Вы находитесь на странице: 1из 9

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO.

1, JANUARY 2011

205

Fault Section Estimation Using Fuzzy Matrix-Based Reasoning Methods


Wen-Hui Chen, Member, IEEE
AbstractA technique of fuzzy reasoning via rule matrix transformations has been developed to estimate fault sections for distribution substations. This study extended the application of fault diagnosis from binary reasoning to fuzzy reasoning. In the inference procedures, the causalities of fault sections and the actions of protective devices were rst represented by the fuzzy cause-effect networks (FCE-Nets). After performing some simple matrix operations, the possible fault sections were estimated. This proposed approach offers a clear framework, rapid reasoning, the ability to handle uncertainty, and it has no problem with convergence during the diagnosis procedure. Index TermsBoolean rule matrix, cause-effect networks, fault diagnosis, fuzzy reasoning.

I. INTRODUCTION HESE DAYS, most power systems are equipped with supervisory-control-and- data-acquisition (SCADA) systems for operational improvement applications. When a fault occurs, it is imperative to limit the impact of outages to the minimum and to restore the fault area as soon as possible. This requires that the fault segment be identied from information provided by protective devices. Various approaches have been proposed to address fault diagnosis problems for power systems. Among those techniques, rule-based expert systems, articial neural networks (ANNs), genetic algorithms (GAs), and their hybrid models are commonly used. Although many works have been performed for power system fault diagnosis, recent research interests are especially focused on how to deal with uncertainty inherent in the operation of protective devices. In [1] and [2], the authors used fuzzy relations represented by sagittal diagrams to deal with uncertainty for power transmission networks, and good results were obtained even in device malfunctions. In order to apply an approach adaptive to network topology change, the authors in [2] extended their work by integrating time-sequence information, protective schemes, and network topology to a 3-D matrix. In [3], the authors presented a hybrid fuzzy rule-based expert system for transmission-level fault diagnosis. The fuzzy concept in this paper is to assign each rule a subjective assigned possibility for the expert system to

perform reasoning. The fuzzy reasoning process was carried out by performing max-product composition. Petri nets are a useful tool for event modeling in a concurrent structure. However, it lacks the ability to handle uncertainty in information. As such, the authors in [4] applied fuzzy Petri nets approach for fault diagnosis in transmission networks. They combined fuzzy logical reasoning with Petri nets to deal with uncertainty existing in the operation of protective devices. The analysis of cause-and-effect relationships for fault diagnosis has been a favorite application in [5] and [6]. In [7] and [8], a fast fault diagnosis algorithm based on cause-effect networks (CE-Nets) was proposed for distribution substations. Fault diagnosis can be considered as a classication task. The authors in [9] have successfully applied a data-mining-based fuzzy classication algorithm to handle fault cause identication problems and demonstrated good results. To summarize, previous research tended to use a fuzzy relation to deal with the uncertainty in the operation of protective devices. In addition, they preferred using subjective possibility given by the operators as a certainty factor in a rule. Note that many previous works assumed that there is no uncertainty between the operations of the relays and circuit breakers (CBs). Fault diagnosis problems for power systems range from power equipment [10], [11]; substations [12], [13]; distribution networks [14]; and transmission networks [15] to power plants [16]. For different applications, the characteristics of a problem, domain knowledge, and concerns would vary accordingly. To endow CE-Nets with inexact reasoning ability, in this paper, a new technique using the fuzzy rule matrix transformation was proposed for fault diagnosis in distribution substations. II. FUZZY CAUSE-EFFECT NETWORKS (FCE-NETS) A. Review of Cause-Effect Networks (CE-Nets) CE-Nets are graphic-modeling methods for representing the causalities between faults and the actions of protective devices [7], [8]. They consist of three kinds of nodes: 1) fault section and node, 2) relay node, and 3) CB node. If two nodes are related, a directed arc is placed on the graph from node to node , indicating the relation. There are three kinds of directed arcs: protected-by arc, cause arc, and backup-by arc, that describe different connections between two nodes. A CE-Net can be represented by a binary matrix. The transpose of the matrix is equivalent to the reversal of all arrows in the associated CE-Net. Therefore, we can use this property to achieve backward reasoning to estimate the possible fault causes. A CE-Net can be easily derived from the system conguration.

