摘要
为了减少风力发电机组齿轮箱故障,确保风电机组持续安全运行,对风电机组运行监控数据在线分析,提出一种结合最小二乘支持向量机(LSSVM)的风机齿轮箱统计过程控制故障预测方法。该方法以支持向量机学习风电机组的正常状态运行模式,利用风电机组实时运行数据来估计正常状态下该时刻齿轮箱油温度和齿轮箱轴承温度,并与实际温度测量值进行比较。随后利用统计过程控制技术分析齿轮箱油温和轴承温度的实际值与估计值的残差,以实现齿轮箱异常状态的预测。
In order to reduce the gearbox failure of wind turbine and ensure a consistently secure operation of wind turbine units, the condition monitoring data of wind turbine operation are analyzed online, and a method of statistical process control (SPC) for failure prediction of wind turbine gearbox based on least squares support vector machines (LSSVM) is proposed. The model for normal behaviour of gearbox is built up by LSSVM, and the real-time data of wind turbine operation are used to predict the temperature of gearbox oil and bearing at certain time in normal condition, and then the prediction data are compared with the actual data. Then, the residual error between predicted values and the actual values is analyzed by SPC. According to the results from SPC analysis, some failures about gearbox can be predicted.
出处
《电力系统保护与控制》
EI
CSCD
北大核心
2012年第13期67-73,共7页
Power System Protection and Control
基金
河北省自然科学基金(E2010001694)~~
关键词
风电机组
齿轮箱
最小二乘支持向量机
统计过程控制
故障预测
wind turbine
gearbox
least square support vector machines
statistical process control
failure prediction