Fault location in active distribution networks using non-synchronized measurements

Electrical Power and Energy Systems 93 (2017) 451–458

Contents lists available at ScienceDirect

Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes

Fault location in active distribution networks using non-synchronized measurements Alireza Bahmanyar, Sadegh Jamali ⇑ Centre of Excellence for Power System Automation and Operation, School of Electrical Engineering, Iran University of Science and Technology, P.O. Box: 16846-13114, Tehran, Iran

a r t i c l e

i n f o

Article history: Received 1 February 2017 Received in revised form 1 June 2017 Accepted 16 June 2017

Keywords: Distributed generation Distribution networks Fault location Non-synchronized measurements

a b s t r a c t Fault location is a necessity to realize the self-healing concept of modern distribution networks. This paper presents a novel fault location method for distribution networks with distributed generation (DG) using measurements recorded at the main substation and at the DG terminals. The proposed method is based on an iterative load flow algorithm, which considers the synchronization angle as an unknown variable to be estimated. Therefore, it obviates the need of synchronized measurements. A new fault location equation is also proposed which is valid for all different fault types, hence the fault type information is not required. The developed method can be simply implemented by minor modifications in any distribution load flow algorithm and it is applicable to different distribution network configurations. The accuracy of the method is verified by simulation studies on a practical 98-node test feeder with several DG units. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction According to failure statistics, approximately 80% of all customer interruptions are due to faults on distribution networks [1]. It is therefore required to manage distribution faults in an efficient and effective manner to maintain the quality of service by minimizing the outage time. Distribution outage management is closely related to operators ability to find the fault location. In branched distribution networks dispersing over vast rural and urban areas, fault location can be a great help in minimizing the inspection and service restoration times. Accordingly, significant research efforts have been devoted to the subject. The proposed approaches can be classified into outage area location methods [2,3], and fault location methods. The first class uses data such as customer calls or fault indicating signals to find the most likely interrupted area, while the second class finds the location of the fault, which caused the resulting outage [4]. Based on the required inputs and their computational process, distribution fault location methods can be categorized into: impedance-based methods [5,6], algorithms based on sparse measurements [7–11], traveling waves-based methods [12–14], learning-based methods [15,16], and integrated methods [17–19]. Although some of these methods have presented

⇑ Corresponding author. E-mail addresses: [email protected] (A. Bahmanyar), [email protected] (S. Jamali). http://dx.doi.org/10.1016/j.ijepes.2017.06.018 0142-0615/Ó 2017 Elsevier Ltd. All rights reserved.

excellent results, most of them are designed for radial networks with a unidirectional power flow. However, the future distribution networks are expected to accommodate a variety of distributed generation (DG) sources, which change the radial nature of distribution systems to multi-source networks with bidirectional power flows. Hence, there is a need for a new class of fault location methods for such systems. Recently, some fault location methods are proposed for distribution networks with DG. In [20] authors present a learningbased scheme with very short online execution time. However, learning-based methods mostly require several actual or simulated fault cases for training and retraining following to any change in network topology. The methods described in [21–24] are modified impedance-based algorithms. These methods solve a set of equations for all line sections, one by one, to find all possible fault locations. In [25,26], authors propose algorithms based on sparse measurements. These algorithms apply the fault at all network nodes, one by one, and calculate the change in three-phase voltages at nodes having measurements. Finally, by comparing the measured and calculated values for all nodes, they identify the node with the minimum difference as the fault location. The genetic algorithm-based technique proposed in [27], optimizes the process of the methods of [25,26] to decrease their computational time. In [28], an integrated method finds all possible fault locations using an impedance-based algorithm and then uses the measured DG terminal voltages to identify the best solution. The authors of [29] propose a method based on a time-domain

