research article intelligent fault diagnosis in a power...

Research ArticleIntelligent Fault Diagnosis in a Power Distribution Network

Oluleke O. Babayomi1 and Peter O. Oluseyi2

1Centre for Space Transport and Propulsion, National Space Research and Development Agency (NASRDA), Epe, Lagos, Nigeria2Department of Electrical and Electronics Engineering, University of Lagos, Lagos, Nigeria

Correspondence should be addressed to Oluleke O. Babayomi; [email protected]

Received 27 June 2016; Revised 15 September 2016; Accepted 26 September 2016

Academic Editor: Pascal Venet

Copyright © 2016 O. O. Babayomi and P. O. Oluseyi. This is an open access article distributed under the Creative CommonsAttribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work isproperly cited.

This paper presents a novel method of fault diagnosis by the use of fuzzy logic and neural network-based techniques for electricpower fault detection, classification, and location in a power distribution network. A real network was used as a case study. Theten different types of line faults including single line-to-ground, line-to-line, double line-to-ground, and three-phase faults wereinvestigated. The designed system has 89% accuracy for fault type identification. It also has 93% accuracy for fault location. Theresults indicate that the proposed technique is effective in detecting, classifying, and locating low impedance faults.

1. Introduction

Fault diagnosis and resolution in a power system networkare essential for clearing faults that manifest in an electricalpower transmission or distribution network. The process offault resolution comprises three stages: first, the detection andidentification or classification of unusual voltage and currentcharacteristics at the affected portions of the network; next,the location of the incidence of the fault to enable quick accessand solution to the problems that arise in the power network;and finally, the fault being cleared within the shortest timepossible to prevent damage to unaffected parts of the network.

1.1. Related Work. Numerous studies have been carried outon the use of intelligent methods for electric fault diagnosisin an electrical system. Some of these methods includeexpert systems, artificial neural networks, and fuzzy logic.The following review of related literature will be delimited toartificial neural network and fuzzy logic applications.

In certain studies, neural network principles were notapplied to the power fault diagnostic process. Thus, theylacked the capabilities to learn from data gathered from theelectrical network.This was the case in [1], where fuzzy logic-based fault identification in an electric power distributionsystem was studied and proven to produce accurate classi-fications of fault types. In addition, the fuzzy logic method

was used in combination with discrete wavelet transform andresulted in accurate fault identification [2]. In [3], the datacollected by alarms and protection relays in a power networkwas analyzed with neurofuzzy techniques. A classificationbased on input signals into faulty component type with ahigh degree of accuracy was achieved in spite of corruptedalarm signals. Petri net and neurofuzzy methods were usedin [4, 5] for fault location in power lines and sections. Theadaptive neurofuzzy inference system (ANFIS)was employedin [6] for accurate fault location for transmission lines andunderground cables. However, the described procedures in[3–6] are not suitable for power distribution networks.

A neurofuzzy means of fault classification, location, andpower restoration plan in an electric power distributionsystem was developed in [7]. Three ANFIS modules wereemployed for fault type classification, 𝑥-coordinates, and 𝑦-coordinates of the fault location, respectively. The resultingsystem performed with a high degree of accuracy. However,it has a shortcoming in the level of accuracy of fault typeclassification which can be improved upon. The robustnessand precision of ANFIS were validated in [8] by testing thecharacteristics of the system after the addition of white noiseto input data.

The accuracy of the fuzzy inference system method offault diagnosis varies with complementary analytical toolsemployed to enhance the capabilities of the system. Reference

Hindawi Publishing CorporationAdvances in Electrical EngineeringVolume 2016, Article ID 8651630, 10 pageshttp://dx.doi.org/10.1155/2016/8651630

2 Advances in Electrical Engineering

[8] reported 78% accuracy in fault type identification whenthe fuzzy inference was used to analyze data derived by thewavelet transform method. 91% accuracy was reported in [9]for fault detection, while 93% accuracy was achieved for faultlocation when the first and third harmonic data sets derivedfrom discrete Fourier transform of fault current were appliedto intelligent fault diagnosis. In [10], awaveletmultiresolutionanalysis was used to extract harmonics generated by currenttransients due to fault incidences. This data was used todevelop an ANFIS model that gave accurate fault locationwith amaximum error of 5%. Furthermore, through aMonteCarlo simulation, the derived algorithm was proven to beimmune to the effects of fault impedance, power angle, faultdistance, and fault inception angle. While past researchershave focused on the application of either fuzzy logic orANFISfor fault detection, classification, and location, we proposethe use of a fuzzy logic controller for fault identification andANFIS for fault location for enhanced accuracies.

