Fault Diagnosis of Metro Traction Power Systems Using A Modified Fuzzy Reasoning Spiking Neural P System

ROMANIAN JOURNAL OF INFORMATION SCIENCE AND TECHNOLOGY Volume 18, Number 3, 2015, 256–272 Fault Diagnosis of Metro Traction Power Systems Using A Modified Fuzzy Reasoning Spiking Neural P System Yangyang HE1, Tao WANG1, Kang HUANG 1, Gexiang X. ZHANG1, Mario J. P ´ EREZ-JIM ´ ENEZ2 1School of Electrical Engineering, Southwest Jiaotong University, Chengdu, 610031, P.R. China E-mail: [email protected], [email protected], [email protected], [email protected] 2Research Group on Natural Computing Department of Computer Science and Artificial Intelligence University of Sevilla, Sevilla, 41012, Spain E-mail: [email protected] Abstract. This paper presents the application of a modified fuzzy reasoning spiking neural P systems (MFRSN P system, for short) to fault diagnosis of metro traction power supply systems. In MFRSN P systems, three types of neurons are used to represent operation information of protection devices including protective relays and circuit breakers; a reasoning algorithm associated with MFRSN P systems is introduced to fulfill fault reasoning; fault diagnosis rules for metro traction power supply systems and their MFRSN P systems are described. Case studies show the feasibility and effectiveness of the presented method. Key-words: Membrane computing, probabilistic fuzzy reasoning spiking neural P system, metro traction power supply system, fault diagnosis. 1. Introduction Membrane computing, formally introduced by Gh. P˘ aun in [1], is an attractive research field of computer science aiming at abstracting computing models, called membrane systems Fault Diagnosis of Metro Traction Power Systems Using A MFRSN P system 257 or P systems, from the structures and functioning of living cells, as well as from the way cells are organized in tissues or higher order structures [2, 3]. As a main type of P systems, a spiking neural P system (SN P system) is a type of P system inspired by the neurophysiological behavior of neurons sending electrical impulses (spikes) along axons from presynaptic neurons to postsynaptic neurons in a distributed and parallel manner [4]. Recently, SN P systems have become a hot topic in membrane computing [7–29], among which there are several investigations focus on the use of SN P systems and their variants to solve fault diagnosis problems [14–22]. An overview of different types of fuzzy reasoning spiking neural P systems (FRSN P systems), differences between FRSN P systems and SN P systems and newly obtained results on these FRSN P systems in solving fault diagnosis problems can be found in [14]. In [15], a fuzzy reasoning spiking neural P system with real numbers (rFRSN P system) was presented to fulfill diagnosis knowledge representation and reasoning and then the rFRSN P system was used for fault diagnosis of a transformer. In [16] and [17], the rFRSN P system was used for fault diagnosis in power systems and several different applications verified its effectiveness. In [18], the rFRSN P system was used to perform fault diagnosis of electric locomotive systems. In [19], an adaptive fuzzy spiking neural P system (AFSN P system) was proposed to fulfill fault diagnosis of a local power system with four cases. Furthermore, a fuzzy reasoning spiking neural P system with trapezoidal fuzzy numbers (tFRSN P system) was proposed in [20] and the fault diagnosis of power transmission networks based on tFRSN P systems was investigated in [20], [21]. Besides, in [22], a weighted fuzzy reasoning spiking neural P system (WFSN P system) was proposed to diagnose faults occurring in a traction power supply system of high-speed railways with three cases. Metro traction power supply systems (MTPSSs, for short) are a kind of special distribution networks with direct current (DC) power supply. As usual, the fault diagnosis of a MTPSS is quite difficult and challenging, due to its two-way feeding power supply approach