Transformer Fault Diagnosis Based On Whale Algorithm Optimized PNN

Transformer is the core equipment used for voltage level transformation in power system,and its working condition will directly affect the normal operation of the whole power system.On the basis of analyzing the current transformer fault diagnosis methods and the problems faced,this paper deeply understands the application of PNN probabilistic neural network in fault diagnosis.Aiming at the influence of its network parameter smoothing factor on the fault diagnosis ability,this paper proposes a fault diagnosis method to optimize the Probabilistic Neural Network(PNN)through improved Whale Optimization Algorithm(WOA).First of all,by analyzing several common fault types of transformers in practical applications and the basic principle of dissolved gas analysis in oil,aiming at the problems of traditional fault diagnosis methods such as three ratio method,such as missing codes or too absolute boundary conditions,PNN neural network in artificial intelligence technology is used for diagnosis.Through experimental simulation analysis,it is found that PNN neural network has ideal classification effect,The application of intelligent algorithm optimization PNN neural network in transformer fault diagnosis has certain advantages.Secondly,in view of the influence of parameter smoothing factor of PNN neural network on the network fault diagnosis accuracy,the whale optimization algorithm proposed in recent years is used to optimize the smoothing factor,and the gray wolf algorithm,which is widely used,is introduced for experimental simulation and comparative analysis,which proves that the algorithm has strong optimization ability,and thus verifies the fault diagnosis ability of WOA-PNN model.Finally,because the random population initialization of the whale optimization algorithm can not effectively obtain all the relevant information in the space,which leads to insufficient global search capability,the initialization strategy combining Skew Tent chaotic sequence and Logistic chaotic sequence is adopted,and the higher search efficiency of Skew Tent chaotic map is used to generate the optimal initialization population,and then the non repeatability of Logistic chaotic map is combined to ensure the diversity of the population,So as to strengthen the global search ability of the algorithm;The linear convergence factor makes the algorithm can not be reduced quickly in the early stage,so that the whale individual can not enter the local development stage quickly.The nonlinear improvement strategy is adopted to ensure a fast convergence speed in the whole process.Six different test functions are used to test and compare the algorithms,and a transformer fault diagnosis model is established for experimental simulation and comparative analysis,which verifies the effectiveness of IWOA-PNN model.