Manuscript received March 26, 2010; revised July 10, 2010; accepted July 13, 2010. Date of publication September 07, 2010; date of current version December 27, 2010. This work was supported by the National Science Council of Taiwan (NSC98-2221-E-027-072 and NSC98-2218-E-002-012). Paper no. TPWRD-00222-2010. The author is with the Graduate Institute of Automation Technology, National Taipei University of Technology, Taipei, Taiwan (e-mail: whchen@ntut.edu.tw). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRD.2010.2061873

0885-8977/$26.00 2010 IEEE

206

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

Fig. 3. Trapezoidal fuzzy number.

Fig. 1. Simple model distribution system.

Fig. 2. CE-Net that represents the model system.

Fig. 4. Membership functions for the linguistic term set.

A simple model distribution system, as shown in Fig. 1, is employed to illustrate the concept. The model system is protected by overcurrent (CO) relays, low-energy over current (LCO) relays, and CBs. Suppose that a fault occurs at the feeder F1 and causes the action of the relay CO1, which trips the circuit breaker CB1. If CB1 fails to open at this moment, the backup protective relay CO3 will operate to trip the circuit breaker CB3. This event is described in the portion of the dotted outline in Fig. 2. After considering all possible fault sections, the CE-Net that represents the model system can then be derived. B. Concept of FCE-Nets Since human knowledge can contain linguistic terms with some degree of uncertainty, it is desired to express the degree of certainty of a rule as a real number between 0 and 1, and incorporate such an expression in the inference. A rule consists of an antecedent part and a consequence part, with the antecedent parts describing causes and the consequent parts describing effects. For example, a rule could be IF (relay CO1 operates) THEN (circuit breaker CB1 tripped). However, in some situations, CB1 cannot be asserted to have tripped, when CO1 operates due to incorrect settings or equipment malfunction. In the classic (nonfuzzy) inference, when the condition in the antecedent part matches the effect, its consequent part is asserted. This is not necessarily true for protective device operation. This uncertainty regarding the operation of protective devices should be considered for practical applications.

The certainty factor refers to the degree of condence that the event will occur. As operators knowledge may contain linguistic terms with some degree of uncertainty, the use of certainty factor is a good way to describe the uncertainty in numerical expression. In this study, the certainty factor is used to represent the uncertain characteristic in conditions and rules. The assignment of a certainty factor can be intuitively given by experts using a linguistic term set [17] or based on some mathematical operations, such as the frequency of occurrences from historical data. For example, the operators may say that a fault causing the main protective relay to operate is extremely true or that the CB that will be tripped is very true. They cannot precisely tell you how true it is in a quantitative way. Table I lists the linguistic terms and their corresponding fuzzy numbers. A fuzzy number can be characterized as a four-tuple , where denotes the interval in which the membership value is equal to 1, and and indicate the left and right width of the trapezoidal distribution. Fig. 3 shows a trapeparameterized by . The corzoidal fuzzy number responding membership functions are given in Fig. 4. The other way to decide the certainty factor is based on the frequency of occurrences from historical data. For example, from historical data, the frequency of occurrences of a fault hitting the feeder F1 was up to 20 times, and only in 19 times, the fault has caused the operation of relay CO1. According to the statistical data, we can calculate the certainty factor of the rule as 19/20 or 0.95. With the certainty factor, the rule can

CHEN: FAULT SECTION ESTIMATION USING FUZZY MATRIX-BASED REASONING METHODS

207

TABLE I A NINE-MEMBER LINGUISTIC TERM SET

Fig. 5. Associated graph for fuzzy implication rules.

