Research Of Transformer Fault Diagnosis Based On Improved Neural Network

Transformer is an important part of power system and its operation state has a great significance for power system stability, safe operation. Researching methods of the transformer fault diagnosis and strengthening the operation of the transformer maintenance, can effectively reduce safety accident caused by the hidden trouble.Dissolved Gases Analysis(DGA) is one of the main technology methods to diagnose the transformer internal fault. Artificial Neutral Network has many advantages, such as parallel distributed processing, self-adapted ability, memory etc. It can accurately express the complex mapping relationship between the transformer internal fault and the dissolved gases in transformer. BP algorithm is a local optimization algorithm which is based on gradient descent rule. This algorithm is easy to fall into local minimum and convergence speed is slow. Particle swarm optimization algorithm can effectively prevent search process converging local optimal solution, and have a high speed of convergence process.In this paper,based on the characteristics of the transformer internal fault type and the dissolved gases in transformer,a transformer fault diagnosis system is formed by neural network and particle swarm optimization algorithm.The result of experiments shows that compared with the Levenberg-marquarat BP algorithm,the training epochs are decreased. The result shows that the algorithm can effectively improve the convergence speed and reduce the training time of neural network. By collecting a large number of representative training samples, the fault diagnosis accuracy of system is above 92%.It shows that the system is proper, feasible and accurate for the transformer fault diagnosis.