基于BP神经网络的风电系统控制器I/O硬件故障自诊断方法  

Self-detecting method for faults in wind turbine controller I/O hardware based on BP neural network algorithm

作　　者：姜寅 任荭葳 陈凯 朱长江 JIANG Yin;REN Hongwei;CHEN Kai;ZHU Changjiang(Zhejiang Windey Company Limited,Hangzhou 310012,China;Key Laboratory of Wind Power Technology of Zhejiang Province,Hangzhou 310012,China)

机构地区：[1]浙江运达风电股份有限公司,杭州310012 [2]浙江省风力发电技术重点实验室,杭州310012

出　　处：《华电技术》2020年第5期55-60,共6页HUADIAN TECHNOLOGY

基　　金：浙江省重点研发计划项目(2019C01050)。

摘　　要：随着风电产业竞争日趋白热化,市场对整机制造商产品交付和运维服务质量的要求不断提高。针对批量库存控制器个别I/O硬件故障影响项目生产维护的问题,介绍一种基于反向传播(BP)神经网络算法的低成本、适用于生产运维现场使用的控制器I/O硬件故障的自诊断方法。通过输入随机顺序故障样本数据集对神经网络模型进行训练,利用待测I/O硬件和继电器构造的自诊断电路采集信号,将其处理为归一化数据的特征矩阵,输入自诊断模型进行故障识别和分类,最终输出参考结果。试验初步验证了该方法在不依赖专用设备的情况下,可有效识别I/O硬件信道故障,具有实际应用价值。With the fierce competition in wind power industry,market′s requirements for the product delivery,operation and maintenance service provided by machine manufacturers are increasingly demanding.A low-cost fault self-detecting method suitable for controller Input and Output(I/O)hardware was proposed based on BP neural network algorithm,in order to mitigate the negative impacts of certain defective I/O hardware of controllers′batch inventory on the operation and maintenance service of the project.The neural network model was trained with the random sequence of fault sample data set.The self-detecting circuit constructed with I/O hardware and relays is designed for collecting sampling signals which are processed into a characteristic matrix of normalized data.Putting the data into the self-detecting model,reference results will be output after fault identification and classification.The experiment preliminarily verified that the method can effectively identify the I/O hardware fault without specified equipment,which is of practical application values.

关 键 词：BP神经网络 故障诊断 控制器 I/O硬件 特征矩阵 风电控制系统 自诊断 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TK32[自动化与计算机技术—控制科学与工程]