Fig. 6. Basic node-arc relations in FCE-Nets.

be expressed as IF (a fault at F1) THEN (relay CO1 operates) 0.95) and are node conditions in a given FCESuppose that Net. The general formulation of a fuzzy implication rule can be denoted as (1) This infers that the truth of condition implies the truth of with a certainty factor . The value of a certainty condition factor is between 0 and 1, which indicates the strength of the belief in the rule. The larger the value is, the more reliable the rule is. The associated graph for (1) is shown in Fig. 5. There are three basic node-arc relations of FCE-Nets, as shown in Fig. 6. Generally, the choice of a good certainty factor needs some expertise. In order to avoid a bias in assigning the certainty factors, it is suggested to determine the value based on experts or senior operators experience and historical data. Since operators knowledge can contain some degree of uncertainty, a rule expressed for the linguistic knowledge can be described as IF (a fault at F1) THEN (relay CO1 operates) For different occasions, the same rule may change to

IF (a fault at F1) THEN (relay CO1 operates) or we can express it in a general form: IF (a fault at F1) THEN (relay CO1 operates) ( or ET or VT or T, ) The determination of which linguistic terms to be used is based on operator experience and the frequency of occurrences. In other words, according to conrmed cases and operator experience, the value of a certainty factor can be different in a rule. Examples of node-arc relations are given below. IF (a fault at F1) THEN (CO1 operates) IF (CO1 operates) THEN (CB1 tripped) . A SCADA system typically consists of a master station, communication networks, remote terminal units (RTUs), and a number of transducers. A transducer is a device which provides a transformed output in response to a specic quantity, such as feeder currents and bus voltages. For example, a current transducer provides the RTU with current information by converting current transformer (CT) signals to the value that can be handled by a RTU, and then the RTU transmits this information to the control center through communication networks. In the signal transmission process, there may be a data mismatch between the measured quantity of a substation and the SCADA computer of the control center due to transducer imperfection or software conversion errors. For example, if a fault occurs at feeder F1 in a substation, the circuit breaker connected to feeder F1 will be tripped, which leads to the current of feeder F1 reaching zero. In normal situation, this zero value will be displayed in the control center. When the current transducer associated with feeder F1 has improper installation or ill-calibration, the practical feeder current could not correctly be displayed in the SCADA computer. As a result, it may show a very small value instead of zero. Conventional rule-based approaches do not allow linguistic variables to appear in the condition of a rule, and have very little power in dealing with imprecise information. In order to deal with this situation, the proposed approach utilizes fuzzy membership functions to represent measured quantities derived from SCADA systems instead of specifying a xed value, and use fuzzy inference techniques to perform inexact reasoning in the proposed inference algorithm. As an opened CB leads to no current passing through it, the degree of truth in the condition circuit breaker CB1 is tripped can be determined by the degree of truth that the value of current passing through CB1 is near zero. A membership function determining the degree of CB in open status is shown in Fig. 7. For example, suppose the value of a current passing through CB1 is 150 A. One can therefore infer that the degree where CB1 tripped is 0. The certainty factor can be viewed as the degree of condence that the event will occur. The assignment of a certainty factor can be intuitively given by experts or based on some mathematic operations, such as the frequency of occurrences from historical data. Generally, the choice of a good certainty factor needs some expertise. In order to avoid a bias in assigning the certainty factors, it is suggested to determine the value based on experts or senior operators experience and historical data.

208

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE II THE SET OF CONDITIONS AND THEIR DESCRIPTIONS

Fig. 7. The membership function for CB in open status.

III. REASONING WITH FUZZY RULE MATRICES In this section, an inexact reasoning algorithm via fuzzy rule matrix transformation for FCE-Nets is presented. Assume that a FCE-Net contains a set of conditions. Each condition contains a high level of linguistic expression for humans to use (e.g., relay CO1 operates). However, this string expression is inconvenient and inefcient for computers to process. Therefore, we dened some vectors to transform string based conditions into numerical vectors for reasoning and computation. The following vectors and matrix operations are dened to develop the inference procedures. : The truth state vector is em1) Truth State Vector ployed to represent the fault symptom with the status of protective devices. This vector contains the truth values for a set of conditions, . Each component is dened by , where is the truth value of condition . : The fault node vector is de2) Fault Node Vector ned to represent the fault section nodes in a given FCE-Net. This vector contains Boolean valued components for a set of conditions. If node condition is associated with a fault secis assigned to 1, otherwise is 0 tion node, the value of if otherwise This vector is dened for extracting the nodes that belong to the fault section node through fuzzy intersection operation. : The backup node vector is 3) Backup Node Vector employed to represent the backup relay nodes in a given FCENet. This vector contains Boolean valued components for a

Fig. 8. The associated FCE-Net for the model system.

