Partial Discharge Measurement in Transmission-Class Cable Terminations

Nezar Ahmed, Oscar Morel and Nagu Srinivas Detroit Edison, 2000 Second Avenue

Detroit, MI 48226

Abstract: This paper describes on-line and off-line partial discharge (PO) measuring techniques applicable to transmission cable terminations. The on-line technique can be utilized for both extruded and pipe-type cable terminations. In pipe-type cable terminations, the on-line technique is only applicable to newly installed terminations, as it requires an internal inductive sensor. In the case of extruded cable terminations, on-line testing is made possible by using an external inductive coupler around the ground connection at the base of termination.

Off-line PD measurements utilize high-frequency capacitive couplers. The coupler is attached to the overhead line within 4-feet from the termination under test. The off-line testing is applicable to new and existing cable terminations.

I: Introduction

Transmission-class cable terminations are classified in two groups depending on the type of cables that terminated to namely, extruded-dielectric cable terminations and pipe-type cable terminations. Terminations are required where cables are connected to overhead lines or electrical apparatus. Terminations are designed to possess the same integrity as their associated cables. Cable terminations are designed to eliminate stress concentration resulting from the ending of cable insulation and shield. The utilization of a stress cone effectively separates the shield away from the insulation surface and distributes the stress. All cable terminations, independently of operation voltage utilize a stress control cone. However, for sub-transmission and transmission type terminations, 60 kV and up, in addition to a stress cone, a secondary stress relief control unit is necessary to distribute dielectric stress over the length of the porcelain insulator. A porcelain stress control unit is utilized in conventional, non- graded type terminations for 69 to 161 kV. This cannot be utilized in higher voltage due to excessive internal diameter requirements. In these cases capacitance graded termination are used. These can be coaxial type and doughnut type. The former consists -of a series of cylindrical electrodes coaxial to the cable formed by intercalating aluminum foils between the paper layers, while the latter consist of a stack of toroidal capacitors connected in parallel.

0-7803-5515-6/99/$10.00 0 1999 IEEE 2

Terminations are the second weakest component in underground transmission cables, cable splices are the weakest cable component. Cable terminations normally failed as the result of aging or improper installations. Failure of terminations, results in chattering of the porcelain insulator with pieces scattered over a wide area in addition to fluid leaks. All high voltage terminations are filled with dielectric fluids with the exception of SF6 indoor terminations and terminations in HPGF cables.

Condition assessment of terminations is a difficult task. Since all transmission type terminations contain a dielectric fluid, sampling of this fluid for dissolved gas analysis presents an effective approach to termination diagnosis. However, only HPFF cable terminations allow direct access to the dielectric fluid. Extruded cable terminations are not designed for fluid sampling and in most cases the terminations must be opened to extract a small volume of fluid. A much less intrusive technique for termination diagnosis is made possible by PD measurements. This paper describes both on-line and off-line PD monitoring in transmission-class cable terminations.

11: Partial Discharge Detection

A: Detection Mechanism

PD in transmission-class terminations often occurs in the stress-relief assembly of the termination. Such PD activity creates high frequency pulses that travel along side the cable conductor and ground connection of the terminations. Typically, at the source, these electromagnetic pulses have bandwidths in the 1.5-ns range with rise-times in the nanosecond range. Depending on the type of defects, these pulses can be either a single pulse or a cascade of fast pulses. A coupling device with bandwidth of a few hundreds megahertz and a fast display are necessary to display these extremely fast pulses.

Either an inductive or a capacitive coupler is needed to detect PD pulses. Inductive couplers are high-frequency current- transformers that are placed around the grounding leads of the termination. These couplers detect the magnetic field disturbance induced by the PD pulses traveling in the ground loop. Inductive couplers are normally utilized during on-line testing. On the other hand, capacitive coupling is accomplished using high-frequency capacitors connected in parallel with the termination. These capacitors filter the 60 Hz component from the very-high frequency pulses associated with PD. As these capacitive couplers are connected directly to the high-voltage side of the termination,

they must be PD free at the test voltage. In theory, on-line testing using capacitive coupler can be made possible by permanently installing the sensor next to the termination. However, permanent installation of capacitive couplers is not economically feasible due to the cost of the couplers and the well knows reliability of transmission cable terminations. Therefore, capacitate coupling techniques become advantageous in off-line testing whereby with a few couplers hundreds of terminations could be scanned every year. Off- line couplers do not need to be weather resistant since such testing would be normally conducted in dry conditions.

