Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 Fault Classfcaton and Locaton on 220kV Transmsson lne Hoa Khanh Hue Usng Anfs Net Vu Phan Huan Electrcal Testng Central Company Lmtted, Da Nang, Vet Nam Emal: vuphanhuan@gmal.com Le Km Hung Danang Unversty of Technology, Da Nang, Vet Nam Emal: lekmhung@dut.udn.vn Nguyen Hoang Vet Ho Ch Mnh Cty Unversty of Technology, Ho Ch Mnh, Vet Nam Emal: nghoangvet2002@yahoo.com nference system (ANFIS) whch have been used to mprove the accuracy n fault locaton [7]. Ths paper presents an applcaton of ANFIS for fault estmaton along wth fault locaton on 220kV transmsson lne Hoa Khanh Hue by Matlab Smulnk. The Anfs has been successfully appled for fault locator (FL) where the nformaton of the voltage and current data of protecton relay and CB are avalable. The effects of varyng fault locaton, fault tme, fault resstance and remote source nfeed have been consdered n ths work. The obtaned results clearly show that the proposed technque can accurately classfy the fault type and locate faults on transmsson lnes under varous fault condtons. Abstract Ths paper presents an applcaton of Anfs approach for fault classfcaton and fault locaton on transmsson lnes usng measured data from one lne termnal. The nput data of the Anfs derved from the fundamental values of the voltage and current measurements usng dgtal sgnal processng va Dscrete Fourer Transform. The Anfs was traned and tested usng varous sets of feld data, whch was obtaned from the smulaton of faults at varous fault scenaros (fault types, fault locatons and fault resstance) of 220kV transmsson lne Hoa Khanh - Hue n Vet Nam usng a computer program based on Matlab/Smulnk. Detaled explanaton and results ndcate that the Anfs can determne the locaton of the fault upon ts occurrence n order to speed up the repar servce and restore the power supply. II. Index Terms protecton relay, fault classfcaton, fault locaton, transmsson lne, anfs, matlab/smulnk I. A. ANFIS Archtecture [8] An adaptve network s a multlayer feedforward network n whch each node performs a partcular functon on the ncomng sgnals, as well as on the set of parameters pertanng to ths node. The formula for the node functons may vary from node to node, and the choce of each node functon depends on the overall nput output functon whch the adaptve network s requred to carry out. The parameter set of an adaptve network s the unon of the parameter sets of each adaptve node. In order to acheve a desred nput output mappng, these parameters are updated accordng to gven tranng data and a gradent-based learnng procedure. In the rest of ths secton, the archtecture of ANFIS as an adaptve network s descrbed. For the purpose of llustraton, the smplfyng assumpton s consdered that the fuzzy nference systemunder consderaton has two nputs x and y and one output f. Suppose that the rule base contans two fuzzy fthen rules of Takag Sugeno s type: INTRODUCTION Nowadays, fault locaton methods of multfuncton dgtal relays base on fundamental frequency voltages and currents measured at one end of the lnes, whch do not have suffcent accuracy. In such technques the nfluence of fault resstance and fault ncepton angle are not taken nto account. Furthermore, ther accuracy s degraded when the lne s fed from other termnal [1]. As a result, the locaton error of Toshba s a maxmum of ±2.5 km for faults at a dstance of up to 100 km, and a maxmum of ±2.5% for faults at a dstance between 100 km and 250 km [2] or accurate fault locaton of Semens s 2.5% of lne length (wthout ntermedate nfeed) [3], or Sel s 2.5% [4], or Abb s 2.5% [5] and Areva s 2.5% [6]. On the other hand, ntellgent computatonal technques such as Fuzzy Inference System (FIS), Artfcal Neural Network (ANN) and adaptve network based fuzzy Manuscrpt receved December 15, 2013; revsed February 4, 2014. do: 10.12720/joace.3.2.98-104 ADAPTIVE NETWORK-BASED FUZZY INFERENCE SYSTEM 98
