Fault Detection Of High Voltage Circuit Breaker Based On Neural Network

With the rapid development of modern society, the importance of the powersystem has become more and more obvious in the national economy and life. Thedemand for electricity sustains growth, which promotes the continuous developmentof power industry. With the enlargement of the grid, the linkages between eachregional power grid become more closely, and the installed capacity and voltage levelincreases. At the same time, the user also put forward higher request to the quality andreliability of the power supply. The normal operation of the high voltage circuitbreaker is related to the security of the whole electric power system, it has importantsignificance to the national production and life. Therefore, with the development ofthe power grid, the fault detection technology of circuit breaker receives more andmore attention. And neural network in fault diagnosis field has outstanding ability, soit has a very in-depth study and application in circuit breaker fault detection.This paper first introduced the types, structure and working mechanism of circuitbreaker, analyzed various monitoring signal and fault information they may contain indetail, laying the foundation for the realization of the high voltage circuit breakerfault detection based on the neural network. Then this paper elaborated the structureand principle of neural network, and analyzed its mathematic model and learningmethod. After that, aiming at the deficiency of neural network in fault diagnosis, twoimproved methods are proposed. One kind is BP neural network based on particleswarm optimization (PSO) algorithm, the other is RBF neural network based onsubtraction clustering algorithm. The two methods are based on two kinds of commonstructure of the neural network, namely back propagation neural network and radialbasis function neural network. With the fusion of new algorithms, the two neuralnetwork both exhibits excellent performance.This study took the MATLAB as the experimental platform, wrote neuralnetwork program based on two new methods, built complete and improved BP andRBF neural network. And this study took the ZN48A-40.5type high voltage vacuumcircuit breaker as the research object, and selected the35sets of input and outputsample as the input of neural network based on the above two methods for simulation.Simulation results showed that both methods have good performance of faultidentification, and the concept of diagnosis accuracy further illustrated the high reliability of diagnosis result vividly. The application of this method is of greatsignificance to improve the maintenance level of circuit breaker and to ensure the safeand stable operation of power system.