Power Transformer Fault Diagnosis Based On Improved Hybrid Leap-frog Algorithm Optimized Support Vector Machine

Power transformer is an indispensable equipment in the process of transportation power in China,which is related to the normal operation of the power system.To ensure the stable operation of the power grid,it is necessary to repair the latent faults of the transformer in time to avoid further accidents.Dissolved gas analysis（DGA）in oil is an extremely effective and common diagnostic method in the field of transformer fault diagnosis.Traditional fault diagnosis methods such as characteristic gas identification method and three-ratio method are all based on the method of dissolved gas in oil.At present,with the development of artificial intelligence technology,transformer fault diagnosis methods develop thowards intelligence through these technologies,which shows better diagnostic performance than traditional diagnosis methods.The amount of transformer fault data is small,and the support vector machine（SVM）can still have better performance characteristics in the case of small samples,which is just suitable for transformer fault diagnosis scenarios.Therefore,the classifier in the fault diagnosis model constructed in this thesis adopts the SVM,the input of the model is the data of dissolved gas in the oil,and the classification result is used as the output.Among them,the classification performance of the SVM is closely related to the parameter combination of the kernel function and the penalty factor.Therefore,the key to constructing a fault diagnosis model with high diagnostic accuracy lies in obtaining the optimal parameter combination.In this thesis,an improved hybrid leapfrog algorithm is proposed to obtain the optimal parameters,and the fault diagnosis model is constructed through the parameters.The standard hybrid leapfrog algorithm is easy to fall into the local optimum in the iterative process,and the optimization accuracy is low.In view of this shortcomings of the standard hybrid leapfrog algorithm,a cross-subgroup communication strategy is proposed to optimize the local search formula,and the position information of the best frog in the adjacent subgroup is introduced into the position update formula of the worst frog to improve the search space.Moreover,the optimal group self-mutation strategy is introduced,and the mutation operation is performed on the optimal group before the position of the worst frog is updated to enhance its guiding ability and reduce the possibility of the algorithm falling into the local optimum.Thus,the improved hybrid leapfrog algorithm is used to optimize multiple standard test functions,and compared with the standard leapfrog algorithm and other improved strategies.The comparison results show that the improved algorithm performs better in terms of convergence speed and optimization accuracy.Through the collected gas data and the improved leapfrog algorithm proposed in this thesis,the parameters of the SVM are optimized,and the optimal parameter combination is obtained to build a transformer fault diagnosis model for performance testing.The comparison of the results show that the fault model classification performance of the SVM optimized by the improved leapfrog algorithm proposed in this thesis is better than that of the particle swarm optimization（PSO）and the genetic algorithm（GA）.Finally,combined with the actual production situation,an online fault diagnosis platform is built through code to provide a visual and information-based diagnosis tool for the staff.