This research work focuses on the detection, identifi-cation, and location of faults in the distribution networkof a power system. Fault detection is achieved by a faultanalysis and the determination of positive, negative, andzero sequence currents and voltages of the power network.Thereafter, a fuzzy controller is included in the networkto identify the fault type on the occurrence of a fault.In addition, a neurofuzzy based fault location model isdeveloped. A distribution network in theNigerian power gridis used as a case study. The ensuing discussion starts with acomputational model, describes the methodology employedin the study, and rounds off with insights drawn from theresults obtained.

2. Computational Model

2.1. Fault Analysis. Electric power faults in a distributionsystem occur randomly and their severity varies in intensity.There are fourmajor types of faults: namely, three-phase fault,single line-to-ground fault (LG), line-to-line fault (LL), anddouble line-to-ground fault (LLG). Amongst these, the three-phase fault is the most severe when it occurs in a power grid,while the line-to-ground fault arises most commonly. LG, LL,and LLG faults cause unbalanced currents to flow through thenetwork and sequence diagrams are employed in the analysisof unbalanced fault incidents in electric networks. The faultcurrents due to three-phase LG, LL, and LLG faults are givenby (1) to (4), respectively. 𝑍𝑓 represents the fault impedance,𝑍𝑘𝑘 represents the impedance at bus 𝑘, 𝐼(0)

𝑓𝑎, 𝐼(1)𝑓𝑎, 𝐼(2)𝑓𝑎

are thezero, positive, and negative sequence fault currents in PhaseA, respectively, and𝑉𝑓 is the per unit voltage at the fault point:

𝐼(1)𝑓𝑎 = 𝑉𝑓𝑍(1)𝑘𝑘 + 𝑍𝑓 , (1)

𝐼(0)𝑓𝑎 = 𝐼(1)𝑓𝑎 = 𝐼(2)𝑓𝑎 = 𝑉𝑓𝑍(1)𝑘𝑘+ 𝑍(2)𝑘𝑘+ 𝑍(3)𝑘𝑘+ 3𝑍𝑓 , (2)

𝐼(1)𝑓𝑎 = 𝐼(2)𝑓𝑎 = 𝑉𝑓𝑍(1)𝑘𝑘+ 𝑍(2)𝑘𝑘+ 𝑍𝑓 , (3)

𝐼(1)𝑓𝑎 = 𝑉𝑓𝑍(1)𝑘𝑘+ 𝑍(2)𝑘𝑘(𝑍(0)𝑘𝑘+ 3𝑍𝑓) / (𝑍(2)𝑘𝑘+𝑍(0)𝑘𝑘 + 3𝑍𝑓) ,

𝐼(2)𝑓𝑎 = −𝐼(1)𝑓𝑎 [ 𝑍(0)𝑘𝑘+ 3𝑍𝑓

𝑍(2)𝑘𝑘+𝑍(0)𝑘𝑘 + 3𝑍𝑓] ,

𝐼(0)𝑓𝑎 = −𝐼(1)𝑓𝑎 [ 𝑍(2)𝑘𝑘𝑍(2)

𝑘𝑘+𝑍(0)𝑘𝑘+ 3𝑍𝑓] .