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 Rule 1: f x s A1 and y s B1 then f1 = p1x + q1y + r1. Rule 2: f x s A2 and y s B2 then f2 = p2x + q2y + r2. of membershp functons on lngustc label A. In fact, any contnuous and pecewse dfferentable functons, such as commonly used trapezodal or trangular-shaped membershp functons, are also qualfed canddates for node functons n ths layer. Parameters n ths layer are referred to as premse parameters. Layer 2: Every node n ths layer s a fxed node labeled π, whch multples the ncomng sgnals and sends the product out. For nstance, w A ( x) B ( y) = 1, 2 (4) Each node output represents the frng strength of a rule. Layer 3: Every node n ths layer s a fxed node labeled N. The th node calculates the rato of the th rule s frng strength to the sum of all rules frng strengths: w w = 1, 2 w1 w2 (5) For convenence, outputs of ths layer wll be called normalzed frng strengths. Layer 4: Every node n ths layer s an adaptve node wth a node functon: O4 w f w ( p x q y r ) (6) where w s the output of layer 3, and {p, q, r} s the parameter set. Parameters n ths layer wll be referred to as consequent parameters. Layer 5: The sngle node n ths layer s a fxed node labeled that computes the overall output as the summaton of all ncomng sgnals,.e.: O5 overalloutput w f Fgure 1. (a) Fuzzy reasonng and (b) equvalent ANFIS Then the fuzzy reasonng s llustrated n Fg. 1(a) and the correspondng equvalent ANFIS archtecture s shown n Fg. 1(b). The node functons n the same layer are of the same functon famly, as descrbed below: Layer 1: Every node n ths layer s an adaptve node wth a node functon: O1 A ( x) (1) the membershp functon of A and satsfes the quantfer A. Usually μa(x) s chosen to be bell-shaped or Gaussan wth maxmum equal to 1 and mnmum equal to 0, such as 1 1 [(( x c ) / a ) 2 ]b (2) x c a 2 (7) E (8) k (9) ( E / ) 2 (3) where E s the output error, η s the learnng rate parameter and k s a parameter whch s automatcally vared durng learnng process to adapt to the learnng rate. k s ncreased f four consecutve learnng epochs reduce output error and s decreased f two consecutve where {a, b, c} s the parameter set. As the values of these parameters change, the bell-shaped or Gaussan functons vary accordngly, thus exhbtng varous forms and or A ( x) exp B. ANFIS Learnng Algorthm ANFIS employs two modes of learnng. Frst, a forward pass s made usng current premse parameters to optmze rule consequent parameters usng least square estmaton based on output error. Ths s possble snce outputs are a lnear functon of consequent parameters. Second, a backward pass s made to alter premse parameters usng gradent-based learnng. Ths process of learnng s named Hybrd Learnng. The backward pass employs learnng n a smlar way as to the backpropagaton n neural networks. For each pass, each rule antecedent parameter α s changed accordng to where x s the nput node, and A s the lngustc label assocated wth ths node functon. In other words, O1 s A ( x) wf w 99
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 learnng epochs result n non-monotonc changes n error. E / s calculated usng the chan rule. III. Hoa Khanh s 4.93% (hgher than 2.5%) and AREVA P437 at 220kV substaton Hue s 13.1% (hgher than 2.5%). POWER SYSTEM UNDER STUDY A. The 110kV Transmsson Lne Dak Ml Dak Nong Ths power system s a double-ended transmsson lne Hoa Khanh - Hue at 300MVA, 220 kv, 50 Hz, 83.2km long uses ABB REL521 and AREVA P437 protecton relay at each end, whch shows n Fg. 2. Fgure 5. Applcaton of ANFIS Approach to Fault Classfcaton and Locaton on Transmsson Lne The queston s whether an ANFIS method s possble to mprove the accuracy of the fault locaton estmaton. In order to use the ANFIS technque for fault detecton, classfcaton and locaton, the nput parameters should be determned precsely. The nput parameters are obtaned from numercal relay ABB REL521 and actual fault locaton as shown n Fg. 5. The output ndcates where the fault occurred and classfed. Due to lmted avalable amount of practcal fault data, t s necessary to generate tranng/testng data usng smulaton. To generate data for the typcal transmsson system, a