基于改进集成KNN回归算法的风电机组齿轮箱状态监测

Wind Turbine Gearbox Condition Monitoring Based on Improved Ensemble KNN Regression Algorithm

风电机组齿轮箱是容易发生故障的重要部件,维修费用高昂,因此有必要对其进行实时状态监测。针对集成K近邻(KNN)算法对随机采样不敏感的问题,提出了一种基于规则采样的改进集成KNN模型。首先利用距离相关系数进行变量选择,然后基于正则化互信息对变量进行排序,将其用于规则采样,构造子训练集,最后基于统计过程控制方法设置预警阈值对实时残差进行分析,根据健康度曲线对风电机组齿轮箱健康度进行监测,并利用某风电机组实际数据对所提方法进行验证。结果表明:所提方法显著提升了模型估计精度,该模型优于常规集成KNN模型,可以实现齿轮箱的早期故障预警。

Wind turbine gearbox is an important component that is prone to failure,and its maintenance cost is high,so it is necessary to monitor its real-time status.Aiming at the problem that the integrated K-nearest neighbor(KNN)algorithm was not sensitive to random sampling,an improved integrated KNN model based on regular sampling was proposed.Firstly,the distance correlation coefficient was used to select variables.Then the variables were sorted based on normalized mutual information and used for regular sampling to construct sub training sets.Finally,the threshold was set to analyze the real-time residual based on the statistical process control method,and the health of gearbox was monitored according to the health rate curve.This method was valified by the actual data of a wind turbine.Results show that the proposed method significantly improves the estimation accuracy of the model,which is better than the conventional ensemble KNN model,and can warn the early failure of the gearbox.