Interpretation of Remote Backup Protection Operation for Fault Section Estimation by a Fuzzy Expert System



Similar documents
Local Back-up Protection for an Electric Power System

ECE 586b Course Project Report. Auto-Reclosing

Developing an Agent-based Backup Protection System for Transmission Networks

EVALUATION OF ALTERNATIVE BACKUP PROTECTION SCHEMES ON A 66KV DISTRIBUTION NETWORK

New Supervisory Control and Data Acquisition (SCADA) Based Fault Isolation System for Low Voltage Distribution Systems

AORC Technical meeting 2014

Teleprotection Schemes and Equipment. James W. Ebrecht Young Power Equipment Scottsdale, AZ

Nuclear Power Plant Electrical Power Supply System Requirements

APPLICATION CASE OF THE END-TO-END RELAY TESTING USING GPS-SYNCHRONIZED SECONDARY INJECTION IN COMMUNICATION BASED PROTECTION SCHEMES

Testing and Evaluating New Software Solutions for Automated Analysis of Protective Relay Operations

ISIO 200. Binary Input/Output (I/O) Terminal with IEC Interface

Distribution Automation Handbook. Section 8.13 Backup Protection

Electric utilities may vary in their application of end-to-end testing

F.C. Chan General Manager, CLP Engineering Ltd., Hong Kong SAR, China

Effect of Remote Back-Up Protection System Failure on the Optimum Routine Test Time Interval of Power System Protection

Effect of Remote Back-Up Protection System Failure on the Optimum Routine Test Time Interval of Power System Protection

CHAPTER 1 INTRODUCTION

Operational Overview and Controls Guide

REQUIREMENTS FOR A REAL-TIME RISK MONITORING TOOL TO REDUCE TRANSMISSION GRID NUCLEAR PLANT VULNERABILITIES

OPTIMAL DISTRIBUTION PLANNING INCREASING CAPACITY AND IMPROVING EFFICIENCY AND RELIABILITY WITH MINIMAL-COST ROBUST INVESTMENT

Transmission Protection Overview

SCADA Controlled Multi-Step Automatic Controlled Capacitor Banks & Filter Banks

VoIP Monitoring Environment based on SNMP - VMES

Application-oriented testing of line differential protection end to end in the field using the corresponding RelaySimTest template

Phase Balancing of Distribution Systems Using a Heuristic Search Approach

A Reactive Tabu Search for Service Restoration in Electric Power Distribution Systems

INTRODUCTION TO SYSTEM PROTECTION. Hands-On Relay School 2012

Bus Protection Considerations for Various Bus Types

Power products and systems. Intelligent solutions for power distribution Zone concept

Electric Power Distribution

Fundamentals of Modern Electrical Substations Part 1: Mission of Electrical Substations and their Main Components

INTELLIGENT BUILDINGS BUS SYSTEMS, MyHOME. Ján Cigánek, Martin Janáček, Stanislav Števo

Application of GA for Optimal Location of FACTS Devices for Steady State Voltage Stability Enhancement of Power System

TA Kahraman Yumak ELK412 - Distribution of Electrical Energy Lab. Notes v Spring web.itu.edu.tr/yumakk. Distance Protection

Study on Differential Protection of Transmission Line Using Wireless Communication

Fuzzy Candlestick Approach to Trade S&P CNX NIFTY 50 Index using Engulfing Patterns

Fuzzy Knowledge Base System for Fault Tracing of Marine Diesel Engine

Industrial Ethernet How to Keep Your Network Up and Running A Beginner s Guide to Redundancy Standards

ISSN: ISO 9001:2008 Certified International Journal of Engineering and Innovative Technology (IJEIT) Volume 3, Issue 3, September 2013

Calculation of Voltage Sag Indices for Distribution Networks

Eaton s E-Series protective relay family

Optimization of Asset Management in High Voltage Substation Based on Equipment Monitoring and Power System Operation