and a complex protection system caused by the DC power supply mode. FRSN P systems are a class of distributed and parallel computing models with good understandability and dynamics [14–22]. The fault occurrence in power systems is a discrete and dynamical process [20]. From the aforementioned studies [14–22], FRSN P systems are feasible and effective in solving fault diagnosis problems of different kinds of power systems as well as show great potential. Thus, this study proposes to use FRSN P systems to diagnose the faults occurring in metro traction power supply systems. In the rFRSN P systems proposed in [15], new ingredients such as fuzzy truth value, new firing rule, pulse value, proposition neuron and rule neuron were added to the original definition of SN P systems. However, two kinds of rule neurons in rFRSN P systems are not enough to well represent status information of protective devices. To use rFRSN P systems to solve fault diagnosis problems in MTPSSs, a modified fuzzy reasoning spiking neural P system (MFRSN P system) is discussed by considering three kinds of rule neurons, a reasoning algorithm associated with MFRSN P systems, fault diagnosis rules for MTPSSs and corresponding MFRSN P systems. The parameter setting with respect to MFRSN P systems is also discussed. Case studies show the effectiveness of the presented method. The remainder of this paper is organized as follows. Section 2 states the problem to solve. The definition of MFRSN P systems and their reasoning algorithm as well as the MFRSN P system models for fault diagnosis rules of MTPSSs are presented in Section 3. In Section 4, 258 Y.Y. He et al. case studies on fault diagnosis of a MTPSS are used to test the effectiveness of MFRSN P systems. Finally, conclusions are drawn in Section 5. 2. Problem Description Metros are important parts of public transport systems. Traction power supply systems are the energy systems of rail transportation systems and have a great significance to ensure the safety and reliable operations of trains. However, impacted by various factors such as equipment failures, interruption of power service may happen in traction power supply systems of metros. Thus, when a fault occurs in a metro traction power supply system, it is crucial to design a method to help dispatchers judge where the faults are and which sections fail so as to restore power supply of the metro as soon as possible. Thus, this study focuses on the application of FRSN P systems to fault diagnosis of metro traction power supply systems. For the convenience of description, fault diagnosis of metro traction power supply systems is abbreviated by FDMTPSS. The framework of fault diagnosis in power systems using reasoning model-based method is shown in Fig. 1 [20]. Metro traction power supply systems (MTPSSs) are a special kind of power systems. Thus, FDMTPSS can be considered in the framework in Fig. 1. In this paper, a modified fuzzy reasoning spiking neural P systems with real numbers (MFRSN P systems) is used for FDMTPSS within this framework. The aim of fault diagnosis is to identify the faulty sections by using status information of protection devices (protective relays and CBs) which are read from SCADA systems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ig. 1. Framework of fault diagnosis in power systems using reasoning model-based method. Fault Diagnosis of Metro Traction Power Systems Using A MFRSN P system 259 2.1. Metro Traction Power Supply Systems To easily understand FDMTPSS based on MFRSN P systems, this section describes basic components and operational principles of metro traction power systems. A metro traction power system consists of two parts: (1) metro traction substations; (2) overhead contact systems (OCS), which are usually set up along with the metro route. We will introduce them in the following. Figure 2 shows metro traction substation electrical connection relationship, where three traction substations (identified by 1, 2 and 3) form