摘要
为了提高风电机组变桨系统的故障识别准确率,提出一种通过提取采集数据的特征参数来构造高维特征参数矩阵,高维特征参数矩阵经过LLE降维得到低维特征参数,低维特征参数矩阵作为SVM模式分类输入进行工况识别的方法。该方法在风机故障识别中的准确率为87.50%,正常状态识别准确率为100%,最终的总识别准确率为93.75%。通过理论分析和实验表明:该方法提高了永磁直驱风电机组变桨系统故障识别的准确率。
In order to improve the fault identification accuracy of wind turbines' variable paddle system, a condition identifying method which has the feature parameter extracted from the collected data to construct the high-dimensional feature parameter matrix was proposed. In which, the high-dimensional feature parameter matrix was reduced by LLE to obtain the low-dimensional feature parameters, and the low-dimensional feature parameter matrix was used as the input of SVM pattern classification to identify the working conditions. The recognition rate of this method for wind turbines' faults is 87.50% and the recognition rate at normal state is 100% along with the final total recognition accuracy of 93.75%. The theoretical analysis and experiments show that, this method improves the accuracy rate of identifying the pitch system's faults of the permanent magnet direct drive wind turbine.
引文
[1] 王正宇,张扬帆,段向阳,等.基于Relief算法的风电机组故障特征参数提取方法[J].华北电力技术,2017,(10):57~62.
[2] 李辉,杨超,李学伟,等.风机电动变桨系统状态特征参量挖掘及异常识别[J].中国电机工程学报,2014,34(12):1922~1930.
[3] 李勇,陈贺新,赵刚,等.基于可变k近邻LLE数据降维的图像检索方法[J].吉林大学学报(工学版),2008,38(4):946~949.
[4] 郭锋,刘丽丽,吕凝.基于LLE和SVM的人像识别方法[J].吉林大学学报(信息科学版),2008,26(1):48~54.
[5] 兰兰,陈万忠,魏庭松.基于融合特征提取与LLE方法的表情识别[J].吉林大学学报(信息科学版),2017,35(4):384~391.
[6] 苏盈盈,马飞,刘兴华,等.基于LLE算法和SVM的旋转机械故障诊断[J].重庆电力高等专科学校学报,2013,18(6):63~66.
[7] 李力,李冕,陈法法.基于LLE和LSSVM的滚动轴承故障诊断[J].煤矿机械,2015,36(7):308~310.
[8] 路敬祎.传感器原理及应用[M].哈尔滨:哈尔滨工程大学出版社,2014:145~149.
[9] 王小祥.增量式旋转编码器的简介与应用[J].数字技术与应用,2016,(10):118~119.
[10] 蒋利勇,李伟光.绝对式光电编码器的接口装置设计与应用[J].装备制造技术,2007,(5):62~64.
[11] 夏亮.绝对式编码器总线传输技术研究与实现[D].武汉:华中科技大学,2011.
[12] 王建忠.高维数据几何结构及降维[M].北京:高等教育出版社,2012:203~220.