would be as follows:

Z1 Input 1 Current = CT-6 (7)Z1 Input 1 Status = CB-1 + drop out delay (8)Z2 Input 1 Current = CT-5 (9)Z2 Input 1 Status = CB-1 + drop out delay (10)

CT

CB

BUSBAR

Over-tripping zoneif CB opened

02 cycles

TIMER

CB closed Status Signal forthe CT current

Fig. 15. Line-side CT configuration.


Page 20 of 26

In other words, shortly after CB-1 gets opened, Z1 contracts to the section-1 pole of theopened CB-1, while Z2 contracts to the section-2 pole of the CB-1. In this way Z1 would not re-spond to faults between CT-6 and CB-1, and Z2 would not respond to faults between CT-5 andCB-1. This is desired for optimum selectivity of protection.

Same approach applies to breakers CB-2, CB-3 and CB-4 in Figure 14.

4.3. Bus-Side CTsSimilar situation occurs for bus-side CTs. Consider an arrangement shown in Figure 16. A

fault between the CB and CT is in a blind spot of the bus protection. To clear the fault the busbarmust be tripped, but the differential zone would not see this fault.

Similarly to the case of the line-side CT, this situation also requires using breaker position asa connection status for the associated current. The fault is cleared sequentially. First, protectionof the circuit – fed from the CT – responds to the fault and opens the breaker. When the breakeropens, the CT current is removed from the differential zone. As a result, the zone expands to thebus-side pole of the opened breaker, the fault becomes internal, and the bus protection clears thebusbar.

4.4. Tie-breaker with a Single CTIdeally two CTs should be used for tie-breakers (see Figure 14). In some situations, however,

a single CT is installed as shown in Figure 17.Two differential zones should be arranged in order to provide for selective protection of sec-

tions 1 and 2, respectively. As a result of having a single measuring point, a fault between the tie-breaker and the CT is internal to Z1, but external to Z2. Z1 would trip the tie-breaker, but thefault would not be cleared.

Using position of the tie-breaker for the connection status of the CT current for Z2 solves theproblem. After TB opens due to operation of Z1, Z2 expands to the opened breaker, the fault be-comes internal to Z2, and Z2 finally clears the fault by tripping section 2 of the busbar.

CT

CB

BUSBAR

Blind spot

02 cycles

TIMER

CB closed Status Signal forthe CT current

Fig. 16. Bus-side CT configuration.


Page 21 of 26

4.5. End Fault ProtectionConsider an arrangement of Figure 18. With the CB-1 opened, the bus zone moves from the

CT point to the CB-1 point leaving the portion between the CB-1 and CT unprotected.A simple overcurrent element responding to the CT current and activated when the breaker is

opened solves the problem. Such End Fault Protection should trip the CB-2 breaker. If CB-2 islocated in the same substation, the application is simple. If CB-2 is located at the remote terminalof a line, Direct Transfer Tripping capability is required. If DTT is not available, CB-2 would betripped from the backup protection functions at the remote terminal of the line.

4.6. Complex Switching StrategiesComplex switching scenarios are often al-

lowed for complex busbar arrangements.Consider a busbar of Figure 14 and assume

a transfer operation that allows closing the twoswitches (such as S-5 and S-6) simultaneouslywith the tie-breaker closed prior to the opera-tion. At the moment the two switches areclosed, the CT-2 current splits in an unknownproportion between sections 1 and 2. As a re-sult it becomes impossible to protect sections 1and 2 separately, and the entire busbar must beprotected as one entity.

First, such a transfer condition must be de-tected by the busbar relay. For a busbar of Fig-ure 14, the following auxiliary flag stands forthe condition of the two sections connected viaswitches:

CT

CB-1 (opened)

BUSBAR

Blind spot: End FaultProtection required

CB-2

Busbar protection zone

Circ

uitp

rote

ctio

nzo

ne

TRIP(local or DTT)

Fig. 18. End Fault Protection.

ZONE 1 ZONE 2

CT

TB

02 cycles

TIMER

TB closed Status Signal forthe CT current, Z2

Fig. 17. Tie-breaker with a single CT.


Page 22 of 26

A1 = (S-5 AND S-6) OR (S-7 AND S-8) OR (S-9 AND S-10) (11)

Second, the two zones must be re-arranged so that sections 1 and 2 are protected as one en-tity. This could be accomplished by one of the following:

• Disable Z1 and expand Z2 accordingly.

• Disable Z2 and expand Z1 accordingly.

• Disable both zones and trip from a check zone (if used).

• Expand both zones to enclose the entire busbar.The last solution is the most elegant because it is symmetrical. The following table shows

zone configurations that account for the transfer scenario.

Table 3. Sample Zone Configuration for busbar of Figure 14 (with transfer capabilities)ZONE 1 ZONE 2

Current CT-6 Current CT-5Z1 Input 1Status NOT (A1)

Z2 Input 1Status NOT (A1)

Current CT-2 Current CT-2Z1 Input 2Status S-5 OR A1

Z2 Input 2Status S-6 OR A1





Assume for example S-5, S-7 and S-10 closed, and S-6, S-8 and S-9 opened. Under this con-dition, the zones are dynamically adjusted as follows:

Z1: (CT-6, CT-2, CT-3) (12)Z2: (CT-5, CT-4) (13)

When subsequently, an action is initiated to transfer circuit C-1 from section 1 to section 2,switch S-6 is closed. As a result of S-5 and S-6 closed simultaneously, variable A1 is asserted(equation (11)). Consequently the zones are re-adjusted as follows (Table 3):

Z1: (CT-2, CT-3, CT-4) (14)Z2: (CT-2, CT-3, CT-4) (15)

In other words, both zones protect the entire bus. If a fault occurs under this configuration, allthe breakers will be tripped. This may include the tie-breaker (not necessary from the fault clear-ance perspective, but beneficial from the subsequent restoration perspective).

When subsequently S-5 gets opened in order to complete the transfer, A1 resets, and thezones re-adjust again as per Table 3:


Page 23 of 26

Z1: (CT-6, CT-3) (16)Z2: (CT-5, CT-2, CT-4) (17)

In this way CT-2 (circuit C-2) got transferred between protection zones 1 and 2.Another operating mode is to transfer all the circuits to one section, say section 1, keep one

circuit on the other section (say number 2) in order to service the breaker. Once the by-passingswitch is closed, the current of the paralleled CT must be ignored. The bus section connected tothe circuit in question becomes part of the circuit and is protected by the circuit relay. In order toaccomplish that protection of the circuit must be transferred to the tie-breaker CT as illustrated inFigure 19.

At the same time bus zone that has the non-valid current assigned to it, must be blocked orotherwise it would misoperate. For the busbar of Figure 19 the following auxiliary flags couldaccomplish this:

A2 = (S-2 AND S-5) OR (S-3 AND S-7) OR (S-4 AND S-9) (18)A3 = (S-2 AND S-6) OR (S-3 AND S-8) OR (S-4 AND S-10) (19)

Flag A2 means Z1 is fed with at least one invalid current and thus it should be blocked. FlagA3 means Z2 is fed with at least one invalid current and thus it should be blocked. Consider con-figuration of Figure 19: A2 = 0 (Z1 is not blocked – equation (18)), A3 = 1 (Z2 is blocked–equation (19)).

CT-5

CT-6

CB-1

S-5

S-6

CT-2

CB-2S-2

S-7

S-8

CT-3

CB-3S-3

S-9

S-10

CT-4

CB-4S-4

SECTION 1

SECTION 2

C-1 C-2 C-3

Fig. 19. Bus configuration that allows servicing CB-2.


Page 24 of 26

In this way:

• Z1 is reconfigured to protect section 1 by monitoring the following currents (CT-6, CT-3,CT-4).

• Z2 is blocked and would not misoperate.

• Section 2 is protected by protection of C-1 transferred to CT-5.

4.7. Check ZoneCheck zone is a mean of achieving extra security. A check zone is used as a supervising ele-

ment for regular zones of protection and mitigates problems of wrong status information (incor-rect switch or breaker position recognized by the relay). It will also – at least partially – mitigateCT trouble conditions.

The check zone is programmed to monitor the overall current balance for the entire busbar. Inthe example of Figure 14 the check zone is configured as follows:

Table 4. Check Zone Configuration for busbar of Figure 14.ZONE 3 (CHECK ZONE)

Current CT-2Z3 Input 1Status OnCurrent CT-3Z3 Input 2Status OnCurrent CT-4Z3 Input 3Status On

Ideally, Z3 should be fed from different cores of the CTs.There are circumstances when the check zone will not work. Consider configuration of Fig-

ure 19. Because the CT-2 current is not valid (due to the S-2 switch closed), it is impossible tocheck the current balance for the entire busbar. In order to ensure that Z1 would trip for internalfaults, the Z3 supervision should by dynamically removed under the circumstances. This could beimplemented using the following logic:

Z1 SUPVERVISION = Z3 OR A2 (20)Z2 SUPVERVISION = Z3 OR A3 (21)

As shown in this section, there are numerous situations when the boundaries of differentialzones of busbar protection should be dynamically and automatically re-adjusted. Such operationis natural, easy to program, and safe when using digital low-impedance relays.


Page 25 of 26

5. Summary

Several new phase-segregated low-impedance digital protection schemes for medium andlarge busbars emerged in 2001 and 2002. Built on existing hardware and firmware platformsthese solutions bring in low-cost, initial maturity and simplicity of application.

At the same time existing busbars are being upgraded to accommodate new generation andnew power equipment. New substations are built as double-busbars or even more complex ar-rangements. In new substation designs driven more by cost and less by engineering considera-tions, motorized switches replace breakers. Sophisticated switching strategies are allowed to copewith various operational conditions. All this creates major problems for high-impedance, or non-digital in general, busbar relays.

A digital busbar protection scheme is a natural choice for protecting such complex re-configurable busbars. Digital relays already reached security of high-impedance schemes [9].Modern relays provide for a sub-cycle tripping time [9]. With the new phase-segregated ap-proach, the digital schemes become also affordable and easy to engineer, opening new applica-tion opportunities.

6. References[1] Peck D.M., Nygaard B., Wadelius K., “A New Numerical Busbar Protection System with Bay-

Oriented Structure”, 5th IEE Developments in Power System Protection Conference, 1993, IEE Pub.No.368, pp.228-231.

[2] Andow F., Suga N., Murakami Y., Inamura K., “Microprocessor-Based Busbar Protection Relay”, 5th

IEE Developments in Power System Protection Conference, 1993, IEE Pub. No.368, pp.103-106.

[3] Funk H.W., Ziegler G., “Numerical Busbar Protection, Design and Service Experience”, 5th IEE De-velopments in Power System Protection Conference, 1993, IEE Pub. No.368, pp.131-134.

[4] Evans J.W., Parmella R., Sheahan K.M., Downes J.A., “Conventional and Digital Busbar Protection:A Comparative Reliability Study”, 5th IEE Developments in Power System Protection Conference,1993, IEE Pub. No.368, pp.126-130.

[5] Sachdev M.S., Sidhu T.S., Gill H.S., “A Busbar Protection Technique and its Performance DuringCT Saturation and CT Ratio-Mismatch”, IEEE Trans. on Power Delivery, Vol.15, No.3, July 2000,pp.895-901.

[6] Jiali H., Shanshan L., Wang G., Kezunovic M., “Implementation of a Distributed Digital Bus Protec-tion System”, IEEE Trans. on Power Delivery, Vol.12, No.4, October 1997, pp.1445-1451.

[7] Pozzuoli M.P., “Meeting the Challenges of the New Millennium: The Universal Relay”, Texas A&MUniversity Conference for Protective Relay Engineers, College Station, Texas, April 5-8, 1999.

[8] Kasztenny B., Sevov L., Brunello G., “Digital Low-Impedance Busbar Protection –Review of Princi-ples and Approaches”, Proceedings of the 54th Annual Conference for Protective Relay Engineers,College Station, TX, April 3-5, 2001. Also presented at the 55th Annual Georgia Tech ProtectiveRelaying, Atlanta, GA, May 2-5, 2001.

[9] B90 Busbar Protection Relay – Instruction Manual. GE publication No.GEK-106241, 2002.


Page 26 of 26

! ! !

Bogdan Kasztenny received his M.Sc. and Ph.D. degrees from the Wroclaw University of Technology(WUT), Poland. After his graduation he joined the Department of Electrical Engineering of WUT. Laterhe taught power systems and did research in protection and control at Southern Illinois University inCarbondale and Texas A&M University in College Station. Currently, Dr. Kasztenny works for GEPower Management as a Chief Application Engineer. Bogdan is a Senior Member of IEEE and has pub-lished more than 100 papers on protection and control.

Gustavo Brunello received his Engineering Degree from National University in Argentina and a Masterin Engineering from University of Toronto. After graduation he worked for the National Electrical PowerBoard in Argentina where he was involved in commissioning the 500 kV transmission system. For sev-eral years he worked with ABB Relays and Network Control both in Canada and Italy where he becameEngineering Manager for protection and control systems. In 1999, he joined GE Power Management asan application engineer. He is responsible for the application and design of protection relays and controlsystems.

1

Abstract-- False operation of the protection system of a

grounding bank has occurred in a substation, in Uruguay, during the energization of the bank. An exhaustive study including ATP simulation determined that lowering the relay sensitivity was not secure. An alternative protection system is proposed which changes the current transformer connections and adds a restricted earth differential function to provide protection against internal ground faults.

Index Terms-- grounding bank - zig-zag transformer - restricted earth differential protective relay.

I. INTRODUCTION or grounding the 30kV distribution network of the Uruguayan interconnected electric power system; two

methods are used depending on the secondary winding of the power transformers: • If the secondary is a wye winding a resistor is connected

between the neutral and ground. • When the secondary is a delta winding, a zig-zag

transformer and a current-limiting resistor are used. The protection of this second grounding method is discussed

in this paper.

A. The traditionally recommended grounding bank protection system

Traditionally the 30kV Uruguayan network had the busbars for distribution feeders isolated. Like grounding systems are a relatively new concept, the systems were grounded some years ago.

As a protection system for the zig-zag transformer and resistor an overcurrent relay was used [1]-[3]. Phase overcurrent functions as zig-zag protection and an inverse time delayed ground overcurrent function as back up protection for the resistor.

Phase current coils of the protective relay are connected to the delta connected secondary windings of the phase current transformers. For external ground faults, only zero-phase-sequence current flow through the primaries of these current transformers and a very small current due to the current transformer errors is supposed to flow through the relay because of the delta connection. Then the protection may be very sensitive and fast for faults inside the bank.

Ground current coil of the relay is connected to the ground current transformer of the resistor. Zero-phase-sequence currents flow through this current transformer and the protective relay will trip in case of external ground faults protecting the resistor and working as a back up protection for distribution feeders.

Fig. 1 shows a diagram of the power system and the grounding bank protection.

B. Settings Instantaneous and inverse time phase overcurrent protection

is provided for the zig-zag transformer. As a very small current flows through the relay when an external fault occurs, these functions are very sensitive for internal short circuits and there is no need to be selective with any other relay. Minimum taps of the relay are selected.

The inverse time ground overcurrent function used for the resistor provides back up protection for the feeders. Pickup current value is equal to the continuous current rating of the resistor and the time characteristic is selective with the distribution feeders relays.

C. The zig-zag magnetizing current causes the relay to trip A typical transmission substation, in which a grounding

bank has been installed, has two similar 150/30kV wye-delta power transformers, working together in parallel. The zig-zag transformer is connected directly to the 30kV busbar without breakers and its neutral is grounded through a 50 ohm current-limiting resistor. Tripping circuits of the grounding bank protective relay are connected to the power transformer 30kV breakers.

Few years ago a transformer back up protection tripped the power transformer 30 kV circuit breakers due to a failure of one of the feeders protective devices. Afterward the energization of the 30 kV busbar and consequently the grounding bank caused the phase overcurrent function of the zig-zag transformer to trip for its first time.

A New Grounding Transformer Protection based on Restricted Earth Differential Principle C. Sena, UTE, J. Zorrilla de San Martín, UTE, J. Alonso, IIE Univ. Mayor, UTE, Uruguay

F

2

Some hypotheses were handled at that time, including an internal zig-zag or resistor fault, lately discarded after some tests. Another hypothesis was that probably the inrush currents of the zig-zag had caused the tripping.

In order to confirm the last, energization tests of the bank were performed, measuring and saving the oscillography data of the magnetizing currents with laboratory equipment. As a result, the inrush currents caused the relay to trip in more than 50% of the cases. Fig. 2 shows the inrush currents of the zig-zag in one of the tests.

In fact, the traditionally recommended grounding bank protection system has a good performance during normal operation and when a fault occurs. But during the energization of the 30kV busbar the zig-zag magnetizing currents flow through the delta connected current transformers and in most of the cases a relay false trip may occur.

II. ATP MODEL OF THE GROUNDING TRANSFORMER A number of different technologies and techniques are

available to test a relay. The technology available in our laboratory is to feed relays with COMTRADE file format using playback equipment. The technique chosen in this research was to simulate the grounding transformer system in

great detail using ATP. The simulation data was then converted into COMTRADE format an injected to the relays.

A detailed ATP model of the grounding transformer, the grounding resistor and surrounding network was created. The statistical switching feature of ATP was used to determine the maximum inrush current. Another purpose of the ATP study was to obtain the waveforms of the phases and neutral currents caused by an internal fault and by an external fault.

Fig. 3 shows the inrush currents obtained in the secondary of the currents transformers, with the simulation in the ATP study [2]. The difference between the magnitude of the currents from the ATP study and the ones saved from the fields experience are due to that the field data should be considered one case from a random phenomenon.

III. PROTECTION SYSTEM CHANGES To rectify the wrong operation problems with the existing

protection of the grounding bank, attention was paid to the protection system and its connection with the current transformers.

It was established that the setting of the existing protection system or the entire protection system with its connections with the current transformers should be changed.

50/51

51N

Protective Relay

Zig-Zag Transformer

Resistor

Distribution Feeder

30kV

R S T

150kV

R S T

Power Transformers

Fig.1. Diagram of the power system and the grounding bank protection.

3

A. Change the settings Changing the settings of the protection system should lower

the relay sensitivity to avoid wrong operations at the expense of reducing its limits of applicability. Also, by raising the setting, the speed of the relay is reduced.

COMTRADE files of the worst cases were created and played back into an off-line relay. It was found that a new secure setting could not be found.

B. Change of the protection-system scheme Otherwise, as it can be shown that the existing protection

system with its connections is not good enough to protect the grounding transformer, it should be necessary to recommend a new protection system for that application.

The proposed protection system should carry on with the problems detected in the existing one using the same overcurrent relay.

Fig. 4 shows the proposed protection system [4]. For external ground faults, zero-phase-sequence currents flow through the primary of the wye-connected current transformers. So, the external ground fault back up is the phase-delay overcurrent function, and its time delay should be long enough to be selective with other relays that should operate for external faults. This function also protects the grounding resistor for overload conditions. The pickup of this function should be the continuous-current rating of the grounding resistor.

The phase-instantaneous overcurrent function provides protection for short circuits on the grounding-transformer side of the current transformers for phase to phase faults. The pickup of this function should be greater than the inrush current but lower than the internal phase to phase short circuit current.

The restricted earth differential function protects the grounding transformer for internal ground faults. The requirement for this function is that the phase-current and

neutral-current transformers should have the same primary and secondary currents rating. The pickup of this function should be lower enough to detect internal ground faults and greater enough not to operate for external ground faults.

IV. PROTECTION SYSTEM TEST In order to observe the performance of the proposed

protection system, a prototype was built in the laboratory that verified the overall design concept.

A. Inrush-current test This first series of tests has the goal to calculate the

suitability of the design of the protection system not to operate when the grounding bank is energized.

Laboratory proposed-protection-system behaviour with the waveforms of the inrush currents generated with the ATP program was observed. Despite many attempts of ATP simulation of the grounding transformer system and replay of the generated waveforms to the protection system in the laboratory, it was not possible to obtain unwanted behaviour of the relay like that has occurred with the existing protection system. The same result was obtained replaying the site-recorded waveforms of the inrush currents.

B. External Short-circuit-current test Having tested the proposed protection system under inrush

currents, it was the task to verify its performance under short circuits currents.

COMTRADE files of external ground faults and phase to phase fault were created and played back into the proposed protection system. It was found that under external phase to phase short circuits the system verified the design and worked properly offering no operation. For external ground faults, the systems showed an operation after a time delay that coordinate with the other relays that should operate for that kind of fault.

0 0.0 0.1 0.1 0.2 0.2 0.3 0.3 0.4 0.4 0.5-3

-2

-1

0

1

2

3

4

Time (s)

Inru

sh

cu

rre

nts

(A

)

Fig.2. Zig-zag transformer inrush currents (recorded)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 -8

-6

-4

-2

0

2

4

6

8

10

12

Time (s)

Inru

sh

cu

rre

nts

(A

)

Fig.3. Zig-zag transformer inrush currents (ATP simulation)

4

C. Internal short-circuit-current test Further targeted ATP simulation was carried out to study the

internal ground faults and internal phase to phase short circuits. These faults were placed in the cable between the current transformers and the grounding transformer. The simulation data was converted into COMTRADE format and played back into the protection system.

The performance of the system for that two kind of fault verifies the design criteria and the relay have worked properly, operating for all the internal faults.

These tests made possibly to carefully check the correct

operation of proposed protection system under the circumstances described above, representing the more usual operation and disturbances of the network.

V. INSTALLATION In December 2000, the proposed protection system was

installed at the grounding bank of a substation. This substation has two 40/32/32 MVA 150/60/30 kV

YNynd5 power transformers. The grounding bank was connected to the 30kV busbar.

The new protection system was connected to trip the both 30kV-power-transformers circuit breakers.

The response of the new protection system to the inrush current was checked in several times. In addition to this, a disturbance recorder was applied to the grounding bank, recording the three-phase currents and the neutral one.

Successful energization of the grounding transformer, in several times, was confirmed by field tests.

VI. CONCLUSIONS After the installation of the proposed protection systems, the

electric power system was exposured to several external abnormal conditions, like short-circuits leading to a general blackout or short-circuits following a reclosing cycle. In all the cases the protection system was submitted to many energization operations. As consequence no incorrect operation of the relay has occurred during the period that it has been in service.

As the restricted earth differential principle has passed all the "exams" and after the experience of more than two-year of use, all the protection systems for grounding banks in the 30kV Uruguayan distribution network will be changed from now on into this new scheme. It seems to be a better solution

50/51

50N

Protective Relay

Zig-Zag Transformer

Resistor

Distribution Feeder

30kV

R S T

150kV

R S T

Power Transformers

Fig. 4. Diagram of the power system and the proposed grounding bank protection

R

5

for a grounding bank protection system than the traditionally recommended.

VII. ACKNOWLEDGEMENT The authors gratefully acknowledge the contributions of the

Maintenance Department and MSc. Claudio Saldaña, from the Protection Engineering Department of the Protection Section, UTE.

VIII. REFERENCES Technical Reports:

[1] R. Mendez Albores, A. Cuenca Romero, and M. A. Mendez Albores, "Aplicaciones de los bancos de tierra, sus esquemas de protección por relevadores y criterio de cálculo de ajustes ," SIPSEP-93-90, pp. 283-297, Nov. 1993.

[2] C. Saldaña, "Estudio de energización, Transformador zig-zag 31.5 kV, Estación Bifurcación", Jan. 2001.

Books:

[3] C. Russell Mason, The art & science of protective relaying, New York: Wiley, 1967.

[4] J. Lewis Blackburn, Protective Relaying: Principles an Applications, Electrical Engineering and Electronics, 1987.

IX. BIOGRAPHIES

Celia Sena was born in Salto, Uruguay, in 1970. She received her degree in Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, in 1997.

She joined the Administración Nacional de Usinas y Transmisiones Eléctricas (UTE) in 1992, where she held the position of engineering in the System Protection Engineering section. Her interests include digital relay modeling, power system protection and power system transients. Juan Zorrilla de San Martín was born in Montevideo, Uruguay, in 1965. He received his Electrical Engineer degree from the Universidad Mayor de la República Oriental del Uruguay, in 1994.

He joined the Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.) in 1989 and worked in protection systems area since then. He is presently Manager of the System Protection Engineering section. Juan is a member of the I.E.E.E. since 1989. Jorge Alonso was born in Montevideo, Uruguay, in 1956. He received the Engineer and MSc. degrees all in Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, in 1979 and 1998, respectively. Since 1980, he has been with the Institute of Electrical Engineering from the Universidad Mayor de la República Oriental del Uruguay, now as an Aggregate Professor. Since 1979, he has been working for the Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.), where he is currently a general manager of control and protection power transmission system section. His main research and development activity has been on the simulation of numerical relays and the dynamic performance of electrical machines. He is a member of the IEEE since 1987.

