基于BP神经网络及蜂群算法的变压器故障诊断  

Transformer Fault Diagnosis Based on BP Neural Network and Artificial Bee Colony Algorithm

作　　者：刘畅[1,2] 吴艳娟[1] 高雅琦[2] LIU Chang;WU Yan-juan;GAO Ya-qi(Tianjin University of Technology,Tianjin 300384;State Grid Tianjin Electric Power Corporation Maintenance Company,Tianjin 300232)

机构地区：[1]天津理工大学,天津300384 [2]国网天津市电力公司检修公司,天津300232

出　　处：《新型工业化》2020年第4期7-12,共6页The Journal of New Industrialization

摘　　要：变压器安全运转夯实了整个电网的根基,本文提出BP神经网络综合蜂群算法的方式诊断变压器故障。利用蜂群算法全局多点搜索功能进行不断优化,找寻最佳解作为BP变压器故障诊断网络中各层的连接权值以及阈值,并构建BP及ABC变压器故障诊断模型。依托实际变电站变压器的绝缘油所含56组气体浓度数据作为诊断依据进行深入探究,仿真结果显示,BP及ABC算法正确率达92.86%,远高于传统BP的64.29%,本研究提升了判别关于油浸式变压器类一次设备所存在故障的精确度。The safe operation of the transformer has consolidated the foundation of the entire power grid.This paper proposes the method of back propagation neural network combined with artificial bee colony algorithm to diagnose transformer fault.Use the global multi-point search function of artificial bee colony algorithm to optimize continuously,find the best solution as the connection weight and threshold value of each layer in BP transformer fault diagnosis network,and build BP and ABC transformer fault diagnosis model.In this paper,56 groups of gas concentration data contained in insulating oil of actual transformer substation are used as the basis of diagnosis,and compared with the diagnosis of transformer fault only through BP mode.The simulation results show that the accuracy of BP and ABC algorithm is 92.86%,which is much higher than that of BP network’s 64.29%.This paper improves the accuracy of identifying faults in oil immersed transformers.

关 键 词：变压器 故障诊断 BP神经网络 蜂群算法 

分 类 号：TM614[电气工程—电力系统及自动化]