Fault Diagnosis Of Power Transformers Based On Membrane Computing

Power transformers are key hub elements in power system. Timely and accuratelygrasping of the operational status and fault of the transformer, and being able to takecorresponding measures timely and effectively is of great significance to the security, stabilityand economic operation of the power system.Based on the relationship of power transformer oiled gas and fault types, a faultdiagnosis method combining membrane computing and BP algorithm is proposed, with BPalgorithm having the defect of converging slowly and easily running into local minimumpoint. The method uses the function of overall optimizing of membrane computing, tooptimize the initial weights and threshold of BP neural network, and uses the function ofstudying and classifying of BP algorithm to achieve the right classification of transformerfault.The main work of this paper is as follows：(1) According to the cell membrane characteristics of penetrating, cell membraneoptimization algorithm uses rules of selection, crossover and rewrite, and introduce the rulesof communication at the same time，which enrich the algorithm structure, strengthen thealgorithm function,and improve the algorithm convergence capability.(2) Tissue membrane optimizing algorithm with wheel topology is used in this paper.Because of cell’s shrink and diastole functions, the rules of shrink and diastole are led in tospeed up the convergence of the algorithm. At the same time, rules such as transposition,crossover, rewrite, communication and so on, are involved.(3) On account of the problem of parameter setting uncertain, in the cell membrane andtissue membrane optimization algorithm, colony entropy is led in. Function Sigmoid hassmoother top and bottom than function Cosine. It appears better balance feature betweenlinearity and nonlinearity. And lead it to evolution generation, combine it and colony entropyto self-adoptive strategy to adjust the probability of crossover, rewrite and transposition tochange according to algorithm evaluating.(4) Combining membrane optimization algorithm with BP neural network，a newmethord to transformer fault diagnosis is proposed and proved effectively by living examples.(5) A methord to transformer fault diagnosis, combining membrane optimizationalgorithm and BP neural network, is proposed and proved effectively by living examples.(6) Compare the application performance in transformer fault diagnosis of adaptiveoptimization algorithm of cell membrane and adaptive optimization algorithm of tissuemembrane, and contrast them to BP neural network. The result proves that two powertransformer fault diagnosis methods based on membrane computing differ to different types of fault in sensitive degree. But both keep high diagnosis accuracy rate. And the convergenceperformance and diagnostic accuracy are higher than that of pure BP algorithm model.