基于协整分析的风力机多工况监测与故障诊断

为降低装备故障率、减小停机损失，提出一种基于协整分析的装备多工况监测与故障诊断方法。基于监控与数据采集系统采集的数据，利用随机森林特征选择算法提取与装备故障相关的关键特征变量；通过对关键特征序列的协整分析，计算协整系数，建立协整残差模型，获得反映装备状态变化的最优残差序列；采用概率图分析最优残差序列，完成了多工况状态的区间划分，确定了每种工况对应的残差预警阈值，实现了状态监测与故障预警。某型号直驱式风力机的研究结果表明：所提出的方法能有效分析非平稳时间序列，利用残差阈值可以监测电机故障、识别风力机工况，提高故障诊断的准确性。

Multi-condition Monitoring and Fault Diagnosis of Wind Turbines Based on Cointegration Analysis

In order to reduce equipment failure rates and downtime loss， a method for equipments multi-condition monitoring and fault diagnosis was proposed herein based on cointegration analysis. