аЯрЁБс>ўџ /1ўџџџ.џџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџьЅС5@ №П[bjbjЯ2Я2 .­X­X[ џџџџџџˆ€€€€€€€”мммм ш ”ЭЖ  ,LNNNNNN$ƒRеdr€444r€€‡4ˆ€€L4L(€€(є РЖ&Ѓ=ХмМ((L0Э(9ф9(””€€€€9€($4444444rr””ФX„њ””XMarco Aurщlio P. Rodrigues "Localizaчуo de Defeitos em Sistemas de Potъncia Utilizando Redes Neurais" In electrical power systems a large number of messages and alarms are transmited to the control center in case of disturbances. These disturbances are associated with faults that may be of different types and can occur anywhere the system. In such situations part of the system is in general isolated in order to eliminate the fault. Protection devices are responsible for detecting the occurrence of a fault and isolating conditionsas soon as possible. Then, it is essential that the fault location is determined in a very short time. In the control center the operator usually has to draw conclusions from a large amount of information, which may be very time-consuming. Other problems suchas: protection devices failures, communication problems, acquisition of corrupted data, etc. can make the fault location a very difficult task. Intelligent system applications to fault diagnosis have been prposed in the technical literature. Many of them are based on the use of expert systems. The major drawback of these methods is the difficulty to deal with new or corrupted alarm patterns. Some methods baseds based on the application of artificial neural networks have also been proposed for system fault diagnosis. These methods usually assume that the protective system is well-defined, with the availability of all messages and alarms at the control center. This situation may not happen for practical power systems. In this work an artificial neural network based methodology is proposed for fault location and diagnosis. Several artificial neural networks are employed, each of them being responsible for detecting faults involving a limited number of components. This strategy leads to the construction of artificial neural etwork classifiers with a reduced number of input variables. Each ANN is trained offline by considering in the training set several alarm patterns associated with fault occurrences involving the monitored components. These alarm patterns also include situations for which protection devices do not work properly. The proposed methodology is tested using a test system and a real Brazilian system. Test results shows the excellent generalization and discrimination capability provided by the ANNs. Correct classifications are obtained even for dificult situations involving protection device failures, loss of data, corrupted data and other alarm patterns not considered in the training phase. It is shown how, in some situations, the ANNs individual diagnosis can be used together in order to produce a final diagnosis. fZ[ђрмиhДF]hсЯ#hсЯhсЯ5B*\mHphџsHhсЯhсЯ5\mHsHfЊ _Z[њњњњњјgdсЯ[ў,1hАа/ Ар=!А"А# $ %ААФАФ Фœ@@ёџ@ NormalCJ_HaJmH sH tH DA@ђџЁD Default Paragraph FontRi@ѓџГR  Table Normalі4ж l4жaі (k@єџС(No ListL^`ђL сЯ Normal (Web)ЄdЄd[$\$ B*ph[ џџџџfЊ_Z ] ˜0€€˜0€€˜0€€˜0€€p˜0€€˜0€€Њ_Z ] O90(€M90(€M90ˆ+Ё[ [ [ џџ maprodrigues] ] eЌЖ чюВИ&,  E M   ] eГХ] 3fх] fh] џџadilsonхДF]јgžсЯџ@€ffдЈўff[ @@џџUnknownџџџџџџџџџџџџG‡z €џTimes New Roman5€Symbol3& ‡z €џArial"qˆ№аhмZ’ІмZ’І‹а‹аб№ ДД24V V 3ƒ№H)№џ?фџџџџџџџџџџџџџџџџџџџџџјgžџџMarco Aurщlio Padilsonadilsonўџр…ŸђљOhЋ‘+'Гй0p˜АМЬифє  , 8 DPX`hфMarco Aurщlio ParcadilsondildilNormaladilson1ilMicrosoft Word 10.0@@0g=Х@0g=Х‹аўџеЭеœ.“—+,љЎ0ј hp|„Œ” œЄЌД М ифUDRV A Marco Aurщlio P Title ўџџџўџџџўџџџ !"#$%ўџџџ'()*+,-ўџџџ§џџџ0ўџџџўџџџўџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџRoot Entryџџџџџџџџ РFp!HЃ=Х2€Data џџџџџџџџџџџџ1TableџџџџWordDocumentџџџџ.SummaryInformation(џџџџџџџџџџџџDocumentSummaryInformation8џџџџџџџџ&CompObjџџџџџџџџџџџџjџџџџџџџџџџџџўџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџџўџ џџџџ РFMicrosoft Word Document MSWordDocWord.Document.8є9Вq