Journal of Science & Technology 101 (2014) 0594)65 Fault Classification and Location on Transmission Line Using Anils Net Le Kim Hung'', Vu Phan Huan^ 'Danang University of Science and Technology No 54 Nguyen Luong Ban : Sir Da Nang City ^Central Electrical Testing Company Limited Received May 28, 2013; accepted: April 22, 2014 Abstract This paper presents an application of ANFIS approach for fault classification and fault location on transmission lines using measured data from one line terminal The input data of the ANFIS was derived via from the fundamental values of the voltage and current measurements, which were pre-processed Discrete Fourier Transform The ANFIS was trained and tested using various sets of data, which was obtained from simulations with 110 kV transmission line Dak Mil - Dak Nong model All simulations were carried out in f^atlab/Slmulink Results indicate that the ANFIS can locate the faults with high accuracy Keywords: Fault classification, Fault locahon, Transmission line, Anfis, Matlab/simulink Introduction Problem formulation Fault location fimction of Ime protection relays normally utilize fundamental frequency voltages and cunents measured at only one line end, however, that function generally does not exhibit sufficient accuracy The drawbacks of that techniques are that fault resistance and fault inception angle are not taken into account Furthermore, their accuracy is degraded when the line is fed from other terminal [I] Statistic data show that enors in fault locating fiinction of some relays in some certain cases may gel up to a couple of kilometers f2-6] On the other band, intelligent computational techniques such as Fuzzy Inference System (FIS), Artificial Neural Network (ANN) and adaptive network based fuzzy inference system (ANFIS) have been used in order to improve accuracy in fault location [7] 2.1 The no kV Transmission line Dak Mil - Dale Nong This paper presents an application of ANFIS for fault classification and location applied on 110 kV transmission line Dak Mil - Dak Nong model The ANFIS has been successfully applied for fault locator (FL) where the information of the voltage and current data of protection relay and circuit breaker are available Effects of varying fault location, fauh time, fault resistance and remote source infeed have been considered in this work The obtamed results clearly show that the proposed technique can accurately identify the fault type and locate faults on transmission lines under various fault conditions ConespondingAuthor: Tel: (+84)914.112 526 Email Iekimhung@dut.udn.vn The 110 kV transmission line Dak Mil — Dak Nong locates in Central Highland of Viet Nam This transmission Ime utilizes Areva P543 relay for protection and fault location purposes as shown in Fig 1Once fault occurs, substation operator retrieves fault records and uses it quickly determine cause and location of the fault, also check for other associated control and protection equipment's responses Further analysis may be earned out for preventing future failures and summarizing all conclusions Actual fauh reports are shown in Table All information shown there are collected over the year from 2012 to 2013 by Central Grid Company (CGC), protective relay is Areva P543 type - < j ^ Fig Schematic diagram of I lOkV Dak Mil - Dak Nong transmission line Journal of Science & Technology 101 (2014) 059-06S Settings of relay is as follow; Line Parameter Line length[km] Line Impedance [Ohm] Line Angle [deg] kZN Residual kZN Res Angle[deg] Value 57.6 14.97 69 62 12 y-ir^ Fig Application of ANFIS approach to fault classification and location in transmission line Parameters of simularion model are as follows: 110 kV transmission Ime with length of 57.6km 2, Three phase section Ime is used to represent the transmission line Line sequence impedances areRi= 0.0931 n/km; Ro= 1688 £J/km Li = 7.7233e-04 H/km, U= 0.0023 H/km C,= 4386e-08 F/km; Co= 4.831e-09 F/km A numeric display block is to indicate the calculated random per unit length of the fault location and fault types Three phase fault block to simulate fault types and other parameters Three-phase measunng blocks s used lo measure line voltages and load currents An ANN based relay FL will use information collected from bus S, detail is in section III Fig Simulation model m MATLAB/Simulink All collected data show that accuracy of fault location function is too low The maximum error of relay at 110 kV substation Dak Md is 24.338% and P543 at 110 kV substation Dak Nong is up to 107,844% The question is whether an ANFIS method is effective to improve the accuracy of the fault location In order to use the ANFIS technique for fauh detection, classification and location, the input parameters should he determined precisely The input parameters are obtained from numerical relay P543 and acmal fault location as shown in figure The output indicates where the fault occuned and classified D 