C. Matrix Representation of FCE-Nets The rules with linguistic certainty factors can be represented as the fuzzy rule matrix . This matrix describes the relations between causes and effects of FCE-Nets. Once the fuzzy rule matrix is established, the diagnosis algorithm can then be performed by matrix operations. A fuzzy rule matrix associated with conditions is a -bymatrix with all ones on the diagonal by reexivity because each means that concondition implies itself. The entry dition implies condition with the certainty factor , and indicates that there is no implication between and . A fuzzy value in the entry of the fuzzy rule matrix gives implies the truth the degree of condence in how condition of condition . The simple model system in Fig. 1 is used again as an example to illustrate the concept of matrix representation of FCE-Nets. The associated FCE-Net for the model system is shown in Fig. 8. The set of conditions is listed in Table II. According to the connectivity of each node in Fig. 8, the fuzzy rule matrix that represents the FCE-Net can be built as: see equation at the bottom of the page

CHEN: FAULT SECTION ESTIMATION USING FUZZY MATRIX-BASED REASONING METHODS

209

set of conditions. If node condition is associated with a is assigned to 1, otherwise backup relay node, the value of is 0. if otherwise This vector is dened for extracting operated backup relays through fuzzy intersection operation. 4) Fuzzy Min-Operator : The fuzzy min-operator for two column vectors and on the corresponding entry is written . For example, as

with . The step of comparing Then, we compared with with was to check if there was a device failure. If equaled with , this means that there was no failure operation did at feeder protection; otherwise, failures did occur. If and assign it not equal with , the process went to update using (3); otherwise, assigned to . to (3) Fuzzy min-operator in (3) is used to remove the status of operated backup relays in the truth state vector when the action of the backup relays is caused by a main relay failure. The fuzzy multiplication operator is used for performing truth state transformation on the transpose of the fuzzy rule matrix and the truth state vector operated on a backup relay node by fuzzy min-operator. As such, the updated transformation vector using (3) is to remove the status of backup relays from fault section candidates. contained information about fault causes, As the vector we selected only fault section nodes with the entry value greater than a threshold as estimated fault sections. The selection can be achieved by computing of fault section nodes from . As suggested in [3], the possibility of a solution below 0.8 would be meaningless. The threshold value for selecting fault sections was set to 0.5 in this study. This value was suitable after testing on many runs. The ow chart of the inference procedures is shown in Fig. 9. IV. CASE STUDY A typical Taipowers secondary substation, as shown in Fig. 10, is employed to illustrate the reasoning process of the proposed approach. The substation is composed of three sub-transmission lines, three main transformers, two tie circuit breakers, one 69 kV primary bus bar, and three 11.4 kV secondary bus bars. Each secondary bus bar contains ve radial distribution feeders. Each feeder is protected by three CO relays and one LCO relay. For example, feeder F1 is protected by three CO relays: CO1-A, CO1-B, and CO1-C as well as one LCO relay: LCO1. The protective relays for 11 kV bus bars also serve as the back-up protection for their connected feeders. This study case is the same as case 2 in [8] except there is missing information in this case. Operated relays: CO3-A, CO3-B, CO3-C, COM1-A, COM1-B, COM1-C, CO8-B, CO8-C Tripped circuit breakers: CB-M1, CB-8 Failure devices: LCO8, CB-3 Missing information: the status of LCO8, CO8-B, CO8-C A. Fault Diagnosis by FCE-Nets Approach In this example, transformer TR #2 has de-energized for maintenance and CB Tie-1 is in closing status. A three-phase fault occurs at the feeder F3. Relays CO3-A, CO3-B and CO3-C operate correctly while the circuit breaker CB-3 fails to trip. Owing to the tripping failure of CB-3, the backup breaker CB-M1 is tripped by the relays COM1-A, COM1-B and COM1-C. Meanwhile, a double-line to ground fault at phases B and C happens at feeder F8, which causes CO8-B and CO8-C operating to trip CB-8. However, the status of relays,