computer program has been desgned to generate tranng data for dfferent faults that wll be presented detal n subsecton B. Fgure 2. Schematc Dagram of 220kV Transmsson Lne Hoa Khanh Hue The relay desred to be checked s havng the followng actvated settngs: Lne length [km] X1 [Ohm] R1 [Ohm] X0 [Ohm] R0 [Ohm] 83.2 20.717 3.744 73.382 11.381 When a fault occurs, the operator stores the current, voltage values and other states of equpment. The cause and locaton of the fault can be quckly determned, and the behavor of the assocated control and protecton equpment can be evaluated. Concrete measures can then be deduced from the analyss of such faults, n order to prevent future falures as well as event summares n spreadsheet. Fault record table collected on the relay ABB REL521, AREVA P437 and actual lne from year 2008 to 2013 by Power Transmsson Company No.2 (PTC2) as shown n Appendx A, Fg. 3 and Fg. 4. B. Power System Smulaton For the analyss of operaton of the proposed ANFIS based relay, the power system shown n Fg. 6 s used. The power system network model s smulated n MATLAB 2012 software. It s a 220 kv, 50 Hz, 83.2km transmsson lne system wth the parameters are as follows: Fgure 3. The Percentage Error of Fault Locaton on AREVA P437 Fgure 6. Power System Model Smulated n MATLAB Smulnk Software. 1) The transmsson lne: three phase secton lne s used to represent the transmsson lne. Lne sequence mpedance: [RL1, RL0] = [0.0450, 0.1368] Ω/km. [LL1, LL0] = [0.7926, 0.28] H/km. [CL1, CL0] = [1.402e-08, 3.969e-09] F/km. 2) A numerc dsplay block s to ndcate the calculated random per unt length of the fault locaton and fault types. Fgure 4. The Percentage Error of Fault Locaton on ABB REL521 Revews: The relay s accuracy s degraded. The maxmum error of ABB REL521 at 220kV substaton 100
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 3) Three phase fault block to deduce fault types and specfy the parameters. 4) Three-phase measurng blocks to measure the three phase lne and load current and voltage values. 5) An ANN based relay FL s located at bus S, whch has been developed for fault detecton and fault dstance locaton. It wll be presented detal n secton IV. 6) Preprocessng of voltage and current sgnals: Preprocessng s a useful method that sgnfcantly reduces the sze of the neural network and mproves the performance and speed of tranng process. Three phase voltages and three phase current nput sgnals were sampled at a samplng frequency of 1 khz and further processed by smple 2nd-order lowpass Butterworth flter wth cut-off frequency of 400 Hz. Subsequently, one full cycle Dscrete Fourer transform s used to calculate the fundamental component of voltages and currents. The nput sgnals were normalzed n order to reach the Anfs nput level (0, 1) [9]. IV. PROPOSED ANFIS BASED FAULT LOCATOR the transmsson lne. The goal of ths paper s to propose an ntegrated method to perform each of these tasks by usng Anfs. For each of the dfferent knds of faults, separate Anfs have been employed for the purpose of fault locaton. Ths fault locator utlzes voltage and current of the fault data at the fault locator end of the lne only. Components of voltages and currents change to lngustc varable and sutable Membershp Functons (MFs) should be chosen for them. The desgn process of the ANFIS fault classfer and locator go through the followng steps [10], [11]: Step 1: Generaton a sutable tranng data To classfy the fault, the followng methodology has been adopted. In order to represent the fault type correctly, a bnary codng system has been developed. The complete bnary codng system and equvalent decmal numbers for representng all possble types of faults are gven n Table I. TABLE I. ANFIS NETWORK DESIRED OUTPUTS Fault type A B C G Output AG 1 0 0 1 1 BG 0 1 0 1 2 CG 0 0 1 1 3 AB 1 1 0 0 4 BC 0 1 1 0 5 AC 1 0 1 0 6 ABG 1 1 0 1 7 BCG 0 1 1 1 8 ACG 1 0 1 1 9 ABC 1 1 1 0 10 To fault locaton, each type of faults at dfferent fault locatons, fault resstance, loadng and fault tmes have been smulated as shown below