452

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

numerical analysis using a data window moving from pre-fault to post-fault in time domain. Although this algorithm considers the dynamic behavior of generators during fault transients, it is only applicable to synchronous machine-based DGs. The above mentioned fault location methods mostly rely on synchronized measurements provided by meters installed at the main substation and at the DG terminals. It is possible to synchronize the measurements using the global positioning system (GPS) or computer networks [30]. However, unavailability of the synchronized measurements, even in some recent smart metering projects [31], limits the application of these methods. Therefore, there is a need for a new method that can employ the nonsynchronized measurements to find the accurate location of faults in distribution networks with DG. To obviate the need of synchronized measurements, this paper presents a new impedance-based fault location method, which considers the synchronization angle as an unknown variable to be estimated. Compared to the previously proposed fault location methods, the main contributions are as follows:
The proposed method can solve the fault location problem in distribution networks with or without DGs using synchronized and/or non-synchronized measurements;
A new fault location equation is proposed which is valid for all different fault types. Therefore, in contrast to the previously proposed impedance-based methods [5,6,21–23], the proposed method does not require fault type information;
Compared to the methods proposed in [21,29], which are designed for synchronous DGs, the presented method is applicable to all DG types without requiring their models and parameters.
The proposed method is fully load flow-based and it can be simply implemented by minor modifications in any distribution load flow algorithm. In the rest of the paper, Section 2 presents the details of the proposed fault location method. Section 3 gives the simulation results for different fault scenarios and finally, Section 4 concludes the paper. 2. Fault location method Following the occurrence of a fault and the protection system operation, the fault locator uses the measurements recorded at the head of the distribution feeder and at the feeder DG terminals to find the fault location. Smart meters, digital relays or fault recorders can provide the required measurements [30]. The proposed method considers the synchronization angle as an unknown variable to be estimated. Therefore, the measurements do not need to be synchronized. However, each meter should still be able to provide the during-fault voltage and current magnitudes, recorded for the same fault event, as well as their angular difference. A simple communication infrastructure such as the one described in [31] can provide the required measurements for the fault locator. System data such as network topological information, line section impedances, and load data are also required, which can be extracted from the distribution system database. Having the required measurements and system data, the fault locator solves the fault location equation for all network line sections to find the fault location. 2.1. Fault location equation Consider a faulted line section of a radial distribution system, as shown in Fig. 1, where [VS] = [Va, Vb, Vc]T and [IS] = [Ia, Ib, Ic]T are the

vectors of the sending end voltages and currents, [IR] = [Ira, Irb, Irc]T is the vector of downstream currents and [IF] = [Ifa, Ifb, Ifc]T is the vector of fault currents. The fault location voltage can be expressed as follows:

½V F  ¼ ½V S   d ½Z ½IS 

ð1Þ

where d is distance to the fault and [Z] is the line impedance matrix:

2

zaa

zab

zac

3

6 ½Z ¼ 4 zba

zbb

7 zbc 5

zca

zcb

zcc

Having the fault location voltage and its upstream and downstream currents, the three-phase current and apparent power flowing to the fault can be calculated using the following equations.

½IF  ¼ ½IS   ½IR 

ð2Þ

SF ¼ PF þ jQ F ¼ ½V F T ½IF 

ð3Þ

By assuming a pure resistive fault, the fault distance can be calculated by splitting the imaginary part of Eq. (3) and equalizing it to zero.

  V ik Irfk  V rk Iifk      d¼P P i i i r r r i r r i k¼a;b;c j¼a;b;c Ifk Ij zk;j  I j zk;j þ Ifk Ij zk;j þ I j zk;j P

k¼a;b;c

ð4Þ

where r and i represent, respectively, the variables real and imaginary parts. This equation is applicable for all different shunt fault types. Therefore, the proposed method does not require any information regarding the fault type. Due to the voltage drop caused by the fault, the during-fault downstream current in Eq. (2) is different from its pre-fault value and it is also an unknown variable. Therefore, for each line section, having the voltage and current phasors at the sending end ([VS], [IS]), the following iterative procedure is carried out to solve the fault location equations, at which [IR] and d are unknown values: (1) Assume the fault at the beginning of the line (d = 0); (2) Calculate the fault location voltage using Eq. (1); (3) Use the voltage obtained in step 2 to calculate [IR] by performing load flow for downstream network; (4) Calculate the fault current ([IF]) using Eq. (2); (5) Substitute [IF] into Eq. (4) to obtain the fault distance; (6) Repeat the above steps until the estimated fault distance converges to a certain value. Fig. 2 shows the flowchart of the proposed fault location method. The fault locator starts from the first line section. For this line, the sending end voltage and current ([VS], [IS]) are equal to the voltage and current measured at the head of the distribution feeder. Fault distance is initially assumed to be zero and using Eq. (1), the fault location voltage ([VF]) is calculated. Then, the calculated [VF] is taken as reference and as described in Section 2.2, a load flow is performed for the downstream network to obtain its current ([IR]). In the next step, having [VS], [IS], [IR] and [Z], the proposed method solves Eq. (4) to find d. If the calculated distance is less than the line section length, fault locator repeats the above procedure until d converges to a certain value. Otherwise, the fault is not in that section and the above steps should be repeated for another line section with the calculated voltage and current phasors at its sending end. The details of the proposed load flowbased algorithms for estimation of the downstream current and calculation of each line section sending end voltage and current are described in the next two subsections.

