风机电动变桨系统状态特征参量挖掘及异常识别  

Conditions Characteristic Parameters Mining and Outlier Identification for Electric Pitch System of Wind Turbine

作　　者：李辉[1] 杨超[1] 李学伟[2] 季海婷[1] 秦星[1] 陈耀君 杨东[1] 唐显虎 

机构地区：[1]输配电装备及系统安全与新技术国家重点实验室(重庆大学),重庆市沙坪坝区400044 [2]国网重庆市电力公司綦南供电分公司,重庆市巴南区401420 [3]重庆科凯前卫风电设备有限责任公司,重庆市渝北区401121

出　　处：《中国电机工程学报》2014年第12期1922-1930,共9页Proceedings of the CSEE

基　　金：国家国际科技合作专项资助项目(2013DFG61520);重庆市集成示范计划项目(CSTC2013JCSF70003);新世纪优秀人才支持计划项目(NCET-10-0878);输配电装备及系统安全与新技术国家重点实验室自主研究项目(2007DA10512710101)~~

摘　　要：为了提高风电机组电动变桨系统运行状态评估的准确性,提出电动变桨系统状态重要参量挖掘及其异常识别方法的研究。论文在阐述风电机组电动变桨系统的结构及控制原理和监测参数特点的基础上,基于特征参数选择的Relief方法,建立变桨系统特征参量挖掘的数学模型,获取了叶片桨距角、发电机转速及变桨电机驱动电流及其IGBT温度的特征参量,并对其故障状态的漏检率指标进行分析。提出基于多特征参量距离的变桨系统运行状态异常识别方法,建立基于风速输入的变桨系统特征参量的支持向量机回归模型,并对距离阈值进行探讨。最后,对实际变桨系统故障状态的异常识别进行实例验证。实验证明,建立的电动变桨系统状态特征参量挖掘模型的有效性,相比单参数绝对阈值评估方法,基于多特征参量距离的电动变桨系统异常识别方法更能及时、准确地识别其异常状态。In order to accurately assess the operating condition of an electric pitch system （EPS） of a wind turbine generator system, methods of characteristic parameter mining and outlier identification for an EPS were studied. After presenting its structure, control principle and monitoring parameters, mathematical model for characteristic parameters mining for an EPS was presented by using a characteristic parameters selection of the Relief method. Characteristic parameters including pith angle, generator rotational speed, current and IGBT temperature of the pitch drive motor system were acquired by using the presented model, and the results are demonstrated by using an index of undetected error rate for its fault condition. Then, an outlier identification method of operational condition of the EPS was proposed based on a distance method of the selected multi-characteristic parameters. Moreover, support vector machine （SVM） based regression model of characteristic parameters was presented by using wind speed variable as input, and the threshold of the distance was also investigated. Finally, the proposed outlier identification method was demonstrated by using the real operating condition of a practical EPS. The results show thatthe proposed model of characteristic parameter mining for EPS is effective, and compared with the condition assessment method using the single parameter and the absolute threshold, the proposed method based on multi-characteristic parameters distance can more quickly and accurately identify the abnormal condition of the EPS.

关 键 词：风电机组 电动变桨系统 状态评估 特征参量 异常识别 

分 类 号：TM614[电气工程—电力系统及自动化] TM315