Fault Diagnosis Of Transformer Based On Oil Chromatogram By SVM

The safe operation of transformer is very important for the stability and reliability of power system.The transformer is an important part of the power system,and its voltage conversion function is very important for the different power demand of the users.If the transformer is damaged,it may lead to the interruption of the power system,resulting in a large area of power failure,which will badly affect social stability and economic development.It is self-evident that ensuring the long-term stable operation of transformer is important for the stability of power system.Aiming at the problem of transformer fault diagnosis and the problem of insufficient transformer fault data,this paper proposes a transformer oil chromatogram fault diagnosis method based on improved particle swarm optimization of SVM parameters and a transformer negative sample generation method based on transformer mechanism,which to some extent solves the problem of low accuracy of transformer fault diagnosis and insufficient fault data available for research.Among the methods of transformer fault diagnosis,gas in oil analysis(DGA)is a method that takes into account both efficiency and cost.However,the DGA method also has defects such as insufficient accuracy of fault classification.In view of the above defects,this paper proposes a Multi-population Particle Swarm Optimization(MPNLPSO)algorithm based on neighborhood learning,and uses the simplified algorithm to optimize the classification parameters of SVM.Finally,SVM completes the fault diagnosis based on the data of the gas dissolved in oil.The experimental results show that on the basis of MPNLPSO’s high efficiency,performance and robustness,the classification parameters of SVM have been optimized,and its classification accuracy has been improved to a certain extent.Multi-population particle swarm optimization based on neighborhood learning(MPNLPSO)is the focus of this paper.Traditional particle swarm optimization algorithms usually ignore the important information of the experience obtained by the optimal particles in the neighborhood.Therefore,in order to fully utilize the experience obtained by the optimal particles in the neighborhood,this algorithm first uses a small world network to construct a neighborhood for each particle,and then completes the neighborhood learning strategy;Next,in order to bring the particles that have fallen into search stagnation back into the effective search process,a learning strategy based on dynamic opposition will be introduced;Combined with the random disturbance strategy,the accuracy and robustness of the algorithm have been improved to a certain extent.The optimization of SVM based transformer fault diagnosis using improved particle swarm optimization algorithm is an important application scenario for the research content of this article.The classification performance of support vector machines largely depends on the penalty parameters and kernel parameters of the model.Choosing appropriate classification parameters based on actual data samples is the main research problem of SVM.And parameter optimization is precisely the advantage of particle swarm optimization algorithm.On the premise of retaining the optimization ability of particle swarm optimization algorithm,the algorithm proposed in this paper is optimized accordingly and used to find the optimal SVM parameters.Finally,fault diagnosis based on transformer oil chromatogram data is realized through the SVM,which basically achieves the goal of accurately identifying transformer faults.