基于SOM神经网络的电力变压器故障诊断方法  

Fault Diagnosis Method of Power Transformer Based on SOM Neural Network

作　　者：严太山 赵琦旭[1] 刘博文 李文彬 YAN Tai-shan;ZHAO Qi-xu;LIU Bo-wen;LI Wen-bin(School of Information Science and Engineering,Hunan Institute of Science and Technology,Yueyang 410019;Key Laboratory of Hunan Province on Intelligent Control and Optimization of Complex Industrial Logistics System,Yueyang 410019)

机构地区：[1]湖南理工学院,信息科学与工程学院,岳阳410019 [2]湖南理工学院,复杂工业物流系统智能控制与优化湖南省重点实验室,岳阳410019

出　　处：《现代计算机》2021年第5期33-36,共4页Modern Computer

基　　金：湖南省自然科学基金项目(No.2017JJ2107、2018JJ2152)。

摘　　要：针对电力变压器故障诊断问题,提出一种改进的SOM神经网络算法(ISOM算法)。该算法根据SOM神经网络权值增量的变化情况对收敛条件进行自适应判断,确保网络的训练速度和精度。将ISOM算法应用于电力变压器故障诊断,以特征气体法为基础对油浸式变压器进行故障诊断实验。仿真结果表明,ISOM算法的训练速度快,训练精度高,对提供的电力变压器各类故障数据诊断准确,该方法对实现电力变压器故障诊断的智能化具有重要参考价值。Aiming at the problem of power transformer fault diagnosis,an improved SOM neural network algorithm(ISOM algorithm for short)is pro⁃posed.The algorithm determines the convergence conditions dynamically according to the weight increment,which ensures the training speed and accuracy of the network.In this paper,ISOM algorithm is applied to power transformer fault diagnosis,and fault diagnosis ex⁃periment of oil immersed transformer is carried out based on characteristic gas method.The simulation results show that the training speed of ISOM algorithm is fast and the training accuracy is high,which can realize the accurate diagnosis of power transformer fault.This meth⁃od has important reference value for realizing the intelligent fault diagnosis of power transformer.

关 键 词：SOM神经网络 电力变压器 故障诊断 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM41[自动化与计算机技术—控制科学与工程]