In the inference procedures, there is a process to remove the status of operated backup relays in the truth state vector if the action of the backup relay is caused by a main relay failure. This process is used to discriminate the source that activates backup relays. This discrimination can be carried out by using the intersection operation. As such, we introduced this operator to perform the required computation for deriving a systematic inference algorithm. : The fuzzy multipli5) Fuzzy Multiplication Operator cation operator for two matrices and is written as The row-by-column matrix product is perfromed by replacing multiplication and addition with the min and max operations, repectively. For example,

The fuzzy multiplication operator is used for performing truth state transformation on the transpose of the fuzzy rule matrix and a truth state vector that contains the degree of truth in its entries. The entry of a fuzzy rule matrix represents an implication between two conditions. As truth state contains information of fault symptoms, the function of this operator is to perform a composition transformation that propagates the truth state leading backward into the fault cause. In this study, we adopted max-min operation to replace real multiplication and addition in arithmetic matrix multiplication. The inference procedures are summarized as below. At rst, we built a -by- fuzzy rule matrix according to the given FCE-Nets, with rows and columns indexed by nodes. presented, we put the certainty factor in cell If an arc ; otherwise we left as zero. denotes a directed arc from node to node . According to fault symptoms, we put the truth value of each condition in the entry to derive the truth state vector. Calculate transformation vector, using (2). The meaning of this transformation is to propagate the truth state of given fault symptoms leading backward contains information that into the fault cause. The vector causes the fault symptoms. (2)

210

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE III THE SET OF NODE CONDITIONS

Fig. 9. The ow chart of the proposed inference procedures.

TABLE IV CONFIRMED EVENTS FROM SCADA SYSTEM DATABASE

Fig. 10. The distribution substation for the study case.

LCO8, CO8-B and CO8-C, is missing. The set of conditions is listed in Table III. The certainty factor is determined by operators according to their knowledge and the frequency of occurrences from conrmed events of historical data, given in Table IV. It is assumed that the event in the same category has the same occurrence frequency.

The corresponding FCE-Net associated with this fault case is shown in Fig. 11. The inference procedures are described step by step as follows. Step 1: Step 1 At rst, we build the fuzzy rule matrix according to Fig. 11. As the dimension of the matrix is 51-by-51, only nonzero entries of the matrix are listed in Table V. Step 2: Step 2 The currents for the related devices are listed in Table VI. The possibility of a condition that a relay operates is evaluated according to the device current and the membership value that the device is in its open status, set in Fig. 7. The nonzero entries of truth state vector are listed in Table VII.

CHEN: FAULT SECTION ESTIMATION USING FUZZY MATRIX-BASED REASONING METHODS

211

TABLE V NONZERO ENTRIES OF FUZZY RULE MATRIX

Fig. 11. The associated FCE-Net of the study case.