1m d gn d d n I hi f ttial f I n g/ g d dd I n ub 2 S mula on mod I Simulation of 110 kV transmission system shown in Fig is used for analysing the proposed ANFIS algonthm Simulation model is implemented in MATLAB environment Pre-processing of voltage and cunent signals: Pre-processing is necessary to significantly reduces the size of the neural network and improves performance and speed of framing process Three phase voltages and three phase cunent input signals were sampled at a sampling frequency of I kHz and fiirther processed by simple 2nd-order low-pass Butterworth filter with cut-off frequency of 400 Hz Subsequently, one fill I cycle Discrete Fourier transform is used to calculate the fundamental component of voltages and currents The input signals were normalized in order to reach the ANN inputlevel(0, 1)[7,8], Proposed anfis based fault locator Main goal of this section is to design, develop, test and implement a complete strategy for the fault diagnosis as shown in Fig, The first step is fault detection, and then next step is to identify the fault type Next task of proposed algonthm is to pin-poinl the location of fault The goal of this paper is to propose an ANFIS based integrated tool to perform all of these tasks This fault locator utilizes vohage and cunenl measured at one end only Components of voltages and cunents change to linguistic variable and suitable Membership Functions (MFs) should he chosen for them Moreover, the rule base contains the Eiizzy ifthen rules of Takagi and Sugeno type, in which the output of each rule is a linear combination of input variables plus a constant term To tune the parameters of the FIS based fault locator, an adaptive network is trained based on with off-line data Then the tuned FIS is used on-line to accurately locate faults on the hne [7], JournalofScience& Technology 101 (2014)059-065 Table Parameter settings for generating training pattems Case No Parameters Fault type Fauit location Lf Itail Fault time f si Fault resistance Rf Loading fMVAl Set value AG, BO, CG, AB BC, AC, ABG, BCG, ACG, ABC 1,5,10,15,20,25,30, 35,40,45, 50, 55 0.06,0.065 1,3,5,7,10 1,10,30,50,70 Fig Flowchart of the proposed scheme Fig The ANFIS Editor displays the traming data for ANFIS A Table I ANFIS network desired outputs Fault type AG BO CO AB EC AC ABO BCG ACG ABC A 0 1 i B I I 1 c0 1 1 1 G 1 0 1 Output 10 The design process of the ANFIS fault classifier and locator go through the following steps [9.10]' Step 1: Generation a suitable training data To classify the fault, the following methodology has been adopted In order to represent the fault type conectty, a binary coding system has been developed The complete binary coding system and equivalent decimal numbers for representing all possible types of faults are given in Table To fauh location, each type of faults at different fault locations, fault resistance, loading and fault times have been simulated as shown below in Table After completing "training data", "training data" by selecting appropriate radio buttons in the Load data portion of the ANFIS Editor GUI m Matiab and then clicking Load Data.,, The loaded data (example tranA) is plotted on the plot region as shown in Fig Step 2; Selection of a suitable ANFIS structure for a given application, with consist of determming number of inputs and outputs, choosing membership functions for each mput and output and defining IfThen rules • Structure of Anfis for fault detection and classification: A single ANFIS for fault detection and classification of all the ten type of faults in the transmission Ime under varying power system operating conditions has been developed The block diagram of the proposed single ANFIS based fault detector and classifier approach is shown in figure 6, The ANFIS's mputs chosen here are the magnitudes of the fundamental components (50 Hz) of three phase cunents measured at the relay location Only the magnitudes recorded at one end of the line are used Thus, the ANFIS's inputs are four: la, Ih, Ic and lo One output should be to 10 in the conesponding phases and/or in neutral according to the types of faults on transmission lines Journal of Science & Technology 101 (2014) 059-06S figure shows the scheme of the system generated by ANFIS Editor GUI in Matiab for fault classification Phase current (la, lb, Ic, IG) Step 3: Training the ANFIS I ANFIS I Fault type Fig Block diagram of Single ANFIS based fault detector and classifier Choosing the FIS model parameter optimization method is the hybrid method, the number of training epochs and the training ermi tolerance Train the FIS model by clicking the Train Now button This action adjusts the membership fimction parameters and displays the enor plots as shown m Fig After the AnfisA is tramed, it needs to save into the folder DakMil that uses for running the simulation Then other type of faults on transmission line combines: FD.fis, AnfisB.fis, AnfisC.fis, AnfisAB.fis, AnfisBC.fis, AnfisAC.fis, AnfisABG.fis, AnfisBCG.fis, AnfisACG fis, and AnfisABC.fis, which are perform similarly as steps above Step 4: Evaluation of the tramed ANFIS using test pattems until its performance is satisfactory Simulation results using data from the power system model are presented in section Fig Membership function of mput variables for fault classification • Structure of ANFIS for fault location: Based on the fault type, appropriate network detects and classifies the fault Ten different ANFIS modules were developed to process different fault type, namely of ANFIS modules based fault location is shown in table Smgle phase to ground faults has inputs; double phase to ground faults and phase to phase faults has inputs; and three phase faults has inputs The mputs are the magnitudes of the fundamental components (50 Hz) of three phase voltages and currents measured at the relay location All modular ANFIS based fault location is output that present distance to fault • Generate an initial FIS model or load an initial FIS mode\ using the options in the Generate FIS portion of the GUI There are two partition methods: grid partitioning for classification fault and subtractive clustering for fault location to initialize your FIS using ANFIS After that, clicking on the Generate FIS button This displays a menu from which you can choose the number of membership fiinctions, MFs, and the type of mput and output membership fimctions Analys test results and discustion The trained ANFIS based fault detector and locator modules were then extensively tested using independent fault data sets which have not been used dunng traming stage Fault type, fault location and fault time were changed to investigate the effects of these factors on the performance of the proposed algonthm In practice, current and voltage data get from fault record of P543 at 110 kV substation Dak Mil in three types of fault categories: AG, CG and ACG that uses to compare the accuracy obtamed of P543 with ANFIS In tables 3,4 and 5, the maximum deviation of the estimated distance Le measured from the relay location and the actual fault location Lf is calculated and the resulting estimated error "e" is expressed as a percentage of total Ime length L of that section as: %Error = II, - L I ^— '—^ v Example for classify the fault, we choose constant for the output membership fimction To view the FIS model structure once an initial FIS has been generated by clicking the Structure button The Fig The error plots Journal of Science & Technology 101 (2014) 059-065 4.1 Test results of AG fault The network was tested by presentmg AG fault case with varying fault locations of total length, fault resistance, fault time and loading which show m Fig and Table The maximum error is about 1.852%, The estimated fault location is 31.93 km as compared to actual fault location of 33 km The estimated fault location is 9.53 km, in contrast the actual fault location is km thus maximum error is 2.656%, 4.3 Test results of ACG fault The test results of the ANFIS based fault classifier and fault locator module for ACG fault is shown m Fig 11 and Table 5, Tbese tests show the maximum enor of about 2.378% only Table5 Test results of ACG" fauh No Fig Test results of "AG" fault Tables Test results of-AG" fauh No 10 11 Fault resistance lOhml S 10 12 Loading |MVA| Lr |km| L |km| Error |%l a 10 15 20 25 30 35 40 45 50 55 60 4.283 8.911 13.81 17.18 22.14 26.94 31.93 37.28 43.24 48.92 49 10 11 13 17 22 28 33 38 43 48 53 53.53 1.58 42 32 25 84 1,85 123 42 59 92 4.2 Test results of CG fault The test results of the ANFIS based fault classifier and fault locator module for CG fauit is shown in Fig 10 and Table 4, Table Test results of "CG"fault No Fault resistance lOliml 5 9 10 10 11 12 l,oading IMVAI L, Iknil L, Ikml error l%i 10 15 20 86 53 0.243 2.656 0.451 0451 25 30 35 40 45 50 55 60 13 17 22 28 33 38 43 48 53 13.26 16.74 21 58 27.33 32.95 36.88 42.78 47.37 52.11 Fault resistance lOhml 10 12 Loading IMVAI Lr jiiml L |km| Error 10 15 20 25 30 35 40 45 50 55 60 13 17 22 28 33 38 43 48 53 3.886 7-736 12.65 17.45 23.04 0.197 28.64 34.05 38.81 44 37 47.37 52 1.111 1.822 406 |%| 0.458 0.607 0.781 1.805 2.378 1093 145 4.4 Comparison between results obtained from actual relay and ANFIS based fault locator In this case, actual cunent and voltage extracted from fault record P543 relay at Dak Mil substation have been used to investigate the effectiveness of proposed algonthm Results obtained from ANFIS based fault locator and those reported by relay are compared as shown in Table 6, Companson results reveal that the prediction capability of ANFIS is extremely good This result may simply reflect the fact that the input (voltage or current) was more directly correlated with the parameter being predicted It can he clearly seen that fault location accuracy as offered by ANFIS based fault locator is outstanding over those