B: PD Coupling Mechanism

In on-line PD testing, the type of termination will dictate the type of coupling sensors. Pipe-type terminations are used to terminate high-pressure oil-filled (HPOF) or high-pressure gas-filled (HPGF) cables. Both cables and termination components are installed in enclosures operating with fluid or gas pressure between 200-300 psi. The steel cable pipe is welded to the termination base plate using a non-magnetic stainless steel, copper or aluminum pipe stub assembly. The semi-stop assembly allows retaining a higher fluid pressure inside the termination in case of leak in the cable, also prevents cable movement. The connection of the cable shield to ground is made inside the termination. In most pipe-type terminations, a braided wire is used to connect the tinned copper shielding braid applied over the stress cone to the termination plate. The ground braid provides a ground path to any PD pulses taking place inside the termination. Technically, the ground braid is the only component that can be utilized to monitor PD. However, since the ground connection is made inside the fluid termination, the inductive sensor for on-line PD monitoring must be installed inside the termination. This could only be possible in new or refurbished terminations.

equipment

Figure 1 Typical inductive PD sensor installation in a pipe-type cable termination.

The inductive sensor for on-line PD measurement consists of a doughnut-shaped high-frequency current-transformer, about

1-inch in diameter. This sensor has a ferrite core with eight- coupling turns and is sensitive to frequencies ranging from 10 kHz to 200 MHz. The PD sensor is installed around the ground braid as shown in Figure 1. Electrical connection to the sensor can be made with an appropriate feed-through, installed on the base of the termination. On-line PD measurement is much easier in extruded-cable terminations since the cable below the termination plate is readily accessible. PD inside the termination generates high- frequency electromagnetic pulses that travel along the cable conductor, stress-cone shield and cable shield in opposite directions. In the vicinity of the PD source, these pulses are carried by a small portion of the shield [l]. However, PD pulses will distribute uniformly along the circumference of the shield at some distance away from the source depending on the wavelength of the discharge pulses. PD pulses in the VHF range can travel a few hundred meters before they are distributed around the entire circumference of the termination or cable shield. This distance is reduced to only few meters in the UHV range. Therefore, the PD current distribution alongside the circumference of the cable underneath the termination plate is not uniform for PD components of up to few hundred MHz. The unbalance current distribution in the shield will result in a stray magnetic field outside the cable. Consequently, inductive couplers can detect this magnetic field.

On-line PD testing in extruded cable terminations is achieved through magnetic coupling to the cable. PD pulses are collected via a clamp-type high-frequency current transformer (HFCT) attached to the cable below the termination plate. The HFCT used in this technique are made in two equal halves. One end of each halve is hinged and the other end butts together when clamped. The mating faces at both ends are machined flush to reduce air gaps. This ensures that the permeability of the toroid, and not the air gap, are the limiting factors in concentrating the magnetic field and reducing the overall path reluctance. The HFCT consists of a ferrite core with eight-coupling turns and is sensitive to a frequencies ranging from 1OkHz to 200MHz. Probes with an opening of 5 to 7 inches are used when the coupling is made around the cable. It should be noted that in some extruded cable terminations, the connection to ground is made outside the termination. In this case, a smaller diameter probe, about one-inch in diameter, can be connected to the external ground braid, which connects the cable sheath to the termination plate. Figure 2 shows the PD coupling arrangement employed in testing extruded cable terminations.

Off-line PD testing is performed through capacitive coupling, refer to Figure 3. The same arrangement is normally used for both pipe-type cable and extruded cable terminations. During testing, a 100 pF coupling capacitor is installed in parallel to the termination under test. The paper insulated, oil filled coupling capacitor rated at 500 kV is a PD free to up to 300 kV. For transport, this capacitor can be of four segments that can be bolted together. The housing of the capacitor is made from fiberglass with all fittings and corona shields are made

3

from aluminum. The weight of each individual segment is less than 40-pounds. The longitudinal capacitive gradient is designed to reduce the electrical stresses near the high voltage segment of the capacitor.

measuring equipment is sufficiently sensitivity and it can discriminate the interference from the overhead lines and busbars. Three coupling capacitors are required, one per each phase, if all three terminations needed to be tested.

Cable Insulation

Conductive Elas

+ + To PD Measuring Equipment

Figure 2 PD sensor arrangement in extruded-cable termination.

