摘要
为提高风力发电机故障诊断的可靠性,结合小波变换,提出一种遗传算法优化BP神经网络(GA-BP)的综合优化算法。利用单子带重构改进小波变换方法对风力发电机的定子电流信号进行分解与重构,提取准确的特征量;通过遗传算法的选择、交叉、变异等操作和全局寻优特性获取稳定的权值和阈值,赋值于BP网络作为初始值;借助BP网络的自学习、寻优具有精确性等特性不断训练网络;最后以BP神经网络的非线性映射能力完成对风力发电机的故障诊断。算法对比和实例分析表明,该算法对风力发电机的故障诊断有良好的实用性。
A sort of the genetic algorithm BP (GA-BP) neural network method was put to heighten the reliability of fault diagnosis in wind turbines with the wavelet transform. With the wavelet transform improved by single sub-band recon-struction, the signals of the stator current of the wind generator were resolved and reconstructed to extract the precise charac- teristic quantity. The stable weight and threshold were selected as the initial value of BP neural network by the selection, crossover, mutation operator and the global optimum capability of GA. The neural network was training repeatedly with the self-learning and precise optimum characteristic of BP network. The fault diagnosis of the wind generator were completed by the input-output nonlinear mapping ability of BP neural network. The algorithm comparison and real case analysis show that the algorithm has a good practicability in the fault diagnosis of the wind generator.
出处
《微特电机》
北大核心
2013年第5期4-8,共5页
Small & Special Electrical Machines
基金
国家自然科学基金项目(60904078)
广东省教育厅专项重点实验室项目(IDSYS200701)
广东高校优秀青年创新人才培养计划项目(2012LYM_0052)
广州市科技攻关项目(11A52081158)
广东省省部产学研(2011B090400046)
关键词
风力发电机
BP神经网络
遗传算法
故障诊断
单子带重构
小波变换
wind turbine
BP neural network
genetic algorithm
fault diagnosis
single sub - band reconstruction
wavelet transform