1

ANÁLISIS DE LAS CORRIENTES EN LA PROTECCIÓN DIFERENCIAL

APLICADO A UN TRANSFORMADOR TRIFASICO


Servicios Especializados de Ingeniería de Protecciones Eléctricas.

Oaxaca Oax. [email protected]

RESUMEN: Para realizar correctamente las pruebas

a los relevadores de protección diferencial de los

transformadores de potencia se requiere conocer el

comportamiento de las corrientes primarias en ambos

lados de un transformador de dos devanados

(delta/estrella) y sus correspondientes corrientes

secundarias, de los transformadores de corriente. Se

presenta un análisis de las corrientes en estado estable y

en condiciones de falla.

INTRODUCCION

Existe una gran cantidad de conexiones para

transformadores de potencia, y además estos pueden ser

desde auto transformadores hasta transformadores de

cuatro devanados, monofásicos o trifásicos.

En este trabajo nos ocuparemos de los de 2 devanados

entre los que también existe una gran variedad.

Los diferentes tipos de conexiones de los transformadores

de potencia, presentan diferentes ángulos de

desfasamiento.

Estos desfasamientos se dan en saltos de 30 °, y pueden

resultar en el rango de 0 a 330° y como esto resulta en 12

divisiones, que se pueden relacionar con las 12 horas de

un reloj de manecillas, se establece la convención de

indicar los desfasamientos con un número horario en

donde cada incremento de un digito representa 30°.

Así por ejemplo una conexión Delta / estrella con

desfasamiento de 30 ° atrasada se representa como Dy1,

mientras que con desfasamiento de 30° adelante se

representa como Dy11.

DESARROLLO

CONDICIONES NORMALES: Para nuestro caso

tomaremos como ejemplo la conexión Dy11, para lo cuál

tendremos los fasores de la tabla No. 1 y conexiones

mostrados en la figura 1. para condiciones normales de

operación.

TABLA No. 1

CORRIENTES

PRIMARIAS

CORRIENTES

SECUNDARIAS

De Línea De Transfor-

mación

De Línea De Transfor-

mación

IA ∠90º IAB ∠120º Ia ∠120º Ia ∠120º

IB ∠-30º IBC∠ 0º Ib ∠ 0º Ib ∠ 0º

IC ∠-150º ICA∠-120º Ic ∠-120º Ic ∠-120º

Las corrientes primarias en el transformador en el lado

delta están definidas por las siguientes ecuaciones.

IA = IAB - ICA 1 IB = IBC - IAB 2 IC = ICA – IBC 3 En el lado estrella las corrientes de línea son las mismas

que las de transformación.

De las conexiones y polaridad Dy11 se establecen las

siguientes ecuaciones.

IAB = Ia / N 4 IBC = Ib / N 5 ICA = Ic / N 6

Donde N es la relación de transformación por fase del

transformador.

Sustituyendo las ecuaciones 4, 5 y 6 en 1, 2 y 3 tenemos.

IAB = Ia / N - Ic / N IBC = Ib / N - Ia / N ICA = Ic / N - Ib / N Simplificando tenemos.

IAB = ( Ia – Ic ) / N 7 IBC = ( Ib - Ia ) / N 8 ICA = ( Ic - Ib ) / N 9

Las ecuaciones anteriores definen el comportamiento del

transformador tanto en condiciones normales de operación

como en condiciones de falla.

2

CONDICIONES DE FALLA MONOFASICA.

La corriente de falla monofásica en el lado de la estrella

del transformador se transfiere al lado de la delta como

una falla entre fases, como se puede apreciar en la figura

No. 2, si la falla en la fase A en el lado estrella, esta se

refleja como falla AB en el lado de la delta.

Las corrientes que circulan a través del transformador se

observan en su diagrama trifilar y su diagrama fasorial de

la figura No. 2,

Ia ≠ 0 Ib = 0 Ic = 0

Sustituyendo en las ecuaciones generales 7,8,9, tenemos

IA = ( Ia - 0 ) / N IB = ( 0 - Ia ) / N IC = ( 0 - 0 ) / N nos queda IA = Ia / N IB = - Ia / N IC = 0

Según la ecuación 4, IAB = Ia / N por lo que se concluye

que:

IAB = IA ademas IA = - IB

CONDICIONES DE FALLA BIFASICA.

Para una falla bifásica entre las fases a y b en el lado de

la estrella se establecen las corrientes cmo se muestran

en la figura No. 3.

Ia = - Ib ≠ 0 IC = 0 Sustituyendo en las ecuaciones generales 7, 8 y 9

tenemos.

IA = ( Ia – 0 ) /N IB = ( Ib - Ia ) /N IC = ( 0 - Ib ) /N

Como Ia = - Ib simplificando nos queda:

IA = Ia /N IB = - 2 Ia /N IC = Ia /N

Según las ecs. 4 y 5 : IAB = Ia / N, e IBC = Ib / N, resulta que: IAB = - IBC, IA = IC , e IB = 2 IA .

REFLEJO DE LAS CORRIENTES PRIMARIAS EN LAS

PROTECCIONES DIFERENCIALES. Las corrientes

primarias se reflejan en el revelador de protección según

el tipo de conexiones que se tengan en los

transformadores de corriente, a continuación analizaremos

las corrientes que se presentan en diferentes condiciones.

Tomaremos para este análisis un transformador trifásico

de potencia con, las siguientes características:

Capacidad: 10,000 KVA

Relación : 115/13.8 KV

Conexión Delta / Estrella ( Dy11 )

Corriente nom. 50.2 / 418.4 A.

a) Condiciones normales

Inom 13.8 = KVA / 3 KV = 10,000 / 3 ( 13.8 ) = 418.4 A.

Inom 115 = 10,000 / 3 ( 115 ) = 50.2 A. a1) Corrientes primarias en forma fasorial de un sistema

balanceado tenemos

Ia = 418.4 ∠120° Ib = 418.4∠ 0° IC = 418.4 ∠ -120°

Aplicando las ecuaciones 7, 8 y 9 para obtener las

corrientes de 115 KV y teniendo una relación de

transformación por fase de 115 / (13.8 3 ) de donde:

N = 14.434

IA = ( 418.4 ∠120° - 418.4 ∠ -120° )/ 14.434 IA = [- 209.2 + j 362.3 – ( - 209.2 – j 362.3 )]/14.434 IA = [ 0 + j 724.6 ]/ 14.434 IA = ( 724.6 ∠90° ) / 14.434 IA = 50.2∠90°

El mismo proceso se usa para calcular IB, e IC

IB = (418.4∠ 0° - 418.4 ∠ -120° ) / 14.434 IB = 50.2 ∠ -30° IC = ( 418.4∠ -120° - 418.4 ∠ 0°) / 14.434 IC = 50.2 ∠ - 150°

De las ecuaciones 4,5 y 6 tenemos :

IAB = 418.4∠120°/ 14.434 IAB = 28.99∠120° IBC = 418.4∠ 0°/ 14.434 IBC = 28.99∠ 0° ICA = 418.4∠-120°/ 14.434 ICA = 28.99∠-120°

Estas corrientes se aprecian en la FIGURA No. 1

3

a2) Corrientes secundarias; Las corrientes secundarias

de los TC’S obtenidos mediante una conexión Estrella /

Delta de los TC’S para la protección diferencial, se

observa en la FIGURA 1.

Definiendo las RTC.- si asumimos en el lado de 115 KV

(delta ) una RTCΔ = 100/5. = 20

RTCy = 3 RTC Δ ( VΔ / Vy )

RTCy = 3 * 20 ( 115/ 13.8) RTCy = 288.67≈ 289 RTCy = 1445 / 5

Esta RTCy es con el propósito de lograr la igualdad en las

corrientes secundarias en la protección solamente con

fines demostrativos ya que en la práctica esta relación de

TC no existe.

iA = IA / RTCΔ = 50.2 ∠ 90° / 20 iB = IB / RTCΔ = 50.2 ∠-30° / 20 iC = IC / RTCΔ = 50.2 ∠-150° / 20 iA =2.51∠ 90° A iB =2.51∠-30° A iC =2.51∠-150° A ia = Ia/ RTCy = 418.4 ∠120° / 289 ib = Ib / RTCy = 418.4 ∠0° / 289

ia

ic - ib ib - ia ic

ia - ic

ib

iA

iC iB

Ia

Ic

Ib

IA IAB

IC

IBC

IB ICA

FIGURA No. 1

87

87

A

B

C

Ic

Ib

Ia IAB

IBC

ICA

Ia

Ib

IC

IB

IA

Ic

ic

ib

ia

1445 / 5 RTC = 100/5

iA

iB

iC

ia - ic

ib - ia

ic - ib

87

4

iC = IC /RTCy = 418.4 ∠-120° / 289 ia = 1.448 ∠120° ib = 1.448∠0° iC = 1.448∠-120°

Estas son corrientes secundarias en los TC´s, mientras

que las corrientes que circulan por la protección diferencial

(87T) es la diferencia de dos corrientes.

ia - iC = 1.448 ∠120° - 1.448∠-120° ia - iC = -0.724 + j1.254 - (-0.724 - j1.254 ) ia - iC = j2.51 ia - iC = 2.51∠90°

De manera similar resulta que:

ib - ia = 2.51∠-30° ic - ib = 2.51∠-150°

Obsérvese en la figura 1 que por las bobinas 87 no

circula ninguna corriente.

b) Falla monofásica externa a la protección

diferencial.

Para facilitar el análisis asumiremos las siguientes

condiciones de falla

Ia = 2000∠0°A, Ib = 0, Ic = 0 b1) Corrientes primarias: Sustituyendo en las ecuaciones

generales del transformador ( 7,8 y 9 )

IA = ( 2000∠0° - 0) / 14.434 = 138.5∠0° IB = ( 0 - 2000∠0° ) / 14.434 = 138.5∠180° IC = ( 0 – 0 ) / 14.434 = 0

b2) Corrientes secundarias :

iA = IA /RTCΔ = 138.5 ∠0°/ 20 iB = IB /RTCΔ = 138.5 ∠-180°/ 20

iC = IC = 0 iA = 6.92 ∠ 0° iB = 6.92∠ 180°

ia = Ia / RTCy = 2000∠0° / 289 ib = 0 ic = 0 ia = 6.92 ∠ 0° Estas corrientes se pueden apreciar en la figura No.2 en

los diagramas trifilar y fasorial.

Nótese que por las bobinas de operación del 87 no circula

ninguna corriente. Por otra parte las corrientes presentan

la siguientes variantes.

Ia pu = Ia / Inom 13.8 = 2000 / 418.4 = 4.78 p.u. IA pu =IA / Inom 115 = 138.5 / 50.2 = 2.76 p.u. Al realizar la siguiente relación tenemos que

Ia pu / IA pu

4.78 / 2.76 = 3 Esto significa que la corriente del

lado estrella del transformador es 3 veces mayor que la

corriente en el lado delta.

c).- Falla Bifásica externa a al protección diferencial .

Para nuestro análisis asumiremos las siguientes

condiciones

Ia = 0, Ib = - Ic = 3000∠0°A c1) Corrientes primarias

Ia = 0 Ib = 3000∠0° Ic= 3000 ∠180° Las corrientes del lado de 115 KV se obtienen de las

ecuaciones 7, 8 y 9.

IA = ( 0 - 3000 ∠180° )/ 14.434 = 207.8∠0° IB = ( 3000∠0° - 0 ) / 14.434 = 207.8∠0° IC = ( 3000∠180° -3000∠0° ) / 14.434 = 415.6∠180° c2) Corrientes secundarias

iA = IA /RTCΔ = 207.8 ∠0°/ 20 iB = IB /RTCΔ = 207.8∠0°/ 20 iC = IC /RTCΔ = 415.6∠180°/ 20 iA = 10.39∠0° iB = 10.39∠0° iC = 20.78∠180° ia = Ia / RTCy = 0 ib = Ib / RTCy = 3000∠0°/ 289 ic= Ic / RTCy = 3000 ∠180° / 289 ia = 0 ib = 10.39 ∠ 0° ic= 10.39 ∠180° Las corrientes de esta sección se muestran en la FIGURA

No. 3. Por las unidades 87 no tenemos ninguna corriente.

Por la otra parte comparando las corrientes primarias de la

delta y la estrella tenemos.

Ic pu = Ic / Inom 13.8 = 3000 / 418.4 = 7.17pu IC pu = IC / Inom 115 = 415.6 / 50.2 = 8.28pu De la relación de ambos tenemos

Ic pu / IC pu = 7.17 / 8.28 = 0.866 Esto significa que la corriente del lado estrella del

transformador es 3 / 2 veces menor que la corriente del

lado delta.

5

IC IA = IB

ic ib

IC iA = iB

Ic Ib

FIGURA No. 3

87

IA

iA

iB

iC

IB

IC IBC

ICA

iA

iB

iC

Ib

Ic

A

B

C

87

87

ib

ic

ic

ib

ib + ic

ib + ic

Ic

Ib

IB IA = IAB

Ia

iA iB ia

FIGURA No. 2

iA

iB

iA

iB

ia

ia

ia

IA

IB IAB Ia

87

87

87 A

B

C

100/5 1445 / 5 1445 / 5 100/5

6

CONCLUSIONES:

La protección diferencial requiere especial atención para

que no opere incorrectamente, mas que opere

correctamente esto implica tener un análisis detallado

del comportamiento de las corrientes primarias del

transformador y las conexiones adecuadas de los TC´s

para lograr un balance de las corrientes secundarias.

El balance de las corrientes secundarias deberá

observarse tanto en condiciones normales de operación

como en condiciones de fallas.

Un análisis matemático de los fasores de corrientes

primarias y secundarias nos permite determinar y prevenir

operaciones incorrectas por fallas externas a la protección

diferencial.

BIBLIOGRAFÍA:

La protección diferencial en los transformadores de

potencia. Meliton Angeles M. Tesis profesional.

Applied Protective Relaying. Westighouse Electric Corp.


BIOGRAFIA

Nació en San Pedro Quiatoni Oaxaca

en 1953, Egresado del Instituto

Tecnológico de Oaxaca en 1980

graduado como Ingeniero Industrial

Electricista.

Trabajó en la Comisión Federal de

Electricidad de Nov/1974 a mayo/1992

desempeñando diferentes cargos en

la Región de Transmisión Sureste.

Miembro del Comité Nacional de Protecciones. Instructor

de diversos cursos sobre su especialidad. Autor de varios

artículos técnicos y ponente en diversos eventos. En 1991

merecedor de la Medalla ADOLFO LOPEZ MATEOS al

mérito electricista por su destacado desempeño dentro de

la C.F.E.

Actualmente se desempeña como consultor e instructor en

la especialidad de protección y medición para México y

Centroamérica a través de la empresa Servicios

especializados de Ingeniería de Protecciones Eléctricas,

en la ciudad de Oaxaca Oax. México.

SISTEMA DE SUPERACIÓN PARA ESPECIALISTAS Y PROFESORES DE PROTECCIONES

Dr. Luis Corrales Barrios Universidad de Camagüey. Dpto. Ingeniería Eléctrica. Carretera Circunvalación km 5 ½ (N); Camagüey, Cuba. Email: [email protected] Dr. Ernesto Vázquez Martínez Universidad Autónoma de Nuevo León. Programa Doctoral en Ingeniería Eléctrica Facultad de Ingeniería Eléctrica y Mecánica San Nicolás de los Garza NL, CP 66450, México. Email: [email protected] MSc. Aurelio Ramírez Granados Universidad Autónoma de Nuevo León. Facultad de Físico Matemáticas. San Nicolás de los Garza NL, CP 66450, México. Email: [email protected] Palabras claves: Protecciones eléctricas, enseñanza, superación. Resumen: Se realiza un análisis del plan de formación para especialistas y profesores que laboran en la rama de las protecciones eléctricas, llegando inclusive hasta la obtención de grado científico. A partir de la experiencia acumulada en la preparación de este personal se propone el sistema de superación Introducción: El avance científico – técnico, los continuos cambios de tecnologías y el desarrollo de equipamiento de nuevo tipo, hacen que sea cada vez más difícil y se necesite un presupuesto de tiempo mayor para lograr especialistas en Protecciones que dominen adecuadamente los principios, métodos, tipos de equipamiento, etc., que intervienen en los procesos de selección, coordinación, mantenimiento y explotación del equipamiento de esta rama. Resulta aún más difícil la preparación de profesores para esta ciencia

debido a que debe tener los conocimientos de un especialista y además saber utilizar los conocimientos de pedagogía, la didáctica, la sicología y otras ciencias relacionadas con la enseñanza. Por criterios de expertos se conoce que se necesitan alrededor de 10 años de superación para formar un especialista en protecciones [1] (sin considerar la obtención de un grado científico que se puede obtener de forma paralela y contribuye en cierta medida a esta especialización), porque para explotar y mantener el equipamiento de protecciones no solamente hay que profundizar en los principios de operación, sino también en el conocimiento específico de un determinado equipamiento a proteger y el conocimiento profundo de las condiciones de explotación del mismo. Es por ello, que es necesario tener profundos conocimientos de matemáticas, computación, comunicaciones, redes eléctricas, del principio

de funcionamiento y regímenes de operación del equipamiento a proteger. Resulta aún más difícil la preparación de profesores que se estima en alrededor de 15 años [1]. Desarrollo: Este tema ha sido discutido en muchas universidades y eventos científicos del mundo [2,3,4,5] y en especial por investigadores de la facultad de Electromecánica de la Universidad de Camagüey, los cuales han desarrollado cierta experiencia al respecto y consideran que debe de buscar colaboración internacional para la preparación adecuada de estos profesionales en toda Iberoamérica [6]. Por otra parte, es necesario, el desarrollo de laboratorios y la obtención de equipamiento para esta preparación, así como la utilización cada vez más creciente de la Educación a distancia y la preparación en el puesto de trabajo. Hay una excelente experiencia en la UANL en el doctorado de sistemas eléctricos, donde se han formado varios doctores en esta rama. Para ambos (profesionales y profesores), el sistema prevé las siguientes acciones de posgrado: Cursos de posgrado. Diplomados. Entrenamientos especializados. Especialidad o maestría. Doctorado. Con lo que se cubren todos los niveles de enseñanza posgraduada, incluyendo la obtención de grados científicos. A continuación se describen los diferentes niveles, considerando sus objetivos y cursos generales con el número de créditos asignados: Para los especialistas: Cursos de posgrado: Horas Créd Protección de sistemas eléctricos

60 4

Protección de sistemas industriales

60 4

Estos cursos se han impartido durante varios años como parte de la continuidad en la formación de los egresados de la carrera de Ingeniería Eléctrica. Constituyen la formación básica para enfrentar algunos problemas de protecciones. Diplomado en Protecciones eléctricas Horas Créd Procesos transitorios en Sistemas eléctricos.

20 2

Protección de sistemas de transmisión de energía.

40 3

Protección de circuitos de distribución.

30 2

Protecciones de la industria. 30 2 Protección de máquinas eléctricas.

30 2

Protección con relés estáticos. 60 4 Trabajo final 30 2 Totales 240 17 Este Diplomado se ha impartido durante varios años y ha permitido que los egresados se especialicen en los temas de Protecciones eléctricas. El egresado alcanza un buen nivel de formación sin llegar a la especialización. Entrenamientos especializados: Se realizan de acuerdo a planes de formación particular y generalmente se utilizan para tratar de dominar una tecnología en particular. Se desarrollan de forma personalizada y preferiblemente en el puesto de trabajo. Especialidad en Protecciones Eléctricas: Horas Cred Diseño, modelación y simulación de sistemas

80 6

Matemática avanzada 60 4 Teoría de redes eléctricas 60 4 Mediciones eléctricas 60 4 Máquinas eléctricas 60 4 Electrónica digital y microprocesadores

80 6

Regímenes de operación de los sistemas eléctricos de Potencia

60 4

Automática de los sistemas eléctricos

60 4

Relés 60 4 Inteligencia artificial aplicada a las protecciones

60 4

Protección de Sistemas de transmisión y distribución de energía

45 3

Protecciones industriales 45 3 Protección de máquinas eléctricas

45 3

Protección de sistemas de comunicaciones y transmisión de datos

45 3

Gerencia empresarial 60 4 Diseño de Experimentos 45 3 Protección contra las sobretensiones

45 3

Trabajo de tesis 225 15 Totales 1195 81 Se deben graduar especialistas que tengan las siguientes posibilidades: Dirigir y participar en el montaje, explotación, puesta en marcha, desarrollo y modernización de las instalaciones donde realiza su trabajo profesional. Dirigir y participar en trabajos técnicos y de desarrollo en la rama de las Protecciones. Esta especialidad ha sido propuesta a impartirse en la Universidad de Camagüey a partir del próximo año y participan como coauspiciadoras otras Universidades y Ministerios del país. Habiendo recibido esta preparación, el especialista puede trabajar en cualquier campo relacionado con las protecciones y se considera que tiene los conocimientos necesarios para desarrollar cualquier labor dentro de este campo. Doctorado en Protecciones eléctricas: Si el egresado adquiere la preparación anterior no es necesario que continúe recibiendo asignaturas, sino desarrollar un trabajo investigativo y preparar una tesis que defenderá en un tribunal creado al efecto y con

preferencia a algunos de los temas que se proponen a continuación: Desarrollo de algoritmos para relés digitales. Diseño y construcción de relés digitales Evaluación del comportamiento de las protecciones eléctricas. Para los profesores: En cuanto a la formación de profesores esta es una tarea mucho más compleja, porque no solamente debe dominar los conocimientos en esta rama, sino además debe aplicar los principios pedagógicos, sicológicos, etc, que le permitan realizar una docencia con la calidad requerida. El profesor debe tener hábitos investigativos y de autosuperación continua para enfrentar estos retos. Cursos de posgrado: Horas Créd Pedagogía básica 60 4 Protección de sistemas eléctricos

60 4

Protección de sistemas industriales

60 4

Estos cursos se imparten continuamente como parte de la formación básica de los profesores y de los recién graduados. Diplomado en enseñanza de las protecciones Horas Créd Generalidades de la pedagogía en las ingenierías

40 3

Metodología de la enseñanza de las protecciones

30 2

Procesos transitorios en Sistemas eléctricos.

20 2

Protección de sistemas de transmisión de energía.

60 4

Protección de circuitos de distribución.