n Table II. TABLE II. PARAMETER SETTINGS FOR GENERATING TRAINING PATTERNS. Case No Parameters Set value 1 Fault type AG, BG, CG, AB, BC, AC, ABG, BCG, ACG, ABC 2 Fault locaton Lf [km] 1, 10, 20, 30, 40, 50, 60, 70, 80 3 Loadng [MVA] 1, 50, 100, 150, 200, 250, 300 4 Fault resstance Rf [Ω] 1, 5, 10, 20, 30 5 Fault tme [s] 0.07, 0.075 Fgure 7. Flowchart Depctng the Outlne of the Proposed Scheme. To mplement a novel applcaton of ANFIS approach to fault classfcaton and locaton n transmsson lnes a proposed computer program based on Matlab software to calculate all ten types of faults that may occur n a transmsson lne. In ths aspect, the man goal of ths secton s to desgn, develop, test and mplement a complete strategy for the fault dagnoss as shown n Fg. 7. The frst step n the process s fault detecton. Once we know that a fault has occurred on the transmsson lne, the next step s to classfy the fault nto the dfferent categores based on the phases those are faulted. Then, the thrd step s to pn-pont the poston of the fault on Step 2: Selecton of a sutable ANFIS structure for a gven applcaton, wth consst of determnng number of nputs and outputs, choosng membershp functons for each nput and output and defnng If-Then rules. Fault curent (Ia, Ib, Ic, Io) ANFIS 1 FD Fgure 8. Block Dagram of Sngle ANFIS Based Fault Detector and Classfer 101
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 Structure of Anfs for fault detecton and classfcaton: A sngle ANFIS for fault detecton and classfcaton of all the ten type of faults n the transmsson lne under varyng power system operatng condtons has been developed. The block dagram of the proposed sngle ANFIS based fault detector and classfer approach s shown n Fg. 8. The ANFIS s nputs chosen here are the magntudes of the fundamental components (50 Hz) of three phase currents measured at the relay locaton. Only the magntudes recorded at one end of the lne are used. Thus, the ANFIS s nputs are four: Ia, Ib, Ic and Io. One output should be 1 to 10 n the correspondng phases and/or n neutral accordng to the types of faults on transmsson lnes. Structure of ANFIS for fault locaton: Based on the fault type, approprate network detects and classfes the fault. Ten dfferent ANFIS modules were developed to process dfferent fault type. Sngle phase to ground faults has 4 nputs; double phase to ground faults and phase to phase faults has 5 nputs; and three phase faults has 6 nputs. The nputs are the magntudes of the fundamental components (50 Hz) of three phase voltages and currents measured at the relay locaton. All modular ANFIS based fault locaton s 1 output that present dstance to fault. In the present study, namely of ANFIS modules based fault locaton s shown n Fg. 9. (Ua, Ub, Uc, Ia) ANFIS 2 AG (Ua, Ub, Uc, Ia, Ib) ANFIS 4 ABG (Ua, Ub, Uc, Ib) ANFIS 3 BG (Ua, Ub, Uc, Ib, Ic) ANFIS 5 BCG Fgure 9. Block Dagram of ANFIS Based Fault Locaton Generate an ntal FIS model usng the optons n the Generate FIS porton of the Matlab GUI. There are two partton methods: grd parttonng for classfcaton fault and subtractve clusterng for fault locaton to ntalze your FIS usng ANFIS. Moreover, the rule base contans the fuzzy f-then rules of Takag and Sugeno type, n whch And Method: prod, Or Method: max, Implcaton Method: mn, Aggregaton Method: max, Defuzzfcaton Method: wtaver (weghted average). On the other hand, the MF s type (gaussmf). For classfy the fault, we choose constant for the output membershp functon. For the fault locaton, we choose lnear for the output membershp functon. For example, the Fg. 10 shows the scheme of the system generated by ANFIS Edtor GUI n Matlab for fault classfcaton. To tune the parameters of the FIS based fault locator, an adaptve network s traned based on wth off-lne data. Then the tuned FIS s used on-lne to accurately locate faults on the lne. Fgure 10. Membershp Functon of Input Varables for Fault Classfcaton. Step 3: Tranng the ANFIS Choosng the FIS model parameter optmzaton method s the hybrd method, the number of tranng epochs (20 epochs) and the tranng error tolerance (0). Tran the FIS model by clckng the Tran Now button. Ths acton adjusts the membershp functon parameters and dsplays the error plots as shown n Table III. After the FIS s traned, t needs to save nto the folder HoaKhanh that uses for runnng the smulaton. TABLE III. STRUCTURE OF ANFIS FOR FAULT CLASSIFICATION AND LOCATION Anfs nformaton Type Input No anfs Inputs Outputs mfs RMSE Epochs 1 AG 4 14 1 2.41e-3 20 2 BG 4 14 1 1.16e-3 20 3 ABG 5 12 1 4.22e-3 20 4 BCG 5 12 1 3.12e-4 20 5 FD 4 3 1 1.14e-3 20 Then other type of faults on transmsson lne combnes: AnfsA.fs, AnfsB.fs, AnfsABG.fs, AnfsBCG.fs, whch are perform smlarly as steps above. Step 4: Evaluaton of the traned ANFIS usng test patterns untl ts performance s satsfactory. Smulaton results usng data from the power system model are presented n secton V. V. ANALYS TEST RESULTS AND DISCUSTION The traned ANFIS based Fault detector and locator modules were then extensvely tested usng ndependent data sets consstng of fault scenaros never used prevously n tranng. Fault type, fault locaton and fault tme were changed to nvestgate the effects of these factors on the performance of the proposed algorthm. In practce, current and voltage get from record of ABB REL521 at 220kV substaton Hoa Khanh that uses to nvestgate the effects of these factors on the performance of the proposed algorthm. The results compare the accuracy obtaned of REL521 wth Anfs that based fault classfer and fault locator module for AG, BG, ABG and BCG fault whch are provded n Table IV. 102
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 TABLE IV. RESULTS COMPARE THE ACCURACY OBTAINED OF RELAY ABB REL521 WITH ANFIS Fault tme ANFIS REL521 Actual Estmated Fault fault fault Error Error type locaton locaton [%] [%] [km] [km] 1/6/2009 ABG 29.36 29.46 0.36 2.24 16/10/2010 BG 27.4 22.95 1.38 3.97 2/8/2010 BG 35.9 36.08 0.03 0.12 12/8/2010 ABG 63.1 61.3 2.16 4.93 17/5/2011 BCG 25.4 27.55 1.38 1.20 20/5/2011 AG 81.8 83.22 0.02 1.68 19/8/2012 ABG 26.4 24.80 1.81 1.44 In Table IV, the maxmum devaton of the estmated dstance L e measured from the relay locaton and the actual fault locaton L f s calculated and the resultng estmated error Error s expressed as a percentage of total lne length L of that secton as: L f - Le % Error 100 (10) L Revews: When usng current and voltage from relay ABB REL521, the predcton capablty of Anfs s extremely good. Output of Anfs for ABG fault on 12/8/2010 s the hghest error. The estmated fault locaton s 61.3km as aganst the actual fault locaton 63.1km, thus t s located accurately wth the max error whch s 2.16% of the lne s length (lower than 4.93% of ABB REL521). Ths result may smply reflect the fact that the nput (voltage or current) was more drectly correlated wth the parameter beng predcted. It can be clearly seen from the test results that the proposed method, whch requres the same amount of measured data, has sgnfcantly outperformed the one-termnal method of ABB REL521. VI. CONCLUSTION Ths paper has presented an applcaton of Anfs for locatng faults n transmsson lne. The proposed method as an Anfs s traned to classfy the fault type, and separate Anfs are desgned to accurately locate the actual fault poston on a transmsson lne. The obtaned results show that the prmary advantages of the proposed algorthm can be summarzed n three aspects. Frstly, t does not depend on the effect of the errors n CT and VT sgnals, fault resstance... Secondly, the accuracy of the fault locaton does not rely on the accuracy of the algorthm type, as n the case of the one-termnal algorthm of ABB REL521. Thrdly, the method yelds accurate results; the errors n locatng the fault are from 0.02% to 2.16%. Anfs s easly traned n personal computer. The only dsadvantage of the method s that the obtaned accuracy s amount of practcal fault data n relay protecton and actual fault locaton, whch may depend on system operatng condtons. Nevertheless, ths ssue can be addressed by the current technology such as supervsory control