Transformer Fault Diagnosis Based On SSA-PNN And PCA Feature Reconstruction

Transformers play a key pivotal role in the power grid.Once there is a fault,there may be a regional power outage,which will bring great troubles to people’s production and life.Therefore,it is imperative to develop a set of stable and reliable transformer fault diagnosis system to diagnose potential transformer faults quickly and accurately,to ensure the safe operation of the transformer,and to effectively contain the power accidents caused by transformer faults in time.This thesis presents a transformer fault diagnosis method based on Sparrow Search Algorithm(SSA)and Probabilistic Neural Network(PNN).Firstly,in view of the problems of the traditional three-ratio method,such as missing coding and too absolute coding limit,the uncoded ratio method is used to process the original fault gas data.Although the uncoded ratio method makes the fault information more complete,it also causes the input of the model to increase from 3-dimensional information to 9-dimensional information in the original IEC three-ratio method,resulting in large information redundancy in the fault data,which increases the calculation difficulty of the model.To solve this problem,Principal Component Analysis(PCA)was used to successfully reduce the information dimension to 5 dimensions,which effectively reduced the information redundancy.Secondly,the probabilistic neural network is applied to transformer fault diagnosis,considering that the smoothing factor is an important parameter that directly affects the accuracy of PNN judgment,and it is often randomly assigned by people,which leads to the low prediction accuracy of PNN model and the model can’t run in the best state.To solve this problem,this thesis uses sparrow search algorithm to optimize the smoothing factor parameters of PNN,selects the best state value to assign to PNN model,and finally builds a transformer fault diagnosis model based on PCA-SSA-PNN.In order to verify the accuracy of the transformer fault diagnosis model based on PCA-SSA-PNN,the same sample data were selected to test the classical PNN,PCA-PSO-PNN and PCA-SSA-PNN optimized by particle swarm optimization.The results show that the transformer fault diagnosis model based on PCA-SSA-PNN has higher diagnosis accuracy and better fault classification effect,which provides a good guarantee for the safe operation of oil-immersed transformers.