基于神经网络的变压器故障诊断技术优化研究

打开文本图片集

关键词：变压器；概率神经网络；海鸥算法；主成分分析；粒子群算法

中图分类号：TM407； TQ336.4+1 文献标志码：A文章编号：1001-5922（2026）1-0221-06

Abstract：To improve the diagnostic accuracy offaults such asaging andoil leakage of rubber sealing rings inoilimmersed transformers，combined with thecharacteristics of DGA data，thisstudy optimizes the parameters of the Probabilistic Neural Network（PNN）byimproving the Sparrow Optimization Algorithm（SOA）on the basis of feature dimensionality reduction using Kernel Principal Component Analysis （KPCA），and applies theoptimized PNN algorithm to the transformer fault diagnosis model.Theresults show that compared with the traditional KPCAPSO-PNN，KPCA -GWO-PNN and KPCA-SOA -PNN models，the KPCA-ISOA-PNN model constructed in this study has significant advantages in both training accuracyandtraining iteration number.Itis thus indicated that the fault diagnosis model established in this research is feasible and suitable for transformerfault diagnosis.

Key words： transformer;probabilisticneural network ；seagull algorithm；principal componentanalysis；particle swarm optimization

变压器作为当前电力的重要设备，其运行的稳定性成为关键。（剩余5299字）