摘要
复杂系统变压器的油中溶解气体分析是变压器绝缘寿命估计和绝缘故障诊断的重要依据,针对传统BP算法收敛速度慢,学习效率低等缺点,运用灰色系统理论,通过缩小神经网络输入样本的灰色区间,避免输入样本矢量的无限制增长,提高了神经网络学习性能,加快了网络的收敛速度;在此基础上,提出了一种基于灰色区间神经网络的变压器故障诊断模型;实例诊断结果表明,该模型能够快速找出故障类型而且能确定故障部位,具有很高的故障诊断率,并且大大提高了收敛速度,验证了其有效性。
Dissolved gas analysis is an effective way to evaluate the transformer insulation and prevent the transformer from further dete- rioration, according to the low speed of convergence and inefficient learning of traditional BP neural network, grey system theory is applied to reduce grey area of neural network input samples, avoid input samples vector increase without limit, and enhance learning ability of neural network; Through optimizing standard BP algorithm, the speed of network convergence is accelerated. Based on these works, a power trans- formers fault diagnosis model based on grey area neural network is proposed. Experimental results show that this model can find out fault types rapidly and locate fault exactly, thus achieving a high diagnostic rate and converging very rapidly, and prove its validity.
出处
《计算机测量与控制》
北大核心
2014年第3期664-666,共3页
Computer Measurement &Control
基金
国家自然科学基金(61001023)
关键词
神经网络
灰色区间
故障诊断
变压器
neural network
grey area
fault diagnosis
transformers