453

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

Downstream network VS

VF IS

IF

IR

Fig. 1. A faulted line section in a distribution network.

2.2. Load flow method for calculation of downstream current

Start

Get the measurements and system data

K=1 th

Select the K line section Calculate the voltage and current of the selected line as described in Section 2.3

K=K+1

d=0 Calculate the fault location voltage ([VF]) using Eq. (1)

As described in Section 2.2, perform a load flow for the downstream network using [VF] and calculate [IR]

In this section [IR] is calculated using the backward/forward load flow method for distribution networks which only requires the voltage at any point of a feeder (e.g. head of the feeder) to find its downstream current. As shown in Fig. 1, having calculated the fault point voltage [VF], by performing the load flow for the downstream network, [IR] can be calculated. However, the load flow method is modified to accommodate DG units and to use nonsynchronized measurements. Taking the main substation voltage (vSub (t)) as reference, the measured current and voltage of each DG can be written as follows:

v DGi ðtÞ ¼ V mDGi cosðwt  hi Þ

ð5Þ

iDGi ðtÞ ¼ Im DGi cosðwt  hi  /i Þ

where as shown in Fig. 3, hi is the phase difference between the main substation voltage with the ith DG voltage and ui is the phase difference between the DG voltage and current. If the DG and substation measurements are synchronized, then m hi, ui, V m DGi and I DGi are known values. Therefore, during the load flow, each DG unit participation can be considered as a current injection and the method does not need any information regarding the DG parameters and its interface model. However, if the meters

Calculate the fault distance (d) using Eq. (4) dold =d

Is the estimated distance less than the line length? Yes |d-dold|<ε Yes Report the estimated fault location

Are all the lines investigated? Yes

q

Vsub j

VDG

IDG

Stop

Fig. 2. Flowchart of the proposed fault location algorithm.

Fig. 3. The phasor diagram of DG current and its voltage taking the main substation voltage as reference.

454

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

are not synchronized, supposing that they still can measure the during-fault voltage and current magnitudes and the phase difference between them (ui), hi will be unknown. In the proposed algorithm, the hi values are initially supposed to m be zero. Then, having the values of ui, V m DGi and I DGi for three phases provided by DG terminal measurements, the algorithm considers each DG unit participation as a current injection. In each iteration, it calculates the line currents in a backward sweep and updates the nodal voltages in a forward sweep. Then, having the updated values of the phase angles of DG terminal voltages (i.e. hi), the algorithm modifies the DG current phasors and repeats this process until convergence is achieved. The steps of the proposed backward/forward sweep load flow using non-synchronized measurements are as follows: (1) Consider the DG units in the downstream network and make an initial guess for the phase angles of their terminal voltages (e.g. hi(0)=0). For each phase, IDGi can be expressed as follows: ð0Þ

ð0Þ

IDGi ¼ Im DGi \ð/i  hi Þ

ð6Þ

(2) Make an initial guess for nodal voltages (e.g. equal to the main substation voltage); (3) Use the nodal voltages to calculate the downstream network load currents; (4) Inject the DG currents at their terminals; (5) Starting from the end sections of the last laterals and moving towards the branches connected to the considered line section, perform a backward sweep and calculate branch currents; (6) Starting from the fault location and moving towards the last sections of the laterals, perform a forward sweep and update nodal voltages; (7) Having the calculated terminal voltages of the downstream network DG units, modify their current phasors:

  ðkÞ ðkÞ IDGi ¼ Im DGi \ /i  hi

ð7Þ

where hi(k) denotes the phase angle of the ith downstream network DG voltage at the kth iteration. (8) Repeat steps 3–7 until convergence is achieved, i.e.:

maxf j½V ðkÞ   ½V ðk1Þ j g < e

ð8Þ

After the convergence of the algorithm, [IR] will be equal to the calculated three-phase branch current flowing to the downstream network from the fault location. 2.3. Line section voltage and current estimation For each line section, prior to locating the fault, the voltage and current phasors of the sending end should be calculated. For each line, the developed algorithm divides the network to downstream and upstream networks linked by the selected line, as shown in Fig. 4. It then performs a load flow in the upstream network to find the current and voltage of the selected line. The steps of the algorithm are as follows: (1) Let the initial voltage for all nodes be equal to the main substation voltage; (2) Make an initial guess for the phase angles of DG terminal voltages (e.g. hi(0)=0) and calculate DG currents using Eq. (6); (3) Having the nodal voltages, calculate the upstream network load currents;

Vsub Isub

IDGi DG i

VDGi Upstream network

VS Selected line

IS

Downstream network Fig. 4. Downstream and upstream networks linked by the line under investigation.