(4)

2.2. Fuzzy Inference System. Fuzzy logic involves reasoningalgorithms that mimic human thinking in a manner describ-able as a kind of gray logic, as opposed to binary logicthat uses only two values. Therefore, fuzzy logic associatesinput data with a range of values between 0 and 1. Thedata is thus processed by a fuzzy controller in three stages:namely, fuzzification, fuzzy processing, and defuzzification.Fuzzification translates input data into a fuzzy form withthe aid of input membership functions. Common member-ship functions include the triangular-shaped, bell-shaped, S-shaped, and Z -shaped functions. Fuzzy processing associatesthe fuzzified inputs via a set of IF. . .THEN rules to determinehow the input membership functions will associate. Finally,defuzzification converts the value from the processing stageinto an output using methods such as the centre of gravitymethod and the maximum value method. There are twocommon fuzzy inference systems (FIS): namely, Sugeno-type FIS and Madami-type FIS. The Sugeno FIS is efficientand works well with mathematical, linear, optimization, andadaptive techniques. On the other hand, the Mandami FIS isintuitive and suitable for human input.

2.3. Adaptive Neurofuzzy Inference System. The differencebetween the adaptive neurofuzzy inference system (ANFIS)and the FIS is that, with the FIS, only fixed membershipfunctions that are chosen arbitrarily are used. However,ANFISmembership functions are adapted to a historical dataset. FIS modeling relies heavily on the user’s interpretationof the relationship between the input and output data. Onthe other hand, ANFIS improves the process by adapting theinput and outputmembership functions to the relationship ofa sample set of input/output data.This adaptation for bespokemembership functions is attained through neuroadaptivelearning. The learning process works similarly to that of neu-ral networks and calculates membership function parametersthat optimally permit the fuzzy inference system to track theinput/output data according to the following steps:

(1) Postulate a model structure that relates inputs tooutputs through membership functions and fuzzyrules using the Sugeno-type FIS.

(2) Collect input/output data for training by ANFIS.(3) Train the initial model with the data provided con-

strained by an error criterion.

2.4. Case Study. Nigeria’s electricity transmission grid is ata voltage of 330 kV, while the distribution network is a

Advances in Electrical Engineering 3

Load 2Load 1

9%

15% 15%

8.5%9%

8% 8% 8% 8%

18kV

330kV

330kV

132kV

90MVA90MVA90MVA90MVA

300MVA 250MVA

500MVA500MVA

150MVA

Figure 1: One-line diagram of distribution network.

Y

Fuzzy rules Normalization

Neuronaddition

Membershipfunctions

(fuzzification)

Defuzzification

M1

M2

M3

M4

X1

X2

Figure 2: Basic ANFIS model for a two-input-one-output system.

33 kV/11 kV system managed by eleven distribution compa-nies across the country. The test network shown in Figure 1 isan extract fromOrile District distribution system, under EkoElectricity Distribution Company (EKEDC).

3. Methodology

3.1. Fuzzy Fault Controller Design. The model for a two-input-one-output ANFIS is illustrated in Figure 2. The dia-gramoutlines the entire structure of the neurofuzzy inferencesystem.The fuzzy diagnostic controller of interest in Figure 3has six inputs: namely, 𝐼𝑎, 𝐼𝑏, 𝐼𝑐, 𝑉𝑎, 𝑉𝑏, and 𝑉𝑐. These are thephase currents and phase voltages, respectively.The inputs areprocessed by the fuzzy diagnostic controller and the output isa number representing the particular fault incidence in thedistribution network.

For the fuzzy inference system (FIS) being designed, themembership functions employed for both input and out-put are triangular-shaped. Each input has two membershipfunctions labelled LW (low) and HI (high), while the outputhas eleven membership functions labelled O0 to O10, asillustrated in Figure 4.

In addition, Table 1 shows the sixty-four rules applicableto the 6-input FIS. However, only ten of these define the faulttypes of interest. The remaining are designed to give outputsof O0, implying that they are nonapplicable situations.