During PD testing, the termination must be de-energized prior to installation of the coupling capacitor. PD testing can be conducted either by using the line voltage as the excitation voltage, or by stressing the termination with an external power supply. Using external power supply allows the termination to be tested at voltages higher than the rated voltage. This will identify smaller defects not clearly exhibited at line voltages. Corona interference from the overhead lines is greatly reduced as the termination is disconnected from the overhead portion of the line. Furthermore, only one coupling capacitor is required to test all three terminations of the cable. However, external power supplies capable of delivering the voltages required to test transmission-class systems are heavy, expensive and hard to design them PD free.

Overhead Line

To PD Meascring Equipment

Figure 3 PD coupling arrangement utilized to conduct off-line testing.

Off-line testing at line voltage can be achieved provided the

C: PD Monitoring Equipment

PD detection under time and frequency domains was applied to monitor transmission cable terminations. The combination of both domains is used to distinguish between PD generated in the termination and external interference resulting from background RF noise. Figure 4 shows schematic diagram of the measuring system.

Frequency domain testing is conducted using a spectrum analyzer, which is capable of conducting measurements in both full and zero-span modes. In full-span mode, the frequency range can be adjusted to examine signals in narrow and wide-frequency bands. The zero-span mode is used to examine single-frequency pulses in a time domain. The sweep time of the zero span is used to search for PD pulses occurring at one or more cycles of the operating voltage.

Time domain measurements are carried out using a pulse- phase analyzer. The pulse-phase analyzer is capable of recording PD pulses sorted by magnitude and phase angle.

A differential noise coupling technique was applied to achieve better noise rejection during field measurements. For this purpose, a pick-up RF antenna is placed near the base of the termination. The signal picked up by the antenna is fed to the inverting input of a differential amplifier while the signal from the PD coupler is connected to non-inverting input, refer to Figure 4.

ToPD ToPickup Sensors Antenna

Differential Amplifier

F l F l figure 4 Schematic Diagram of the PD measurement system.

D: Noise Treatment

The biggest challenge of PD measurement in cable terminations is the electrical interference pickup from the

4

overhead system. The main sources of electrical interference in cable terminations are:

Radio and TV station broadcasts, and mobile

Aerial RF energy (corona) emitted from other high telecommunication.

voltage components such as overhead insulators. PD produced in the overhead lines.

The noise-discrimination approach presented in this paper is based on the fundamental properties of PD pulses. In solid- insulation defects, PD pulse shapes differ, and several PD pulse with different shapes can be generated simultaneously from the same defect [2]. PD in solid insulation gives rise to electromagnetic pulses having a rise time ranging from few ns to several 10's ns. These pulses become attenuated as they propagate. Strong frequency discrimination occurs as pulses of high frequency are attenuated more intensively than pulses of low frequency [3]. As a result, the rise time of a pulse degrades rapidly after a few feet of propagation, suffering a slower degradation thereafter. For example, the rise time of a 2-ns pulse can increase to 10-ns in the first 300 feet, but only to 30-11s after an additional 1,OOO feet of propagation.

In cable terminations, PD is measured within less than 5-feet from the source. As a result, PD pulses are expected to maintain their initial shapes due to the short proximity to the source. Therefore, it can be assumed that if the rise time of the detected pulse is less than 10 ns, it is very likely that this pulse has originated at the termination. Likewise, PD pulses originated within the cable will have much wider rise times when measured at the termination end of the cable. Noise discrimination based on the rise time properties of the pulse requires equipment with detection bandwidths greater than 300 MHz.

In this work, PD was measured in both time and frequency domains. Time domain detection is achieved using a 500 MHz-analog bandwidth pulse-height detector. The device is programmed to search for pulses with a rise time of less than 10 ns and while rejecting wider pulses. Frequency-domain measurements are made using a spectrum analyzer capable of discriminating frequency components up to 1.5 GHz. The device utilizes an analog technique to deconvolute the incoming signal into individual frequency components. PD originated in the termination will have frequency components as high as 500 MHz if measured within 5-feet of the source. On the other hand, the higher-frequency components of the PD originated in the cable system will be completely attenuated before they reach the termination. (Only the frequency components of 100 MHz or less of these pulses will be detected at the termination.)

Aerial RF energy (corona) emitted from other high voltage components such as overhead insulators are the most difficult to discriminate. Aerial-corona pulses are introduced into the termination either through direct emission or through the overhead-line conductors. In either way, the air media

controls the pulse attenuation. The RF-signal attenuation rate through air is slower than that through solid insulation. Corona noise effect was investigated by placing a corona source emitting different RF energy levels at different distances from the termination under test. It was found that no RF pulse having a rise time of less than 10 ns or frequency component greater than 200 MHz can be detected at the termination. In the mean time, RF sources located within less than 50-feet induce pulses that have almost the same rise time as those originated at the termination. In this case, corona noise discrimination must be based on pulse duration and repetition. In general, corona pulses have smaller width than PD pulses. Pulse repetition in air is normally higher than in solid insulation. The magnitude of the individual pulses is almost identical when they are produced in air. Furthermore, PD in air are more pronounced during the negative cycle. The difference of zero-span pattern between a PD signal generated in cable termination and a corona signal induced in the termination from a nearby corona source is shown in Figure 5 .

