Research on Fault Location in Distributed Network with DG Based on Complex Correlation Thevenin Equivalent and Strong Tracking Filter

Transcription

1 ensors & ransducers, Vol. 62, Issue, January 204, pp ensors & ransducers 204 by IFA Publishing,.. Research on Fault ocation in Distributed Networ with DG Based on Comple Correlation hevenin Equivalent and trong racing Filter Yanyan FENG, Zhongjian KANG College of Information and Control Engineering, China University of Petroleum, Qingdao , China el.: , fa: Received: 22 October 203 /Accepted: 9 January 204 /Published: 3 January 204 Abstract: With the introduction of distributed generators (DG), the traditional distribution system characterized by radical networ becomes a multi-source one. herefore the accuracy of the equivalent model of distributed generators directly affects the precision of fault location. o solve this problem, this paper proposed a fault location method based on comple correlation hevenin equivalent and strong tracing filter (F) in distribution networ with DG. trong tracing filter (F) can be used for real-time etraction of fundamental wave phase and amplitude of single phase voltage and current, and fast trac the power parameters mutation. he method of comple correlation hevenin equivalent taes into account the randomness of measurement data, DG impedance and load to overcome measurement error caused by the uncertainty of the data, so that the construction of the DG impedance model is more accurate. imulation results show that the method of fault location proposed in this paper has advantages of high accuracy and strong robustness. Copyright 204 IFA Publishing,.. Keywords: Distributed generator (DG), Comple correlation hevenin equivalent, trong tracing filter (F), Fault location, Impedance model.. Introduction Research on fault location has always been an important issue for the power system []. Due to the compleity of power distribution networ topology structure, prone to failure, the distribution networ fault diagnosis and repair at home and abroad has been a focus in the study of electrical worers [2-4]. With the introduction of the distributed generators (DG), the traditional radial distribution networ evolved into a multi-source one, which made the feeder protection and fault location analysis based on the feeder terminal unit (FU) of the distribution automation become more comple [5-8]. Fault section location is the basis of precise fault location. As a result of the DG access, fundamental changes have taen place in the structure of the traditional distribution networ, fault location methods must adapt to the change. In distributed networ with DG, multi-layer perception (MP) neural networ is used to determine the eact fault type and location first, with the normalization of main power source fault currents, then input to the trained neural networ to get the eact fault location, which is proposed in literature [9]. he article [0] mainly studies the effects of DG on automatic 52 Article number P_763

2 ensors & ransducers, Vol. 62, Issue, January 204, pp reclosing switch and fuse coordinated ability in the distribution system. It uses the classification techniques to classify them under the condition of fault, and according to the current direction and the amplitude value that FU monitored at the point DG access to distribution networ to determine fault section. he article [] proposed a new automatic fault localization method based on radial basis function neural networ (RBFNN) in distribution networ with DG, this method using RBF neural networ can accurately determine the fault type and fault location. iterature [2] proposed a multi-agent based fault diagnosis method in distribution networ with DG, distribution networ is divided into several parts by using the relay agent, and measure bus current, the fault type is detected and fault location is determined. he monotonicity and continuity of the short-circuit current produced by each power source has nothing to do with other DGs, on the basis of this feature, a new fault line locating method is proposed, which includes the formation of the fault matching table by pre-fault off-line calculation and matching process. his is studied in the article [3]. However, the above method requires a specific topology of the distribution networ with DG, and in some operation mode, considerable pre-accident simulation needs to be done. When distribution structure or operation mode changes, it will not apply, the versatility and practicality of this method needs to be improved. Above methods all have certain disadvantages, this paper puts forward a fault locating method based on the comple correlation hevenin equivalent and strong tracing filter (F) in distribution networ with DG. his method put forward the definition of fault characteristic value, built the three phase impedance model of distribution networ with DG,analyze fault characteristic value of each node, and search for the node whose fault characteristic value is smallest, so as to decide the fault section, and tested by simulations. 2. Real-time Etraction of ingle-phase Voltage Fundamental Phase Based on F 2.. Comple tate pace Description of Voltage ignal Discrete form of single phase voltage signal can be epressed as N sin,,2, () y A v N n n s n n where An and n is the n-th sinusoidal component amplitude and initial phase angle in single-phase voltage signal respectively; n is the frequency of the n-th sinusoidal component; s is the sampling time interval, v is the white Gaussian noise added to the single phase voltage. Assumes that the single phase voltage signal contains only fundamental component, Equation () can be simplified as follows: y Asin( s ) j( s) j( s) ( 0.5 j)( Ae ) (0.5 j)( Ae ) (2) where is the fundamental frequency, A is the fundamental amplitude, is the fundamental initial phase. he state variable of voltage signal is defined as js j( s ) 2 Ae j( s ) 3 Ae e, (3) herefore, the nonlinear state space description of voltage signal is given by equation (4). f( ) w y H v, (4) where ; f ; H 0 0.5j 0.5j. Ecitation process noise w and measurement noise v is not related and zero-mean Gaussian white noise, and they satisfies the condition as follows. E wwj Q (, j),, j 0, (5) E v v R (, j) j E is the mathematical epectation; Q is where the covariance matri of process noise w ; R is the covariance matri of measurement noise v. (, j) is the Kronnecer delta function, its definition is given as follows. 0, j (, j), (6), j When voltage signal contains harmonic components, state variables can be epanded. For eample, if voltage signal contains the fifth harmonic, formula (7) can be added to the state variables defined in formula (3). j 5s 5 4 Ae 5 j 5s 5, (7) 5 Ae 5 53