(4) Inject the DG currents at their terminals; (5) Calculate the current of the line under investigation as follows: m h n h h i i X i X ðkÞ ðkÞ ðkÞ IS ¼ ½ISub  þ IDGi  ILj i¼1

ð9Þ

j¼1

where [ISub] is the three-phase measured during fault current of the main substation, n and m are the total number of upstream network loads and DGs respectively and k denotes the iteration; (6) Perform a backward sweep and calculate the upstream network branch currents; (7) Perform a forward sweep and update the upstream network nodal voltages; (8) Having the calculated DG terminal voltages, modify the DG currents using Eq. (7); (9) Repeat steps 3–8 until convergence is achieved, i.e.:

h i h i ðkÞ ðk1Þ maxf j IS  IS jg
ð10Þ

3. Case study Simulation tests are performed on a simplified model of an overhead, three-phase, 20 kV, real life distribution feeder in Iran, with a total installed power of 5.47 MVA. As shown in Fig. 5, this feeder contains several laterals. The average distance between two neighboring nodes is 441 m and the furthest node is located 19,750 m from the main substation. The test system is simulated in MATLAB/Simulink where the loads are modeled as constant impedances. In order to consider the presence of DG units in the distribution feeder, an average model of a 1.5 MW doubly fed induction generator (DFIG) driven by a wind turbine, and three similar average models of a 400 kW PV farm (4  100 kW) are connected to the simulated grid through step/up transformers. Several fault scenarios are considered for different fault types, locations, and resistances. For each scenario, the voltage and current waveforms are recorded at each source terminal assuming that the measurements have a sampling rate of 256 samples per cycle. Full cycle discrete Fourier transform is used to calculate the fundamental voltage and current magnitudes or phasors, which are fed into the developed method to obtain the fault location. In each test case, the fault location error is defined as the devi-

455

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

Feeder 1 HV/MV

2

3

4

5

6 7

M

9

8 10

11

12

13

14

15

16

22 26

27

28 19

33

32

31

34

37

38

18

25

29

30

17

23

24

20 21

M

35

36 39 40

41

64

42

M

82

75

52

51 69

74

53

73 87

86

93

94

85

55 54 71

70

90 89

56

57

72 84

83 58

96 92

61

60

50

78

M

59

77

66

62

76

79

88

65

67

47

48 80

63

68

46

81

M

44

45 49

91

43

97 98

M Metering equipment

95 Fig. 5. Topology of the 98-node real-life Iranian distribution feeder modified by adding DG units.

ation between the estimated fault distance (de), and the real fault distance (dr) from the main substation:

It should be noted that similar to any other impedance-based fault location method, the proposed method estimates multiple locations for each fault scenario. The best estimated fault location among the multiple candidates can be determined using the voltage matching techniques proposed in [18,28]. In the following sections, two sets of simulation tests are carried out. The first test set is performed under ideal conditions to study the effect of fault resistance, type, and distance on the fault location accuracy. In the second test set, the impact of load data errors and measurement inaccuracies are studied. 3.1. Performance under ideal conditions To study the effect of fault type, fault resistance and fault distance variations on the accuracy of fault location estimates, faults with different resistances from 2 X to 50 X are simulated at different locations. Fig. 6 shows the deviation between the actual and

Rf=2 Rf=5

ð11Þ 15

Error (m)

Error ¼ jde  dr j

20

Rf=10 Rf=20 Rf=50

10

5

0 4550

7975

10950

14000

16450

Fault distance (m) Fig. 6. Fault location error for single-phase-to-ground faults (AG) with different fault resistance values.

estimated locations for single-phase-to-ground faults. It can be seen that the proposed method provides accurate and almost similar results for different fault resistances. The method is not very