Figures 5(a) and 5(b) are plots of Load 1 three-phasecurrents and voltages, respectively, under normal conditions.The calibration of the Mandami-type fuzzy fault controllerinvolved the use of no-load RMS values of current (994A)and voltage (18.56 kV) as the base design. Values of currentand voltage for each of the 10 faults were recorded. Generally,any value of current and voltage 1% above the base valueswas set as high (HI) and values from 0 to 1.0 pu were low(LW). Table 1 was drawn by recording the characteristics ofeach phase current and voltage at the incidence of the faults.In Figure 6, the characteristics of currents and voltages areshown for four types of faults that affect Phase A. Theseplots corroborate the information in Table 1: for example,an LG fault on Phase A results in a high Phase A current,relatively low Phases B and C currents, low Phase A voltage,and relatively high Phases B and C voltages.

3.2. ANFIS Model Design. The given distribution networkwas modeled and simulated with SIMULINK for Load 1(50MW, 30MVar) and Load 2 (60MW). At Load 1 and Load2, each of the 10 faults was simulated over a range of valuesof the fault impedance from 0.001Ω to 10Ω for 150ms. Theresulting values of phase currents and voltages were recordedand this data (3,454 data points) was later used to train theFIS. Data for checking the model results was collected at afault resistance of 6.5Ω and this set of data was not used intraining the FIS. Fault samples were taken 25ms after theincidence of the fault.

The initial Sugeno fuzzy inference system which has sixinputs, namely, 𝐼𝑎, 𝐼𝑏, 𝐼𝑐, 𝑉𝑎, 𝑉𝑏, and 𝑉𝑐, was chosen to have 3membership functions of the generalized bell type for eachof the inputs. The use of three input membership func-tions resulted in lower errors than when two membershipfunctions were applied. The output membership functionwas set as constant, while for optimization hybrid combina-tion of least-squares estimation with back propagation wasemployed. The initial input membership functions prior totraining and after training are illustrated in Figures 7(a) and7(b). As shown, the training process significantly alteredthe voltage membership functions. The resulting ANFISmodel has 1,503 nodes, 729 linear parameters, 53 nonlinearparameters, and 783 total parameters. 2,570 data pointswere used for training, while 884 data points were used forchecking the model. A parameter-to-training data pair ratioof 3.3 indicates that the training data was sufficient to capturethe general characteristics of the physical system.

The ANFIS designed has twenty outputs. Outputs 10to 19 represent ten fault types at Load 1, while outputs 20to 29 represent ten fault types that occur at Load 2. The


Fault fuzzycontroller

(Mandami)

Fault ID

Ia

Ib

Ic

Va

Vb

Vc

Figure 3: Block diagram of 6-input fuzzy inference system.

0 2 4 6 8 10

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8

1

Deg

ree o

f mem

bers

hip

LW HI

0 2 4 6 8 10

LW HI

0 2 4 6 8 10

LW HI

0 1

LW HI

0 1

LW HI

0 1

LW HI

Ia Ib Ic

VaVb Vc

(a)

0 1 2 3 4 5 6 7 8 9 10Output

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1 O0 O1 O2 O3 O4 O5 O6 O7 O8 O9 O10

(b)

Figure 4: (a) Fuzzy input membership functions. (b) Fuzzy output membership functions.


0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2−1500

−1000

−500

0

500

1000

1500

Time (sec)

Normal load condition

Ia

Ib

Ic

Iabc

(A)

(a)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

Normal load condition

−3

−2

−1

0

1

2

3

Va

Vb

Vc

×104

Vabc

(V)

(b)

Figure 5: Load current and voltage during normal conditions: (a) phase currents and (b) phase voltages.

Table 1: Fuzzy inference rules for 6-input fault system.

Rule 𝐼𝑎 𝐼𝑏 𝐼𝑐 𝑉𝑎 𝑉𝑏 𝑉𝑐 Output Fault type

1 HI LW LW LW HI HI 1 LG fault, Phase A

2 LW HI LW HI LW HI 2 LG fault, Phase B

3 LW LW HI HI HI LW 3 LG fault, Phase C

4 HI HI LW HI HI HI 4 LL fault, Phases A-B

5 LW HI HI HI HI HI 5 LL fault, Phases B-C

6 HI LW HI HI HI HI 6 LL fault, Phases A-C

7 HI HI LW LW LW HI 7 LLG fault, Phases A-B

8 LW HI HI HI LW LW 8 LLG fault, Phases B-C

9 HI LW HI LW HI LW 9 LLG fault, Phases A-C

10 HI HI HI LW LW LW 10 3-phase fault

11 to 64 All other input combinations 0 N/A

interpretations of the twenty different outputs are highlightedin Table 2.