A HYBRID RULE BASED FUZZY-NEURAL EXPERT SYSTEM FOR PASSIVE NETWORK MONITORING

GE Protection and Control Business Department. Page 1 Date 5/11/99

New standardized approach to arc flash protection

Product Description Full Voltage Starting Electric Fire Pump Controllers FTA1000

APPLICATION OF ADVANCED SEARCH- METHODS FOR AUTOMOTIVE DATA-BUS SYSTEM SIGNAL INTEGRITY OPTIMIZATION

For a phase-to-phase voltage between 100 V and 1000 V. The standard ratings are: 400 V V V (at 50 Hz)

KNOWLEDGE-BASED IN MEDICAL DECISION SUPPORT SYSTEM BASED ON SUBJECTIVE INTELLIGENCE

Arc Flash Energy Mitigation Techniques

Chapter 5. System security and ancillary services

Study of Lightning Damage Risk Assessment Method for Power Grid

Secure your electrical network. PACiS The Safety Solution for Healthcare Institutions

2Azerbaijan Shahid Madani University. This paper is extracted from the M.Sc. Thesis

Medium Voltage (MV) & High Voltage (HV) Training Module 6. Date of the presentation

PSS E. High-Performance Transmission Planning Application for the Power Industry. Answers for energy.

Increased power protection with parallel UPS configurations

Reverse-feed applications for circuit breakers

Suitable for Emergency, Peak and Prime Power System Applications. Low Voltage Automatic Transfer Switch Systems

ESB Networks Response. ERGEG Consultation. Voltage Quality Regulation in Europe

OUTSOURCING MAINTENANCE PROCESSES IN ELECTRICITY UTILITIES. Philip Wester. NUON InfraCore, The Netherlands

DEMAND FORECASTING AND STRATEGIC PLANNING IN ELECTRICITY DISTRIBUTION COMPANIES: A SYSTEM DYNAMICS APPROACH

UNIVERSITY OF WASHINGTON Facilities Services Design Guide. Electrical. Switchboards. Basis of Design. Design Evaluation

A Multiagent Model for Intelligent Distributed Control Systems

ADMS(Advanced Distribution Management System ) in Smart Grid

ELECTRICAL ENGINEERING DESIGN CRITERIA APPENDIX F

IDENTIFYING BANK FRAUDS USING CRISP-DM AND DECISION TREES

Generator Stator Protection, under/over voltage, under /over frequency and unbalanced loading. Ramandeep Kaur Aujla S.NO

Mladen Kezunovic Texas A&M University

SENTRON Switching and Protection Devices Air Circuit Breakers

*.ppt 11/2/ :48 PM 1

DEVELOPING FUTURE SUBSTATION AUTOMATION STRATEGIES: SELECTING APPROPRIATE IEDs AND DEVELOPING NEW APPLICATIONS

Mathematical models to estimate the quality of monitoring software systems for electrical substations

Dr Jamali Publications in English

May, ArACBMENT I. Review of. Diesel Generator Alan and Control Circuitry. Indian Point Units No. 2 and 3. Docket Nos.

DEVELOPING A MODEL FOR PLANNING AND CONTROLLING PRODUCTION IN SMALL SIZED BUILDING FIRMS

IEC SCL - MORE THAN INTEROPERABLE DATA EXCHANGE BETWEEN ENGINEERING TOOLS

Tracking Groups of Pedestrians in Video Sequences

A FUZZY LOGIC APPROACH FOR SALES FORECASTING

Design of Prediction System for Key Performance Indicators in Balanced Scorecard

Digital Energy ITI. Instrument Transformer Basic Technical Information and Application

A MODERN DISTRIBUTION MANAGEMENT SYSTEM FOR REGIONAL ELECTRICITY COMPANIES

Impact of Remote Control Failure on Power System Restoration Time

REQUIREMENTS FOR AUTOMATED FAULT AND DISTURBANCE DATA ANALYSIS

Substation Automation and Smart Grid

A Direct Numerical Method for Observability Analysis

Electricity distribution network design

Introduction to Fuzzy Control

Adi Armoni Tel-Aviv University, Israel. Abstract

Fusible Disconnect Switch

MANAGING QUEUE STABILITY USING ART2 IN ACTIVE QUEUE MANAGEMENT FOR CONGESTION CONTROL

SUBJECT: How to wire a motor starter Number: AN-MC-004 Date Issued: 2/08/2005 Revision: Original

2695 P a g e. IV Semester M.Tech (DCN) SJCIT Chickballapur Karnataka India

..OR How To Protect your 3-Phase Equipment Investment with 3-Phase Monitors from Time Mark...

