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

During the operation of the transformer,the content of gas in the oil is easy to detect and can accurately reflect the operating state of the power transformer.The dissolved gas analysis technology in the oil can realize real-time online monitoring without power failure.It is one of the most effective means of transformer fault diagnosis..In the field of transformer fault diagnosis,artificial intelligence optimization algorithms are used to build models.The bat algorithm is a new type of intelligent optimization algorithm,but the algorithm still has the problems of slow convergence in the early stage and easy to fall into local optimum.Aiming at the shortcomings of traditional bat algorithm,such as slow convergence speed,low solution precision and easy to fall into local optimum,this paper proposes a method to introduce the self-learning factor and introduce the proportional weight coefficient in the position update formula to the traditional bat algorithm.Improvement,and then through a series of basic test functions to simulate the improved bat algorithm before and after,and finally based on the simulation results for solution accuracy and convergence analysis.The simulation results show that the improved bat algorithm is superior to the traditional bat algorithm in solving accuracy and convergence.In this paper,according to the support vector machine,it can solve the problems of nonlinear,high-dimensional,local minimum and small problems in the case of small samples.A transformer fault diagnosis model using support vector machine as classifier is proposed.Since the fault diagnosis result is mainly determined by the parameters of the support vector machine and the current method of obtaining the parameters is conservative,in order to obtain the best support vector machine parameter model,this paper also proposes to improve on the basis of the traditional bat algorithm.The latter algorithm optimizes the two main parameters affecting the classification accuracy of support vector machine,and optimizes it through the Wine dataset in UCI.The method of optimizing support vector machine proposed in this paper has certain superiority in accuracy.Finally,the original DGA is used for training and prediction.The simulation results show that the transformer fault diagnosis results based on the improved bat algorithm optimization support vector machine are better than those obtained by the grid search method,genetic algorithm and particle swarm optimization algorithm.