基于LM算法建立风电机组神经网络故障预警诊断模型  

ESTABLISHMENT OF FAULT WARNING DIAGNOSIS MODEL FOR WIND POWER UNITS BASED ON LM ALGORITHM OF NEURAL NETWORK

作　　者：巨林仓[1] 史贝贝[2] 杨清宇[2] 宋德宽[1] 

机构地区：[1]西安交通大学能源与动力工程学院,陕西西安710049 [2]西安交通大学电子与信息工程学院,陕西西安710049

出　　处：《热力发电》2010年第12期44-49,共6页Thermal Power Generation

基　　金：陕西省自然科学研究计划项目(2007E236)

摘　　要：基于人工神经网络LM算法,建立了风力发电机组故障预警诊断模型。神经网络LM算法是一种BP的改进算法,但同时也存在易陷入局部极小的问题,对此采用施加动量使其跳出局部极小的方法,取得了良好的效果。应用Matcom工具实现VC++与Matlab软件混合编程的方法,解决了算法程序化过程中遇到的复杂矩阵运算问题,提高了算法性能。将改进后的算法用于结构为15-22-4的风电机组故障预警诊断模型训练和检验,结果证明网络收敛性能良好。该算法模型已嵌入风电机组故障预警诊断软件中。Based on LM algorithm of an artificial neural network,a model for fault warning diagnosis of wind power units has been established.The LM algorithm of neural network is an improved BP algorithm,but the problem of easily falling into local minimum is still existing.For this,a momentum method has been applied to bounce out from the local minimum,achieving good effectiveness.By using Matcom tool,the hybrid programming method between VC＋＋ and Matlab softwares has been realized,the problem of complex matrix calculation encountered in the process of algorithm programming being resolved,enhancing the performance of algorithm.The improved algorithm has been used in training and inspection of the fault warning diagnosis model for a wind power unit with structure of 15-22-4,result shows that the converge performance of the network is good.The model of said algorithm has already been embedded into software of fault warning diagnosis for wind power units.

关 键 词：风力发电 LM算法 局部极小 故障预警诊断 模型 Matcom软件 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]