3 ensors & ransducers, Vol. 62, Issue, January 204, pp tate Variables Estimation with F F can recursively estimate the state variables defined in formula (4), the process is shown below. ) Prediction stage. his step is to calculate the value of the state prediction and covariance matri of the prediction error. he value of the state prediction is f ( ), (8) he covariance matri of the prediction error is P F P F Q, (9) where 0 0 f( ) F is the suboptimal fading factor. It is calculated as follows: where, tr N, 0 0 0, 0 tr M N V HQ H R M HF P F H, (0), () tris the matri trace operator, is the V selected weaening factor, is the covariance matri of the actual output residuals, it is unnown in practice, and can be estimated by equation (2). V,, (2), V where 0 is the forgetting factor, and generally ) Updating stage. his step is to calculate the constructed gain matri, and update the state estimation value and the covariance matri of estimation error. he gain matri is K P H HP H R, (3) he state estimation value is K, (4) where actual output residuals is ( y y ), y H R. he covariance matri of estimation error is P ( I K H) P, (5) he state variable is recursively estimated with F, by formula (6) formula (8) the fundamental frequency, amplitude and phase at ( ) s can be obtained [4]. Fig. is the flowchart of the Fbased fundamental phase real-time etracting method for single-phase grid voltage. Read single -phase voltage simulation data tate prediction value Initialization Calculate the output residual =+ Calculate suboptimal fading factor Calculate the covariance matri of the prediction error P Calculate Kalman gain matri K tate estimation value Formula (6-8) Calculate the covariance matri of the estimation error P Fundamental amplitude and phase Fig.. he flowchart of the F-based fundamental phase real-time etracting method for single-phase grid voltage. 54

4 ensors & ransducers, Vol. 62, Issue, January 204, pp f Im(ln( )), (6) 2 s A, (7) 2 2 Im ln( ) 2 ( ) 3. Comple Correlation hevenin Equivalent Model of Main ource and DG, (8) By the hevenin's theorem, the impedance model of main feeder source and DG can respectively be equivalent to an ideal voltage source with an internal resistance. Assuming that each phase current and voltage of the main feeder source and DG can be measured simultaneously (this is feasible under the condition of eisting technology), then the positive sequence, negative sequence and zero sequence of them can be measured before fault and during fault. he modeling process of main feeder source and DG impedance model is the same, tae some DG as an eample to describe its modeling process in detail: According to the asymmetry theory of electrical networ, DG can be equivalent to positive sequence, negative sequence and zero sequence networs shown in Fig. 2 Fig. 2. DG equivalent sequence networ diagram. In Fig. 2, the positive sequence impedance of the load impedance is constant, then the positive sequence voltage equations of the hevenin equivalent circuit can be rewritten as follows: V EIZ, V IZ Z (9) where V is the positive sequence voltage of access point; I represents sequence current to the positive sequence load; Z is the equivalent positive sequence impedance of DG; E is the equivalent positive sequence voltage of DG; Z represents the positive sequence impedance of the load. Adding a small signal variation to voltage equation (9), it can be epressed as follows: or V V ( EE) ( I I)( Z Z) V V ( I I)( Z Z ) V V Z I, 2 (20) I I E ( E V ) Z V I, (2) 2 I I I Multiply (2) by (20) and tae the average over a large number of measured voltages and currents, equation (20) can be epressed as: I R EV R V I VV II V R I VI VI R I, (22) where Ry is the covariance of variables and y. he covariance of two variables, and y, provides a statistical measure of how strongly correlated these variables are. he definition of the covariance of comple variables and y, R is given as follows: u where i i and Z y Ry ( u )( yuy ) y i iuu y, and uy yi is the mean value of y i. Note that the impedance of the load side,, has no statistical correlation with the impedance of source side, Z. In other words, the impedance of the load size is statistically independent from the impedance of the source side.hence, the covariance of variables Z and Z is zero ( RZ 0 Z ). ame reason can be applied to source voltage, E, and load impedance, Z.he covariance of variable E and Z is zero( R 0 ). From (9) and (22), the equivalent positive sequence impedance of the system, Z is Z IR VR EZ VR VV VI II IRVI, (23) where I represents the comple conjugate of I. ince each parameter type is comple, so the result 55