a traction power supply section. Traction substation 1 receives 35 kV alternating current (35 kV AC) from the main substation, then feeds 35 kV AC power to traction substation 2. Traction substation 2 receives 35 kV AC power from traction substation 1 and then feeds 35 kV AC power to traction substation 3. The electrical principle schematic illustration of metro traction substation is shown in Fig. 3, where metro traction substations use single-bus sections which can be partitioned into two parts: Bus I and Bus II. Bus I and Bus II access different powers, respectively. Two traction rectifier transformers are connected to Bus II. These transformers receive three phases 35 kV AC power from section Bus II and then step-down and rectification is down to convert 35 kV AC power to 1500 V DC power. The OCSs are connected to the 1500 V DC Bus. It is worth pointing out that metro traction power systems adopt two-way feeding. So OCSs should be accessed to a left traction substation and a right traction substation at the same time. Traction substation 1 Traction substation 2 Traction substation 3 110kV/35kV L 1 L 2 L 3 L 4 L 5 L 6 CB 1 CB 7 CB 2 CB 8 CB 3 CB 9 CB 4 CB 5 CB 6 CB 10 CB 11 CB 12 CB 13 CB 14 CB 15 CB 16 CB 17 CB 18 B ĉ B Ċ B ĉ B ĉ B Ċ B Ċ B ĉ B Ċ Fig. 2. Metro traction substation electrical connection relationship. 2.2. Protection Devices of Metro Traction Power Supply Systems The fault diagnosis of FDMTPSS in this paper is based on SCADA data (statuses of protective relays and CBs). According to SCADA data, fault diagnosis models based on MFRSN P systems reason out fault confidences of candidate fault sections. So it is necessary to describe operational rules of protective relays and CBs. In this study, the protective relays consist of main protective relays (MPRs), first backup protective relays (FBPRs) and second backup protective relays (SBPRs). It is worth pointing out that there is no FBPR for buses. When a section has a fault, MPRs of this section operate immediately to trip their associated CBs. If these MPRs fail to operate or their associated CBs 260 Y.Y. He et al. do not trip, then FBPRs of this section operate to trip their associated CBs. If both MPRs and FBPRs of this section fail to operate, then SBPRs of this section operate to trip the CBs of its adjacent section. It is worth pointing out that the MPRs and FBPRs control the operation of CBs associated with the faulty sections while SBPRs control the operation of CBs associated with adjacent sections of the faulty sections. We will take some examples to illustrate operation rules of aforementioned protection devices. (1) If L3in Fig. 2 has a fault, then MPRs of L3will operate to trip CB3and CB4to protect L3. If MPRs of L3fail to operate, then FBPRs will operate to trip CB3to protect L3. If MPRs and FBPRs fail to operate or the CB3does not trip, then SBPRs of L3operate to trip CB1to protect L3. (2) If the DC 1500 V Bus in Fig. 4 has a fault, then the MPRs of the DC 1500 V Bus will operate to trip CB1,CB7and CB9to protect this Bus. (3) For each level of protective relays for a transformer, if a protective relay operates, then it will trip the CBs of the transformer on both its primary winding and secondary winding. If T1in Fig. 3 has a fault, then its MPRs or FBPRs or SBPRs will operate to trip CB5 and CB9. 35kV 1 # Input 35kV 1 # Output 0.4kV Busĉ0.4kV DC1500V 35kV 2 # Input 35kV 2 # Output BusĊCB 9 CB 1 CB 2 CB 3 CB 4 CB 5 CB 6 CB 7 CB 8 CB 11~14 CB 10 CB 15 T 1 T 2 T 3 T 4 Fig. 3. Schematic illustration for traction substation of metro. A metro traction network has transfer inter-trip (TIP) protection devices (main protection devices) because its power supply technique is two-way feeding. When a TIP protection device of a DC feeder line detects a tripping single of a CB associated with this TIP protection device, it will operate to trip all its associated CBs. For example, if the protective relays associated with CB1in Fig. 4 operate to trip CB1, then the TIP protection device T IP1of OCS will operate to trip CB2. Fault Diagnosis of Metro Traction Power Systems Using A MFRSN P system 261 OCS + _ DC1500V DC1500V Rail + _ CB 3 CB 2 T 1 T 2 T 3 T 4 CB 5 CB 4 CB 6 CB 7 ~CB 10 CB 11 ~CB 14 Traction subtation A Traction substation B CB 1 Fig. 4. Metro two-way feeding diagram 3. MFRSN P systems In this section, a modified fuzzy reasoning spiking neural P systems (MFRSN P systems, for short) are first presented, and then a reasoning algorithm for MFRSN P systems is described. Finally, fault diagnosis rules for MTPSS and a MFRSN P system model are discussed. 