摘要
BP算法基于梯度下降原理是一种局部寻优算法,在变压器故障诊断应用中网络学习过程收敛速度慢,且易陷入局部极小值。而遗传算法(GA)具有并行计算的特点,可以有效防止搜索过程收敛于局部最优解。将二者结合起来,由GA寻找最优的BP神经网络权值与相应节点的阈值。仿真结果表明:此方法既能快速收敛,又能大大提高避免陷入局部极小的能力,改善了故障诊断的精度和速度。
The BP algorithm is a local optimization algorithm which is based on gradient descent rule.This algorithm apply in transformer fault diagnosis is easy to fall into local minimum and convergence speed is slow.Genetic Algorithm can effectively prevent search process converging local optimal solution,which has characteristics of parallel computation;this paper combines two aspects,using GA to find optimal weights of BP network and corresponding node threshold.The results show that improved algorithm has advantages of fast convergence and avoiding falling into local minimum,and improves accuracy and efficiency of fault diagnosis.
出处
《煤矿机械》
北大核心
2012年第8期257-259,共3页
Coal Mine Machinery
关键词
BP神经网络
遗传算法
变压器
故障诊断
back propagation network
genetic algorithm
power transformer
fault diagnosis
作者简介
张明慧（1987-），女，安徽宿州人，在读硕士，研究方向：智能控制。电子信箱：zmhwk4500@163．com．