Radio and TV station broadcast noises are generally stationary and can be dealt with filtering techniques. Frequency-domain PD-detection in cable termination is conducted within a frequency window of 100 to 400 MHz. Fortunately, this window is free of broadcast radio RF.

During field testing, the differential noise coupling technique is also used to deal with both the aerial and broadcast interference. However, since the gain of the antenna and the PD sensor are not identical, the differential coupling technique can greatly reduced the RF aerial noise without completely eliminating it.

2.0

E 9 v e

2no

E

0.0

> E v

0.0 0.0 Time (ms) 50.0

Figure. 5 Zero span of A: PD signal. B: corona si,--'

111: Results

PD testing was conducted on two types of 115 kV cable terminations, namely, HPGF type and extruded cable type terminations. In both sets, PD measurements were made through both capacitate and inductive coupling. PD signals from both coupling techniques are identical to each other with the exception of the amplitude, as both sensors posses different coupling gains. PD signal picked up by the capacitor coupler is twice the magnitude of that picked by the inductive coupler.

5

A: HPGF Terminations

The test setup consisted of a 30 feet, 5 inch-diameter stainless steel pipe. Two terminations rated 115 kV were installed at both ends of the loop. Both terminations were fitted with an inductive sensor placed at the ground braid, refer to see Figure 1 . A 115 kV rated, HPGF cable was used to connect the terminations together. The cable shield was insulated from the pipe through a 0.5-inch polyester rope, which was helically wrapped over the cable insulation shield. Two insulated wires were used to provide access to the shield of the cable. These were soldered to the shield at the middle of the cable and penetrated the pipe wall through a feedthrough. These leads were used to monitor any PD that might originate in the cable itself. Prior to each sequent of testing, the system was evacuated to a vacuum level of 60 mtorr and maintained at that pressure for 7 days. The system was then flushed with nitrogen and then re-vacuumed again before it was finally pressurized to the required pressure of 200 psi with high grade, pure nitrogen.

In service, PD occurs in high-pressure gas filled terminations due to several factors. The two most dominant factors are aging resulting from load cycling and improper installation of the termination.

The load cycling effect is studied by subjecting the terminations to fifteen thermal load cycles. The thermal load cycling was achieved by means of induced current in the cable conductor. In each load cycle, the current was turned on for the first 8 hours and turned off for the remaining 16 hours. The conductor temperature was maintained at 85°C for the last four hours of the current period. A voltage of 85 kV was applied continuously during the load cycling, except during the determination of the PD inception voltage. For each cycle, the PD inception voltage was determined in cold condition, prior to the start of the load cycle, and at hot condition, immediately prior to the end of the current loading period, see figure 6. The PD onset voltage significantly decreased with load cycling. It reached a value of 35 kV after 15 load cycles. Such decrease was more pronounced after four load cycles. The PD characteristic was similar throughout the load cycling period. Exception from that is the sharp increase in the PD repetition as load cycling progresses, see figure 7.

100 , 2 80

f 60

40

- 0

6 g 20

0 0 2 4 6 8 10 12 14 16

Cycle Number

Figure 6 PD onset voltage during load cycling for HPGF cable type termination.

Phase Angle Figure 7 PD pulses after 7-load cycles (above) and 15-load cycles (bottom).

Installation defects is studied as follows: Following the installation of the termination stress-cone, the cable is allowed to slide down for 2-inches and then pulled back to its position, a mistake is often encountered during installation. This resulted in severe damage to several paper layers of the stress cone.

Small amount of intermediate PD was observed to occur at an applied voltage of 59 kV, which is about 8 kV below the line- to-ground rated voltage of the termination. However, increasing the applied voltage to 65 kV resulted in increase in PD frequency, pulse repetition and magnitude. It should be mentioned here that the characteristic of the PD resulted from the induced defect is somewhat similar to those resulting from aging induced by load cycling, see figure 8. PD of frequencies of up to 150 MHz was observed. PD pulses occurred at the positive cycles of the applied voltage was more intense than those of the negative cycles. After the initial assessment, the defected termination is then subjected to qualification test according to AEIC CS2-90. Tests included power factor determination at room temperature, 80°C and 90"C, ac withstand and hot impulse tests. After the termination successfully completed all the qualification tests, PD determination was made. The PD onset voltage dropped from 59 kV, prior t o the qualification tests, t o about 30 kV, after the qualification test. PD components ranging in frequency from 1 to over 400 MHz were observed. PD equally occurred at the positive and negative cycles of the applied voltage, with a pulse repetition of over lo00 pulse per a second, see figure 9.