5 ensors & ransducers, Vol. 62, Issue, January 204, pp Z is a comple number. he positive sequence current and voltage data in Formula (23) is short circuit data. Comple correlation hevenin equivalent method can be applied to calculate for the positive, negative and zero sequence impedance of DG, then impedance model of DG can be obtained. imilarly, comple correlation hevenin equivalent method can be used to obtain the impedance model of the main source. he specific flowchart is shown in Fig. 3. RVV, RVI, RII Fig. 3. he flowchart of building DG impedance model based on comple correlation hevenin equivalent and F. 4. Fault ocation Method Based on Comple Correlation hevenin Equivalent and F 4.. Fault Feature Etraction Assuming that the DGs are located in nodes Bus (), Bus (2)... Bus (m), the falut occurred in noad j and the voltage and the injection current of each power supply can be synchronously measured. he measured voltage and current signals before the fault are VBU (), VBU (2) VBU ( m) and IBU (), IBU (2) IBU ( m) ; he measured voltage and current signals when the ' ' ' fault occurred are V (), V (2) V ( m) and ' ' ' BU BU BU BU BU BU I (), I (2) I ( m). he voltage of the fault point before fault, noted as oc V j, can be calculated by the system node fault equation. When the fault occurred in node j, he fault current can be calculated as equation (24) by the fault voltage component in source measurement point. ' fj ()= (,)(V BU ()-V BU ())(=,2,,, ) I i Y i j i i i m, (24) he fault current generated by all power supplies is I j j R V, (25) =(Z(, )+ ) - oc fj0 g j Here, m = Rg oc V = Z( j, ) I ( ). j is the fault grounding resistance, BU he error between the fault current of the fault phase calculated by each power supply and the fault current of the fault phase calculated by all power supplies is defined as fault characteristic value. Its mathematical epression is as following. m Ej ()= I ()- i I fj fj0, (26) i= 4.2. Fault ocation Method Based on F and Comple Correlation hevenin Equivalent According to the electrical networ theory, the fault current at fault point generated by the common interaction of power sources is equal to that calculated by fault voltage component at measurement points. According to the theory, the fault eigenvalue at fault point in distribution networ with DG is 0. o a node with smallest eigenvalue is the associated node of fault branch. Differential evolution algorithm is used to obtain precise fault distance after the fault section is identified. he specific flowchart of fault location method based on F and comple correlation hevenin equivalent is shown in Fig.4. 56

6 ensors & ransducers, Vol. 62, Issue, January 204, pp I I ( ) fi fi0 j e rr( j) norm I ( j) I fi0 fi Fig. 4. he flowchart of fault location method based on F and comple correlation hevenin equivalent. 5. imulations in Different Fault Conditions 5.. he imulation Model he method proposed in this paper is tested in an actual 6 KV substation AiDing in XinJiang. he distribution networ simulation system model is built in Matlab/imulin, which is shown as Fig. 5. he total load is 3.6 MVA, the line number is shown in able, the DG unit capacity and access point is shown in able 2, all the DG units account for 47 % of the total load imulation Results he precision of the DG impedance model built by the method proposed in this paper is evaluated by the ranging results based on three-phase impedance model and differential evolution algorithm, and compared with that built by traditional hevenin equivalent, while the model of other components (transformer, feeder, load and so on ) in distribution networ with DG is the same. 57

7 ensors & ransducers, Vol. 62, Issue, January 204, pp able. ine number. ine tart End tart End tart End ine ine point point point point point point able 2.he capacity and access point of DG unit. Name Access Capacity Access Capacity Name point /MVA point /MVA DG DG DG DG DG Fig. 5. he topology of simulation model. Under different operation modes, set different fault types, and fault location is carried out on the basis of the method shown in Fig. 3, the results are shown in able 3 (length is limited, only part of the fault location results is listed). Note that, in able 3, Ranging Result A means that the result is obtained by the method proposed in this paper, Ranging Result B means that the result is obtained by traditional hevenin equivalent. From the ranging results A shown in able 3, it is more accurate than that by traditional hevenin equivalent. In other words, the precision of DG impedance model based on F and comple correlation hevenin equivalent is better than traditional methods. imulation results show that the method proposed in this paper is effective. 58