3.1. MFRSN P systems On the basis of the work in [15] and [22], we introduce a MFRSN P system, which considers one more kind of rule neurons, general rule neurons, compared with rFRSN P systems in [15], and does not consider weighted synapses compared with WFRSN P systems in [22]. Definition 1. AMFRSN P system of degree mis a tuple Π = (O, σ1, ..., σm, syn, in, out), where: (1) O={a}is a singleton alphabet (ais called spike); (2) σ1, . . . , σmare neurons such that σi= (θi, ci, ri, λi),1≤i≤m, where: (a) θiis a real number in [0, 1] representing the potential value of spikes contained in neuron σi; (b) ciis a real number in [0, 1] representing the truth value of neuron σi; (c) rirepresents a firing (spiking) rule associated with neuron σiof the form E/aθ→ aβ, where θand βare real numbers in [0,1], E={aθ, θ > λi}represents the firing condition. The firing condition means that if and only if neuron σireceives at least nspikes and θ > λi, then the firing rule contained in the neuron can be 262 Y.Y. He et al. applied; otherwise, the firing rule cannot be applied. If the firing rule is applied, then it means that neuron σiwill consume a spike awith potential value θand then send out a spike awith potential value β; (d) λiis a real number in [0,1) representing the firing threshold of neuron σi; (3) syn ⊆ {1, ..., m}×{1, ..., m}with i6=jfor all (i, j)∈syn,1≤i, j ≤m; that is, syn is a directed graph of synapses that provides the links between neurons; (4) in, out ⊆ {1, . . . , m}indicate the input neuron set and the output neuron set of Π, respectively. MFRSN P systems include two types of neurons: proposition neurons and rule neurons. For both proposition neurons and rule neurons, their firing rules have an identical form and there is only one firing rule in each neuron. According to different computing methods of spikes, rule neurons are divided into three different kinds: General rule neurons, And rule neurons and Or rule neurons, which are shown in Figure 5 (a), (b) and (c), respectively.                                                         Fig. 5. Rule neurons. (a) General rule neurons; (b) And rule neurons; (c) Or rule neurons. 3.2. Fault diagnosis rules for MTPSSs and their MFRSN P systems In this subsection, we will first describe three different kinds of fault diagnosis rules for MTPSS and then we propose their MFRSN P system models, which are shown in Fig. 6. (1) General Rule Ri: IF pj(θj)THEN pk(θk)(CF = ci), where pjand pkare propositions, ciis a real number in [0,1] representing the certainty factor of such rule Ri,θjand θkare real numbers in [0,1] representing the truth values of pjand pk, respectively. The truth value of pkis θk=θj∗ci. (2) And Rule Ri: IF p1(θ1)and . . . and pk−1(θk−1)THEN pk(θk)(CF = ci), where p1, . . . , pkare propositions, Ciis a real number in [0,1] representing the certainty factor of such rule Ri,θ1, . . . , θkare real numbers in [0,1] representing the truth values of p1, . . . , pk, respectively. The truth value of pkis θk= [(θ1+. . . +θk−1)/(k−1)] ∗ci. (3) Or Rule Ri: IF p1(θ1)or . . . or pk−1(θk−1)THEN pk(θk)(CF = ci), where p1, . . . , pk are propositions, ciis a real number in [0, 1] representing the certainty factor of such rule Ri,θ1, . . . , θkare real numbers in [0, 1] representing the truth values of p1, . . . , pk, respectively. The truth value of pkis θk=max{θ1∗ci, . . . , θk−1∗ci}. Fault Diagnosis of