The Method Of Partial Discharge Pattern Recognition Based On T-S Model Fuzzy Neural Network

Partial discharge is important parameter for evaluating insulation ageing of high voltage electric appliance and equipment. Operating reliability of high voltage electric appliance and equipment depends on reliability of insulation, so the researching of partial discharge pattern recognition supplies reference data for on-line monitoring and fault diagnosis of high voltage electric appliance.In this paper, after reading lots of literature home and abroad, fuzzy logical and neural network are combined and gathered merits of each network to find a new effective method for pattern recognition-fuzzy neural network (FNN), and then the new method is used to partial discharge pattern recognition. Firstly, three models of partial discharge for different electric field are designed and made, and then experiment data of partial discharge are collected. After researching and analyzing T-S model of fuzzy neural network topological structure, T-S model of fuzzy neural network is applied in partial discharge pattern recognition for the first time, and then compared pattern recognition of different membership function and parameters to seek a befitting structure of network. Secondly, to integrate the characteristic of partial discharge and fuzzy neural network, fractal characteristic and original characteristic are extracting separately as the input characteristics of fuzzy neural network. At the same time, fuzzy hierarchical clustering and fuzzy equivalence matrix method based on fuzzy logical are designed for partial discharge pattern recognition as contrasting experiment.Fuzzy neural network based on T-S model can overcome many problems and limitations of the traditional expert system. This method also combines the advantage of neural network such as parallel processing, distributed memory, self-learning and self-organization with the advantage of fuzzy system in solving the uncertainty fuzzy problem. Then compared the method with fuzzy cluster, a new pattern recognition method with flexible and high astringency is found for partial discharge.