8 ensors & ransducers, Vol. 62, Issue, January 204, pp able 3. he contrast of the fault location result. Fault Ranging Ranging Operation Mode Fault ype Fault ine Distance Result A Result B Direct Grounding ingle phase Direct Grounding ingle phase Direct Grounding ingle phase Direct Grounding ingle phase Direct Grounding ingle phase Direct Grounding wo phase Direct Grounding wo phase Direct Grounding wo phase Direct Grounding wo phase Direct Grounding wo phase Direct Grounding hree phase Direct Grounding hree phase Direct Grounding hree phase Direct Grounding hree phase Direct Grounding hree phase Non-grounding wo phase Non-grounding wo phase Non-grounding wo phase Non-grounding wo phase Non-grounding wo phase Non-grounding hree phase Non-grounding hree phase Non-grounding hree phase Non-grounding hree phase Non-grounding hree phase By arc-suppressing coil wo phase By arc-suppressing coil wo phase By arc-suppressing coil wo phase By arc-suppressing coil wo phase By arc-suppressing coil wo phase By arc-suppressing coil hree phase By arc-suppressing coil hree phase By arc-suppressing coil hree phase By arc-suppressing coil hree phase By arc-suppressing coil hree phase Conclusions he fault occurring in distribution networ with DG can affect the system greatly. his paper proposed a fault location method based on comple correlation hevenin equivalent and F in distribution networ with DG. F is used to etract the fundamental amplitude and phase of singlevoltage real-time. On this basis, comple correlation hevenin equivalent is applied to the establishment of DG impedance model. he simulation results show that this method using the correlation between multiple measurement information can overcome the error from the randomness of measurement information. Compared with the traditional hevenin equivalent, it has a higher accuracy and wider application prospect. Acnowledgements his paper is sponsored by the National Nature cience Fund Project of China (62700). he authors are grateful for all the reviewers for valuable suggestions to improve the quality of this paper. References []. Zhou Xiaoin, o develop power system technology suitable to the need in 2 century, Power ystem echnology, Vol. 2, Issue, 997, pp

9 ensors & ransducers, Vol. 62, Issue, January 204, pp [2]. Xu Hao, Miao hihong, Jiang Zhen, Zeng Fei, Zhang ei, iu Pei, A new fault location algorithm based on fault component from finite synchronized phasor measurement unit, Automation of Electric Power ystems, No. 36, 202, pp. -6. [3]. Wang in-chuan, i Qing-Xin, iu Xin-Quan, Zhang Wei, Pan Wen-Ming, Distribution networ fault location based on the improved ant colony algorithm, Power ystem Protection and Control, Vol. 36, Issue 22, 2008, pp [4]. u ing-min, Peng Min-Fang, Wang Yao-Nan, Distribution networ fault location based on data of sensor FU, ransducer and Microsystem echnologies, Vol. 3, Issue, 202, pp [5]. V. Calderaro, A. Piccolo, V. Galdi, P. iano. Identifying fault location in distribution systems with high distributed generation penetration, in Proceedings of the IEEE Annual African Conference (AFRICON), 2009, pp. -6. [6]. un Ming, Wang ei, Wang Zhigu, plitting and paralleling research of the distribution system which contain distributed generation (DG) under the power system faults, in Proceedings of the International Conference (CICED), echnical ession 4, 2008, pp. -4. [7]. in Xia, u Yuping, Wang ianhe, New fault region location scheme in distribution system with DGs, ransactions of China Electrotechnical ociety, Vol. 23, Issue, 2008, pp [8].. M. Debritto, D. R. Morais, M. A. Marin, Distributed generation impacts on the coordination of protection systems in distribution networs, in Proceedings of the IEEE/PE ransmission and Distribution Conference and Eposition, 2004, pp [9].. A. M. Javadian, A. M. Nasrabadi, M.-R. Haghifam, J. Rezvantalab, Determining fault's type and accurate location in distribution systems with DG using MP neural networs, in Proceedings of the International Conference on Clean Electrical Power, 2009, pp [0]. A. F. Naiem, Y. Hegazy, A. Y. Abdelaziz, M. A. A. Elsharawy, Classification technique for recloser-fuse coordination in distribution systems with distributed generation, IEEE ransactions on Power Delivery, Vol. 27, Issue, 20, pp []. Zayandehroodi Hadi, Mohamed Azah, hareef Hussain, Mohammadjafari Marjan, Determining eact fault location in a distribution networ in presence of DGs using RBF neural networs, in Proceedings of the IEEE International Conference on Information Reuse and Integration (IRI), 20, pp [2]. A. M. El-Zonoly, Fault diagnosis in distribution networs with distributed generation, Electric Power ystems Research, Vol. 8, Issue 7, 20, pp [3]. Ma Jinjie, u Yuping, Du Jiao, in Xia. A new fault location scheme based on distributed short-circuit current in distribution system with DGs, in Proceedings of the IEEE International Conference on ustainable Energy echnologies, 2008, pp [4]. K. P. Dash, K. A. Pradhan, G. Panda, An etended comple Kalman filter for frequency estimation of distorted signals, IEEE ransactions on Instrumentation and Measurement, Vol. 49, Issue 9, 2000, pp Copyright, International Frequency ensor Association (IFA) Publishing,.. All rights reserved. ( 60