Transcription:

Interpretation of Remote Backup Protection Operation for Fault Section Estimation by a Fuzzy Expert System G. Cardoso Jr., J. G. Rolim and H. H. Zürn. Abstract-- A hybrid intelligent system, combining neural network modules with a fuzzy expert system, is employed for fault diagnosis in power transmission systems. The artificial neural networks model the protection system of every equipment and the fuzzy expert system analyses their outputs in order to identify the power system section where the fault occurred. Each neural module classifies the fault, according to the information on protection devices and circuit breaker events, in internal, external or lack of information. External fault classification is related to the operation of remote backup protection or pilot relaying starting units responsible for detecting faults outside the protected transmission line. The expert system s inexact reasoning deals with the imprecision of distance relay reach, besides considering information on the operative state of the main protection, for instance, if it is in maintenance. The hybrid method allows to infer the faulted component by mapping the fault direction determined by the relay operation. Index Terms Fault section estimation, backup protection, fuzzy sets. A I. INTRODUCTION fter contingencies with permanent outages, the control system operator should estimate the initial cause of the occurrence before starting the restoration procedures. When the protection system operates correctly and the disconnected area is small, this task is straightforward. However, when relays or circuit breakers fail to operate, remote backup protection operation is requested and larger areas are disconnected. Besides, the initial occurrence may cause other problems throughout the system, overloads or overvoltages for instance, causing other outages. In those cases, the operator may be overwhelmed by the number of received alarms, and the availability of a computational tool to support the decision making task can speed up the restoration process, besides reducing the risks of a bad interpretation of the signaled alarms. G. Cardoso Jr is with the Federal University of Pará (ghendy@ufpa.br). His work was supported by the Brazilian Agency for the Improvement of Higher Education (CAPES). J. G. Rolim and H. H. Zürn are with the Federal University of Santa Catarina and their research was supported by the Brazilian Research Council (CNPq) (jackie@labspot.ufsc.br; hans@labspot.ufsc.br). The fault diagnosis task may be defined as a decisionmaking problem, where several hypotheses (sections in fault), previously formulated, compete amongst themselves. The operator or the computational support tool is responsible for selecting the most probable one [1]. The literature proposes the use of intelligent systems for the task of fault diagnosis. Expert systems [2], neural networks [3], fuzzy logic [1] and genetic algorithms [4] constitute the main techniques applied to the problem. Other methods, such as Tabu search [5] and Petri nets [6], are also used. Most of the methods presented in the literature do not have difficulties in diagnosing the faulted component when the circuit breakers receive a tripping signal from relays used as main protection, for example: Buchholz, first zone distance relays, instantaneous overcurrent, pilot relaying schemes and differential relays. Generally, this kind of relays is very selective, and its operation allows to infer with great success the fault possibility of the component it protects, even when the deenergized area reaches larger proportions caused by circuit breaker failure, overloads, oscillations and overvoltages. The great challenge is to estimate the faulted component when the units used as main protection fail or do not operate, and the whole tripping process occurs by remote protection schemes. This paper proposes a methodology for interpreting the outputs of modules specially designed for identifying internal or external equipment faults by considering the operation of protection systems. In case of external fault classification, these modules also indicate the fault side. Therefore, the objective is to develop a simple computational tool with fast execution and ability to combine the information about the system topology and the neural module outputs, in order to estimate the faulted section, even in case of multiple faults, partial lack of information and/or failure of the protection system. For such, the problem employs multi-objective decision making theory. II. TRANSFORMER, LINE AND BUS PROTECTION SCHEMES This research work deals with philosophies commonly used in electric power systems for transmission lines,