456

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

sensitive to fault resistance variations and the highest error for all the considered scenarios is 18 m, which is not significant. Fig. 7 shows the fault location error for different fault types simulated at different locations. Comparison of the results indicates that the proposed algorithm presents satisfactory results for different fault types. It can be seen that the fault location accuracy decreases as the fault distance from the main substation increases. However, the average error for the furthest considered fault, which is located 16,450 m away from the main substation, is less than 7 m. Therefore, generally, it can be concluded that fault distance variations do not affect the accuracy of the proposed method significantly. 3.2. Performance under non-ideal conditions In this part, the performance of the proposed algorithm is evaluated under load data errors and measurement inaccuracies. In modern distribution networks, smart meters enable real-time or near real-time registration of consumption data. However, due to communication infrastructure limitations, it would be impossible to retrieve all meters data within short time intervals [31], and the available load data is always prone to errors. In order to evaluate the influence of load data errors on the performance of the proposed method, the load data of the simulated feeder is taken as the actual data and the erroneous load data is created by random variation of this data:

  Si ¼ Sact 1 þ eLi rL i

ð12Þ

where Sact and eLi are the actual value of the load apparent power of i the ith node and a random number between 1 and 1, respectively and rL is the range of deviation considered. Measurements are also prone to errors due to noises, meter inaccuracies, etc. In order to study the effect of measurement errors on the proposed method, random errors are added to measured voltages and currents:

  m V Mi ¼ V act Mi 1 þ ei rV   m IMi ¼ Iact Mi 1 þ ei rI

ð13Þ

act m where Iact Mi , VMi and ei are the actual values of measured voltage and current and a random number between 1 and 1 for the ith meter, respectively and rV and rI are the ranges of deviation considered. Five cases are considered to evaluate the performance of the proposed algorithm:

Case (2) Random variation of all measured voltages and currents within 2% of deviation (rI = rV = 0.02); Case (3) Random variation of all loads within 10% of deviation (rL = 0.1); Case (4) Random variation of all loads within 30% of deviation (rL = 0.3); Case (5) Random variation of all loads and DG measurements within 30% and 2% of deviations respectively. The results are summarized in Fig. 8. Comparison of the results obtained for Case 1 and Case 2 with the ideal condition (Fig. 6) reveals that random errors within 1% of measured values do not affect the method accuracy significantly, but higher errors may slightly decrease its performance. In Case 3 and Case 4, the method accuracy under load data uncertainties is evaluated. The results indicate a correlation between methods precision and load data errors. However, even with random variation of all loads within 30% of deviation, the method still provides acceptable results. In all the considered scenarios, as the fault resistance increases, the effect of measurement and load data errors becomes more considerable. Case 5 is the worst case considered where both the load data and measurements are randomly changed. As shown in Fig. 8, the estimation errors are not significant and the maximum error is about 300 m of a 20 km feeder length. 3.3. Performance using phasor measurements The proposed method does not require phasor measurements and the results presented in the previous sections validated its good performance even without using such measurements. Synchronized metering infrastructures require more complex meters with capabilities such as GPS time tagging. However, the metering and communication infrastructures of distribution systems are continuously improving and in the near future, current meters may be replaced with meters capable of providing phasor measurements. In this section, the effect of using such measurements on the performance of the proposed method is studied. As shown in Fig. 9, when compared to Fig. 8, the performance of the algorithm is generally improved. In Case 1, and Case 2 at which random errors are added to the measured values, the method provides accurate estimations and its average errors are 24.8 m and 51.4 m, respectively, compared to 29.3 m and 73.62 m in Fig. 8. As can be seen in Fig. 9, although the performance reduces with random variation of loads, the results are satisfactory and the use

Case (1) Random variation of all measured voltages and currents within 1% of deviation (rI = rV = 0.01); (a) Error (m)

Case 1

Phase to phase fault (AC, rf=5 )

6

Error (m)

Error (m)

20 4500

7975

Case 1

2

10950

10950

14000

16450

14000

16450

Case 2

Case 3

Case 4

Case 5

300 200 100 0

7975

Case 5

(b)

Single phase to ground fault (AG, Rf=5 )

0 4550

Case 4

40

Three phase to ground fault (ABCG, rf=5 , Rf=5 )

4

Case 3

60

0

Two phase to ground fault (BCG, rf=5 , Rf=5 )

Case 2

4500

7975

10950

14000

16450

Fault Distance (m)

Fault distance (m) Fig. 7. Fault location error for different fault types.