30 2

Protecciones de la industria. 30 2 Trabajo final 30 2 Totales 240 17

En cada una de las asignaturas del diplomado se analiza la preparación e impartición de cada uno de los temas de las mismas. Este Diplomado se impartió por una vez a un grupo de profesores que trabajan en la Educación superior y en la Enseñanza técnico profesional con muy buenos resultados para los Centros de estos niveles Entrenamientos en enseñanza de las protecciones: Se realizan de acuerdo a planes de formación particular y generalmente se utilizan para tratar de dominar un tema en particular. Se desarrollan de forma personalizada y preferiblemente en el puesto de trabajo. Maestría en Enseñanza de las Protecciones Eléctricas: Horas Cred Metodología de la enseñanza de las ingenierías

60 4

Metodología de la enseñanza de las protecciones

40 2

Diseño, modelación y simulación de sistemas

60 4

Matemática avanzada 60 4 Teoría de redes eléctricas 60 4 Mediciones eléctricas 40 3 Máquinas eléctricas 40 3 Electrónica digital y microprocesadores

60 4

Regímenes de operación de los sistemas eléctricos de Potencia

60 4

Automática de los sistemas eléctricos

60 4

Relés 60 4 Inteligencia artificial aplicada a las protecciones

60 4

Protección de Sistemas de transmisión y distribución de energía

45 3

Protecciones industriales 45 3 Protección de máquinas eléctricas

45 3

Protección de sistemas de 45 3

comunicaciones y transmisión de datos Gerencia empresarial 60 4 Diseño de Experimentos 45 3 Protección contra las sobretensiones

45 3

Trabajo de tesis 225 15 Totales 1215 81 Los graduados de Master deben de tener las siguientes posibilidades: Impartir las asignaturas relacionadas con la rama de las protecciones eléctricas. Desarrollar investigaciones y participar en trabajos técnicos y de desarrollo en la rama de las Protecciones. Esta maestría ha surgido del diseño curricular realizado por un grupo de profesores de la Facultad de Electromecánica de la Universidad de Camagüey y participan como coauspiciadoras otras Universidades del país. Doctorado en Enseñanza de las Protecciones eléctricas: Después de haber adquirido la preparación anterior, el profesor no necesita recibir más asignaturas, sino dedicarse a la autosuperación en temas determinados para desarrollar un trabajo investigativo y preparar una tesis que defenderá en un tribunal creado al efecto y con preferencia a algunos de los temas que se proponen a continuación: Perfeccionamiento de la impartición de las protecciones. Elevación de la calidad de la enseñanza de las protecciones, introduciendo nuevas técnicas, contenidos y medios de enseñanza. Elaboración de modelos pedagógicos en la impartición de la asignatura Formación de habilidades en los egresados de pre y posgrado. Otros planes de formación de profesores: Independientemente de este plan de formación adecuado para los profesores de protecciones hay otras posibilidades de superación en la rama de la enseñanza de las ingenierías para

formar profesores con conocimientos más profundos en la rama pedagógica, que también forman un sistema: Diplomado en enseñanza de la ingeniería y la tecnología: Horas Créd Fundamentos de Didáctica de la Educación Superior

60 4

Diseño curricular en la ingeniería y las carreras tecnológicas

60 4

Metodología de la enseñanza de la ingeniería y la tecnología

60 4

Métodos activos de enseñanza para la ingeniería y la tecnología

60 4

Trabajo práctico 30 2 Totales 270 18 Maestría en enseñanza de la Ingeniería: Horas Créd Fundamentos de didáctica de la Educación Superior

60 4

Fundamentos sicológicos del proceso docente educativo

60 4

Pedagogía general y tendencias pedagógicas contemporáneas

60 4

Diseño curricular en las ingenierías

60 4

Informática para ingenierías 60 4 Formación de profesores en Ingenierías

60 4

Metodología de la enseñanza de la Ingeniería

60 4

Metodología de la investigación pedagógica

60 4

Diseño experimental 60 4 Taller de investigación científica

60 4

Totales 600 40

Conclusiones El sistema de superación en Protecciones eléctricas concibe varios niveles a alcanzar por el egresado de Ingeniería Eléctrica y ramas afines, lo que garantiza su especialización y competitividad en las labores de la industria, pudiendo obtener el grado científico de Doctor en Ciencias Técnicas El plan de formación de profesores de Protecciones eléctricas abarca una alta preparación pedagógica para enfrentar los retos actuales que impone la impartición de las asignaturas y la preparación técnica en la rama de las protecciones que garantiza tener un alto conocimiento en esta disciplina. Referencias: [1] Colectivo de autores. Encuesta a graduados y especialistas de la producción, los servicios y la docencia. Facultad de Electromecánica, 1992. [2] Corrales, L., Nápoles, E. La formación de profesores en Ingeniería. Evento Internacional Ingered: Julio de 1997. [3] Corrales, L., Nápoles, E. Proyecto de centro formación de profesores de ingeniería. COPIMERA: Setiembre de 1997. [4] Corrales, L. Proyecto de superación posgraduada en la enseñanza de la Ingeniería. Evento Internacional La Educación en Ingeniería: Perspectivas al inicio del tercer Milenio. Mayo de 1999. [5] Corrales, L., Nápoles, E. Un modelo de superación posgraduada en la enseñanza de la Ingeniería. 1ra Conferencia Internacional “La eficiencia Energética y el Medio ambiente”. Mayo 2001. [6] Corrales, L., Nápoles, E. Informe final de la investigación del Premio UDAL 1999: Un modelo de superación posgraduada en la enseñanza de la Ingeniería. Julio 2000.

Autores: Luis Corrales Barrios Se graduó en 1978 de Ingeniero Electricista en la Universidad de Camagüey y obtuvo el título de Doctor en Ciencias Técnicas en la Universidad Central de Las Villas. Es Profesor Titular del Dpto. de Ingeniería Eléctrica de la Facultad de Electromecánica de la Universidad de Camagüey. Ha impartido 22 cursos de posgrado y ha asesorado 6 entrenamientos en las ramas de las protecciones, la computación y la pedagogía. Ha participado en 48 Eventos científicos con carácter nacional e internacional con más de 60 trabajos. Tiene 58 artículos publicados o aprobados a publicar Ha desarrollado y dirigido investigaciones para el desarrollo de relés estáticos. Tiene 2 normas y 9 patentes de invención. Tiene registrado los programas de computación: ProMaq (versión 1,0) y Cálculo de las Protecciones de máquinas eléctricas. Ha recibido 40 cursos de posgrado y 3 entrenamientos (de ellos uno en el Instituto Politécnico de Bielorrusia) de su especialidad y de perfil pedagógico. Es miembro del Consejo Científico y de la Comisión de Grados Científicos de la Universidad de Camagüey. Ernesto Vázquez Martínez Se graduó de Ingeniero en Electrónica y Comunicaciones en 1988, y obtuvo su Maestría y Doctorado en Ingeniería Eléctrica en 1991 y 1994 respectivamente, en la Universidad Autónoma de Nuevo León, México. Desde 1996 es Profesor Investigador del Programa Doctoral en Ingeniería Eléctrica de la Universidad Autónoma de Nuevo León, México, y de 2000 a 2001 realizó una estancia posdoctoral en la Universidad de Manitoba, en Canadá. Actualmente es el Coordinador del Postgrado en Ingeniería Eléctrical de la misma universidad. Es miembro del Instituto de Ingenieros en Electricidad y Electrónica (IEEE) de Estados Unidos y miembro del

Sistema Nacional de Investigadores de México, Nivel I. Sus áreas de investigación son la protección de sistemas eléctricos de potencia y la aplicación de técnicas de inteligencia artificial en sistemas eléctricos de potencia. Aurelio Ramírez Granados Se graduó en 1980 de Ingeniero en Electrónica y Comunicaciones en el Instituto Tecnológico de Estudios Superiores de Monterrey (ITESM), obtuvo el grado de Maestría en el Postrado de la Facultad de Ingeniería Mecánica y Eléctrica de la Universidad Autónoma de Nuevo León (1998), participó como investigador en el Instituto de Investigaciones Eléctricas (IIE) de Cuernavaca Morelos México, ha impartido cursos de Matemáticas y Comunicaciones vía microondas en el ITESM, catedrático en Eléctrica e Informática en la Universidad de Monterrey (UDEM). Ha participado en eventos nacionales (5) e internacionales (1), tiene asesoradas más de 60 tesis a nivel licenciatura en las áreas de la Física, Matemáticas y Computación, actualmente catedrático de tiempo completo por más de 20 años de antigüedad en la Facultad de Ciencias Físico Matemáticas de la Universidad Autónoma de Nuevo León y se encuentra desarrollando el doctorado en “Protecciones de Generadores mediante Laboratorios Virtuales” en la Universidad de Camaguey, Cuba.

1

Resumen--Se describe TREMA, una herramienta informática

fundamental para la gestión de la información de interés para los especialistas que realizan análisis de perturbaciones. TREMA realiza la recolección en forma automática de la información generada en los relés de protección: los registros oscilográficos y de eventos. La procesa, la organiza en una base de datos y la presenta al analista en forma clara y fácil de trabajar, minimizando el tiempo invertido en la recopilación manual de la misma. El desarrollo enfrenta la diversidad de modelos y protocolos de relés definiendo módulos para la interrogación de cada uno y un formato de almacenamiento normalizado en una base de datos.

Palabras Clave--análisis de perturbaciones - relés de protección - eventos - registros oscilográficos - base de datos - recolección automática.

I. INTRODUCCIÓN L análisis de perturbaciones, es decir, el estudio sistemático de las faltas que ocurren en el sistema eléctrico

de potencia de alta y extra alta tensión, juega un papel fundamental en la vida de una empresa eléctrica. A través de él se evalúa el comportamiento en régimen transitorio del sistema de potencia, se mide el desempeño de sus protecciones, se generan acciones correctivas y preventivas para mejorar la calidad del servicio y se determinan responsabilidades sobre los sucesos ocurridos.

El proceso del análisis de perturbaciones, a los efectos de su mejor explicación y comprensión, puede dividirse en dos subprocesos que se suceden uno a continuación del otro: el Análisis de Perturbaciones en Tiempo Real (ATR) y el Análisis de Perturbaciones en Tiempo Diferido (ATD).

El primero, el subproceso ATR, comienza cuando una perturbación es detectada por el accionamiento del sistema de alarmas en el Centro de Control. El operador analiza la información mínima necesaria para comprender lo sucedido y

realiza las maniobras correspondientes para la reconexión del sistema afectado en forma segura y rápida, minimizando el tiempo de corte y aislando los componentes averiados para su posterior recuperación y mantenimiento.

El siguiente subproceso, el ATD, se desencadena una vez que el operador informa de la perturbación ocurrida y de las maniobras efectuadas para restablecer el sistema eléctrico. El ATD consiste en un análisis pormenorizado de los hechos que busca determinar exactamente lo que sucedió para luego establecer por qué ocurrió y cómo evitarlo en el futuro. Los objetivos principales del ATD son evaluar el comportamiento del sistema eléctrico de potencia y el desempeño de los relés de protección, y definir acciones de mantenimiento correctivo y preventivo que minimicen los daños en los equipos ante una futura perturbación de características similares. Otro objetivo, que adquiere cada vez más relevancia, es determinar las responsabilidades sobre los hechos ocurridos y los perjuicios ocasionados.

El ATD es realizado por un equipo de especialistas, principalmente integrado por profesionales con probada experiencia en protecciones y gran conocimiento del sistema de potencia. Los registros oscilográficos y de eventos generados por los relés y registradores de perturbaciones cuando ocurre la falta en el sistema eléctrico son la materia prima fundamental para dicho análisis. Obtener esta información en forma confiable, rápida y segura es todavía una tarea ardua y difícil de realizar a pesar de la renovación tecnológica de los últimos años. Los especialistas en protecciones, quienes son responsables de la recopilación de la información, son plenamente conscientes de ello.

El advenimiento de los dispositivos electrónicos inteligentes, que integran funciones de protección, control y registro, ha dejado obsoletos hace años ya a los antiguos registradores oscilográficos de papel y a otros mecanismos electromecánicos de registro. Los nuevos equipos numéricos generan abundante y detallada información que permite comprender más a fondo el comportamiento de la red. Sin embargo, la diversidad de marcas y modelos, los diferentes protocolos de comunicación propietarios de cada fabricante, las dificultades y demoras en lograr un acuerdo en la normalización de los formatos de la información, la nomenclatura diversa y la disparidad de software para la comunicación de los dispositivos, hace que la interpretación de la información generada sea todavía un problema y que la recopilación de la misma continúe siendo un dolor de cabeza y un empleo de recursos excesivo.

TREMA: Una Herramienta Fundamental para el Análisis de Perturbaciones

J. Zorrilla de San Martín, U.T.E., y V. González Barbone, IIE Universidad de la República, Uruguay

E

2

El sistema de Transmisión de Registros, Medidas y Ajustes (TREMA) es un desarrollo realizado en el Instituto de Ingeniería Eléctrica de la Universidad de la República Oriental del Uruguay para la Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.), la Empresa Eléctrica del Uruguay.

TREMA es un sistema automático de recolección de registros oscilográficos y de eventos que integrará paulatinamente a todos los relés de protección y registradores instalados en el país para el sistema de transmisión. Enfrenta la diversidad de modelos y protocolos definiendo módulos para la interrogación de cada uno, un formato común de almacenamiento en base de datos y una lista de eventos normalizados hacia la cual se mapean las múltiples denominaciones asignadas por cada fabricante a los mismos eventos. Este formato único de registro, donde se guarda además en su totalidad el registro original generado en el dispositivo correspondiente, sin pérdida de información, inclusive si tiene asociados archivos binarios de registros oscilográficos u otros, permite la consulta sencilla de los eventos del sistema eléctrico sin necesidad de atender a las denominaciones propias de cada fabricante y modelo.

El sistema mantiene información sobre las entidades del sistema eléctrico tales como subestaciones, equipos de potencia y relés de protección. Un esquema de roles definidos por el usuario administrador permite autorizar selectivamente a diferentes personas (especialistas en protecciones, operadores de Centro de Control, técnicos, gerentes, etc.) la consulta de registros, la operación del sistema modificando o agregando entidades nuevas, o enviar información en forma automática vía correo electrónico de sucesos relevantes. Las acciones del sistema se programan en el tiempo mediante una agenda donde se ingresa el tipo de acción deseada (interrogar subestación, informar eventos, transferir configuraciones, etc.) y el momento o la frecuencia de la ejecución de la tarea.

II. ANTECEDENTES La automatización de la recopilación de registros para el

análisis de perturbaciones es una idea que se remonta a principios de los años noventa. Surge con la instalación y puesta en servicio de los primeros relés de protección numéricos en nuestro sistema eléctrico. La completa y detallada información generada en dichos equipos preveía el logro de mejores resultados en el análisis, comprendiéndose en aquél entonces la necesidad de incrementar la dedicación de recursos para realizar esa tarea. Para obtener esos recursos adicionales era imprescindible reducir el tiempo invertido en la búsqueda y recopilación de la información generada en cada perturbación del sistema eléctrico. Automatizar este trabajo se transformó entonces en una meta prioritaria.

Ya no se estaba hablando de registros oscilográficos en papel, los que tardaban varios días en llegar al escritorio del ingeniero, ni de datos verbales o planillas confeccionadas por los operadores de subestaciones, sino de información de eventos y archivos de registros oscilográficos que podían obtenerse a través de una comunicación digital y un medio

informático. Las primeras iniciativas concretas se remontan al año 1995,

cuando se realizaron en la Empresa las primeras especificaciones de un sistema automático de recolección de registros generados por los relés de protección y registradores de perturbaciones [1], el cual consistía básicamente en un desarrollo de software y computadores que, instalados en las subestaciones, se conectaban con los relés para extraerles la información por ellos generada, la que luego era transmitida a un computador central ubicado en el Centro de Control de Protecciones, lugar de trabajo habitual de los especialistas responsables de realizar los estudios correspondientes.

Estas especificaciones teóricas se vieron materializadas por primera vez en el año 1999, mediante un desarrollo realizado en exclusiva para U.T.E. por el fabricante de relés ALSTOM [2]. A través de su uso se comenzaron a visualizar las ventajas de un sistema automático con tales características, que permitía disponer rápidamente de información para el análisis de lo sucedido. Con el paso del tiempo y debido a la experiencia adquirida se fueron sugiriendo nuevas prestaciones al sistema, planificándose un nuevo desarrollo que además organizara los registros oscilográficos y de eventos en una base de datos, que se generalizara para todas las marcas y modelos de los relés numéricos instalados, que fuera flexible y se adaptara fácilmente a la expansión de la red eléctrica y a la renovación de equipamientos, que tuviera diferentes perfiles de usuario con mayores niveles de seguridad y que mejorara la interfaz de usuario para facilitar la búsqueda de información, visualización y consulta a través de páginas web.

Esta nueva versión del desarrollo se ha logrado recientemente en un esfuerzo conjunto entre U.T.E. y el Instituto de Ingeniería Eléctrica de la Universidad de la República Oriental del Uruguay a través de un convenio para la elaboración de un sistema que tuviera en cuenta la especificación original incorporando además las mejoras descritas. TREMA es el resultado de este trabajo mancomunado, que luego de superadas algunas dificultades administrativas, pudo iniciarse a principios del año 2001 y hoy ha dejado de lado ya la fase experimental para comenzar a explotarse en forma definitiva. Incluye modelos de relés de protección marca General Electric (MOR y DLP), ALSTOM (K-serie, MiCOM y EPAC) y ABB (RELxxx y línea SPACOM). Una próxima etapa de desarrollo incorporará nuevos modelos al sistema y también a los registradores de perturbaciones que prestan servicio en la red eléctrica uruguaya.

III. UNIFORMIZACIÓN DE LA DIVERSIDAD TREMA enfrenta la diversidad de formatos de registro y

protocolos de comunicación empleados por los diversos fabricantes en dos planos: • Manejo separado de cada protocolo de comunicación, por

módulos de software independientes, con una interfaz común hacia el núcleo del programa encargado de realizar la interrogación de los relés y registradores. Esto habilita el agregado de nuevos protocolos sin fisuras, al hallarse

3

determinada la forma en que el nuevo módulo deberá interactuar con el núcleo de interrogación.

• Definición de un formato unificado para presentar la información recogida por los relés, independiente del formato propietario de origen, con un conjunto de datos de interés general para todo tipo de registro. Este formato unificado es soportado en la estructura de la base de datos, permitiendo búsqueda y selección de registros por diversos criterios. Este enfoque configura un nivel lógico unificado por encima de la diversidad física de los modelos. El analista de perturbaciones opera en este nivel lógico común. La diversidad de protocolos alcanza aún la denominación de

los fenómenos: faltas, eventos, registros, informaciones correspondientes a las mismas medidas o hechos físicos aparecen denominadas diferente en los distintos modelos de relé. El análisis de comportamiento de la red apunta, en definitiva, a los hechos físicos. No es el analista quien debe adaptarse a las múltiples denominaciones de un mismo fenómeno, sino él mismo quien debe poder caracterizar los fenómenos de su interés con una denominación significativa para él, siguiendo el modelo corporativo consagrado en la práctica de la Empresa Eléctrica. TREMA permite definir un conjunto de "eventos normalizados", donde una denominación única para un mismo fenómeno físico puede ser mapeada desde las múltiples denominaciones de ese mismo fenómeno empleadas por los diferentes protocolos en sus tipos de evento propietario. Las denominaciones propietarias no pueden cambiarse: han sido fijadas por el fabricante del modelo; la solución es asignar a cada uno de estos eventos propietarios un evento normalizado. Procediendo de ese modo con todos los modelos manejados, el operador puede realizar su búsqueda o selección de registros basado en un tipo de evento conocido, cuya denominación es la habitual en la empresa donde trabaja.

El formato único de registro en la base de datos contiene un encabezamiento común a todos los tipos de registro, con fecha, hora, subestación, equipo, relé y tipo de evento normalizado, pero además guarda el registro en su formato propietario original, en las denominaciones propias del fabricante, sin pérdida de información, inclusive si tiene asociados archivos binarios de registros oscilográficos u otros. TREMA permite extraer de la base de datos y grabar en disco estos archivos binarios asociados, para ser examinados posteriormente mediante el empleo de software específico. Véase Fig. 1.

IV. ARQUITECTURA DE TREMA El despliegue (deployment) del sistema, como se muestra en

el esquema de la Fig. 2, incluye los siguientes elementos: • TremaSE, programa ejecutable residente en el computador

personal de la subestación. Este módulo conoce la configuración de la subestación (equipos, buses, relés, protocolos, eventos propietarios y normalizados) y la forma de interrogar cada modelo de relé. Realiza la interrogación de los relés de la subestación en forma secuencial, almacenando los registros en formato normalizado.

• TremaCC, programa ejecutable residente en un computador del Centro de Control. Siguiendo los dictados de una Agenda armada por el operador, a las horas indicadas establece comunicación con la subestación a interrogar, transfiere los registros desde la subestación al Centro de Control y los ingresa en la base de datos.

• La base de datos, instalada en un computador del Centro de Control, eventualmente el mismo donde corre TremaCC.

• El servidor web, instalado en un computador del Centro de Control, eventualmente el mismo donde corren la base de datos y TremaCC.

• Estaciones de operador o consultante, ubicadas en el Centro de Control, la subestación o cualquier punto de la red interna de la Empresa, o aún fuera de ella a través de la Internet, con un navegador. La estación de trabajo no requiere instalación de ningún software especial, basta con un navegador.

La instalación de TremaSE en el computador de la

subestación es sencilla, no requiriendo capacitación especial en informática más allá del nivel de usuario. La transferencia de las configuraciones de subestación se programan en la Agenda, se realizan automáticamente, y el módulo TremaSE la relee cuando ha cambiado.

La transferencia de registros desde la subestación al Centro de Control contempla especialmente algunos puntos delicados, como ser: • Bloqueo del acceso concurrente a los registros por el

ingreso proveniente de la interrogación de los relés y su preparación para la transmisión.

• Previsiones de recuperación ante cortes de energía, con particular consideración a evitar la pérdida de datos.

• Compresión de registros en un archivo único para la transferencia, que no se considera realizada hasta que el archivo transferido ha sido recibido exitosamente en el Centro de Control.

Fig. 1. Visualización de un Registro Oscilográfico.

4

Consideraciones similares se atienden al descomprimir el archivo de registros en el Centro de Control e ingresar los registros en la base de datos.

La introducción de redundancia controlada ha sido otro de los paradigmas empleados en la conservación de la información: al crearse el registro en formato normalizado, en la subestación, ya contiene todos los datos necesarios para su

inclusión en la base de datos, sin hacer referencia a información ajena al mismo. Los registros que ingresan a la base de datos lo hacen tal cual han sido generados.

En suma, el diseño apunta, en todos sus términos, al aseguramiento en la recolección de información, a la conservación de su integridad, a la facilidad de consulta y al control de acceso.

Fig. 2. Diagrama de Software y Formatos de Intercambio.

5

V. OPERACIÓN

A. La consulta. La consulta de registros almacenados en la base de datos

puede realizarse seleccionando: • Fecha y hora inicial (desde). • Fecha y hora final (hasta). • Tipos de evento normalizado, hasta 5; muestra todos los

eventos correspondientes a cualquiera de los elegidos. • Hasta 2 subestaciones. • Hasta dos equipos. • Hasta dos relés, posiblemente ubicados en diferente

subestación, en diferente o el mismo equipo (si se trata de una línea, que aparece en dos subestaciones diferentes). La Fig. 3 muestra la pantalla inicial del sistema, a través de

la cual el usuario accede al mismo y la Fig. 4 el perfil de consulta.

La consulta muestra una lista de todos los registros correspondientes a la selección. Véase detalle en la Fig. 5. La omisión de un elemento cualquiera es interpretada como "todos". En particular, si no se elige tipo de evento normalizado, se muestran todos los registros de cualquier tipo que cumplan las restantes condiciones.

La lista de registros muestra un identificador único para el registro, fecha y hora hasta el milisegundo, subestación, equipo, relé, tipo de evento normalizado y texto crudo, campo donde se muestra la información del registro tal como la presenta el fabricante del relé, con un mínimo de elaboración para hacerla legible.

El operador puede elegir así un conjunto de registros bien específico en torno al fenómeno que le interesa analizar. Un caso de interés especial es el análisis de falta en una línea, donde pueden seleccionarse registros de los relés asociados a la línea en cada una de las subestaciones extremas, o restringir a lo registrado por un cierto relé en un extremo y un relé acompañante en el otro extremo.

Un clic sobre el identificador de un registro abre una ventana de detalle donde se presenta toda la información colectada para ese registro en particular, a pantalla completa. Ver Fig. 6.

La lista de registros, presentada en forma de tabla, habilita la transferencia a una planilla electrónica por la simple operación de seleccionar en la ventana del navegador, copiar y pegar en la ventana de la planilla electrónica.

B. Mantenimiento de entidades. El sistema permite al operador ingresar todos los datos

correspondientes a las entidades de la red eléctrica sobre la cual trabaja: subestaciones, equipos, buses de comunicación, relés. Puede también definir los Tipos de Evento normalizados según la denominación habitual en su empresa, y mapear los Tipos de Evento Propietario de los distintos protocolos hacia las denominaciones normalizadas correspondientes.

C. Autorizaciones y roles. El control de acceso es por usuario autorizado en la base de

datos. El Administrador, un usuario plenipotenciario, puede ingresar nuevos usuarios, quienes deberán validarse ante el sistema con una contraseña. Un nuevo usuario puede tener privilegios de administrador, si así se determina. A los usuarios no privilegiados puede asignárseles un Rol de Consulta, en el cual se limita su potestad de consulta a un cierto conjunto de subestaciones y tipos de evento. Si se trata de un operador, puede asignársele un Rol de Operación, consistente en una lista de operaciones y opcionalmente subestaciones sobre las que le es permitido modificar entidades (relés, equipos, tipos de evento y otros). El Administrador puede definir tantos roles de operación y de consulta como necesite, para que cada uno vea o haga lo que puede ver o hacer, y sólo eso.

Un tercer tipo de rol, el Rol Informar, permite definir qué eventos, relativos a qué subestaciones, deben ser informados a un usuario por correo electrónico, según las horas a las que esta tarea haya sido incluida en la Agenda.