transformers and busbars protection. The breaker failure protection is included in the busbar protection system, and has the objective of tripping all breakers connected to the bus if one of them fails to open after a protection request. The autotransformer protection is composed of main and backup protection, as shown in Fig. 1. Fig. 1. Autransformer protection scheme. The transformer main protection is composed of the following relays: 87-differential protective relay; 63 T- Buchholz relay; 63 VS- safety valve; 63 C- pressure relay of the under-load-tap-switch; 86-blocking relay; The backup protection is composed of: 94-trip relay; 51 HV-time overcurrent relay, high voltage side; 51 MV- time overcurrent relay, medium voltage side; 51 N ground time overcurrent relay; The transmission line main protection is based on a directional comparison carrier scheme, with directional phase distance (21P) and directional overcurrent ground relays without time delay (67NP). On the other hand, the alternative protection is based on directional distance relays with three protection zones and instantaneous ground overcurrent (67NI) and time overcurrent (67NT) relays. Fig. 2 shows a description of the line protection. Fig. 2. Transmission line protection. The first zone of the alternative protection is composed of a phase distance directional (21-1) and ground (21N-1) relay without time delay, adjusted at 80% of the length of the line. The second zone is composed of distance relays (21-2) adjusted at 120% of the length of the line, with time delay. Finally, the third zone (21S) is reverse and its main function is to start carrier signal, blocking the operation of the main protection for external faults, being therefore adjusted at a value larger than the reach of the distance relay used as main protection on the other end. This zone is also equipped with a time delay (TU3) and may trip the breaker in case other protections do not clear the fault. Bus protection is accomplished through a differential relay (87), which would excite the auxiliary unit (86) which sends a trip command to all breakers connected to the bus. Besides these relays, the bus also has overvoltage units (59), which also cause tripping. The breakers are equipped with breaker failure protection (86BF), which in case of a breaker failing to open after a time interval set by a timer (62BF), sends a trip command to all circuit breakers linked to the same bus. III. PROPOSED METHODOLOGY The solution strategy consists of using the information resulting from the execution of neural modules based on the observed alarms. There are three main types of neural networks used to model the philosophy of busbar, transmission line and transformer protection schemes. Evidently, there should be more models if the system being studied has different protection philosophies, for instance, for transmission lines operating at different voltage levels. Each neural network is trained to classify the received alarms in the following categories: fault, no fault, fault towards side S (sending end), fault towards side R (receiving end). The only exception is the busbar fault classification module, which does not classify the fault as external, since it does not have the function of providing remote protection for other equipment. After an occurrence with definitive outages, the neural module of each element with signaled events is executed. The connection between the information generated by neural modules is accomplished with the aid of an expert system. This expert system has two tasks. First it identifies the sections involved by the fault to form a set of suspected elements, according to the deenergized area. Each section included in this set represents one hypothesis (suspected component) of the cause of the events sequence being analyzed. The expert system second task is performed after it is supplied with the information resulting from the neural models (classifications), when it combines this information to give a final result. The expert system was implemented in CLIPS 6.05 (C Language Integrated Production System) due to its readiness, since it is a public domain program [7].