4. Results

Figure 8 illustrates the per-unit RMS phase currents andvoltages at the occurrence of line-to-line and double line-to-ground faults on lines B and C. The fuzzy fault controlleraccurately deciphered the fault numbers as 5 and 8, respec-tively. For line-to-ground faults on Phases A and C, Figure 9shows that the outputs of the controller are 1 and 3, respec-tively, which are accurate codes for the respective fault types.

Training of the ANFIS fault locator and identifier wascarried out in 120 epochs. Figure 10(a) indicates that the

ANFIS training error reduced as the number of epochsincreased. The curve shows the training error declining over40 epochs (the third 40 epochs of training; epochs 80–120).

Model validation is an important part of the entire designprocess. The validation of the developed ANFIS model wasdone by testing the trained FIS model with input-out data setthat was not used in the neurofuzzy training process.Thiswasachieved with 884 pairs of input/output vectors. Figure 10(b)illustrates the ANFIS output and actual expected output for884 data points of the checking data. From the resulting plots,we can deduce that the ANFIS model has 51% accuracy foridentification of fault type and 93% accuracy for the locationof faults that occur at either Load 1 or Load 2. Furthermore,the system was tested for situations when similar faults occur


0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2−4000−2000

02000400060008000

100001200014000

Time (sec)

LG fault, Phase A

Ia

Ib

Ic

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2

−3−2−1

0123

Time (sec)

LG fault, Phase A

Va

Vb

Vc

×104

×104

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2−1.5−1

−0.50

0.51

1.5

Time (sec)

LLG fault, Phases A-C

Ia

Ib

Ic

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2

−3−2−1

0123

Time (sec)

LLG fault, Phases A-C

Va

Vb

Vc

×104

×104

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2−1.5

−1

−0.5

0

0.5

1

Time (sec)

LL fault, Phases A-C

Ia

Ib

Ic

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2−3−2−1

0123

Time (sec)

LL fault, Phases A-C

Va

Vb

Vc

×104

−1500−1000

−5000

50010001500

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2

Time (sec)

3-phase fault

Ia

Ib

Ic

0

0.02

0.04

0.06

0.08 0.

1

0.12

0.14

0.16

0.18 0.

2

−3−2−1

0123

Time (sec)

3-phase fault

Va

Vb

Vc

×104

Vabc

(V)

Vabc

(V)

Vabc

(V)

Vabc

(V)

Iabc

(A)

Iabc

(A)

Iabc

(A)

Iabc

(A)

Figure 6: Plots of Phase A current and voltage during different faults.


1 2 3 4 5 6 7

in1mf1 in1mf2 in1mf3

in4mf1 in4mf2 in4mf3 in5mf1 in5mf2 in5mf3 in6mf1 in6mf2 in6mf3

in2mf1 in2mf2 in2mf3 in3mf1 in3mf2 in3mf3

1 2 3 4 5 6 1 2 3 4 5

1 1 1

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8

Ia Ib Ic

Va Vb Vc

(a)

1 2 3 4 5 6 7

in1mf1 in1mf2 in1mf3

in4mf1 in4mf2 in4mf3 in5mf1 in5mf2 in5mf3 in6mf1 in6mf2 in6mf3

in2mf1 in2mf2 in2mf3 in3mf1 in3mf2 in3mf3

1 2 3 4 5 6 1 2 3 4 5

1 1 1

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0

0.2

0.4

0.6

0.8

1

Deg

ree o

f mem

bers

hip

0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8

Ia Ib Ic

Va Vb Vc

(b)

Figure 7: (a) Initial input membership functions. (b) Trained input membership functions.