Fig. 3. Pantalla Inicial.

Fig. 4. Perfil de Consulta.

6

D. Agenda. La Agenda permite controlar la realización automática de

tareas, ya sea en forma periódica o por única vez. El operador puede ingresar mes, día del mes, día de la semana, hora y minuto, con opciones de indicar "todos" en cada uno de los campos, y la tarea a realizar. Ejemplos típicos de tareas en agenda son transferir registros desde una subestación al Centro de Control, transferir configuración de subestación desde el Centro de Control a la subestación correspondiente, enviar correo electrónico de eventos especiales de la última jornada a los usuarios incluidos en Roles Informar según lo especificado en cada rol.

VI. HERRAMIENTAS DE DESARROLLO El análisis y diseño del sistema se realizó empleando el

lenguaje de modelado UML. La base de datos empleada fue Oracle 8i, aunque la instalación final se realizó en Oracle 9i. El servidor web es Apache, incluido en el software de Oracle, así como el controlador de acceso a la base de datos desde Java (JDBC), y el motor para las Java Server Pages (JSPs) de las páginas en el servidor web. La configuración del sistema, los datos de las subestaciones, el mapeo de eventos y otros datos de uso en el Centro de Control o transferidos a la subestación emplean el lenguaje estándar XML. Los módulos TremaCC y Trema SE están escritos en Visual C++ de Microsoft. La plataforma de desarrollo fue MS-Windows 2000. Tanto la plataforma de desarrollo como el manejador de base de datos fueron requerimientos del cliente.

VII. POSIBILIDADES DE EXPANSIÓN La dirección más inmediata de expansión es la

incorporación de nuevos protocolos. La especificación de una interfaz hacia el núcleo del módulo TremaSE permite programar un nuevo módulo de soporte de protocolo de interrogación de relés o registradores integrándolo sin fisuras al sistema. Esta expansión ha sido prevista desde el primer momento, y si bien la incorporación de un nuevo módulo

requiere una recompilación, no va más allá de eso. Se espera como consecuencia natural de la utilización del

programa en producción la sugerencia de mejoras operativas en diversos sentidos, difíciles de adelantar al día de hoy, que serán incorporadas seguramente en futuras versiones.

En cuanto a la incorporación de nuevas plataformas de desarrollo, el punto más laborioso es la migración del código MS Visual C++, que tropieza con las limitaciones impuestas por las particularidades propietarias de Microsoft. No se trata de una valla insalvable, pero requiere horas de programador. El resto del sistema cambia de plataforma fácilmente: tanto la base de datos Oracle como el servido web Apache están soportados en múltiples plataformas, y las páginas JSP corren en cualquier plataforma donde esté instalado Apache con el módulo correspondiente.

En un criterio conservador, aún tratándose de un software en funcionamiento, al no contar con un tiempo prudencial en producción, debería concebirse la versión 1 de TREMA como un "prototipo terminado", un software utilizable pero donde seguramente cabrán muchas mejoras. El proceso de desarrollo ha aportado un conocimiento bastante profundo de la naturaleza del problema; el diseño modular, la separación "lógico-física" adoptada desde el inicio, han permitido a lo largo del desarrollo enfrentar exitosamente las dificultades que fueron surgiendo. Este hecho, más que ningún otro, aumenta nuestra confianza en la capacidad de evolución de este producto.

VIII. REFERENCIAS Especificaciones Técnicas:

[1] J. Alonso, R. Normey, L. Gaggero y J. Zorrilla de San Martín, “Especificación Técnica del Sistema Automático de Relés de Protección y Registradores de Perturbaciones”, U.T.E., 1995.

Ponencias Técnicas Publicadas:

[2] J. Zorrilla de San Martín, “El Sistema Automático de Gestión de Relés de Protección y Registradores de Perturbaciones de U.T.E.” SIPSEP-00-21, pp. 299-308, Nov. 2000.

Fig. 5. Lista de Registros.

Fig. 6. Detalle de Registro.

7

IX. BIOGRAFÍAS Juan Zorrilla de San Martín nació en Montevideo, República Oriental del Uruguay, el 28 de setiembre de 1965. Recibió el título de Ingeniero Electricista en la Facultad de Ingeniería de la Universidad de la República Oriental del Uruguay en el año 1994. Ingresó a la Administración Nacional de Usinas y Transmisiones Eléctricas (U.T.E.) en 1989 y a partir de ese año comenzó a desarrollarse en la temática de protecciones. Ha asistido a seminarios y cursos de especialización al respecto y publicado trabajos relativos a su área de trabajo. Actualmente ocupa el cargo de Subgerente de Ingeniería y Desarrollo de Protecciones en la Administración. Es miembro de la I.E.E.E. desde 1989. Víctor González Barbone nació en Montevideo, República Oriental del Uruguay, el 13 de noviembre de 1949. Luego de haber cursado estudios en Electrónica, se orientó hacia la informática, graduándose como Ingeniero de Sistemas en Computación.

Trabajó en mantenimiento eléctrico en la refinería de A.N.C.A.P. en La Teja, Montevideo, siendo destinado luego a programación de mantenimiento. Su desempeño en la Gerencia de Protecciones de U.T.E., la empresa eléctrica uruguaya, le permitió familiarizarse con los relés, el análisis de perturbaciones y sus problemas asociados. Ha actuado también como administrador de sistemas y encargado de redes en Centro de Informática, institución de enseñanza privada. Actualmente es Profesor Adjunto en el Instituto de Ingeniería Eléctrica de la Facultad de Ingeniería, Universidad de la República (UDELAR). Sus intereses profesionales se orientan hacia los sistemas operativos, la administración de redes, el análisis y la gestión de proyectos informáticos. Como parte del equipo docente del proyecto TREMA intervino en las etapas de análisis y diseño del sistema, manejo de la base de datos, supervisión del diseño y desarrollo de la interfaz de usuario, y en la instalación y puesta en marcha.

SOFTWARE PARA LA DETERMINACIÓN DE LOS NIVELES DE CORTOCIRCUITO EN SISTEMAS INDUSTRIALES

Alexis Martínez del Sol Juan J. Sánchez Jiménez Juan M. García Martínez José A. Gómez Reyna Enrique Cisneros Sedano Ismael Díaz Verduzco Mariano Zerquera Izquierdo

Universidad de Guadalajara Centro Universitario de Ciencias Exactas e Ingenierías

Departamento de Mecánica-Eléctrica. Av. Revolución 1500 CP 44430 . Tel: (01)(33) 3619-8367.

Fax (01) (33) 3619-9973 Guadalajara, Jal, México.

Resumen: En este trabajo se presenta la metodología de para la determinación de los niveles de cortocircuito en sistemas eléctricos industriales y un software diseñado al efecto, el cual permite la determinación de las corrientes y voltajes en todos nodos y elementos del diagrama unifilar del sistema, para los diferentes tipos de cortocircuitos que se pueden presentar. Además realiza una comprobación de los elementos desconectivos ante un cortocircuito máximo, mostrando los parámetros necesarios para que puedan soportar esta condición. Actualmente se trabaja para que el software tenga la capacidad de realizar la coordinación de protecciones.

I. INTRODUCCIÓN Un aspecto importante a considerar en la operación y planeación de los Sistemas Eléctricos (SE) es su comportamiento en condiciones transitorias y un caso de interés especial lo representa el comportamiento en condiciones de cortocircuito. La condición normal de operación de un SE es sin falla, no obstante esto, es imposible evitar la presencia de fallas en las instalaciones por distintas causas fuera del control humano. Debido a lo anterior se debe considerar que un equipo o sistema en condiciones de falla puede sufrir daños que en ocasiones son graves por lo que es necesario diseñar las instalaciones en tal forma que contengan los elementos necesarios de protección y señalización. El cálculo de las corrientes de cortocircuito es uno de los pasos principales en el diseño de un sistema de protección. Sus resultados son necesarios para la selección de equipos (no sólo de protección) y para el cálculo de parámetros de ajuste de protecciones. En forma general un estudio de corto circuito brinda información que permite:

a) Determinar las características interruptivas de los elementos de desconexión de las corrientes de corto circuito como son interruptores, fusibles, restauradores y fusibles de potencia principalmente.

b) Realizar un estudio para la selección y coordinación de los dispositivos de protección contra las corrientes de corto circuito.

c) Realizar los estudios térmicos y dinámicos debidos a los efectos de las corrientes de corto circuito en algunos elementos de las instalaciones como son sistemas de barras, tableros, cables, buses de fase aisladas, etc.

d) Relacionar los efectos del corto circuito con otros estudios de sistema como por ejemplo los estudios de estabilidad de las redes eléctricas en sistemas de potencia.

Las fuentes básicas de corrientes de cortocircuito en un sistema eléctrico son a) El sistema eléctrico de potencia exterior. b) Los generadores. c) Los motores sincrónicos d) Los motores asincrónicos. II. REPRESENTACIÓN DE ELEMEN-TOS Un estudio de cortocircuito se inicia siempre con un diagrama unifilar del sistema por estudiar, donde se indiquen todos los elementos que van a intervenir, especialmente las fuentes y los elementos pasivos donde los valores de potencia, tensiones e impedancias siempre que sea posible. El diagrama unifilar debe transformarse en un diagrama de impedancias que muestre el circuito equivalente de cada componente del sistema referido al mismo lado de uno de los

transformadores para estudiar el comportamiento en condiciones de carga o al presentarse un cortocircuito.

Los circuitos equivalentes para el estudio de cortocircuito de los distintos componentes del sistema son los siguientes: - Generadores y Motores: La representación

elemental de una máquina sincrónica es una fuente de tensión en serie con una impedancia. Los motores de inducción se representan igual que las máquinas sincrónicas pero se considera su contribución al cortocircuito solo en los primeros ciclos.

- Transformadores: Generalmente se

representan por su circuito equivalente “T” ignorando su rama magnetizante.

- Líneas de transmisión y Cables: El

circuito equivalente a utilizar depende de la longitud de la línea, usándose el modelo “π” para líneas largas y medias, mientras que para las líneas cortas y cables se representan como una resistencia en serie con una inductancia.

- Cargas: Se pueden modelar como

impedancias de valor constante que consumen potencia activa y reactiva. En estudios de cortocircuito se representan como circuitos abiertos.

- Sistemas externos: Se modela por el

circuito equivalente de Thevenin donde la tensión equivalente depende de las tensiones internas de los generadores y la impedancia equivalente depende del resto de elementos del sistema.

Es común que en algunos sistemas no se conocen los valores de impedancias, a tal efecto se ha desarrollado una metodología de estimación de parámetros, la cual se describe a continuación. III. ESTIMACIÓN DE PARÁMETROS Toda la metodología de estimación se basa en la introducción en el programa de catálogos completos de elementos que permiten en algunos casos calcular y en otros interpolar valores para los parámetros que no se suministren. En todos los casos los parámetros que son estimables o calculables aparecen en el programa como parámetros “opcionales”. Esto

significa que si los parámetros se dan por el usuario se respeta ese valor, de lo contrario se calculan o estiman según sea el caso. Se prefiere siempre la introducción de los datos reales por parte del usuario y usar la estimación como recurso cuando no se dispongan de datos. Desde el punto de vista visual los parámetros “opcionales” se muestran en blanco y para estimar su valor tienen que estar en blanco (el “cero” es un valor que el sistema lo considera como entrado por el usuario y por tanto no estima el valor de los parámetros puestos en “cero”). Una vez estimado un parámetro para volver a estimarlo hay que borrar (dejar en blanco su valor). Se estiman parámetros solo en los siguientes elementos: • Generadores sincrónicos. • Motores sincrónicos. • Motores de inducción. • Transformadores. • Líneas. A continuación la metodología para determinar cada parámetro estimable. Generadores sincrónicos: Para estimar los parámetros del generador sincrónicos se parte de haber introducido literalmente los datos: • Tipo: Turbogenerador ó Polos Salientes. • Velocidad • Potencia • Factor de Potencia Los parámetros estimables son: • Eficiencia • Relación X/R • Reactancia subtransitoria • Reactancia de secuencia cero Los parámetros pueden ser estimados explícitamente en el diálogo de entrada de datos de cada uno de estos elementos a través del botón “Calcular”. No Obstante si no se entran explícitamente estos valores a la hora de ejecutar cualquier estudio los parámetros son estimados de forma transparente por el sistema, o sea, siempre están presentes cuando se realiza el estudio.

Motor Sincrónico. Para estimar los parámetros del motor sincrónico se hace similar al generador sincrónico. En el motor sincrónico se parte de haber introducido literalmente los datos: • Tipo: Rotor cilíndrico ó Polos Salientes. • Velocidad • Potencia • Factor de Potencia Los parámetros estimables son: • Eficiencia • Relación X/R • Reactancia subtransitoria • Reactancia de secuencia cero Motor Inducción. Para estimar los parámetros del motor de inducción se sigue más o menos la misma metodología. Datos imprescindibles de entrar por el usuario: • Tipo: Estándar ó Alta Eficiencia • Velocidad • Potencia Los parámetros estimables son: • Factor de Potencia • Eficiencia • Reactancia a rotor bloqueado. • Reactancia de secuencia cero Líneas: En las líneas hace falta para estimar los siguientes datos: • Tipo: Puede ser

o Cable 3f (mag): Cable trifásico en conducto magnético

o Cable 3x1 (mag): tres cables monofásicos en conducto magnético

o Cable 3f: Cable trifásico sin conducto magnético

o Cable 3x1: tres cables monofásicos sin conducto magnético

o Línea Aérea o Barra

• Longitud • Calibre • Material

Se estiman: • R de secuencia positiva y cero (ohm km) • X de secuencia positiva y cero (ohm km) • B de secuencia positiva y cero (1/ohm km) Transformadores La estimación de parámetros en los trasformadores también obedece a la existencia de tablas con valores redeterminados y se hacen interpolaciones para los valores no encontrados en las tablas. Los parámetros necesarios para estimar los parámetros en los transformadores son:

• Potencia • Voltajes por alta y por baja.

Se determinan por estimación los parámetros:

• Impedancia equivalente • Resistencia equivalente

FaultChecker es un software a la medida diseñado para analizar el comportamiento de los sistemas industriales ante la presencia de las corrientes de cortocircuitos. Constituye una herramienta muy útil para el personal dedicado al diseño, explotación y mantenimiento de los sistemas eléctricos industriales. Las facilidades fundamentales que brinda el sistema son las siguientes: • Diseño gráfico visual del unifilar de la red

industrial. El software se soporta sobre una plataforma gráfica donde de forma natural y sencilla se dibuja y describe toda la red eléctrica de la industria.

• Introducción directa de los datos de los elementos eléctricos sobre el propio unifilar del circuito sin necesidad de utilizar grandes tablas. Cada elemento de la red tiene asociado un diálogo de datos donde se introducen todos los parámetros asociados a ese elemento.

• Facilidades de edición visual altamente sofisticadas al estilo de los programas visuales más renombrados.

• Exploración sencilla y rápida por todos los elementos de la red a través del explorador de elementos.

• Chequeo riguroso de los datos para evitar cualquier omisión que pueda afectar los cálculos.

• Estimación de muchos parámetros de los elementos de la red cuando no se dispone de ellos a través de bases de datos internas totalmente actualizadas.

• Estudio de fallas que incluye todos los tipos posibles (trifásica, bifásica, monofásica y bifásica a tierra). Se muestran los resultados de las corrientes en el punto de falla así como en los demás elementos del sistema.

• Muestra de las corrientes circulantes por cada punto de la red (módulo y ángulo) y voltajes en los nodos o barras del sistema, para la falla seleccionada, con solo dar un clic sobre el elemento seleccionado.

• Estudio de régimen máximo de cortocircuito en todos los nodos de la red con el objetivo de chequear el funcionamiento correcto ó no de los desconectivos y alimentadores existentes. Se prueba en cada desconectivo ó alimentador su funcionamiento una vez calculado el nivel de cortocircuito máximo.

• Facilidades de impresión de reportes tal como se muestran en pantalla con mínima utilización de recursos.

• Posibilidades de análisis de múltiples redes sin restricción de tamaño.

• Sencillez máxima en la manipulación del sistema respetando estándares internaciones de programación.

• Utilización de pocos recursos del sistema. La aplicación FaultChecker es del tipo MDI (Múltiple Document Interface) lo que permite que en una sola instancia del programa se puedan procesar a la vez muchos documentos (circuitos). Para cada circuito existe una ventana MDI y las acciones de menú en muchas ocasiones son válidas solo para la ventana MDI activa.

Fig. 1 Ventana principal de FaultChecker

Los requerimientos del sistema de este programa son mínimos:

Windows 98/ME/NT/2000 Procesador Pentium ó superior 16 MB RAM (32MB Recomendado) 5 MB Libres en Disco Duro Tarjeta VGA o superior Dispositivo apuntador(Mouse)

IV. EDITOR GRAFICO Todo el programa FaultChecker gira alrededor de un editor gráfico propio con la mayoría de las facilidades que brindan otros simuladores visuales eléctricos. Cuando se comienza a editar un nuevo circuito aparece una nueva ventana formada por un panel de inserción y un panel explorador como se muestra en la Figura 2 La barra de herramientas de elementos está dividida en tres segmentos: Nodos o Barras: Elementos básicos de la red

que permiten la colocación de los demás componentes.

Elementos Series:Elementos que enlazan dos o más nodos.

Elementos paralelos: Elementos que se conectan directamente a los nodos. Además de la inserción de elementos el Editor Gráfico con que cuenta FaultChecker brinda un gran número de facilidades de edición similares a la de los mejores ejemplos de editores visuales que existen en el mercado. Para realizar las acciones que a continuación se explican es necesario que el editor se encuentre en estado de “Edit” y para ello se debe seleccionar el elemento Cursor de la barra de elementos.

Fig 2 Editor gràfico

V. PANEL EXPLORADOR Con el objetivo de acelerar al máximo el proceso de edición y de reportes se habilitó el sistema con un explorador de elementos muy al estilo de la plataforma Windows para la que está diseñado el programa. En la medida que se insertan elementos eléctricos en el panel de inserción en este panel explorador se adicionan

los elementos organizados de manera análoga a la barra de elementos.

Este panel de exploración (Fig. 3) es de solo lectura por lo que desde él no se pueden adicionar elementos ni cambiar los nombres de las categorías o grupos de elementos ni el nombre del elemento.

VI. ELEMENTOS ELECTRICOS Los elementos eléctricos que se necesitan insertar para formar el unifilar de la red industrial se muestran en una barra de herramientas conocida como barra de elementos eléctricos.

Esta barra está formada de tres partes: la primera es una caja de diálogo desplegable para mostrar la escala en que se visualiza el editor, la segunda parte es otra caja de diálogo desplegable para seleccionar el tipo de elemento a insertar y la tercera es la dedicada precisamente a seleccionar el elemento a insertar, como se muestra en la Fig.4

Fig. 3 Panel explorador.

Fig. 4 Barra de elementos

La lista de elementos organizados por categoría se muestra a continuación: Nodos o Barras:

Barras o tableros Terminales

Series: Líneas Transformadores Transformadores de 3 devanados Capacitores Series Reactores Series

Paralelos: Sistema Generador Motor de Inducción Motor Sincrónico

Capacitor Grupo de Cargas

A modo de ejemplo veremos el dialogo que se muestra para las líneas.

VII. LINEAS Elemento tipo serie cuyo símbolo en la barra de elementos es Cuando se insertan las líneas se coloca un interruptor de bajo voltaje por el envío y el recibo no tiene interruptor. Al dar doble clic sobre la línea se muestra el siguiente diálogo (ver Fig.5) Los datos del elemento tipo Línea son: Nombre: Nombre interno totalmente configurable. Por defecto el sistema le coloca el prefijo Línea seguido de un número consecutivo que se muestra en el unifilar. Voltaje: Voltaje nominal al que se encuentra trabajando la línea, tiene que coincidir con los voltajes de los nodos a los que se conecta. Esto lo verifica internamente el programa. Tipo: Tipo de línea que une dos nodos. Puede ser de los siguientes tipos: - Cable 3f (mag). Cable trifásico en conducto

magnético. - Cable 3*1 (mag). Cable de tres conductores

monofásicos por conducto magnético. - Cable 3f. Cable trifásico en conducto no

magnético. - Cable 3*1. Cable de tres conductores

monofásicos por conducto no magnético. - Línea aérea. - Barra Metros: Longitud en metros del tramo de línea. Circuitos: Cantidad de circuitos en paralelos que conforman la línea. Cuando entre dos tableros se colocan más de una línea en paralelo con el mismo calibre no hay necesidad de dibujar ambas líneas, colocando la cantidad de líneas en este campo se obtiene el mismo resultado.

Material: Material constructivo de la línea (Cobre o Aluminio). Calibre: Calibre de la línea. En caso de ser cable o línea aérea coincide con el calibre de la línea, se puede además indicar la unidad de norma utilizada para denominar el calibre (AWG, MCM ó mm2). Se incluye hasta el calibre 700 MCM o su equivalente en mm2. En el caso de las barras se introduce el valor de la corriente de la barra.

Fig. 5. Dialogo correspondiente a las Líneas El significado de los parámetros de la línea que aparecen en la página “Parámetros” es: R (Ohm/km): Resistencia serie en Ohm/km de secuencia positiva y de secuencia cero. X (Ohm/km): Reactancia serie en Ohm/km de secuencia positiva y de secuencia cero. B (1/Ohm·km): Susceptancia capacitiva paralelo en 1/Ohm·km de secuencia positiva y de secuencia cero. Similar a estos parámetros, se introducen los datos para los demás elementos del diagrama unifilar del sistema industrial. El programa internamente estima muchos parámetros, como eficiencia de motores y transformadores, reactancias de conductores, etc.

VIII INTERRUPTORES Aunque en el cálculo de los niveles de cortocircuito no intervienen los interruptores, el programa permite la edición de estos, agrupados en las siguientes categorías: (Alto voltaje, Bajo voltaje, Fusible y la categoría especial “Sin

Interruptor” que es equivalente a eliminar el interruptor desde el unifilar). Los datos generales son: V: Voltaje nominal del interruptor en V. A: Corriente nominal del interruptor en A. kA(Int): Corriente de interrupción en kA. En función de la categoría de interruptor escogida pueden aparecer diferentes parámetros. Alto Voltaje kA(Max): Corriente de interrupción máxima en kA. Rating: Simétrico: Capacidad interruptiva simétrica. Asimétrico: Capacidad interruptiva asimétrica Ciclos: 2 : Disparo instantáneo a 2 ciclos. 3 : Disparo instantáneo a 3 ciclos. 5 : Disparo instantáneo a 5 ciclos. 8 : Disparo instantáneo a 8 ciclos. Bajo Voltaje Tipo: Tipo de interruptor de bajo voltaje.

Caja moldeada: Interruptor de caja moldeada

Potencia (fusible): Interruptor de aire con fusible)

Potencia (fusible): Interruptor de aire sin fusible)

FP(%): Factor de potencia de prueba del interruptor.

Fusible FP(%): Factor de potencia de prueba del

fusible. (no es necesario )

IX ESTUDIOS Este programa está concebido para el cálculo de los niveles de cortocircuitos en una industria y para la comprobación de los interruptores existentes ante la presencia de fallas máximas. Básicamente se realizan dos estudios distintos: Calculo de Fallas y Pruebas de CC Máximo. Al estudio de fallas se puede acceder a través del comando Análisis de Fallas del menú Estudios ó a través de un botón equivalente en la barra de acceso rápido. Con este estudio se pueden determinar las corrientes de cortocircuito ante la presencia de una falla específica en un nodo y las contribuciones de cada elemento de la red así como las transferencias de corriente por cada sección del circuito. Además determina los voltajes en cada uno de los nodos del circuito. Cuando se invoca este estudio se realiza previamente un chequeo de datos si no fue realizado explícitamente con anterioridad. Una

vez verificados todos los datos necesarios se muestra un diálogo inicial para especificar la falla, como se muestra en la Fig. 6 Este diálogo consta de tres partes; la primera destinada a ubicar la falla, se especifica el nodo donde ocurre la falla así como su ubicación específica en ese nodo (puede ser en el propio nodo ó a la salida de cualquier interruptor conectado a ese nodo).