IV. TREATMENT OF THE EXTERNAL FAULT CLASSIFICATIONS The neural networks are trained in a way to interpret a group of alarms and to classify them according to the outputs used in the protection scheme model developed for each component. The neural modules which classified the fault as external are associated with electric components whose relays possibly operated as remote backup or are start relays used to indicate an external fault. The deenergized area mapping facilitates to infer the cause of the relay operation, using the indication of the fault direction, considering that the adjacent components are identified by the expert system. Consider that the 4 bus system presented in Fig. 3 corresponds to a deenergized area, whose topology was properly found. The suspected components are transmission lines or transformers ELM-1, ELM-2, ELM-3 and busbar B1. The buses B2, B3, and B4 are not part of the suspected list, because they are still energized. V. MULTI OBJECTIVE DECISION MAKING Multi-objective problems are characterized by a finite solution set and an objective set [8]. In applications of optimization techniques, the solution of this kind of problem is usually found combining the set of objectives into a single one by some computation. In rule based systems the various objectives may be combined using fuzzy inference, then the optimal solution a* represents the most plausible one, regarding all possibilities. To develop these calculations it is necessary to have some definitions in mind. Consider a universe with n solution alternatives, A = {a1,a2... a n }, and a set of r objectives, O = {O 1,O 2...Or}. If O i is the i th objective, the membership grade of an alternative a with respect to O 1 is denoted by µo 1 (a), which means the degree of how much alternative a satisfies the specified objective. Therefore, it is necessary to define a decision function that satisfies all the objectives of the decision process simultaneously [9]. The decision function, D, is given by the intersection of all objectives, that is, D = O 1 O 2... O r (1) Consequently, the membership degree of the decision function, D, for each alternative a is given in (2). µ D (a) = min[µ O1 (a), µ O2 (a),..., µ Or (a)] (2) The optimum decision, a*, is the alternative that satisfies (3). µ D (a*) = max (µ D (a)) a A (3) Fig. 3. Levels of backup protection and start relays of adjacent lines operation. ELM-x represents the element x (transmission line or transformer) that integrates the suspected list, whose neural module classified the fault as external. Therefore, x varies from 1 to the total number of suspect elements classified as external fault. On the other hand, Level-y corresponds to the level in which the adjacent element relates with the relay that operated or started, being y the number of the level. The branches and buses are disposed in levels according to the element whose neural module classification was external fault. Each one of these levels is associated with a membership function, whose values vary from 0 to 1. The larger this value the higher is the possibility of relays at ELM-x having operated due to a fault in the component belonging to the level action area. Thus, during the analysis, each section is classified with a level by each element that had the external fault indication. The multi-objective decision making theory is applied to infer the most plausible solution. In the case of the problem dealt in this paper one defines: A set of suspected components (deenergized components); objectives - evaluation of the alternatives (a) according to the levels assigned to it; level represents a neighbor equipment with some possibility of being the cause of the remote relay operation (VL, very likely; L, likely; LL, less likely; UL, unlikely; VUL, very unlikely); number of objectives equal to the number of components whose neural modules classified the fault as external. Figure 4 shows several levels and their importance with respect to a line or transformer for which the fault was classified as external. Therefore, the number of objectives r vary according to the number of elements whose neural net classified them as external fault. For instance, [µ O1 (a), µ O2 (a),..., µ Or (a)] represents the membership grade for each potential solution (alternative a) regarding the levels (objectives). Fig. 4 a) shows the case of a transmission line where the protection system operated due to a fault classified as external in one direction (shown in the figure). The cause of the

outage may be the transmission line (in case the main protection has failed), or busbar B1 (in case the main protection of the busbar fails). These first two hypotheses (TRANSMISSION LINE and B1) are considered very likely (VL). Fig. 4. Treatment of the external fault information accomplished for each line or transformer in agreement with the direction of the fault. There are still other hypotheses. It is likely (L) that the TRANSMISSION LINE protection relays have operated due to a fault in ELM-2, but in this case the main and local backup protection of ELM-2 would have failed and ELM-1 backup protection may be at the border of its reach (for example, a fault occurring at the end side of ELM-2). A fault at busbar B2 is less likely (LL) because in that case the primary protection of B2 and the remote backup protection provided by ELM-2 protection system would have failed. Besides, this would be a case of overreach of ELM-1 protection relays. Finally, it is unlikely (UL) that the protection of ELM-1 operated due to a fault at ELM-3 and very unlikely (VUL) for B3, considering the large number of relay failures and the distance of this equipment from ELM-1 All levels out of the direction and zone covered by the directional unit are considered as very unlikely (VUL, level 6). Fig. 4b) shows an example of a TRANSFORMER protection system. The procedure is similar to the one previously presented for a transmission line, but it is executed for the two sides of a TRANSFORMER, because its overcurrent protection is non directional. Based in the considerations presented above, the graph of the membership function µ Or (a) can be drawn for transmission lines (µ LT (a)) and transformers (µ TR (a)) as shown in Fig. 5. Level-1, which is the one with the most plausibility among all levels, reaches a maximum of 0.75, because it is supposed that if the neural module classifies a certain component as external fault, relays or circuit breakers have failed to operate. Value 1.00 (C) is only assigned for the membership function in cases when the backup protection operates correctly, explicitly cases with no relay or circuit breaker failure, as for instance, a busbar without primary protection (maintenance). Therefore, in case the primary protection of a certain component is in maintenance, the membership function will skip to the value above, that is: level 1 and level 2 pass from 0.75 (VL) to 1.00 (C); level 3 passes from 0.5 (L) to 0.75 (VL); level 4 from 0.25(LL) to 0.50 (L). The exceptions are level 5 and 6 that remain 0.15(VU) and 0.0(VUL), respectively, due to its distance with respect to the element that indicates external fault. Fig. 5. Membership function for transmission lines and transformers in function of their levels. The expert system developed to execute the tasks pertinent to this routine is composed of 5 main rules: RULE 1: Listing disconnected components. This rule finds all components which where turned off by the fault, but belonging to the same deenergized subsystem.