2.0

s E -

0.0 Frequency (MHz) 200.0 2.0

E 9 v

Time (ms) 50.0 2

0.0

Phase Angle - i!

Figure 8 PD activity occur at 65 kV of a defected termination. A: Full span B: zero span C: Pulse count rate.

6

figure 9 PD activity of defected termination following qualification tests. A: Full span (0-200 MHz) B: Full span (200-500 MHz).

B: Extruded-Cable Termination

A test setup consisted of two 115 kV rated extruded cable terminations and 80-feet of XLPE cables were assembled. PD monitoring of the termination was conducted using both capacitive and inductive coupling. Inductive coupling was achieved by placing HFCT at the ground braid and at the cable, underneath the termination, see figure 2. For unaged terminations, the PD inception voltage was about 130 kV. However, the PD inception voltage decreased to a value of 70 kV when the terminations were subjected to 30 load cycles, see figure 10. In extruded cable termination PD frequency components of up to 500 MHz was observed. Figure 11 shows the PD signal detected from an inductive coupler placed at the ground braid, inductive coupler placed around the cable and capacitive coupler connected in parallel with the termination.

140 I

2oz 0 0 5 10 15 20 25 30

Cycle Number

figure 10 PD inception voltage during load cycling for extruded cable type termination.

IV: Conclusion

This paper discusses the possibility of conducting on-site PD measurements in transmission-class cable terminations. Both on-line and off-line approaches were presented. It demonstrated that utilizing the line voltage as the excitation voltage both on-line and off-line testing can be achieved.

V: References

[2] EH. Kreuger, ”Partial Discharge Part XVIII”, EEE Electr. Insul. Magazine, November, 1993, p.p.14-24.

S. Boggs,”Partial Discharge XXII”, IEEE Electr. Insul. Magazine, [3] January ,1996, p.p.9-16.

Figure 11 PD detected from extruded cable termination following 30 load cycle using (a): inductive coupler placed at the ground braid (b): inductive

coupler placed around the cable (c) capacitive coupler.

VI: Biographies

Nezar H. Ahmed received his B.S. degree from the University of Amman, Jordan, the M.S. and Ph.D. degrees from the University of Strathclyde, Glasgow, Scotland , all in electrical engineering in 1982, 1987, 1990 respectively. Dr. Ahmed‘s industrial experience is with Center for Electrical Power Engineering, Glasgow, Scotland and currently with Detroit Edison Company. Dr. Ah“s areas of interest are GIS, insulation coordination, high voltage testing, cables, and composite insulators. He has published over 20 papers on GIS systems, surface and space charges, partial discharges, field probing and polymeric insulating materials.

Oscar E. Morel received his M.S. and Ph.D. from the Pennsylvania State University in 1980 and 1987, respectively. Dr. Morel is a full-time research consultant with Detroit Edison Co. His industrial experience includes failure analysis, condition assessment and life expectancy of underground transmission cables and accessories. He is currently involved with several EPRI sponsored projects in the area of aging and condition monitoring of underground cables. Dr. Morel’s other interests include chemical and physical testing of dielectric materials utilized in high voltage applications.

Naeu Srinivas is an Electrical Engineer with 30 years experience in the field of cables, high voltage testing, and power equipment evaluation. Working with Detroit Edison as Engineering Support Leader. Electrical and Instrumentation Testing, he supervises a team of professionals working as testing, evaluation and reliability measurements of electrical products used in power plants and line hardware. Prior to 1982. he was working with Phelps Dodge Corp. as Chief Engineer (tests) in the R&D department. He was the project manager for several EPRI sponsored projects on treeing, life estimate and dc testing effect on cable life. He is the chairman of IEEE-ICC # 12 subcommittee, which deals with tests and measurements. He is a voting member of 1CC and a member of PES and DEIS. He has published several papers on cable aging and treeing.

[ I ] N. Ahmed and N. Srinivas, “On-line Partial Discharge Detection in Cables”, IEEE Trans. On Dielectric and Electr. Insul., Vol. 5, No. 2. 1998.

7