TABLE VI DEVICE CURRENTS OF THE STUDY CASE

Step 3: Step 3 The fuzzy truth state transformation can be obtained using (2), and the results are listed in Table VIII. and , are not Step 4: Step 4 As these two vectors, equal, the inference procedure goes to update the fuzzy truth state transformation using (3). Then, we assign the updated values to . The value of vector is listed in Table IX. contains information about Step 5: Step 5 As the vector fault causes, the fault sections can be retrieved by with entry selecting fault section nodes from value greater than 0.5. This can be achieved by rst , and then selecting the entry performing value greater than 0.5 as estimated fault sections. The nonzero entries of the fault node vector are listed in Table X. The results after selecting the fault section nodes by are listed in Table XI performing From Table XI, the estimated fault sections can be obtained by selecting the entry value greater than 0.5. Therefore, feeders F3 and F8 are selected as fault sections, i.e., multiple faults at F3 and F8 are estimated. The inference results are summarized in Table XII. B. Fault Diagnosis by CE-Nets Approach In order to make comparisons with CE-Nets based approach, the method presented in [8] was applied to this fault case. The associated CE-Nets and the rule matrix are the same as those of case 2 in [8]. The inference results using CE-Nets approach are summarized in Table XIII. Note that the fault situation has missing signals, which is different and more complicated than that in case 2 of [8]. From the experimental results, as shown in Tables XII and XIII, the proposed FCE-Nets can correctly identify the fault sections, F3 and F8, while the approach using CE-Nets fails to nd the fault section F8 due to incomplete information.

TABLE VII NONZERO ENTRIES OF TRUTH STATE VECTOR

TABLE VIII ENTRY VALUES OF TV

212

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE IX ENTRY VALUES OF

TABLE XII ESTIMATED RESULTS USING FCE-NETS APPROACH

TABLE XIII ESTIMATED RESULTS USING CE-NETS APPROACH

TABLE X NONZERO ENTRIES OF FAULT NODE VECTOR

TABLE XI ENTRY VALUES OF

T ^F

CE-Nets approach with the ability of handling uncertain and incomplete information that may happen in power system operation. This example demonstrates that the proposed FCE-Nets approach outperforms the method presented in [8]. When the information is complete, both fuzzy (FCE-Nets) and non-fuzzy (CE-Nets) approach can obtain the correct results. In the situation where a status signal is missing, CE-Nets approach fails to nd the fault sections. From this study, it is observed that the proposed algorithm has the ability to infer multiple faults even when a failure device and incomplete information or missing signals occur. The merits of the proposed approach compared to the approach in [8] are summarized below. 1) The proposed approach can handle linguistic terms in conditions and rules. It is more exible than that in [8] as it allows the use of certainty factors in a rule rather than crisp numerical values. 2) The proposed approach provides the results with quantitative condence level for all estimated fault sections, which makes the results more conclusive. 3) The proposed approach successfully integrates fuzzy reasoning with graphical model representation. Therefore, the proposed approach has the ability to handle incomplete and uncertain information in practical problems with fast computation speed. V. CONCLUSION In this study, we developed fuzzy CE-Nets and presented a new reasoning algorithm for fault diagnosis in distribution substations. The proposed approach is capable of representing uncertain knowledge and performing fuzzy reasoning through matrix based transformation. Since knowledge representation with the fuzzy CE-Nets model is based on graphical methodology, it

In general, CE-Nets approach has some advantages in fault diagnosis but only work well when the fault information is complete and certain. FCE-Nets approach is an extension of