and data acquston (SCADA), wde area montorng (WAM), and Automaton Substaton. Future work wll focus on onlne takng of the fault nformaton on relay protecton, and actual fault locaton n central control company, to obtan even better fault locaton accuracy. No Fault tme APPENDIX A. RESULTS DISTANCE TO FAULT COLLECTS ON THE RELAY AND ACTUAL LINE FROM YEAR 2008 TO 2013 220kV Hoa Khanh Substaton Estmated fault locaton on REL521[km] Actual fault locaton [km] Error [%] Estmated fault locaton on P437[km] 220kV Hue Substaton Actual fault locaton [km] Error [%] Reason/ Fault type 1 08/05/2008 22,33 22 0.40 - - - BG 2 01/06/2009 27.5 29.36 2.24 49.4 53.84 5.34 ABG 3 30/08/2009 6.9 7.9 1.20 64.4 75.3 13.1 BG 4 10/11/2009 18.3 can't fnd - 62 can't fnd - lghtnng 5 16/10/2010 24.1 27.4 3.97 48.3 55.8 9.01 BG 6 02/08/2010 35.8 35.9 0.12 46.6 47.3 0.84 BG 7 12/08/2010 67.2 63.1 4.93 14.7 20.1 6.49 ABG 8 16/12/2010 10.2 can't fnd - - - - lghtnng 9 17/5/2011 25.4 26.4 1.20 51 56.8 6.97 BCG 10 20/5/2011 83.2 81.8 1.68 0.9 1.4 0.60 AG 11 12/9/2011 38.4 39.4 1.20 - - - ABG 12 26/9/2011 52.8 can't fnd - 59.72 can't fnd - lghtnng 13 26/9/2011 52.8 can't fnd - - - - storm 14 19/8/2012 25.2 26.4 1.44 - - - ABG 15 30/09/2013 11.7 can't fnd - - - - storm ACKNOWLEDGMENT The authors acknowledged the actual data collecton on 220kV Hoa Khanh - Hue was supported by Power Transmsson Company No.2 n Vet Nam. REFERENCES [1] A. Yadav and A. S. Thoke, Transmsson lne fault dstance and drecton estmaton usng artfcal neural network, Internatonal Journal of Engneerng, Scence and Technology, vol. 3, no. 8, pp. 110-121, 2011. [2] Toshba, Instructon Manual Dstance Relay GRZ100, 2006. [3] Semens, Instructon Manual Numercal Dstance Protecton Relay Sprotec 7SA511, 1995. [4] Schwetzer Engneerng Laboratores, SEL-421 Relay Protecton and Automaton System, 2011. [5] Abb, REL 521 V2.3 Lne Dstance Protecton Termnal, 2003. 103
Journal of Automaton and Control Engneerng Vol. 3, No. 2, Aprl 2015 [6] Areva, Techncal Manual, Fast Multfuncton Dstance Protecton Relays P43x, 2010. [7] M. Jooraban and M. Monad, Anfs based fault locaton for EHV transmsson lnes, Aupec2005 Australa, 2005. [8] J. Sadeh and H. Afrad, A new and accurate fault locaton algorthm for combned transmsson lnes usng adaptve network-based fuzzy nference system, Electrc Power Systems Research, vol. 79, pp. 1538-1545, 2009. [9] A. A. Elbaset and T. Hyama, A novel ntegrated protectve scheme for transmsson lne usng ANFIS, n Internatonal Journal of Electrcal Power and Energy Systems, No.: IJEPES-D- 08-00112, 2009. [10] T. S. Kamel and M. A. M. Hassan, Adaptve neuro fuzzy nference system (ANFIS) for fault classfcaton n the transmsson lnes, The Onlne Journal on Electroncs and Electrcal Engneerng, vol. 2, no. 1, 2009. [11] Mathworks. [Onlne]. Avalable: http://www.mathworks.com/help/fuzzy/anfs-and-the-anfs- Edtor-Gu.Html. Le Km Hung receved the B.E. (1980) degree n electrcal engneerng from Da Nang Unversty of Technology, Vet Nam, and the M.E. (1991), D.E. (1995) degrees n electrcal engneerng from INPG, Grenoble, France. Currently, he s an Assocate Professor at the Electrcal Engneerng Department at Da Nang Unversty of Technology, Vet Nam. Hs research nterests nclude power system protecton and control. energy. Nguyen Hoang Vet receved the B.E. (1977) degree n electrcal engneerng from Ho Ch Mnh Cty Unversty of Technology, Vet Nam, and D.E. (1988) degrees n electrcal engneerng from Poltecncal Unversty, Lenngrag, Russa. Currently, he s an Assocate Professor at the Power System Department at Ho Ch Mnh Cty Unversty of Technology, Vet Nam. Hs research nterests nclude power system protecton and new Vu Phan Huan receved the B.E. (2002), M.E. (2009) degrees n Electrcal engneerng from Da Nang Unversty of Technology. He s 10 years wth Electrcal Testng Center n Relay protecton and Automaton substaton. He started n 2002 workng on the test of relay protecton. Snce 2009, he held the poston of a team leader n the feld of test relays n Central electrcal testng Company n Vet Nam. 104