Metro Traction Power Systems Using A MFRSN P system 263                                                                  Fig. 6. Diagnosis models for MTPSS based on MFRSN P systems. (a) General Rules; (b) And Rules; (c) Or Rules 3.3. Reasoning algorithm To make MFRSN P systems have better knowledge representation and reasoning ability, a reasoning algorithm for MFRSN P systems is introduced in this subsection. In order to clearly describe the reasoning algorithm, we will introduce some parameter vectors and matrices as follows where srepresents the number of proposition neurons, trepresents the number of rule neurons and s+t=m. (1) Bp= (bji)s×t. If there is a directed arc from rule neuron σito proposition neuron σj, then bji = 1; otherwise, bji = 0; (2) B1= (bij)t×s. If there is a directed arc from proposition neuron σjto General rule neuron σi, then bij = 1; otherwise, bij = 0; (3) B2= (bij)t×s. If there is a directed arc from proposition neuron σjto And rule neuron σi, then bij = 1; otherwise, bij = 0; (4) B3= (bij)t×s. If there is a directed arc from proposition neuron σjto Or rule neuron σi, then bij = 1; otherwise, bij = 0; (5) θ= (θ1, ..., θs)T.θi(1≤i≤n) is a real number in [0, 1] representing the potential value of spikes in proposition neurons σ1, . . . , σn; (6) δ= (δ1, ..., δt)T.δi(1≤i≤n) is a real number in [0, 1] representing the potential value of spikes in rule neurons σ1, . . . , σn; (7) C=diag(c1, ..., ct).ci(1≤i≤n) is a real number in [0, 1] representing the certainty factor of rule neurons σ1, . . . , σn; (8) λp= (λp1, ..., λps)T.λpi (1≤i≤n) is a real number in [0, 1] representing the firing threshold of proposition neuron σi; (9) λr= (λr1, ..., λrt)T.λri (1≤i≤m) is a real number in [0, 1] representing the firing threshold of rule neuron σi. The reasoning algorithm contains three types of multiplication operations as follows: 264 Y.Y. He et al. (1) ⊗:(bij)t×s⊗(θ1, ..., θs)T= (a1, ..., at)T, where ai=bi1∗θ1+... +bis ∗θs; (2) ⊕:(bij)t×s⊕(θ1, ..., θs)T= (a1, ..., at)T, where ai=bi1∗θ1+...+bis ∗θs bi1+...+bis ; (3) :(bij)t×s(θ1, ..., θs)T= (a1, ..., as)T, where ai=max{bi1∗θ1, ..., bis ∗θs}. Now the reasoning algorithm is presented as follows. Step 1: Initialization parameters. Set zero-matrix Otermination condition. According to MFRSN P system fault diagnosis models rules for MTPSS, information of protection devices from SCADA systems and Table 1, vectors θ0,δ0,λp,λrand matrices Bp,B1,B2,B3,C are initialized; Step 2: Set g= 0, which presents the reasoning step; Step 3: If the firing condition of a proposition neuron is satisfied and there is a postsynaptic neuron, then the proposition neuron will transmit a new spike to the next rule neuron through directed arc; Step 4: According to fault diagnosis rules for MTPSS, rule neurons compute the vector δg+1 according to δg+1 = (B1⊗θg)+(B2⊕θg)+(B3θg); Step 5: If δg+1 =O, then the reasoning algorithm stops and outputs the reasoning results; Step 6: g=g+ 1; Step 7: The new spikes are transmitted to the next proposition neurons through directed arc and the pulse value of each spike is equal to θg+1, where θg+1 =Bp(C⊗δg); Step 8: Go to Step 3. 4. Case Studies 4.1. Parameter setting The operation information of protective relays and CBs may contain uncertainty or incompleteness. So it is necessary to use a parameter to describe the accuracy of the information and confidence levels of protection devices. Table 1 shows the confidence levels of operated and non-operated feeder lines, buses and transformers [22]. According to the protection devices operation information from SCADA systems and confidence levels for sections in Table 1, we can initialize the input spikes of MFRSN P system models. It is worth pointing out that both main protective relays (MPRs) and first backup protective relays (FBPRs) control the same CBs. Thus, if both MPRs and FBPRs operate, then the values of the CBs is set according to their corresponding MPRs. Table 1. Operation and non-operation confidence levels of