自适应因子神经网络的变压器故障诊断研究  

Research on Power Transformer Fault Diagnosis Based on Adaptive Factor Neural Network

作　　者：代艳霞[1] 王洪益[1] 伍倪燕[1] DAI Yan-xia WANG Hong-yi WU Ni-yan(The Department of Modern Manufacturing, Yibin Vocational and Technical College, Sichuan Yibin 644003, Chin)

机构地区：[1]宜宾职业技术学院现代制造系,四川宜宾644003

出　　处：《机械设计与制造》2017年第7期175-178,共4页Machinery Design & Manufacture

基　　金：宜宾职业技术学院院级项目(ybzysc15-40)

摘　　要：电力变压器是是电力系统的核心设备。为预防并降低电力变压器发生故障概率,设计了小波神经网络对电力变压器进行故障诊断。为提高迭代计算速度及计算精度,提出一种基于自适应修正因子的模型优化方法,通过自适应修正因子可以忽略模型中的局部极值,进而消除微小变化特性,排除杂波干扰。基于自适应修正因子设计变压器故障诊断小波神经网络模型训练方法,从而提高迭代计算效率及精度。通过与传统的神经网络模型及粒子群小波神经网络的故障分析结果及误差对比分析,验证所设计的电力变压器故障诊断模型具有较高的可用性。研究结果为电力变压器故障诊断分析提供理论基础。The paramount apparatus in power system is power transformer. In order to prevent and reduce the probability of power transformer failure, wavelet neural network is appraved for fault diagnosis of power transformer. For the sake of the improvement of iterative calculation speed and precision, a model optimization method based on adaptive correction factor is presented in this paper. Local minimum in model can be neglected through adaptive correction factor, and the characteristics of subtle changes can be further eliminated. Wavelet neural network training method of fanlt diagnosis on power transformer based on adaptive correction factor is designed and presented, which advances efficiency and precision of iterative calculation. The proposed model of power transformer fault diagnosis is proved to be of high availability through comparison of fault error and analysis of fault results, which are based on fanlt results and error of traditional neural network model and particle swarm optimization wavelet neural network. Theoretical basis for fault diagnosis of power transformers can be provided by the research results.

关 键 词：自适应修正因子 小波神经网络 电力变压器 故障诊断 迭代计算 

分 类 号：TH16[机械工程—机械制造及自动化]