performed by P543 relay Table Results comparison Time fault Type fault Lf Ikm] ANFIS Le Error Ikml |%1 P543 [Vol 0.729 1.163 0.087 17/5/2013 AG 44.64 46 27 2.74 0-43 6/06/2013 ACG 26 243 27 33 1.88 786 1.944 0.382 1.094 10/6/2013 CG 40.029 39-23 I 38 24.34 6/9/2013 AG 27,69 26.11 2,82 92 545 Journal of Science & Technology 101 (2014) 059-065 J-MUl^il U ml|||llli Fig 10 Test results of "CG" fault Fig II Test results of "ACG" fault Table Distance to fault table collected on the relay and actual line fiom year 2012 to 2013 No Time fault 110 kV Substation Dak Nong Estimated Actual fault fault Error location on location P543 |kra| Ikml 30 083 26.72 |%| 110 kV Substation Dak Mil Estimated Actual fault fault Error location on location P543 (km) Ikml 0.5903 26.46 26.8 13/04/2012 11/05/2012 -1.697 12.126 23.998 44.73 45,474 1.2917 16/07/2012 17/05/2013 06/06/2013 10/06/2013 06/09/2013 08/10/2013 69.6 10.2 34.86 30.5 30.5 -2.686 7.482 12 959 31 357 17.571 29.91 28 53 107.844 7899 0816 22 446 1.0243 54.194 48.98 44.89 29 26.01 26.01 29.6 50 118 44,641 26,243 40,029 27,69 29.07 1.9757 0.4323 4.786 24.338 2.9167 0.92014 ReasoD, Type fault Lightning, AG Lighming, ABG Lightning, AG Lightning, AG Lightning, AC Lightning, CG Lightning, AG Lightning, BG Table Structure of ANFISfor fault classification and location No Type fault 10 11 AG BG CG AB BC AC ABG BCG ACG ABC FD Numbers 4 5 5 5 Anfis information Inputs Input mfs Description Ua, Ub, Uc, la 14 Ua, Ub, Uc, lb 12 Ua, Ub, Uc, Ic 12 Ua, Ub, Uc, la, lb 12 Ua, Ub, Uc, lb, Ic 12 Ua, Ub, Uc, la, Ic 11 Ua, Ub, Uc, la, lb 12 Ua, Ub, Uc, lb, Ic 11 Ua, Ub, Uc, la, Ic 12 Ua, Ub, Uc, Ia,Ib, Ic 14 la, lb, Ic, In Conclustion This paper has presented an application of ANFIS for locating faults in transmission line The ANFIS based proposed method is trained to classify the fault type, and separate ANFIS are designed to accurately locate the actual fault position on a transmission line The obtained results show that the primary advantages of the proposed algorithm can be summarized in three aspects First, it does not depend Outputs RMSE Epochs 3.01e-3 1.41e-3 2.82e-3 03e-3 2.47e-3 4.55e^ 2.47e-3 S.22e-4 5.47e-4 8.337e-3 1.14e-3 30 30 30 30 30 30 30 30 30 30 30 on the effect of the enors in CT and VT signals, fault resistance , Second, the accuracy of the fault location does not rely on the accuracy of the algorithm type, as in the case of the one-terminal algorithm of P543, Third, the method yields accurate result, ANFIS is easily trained in personal computer The only disadvantage of the method is that the obtained accuracy is amount of practical fault data in relay protection and actual fault location, Journal of Science & Technology 101 (2014) 0594)65 which may depend on system operating conditions Nevertheless, this issue can be addressed by the cunent technology such as supervisory control and data acquisition (SCADA), wide area monitoring (WAM), and Automation Substation Future work will focus on online taking of the fault information on relay protection to obtain even better fault location accuracy References [1} Anamlka Yadav, A S Thoke, 'Transmission line fault distance and direction estimation using artificial neural network" International Joumal of Engineering, Science and Technology, (2011) ] 10121 [2] Toshiba, Introducton manual distance relay GRZIOO, (2006) [3] Siemens, Introducton manual numencal distance protection relay Siprotec 7SA511, (1995) [4] Schweitzer Engineenng Laboratories, SEL-421 Relay Protection and Automation System, (2011) [5] Abb, REL 521 Line distance protection terminal, (2003) [6] Areva, Technical Manual, Fast Multifunction Distance Protection Relays P43, (2010) [7] M.joorabian, M monadi,"Anfis based fauit location for EHV transmission lines", aupec2005 - australia, (2005), [8] Tamer S Kamel M A, Moustafa Hassan, "AdapUve Neuro Fuzzy Inference System (ANFIS) For Fault Classification in the Transmission Lines",The Online Journal on Electronics and Electncal Engineenng (OJEEE) Vol, (2) - No,(l), (2009), [9] Adel A Elbaset and Takashi Hiyama "A Novel Integrated Protective Scheme for Transmission Line Using ANFIS" Sent lo International Journal of Electncal Power and Energy Systems, with Ms Ref No • IJEPES-D-08-00112, (2009) [ 10] Hitp://www.Mathworki Com/Help/Fuzzy/Anfis-AndThe-Anfis-Editor-Gui Html ... AnfisC.fis, AnfisAB.fis, AnfisBC.fis, AnfisAC.fis, AnfisABG.fis, AnfisBCG.fis, AnfisACG fis, and AnfisABC.fis, which are perform similarly as steps above Step 4: Evaluation of the tramed ANFIS using test... the AnfisA is tramed, it needs to save into the folder DakMil that uses for running the simulation Then other type of faults on transmission line combines: FD.fis, AnfisB.fis, AnfisC.fis, AnfisAB.fis,... system generated by ANFIS Editor GUI in Matiab for fault classification Phase current (la, lb, Ic, IG) Step 3: Training the ANFIS I ANFIS I Fault type Fig Block diagram of Single ANFIS based fault