Fig. 8. Fault location error for single-phase-to-ground faults (AG) under non-ideal conditions (a) Rf = 2 X (b) Rf = 20 X.

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

(a)

Error (m)

Case 1

Case 2

Case 3

Case 4

Case 5

45

457

future modern distribution networks with different configurations and metering infrastructures. Acknowledgment

30 15 0 4500

7975

10950

14000

16450

The first author gratefully acknowledges the research facilities provided during his Visiting Researcher Position at the Energy Department of Politecnico di Torino by Professor Ettore Bompard.

(b) Error (m)

Case 1

Case 2

Case 3

Case 4

Case 5

References

210 140 70 0 4500

7975

10950

14000

16450

Fault Distance (m) Fig. 9. Fault location error for single-phase-to-ground faults (AG) under non-ideal conditions using synchronized measurements (a) Rf = 2 X (b) Rf = 20 X.

of phasor measurements improves the method accuracy. In all of the considered scenarios, the fault location accuracy generally increases as the fault resistance decrease. 4. Conclusions In future modern distribution grids, fault location will be a necessity to realize the prescribed requirements for continuity of supply. The presence of DGs, however, changes the unidirectional nature of distribution networks and complicates the fault location. Considering the unavailability of synchronized measurements, even in some of the recent smart metering projects [31], this paper presents a new fault location method, for distribution networks with DG, which does not require such measurements. The proposed method is applicable to all DG types without knowing their model and parameters. It solves a single equation for all line sections to find the fault location and obviates the need for fault type information. A simplified model of a real life distribution feeder in Iran is used to test the proposed method. In order to consider the presence of DG units, average models of a DFIG and three PV farms are connected to the simulated network through step-up transformers. The simulation test results verify the accuracy of the proposed method for different fault types, fault resistances, and fault distances. Under ideal conditions of precise measurements, the maximum error is 18 m (i.e. only 0.11% of the distance from the main substation). Performance tests under non-ideal conditions show that random errors within 1% of measured values and random variation of all loads within 10% of deviation do not affect the fault location results significantly. Although the results indicate that higher rates of errors may decrease the accuracy, the proposed method provides acceptable results for all the studied scenarios, with a fault distance error below 315 m of a 20 km feeder length. Comparative tests indicate that by using synchronized measurements, if available, higher accuracies can be achieved. On a personal computer with 2-GHz Intel Core 2 Duo processor and 2 GB of RAM, the mean required computational time for the proposed method is about 16 s. In conclusion, the proposed method can solve the fault location problem in distribution networks with or without DGs using synchronized and/or non-synchronized measurements. Therefore, considering its general network application, simplicity, and practicality, the method can help to satisfy reliability requirements of

