Research On Power Transformer Fault Diagnosis Based On Bat Algorithm Optimized Support Vector Machine

Power transformer is the key electrical equipment in power system.It is an effective method to find out the overheating or discharge fault in transformer in time by analyzing dissolved gas in oil.At the same time,it has a certain role in ensuring the normal operation of the transformer,intervening and preventing various possible faults of the transformer in advance and improving the power supply reliability of the power system.Through the deep research of a large number of intelligent diagnosis methods,in order to solve the problem that ratio judgment is difficult to deal with the problem of the feature boundary is not obvious,and to make up for the problem that the algorithm is difficult to fit under the condition of less data,this thesis uses bat algorithm to improve the support vector machine to diagnose the transformer fault.The combination of support vector machine(SVM)and artificial intelligence can effectively solve the problem of small sample,high dimension and nonlinear classification.Compared with the fixed parameter and particle swarm optimization support vector machine,it has a better advantage in fault diagnosis accuracy.Through the comparison and analysis of the methods of commonly used SVM classification,the tree structure is used to construct multiple two classifiers to realize the diagnosis of various faults.According to the relevant documents and guidance instructions,the input characteristic quantity and fault type of this model are divided.According to the analysis of the example data,it is proved that the output type and the selected input can establish the relation.By comparing,Gauss kernel function is used to map the input feature in high dimension,and the important parameters affecting classification are determined by reference grid search.At the same time,the selection of parameters is changed from manual fixed to dynamic adjustment based on sample data and different tasks.The example shows that the intelligent algorithm is helpful to improve the accuracy of fault diagnosis.In order to improve the anti-interference ability of bat algorithm,to find the parameters of SVM which is most conducive to classification,to improve the accuracy of diagnosis,the speed and displacement updating mode of bat algorithm are improved.The inertia weight is used to improve the speed update mode of bat algorithm,which is conducive to the transition between global optimization in the early stage and local optimization in the late stage of the algorithm.The heavy tail feature of Levy flight is conducive to the algorithm jumping out of the local optimal solution.The convergence analysis of the optimized bat algorithm is carried out.Through the optimization calculation of the standard test function,the optimized algorithm is more stable and closer to the global optimal value than the algorithm before optimization,particle swarm optimization and genetic algorithm,and has a great advantage in convergence speed.The SVM based on binary tree is used to realize multi classification model.Bat algorithm with inertia weight and Levy flight characteristics is used to calculate the variables of SVM.Comparing the other two parameter selection methods,the fault diagnosis model is established and verified by combining sample data.The verification table is compared with the multi classification SVM model optimized by artificial parameter setting and particle swarm optimization.The fault diagnosis model of transformer is more accurate than other kinds of models,and it can reduce the error judgment when processing complex samples.