Fig. 6 Determinación del tipo de cortocircuito. La segunda parte está destinada a especificar el tipo y parámetros de la falla. Estos parámetros son: Tipo: Tipo de falla (trifásica, bifásica, monofásica y bifásica a tierra). Para las fallas que involucran tierra debe preverse la introducción de los datos de las impedancias de neutro de los elementos conectados en estrella. Tiempo: Tiempo al que se calcula la falla. ½ ciclo: Cálculo a ½ ciclo (corriente subtran-sitoria) 1½ - 4 ciclosCálculo a 1½ - 4 ciclos (corriente transitoria) 30 ciclos Cálculo a 30 ciclos (corriente estable) Régimen: Si la falla se analiza para régimen de generación máxima o mínima. R arco: Resistencia de arco en Ohm si existe. La tercera parte del diálogo está dedicada a los botones de manipulación y las barras de progreso. Una vez concluido el estudio se muestra una etiqueta de selección para brindar la posibilidad de mostrar o no los resultados del

estudio de fallas. Cuando se sale del estudio, los elementos calculados del unifilar se dibujan de color azul. El estudio o Prueba de CC Máximo permite determinar para un cortocircuito máximo en todos los nodos del sistema las siguientes corrientes para todos los elementos de la red: • Corriente subtransitoria pico • Corriente subtransitoria total. • Corriente subtransitoria simétrica • Corriente transitoria simétrica Con estas corrientes calculadas se pasa a desarrollar un algoritmo de pruebas para interruptores y alimentadores. Con el resultado de estas pruebas se determina si los dispositivos colocados en la red están preparados para resistir u operar ante la presencia de un cortocircuito máximo. Para ello hay que determinar la corriente de cortocircuito ajustada a los valores reales de X/R a los que se hizo la prueba de los interruptores por los fabricantes. En realidad los calores de los parámetro de los interruptores y fusible (Corriente de interrupción, voltaje, etc.) están referidos a una relación R/X de prueba ideal de laboratorio que no tiene por qué coincidir con la rea en que se encuentra trabajando en el circuito que se calcula con FaultChecker. Por ejemplo en la Fig. 7 se muestra las pruebas realizadas a un interruptor que paso satisfactoria las pruebas y la Fig. 8 a uno que no la soporto.

X REPORTES El programa FaultChecker después de efectuar sus estudios brinda un grupo de reportes para cada estudio agrupado en dos categorías:

• Reportes generales para el estudio • Reportes individuales de los elementos

para el estudio. Cuando se realiza un determinado estudio los resultados de los estudios anteriores se pierden por lo que no coexisten los resultados de dos estudios consecutivos, solo se mantienen los resultados del último estudio realizado. Los reportes obtenidos se pueden exportar a formato de Word, Excel y Txt, además de un formato interno del programa.

Fig. 7 Prueba exitosa de un interruptor.

FaultChecker cuenta además con un visualizador propio de archivos de resultados.

Fig. 8 Prueba fallida a un interruptor.

XI. CONCLUSIONES

1. Se dispone de un software para el Cálculo de los diferentes tipos de cortocircuito en sistemas industriales.

2. Se dispone de un editor con todas las facilidades de edición de Windows, el cual permite la edición rápida y sencilla del unifilar del sistema.

3. El software permite además el calculo de los niveles de cortocircuito máximo y el chequeo de los elementos desconectivos

(interruptores y alimentadores) ante esta magnitud de corriente.

4. Todos los resultados pueden ser exportados a formatos estandares, incluyendo el diagrama unifilar del sistema industrial.

5. El software realiza la estimación de muchos parámetros a partir de una poderosa base interna de datos.

6. Se dispone de una ayuda on line en cualquier parte del programa.

7. Actualmente se trabaja para incluir la posibilidad de que el software realice la coordinación de protecciones.

X. BIBLIOGRAFÍA

1. Willian D. Stevenson Jr., “Elementos de Análisis de Sistemas de Potencia”

2. Arun G. Phadke, James S. Thorp, “Computer Relaying for Power Systems”, Research Studies Press LTD., John Wiley & Sons Inc., 1988.

3. Altuve H. Ferrer, “Protección de Sistemas Eléctricos de Potencia”, Universidad Autónoma de Nuevo León.

4. Altuve H. Ferrer, “Introducción a los Relevadores y Sistemas Digitales de Protección”, Universidad Autónoma de Nuevo León.

5. Altuve H. Ferrer, “Tópicos Selectos de Protección Digital de Sistemas Eléctricos de Potencia”, Universidad Autónoma de Nuevo León.

6. Altuve H. Ferrer, “Apuntes sobre Fundamentos Matemáticos de los Algoritmos de Protección Digital”, Universidad Autónoma de Nuevo León.

XII. BIOGRAFÍAS

Alexis Martínez del Sol nació en Cienfuegos Cuba. Recibió el grado de Doctor en Ingeniería Eléctrica en la Universidad Central de Las Villas, Cuba en 1997. Desde 1987 trabajó en la Facultad de Ingeniería Eléctrica de la Universidad Central de Las Villas donde fue profesor Asistente y jefe del departamento de Electroenergética entre 1997 a 1999. Desde 1999 trabaja como Profesor Investigador C del Departamento de Ingeniería Mecánica Eléctrica de la Universidad de Guadalajara. Su área de investigación es en el control, diseño y protección de motores eléctricos. E-Mail: [email protected] Juan José Sánchez Jiménez nació en Placetas, Cuba, en 1942.Recibió el grado de Doctor en Ingeniería Eléctrica en la Universidad Central de Las Villas. Cuba en 1990. Desde 1969 trabaja en la Facultad de Ingeniería Eléctrica de la Universidad Central de Las Villas. Donde es Profesor Titular y Jefe de la Disciplina Ingeniería Eléctrica. Su área de investigación es el diseño, explotación y protección de sistemas eléctricos industriales E-mail: [email protected]

CONDICIONES OPTIMAS DE OPERACIÓN DE LA SUBESTACION LA QUEBRADA, POR LA PUESTA EN SERVICIO DEL TERCER BANCO T82C, DE 85/23 kV, 30 MVA

Ing. Tomás Castellanos López

Gerencia de Aseguramiento de la Calidad Luz y Fuerza del Centro

Av. Pie de la Cuesta No. 273, San Andres Tetepilco, Iztapalapa C.P. 09440, Mexico, D.F. Teléfonos: 56-34-50-38, Fax: 56341154

Email:[email protected] RESUMEN El presente estudio se realiza para analizar las condiciones óptimas de operación de la Subesta-ción La Quebrada de 85/23 kV, debido a la insta-lación del tercer banco (T82 C) y dos alimentado-res de 23 kV ( LAQ-27 y LAQ-28), con una car-ga estimada de 9 MW (251 A) según estudios de la Gerencia de Distribución. Para ello, se calculan los niveles de cortocircuito trifásico y de fase a tierra en 85 y 23 kV, para cuatro diferentes condi-ciones de operación de los tres bancos de la subes-tación. Además se efectúa un análisis de cargas máximas de cada alimentador, para determinar el ajuste de las protecciones de sobrecarga máxima, de los bancos en el lado de 85 kV. INTRODUCCION La Subestación La Quebrada forma parte del sistema de transmisión de 85 kV de LyFC. El suministro de energía eléctrica para esta subesta-ción proviene de dos fuentes de generación fuertes (Lechería y Cerro Gordo) y conforma un arreglo en anillo a través de otra línea paralela entre estas dos subestaciones, tal y como se indica en el dia-grama unifilar de la Fig. 1. La Quebrada es una subestación de transforma-ción de 85/23 kV, encapsulada en SF6, con arre-glo de doble barra con interruptor de amarre, tanto en 85 como en 23 kV, ver Fig. 2.

Las condiciones que deben cumplir las subesta-ciones de 85/23 kV, en el Sistema Central de LyFC, para dar una mejor calidad de servicio a los consumidores deben ser las siguientes. • Continuidad de servicio. Esta condición se cumple de acuerdo con las con-sideraciones siguientes:

Recurso económicos disponibles Espacio adecuado para la construcción de la

S.E. Derechos de vía existentes para la instalación

de líneas Definición del arreglo de la S.E. Características propias del sistema del cual va

a formas parte Exigencias del servicio para el que está desti-

nado Necesidades de operación y mantenimiento

del equipo que lo constituye • Limitación del cortocircuito. El cortocircuito máximo trifásico llegó a alcanzar en algunos puntos de nuestro sistema de 23 kV (Cerro Gordo y Lechería) valores próximos a 1000 MVA (25100 A) a medida que se fueron ampliando las subestaciones con la instalación de nuevos bancos de transformadores que se conec-taban en paralelo con los existentes. Este valor de cortocircuito es excesivo para un sistema de dis-tribución de 23 kV, ya que no se fabrica equipo de interrupción normalizado para esta capacidad interruptiva, lo cual obliga a usar interruptores para 34.5 kV, lo que encarece los servicios conec-tados en 23 kV. Por lo anterior se decidió limitar el cortocircuito máximo en 23 kV a 500 MVA(12550 A), que es un valor más adecuado para este nivel de voltaje. Los bancos 82 que LyFC adquiere en el mercado nacional e internacional, lo hace de acuerdo a la norma LFC-ING-039, en ésta se especifica que los bancos deben ser trifásicos de 85/23 kV, 20/25/30

MVA y una impedancia nominal de 14.4 % con un rango de 15.48 % y 13.32 % referida a la base de 30 MVA; por lo tanto, para el caso de un solo transformador, el cortocircuito trifásico en 23 kV queda limitado por la impedancia del banco aproximadamente a 250 MVA, ya que la impe-dancia de secuencia positiva equivalente del sis-tema en las barras de 85 kV es de valores en dé-cimas de porciento, por lo que la impedancia del banco es la que predomina. El cortocircuito de fase a tierra se limita a un valor aproximado de 250 MVA (6275 A) conectando el neutro del transformador a tierra a través de un reactor de 0.4 Ω. En el caso de instalaciones de 23 kV, con arreglo de doble barra con interruptor de amarre, pueden conectarse dos o tres transformadores en paralelo, lo que provoca que el cortocircuito trifásico de fase y tierra alcancen valores mayores de 500 MVA. • Regulación de voltaje Los transformadores trifásicos instalados deben tener cambiador automático de derivaciones bajo carga, controlado por un sistema de regulación de voltaje que mantenga el voltaje regulado en 23 kV de 24.726 kV a 21.274 kV, en función de las con-diciones de la carga conectada. • Posibilidades de ampliación Se debe elegir un arreglo en las instalaciones físicas, que permitan instalar en la primera etapa dos transformadores conectados en forma inde-pendiente y en la segunda etapa se instala un ter-cer transformador que trabaje de reserva con los demás transformadores. • Flexibilidad de operación y mantenimien-

to. En las subestaciones de transformación de 85/23 kV, que alimentan directamente al sistema de distribución, en el caso de una desconexión de toda la subestación por una contingencia en las barras de 85 kV, no afecta a la red de alta tensión, sino únicamente a una porción del sistema de distribución. Además las subestaciones con doble circuito de alimentación por 85 kV, en caso de una contin-gencia o mantenimiento de una de las líneas, tie-nen la capacidad para llevar por un solo circuito, la carga de las dos líneas de 85 kV. En el arreglo de doble barra con interruptor de amarre en 85 kV, las condiciones normales de operación de éste es cerrado, lo que permite que cada juego de barras colectoras quede protegido por una protección diferencial independiente, de manera que en caso de una falla en cualquiera de las dos barras, no se pierda más de la mitad de la subestación.

Dentro del sistema de distribución de LyFC se encuentran instalados diferentes equipos como son : transformadores, capacitores, reguladores y conductores, los cuales cumplen una función específica. En este trabajo se hace énfasis en el comportamiento de los transformadores de poten-cia ante condiciones anormales de operación y de cortocircuito, estableciéndose los criterios norma-lizados para brindarles una adecuada protección. En cuanto al ajuste de la protección de sobrecarga máxima (51-1,2) en los bancos T82 por el lado de 85 kV, el criterio utilizado en Luz y Fuerza en un principio fue el de soportar una sobrecarga no mayor al 200 %, con lo cual se ajustaban la mayo-ría de estos bancos , sin embargo a raíz de los cambios en la norma IEEE std C57.91-1995, las protecciones de sobrecarga se calculan para un 160 % de sobrecarga máxima, sin causar daño a los aislamientos y devanados de estos transforma-dores. El tiempo de operación de esta protección se ajusta a un tiempo t=1.2 a 1.4 seg con falla trifásica en las barras de 23 kV, con la finalidad de coordinar con las protecciones de los sistemas de distribución de 23 kV y transmisión de 85 kV. DESARROLLO Los cálculos de cortocircuito en 23 y 85 kV y las corrientes de carga máxima medidas en las barras de 23 kV de la S.E. La Quebrada, se utilizan para el cálculo de ajustes de los relevadores de sobre-carga de fases (51-1,2) de los bancos T-82 A, B y C, de esta manera se verifica su comportamiento para las cuatro condiciones de operación propues-tas en este estudio y se muestran en las figuras 3,4,5,6,7 y se resume en la tabla 1. Los transformadores T82 A,B y C tienen una capacidad nominal de 30 MVA c/u. La corriente nominal (In) por 85 kV es: 30000 In=-----------=203 A √3*85 La corriente nominal (In) por 23 kV es: 30000 In=-----------=753 A √3*23 El ajuste de la protección de sobrecarga(51-1,2) de los bancos en el lado de 85 kV se calcula como sigue: Tap=In*FS/RTC RTC=400/5 A Tap=203*1.6/80=4.06 A

Por lo tanto se selecciona el Tap =4 A La sobrecarga con tap=4, está ajustada a 320 A El factor de sobrecarga (FS) expresada en por-ciento es: 320 FS=-----------=157 % 203 Considerando este factor de sobrecarga (FS) por el lado de 23 kV, tenemos: FS=1.57*753=1182 A Según datos de cargas máximas en los alimenta-dores de la subestación LAQ21=368 A LAQ22=290 A LAQ23=347 A LAQ24=72 A LAQ25=297 A LAQ26=0 A LAQ27+LAQ28= Carga futura, 251 A (9 MW) La suma total de las cargas es igual a 1625 A Distribuyendo la carga de los alimentadores en forma balanceada entre los bancos T 82 A y B tenemos Banco T82 A: LAQ21, 25, 27 y 28 Esta carga es igual a 916 A Banco T82 B: LAQ22, 23, 24 y 26 Esta carga es igual a 705 A Debido a que los alimentadores LAQ25 y LAQ26 alimentan a un interruptor de transferencia (TACI) se conectan a bancos diferentes. En caso de falla del alimentador LAQ25 automá-ticamente se transfiere la carga al LAQ26. En esta condición la carga del banco T82 B es de 1002 A. Ver tabla 2. CONCLUSIONES Del análisis obtenido en el desarrollo de esta po-nencia, el orden que se recomienda para las con-diciones óptimas de operación en la S.E. La Que-brada por 23 kV, son las siguientes. • Condición normal de operación T -82 A conectado a Barras 1 de 23 kV T-82 B conectado a Barras 2 de 23 kV T-82 C excitado y de reserva Interruptor de amarre de 23 kV abierto

• Condiciones especiales de operación Primera: T-82 A y C conectados en paralelo a Barras 1 de 23 kV T-82 B conectado a Barras 2 de 23 kV Interruptor de amarre de 23 kV abierto Segunda: T-82 A conectado a Barras 1 de 23 kV T-82 B yC conectados en paralelo a Barras 2 de 23 kV Interruptor de amarre de 23 kV abierto Tercera: T-82 A, T82-B y T82-C conectados a Barras 1 de 23 kV Interruptor de amarre cerrado La protección de sobrecarga de fases (51-1,2) de los bancos T-82 se utiliza como respaldo en el lado de 85 kV y resulta segura, simple y económi-ca, sin embargo es difícil de aplicar en el caso de la S.E. La Quebrada, ya que se requiere verificar continuamente las condiciones normales y espe-ciales de operación en la S.E. y dependiendo de los requerimientos de demanda de energía en los alimentadores de 23 kVen horas de carga máxima, los Operadores del Sistema se ven obligados a aplicar condiciones especiales de operación, co-nectando dos bancos en paralelo en esta S.E. • Los interruptores en SF6 de 23 kV tienen

suficiente capacidad para interrumpir las co-rrientes asociadas a fallas trifasicas y monofá-sicas, aun en la condición más desfavorable, que sería la conexión de los 3 transformado-res en paralelo.

BIBLIOGRAFIA: 1.-Diseño de Subestaciones Eléctricas José Raúll Martin Ed. McGraw Hill, 1990 2.-Protección de Instalaciones Eléctricas Indus-triales y Comerciales. Gilberto Enríquez Harper 3.-Manual de Diseño de Subestaciones Introducción. Compañía de Luz y Fuerza del Centro S.A. 1973 4.-Manual de Diseño de Subestaciones Capítulo 1, Diagramas de Conexiones Compañía de Luz y Fuerza del Centro S.A. 1978 5.-Transformadores y Autotransformadores de Potencia de 5 hasta 330 MVA Especificación LFC-ING-039 Luz y Fuerza del Centro Abril/2000 6.-Especificaciones de Transformadores de Poten-cia, Trifásicos, de 30 MVA, 85/23 kV LFC-ING-072 Luz y Fuerza del Centro Abril/1998

7.-Procedimiento para Coordinación de Protec-ciones de sobrecorriente en Sistemas de Distribu-ción. Comisión Federal de Electricidad Subdirección de Distribución, 1996 CURRICULUM VITAE

Tomás Castellanos López. Es Ingeniero Mecánico Elec-tricista, egresado de la Facul-tad de Ingeniería de la Uni-versidad Nacional Autónoma de México. Ingresó a laborar a Luz y Fuerza del Centro en el año de 1984, donde ha

desempeñado diferentes puestos. Actualmente ocupa el puesto de ayudante de jefe en la Superin-tendencia de Estudios de la Gerencia de Asegu-ramiento de la Calidad.

NIVEL DE CORTO CIRCUITO [A]

Barras 23 kV Barras 85 kV

CONDICION

DE OPERACIÓN

CONEXIÓN

DE LOS BANCOS

3ø øT 3ø øT

T82-A BARRAS 1 4907 4260

T82-B BARRAS 2 4949 4292

NORMAL

1 AMARRE ABIERTO

T82-C RESERVA - -

T82A y C BARRAS 1 9001 8066 2

AMARRE ABIERTO T82-B

BARRAS 2 4949 4292

T82-A BARRAS 1 9036 8095 3

AMARRE ABIERTO T82-ByC

BARRAS 2 4707 4260

E S P E C I AL 4

AMARRE CERRADO

T82 A,B y C BARRAS 1y2 12702 11685

17364

11397

TABLA 1.- Valores de corto circuito en los Ban-cos de la S.E. La Quebrada para cuatro diferentes condiciones de operación.

INTERRUPTOR TACI POR ALIM.

PREFERENTE EMERGENTE BANCO ALIMENTADOR LAQ

I max. [A]

VECES In

I max. [A]

VECES In

T82-A 21, 25, 27 y 28 916 1.20 594 0.78

T82-B 22, 23, 24 y 26 705 0.93 1002 1.30

TABLA 2.- Comparación de la corriente máxima y número de veces la corriente nominal en los Bancos 82 A y B.

CRG

LUQUE 2 LECHERIA 2

LUQUE 1 LA QUEBRADA LECHERIA 1

FORD JASSO 2 CUATITLAN

JASSO 1 COYOTEPEC CUATITLAN 2

VIC CTT CYO JAS

LAQ LEC

Figura 1.- Diagrama de la zona de 85 kV, S.E. La Quebrada

Figura 2.- Diagrama unifilar de la S.E. La Quebrada

Figura 3.- Condición de operación 1



Figura 6.- Condicion de operación 4

1

CÁLCULO DE AJUSTES, COORDINACIÓN DE PROTECCIONES

Y PUESTA EN SERVICIO DE UN BANCO DE CAPACITORES EN

85 kV, 31.2 MVAR, INSTALADO EN LyFC.

Ing. Miguel Angel Méndez Albores Ing. José Marcelino Santiago Jiménez

LUZ Y FUERZA DEL CENTRO

RESUMEN

Como consecuencia del aumento de la

carga y cambios en los niveles de tensión

en el sistema eléctrico de potencia en la

zona central del país, se hace necesario la

compensación reactiva en los voltajes de

transmisión de 230 y 85 kV. En el año

2001, Luz y Fuerza del Centro instaló su

primer Compensador Estático de Vars en

230 kV, de 300 MVAR capacitivo 75 MVAR

inductivo, en la Subestación Cerro Gordo y

se están instalando los primeros Bancos

de Capacitores en 85 kV en varias

Subestaciones, por lo que se hace

necesario establecer un criterio para el

cálculo de ajustes, coordinación de

protecciones, energización y pruebas con

las protecciones, así como realizar algunas

consideraciones para el comportamiento

del voltaje en función de los flujos de

potencia al momento de la energización.

INTRODUCCIÓN

Luz y Fuerza del Centro cuenta con una

red de transmisión en 85 kV, con una gran

cantidad de líneas y subestaciones, tanto

para alimentar bancos de transformación

de 85 a 23 kV, así como una gran cantidad

de clientes en este voltaje y que finalmente

se crea la necesidad de compensación

reactiva en el sistema.

Una solución a este problema ha sido la

instalación de Bancos de Capacitores en

85 kV, de 31.2 MVAR, conectados en

estrella no aterrizada (aislada) y a las

barras colectoras en este tipo de

Subestaciones. El primer banco se instaló

en la Subestación Magdalena en el año de

2001 y a la fecha se han seguido

instalando más bancos.

DESARROLLO

Estos bancos de capacitores en 85 kV

están conectados en una sola estrella no

aterrizada (aislada); construida de cuatro

grupos de capacitores conectados en serie

por fase y cada grupo está formado con 13

elementos (capacitores) en paralelo. Por lo

que de acuerdo a los niveles de tensión,

cumple con lo recomendado en la norma

IEEE C37.99-1990, como se muestra en la

tabla 1 y 2 .

2

Se tiene un total de 52 elementos por

fase y 156 en todo el banco, cada

elemento tiene una potencia reactiva de

200 kVAR; siendo la potencia reactiva total

del banco de 31.2 MVAR.

El banco se encuentra montado sobre

estantes con aisladores de porcelana

entre grupos y fases, está equipado con

cuchillas de puesta a tierra que conectan

todos los elementos (el neutro y los

estantes a tierra) al momento de operarlas.

Además del equipo propio del banco, se

instalan un interruptor de potencia, dos

juegos de cuchillas seccionadoras de

operación sin carga, transformador de

corriente, de potencial y apartarrayos

(como se muestra en la figura 1), en

Subestaciones con arreglo de doble barra

colectora con interruptor de amarre.

CONSIDERACIONES BÁSICAS PARA EL

CÁLCULO DE AJUSTES DE LAS

PROTECCIONES

Los bancos de capacitores deben de ser

capaces de operar en forma continua hasta

un 110% del voltaje nominal (rms) y hasta

un 135% de su potencia reactiva nominal,

incluyendo potencia reactiva debido a la

tolerancia de fabricación, así como los

voltajes fundamentales y todos los voltajes

armónicos que se presenten.

Deben de soportar una sobrecarga del

180% de la corriente nominal, incluyendo

la corriente fundamental y corrientes

armónicas, que se presentan en el

momento de energizar el banco.

TABLA 1.- BANCO DE CAPACITORES CONECTADOS EN ESTRELLA, NUMERO DE

GRUPOS SERIE

C37.99-1990

VOLTAJES DE LOS CAPACITORES DISPONIBLES (kV POR UNIDAD) VLL kV

VLN kV 21.6 19.92 14.4 13.8 13.28 12.47 9.96 9.54 8.32 7.96 7.62 7.2 6.64

500.0 288.7 14 15 20 21 22 29 30 35 36 38 345.0 199.2 10 15 16 20 21 24 25 27 230.0 132.8 10 14 16 17 18 20 161.0 92.9 7 13 14 138.0 79.7 4 6 6 6 8 10 11 12 115.0 66.4 5 7 8 9 9 10 69.0 39.8 2 3 3 4 5 6 46.0 26.56 2 4 34.5 19.92 1 2 3 24.9 14.4 1 2 23.9 13.8 1 23.0 13.28 1 2 14.4 8.32 1 13.8 7.96 1 13.2 7.62 1

12.47 7.2 1 Esta tabla muestra, para un voltaje del sistema en particular, el número de capacitores conectados en serie por fase, de un banco conectado en estrella que opera cerca del voltaje nominal de cada elemento (capacitor).

NOTAS:

Cada símbolo representa 13

capacitores en paralelo.