RULE 2: Determining adjacent elements. This rule has the function of determining the adjacency (first neighbours) of all components associated with the same deenergized area, found in Rule 1. RULE 3: Determines the levels of protection operation. This rule assigns levels for elements along the direction of the fault. This task is executed step by step, from the lowest level to the highest. This is done using 5 sub-rules that are run as many times as the number of external fault classifications. Rule 3.1: Level-1. Only lines or transformers are considered as level-1, in other words, electrical elements whose protection can operate for external faults. Fig. 6 shows the recognition of a level-1 element. Rule 3.4: Level-4. The adjacency of each component defined as Level-3 is found, according to Fig. 9, and assigned Level-4, except if the component has been previously classified as Level 1,2 or 3. Fig. 9. Definition of Level-4. Rule 3.5: Level-5. The adjacency of each component classified as Level-4 is found and labeled Level 5, except if the component has been previously classified as Level 1, 2, 3 or 4 (see Fig 10). Fig. 6. Definition of Level-1. Rule 3.2: Level-2. The adjacency for each element identified as Level-1 is found. Fig. 7 shows the recognition of a Level-2 component. Fig. 7. Definition of Level -2. Rule 3.3: Level-3. The adjacency of each component defined as Level-2 is found, according to Fig. 8, and assigned Level 3, except if the component has been previously been classified as Level 1 or 2. Fig. 8. Definition of Level-3. The direction of the fault shown in Fig. 8 is inherited from the previous level. Fig. 10. Definition of Level-5. RULE 4: Treatment of non directional units. All rules that compose Rule 3 are executed according to the indication of the fault direction. Therefore, an additional treatment for relays without directional units is necessary. In this case, Rule 3 is executed twice, once for each direction. This happens only if the equipment is protected by relays without directional units, for instance, transformers. RULE 5:Assigns Level-6 to other components. This rule assigns Level 6 to all components in the suspected list that have not yet been associated with a level. The components classified as Level-6 are those that are very far away from those classified as external fault. VI. TEST CASE Suppose that a fault occurrence and the subsequent operation of protection devices have caused the deenergized subsystem of Fig. 11. The alarms of transformer T1, transmission lines L_C-A, L_B-E and L_A-E started the neural module associate with each of these components. In all cases the fault was considered external in the direction indicated by the arrows in Fig. 11. Notice that it is supposed that no relay with main function has operated, therefore, all circuit breakers were tripped by remote backup relays. The suspected set to be analyzed is formed by T1, B, A, L_B-A, L_B-E and L_A-E. The other components (C, D, E and L_C-A) are not in this set, because they are energized. Table 1 displays the membership degree level associated with each component belonging to the suspects set after the