[1] Gonen T. Electric power distribution engineering. CRC Press; 2014. [2] Liu Y, Schulz NN. Knowledge-based system for distribution system outage locating using comprehensive information. IEEE Trans Power Syst 2002;17 (2):451–6. [3] Sun K, Chen Q, Gao Z. An automatic faulted line section location method for electric power distribution systems based on multisource information. IEEE Trans Power Del 2016;31(4):1542–51. [4] Bahmanyar A, Jamali S, Estebsari A, Bompard E. A comparison framework for distribution system outage and fault location methods. Electr Pow Syst Res 2017;145:19–34. [5] Salim RH, Resener M, Filomena AD, De Oliveira KRC, Bretas AS. Extended faultlocation formulation for power distribution systems. IEEE Trans Power Del 2009;24(2):508–16. [6] Salim R, Salim K, Bretas AS. Further improvements on impedance-based fault location for power distribution systems. IET Gener Transm Distrib 2011;5 (4):467–78. [7] Pereira RAF, Silva LGWD, Kezunovic M, Mantovani JRS. Improved fault location on distribution feeders based on matching during-fault voltage sags. IEEE Trans Power Del 2009;24(2):852–62. [8] Jamali S, Bahmanyar A, Bompard E. Fault location method for distribution networks using smart meters. Measurement 2017;102:150–7. [9] Majidi M, Arabali A, Etezadi-Amoli M. Fault location in distribution networks by compressive sensing. IEEE Trans Power Del 2015;30(4):1761–9. [10] Majidi M, Etezadi-Amoli M, Fadali MS. A novel method for single and simultaneous fault location in distribution networks. IEEE Trans Power Syst 2015;30(6):3368–76. [11] Trindade FCL, Freitas W, Vieira JCM. Fault location in distribution systems based on smart feeder meters. IEEE Trans Power Del 2014;29(1):251–60. [12] Borghetti A, Bosetti M, Nucci C, Paolone M, Abur A. Integrated use of timefrequency wavelet decompositions for fault location in distribution networks: theory and experimental validation. IEEE Trans Power Del 2010;25 (4):3139–46. [13] Robson S, Haddad A, Griffiths H. Fault location on branched networks using a multiended approach. IEEE Trans Power Del 2014;29(4):1955–63. [14] Goudarzi M, Vahidi B, Naghizadeh R, Hosseinian S. Improved fault location algorithm for radial distribution systems with discrete and continuous wavelet analysis. Int J Electr Power Energy Syst 2015;67:423–30. [15] Thukaram D, Khincha H, Vijaynarasimha H. Artificial neural network and support vector machine approach for locating faults in radial distribution systems. IEEE Trans Power Del 2005;20(2):710–21. [16] Rafinia A, Moshtagh J. A new approach to fault location in three-phase underground distribution system using combination of wavelet analysis with ANN and FLS. Int J Electric Power Energy Sys 2014;55:261–74. [17] Gazzana DS, Ferreira GD, Bretas AS, Bettiol AL, Carniato A, Passos LFN, et al. An integrated technique for fault location and section identification in distribution systems. Electr Pow Syst Res 2014;115:65–73. [18] Bahmanyar A, Estebsari A, Pons E, Jamali S, Bompard E. An improved fault location method for distribution networks exploiting emerging LV smart meters. In: IEEE workshop on environmental, energy, and structural monitoring systems (EESMS), Bari; June 2016. p. 1–6. [19] Trindade FCL, Freitas W. Low voltage zones to support fault location in distribution systems with smart meters. IEEE Trans Smart Grid 2016;PP:1–10. [20] Javadian S, Haghifam MR, Rezaei N. A fault location and protection scheme for distribution systems in presence of DG using MLP neural networks. In: IEEE power & energy society general meeting, Calgary; July 2009. p. 1–8. [21] Jamali S, Talavat V. Accurate fault location method in distribution networks containing distributed generations. Iran J Elect Comput Eng 2007;10(1):27–33. [22] Orozco-Henao C, Bretas A, Chouhy-Leborgne R, Herrera-Orozco A, MarínQuintero J. Active distribution network fault location methodology: a minimum fault reactance and Fibonacci search approach. Int J Electric Power Energy Sys 2017;84:232–41. [23] Mora-Flórez JJ, Herrera-Orozco RA, Bedoya-Cadena AF. Fault location considering load uncertainty and distributed generation in power distribution systems. IET Gener Transm Distrib 2015;9(3):287–95. [24] Grajales-Espinal C, Mora-Flórez J, Pérez-Londoño S. Advanced fault location strategy for modern power distribution systems based on phase and sequence components and the minimum fault reactance concept. Electr Pow Syst Res 2016;140:933–41. [25] Chen PC, Malbasa V, Dong Y, Kezunovic M. Sensitivity analysis of voltage sag based fault location with distributed generation. IEEE Trans Smart Grid 2015;6 (4):2098–106.

458

A. Bahmanyar, S. Jamali / Electrical Power and Energy Systems 93 (2017) 451–458

[26] Brahma SM. Fault location in power distribution system with penetration of distributed generation. IEEE Trans Power Del 2011;26(3):1545–53. [27] Jamali S, Bahmanyar A, Borhani-Bahabadi H. A fast and accurate fault location method for distribution networks with DG using genetic algorithms. In: Smart grid conference (SGC), Tehran; November 2015. [28] Alwash SF, Ramachandaramurthy VK, Mithulananthan N. Fault-location scheme for power distribution system with distributed generation. IEEE Trans Power Del 2015;30(3):1187–95.

[29] Jamali S, Talavat V. Dynamic fault location method for distribution networks with distributed generation. Electr Eng 2010;92(3):119–27. [30] Kezunovic M. Smart fault location for smart grids. IEEE Trans Smart Grid 2011;2(1):11–22. [31] Bahmanyar A, Estebsari A, Pons E, Jamali S, Bompard E, Patti E, et al. Emerging smart meters in electrical distribution systems: opportunities and challenges. In: 24th Iranian conference on electrical engineering (ICEE), Shiraz; May 2016. p. 1082–7.