Fusibles individuales tipo

expulsión de 30 A, tipo k.

Apartarrayos.

Cuchilla de puesta a tierra.

Protección de sobrecorriente de

fases y tierra (50/51-1, 2, 3 y N)

Protección de sobrevoltaje (59).

A B C

50/51-1 50/51-2

50/51-3

50/51-N

85 kV

85 kV 85 kV

12.26 kV

12.26 kV

12.26 kV

12.26 kV

49 kV

59

TP 400:1

58

Barras 85 kV

1200/5A

SECUENCIA DE DISPAROS:

50/51-1, 2, 3 y N 59

DISP INT 58 ALARMA 50/51 F y N DISP INT 58 ALARMA DESBALANCEO DE VOLTAJE BCO. CAP. 85 kV

Icc 3φ = 13790 A

I1φ = 6010 A

212 A 212 A 212 A

FIGURA 1. DIAGRAMA DE PROTECCIÓN DE UN BANCO DE CAPACITORES EN 85 kV,

31.2 MVAR, CONECTADO EN ESTRELLA NO ATERRIZADA (AISLADA).

(4 GRUPOS DE CAPACITORES CONECTADOS EN SERIE POR FASE, CADA GRUPO CON 13 CAPACITORES EN PARALELO)

3

4

En casos especiales, debido a la

existencia de corrientes armónicas

relativamente altas en el punto de

instalación, la corriente total que toma el

banco puede ser superior del 135% de su

corriente nominal. Los capacitores no

generan armónicas, pero proporcionan una

trayectoria para las posibles condiciones

locales o generales de resonancia.

Se permite una tolerancia en la

capacitancia de más 15%, como se indica

en la norma IEEE C37.99-1990 y 2000.

TAMAÑO Y NÚMERO DE BANCOS

Se requiere realizar un estudio de flujos

de carga y estabilidad en el sistema de

transmisión para determinar, el punto de

localización y la capacidad del Banco de

Capacitores. Los capacitores minimizan las

pérdidas del sistema, incrementan el

voltaje del sistema y aumentan los

márgenes de estabilidad. Después de que

se conocen los requerimientos de potencia

reactiva, se determinan los MVAR de los

bancos de capacitores individuales y el

número de pasos.

Se recomienda que los bancos de

capacitores instalados en alta tensión

entren en pasos, para una mejor variación

del voltaje cuando se inserte en el sistema.

El tamaño máximo del banco, está

influeciado por los factores siguientes:

• Cambios en el voltaje del sistema sobre

la operación del banco de capacitores.

• Limitaciones en la corriente de servicio

del equipo de interrupción.

Cuando un banco de capacitores es

energizado, el voltaje del sistema se

incrementa y cuando es desenergizado el

voltaje disminuye. Para tener un efecto

mínimo en las cargas de los usuarios, este

cambio de voltaje se limita a valores entre

2 a 3% del voltaje nominal.

Este cambio de voltaje ( V) se calcula

con la expresión siguiente:

donde:

Q = Potencia reactiva trifásica total del

banco de capacitores (MVAR).

Scc = Potencia aparente de cortocircuito

trifásica, en el punto de instalación

del banco de capacitores (MVA).

El tamaño mínimo del banco está

influenciado por los factores siguientes:

• Consideraciones de desbalance en el

banco de capacitores.

• Coordinación de fusibles.

Cuando un fusible opera para proteger

un capacitor fallado, se puede presentar

una condición de desbalanceo que sujeta a

las unidades del mismo grupo conectadas

en serie a una sobretensión de 60 Hz. Un

criterio comúnmente aplicado es limitar

esta sobretensión al 110%, con una unidad

fuera. Esto requiere de un número mínimo

de unidades en paralelo que se da en la

tabla 2 (tomada de la norma IEEE C37.99-

1990).

Cuando un capacitor se cortocircuita

completamente, otros circuitos agrupados

en serie dentro del banco se sujetan a

sobretensiones de 60 Hz hasta que el

fusible opera. El fusible debe de operar

suficientemente rápido, para no dañar los

elementos en buen estado debido a esta

sobretensión. La norma IEEE std 18-1992

indica que se puede esperar que dentro de

su vida normal de servicio de un capacitor,

este soporte razonablemente una

Q V = ⎯⎯⎯ X 100 Scc

5

combinación total de 300 aplicaciones de

sobretensiones de frecuencia fundamental

entre terminales sin superposición de

transitorios o de contenido armónico, de

las magnitudes y duraciones como se

muestra en la tabla 3.

TABLA 3.- LÍMITES DE SOBREVOLTAJE

DE TIEMPO CORTO A FRECUENCIA

FUNDAMENTAL Y TEMPERATURAS

BAJO CERO.

Duración

Voltaje máximo permisible

(factor aplicado a Vnom

rms)

0.5 ciclos 3.0

1.0 ciclo 2.7

6.0 ciclos 2.2

15.0 ciclos 2.0

1.0 seg 1.7

15.0 seg 1.4

1.0 min 1.3

30.0 min 1.25

PROTECCIONES DE UN BANCO DE

CAPACITORES EN 85 kV

En los bancos de capacitores se utilizan

las protecciones siguientes:

Fusibles externos

Fusibles de expulsión de 30 A, tipo k, para

proteger individualmente cada capacitor

contra cortocircuito. Tiene la función de

aislar el capacitor fallado, previniendo la

ruptura del tanque y evitar la falla de un

capacitor que pueda causar daños a otros

capacitores del mismo banco; a otros

equipos instalados en las proximidades del

banco de capacitores o incluso accidentes

al personal. Además se tiene la ventaja

que se puede detectar visualmente él o

(los) elemento(s) fallado(s), tomando en

consideración que se deben formar

bancos de por lo menos 10 elementos en

paralelo por grupo/fase (ver tabla 2), para

TABLA 2.- NÚMERO MÍNIMO RECOMENDADO DE UNIDADES EN PARALELO POR

GRUPO SERIE PARA LIMITAR EL VOLTAJE AL 110 %, EN LAS UNIDADES

RESTANTES, CUANDO FALLA UN ELEMENTO DEL GRUPO

C37.99-1990

NUMERO

DE GRUPOS

SERIE

ESTRELLA CON

NEUTRO A TIERRA

O EN DELTA

ESTRELLA CON

NEUTRO NO

ATERRIZADO (AISLADO)

ESTRELLA DOBLE NO

ATERRIZADA DIVIDIDA EN

SECCIONES IGUALES

1 - 4 2

2 6 8 7

3 8 9 8

4 9 10 9

5 9 10 10

6 10 10 10

7 10 10 10

8 10 11 10

9 10 11 10

10 10 11 11

11 10 11 11

12 Y MAS 11 11 11

6

que la sobretensión en los elementos del

banco no exceda del 110% como se indica

en la norma.

También se debe tomar en cuenta que

cuando falla un capacitor en un grupo, el

fusible opera y disminuye la capacitancia

total del grupo.

Los fusibles deben seleccionarse con una

corriente nominal mínima del 165% de la

corriente nominal de cada capacitor en el

caso de protección individual, o de la

corriente nominal del banco, en caso de

protección en grupo.

Apartarrayos

Sirven para proteger el Banco de

Capacitores, cuando existan

sobretensiones por descargas

atmosféricas o por maniobra, etc.

Protección de sobrecorriente de fase y

tierra

Protegen al Banco de Capacitores contra

sobrecargas debidas a corrientes

armónicas y fallas de cortocircuito de una

fase a tierra, dos fases a tierra, bifásica o

trifásica.

Protección de Sobrevoltaje

En el neutro del Banco de Capacitores se

encuentra un transformador de potencial

de relación 400:1 y en el devanado

secundario se conecta un relevador de

sobrevoltaje. Para proteger contra

desbalance del neutro, ocasionado por la

falla de uno o más elementos

(capacitores); en algún grupo y fase del

banco.

Cálculos del banco previos a los ajustes

Potencia reactiva trifásica total del banco

de capacitores (Q) 31.2 MVAR, a 85 kV

Potencia reactiva por fase (QF)

Potencia reactiva por elemento (QE)

Corriente nominal del banco (IN)

Voltaje aplicado de línea a tierra (VN)

Voltaje aplicado en los elementos de cada

grupo (VE)

Corriente nominal de cada elemento (IE)

31200

3 QF = ———— = 10400 kVAR/fase

10400

52 QE = ————— = 200 kVAR/elem.

31200

3 X 85

Q

3 X VL-L

IN = ————— = ————— = 212 A

85

3

VL-L

3

VN = ————— = ———— = 49.07 kV

49.07

4

VE = ————— = 12.26 kV

200

12.26

QE

VE

IE = ————— = ———— = 16.3 A

7

Reactancia capacitiva de cada elemento

(XcE)

Capacitancia de cada elemento (CE)

Capacitancia de cada grupo por fase (CG)

Reactancia capacitiva de cada grupo por

fase (XcG)

Capacitancia total por fase del banco de

capacitores (CT)

Reactancia capacitiva total por fase del

banco de capacitores (XcT)

CÁLCULO DE AJUSTES Y

COORDINACIÓN DE PROTECCIONES

(50/51-1, 2, 3 y N)

En Luz y Fuerza del Centro se ha

adoptado como norma interna, para

proteger los bancos de capacitores en 85

kV con un relevador trifásico de

sobrecorriente de fases y de tierra con

elementos de tiempo e instantáneo, y un

relevador de sobretensión por

desplazamiento del neutro con un

elemento de tiempo definido; además trae

por diseño listones fusibles de 30 A para la

protección de cada elemento del banco de

capacitores.

El esquema debe operar en forma

instantánea y con retardo de tiempo

mínimo, para fallas comprendidas entre el

interruptor de potencia y el banco de

capacitores, teniendo la correcta

coordinación con los listones fusibles de 30

A para los diferentes tipos de fallas.

No deben de operar los relevadores de

fases con elementos instantáneos (50-1, 2

y 3), cuando se energice el banco de

capacitores (corriente de energización o

de inrush). La corriente de inrush tiene una

duración aproximada de 0.1 seg (6 ciclos)

y se puede calcular utilizando la expresión

siguiente:

y la frecuencia de inrush es :

ISC

———

IN

Imax = 2 ISC IN

f = fs

12260

16.3

VE

IE

XcE = ——— = ———— = 752.2 Ω

1

2 π (60) (752.2)

1

2 π f XcE

CE = ————— = ———————

CE = 3.5 X 10-6

F

CG = 13 X 3.5 X 10-6

= 4.6 X 10-5

F

1

2 π (60) 4.6X10-5

1

2 π f CG

XcG = ————— = ———————

XcG = 57.7 Ω

1

CT = ———————————————

+ + +

1

————

4.6X10-5

1

————

4.6X10-5

1

————

4.6X10-5

1

————

4.6X10-5

CT = 1.15 X 10-5

F

1

2 π (60) 1.15X10-5

1

2 π f CT

XcT = ————— = ————————

XcT = 230.7 Ω

8

donde:

Imax = Corriente máxima (pico) de

energización o de inrush.

ISC = Corriente de cortocircuito trifásica

(rms) en las barras, donde está

instalado el banco de capacitores.

IN = Corriente (rms) nominal del banco de

capacitores.

fs = Frecuencia del sistema (60Hz).

Criterios para el cálculo de ajuste del

elemento instantáneo de fases

(50-1, 2 y 3)

Para asegurar una buena sensibilidad

de los elementos instantáneos de los

relevadores de sobrecorriente de fase y

tierra, el ajuste se realiza considerando

que la impedancia del banco de

capacitores es mucho mayor que la del

sistema, debido a que la conexión del

banco es con neutro aislado,

prácticamente ocurre que las magnitudes

de las fallas de cortocircuito son a lo

mucho 3 veces la corriente nominal del

banco de capacitores, cuando falla un

elemento (capacitor) y además que las

magnitudes de las fallas entre el interruptor

y el tramo de barras para la conexión del

banco de capacitores son prácticamente

las mismas que en las barras de 85 kV.

Es práctica de Luz y Fuerza del Centro

dejar sensible los elementos instantáneos

de fases (tomando en cuenta la corriente

de energización o de inrush del banco) y

de tierra, dependiendo de la impedancia

del sistema; con esto se garantiza que el

relevador opere correctamente en

coordinación con el listón fusible de los

capacitores del banco y quede sensible

para fallas cercanas en la barra de 85 kV.

Los relevadores de sobrecorriente de

fases con elemento instantáneo, no deben

de operar en condiciones de energización

del banco de capacitores y deben quedar

coordinados con el listón fusible de los

elementos del banco.

El ajuste del relevador de sobrecorriente

de fases con del elemento instantáneo,

debe ser mayor que la corriente de

energización (inrush) o tener un retardo de

tiempo para evitar que el interruptor del

banco de capacitores se dispare en forma

incorrecta.

Ajuste de los relevadores de

sobrecorriente de fases, con elemento

instantáneo (50-1, 2 y 3).

Se calcula la corriente de cortocircuito

para una falla trifásica en las barras de 85

kV, siendo de :

I3φ=13790 A, RTC=1200/5 A.

La corriente de energización o de

inrush.

El criterio del instantáneo de fases es

dejarlo sensible para coordinar mejor con

el fusible, cuidando que no opere por

corriente de energización. Por lo que el

instantáneo se ajusta a un valor de :

Instantáneo de fases = 25 A

13790

240

I3φ

RTC

= ——— = ———— = 57.4 A Instantáneo

por fases

Imax = 2 ISC IN = 2 (13790) (212)

Imax = 2418 A

2418

240

Imax

RTC

Imaxsec = ——— = ———— = 10 A

9


elemento instantáneo de tierra (50-N)

El relevador de sobrecorriente de tierra,

no opera con las corrientes de

energización (inrush) del Banco de

Capacitores, por lo que puede quedar

sensible para las fallas de fase a tierra en

el interior del banco, el ajuste del elemento

instantáneo, se realiza considerando un

factor de seguridad del 35 al 50 % de la

corriente secundaria de cortocircuito de

fase a tierra en las barras de 85 kV.


sobrecorriente de tierra, con elemento

instantáneo (50-N).

Se calcula la corriente de cortocircuito

para una falla monofásica en las barras de

85 kV, siendo de :

I1φ = 6010 A

Considerando un factor de seguridad

(ks) del 35 al 50% de la corriente

secundaria de cortocircuito de fase a tierra,

se obtiene lo siguiente:


elemento de tiempo de fases (51-1,2,3)

Para lograr una buena sensibilidad de

los elementos de tiempo de los relevadores

de sobrecorriente de fase, el ajuste de la

corriente de arranque (tap), se realiza

considerando un factor de sobrecarga (FS)

del 160 % de la corriente nominal, para

evitar disparos indeseados por la presencia

de corrientes armónicas que provocan

sobrecargas en el banco.

Este factor de sobrecarga cumple con la

norma IEEE C37.99-1990, que considera

que los bancos de capacitores pueden

soportar una sobrecarga del 180 % de la

corriente nominal.

Selección del dial o palanca de los

elementos de tiempo (51-1,2 y 3).

El ajuste del dial o palanca de los

elementos de tiempo de los relevadores de

sobrecorriente de fase, se realiza para una

falla de cortocircuito trifásico en las barras

de 85 kV, considerando un tiempo de

operación del relevador de sobrecorriente

de tierra de 0.2 a 0.3 seg más un margen

de coordinación (Δt) que es de 0.3 seg, de

modo que el tiempo de operación del

relevador de fase, es la suma de los

tiempos y es de 0.5 a 0.6 seg ; para esto

es necesario conocer la curva de tiempo

de los relevadores y la relación de los

transformadores de corriente.


sobrecorriente de fases, con elemento

de tiempo (51-1,2 y 3)

La corriente de cortocircuito trifásico es

I3φ=13790 A, RTC=1200/5 A.

Se selecciona el tap fase= 1.4 A

6010

240

I1φ

RTC

IS = ——— = ———— = 25 A

= ks X IS = 0.5 X 25 = 12.5 A Instantáneo

de tierra

31200

3 X 85

Q

3 VLL

IN = ————— = ———— = 212 A

IN X FS

RTC

212 X 1.6

240

= ———— = ———— = 1.4

Corriente de

arranquefases

(tap)

= 12.5 A Instantáneo

de tierra

10

Se determina el dial o palanca de los

elementos de tiempo de los relevadores de

sobrecorriente de fase (51-1,2 y 3),

tomando en cuenta la corriente secundaria

de cortocircuito trifásica, que circula por

dichos relevadores y se selecciona un

tiempo de operación del relevador de fases

de 0.5 seg. El múltiplo de tap (MT), se

calcula de la manera siguiente:

Utilizando la curva muy inversa (U3) del

relevador, con un tiempo de coordinación

de 0.5 seg y un MT=41.6 veces se tiene un

dial o palanca de 5.0

Dial de fase=5.0

Criterios para el cálculo de ajustes del

elemento de tierra (51-N).

Para lograr una buena sensibilidad del

elemento de tiempo del relevador de

sobrecorriente de tierra, el ajuste de la

corriente de arranque (tap), se realiza

considerando un factor de desbalance (Fd)

del 10 al 30 % de la corriente de arranque

de fase (tap de la corriente mínima de

disparo de fase).

Selección del dial o palanca del

elemento de tiempo (51-N)

El ajuste del dial o palanca del elemento

de tiempo del relevador de sobrecorriente

de tierra, se realiza para una falla de

cortocircuito de fase a tierra en las barras

de 85 kV, considerando un retardo de

tiempo del 0.2 a 0.3 seg, para coordinar

con el elemento instantáneo de tierra y el

relevador de fases; para esto es necesario

conocer la curva de tiempo del relevador y

la relación de los transformadores de

corriente.


sobrecorriente de tierra, con elemento

de tiempo (51-N)

La corriente de cortocircuito monofásica

es I1φ = 6010 A.

Corriente de arranque de tierra (tap)tierra

(tap)tierra = Fd X tapfase = 0.3 X1.4 = 0.42

Se selecciona el tap de 0.5, porque es el

mínimo que tiene el relevador.

Tap tierra = 0.5 A

Siguiendo el criterio de selección del dial

(51-N), se calcula el MT.

Utilizando la curva muy inversa (U3) del

relevador, con un tiempo de coordinación

de 0.25 seg y un MT=50 veces.

Se tiene un dial o palanca de 2.5 .

Dial de tierra = 2.5

Cálculo de ajuste de la protección de

sobrevoltaje (59), por desbalanceo del

neutro.

Para lograr una buena sensibilidad del

relevador de sobretensión del neutro del

banco de capacitares, el ajuste de tensión

(tap) del relevador se realiza considerando

un sobrevoltaje del 110% del voltaje

nominal de fase a tierra, de acuerdo a la

norma IEEE C 37.99-1990 y 2000.

Se recomienda un retardo de tiempo de 2

segundos, este se debe verificar con el

tiempo de operación de la protección de

respaldo local y remoto para fallas a tierra,

para ver la manera de reducirlo.

13970

240 X 1.4

I3φ

KTC X tapfases

MT = ————— = ———— = 41.6 veces

6010

240 X 0.5

I1φ

RTC X taptierra

MT = ————— = ————— = 50 veces

11

Cálculo del número de elementos

(capacitores), que pueden fallar en un

solo grupo

Los Bancos de Capacitores en 85 kV,

conectados en estrella no aterrizada y en

el neutro y tierra se encuentra un

transformador de potencial de relación

400:1; el número de elementos que

pueden fallar en un solo grupo, se calcula

tomando en cuenta el 110% de

sobrevoltaje o voltaje de desbalanceo en el

neutro.

Se puede calcular con la expresión

siguiente:

Sabemos que:

donde:

%V = Incremento de sobretensión por

desplazamiento del neutro (110%).

P = Número de elementos (capacitores) en

paralelo, en un grupo serie.

S = Número de grupos de capacitores en

serie por fase.

F = Número de elementos (capacitores)

fallados en un grupo serie.

Para calcular el número de elementos

fallados (F), se aplica la expresión

siguiente:

Para este banco :

P = 13 capacitores / grupo

S = 4 grupos serie / fase

% V = 110

Por razones prácticas se consideran 2

elementos:

Considerando el tiempo de disparo de 2

seg, este tiempo se encuentra dentro del

rango permisible, para la sobretensión del

114.7%, como se indica en la tabla 3.

La relación de los elementos (capacitores)

fallados en un grupo serie, respecto al total

de elementos del mismo grupo, resulta ser:

Esta información sirve para corroborar la

sobretensión que se presenta a fallar 2

elementos %V = 114.7, como se indica en

la gráfica de la figura 25 (página 34) de la

norma IEEE C 37.99-1990.

Cálculo del voltaje de desplazamiento

del neutro.

Para determinar el voltaje de

desplazamiento del neutro, cuando fallan 2

elementos se utiliza la expresión siguiente:

donde:

VNG = Voltaje de desplazamiento del neutro

VLG = Voltaje aplicado de línea a neutro

300 PS

%V= —————

3S (P-F) + 2 F

3 PS 100

F = ————— 1 - ————

3S–2 % V

3 (13) (4) 100

F = ————— 1 - —— = 1.43 elem

3 (4) – 2 110

300 (13) 4

%V = ————————— = 114.7

3 (4) (13 –2) + 2 (2)

F 2

%E = —— X 100 = —— X 100 = 15.4

P 13

VLG (F)

VNG = —————— y

3 S (P-F) + 2 F

VLL

VLG = ———

3

12

VLL = Voltaje aplicado de línea a línea

RTP = Relación del Transformador de

Potencial (400:1)

El ajuste del tap del relevador de

sobrevoltaje (59) queda en 2 V, con un

retardo de tiempo de 2 s.

ENERGIZACIÓN Y PROCEDIMIENTO DE

PRUEBA

Cuando todos los requisitos para la

energización de un equipo nuevo, han sido

realizado y satisfacen los valores de

pruebas de aceptación que establece la

norma de LyFC, se procede a energizar

por primera vez el banco; con todas las

protecciones habilitadas para que en caso

de falla de algún dispositivo que forma

parte del arreglo del banco, operen las

protecciones 50/51 1, 2, 3 y N o 59.

Se inicia la maniobra asegurándose que

todo el equipo se encuentra en condición

de operación remota, para evitar la

cercanía del personal durante la

energización. Se asegura de modo visual

que las cuchillas de puesta a tierra no

estén cerradas, y se procede a cerrar las

cuchillas seccionadoras que conectan las

barras de 85 kV hacia el interruptor propio

del banco y posteriormente se cierra el

interruptor para energizar el banco.

En caso de que no haya problemas por

disparos de los relevadores 50/51-1, 2, 3 y

N que indican la falla en el aislamiento

(entre fases y/o de fase a tierra) y

posteriormente se observa el banco

tomando las mediciones de potencias,

corriente y voltaje. Si resultan correctos se

procede a desenergizar el banco,

permitiendo que se autodescarguen los

elementos a través de la resistencia de

descarga, por un periodo de 15 minutos y

después se cierran las cuchillas de puesta

a tierra que cortocircuitan todos los

elementos.

Para la prueba de desbalanceo se

procede a desconectar un elemento de un

grupo de una fase y se repite el

procedimiento para energización del

banco, tomando nuevamente todos los

valores de medición (potencias, corriente y

voltaje) y se mide el voltaje en el

secundario del transformador de potencial

y no debe operar el relevador 59, se

procede a desenergizar el banco de la

forma antes mencionada y se desconecta

otro elemento del mismo grupo,

suponiendo que ambos elementos han

fallado en la misma fase y grupo. Con este

procedimiento se energiza nuevamente el

banco, pero en esta ocasión debe operar el

relevador 59 mandando disparo al

interruptor de potencia del banco y alarma

de desbalanceo de voltaje.

Finalmente se normaliza el banco y se

deja en servicio para disposición de

Operación Sistema.

ASPECTOS A CONSIDERAR

Durante la energización de estos

bancos, es importante tomar en cuenta las

condiciones que guarda el sistema en

cuanto al flujo de potencia y la regulación

de voltaje.

En cuanto al flujo de potencia, es

importante asegurarse que la potencia que

49075 (2)

VNG = ———————— = 722 V

3 (4) (13-2) + 2 (2)

85000

VLG = ——— = 49075 V

3

VNG 722

VNG SEC = —— = —— = 1.8 V

RTP 400

13

fluye por las barras colectoras de la

subestación, sea lo mayor posible para

evitar que se presenten fenómenos

transitorios que no estabilicen

rápidamente; ocasionando que operen los

fusibles individuales de los capacitores. De

esta forma las fluctuaciones de voltaje

consecuencia de las variaciones de la

potencia reactiva del sistema serán menos

severas.

