Research on transformer fault diagnosis based on sparrow algorithm optimization probabilistic neural network

Abstract

∗To solve the problems such as the inability to monitor the operation of oil-immersed power transformers in real-time, low accuracy and slow speed of fault diagnosis, a fault diagnosis based on a sparrow algorithm (SSA) optimizing the probabilistic neural network (PNN) was proposed. PNN with its some advantages such as simple training, additional sample ability is widely used in multiple fault diagnosis field, but smooth factors of PNN optimization is a difficult problem, this article uses three ratio methods to deal with the collected raw dissolved gas analysis (DGA) data, optimize the input data, then SSA is used to optimize the smooth factor of PNN, to establish an optimized fault diagnosis model SSA-PNN. The simulation results show that compared with the network before optimization and the traditional back propagation (BP) neural network, the accuracy of transformer fault diagnosis is significantly improved and the convergence speed is faster.