CHEN: FAULT SECTION ESTIMATION USING FUZZY MATRIX-BASED REASONING METHODS

213

is easy to understand the relationship between the rules and conditions. Also, it is possible to predict the inference results in advance by observing the ow of truth state in the FCE-Nets when some conditions are specied. Since the proposed reasoning algorithm requires only simple matrix operations, it is well suited for integration with existing SCADA systems for on-line applications. ACKNOWLEDGMENT The author would like to thank Prof. C.-W. Liu of National Taiwan University for his guidance and Prof. C.-J. Chou of National Taipei University of Technology for his insightful conversations. REFERENCES
[1] H. J. Cho and J. K. Park, An expert system for fault section diagnosis of power systems using fuzzy relations, IEEE Trans. Power Systems, vol. 12, no. 1, pp. 342348, Feb. 1997. [2] S. W. Min, J. M. Sohn, J. K. Park, and K. H. Kim, Adaptive fault section estimation using matrix representation with fuzzy relations, IEEE Trans. Power Systems, vol. 19, no. 2, pp. 842848, May 2004. [3] H. J. Lee, D. Y. Park, B. S. Ahn, Y. M. Park, J. K. Park, and S. S. Venkata, A fuzzy expert system for the integrated fault diagnosis, IEEE Trans. Power Delivery, vol. 15, pp. 833838, Apr. 2000. [4] J. Sun, S. Y. Qin, and Y. H. Song, Fault diagnosis of electric power systems based on fuzzy Petri nets, IEEE Trans. Power Systems, vol. 19, no. 4, pp. 20532059, Nov. 2004. [5] Y. Sekine, H. Okamoto, and T. Shibamoto, Fault section estimation using cause-effect network, in Proc. 2nd Symposium on Expert System Application to Power Systems, Seattle, WA, 1989, pp. 277282. [6] Y. Sekine, Y. Akimoto, M. Kunugi, C. Fukui, and S. Fukui, Fault diagnosis of power systems, IEEE Proceedings, vol. 80, no. 5, pp. 673683, May 1992. [7] W. H. Chen, C. W. Liu, and M. S. Tsai, On-line fault diagnosis of distribution substations using hybrid cause-effect network and fuzzy rule-based method, IEEE Trans. Power Delivery, vol. 15, pp. 710717, Apr. 2000. [8] W. H. Chen, C. W. Liu, and M. S. Tsai, Fast fault section estimation in distribution substations using matrix-based cause-effect networks, IEEE Trans. on Power Delivery, vol. 16, no. 4, pp. 522527, Oct. 2001.

[9] L. Xu and M. Y. Chow, Power distribution system fault diagnosis using hybrid algorithm of fuzzy classication and articial immune systems, in Soft Computing Applications in Industry. : Springer-Verlag Berlin Heidelberg, 2008, vol. 226, pp. 357372. [10] Z. Wang, Y. Liu, N. C. Wang, T. Y. Guo, F. T. C. Huang, and P. J. Grifn, Articial intelligence in power equipment fault diagnosis, in IEEE PES Winter Meeting, 2000, vol. 1, pp. 128133. [11] C. Pan, W. Chen, and Y. Yun, Fault diagnostic method of power transformers based on hybrid genetic algorithm evolving wavelet neural network, IET Electric Power Applications, vol. 2, no. 1, pp. 7176, Jan. 2008. [12] A. P. Silva et al., Neural networks for fault location in substations, IEEE Trans. Power Delivery, vol. 11, pp. 234239, Jan. 1996. [13] H. J. Lee, B. S. Ahn, and Y. M. Park, A fault diagnosis expert system for distribution substations, IEEE Trans. Power Del., vol. 15, no. 1, pp. 9297, Jan. 2000. [14] R. H. Salim, K. de Oliveira, A. D. Filomena, M. Resener, and A. S. Bretas, Hybrid fault diagnosis scheme implementation for power distribution systems automation, IEEE Trans. Power Del., vol. 23, no. 4, pp. 18461856, Oct. 2008. [15] F. S. Wen and C. S. Chang, Probabilistic approach for fault-section estimation in power systems based on a rened genetic algorithm, Proc. Inst. Elect. Eng., Gen. Transm. Distrib., vol. 144, no. 2, pp. 160168, 1997. [16] S. C. Tan and C. P. Lim, Application of an adaptive neural network with symbolic rule extraction to fault detection and diagnosis in a power generation plant, IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 369377, Jun. 2004. [17] S. H. Wei and S. M. Chen, A new approach for fuzzy risk analysis based on similarity measures of generalized fuzzy numbers, Expert Syst. Appl., vol. 36, pp. 589598, 2009. Wen-Hui Chen (M04) was born in Taiwan in 1965. He received his B.S. degree from National Taiwan University of Science and Technology, and the M.S. and Ph.D. degrees from National Taiwan University, all in electrical engineering. From year 1992 to 2000, he held the position as a senior engineer and received numerous employee outstanding awards at Taiwan Power Company (Taipower Co.), Taiwans largest power utility company. He is currently an associate professor at the Graduate Institute of Automation Technology, National Taipei University of Technology.

Вам также может понравиться