En cuanto al voltaje, debe ser registrado

antes, durante y posterior a la energización

y sobre todo tener previsto la operación de

los reguladores o cambiadores

automáticos con carga de los bancos de

transformación, para regular el voltaje de

las barras de 85 kV.

CONCLUSIONES

• La instalación de Bancos de

Capacitores en 85 kV no regulados, ha

representado para LyFC la forma más

viable tanto económica como

técnicamente, para atender los

requerimientos del sistema en cuanto a

compensación de reactivos, regulación

de voltaje y estabilidad del sistema.

• LyFC como no cuenta con líneas de

transmisión largas, ha solucionado el

control de reactivos mediante la

instalación de bancos de Capacitores,

en 23 kV y debido al crecimiento de

carga en el sistema se han instalado

Bancos de Capacitores en 85 kV.

• Se recomienda que el procedimiento

para la prueba y energización se

normalicen, tomando en cuenta las

condiciones del sistema durante dichas

pruebas.

• Se deben de realizar mediciones de

corrientes armónicas (3a, 5a, 7a, etc).

antes de energizar un banco de

capacitores. En caso de que se

presenten valores altos se deben

instalar inductancias en serie con el

banco de capacitores, cuya reactancia

inductiva represente una pequeña

fracción de la reactancia capacitiva, por

fase del banco para la frecuencia

fundamental, constituyendo un filtro

para las altas frecuencias. La

inductancia de diseño de estos

reactores depende del orden de las

armónicas que causen problemas en el

sistema.

• No se debe energizar un banco de

capacitores, cuando el sistema de

potencia no requiere de compensación

reactiva, es decir, cuando la carga es

baja, ya que podrían presentarse

fenómenos de resonancia al momento

de la energización.

• Se recomienda antes de la

energización del banco de capacitores

en 85 kV, se desconecten los bancos

de capacitores en 23 kV, a fin de

disminuir el fenómeno de sobretensión

transitoria, ya que la variación de

voltaje no debe ser más del 3% del

voltaje nominal.

• Es conveniente realizar la prueba del

banco cuando Operación Sistema

considere necesaria la compensación o

realice las maniobras necesarias para

hacer fluir la mayor cantidad de

potencia activa en esa Subestación.

BIBLIOGRAFÍA

• Norma IEEE C 37.99-1990.

• Norma IEEE C 37.99-2000.

• Norma ANSI/IEEE std 18-1989.

• Norma IEEE std 18-1992. Standard for

shunt power capacitors.

14

• Capacitores de potencia. Dr. Alfredo

Navarro Crespo Balmec.

• Manual del Banco de Capacitores en 85

kV de Alstom.

• Análisis y Aplicaciones de los Bancos

de Capacitores, sus Esquemas de

Protección por Relevadores y Criterio

de Cálculo de los Ajustes, LyFC

ponencia IEEE RVP 1994. Acapulco,

México.

• ANSI/IEEE std. 18-1989, Standard for

shunt Power Capacitors.

• John E. Harder, Capacitor bank

Protection, Asea Brown Bovery, July

1990.

AUTORES

Ing. Miguel Ángel Méndez Albores

Es Ingeniero

Electricista, egresado

de la ESIME del

Instituto Politécnico

Nacional en el año de

1986; ingresó a laborar

a Luz y Fuerza del

Centro en el año de

1982, donde ha venido

desempeñando diferentes puestos y

actualmente ocupa el puesto de

Superintendente de Estudios, de la

Gerencia de Aseguramiento de la Calidad.

Desde 1994 es profesor de la Academia

de Electrotecnia, del Departamento de

Ingeniería Eléctrica de la ESIME-IPN;

donde actualmente imparte las materias de

Electrotecnia I, II y Laboratorio; en el

Departamento de Ingeniería en Control y

Automatización las materias de Circuitos

Eléctricos I, II y Laboratorio.

En 1998 cursó el Diplomado en

Sistemas Eléctricos de Potencia, impartido

por General Electric.

Ing. José Marcelino Santiago Jiménez

Ingeniero Electricista

egresado de la UAM

Azcapotzalco. Ingresó a

laborar a Luz y Fuerza del

Centro en el año de 1998,

donde ha venido

desempeñando diferentes

puestos y actualmente ocupa el puesto de

Ingeniero Cl. 20 A en la Superintendencia

de Estudios de la Gerencia de

aseguramiento de la Calidad. De 1996 a

1999 realizó estudios de Maestría con

Especialidad en Sistemas Eléctricos de

Potencia en la Sección de Estudios de

Posgrado e Investigación de la ESIME-IPN

Zacatenco. Es Profesor de la carrera de

Profesional Técnico en Mantenimiento de

Equipo de Cómputo y Control Digital en

CONALEP.

1

ANÁLISIS DE REGISTROS OSCILOGRAFICOS REALES Y SIMULADOS DE LA FALLA DE UN

INTERRUPTOR DE 400 KV EN LA S.E. ACATLAN DE CFE-MÉXICO

Antulio Jarquín Hatadis y Javier De J. Angel León

Subgerencia Nacional de Protecciones

Subgerencia De Protección y Medición ATOCC

CFE-MEXICO

2002

I.-INTRODUCCION

Los registros oscilográficos de los

cortocircuitos y fallas de un sistema eléctrico

de potencia, constituyen una importante

herramienta de análisis para el Ingeniero de

Protecciones.

En este trabajo se pretende revisar desde la

óptica de la especialidad de Ingeniería de

protecciones, los oscilogramas de la falla de

un interruptor de 400 KV Figura 1a,

Figura. 1a.-Modelo de interruptor 400 KV.

en el momento en que al disparar para librar

una falla de fase a tierra en una línea de

transmisión, el interruptor se ve de tal forma

involucrado en el evento, que falla y provoca

un disturbio en el sistema de potencia.

Como un importante complemento de

análisis, en este trabajo se incluye un

estudio del caso por medios digitales de

simulación, del cual se llega a interesantes

conclusiones del evento.

II.-DESCRIPCION SIMPLIFICADA DEL

EVENTO

Al ocurrir una falla transitoria de la fase B a

tierra en la Línea de transmisión de 400 KV

(ATN) ACATLAN-A3230-

MANZANILLO(MNZ) Junio del 2001, para

librar la fase fallada se dispara la fase B

monopolarmente, operando adecuadamente

la protecciones de ambos extremos de la

línea, con esta acción se arranca el relé de

recierre monopolar, recerrando bajo falla el

polo de la fase B en la terminal de línea de

S.E. ATN en 960 ms., y con esto operan

nuevamente las protecciones P1( PLS) Y

P2(SEL 321) disparando nuevamente la fase

fallada, operando el disparo tripolar del

interruptor ATN-A3230 , con esta operación

explota el polo de la fase C de este mismo

interruptor operando la protección Diferencial

de Barrras (87B) para liberar la falla.

III.-CUESTIONAMIENTOS ACERCA DE LA

FALLA.

1.-¿ Porque Explotó la fase C del interruptor

si la falla en la línea ocurrió en la fase B?

2.-¿Qué ocurrió en el Interruptor?

3.-¿Ocurrió algo en el sistema que provocó la

falla?

Derivado de estas preguntas, se procederá a

efectuar un análisis de los oscilogramas

obtenidos para este evento, con la finalidad

de contestar a estos cuestionamientos.

IV.-ANÁLISIS DE LOS OSCILOGRAMAS

DEL EXTREMO EN LA S.E. ACATLAN

En el tiempo t1 ocurre una falla en la fase B

a 73 Km. De distancia de la S.E. Acatlán , la

falla es liberada por las protecciones

respectivas de la línea en 2.5 ciclos en el

tiempo t2 en forma monopolar.

Nota: Esta línea tiene esquema de disparo y

recierre monopolar con un recierre de tipo

secuencial con tiempos de recierre de:

MANZANILLO=1.2 Seg. y ACATLAN = 1.0

Seg.

Al abrir el polo B de la línea se observa en el

canal analógico de voltaje de la fase B que el

voltaje de recuperación aparece y después

vuelve a desaparecer, por lo que al terminar

el tiempo muerto se recierra bajo falla el polo

2

de la fase B en el tiempo t3, disparando en

foma tripolar el interruptor ATN-A3230 la falla

se libera en 5.0 ciclos, en el momento t4.

t1 t2 t3 t4

Oscilograma No. 1

t4

t5

t6

Oscilograma No. 2

Oscilograma No. 2

En el tiempo t4 (oscilograma 2) se aprecia

que el voltaje de la fase B no esta presente

debido a la propia falla, sin embargo los

voltajes de las fases A y B desaparecen, esto

es debido a que abre en forma tripolar el

extremo de la S.E. MANZANILLO.

En el tiempo t5 ocurren detalles importantes,

aparece el voltaje de la fase B sin existir

corriente en esta fase B, se presupone que

existió un recierre de esta fase en el extremo

de la S.E. MANZANILLO debido a que

coincide con el tiempo de ajuste de recierre

secuencial a 1.2 segundos.( en la gráfica son

1.1 segundos reales) coincidiendo con este

evento aparece también el voltaje en la fase

C así como la corriente en la misma fase C.

Aquí se piensa que el polo de la fase C del

interruptor ACATLAN-A3230 empieza a

conducir corriente por algún medio, con

FLASHOVERS adicionales; especialistas de

CFE argumentan que falló el mecanismo de

la resistencia de preinserción en una de las

secciones del interruptor. El valor que se

observa es la corriente de carga capacitiva

de la línea en vacío de aproximadamente 100

Amperes.

Oscilogramas 2,3 y 4.

En estos oscilogramas se presentan los

tiempos t6, t7, t8, t9 y t10. En estos tiempos

se observa básicamente el tiempo en que

permaneció el polo de la fase C conduciendo

la corriente capacitiva de la línea este lapso

fue de 6 seg. con 270 ms.

Posteriormente desaparece el voltaje de la

fase C y vuelve aparecer 10 seg. con 370

ms. De manera que el polo de la fase C

vuelve a conducir con el interruptor en

posición de abierto durante 3 seg. con 820

ms. Hasta la explosión del interruptor, ver los

tiempos t11 y t12 del Oscilograma No. 5

t7 t8

Oscilograma 3

Oscilograma No. 3

t9

t10

Oscilograma No. 4

Oscilograma No. 4

3

t11 t12

Oscilograma No. 5

Oscilograma No. 5

V.-OPERACIÓN DE PROTECCIONES

En cuanto a la operación de Protecciones,

estas operaron adecuadamente para esta

falla, ya que operó tanto la Protección

primaria 1 como la Protección Primaria 2, y

sus correspondientes canales de

comunicación de Comparación direccional.

VI.-ESTUDIO DE SIMULACIÓN DIGITAL DE

LA FALLA.

VI.1.-Antecedentes

Por motivo de que el fabricante del interruptor

argumentó que la falla del interruptor ocurrió

por haber sufrido el equipo el fenómeno de

rayo, esto causado por el sobrevoltaje

generado por el transitorio de la falla

ocurrida, fue necesario hacer estudios de

simulación del caso procediendo de la

siguiente forma:

VI.2.-Objetivo del estudio.

El objetivo del estudio fue determinar las

causas de la falla y de la explosión de una de

las cámaras del polo fase “C” del interruptor

ATN-A3230, para esto se obtendrán las

sobretensiones a las que se vió sometido el

aislamiento en cada una de sus cámaras y

con esto poder determinar si de alguna forma

se excedieron los parámetros de diseño del

interruptor, provocando así la falla del mismo.

Por otra parte también concluir que estas

sobretensiones no fueron lo suficientemente

grandes y no rebasaron los límites

admisibles de sobretensión, de ser así

entonces la falla del interruptor se debió a

otras causas.

Figura No.1b.-Diagrama del enlace fallado y

del modelo de estudio

VI.3.-Modelo para el estudio

En el estudio de simulación se uso el circuito

mostrado en la Figura 1b donde la línea de

interés es la MANZANILLO-A3230-

ACATLAN, la red se modeló con los

equivalentes en los buses más próximos al

elemento estudiado, es decir en las barras de

S.E. ATEQUIZA, S.E. MAZAMITLA y en la

S.E. MANZANILLO.

Se modelaron los interruptores de S.E.

MANZANILLO con resistencias de

preinserción de 400 Ohms, y para ACATLAN

se modelaron los interruptores con dos

cámaras por polo, cada una con con

resistencias de preinserción de 200 Ohms,

así mismo se agregaron los capacitores de

distribución de potencial para hacer más

realístico el estudio.

Para la reproducción del estudio se tomaron

como referencia los oscilogramas mostrados

descritos en el inciso III anterior, con esto se

elaboró una secuencia de los eventos

ocurridos de manera que al efectuar el

estudio se compararon los resultados de la

simulación con los registros oscilográficos,

con la finalidad de validar los resultados

obtenidos con el estudio, es decir obtener

4

gráficas de la simulación similares al registro

oscilográfíco del evento real de la falla.

VII.-EL ESTUDIO

Para la reproducción del evento se

cuenta con el registro de la falla en la

subestación Acatlán, con esta referencia se

hizo la secuencia de los eventos ocurridos

mismos que se muestran en la Tabla 1. Para

la obtención de la secuencia de eventos se

comparó el resultado de la simulación con el

registro obtenido, de tal forma que se

obtuviera la gráfica de la simulación similar a

la del registro.

Para efecto de verificación de la

secuencia de eventos ocurridos,

mostraremos algunas de las gráficas

obtenidas del registro de la falla y en la

simulación del evento. Como el registro de

los voltajes se obtiene a través de

dispositivos de potencial, fue necesario

agregar el modelo de éste para realizar la

comparación de los registros, y una vez

aceptado que la secuencia de eventos es la

correcta la medición de los voltajes a que se

somete el equipo se realiza directamente en

alta tensión.

Tabla No 1.- Secuencia de eventos ocurridos en el disparo y recierre monopolar

No TIEMPO

(ms) DESCRIPCION DEL EVENTO 0 0 Se inicia el registro

1 49 Aparece falla en fase ‘B’ de la línea

2 92 Abre polo ‘B’ de interruptor en Acatlán

3 100 Abre polo ‘B’ del interruptor en Manzanillo

4 175 Se extingue el arco secundario

5 429 Reenciende la falla (polo ‘b’ abierto en ambos extremos)

6 1058 Recierra con falla polo ‘B’ de resistencia de preinserción en Acatlán

7 1064 Recierra con falla polo ‘B’ de contacto principal en Acatlán

8 1141 Abre interruptor en forma tripolar en Acatlán

9 1164 Abren fases ‘A’ y ‘C’ en Manzanillo

10 1172 Se extingue el arco secundario de la fase ‘B’

11 1213 Recierra en Manzanillo el contacto de la fase ‘B’ correspondiente a la

resistencia de preinserción del interruptor (sin falla)

12 1221 Recierra en Manzanillo contacto principal de la fase ‘B’ del interruptor (sin falla)

13 1221 Arquea el contacto del interruptor de Acatlán, correspondiente a la resistencia

de preinserción de una de las cámaras en la fase ‘C’ y se mantiene el arco.

14 1221 Arquea externamente la otra cámara de la fase ‘C’ del interruptor de Acatlán y

se mantiene el arco

15 1273 Abre la fase ‘B’ del interruptor en Manzanillo

VII.1.-COMPARACION DE LAS

GRAFICAS AL INICIO DE LA FALLA

Las siguientes gráficas

corresponden a la primera parte del

registro, en ellas se comparan los voltajes y

corrientes medidos en la subestación

Acatlán por el registrador a través del

dispositivo de potencial contra las gráficas

obtenidas de la simulación. Se puede ver la

incidencia de la falla y el libramiento

correcto en forma monopolar, se puede

notar también en la gráfica de voltaje la

inestabilidad del arco durante el evento. En

las corrientes se aprecia el libramiento de

la falla en Acatlán en tres ciclos.

5

Figura 2.- Voltajes del registro en ATN

Figura 3.- Voltajes de la simulación

Figura 4.- Corrientes del registro de ATN

Figura 5.- Corrientes de la simulación

Figura 6.- Voltajes del registro en ATN

Figura 7.- Voltajes de la simulación

6

Figura 8.- Corrientes del registro de ATN

Figura 9.- Corrientes de la simulación

VII.2.-COMPARACIÓN DE GRAFICAS

DURANTE EL RECIERRE CON FALLA Y

EL DISPARO TRIPOLAR

De manera similar para la segunda

parte del registro tenemos las siguientes

gráficas, en éstas se aprecia el recierre con

falla en Acatlán y el disparo tripolar de los

interrruptores. En la gráfica de voltaje se

puede ver que después del disparo tripolar

en Acatlán, las fases ‘A’ y ‘C’ del

interruptor en Manzanillo abren

correctamente y posteriormente aparece el

recierre de la fase ‘B’.

En las gráficas de corriente se

observa el recierre con falla, con

libramiento de 4 ½ ciclos y al final aparece

la corriente del arqueo del polo ‘C’ del

interruptor; este arqueo coincide con el

transitorio producido por el recierre del

contacto principal del polo ‘B’ del interruptor

de Manzanillo.

Los resultados obtenidos son

aceptables y podemos considerar que la

simulación del evento corresponde a lo

ocurrido durante la falla, el recierre con falla

en Acatlán y el recierre del polo ‘B’ en

Manzanillo. La parte interesante del evento

para el estudio es la última, donde ocurre el

arqueo del interruptor; en esta parte del

evento podemos analizar lo que ocurre

específicamente en las terminales del

interruptor de la subestación Acatlán y los

voltajes a los que fue sometido.

Fig. 10.- Voltajes lado Bus en ATN

A continuación analizaremos los

voltajes de la fase ‘C’ de alta tensión. La

Figura 10 muestra el voltaje de lado Bus y

la Figura 11 el voltaje lado línea del

interruptor en alta tensión, se indica en las

gráficas el instante en que se inicia el arco

en la cámara, en este punto los voltajes de

la barra y la línea se encuentran con

polaridades contrarias.

Figura 11.- Voltajes lado LINEA en ATN

7

El voltaje en el interruptor de Acatlán

es el resultado de la suma del presentado

en las gráficas de las Figuras 10 y 11

anteriores, y que a continuación se

muestra, y es al que fue sometido el

interruptor en cada cámara de la fase ‘C’

durante el evento Ver Fig. 12

Figura 12.- Voltaje en las cámaras del polo

‘C’ del interruptor de ATN

El voltaje al que ocurre el arqueo del

interruptor según la reproducción del

evento son 360 kv, punto que se indica en

la misma gráfica de la Figura 12. Como

podemos ver el interruptor no fue sometido

a esfuerzos excesivos de voltaje, y éste

debió haber resistido este esfuerzo sin

ningún problema.

Ahora si en la simulación quitamos

el arqueo del interruptor para ver a que

magnitud se incrementa el voltaje al que se

exponen las cámaras del interruptor, el

oscilograma que se obtiene es el de la

Figura 13 .

Figura 13.- Voltaje en las cámaras del polo

‘C’ del interruptor de ATN sin la ocurrencia

del arco en la cámara.

Figura 14.- Voltaje en las cámaras de los

tres polos del interruptor de ATN sin la

ocurrencia del arco en la cámara.

Como podemos observar no existe

ningún transitorio severo durante el evento,

el voltaje máximo que se obtiene en

terminales de la cámaras del polo ‘C’ del

interruptor de Acatlán en el cierre del polo

‘B’ del interruptor de Manzanillo es 368 kv

Adicionalmente, mostramos los

voltajes en las cámaras de los polos del

interruptor en las tres fases, éstos se

muestran a continuación en la Figura 14

donde, se ve que la fase ‘A’ se expone

también a un voltaje de 368 kv muy similar

al de la fase ‘C’ y este polo no tuvo ningún

problema de arqueo durante el evento.

8

Figura 15.- Corriente del registro de ATN durante el arqueo de la resistencia de preinserción.

Se concluye por lo tanto que el

voltaje al que fue sometido el interruptor y

específicamente en la cámara de la

resistencia de preinserción no es de

consideración, sobre todo porque sabemos

que debe resistir cuando menos 395 kv

bajo estas condiciones.

Ahora bien en el registro se aprecia

que arquea la fase ‘C’ del interruptor y

decimos que esto ocurre internamente en

la cámara de la resistencia de preinserción

y en forma externa en la otra cámara,

porque para obtener en la simulación los

niveles de corriente al inicio del arqueo

(1039 A), se requiere que esto ocurra a

través de una resistencia de 200 ohms

(1054 A). Si se hace con 400 ohms lo cual

representaría un arco en ambas cámaras

de las resistencias, los niveles de corriente

son menores (811 A). Otra posibilidad es

que el arco ocurriera en el contacto

principal de la otra cámara, lo cual

consideramos improbable ya que la rigidez

dieléctrica dentro de ésta es mayor. Las

gráficas del inició del arqueo se muestran a

en las Figuras 15, 16 y 17..

Figura 16.- Corriente de la simulación

durante el arqueo de la cámara de una sola

resistencia de preinserción (200 ohms).

Figura 17.- Corriente de la simulación

durante el arqueo de las cámaras de las

dos resistencias de preinserción (400

ohms)

VIII.-CONCLUSIONES

• Las sobretensiones a que fue sometido

el interruptor de la subestación Acatlán

durante el recierre en Manzanillo no

son de consideración y el interruptor

debió soportar sin presentar ningún

problema.

• Algún defecto interno en el contacto de

la resistencia de preinserción de una

de las cámaras provocó que se

perdiera el aislamiento nominal y por lo

tanto arqueara internamente y la otra

cámara arqueó externamente, de tal

forma que circuló permanentemente

una corriente de aproximadamente 234

A, con la consecuente destrucción de

la cámara de la resistencia de

preinserción y ésta misma.

9

• La falla del interruptor es imputable a él

mismo y no a sobretensiones en la red

durante el recierre del polo ‘B’ del

interruptor de Manzanillo ocurrido

durante el evento.

• Que los oscilogramas reales son de

alto valor ya que sirven como base

para realizar los estudios adecuados, y

con ellos poder determinar las causas

que originaron las fallas en los equipos.

IX.-BIOGRAFÍA DE LOS AUTORES

Antulio Jarquín Hatadiz.- Efectuó sus

estudios de Ingeniería en El Instituto

Tecnológico Regional de Veracruz

graduándose en 1979, a trabajado en CFE-

MEXICO durante 21 años, entre otras

actividades destacan sus trabajos en

Compensadores Estáticos de VARS y

Bancos de Capacitores Serie en Líneas de

Extra Alta Tensión a Nivel Nacional,

Trabajo durante 7 años en la Unidad de

Ingeniería Especializada de CFE donde

realizo importantes estudios en el TNA

( Analizador Transitorio de Redes), Es un

experimentado ingeniero en la ejecución de

estudios de Transitorios Electromagnéticos

del Sistemas Eléctricos de Potencia. Desde

1992 esta al frente del Departamento De

Análisis de Redes de la Subgerencia de

Protección y Medición Nacional.

Javier de J. Angel León.- Es Ingeniero

Electricista, egresado de la Escuela de

Ingeniería Eléctrica de la Universidad

Michoacana de San Nicolás de Hidalgo

(1976) MEXICO, tiene amplia experiencia

en la protección de Sistemas Eléctricos de

Potencia en CFE (23 AÑOS), posee un

Diplomado en Fibras Ópticas en el ITESM y

LATINCASA, es instructor titular de CFE-

MEXICO en el Centro de Capacitación de

Occidente en Protección de Líneas de

Extra Alta Tensión y Protección de

Generadores, a participado como ponente

en el “MEXICON DEL IEEE” en 1992 con el

tema “Utilización de Equipo computarizado

para Prueba de Protecciones”, trabaja

actualmente como Jefe del Departamento

de Ingeniería de Protecciones en el Área

de Transmisión y Transformación.

Occidente de CFE.

Protección por Comparación Direccional de Líneas Cortas de 115 kV con Canal de

Comunicación por Radio

Martín R. Monjarás Méndez Alfredo Dionicio Barrón J. Ignacio Muñoz González División de Distribución Bajío, CFE

Guanajuato, GTO, México División de Distribución Bajío, CFE

Querétaro, QRO, México Schweitzer Engineering Laboratories, S.A. de C.V.

