Método para estimação da distância de faltas de alta impedância em redes de distribuição de energia elétrica considerando diferentes tipos de solo
This work proposes a time-domain methodology to locate high impedance faults in overhead distribution systems. One of the innovative aspects of the method is the proposition of a single mathematical model to represent the different V x I curves generated during a high impedance fault in different ty...
| Autor: | |
|---|---|
| Formato: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2017 |
| País: | Brasil |
| Recursos: | Universidade Federal de Santa Maria (UFSM) |
| Repositorio: | Manancial - Repositório Digital da UFSM |
| Idioma: | portugués |
| OAI Identifier: | oai:repositorio.ufsm.br:1/14035 |
| Acesso em linha: | http://repositorio.ufsm.br/handle/1/14035 |
| Access Level: | acceso abierto |
| Palavra-chave: | Localização de faltas Curtos-circuitos Alta impedância de falta Redes neurais Treinamento continuo Tensão no ponto de falta Fault location Short circuits High impedance fault Neural networks Continuous training Voltage at the Fault point CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA |
| Resumo: | This work proposes a time-domain methodology to locate high impedance faults in overhead distribution systems. One of the innovative aspects of the method is the proposition of a single mathematical model to represent the different V x I curves generated during a high impedance fault in different types of soils. The feeder behavior is modeled by the distance of the fault, the network parameters and the currents and voltages measured at the substation. Therefore, the proposed method does not require the installation of any additional measurement equipment in the network. The feeder capacitances were also considered in the system model, making it closer to a real feeder. Another innovative aspect is the use of an artificial neural network to estimate the unknown parameters of the nonlinear equations that model the feeder behavior during high impedance faults. This network is trained continuously, and only after the fault starts, through the data generated by the own fault. Thus, it is not necessary to simulate several cases for the previous training of the network. The performance of the proposed method was evaluated in IEEE 34 node test feeder through the variation of soil type, fault incidence angles and load feeder. Furthermore, the influence of the current estimation methodology on the fault point was also evaluated. Finally, the performance of the method proposed was compared with another article recently presented. In general, in 86% of the cases tested, the algorithm obtained an error less than 2.5% in the estimation of the fault distance, and the maximum error obtained was 4%. In the comparative analysis with the other method, the proposed algorithm obtained better results in all cases tested, regardless of the soil type in which the fault occurred and its distance. The good results obtained, combined with its simplicity and low cost of implementation, make the method proposed in this work promising for the application in a real feeder. |
|---|