024

RMS current for LL fault on Phases B-C

00.5

11.5

RMS voltage for LL fault on Phases B-C

0246

Faul

t typ

enu

mbe

r

Fault number plot for LL fault on Phases B-C

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

Va

Vb

Vc

Ia

Ib

Ic

Iabc

(p.u

)Vabc

(p.u

)

(a)

048

RMS current for LLG fault on Phases B-C

00.5

11.5

RMS voltage for LLG fault on Phases B-C

048

Faul

t typ

enu

mbe

r

Fault number plot for LLG fault on Phases B-C

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

Va

Vb

Vc

Ia

Ib

Ic

Iabc

(p.u

)Vabc

(p.u

)

(b)

Figure 8: Fuzzy fault controller output for (a) LL fault on Phases B-C and (b) LLG fault on Phases B-C.

048

RMS current for LG fault on Phase A

00.5

11.5

RMS voltage for LG fault on Phase A

024

Faul

t typ

enu

mbe

r

Fault number plot for LG fault on Phase A

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

Va

Vb

Vc

Ia

Ib

Ic

Iabc

(p.u

)Vabc

(p.u

)

(a)

0

42

6RMS current for LG fault on Phase C

00.5

11.5

RMS voltage for LG fault on Phase C

024

Faul

t typ

enu

mbe

r

Fault number plot for LG fault on Phase C

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2Time (sec)

Va

Vb

Vc

Ia

Ib

Ic

Iabc

(p.u

)Vabc

(p.u

)

(b)

Figure 9: Fuzzy fault controller output for (a) LG fault on Phase A and (b) LG fault on Phase C.


0 5 10 15 20 25 30 35 400.375

0.38

0.385

0.39

0.395

0.4

0.405

0.41

Epochs

RMSE

(roo

t mea

n sq

uare

d er

ror)

Error curves

(a)

0 100 200 300 400 500 600 700 800 9005

101520253035

Plot of actual faults and ANFIS analysis

0 100 200 300 400 500 600 700 800 900−10

−505

10ANFIS model errors for fault type and location

Index of measured inputs


ANFIS outputActual fault

(b)



ANFIS outputActual fault

0 500 1000 1500 2000 2500 3000 3500−15−10

−505

101520

Prediction errors for fault type and location

0 500 1000 1500 2000 2500 3000 35005

10152025303540

Plots of actual faults and ANFIS outputs(multiple simultaneous faults)

(c)

Figure 10: (a) Error curve for ANFIS model training process. (b) Plots of ANFIS output and error. (c) Plots of ANFIS output and error formultiple simultaneous faults.

at both loads simultaneously. 3,188 data points were collectedfor this purpose. In this case, the results shown in Figure 10(c)indicate that the ANFIS correctly identified the fault type in27% of cases and accurately reported one location of the faultin 78% of the cases.

Thus, the ANFIS model has a high accuracy of faultlocation when the fault occurs at a single point. However, thefault location accuracy is significantly reducedwhen the samefault occurs at multiple points in the distribution network.Furthermore, for the developed ANFIS model, fault typeidentification is relatively low for both single and multiple-point fault locations in the distribution network. Hence,for fault type identification, the fuzzy fault controller afore-described, which has 89% accuracy, is better suited for thepurpose. In comparison with the ANFIS reported in [9, 10],our designed ANFIS has similar levels of accuracies of faultlocation without the extraction of additional features likeharmonics from the input signals. In addition, the accuracy

of our fuzzy logic fault classification has similar levels ofperformance with the designs in [9, 10].

5. Conclusion

In this paper, an adaptive neurofuzzy method of achievingfault diagnosis in a power distribution system was presented.Data was first obtained from the network for ten typesof faults. The data collated was utilized in training thefuzzy inference system for both fault identification and faultlocation. The output of the neurofuzzy model is representedwith integers from 10 to 19 and 20 to 29, representing faulttypes for Loads 1 and 2 in the distribution system, respectively.The accuracy of fault location was high when only one typeof fault occurs in the power system network. However, whenthe same fault type occurs simultaneously at more than onepoint in the power system, the fault location accuracy issignificantly reduced. Therefore, complementing the ANFIS


Table 2: Interpretation of ANFIS model output.