the protective devices Sections Protective devices (operated) Protective devices (non-operated) Main Primary backup Remote backup Main Primary backup Remote backup Relays CBs Relays CBs Relays CBs Relays CBs Relays CBs Relays CBs FL 0.9913 0.9833 0.8 0.85 0.7 0.75 0.2 0.2 0.2 0.2 0.2 0.2 B0.8564 0.9833 - - 0.7 0.75 0.4 0.2 - - 0.4 0.2 T0.7756 0.9833 0.75 0.8 0.7 0.75 0.4 0.2 0.4 0.2 0.4 0.2 Fault Diagnosis of Metro Traction Power Systems Using A MFRSN P system 271 [12] WANG J., ZOU L., PENG H., ZHANG G.X., An extended spiking neural P systems for fuzzy knowledge representation, International Journal of Innovative Computing, Information & Control, 7(7), pp. 3709–3724, 2011. [13] WANG J., PENG H., Adaptive fuzzy spiking neural P systems for fuzzy inference and learning, International Journal of Computer Mathematics, 90(4), pp. 857–868, 2013. [14] WANG T., ZHANG G.X., P ´ EREZ-JIM ´ ENEZ M.J., Fuzzy membrane computing: theory and applications, International Journal of Computers Communications & Control, 10(6), pp. 861–892, 2015. [15] PENG H., WANG J., P ´ EREZ-JIM ´ ENEZ M.J., WANG H., SHAO J., WANG T., Fuzzy reasoning spiking neural P system for fault diagnosis, Information Sciences, 235, pp. 106–116, 2013. [16] XIONG G.J., SHI D.Y., CHEN J.F., Implementing fuzzy reasoning spiking neural P system for fault diagnosis of power systems, General Meeting of the IEEE-Power-and-Energy-Society (PES), Vancouver, Canada, Article ID 5970635, 5 pages, 2013. [17] XIONG G.J., SHI D.Y., ZHU L., DUAN X.Z., A new approach to fault diagnosis of power systems using fuzzy reasoning spiking neural P systems, Mathematical Problems in Engineering, Article ID 815352, 13 pages, 2013. [18] WANG T., ZHANG G.X., P ´ EREZ-JIM ´ ENEZ M.J., Fault diagnosis models for electric locomotive systems based on fuzzy reasoning spiking neural P systems, Lecture Notes in Computer Science, 12(7), pp. 385–395, 2014. [19] TU M., WANG J., PENG H., SHI P., Application of adaptive fuzzy spiking neural P systems in fault diagnosis of power systems, Chinese Journal of Electronics, 23(1), pp. 87–92, 2014. [20] WANG T., ZHANG G.X., ZHAO J.B., HE Z.Y., WANG J., P ´ EREZ-JIM ´ ENEZ M.J., Fault diagnosis of electric power systems based on fuzzy reasoning spiking neural P systems, IEEE Transactions on Power Systems, 30(3), pp. 1182–1194, 2015. [21] WANG T., ZHANG G.X., RONG H.N., P ´ EREZ-JIM ´ ENEZ M.J., Application of fuzzy reasoning spiking neural P systems to fault diagnosis, International Journal of Computers Communications & Control, 9(6), pp. 786–799, 2014. [22] WANG T., ZHANG G.X., P ´ EREZ-JIM ´ ENEZ M.J., CHENG J.X., Weighted fuzzy reasoning spiking neural P systems: application to fault diagnosis in traction power supply systems of high-speed railways, Journal of Computational and Theoretical Nanoscience, 12(7), pp. 1103–1114, 2015. [23] ZHANG G.X., CHENG J.X., GHEORGHE M., MENG Q., A hybrid approach based on differential evolution and tissue membrane systems for solving constrained manufacturing parameter optimization problems, Applied Soft Computing, 13(3), pp. 1528–1542, 2013. [24] P ˘ AUN GH., ROZENBERG G., A guide to membrane computing, Theoretical Computer Science, 287(1), pp. 73–100, 2002. [25] HUANG L., SUH H., ABRAHAM A., Dynamic multi-objective optimization based on membrane computing for control of time-varying unstable plants, Information Science, 181(11), pp. 2370– 2391, 2011. [26] P ˘ AUN GH., P ´ EREZ-JIM ´ ENEZ M.J., ROZENBERG G., Spike trains in spiking neural P systems, Information Sciences, 17(4), pp. 975–1002, 2006. 272 Y.Y. He et al. [27] PAN L.Q., P ˘ AUN GH., Spiking neural P systems: an improved normal form, Theoretical Computer Science, 411(6), pp. 906–918, 2010. [28] PAN L.Q., ZENG X.X., Small universal spiking neural P systems working in exhaustive mode, IEEE Transactions on Nanobioscience, 10(2), pp. 99–105, 2011. [29] ZHANG X.Y., LUO B., FANG X.Y., Sequential spiking neural P systems with exhaustive use of rules, Biosystems, 108(1–3), pp. 52–62, 2012.