León, GTO, México

Héctor J.Altuve Ferrer

Schweitzer Engineering Laboratories, S.A. de C.V. Monterrey, N.L., México

Resumen: Este artículo analiza la aplicación de un esquema de protección para líneas cortas de 115 kV, con relevadores de protección direccional de sobrecorriente (67) y lógica de teleprotección POTT, que utiliza un canal de comunicación por radio. La necesidad de utilizar estos esquemas de protección se presenta en líneas cortas que forman parte de anillos de subtransmisión de 115 kV, en que no se dispone de fibra óptica u otro medio de comunicación adecuado para aplicar una protección diferencial de línea. En el artículo se presentan los criterios de aplicación y ajuste de los elementos y funciones de protección, y se hace un análisis de operación para fallas reales internas y externas a la línea protegida. Palabras clave: Relevadores de protección, protección de líneas, teleprotección, canal de comunicaciones, oscilogramas.

I. INTRODUCCIÓN La especificación de la Comisión Federal de

Electricidad de México (CFE) referente a esquemas aplicables para protección de líneas, indica que para líneas cortas (longitud menor que 10 km) en voltajes de 161 kV y menores, la protección primaria debe ser diferencial de línea (87L) y la protección de respaldo debe ser direccional de sobrecorriente de tierra (67N).

Cuando entran en servicio nuevas subestaciones en ciudades o áreas densamente pobladas con anillos de 115 kV, la longitud de algunas líneas de interconexión generalmente se hace menor que 10 km. En la mayoría de los casos, debido a la rapidez del crecimiento de la demanda, no es posible dotar con fibra óptica y 87L a las líneas que se acortan, lo que hace difícil la protección de la línea. Esta protección típicamente se brinda por medio de relevadores de distancia (21) como protección primaria, con un relevador direccional de sobrecorriente de tierra (67N) como protección de respaldo.

En esta situación, en vista de la reducida longitud de la línea, es necesario acortar los alcances de la primera zona de los relevadores de distancia o, en casos extremos, dejarla bloqueada. Ambas opciones conducen a la pérdida de la operación en alta velocidad (menos de tres ciclos) de las protecciones de la línea. Esto pone en riesgo la integridad de los componentes del sistema de subtransmisión, y compromete la coordinación de protecciones entre las líneas del anillo de subtransmisión y los transformadores de las subestaciones de transmisión (230/115 kV).

El esquema apropiado para la protección de líneas en esta situación debe tener las siguientes características:

Operación en alta velocidad (menos de tres ciclos) para fallas en cualquier punto de la línea. Canal de comunicaciones (no necesariamente por medio de fibra óptica). Lógica de teleprotección (preferentemente POTT con detectores de sobrecorriente de secuencia negativa para mejor cobertura de fallas con resistencia de arco). La solución alternativa a la aplicación de protección

diferencial de línea debe contemplar, en lo posible, utilizar los relevadores convencionales para protección de líneas (21/67N) que se tienen en operación, aumentando sus funciones por medio de programación, y añadir un mínimo de componentes externos.

En este trabajo se propone una solución al problema de protección de líneas cortas de 115 kV utilizando un esquema de comparación direccional con lógica POTT y un canal de comunicación por radio.

II. PLANTEAMIENTO DEL PROBLEMA

Un caso real de aplicación de esta solución es en el anillo de 115 kV de Querétaro. El diagrama unifilar parcial, antes de la conexión de la subestación Satélite entre las subestaciones Querétaro (QRO) y Querétaro Poniente (QPE), se muestra en la Fig. 1.

This frame provides a blank space 2 inches high in the lower left hand corner of the first page (for use by IEEE Headquarters).

QPE

QRP

1BCO-225 MVA230/ 115/ 13.8 KV

QRS

QRO

2 BCOS-100 MVA230/ 115/ 13.8 KV

K - 8 + 300

Fig. 1. Diagrama unifilar parcial antes de la conexión de la subestación Satélite.

Inicialmente, la longitud de la línea entre las

subestaciones QRO y QPE era de 8.3 km, por lo que, según lo especificado por CFE, sería necesario tener una protección diferencial; en la práctica se ha observado que para esta longitud de línea se tiene una coordinación confiable con relevadores de distancia y de sobrecorriente direccional, aún cuando ocurran fallas con resistencia de arco. Las protecciones que estaban habilitadas en ambos extremos de la línea QRO-QPE, eran las siguientes:

Protección primaria: Relevadores de distancia (21). Protección de respaldo: Relevadores direccionales de sobrecorriente (67). Debido a que el crecimiento de la demanda fue mayor

que lo esperado, se decidió construir la subestación Satélite (SAT), que se conectó entre las subestaciones QRO y QPE, con un banco de transformación de 30 MVA, 115/13.8 kV. En la Fig. 2 se muestra la ubicación de la subestación SAT dentro del enlace de QPE-QRO.

K - 3 + 300 K - 5 + 000QPE

2 BCOS-100 MVA230/115/13.8 KV

1BCO-225 MVA230/115/13.8 KV

QRP

QRS SAT

QRO

Fig. 2. Diagrama unifilar parcial mostrando la inserción de la subestación Satélite (SAT).

Al entrar en servicio la subestación Satélite, la línea entre las subestaciones Querétaro Poniente (QPE) y Satélite (SAT) quedó con una longitud de 3.3 km, dificultando el ajuste y coordinación de las protecciones existentes, principalmente para fallas a tierra con alta resistencia de arco.

III. SOLUCIÓN PROPUESTA

La solución contemplada inicialmente fue la instalación de fibra óptica y relevadores de protección diferencial, pero esto implicaba un costo elevado y el reemplazo de los esquemas existentes de protección de línea (21 y 67). Se observó que los relevadores 67 instalados en esta línea tienen una función programable que permite transmitir y recibir estados de elementos internos del relevador para aplicar funciones de protección y control a través de un enlace de comunicación por radio, fibra óptica u otros medios.

La solución propuesta contempló: Utilizar los relevadores de protección de línea existentes y activar varios elementos direccionales de sobrecorriente de fase, de secuencia negativa y de secuencia cero en los relevadores 67 como detectores para las funciones de disparo permisivo. Habilitar la lógica de teleprotección POTT integrada en los relevadores 67. Agregar radios para proporcionar el enlace de comunicación. Ajustar los puertos de los relevadores al protocolo de comunicación de elementos internos para poder utilizar los radios conectados directamente al puerto RS-232 de cada relevador.

Al aplicar estos cambios, la protección primaria de la línea es brindada por los relevadores direccionales de sobrecorriente en esquema POTT con radios, y la protección de respaldo queda a cargo de los relevadores de distancia.

En lo relativo a componentes, para implementar esta solución solamente fue necesario añadir dos radios, cable coaxial y antenas. Los criterios de ajuste de los elementos direccionales de sobrecorriente para utilizar la lógica de teleprotección POTT disponible en los relevadores 67 son los siguientes:

Nivel 1: Bloqueado, por ser una línea muy corta. Nivel 2: Detección de fallas hacia adelante entre fases y de fase a tierra (sensibilidad para sobrealcanzar el bus remoto). Función de disparo permisivo y generación de un disparo local si no es recibida o no está activada la señal de bloqueo. Nivel 3: Detección de fallas hacia atrás entre fases y de fase a tierra (sensibilidad mayor que la del Nivel 2 del otro extremo). Transmite la señal de bloqueo al otro extremo de la línea.

Nivel 4: Detección de fallas hacia adelante entre fases y de fase a tierra (igual alcance que el Nivel 2). Función de disparo temporizado a 18 ciclos, usada como respaldo remoto y/o cuando el canal está fuera de servicio.

Una de las ventajas de esta protección es que no se bloquea por pérdida del canal; solo queda fuera el disparo por comparación direccional, quedando habilitados el Nivel 4 temporizado a 18 ciclos y las funciones de sobrecorriente de tiempo inverso; se conserva la direccionalidad de las funciones de protección. En la Fig. 3 se muestran de manera gráfica los alcances de los elementos direccionales de sobrecorriente. Para utilizar la lógica de teleprotección [1,2] se utilizaron los elementos internos 1 y 2 (MB1A y MB2A) de los relevadores 67, de la forma siguiente:

MB1A: Envío (TX) / recepción (RX) de disparo permisivo. o TMB1A: Envío de disparo permisivo. o RMB1A: Recepción de disparo permisivo. MB2A: Envío (TX) / recepción (RX) de bloqueo de disparo. o TMB2A: Envío de bloqueo de disparo. o RMB2A: Recepción de bloqueo de disparo.

Los elementos internos MB4A no intervienen en la lógica de teleprotección; solamente se utilizan para fines de información en los registros:

MB4A: Envío (TX) / recepción (RX) de estados de interruptores local y remoto. o TMB4A: Envío de estado de interruptor local. o RMB4A: Recepción de estado de interruptor remoto.

Otros elementos que intervienen en los registros de falla:

ROKA: Elemento que indica que el enlace de comunicación está trabajando en forma correcta. TRIP: Salida de disparo local. 67G2: Elemento direccional de sobrecorriente de secuencia cero para fallas hacia adelante (activa la transmisión del MB1A).

QPE Nivel 3

TX

RX

TX

RX

SAT 115 KVQPE 115 KV

RS 232 351351

RS 232

QPE Nivel 4

SAT Nivel 2

SAT Nivel 4

QPE Nivel 2

SAT Nivel 3

Fig. 3. Alcances de los elementos direccionales de sobrecorriente.

67G3: Elemento direccional de sobrecorriente de secuencia cero para fallas hacia atrás (activa la transmisión del MB2A). 67Q2: Elemento direccional de sobrecorriente de secuencia negativa para fallas hacia adelante (activa la transmisión del MB1A). 67Q3: Elemento direccional de sobrecorriente de secuencia negativa para fallas hacia atrás (activa la transmisión del MB2A).

IV. PRUEBAS AL SISTEMA DE PROTECCIÓN

A. Medición del Tiempo del Canal La medición del tiempo del canal se realizó activando la transmisión del elemento interno MB1A en la subestación Satélite (SAT) y haciendo un puente entre la recepción y la transmisión del mismo elemento en la subestación Querétaro Poniente (QPE). El tiempo medido entre la transmisión y la recepción del elemento MB1A en la subestación Satélite es igual a dos veces el tiempo del canal en un solo sentido de comunicación. Los resultados obtenidos son los siguientes: =>>SER 4 SAT-73110 -QPE Date: 09/29/00 Time: 01:46:38.630 351-MB-RADIO-POTT-GPO1 FID=SEL-351-1-R205-V0-Z001001-D19990827 # DATE TIME ELEMENT STATE 4 09/29/00 01:46:29.475 TMB1A Asserted 3 09/29/00 01:46:29.479 TMB1A Deasserted 2 09/29/00 01:46:29.504 RMB1A Asserted 1 09/29/00 01:46:29.512 RMB1A Deasserted Con los registros de transmisión y recepción del elemento MB1A del esquema de la subestación Satélite, se determina el tiempo total del lazo, que fue de 29 ms. Este tiempo se divide entre dos, para obtener el tiempo de canal de comunicaciones en una sola dirección (igual a 14.5 ms). B. Operación para una Falla Interna Para evaluar el comportamiento de los relevadores en esta aplicación, a continuación se analizan oscilogramas de dos fallas reales, una interna y otra externa. Se muestran los registros de los relevadores en ambos extremos de la línea.

La falla interna ocurrió el 31 de octubre de 2001 en la línea SAT-73110-QPE (apartarrayo dañado en la fase B, a la salida de la subestación QPE), con una aportación de 3,950 A por el lado de QPE y de 5,517 A por el lado de SAT (ver Fig. 4). En la Fig. 5 se muestra el oscilograma obtenido por el relevador de la Subestación QPE. Su análisis permite concluir lo siguiente:

El inicio de la falla es en el ciclo 4.0 del registro. A los 4.0 ciclos se activan los elementos 67G2 (falla delante, secuencia cero) y TMB1A (transmisión de disparo permisivo). A los 4.5 ciclos se activa el elemento 67Q2 (falla delante, secuencia negativa). A los 5.5 ciclos se activa el elemento de recepción de disparo permisivo RMB1A y se completa la condición de disparo local TRIP, el cual se activa también a los 5.5 ciclos del registro.

21

SEL351

21

SEL351

QROSATQPE

1

3,950 Amp 5,517 Amp

Fig. 4. Aportaciones para falla en la línea SAT-73110-QPE.

Fig. 5. Oscilograma del registro del relevador ubicado en la

subestación QPE (falla interna).

A los 9.25 ciclos el interruptor local señaliza �“abierto�”. A los 11.0 ciclos el estado del interruptor remoto señaliza �“abierto�”. Desde el inicio de la falla (4.0 ciclos) hasta la salida de disparo local (5.5 ciclos) transcurren 1.5 ciclos.

En la Fig. 6 se muestra el oscilograma del relevador ubicado en la subestación SAT para esta falla interna. Puede concluirse lo siguiente:

El inicio de la falla es en el ciclo 4.0 del registro. A los 4.25 ciclos se activa el elemento 67Q2 (falla delante, secuencia negativa). A los 4.5 ciclos se activan los elementos 67G2 (falla delante, secuencia cero) y TMB1A (transmisión de disparo permisivo). A los 5.0 ciclos se activa el elemento de recepción de disparo permisivo RMB1A y se completa la condición de disparo local TRIP, el cual se activa también a los 5.0 ciclos del registro. A los 9.75 ciclos el interruptor local señaliza abierto. A los 10.25 ciclos el estado del interruptor remoto señaliza abierto. Desde el inicio de la falla (4.0 ciclos) hasta la salida de disparo local (5.0 ciclos) transcurre 1.0 ciclo.

En resumen, el esquema de protección operó correctamente para esta falla monofásica a tierra interna. El tiempo total de operación del esquema de protección de la línea para esta falla es de 1.5 ciclos (25 ms).


subestación SAT (falla interna). C. Operación para una Falla Externa La falla externa ocurrió el 2 de febrero de 2002 en la línea SAT-73540-QRO (daño en estructura por choque de vehículo), con una corriente de aportación de 2,419 A desde las subestaciones QPE y SAT (ver Fig. 7).

21

SEL351

21

SEL351

QROSATQPE

2

2,419 Amp 2,419 Amp 10,313 Amp

Fig. 7. Aportaciones para falla en la línea SAT-73540-QRO.

La Fig. 8 muestra el oscilograma registrado por el relevador de la Subestación QPE. Puede concluirse que:

El inicio de la falla es en el ciclo 4.0 del registro. A los 4.25 ciclos se activan los elementos 67G2 (falla delante, secuencia cero) y TMB1A (transmisión de disparo permisivo). A los 4.75 ciclos se activa el elemento 67Q2 (falla delante, secuencia negativa). A los 5.0 ciclos se activa el elemento de recepción de bloqueo de disparo permisivo RMB2A y debido a esto no se completa la condición de disparo local TRIP, el cual no se activa durante el registro. El estado de ambos interruptores de línea indica �“cerrado�” durante todo el registro.

En la Fig. 9 se presenta el oscilograma del relevador de la subestación SAT. Puede concluirse que:

El inicio de falla es en el ciclo 4.0 del registro.


subestación QPE (falla externa).

A los 4.0 ciclos se activan los elementos 67G3 (falla atrás, secuencia cero), 67Q3 (falla atrás, secuencia negativa) y TMB2A (transmisión de bloqueo de disparo permisivo). A los 5.25 ciclos se activa el elemento de recepción de disparo permisivo RMB1A, pero, dado que están activados los elementos direccionales de sobrecorriente locales orientados hacia atrás (67G3 y 67Q3), no se completa la condición de disparo local TRIP, el cual no se activa durante el registro. El estado de ambos interruptores de línea indica �“cerrado�” durante todo el registro. A los 9.75 ciclos se desactivan los elementos 67G3 (falla atrás, secuencia cero), 67Q3 (falla atrás, secuencia negativa); esto ocurre cuando desaparece la corriente de falla.

Nuevamente el esquema de protección funcionó correctamente, al no operar para esta falla externa.

V. COMPARACIÓN DE ESQUEMAS DE PROTECCIÓN DE LÍNEAS

Como se indicó en la Sección III, para la protección de líneas cortas de 115 kV existen dos soluciones que garantizan alta velocidad de operación: protección diferencial de línea (87L) y protección piloto por comparación direccional. La protección diferencial de línea requiere un canal digital, preferentemente de fibra óptica. La protección por comparación direccional puede utilizarse con distintos tipos de canales de comunicación. La solución utilizada en este trabajo para la protección de líneas cortas de 115 kV es un esquema de comparación como canal de comunicación.

Fig. 9. Oscilograma del registro del relevador ubicado en la subestación

SAT (falla externa).

La Tabla I presenta una comparación entre los canales de fibra óptica y de radio. La principal ventaja del enlace por radio es su economía, que permite aplicar esta solución a líneas de niveles bajos de voltaje. Es necesario garantizar la existencia de una línea de vista entre los extremos de la línea para la aplicación del enlace por radio. En la Tabla II se presenta una comparación entre los esquemas de protección diferencial de línea con fibra óptica y de comparación direccional (POTT) con enlace por radio. Esta comparación ha sido restringida a aquellos aspectos que se consideran de interés para esta aplicación a líneas de 115 kV. Puede observarse en la Tabla II que la principal ventaja de la protección diferencial de línea para esta aplicación es una mayor cobertura de resistencia de falla. La cuantificación de esta cobertura está fuera del alcance de este artículo. El esquema de comparación direccional con enlace por radio representa una solución económica y confiable, que garantiza una alta velocidad de operación y un respaldo inherente para pérdidas del canal.

En la actualidad existe un relevador digital multifuncional que incluye funciones de protección diferencial de línea, protección piloto por comparación direccional y elementos de distancia y direccionales para funciones de respaldo. Un relevador de este tipo permite aplicar una solución escalable a la protección de líneas de 115 kV; los pasos pueden ser:

Comenzar con un esquema de protección de distancia/direccional (21/67N) sin canal de comunicación. Añadir un enlace por radio cuando la línea lo requiera, y conformar un esquema piloto por comparación direccional, que incluye funciones de respaldo. Añadir un canal de fibra óptica e implementar la función de protección diferencial de línea. El enlace por radio puede quedar como un segundo canal de comunicación en el esquema, o puede ser trasladado a otra línea.

TABLA I

COMPARACIÓN DE CANALES DE COMUNICACIÓN

CONCEPTO FIBRA ÓPTICA

RADIO

COSTO $ 350,000 USD

(FIBRA ÓPTICA PARA 3.3 KM)

$ 5,000 USD

RESTRICCIONES SOLAMENTE COSTO

SE REQUIERE LÍNEA DE VISTA ENTRE EXTREMOS DE LA LÍNEA

ACCESORIOS DISTRIBUIDOR ÓPTICO, MULTIPLEXOR

ANTENAS, CABLE COAXIAL, FUENTES DE ALIMENTACIÓN PARA RADIOS

TABLA II COMPARACIÓN DE ESQUEMAS DE PROTECCIÓN DE LÍNEAS

CONCEPTO 87L CON FIBRA

ÓPTICA POTT CON RADIO

TIEMPO TOTAL DE OPERACIÓN DEL ESQUEMA

25-29 mseg. 25-32 mseg.

ZONA DE PROTECCIÓN

PROTEGE SOLO LA LÍNEA

PROTEGE LA LÍNEA Y RESPALDA LÍNEAS ADYACENTES

COMPORTAMIENTO ANTE PÉRDIDA DEL CANAL DE COMUNICACIONES

SE BLOQUEA LA PROTECCIÓN DIFERENCIAL (87L) Y OPCIONALMENTE SE HABILITA LA PROTECCIÓN DE SOBRECORRIENTE

SE BLOQUEA LA FUNCIÓN DE TELEPROTECCIÓN, PERO QUEDA HABILITADA LA PROTECCIÓN DIRECCIONAL DE SOBRECORRIENTE TEMPORIZADA

REQUIERE CAMBIOS MAYORES EN TABLERO

GENERALMENTE SÍ NO

COBERTURA DE RESISTENCIA DE FALLA

MAYOR MENOR

VI. CONCLUSIONES

El esquema de protección a base de elementos direccionales de sobrecorriente utilizando un canal de comunicación por radio es una opción viable para aplicación en líneas cortas de 115 kV entre subestaciones con línea de vista, donde no se disponga de fibra óptica y/o esquemas de protección diferencial de línea. El tiempo total de operación de este esquema es solo ligeramente mayor que el que se obtiene con esquemas de protección diferencial de línea. Esta solución es fácil de implementar, ya que es posible utilizar relevadores de protección de distancia y direccionales de sobrecorriente existentes en la línea. Es necesario añadir solamente los radios y sus accesorios. Es recomendable utilizar elementos direccionales de sobrecorriente de secuencia negativa en el esquema de disparo permisivo para detección de fallas desbalanceadas con resistencia de arco. Un relevador multifuncional de protección de líneas que incluye protección diferencial, de comparación direccional y elementos de respaldo, permite aplicar una solución escalable. Esta protección puede comenzar a funcionar sin canal de comunicación, y ser completada posteriormente con canales de radio y, finalmente, de fibra óptica.

VII. REFERENCIAS [1] K. Behrendt, �“Relay-To-Relay Digital Logic Communication For

Line Protection, Monitoring and Control�”, Schweitzer Engineer-ing Laboratories, Inc., 1996.

[2] K. Behrendt, �“Applying the New SEL-321-1 Relay-To-Relay Logic Communication To a Permissive Overreaching Transfer Trip (POTT) Scheme�”, Schweitzer Engineering Laboratories, Inc. AG-9613, 1996.

VIII. BIOGRAFÍAS

Martín R. Monjarás Méndez se graduó de Ingeniero Electricista en la Facultad de Ingeniería Mecánica, Eléctrica y Electrónica de la Universidad de Guanajuato en 1990. Ingresó a la División de Distribución Bajío de la CFE en el año 1991, donde ha desempeñado los siguientes cargos: Superintendente de Área de Distribución de 1991 a 1994 en la Zona San Juan del Río, QRO; Jefe de Oficina de Subestaciones, Líneas y Protecciones de

1994 a 1997 en la Zona San Juan del Río, QRO; desde 1997 ocupa el puesto de Jefe Ofna. Protecciones Divisional en Guanajuato, GTO.

Alfredo Dioncio Barrón se graduó de Ingeniero Industrial Eléctrico en el Tecnológico Regional de Querétaro en 1991; recibió un Diplomado en Sistemas de Potencia en la Facultad de Ingeniería Mecánica, Eléctrica y Electrónica de la Universidad de Guanajuato en 1994. Ingresó a la División de Distribución Bajío de la CFE en el año 1991, donde

ha desempeñado los siguientes cargos: Supervisor de Electrificación Rural de 1991 a 1992 en Guanajuato, GTO; Superintendente de Área de Distribución de 1992 a 1994 en la Zona Celaya, Regional San Miguel de Allende, GTO; desde 1994 ocupa el puesto de Jefe de Oficina de Protecciones en la Zona Querétaro, QRO.

Juan Ignacio Muñoz González se graduó de Ingeniero Electricista en la Facultad de Ingeniería Mecánica, Eléctrica y Electrónica de la Universidad de Guanajuato en 1978. De 1981 a 1997 trabajó en CFE en Protección de Sistemas de Distribución y Transmisión. De 1997 a 2000 trabajó en INELAP-PQE como Ingeniero de Soporte Técnico. Desde el 2000 trabaja en Schweitzer Engineering

Laboratories, S.A. de C.V., donde actualmente ocupa el puesto de Director de Servicios Técnicos. De 1988 a 1997 trabajó también como profesor de asignatura en la Universidad Iberoamericana, Plantel León, Departamento de Ciencias de Ingeniería. Es miembro del IEEE. Héctor Jorge Altuve Ferrer se graduó de Ingeniero Electricista en la Universidad Central de Las Villas, Cuba, en 1969. Recibió el grado de Doctor en Ingeniería Eléctrica en el Instituto Politécnico de Kiev, URSS, en 1981. De 1969 a 1993 trabajó como Profesor Titular en la Facultad de Ingeniería Eléctrica de la Universidad Central de Las Villas. De 1993 a 2001 fue Profesor Titular del Programa Doctoral de la Facultad de Ingeniería Mecánica y Eléctrica de la Universidad Autónoma de Nuevo León, México. Fue Profesor Visitante de Washington State University en el curso 1999-2000. Desde 2001 trabaja en Schweitzer Enginering Laboratories, S.A. de C.V., donde actualmente ocupa el puesto de Director General. Su área de investigación es la protección, control y supervisión de sistemas eléctricos de potencia. Es Senior Member del IEEE.

capabilities of modern numerical differential … · design numerical differential protections with...

Documents