The Format Criterion Of Master Degree Thesis Of XIHUA

Transformer fault diagnosis and prediction technology is one of the important technical means to guarantee safe operation of power transformer, in this paper, the fault diagnosis model and fault forecasting model are proposed by the advantage of support vector machine,which is support vector machine can still solve nonlinear, high dimension, the local minimum points problems in the small samples. Because the choice of parameters influence fault diagnosis, therefore, it is proposed that the particle swarm algorithm is used to optimize the parameters of support vector machine. Through the simulation experiments show that the transformer fault diagnosis and forecast rate had increased. In this paper, the innovation points are expressed as follows:Firstly, in the fault diagnose model of transformer, the gauvey question is saved through the core functions, in the three core functions, the radial core function is better than the polynomial function and nerve network core functions of gender through the instance emulation and analysis .Secondly, in the fault diagnose model of transformer, the punishing factorC and nuclear function parameterσis getted by particle swarm optimize support vector machine, in order to improve the optimize ability of the particle swarm algorithm, the three way improvement of the particle swarm algorithm is proposed, that is dynamic adjustment of inertia weightω,the convergence factor and v_max, and analyze the convergence of the standard particle swarm algorithm,basic particle swarm algorithm.Thirdly, in the fault diagnose model of transformer, the the classification model of support vector machine based on the binary tree is established, the parameters of support vector machine are optimized by the particle swarm algorithm of inertia weightω.finally, the simulation results of classification C ? SVC algorithm,V ? SVC algorithm,IEC three ratios and the nerve network algorithms are comparised through the instance emulation and analysis.The last but not least, in the fault prediction model of transformer, the model based on support vector machine regression algorithm and time series is established, the embedding dimension m of the training model is determined by the MSE minimum standard, the punishment factorC,nuclear function parameterσand insensitiity loss function parameterεare determined by the particle swarm algorithm optimize support vector machine, the predictive accuracy of transformer is evaluated by the average absolute value relative error. Finally, The transformer fault forecasting model based on particle swarm optimize support vector machine is compared with grey forecasting model.