execution of the five rules previously described for each case of external fault classification (four columns). Fig. 11. Test system TABLE 1 MEMBERSHIP DEGREE FOR EACH COMPONENT REGARDING THE EXTERNAL FAULT INDICATIONS T1 L_B-E L_A-E L_C-A T1 0.75 0.50 0.15 0.15 B 0.75 0.75 0.25 0.25 A 0.25 0.25 0.75 0.75 L_B-A 0.50 0.50 0.50 0.50 L_B-E 0.50 0.75 0.15 0.15 L_A-E 0.15 0.15 0.75 0.50 Applying equation (2) for each line of Table 1 results in: µ D (L_B-A)=0.50; µ D (B)=0.25; µ D (A)=0.25; µ D (T1)=0.15; µ D (L_B-E)=0.15; µ D (L_A-E)=0.15. Therefore, by solving (3) we have D(a*) =0.50, thus transmission line L_B-A is the optimum solution (a*). It is interesting to notice that although the best solution is line L_B-A, busbars B and A are the second more likely hypotheses, and so on. VII. CONCLUSIONS The proposed methodology is quite simple and of low computational cost, being indicated to deal with cases where relays with primary function fail to operate, causing the outage of larger areas. The most plausible solution is found by mapping the fault directions indicated through the operation of overreaching relays, analysed by neural network modules, besides the determination of the deenergized area topology. The method may still be improved, considering different membership functions or adopting priority indexes for each electric component. [4] F. S. Wen, and C. S. Chang, Probabilistic approach for fault-section estimation in power systems based on a refined genetic algorithm, IEE Proc.-Generation Transmission Distribution, vol. 144, nº 2, pp. 160-168, March 1997. [5] F. S. Wen, and C. S. Chang, A tabu search approach to fault section estimation in power systems, Electric Power Systems Research, 40, pp. 63-73, 1997. [6] K. L. Lo, H. S. Ng, D. M. Grant, and J. Trecat, Extended Petri net models for fault diagnosis for substation automation, IEE Proc.-Generation Transmission Distribution, vol. 146, nº 3, pp. 229-234, May 1999. [7] J. Giarratano & G. Riley. Expert Systems: Principles and Programming. 3. ed. Boston: PWS- Publishing Company, p. 597, 1998. [8] A. H. F. Insfran; A. P. Alves da Silva; G. L. Torres, Fault Diagnosis Using Fuzzy Sets. In: INTELLIGENT SYSTEM APPLICATION TO POWER SYSTEMS (ISAP 99 April 4-8, 1999: Rio de Janeiro, Brazil). p. 162-166, 1999. [9] T. J Ross. Fuzzy Logic with Engineering Applications. MCGraw-Hill, 1995. IX. BIOGRAPHIES Ghendy Cardoso Junior received his B.Sc. degree in Electrical Engineering from the Federal University of Santa Maria, in 1995. In 1997 he received the M.Sc. degree in Power Systems, from Federal University of Pará (UFPA), where he is working as a lecturer, since 1997. From 1999 to 2003 he was a doctoral student at Federal University of Santa Catarina (UFSC). He is presently an associate professor at UFPA and his area of interest are short circuit, protection systems, and artificial intelligence techniques applications to power systems. Jacqueline Gisèle Rolim received her BSEE and M.Sc. in Power Systems from the Federal University of Santa Catarina (UFSC), Brazil, in 1982 and 1988, respectively. From 1985 to 1991 she worked in HV and EHV substation projects. She was in a split Ph.D. program at UFSC and Brunel University (England) and received her degree in 1995. She is an associate professor at UFSC and works with artificial intelligence applications to power system operation. Hans Helmut Zürn received his first engineering degree from the Federal University of Rio Grande do Sul in 1966. In 1969 e 1976 he earned his M.Sc. (University of Houston) and Ph.D. (University of Waterloo) respectively. Since 1967 he lectures at the Federal University of Santa Catarina (UFSC) with emphasis on power systems, dispersed/renewable generation and stochastic methods. VIII. REFERENCES [1] D. Y. Park, B. S. Ahn, S. H. Kim,H. J. Lee, Y. M. Park, J. K. Park, and J. R. Shin, Dealing Uncertainties in the Fault Diagnosis System, Intelligent System Application to Power Systems, pp. 273-277, Brazil, April 1999. [2] T. S. Sidhu, O. Cruder, and G. J. Huff, An abductive inference technique for fault diagnosis in electrical power transmission networks, IEEE Transaction on Power Delivery, vol. 12, pp. 515-522, Jan. 1997. [3] C. Rodriguez, S. Rementería, J. I. Martín, A. Lafuente, J. Muguerza, and J. Pérez, Fault analysis with modular neural networks, Electric Power & Energy Systems, vol.18, nº 2, pp. 99-110, 1996.