ANFIS output Fault type10 LG, Ph. A (Load 1)11 LG, Ph. B (Load 1)12 LG, Ph. C (Load 1)13 LL, Ph. A-B (Load 1)14 LL, Ph. B-C (Load 1)15 LL, Ph. A-C (Load 1)16 LLG, Ph. A-B (Load 1)17 LLG, Ph. B-C (Load 1)18 LLG, Ph. A-C (Load 1)19 3-Ph. fault (Load 1)20 LG, Ph. A (Load 2)21 LG, Ph. B (Load 2)22 LG, Ph. C (Load 2)23 LL, Ph. A-B (Load 2)24 LL, Ph. B-C (Load 2)25 LL, Ph. A-C (Load 2)26 LLG, Ph. A-B (Load 2)27 LLG, Ph. B-C (Load 2)28 LLG, Ph. A-C (Load 2)29 3-Ph. fault (Load 2)

with a fuzzy logic-based fault identifier improves the accuracyof fault identification.

Competing Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

References

[1] W.-H. Chen, C.-W. Liu, andM.-S. Tsai, “On-line fault diagnosisof distribution substations using hybrid cause-effect networkand fuzzy rule-based method,” IEEE Transactions on PowerDelivery, vol. 15, no. 2, pp. 710–717, 2000.

[2] M. Jamil, R. Singh, and S. K. Sharma, “Fault identification inelectrical power distribution system using combined discretewavelet transform and fuzzy logic,” Journal of Electrical Systemsand Information Technology, vol. 2, no. 2, pp. 257–267, 2015.

[3] J. C. Stacchini De Souza, E. M. Meza, M. T. Schilling, and M.B. D. C. Filho, “Alarm processing in electrical power systemsthrough a neuro-fuzzy approach,” IEEE Transactions on PowerDelivery, vol. 19, no. 2, pp. 537–544, 2004.

[4] P. T. T. Binh and N. D. Tuyen, “Fault diagnosis of power systemusing neural petri net and fuzzy neural petri net,” in Proceedingsof the IEEE Power India Conference, pp. 554–558, New Delhi,India, April 2006.

[5] J. Sadeh and H. Afradi, “A new and accurate fault locationalgorithm for combined transmission lines using AdaptiveNetwork-Based Fuzzy Inference System,” Journal of ElectricPower Systems Research, vol. 79, no. 11, pp. 1538–1545, 2009.

[6] A. Ashouri, A. Jalilvand, R. Noroozian, and A. Bagheri, “Anew approach for fault detection in digital relays-based power

system using petri nets,” in Proceedings of the Joint InternationalConference on Power Electronics, Drives and Energy Systems(PEDES ’10), pp. 1–8, Christchurch, New Zealand, December2010.

[7] Rasli, Hussain, and Fauzi, “Fault diagnosis in power distribu-tion network using Adaptive Neuro-Fuzzy Inference System(ANFIS),” in Fuzzy Inference System—Theory and Applications,M. F. Azeem, Ed., chapter 15, pp. 315–336, InTech, Rijeka,Croatia, 2012.

[8] T. S. K. M. M. Hassan, “Adaptive neuro fuzzy inference system(ANFIS) for fault classification in the transmission lines,”Journal of Electrical and Electronic Engineering, vol. 2, pp. 2551–2555, 2010.

[9] M. S. Abdel Aziz, M. A. Moustafa Hassan, and E. A. Zahab,“High-impedance faults analysis in distribution networks usingan adaptive neuro fuzzy inference system,” Electric PowerComponents & Systems, vol. 40, no. 11, pp. 1300–1318, 2012.

[10] M. J. B. Reddy, D. V. Rajesh, P. Gopakumar, and D. K. Mohanta,“Smart fault location for smart grid operation using RTUs andcomputational intelligence techniques,” IEEE Systems Journal,vol. 8, no. 4, pp. 1260–1271, 2014.

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of


Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks


research article intelligent fault diagnosis in a power...

Documents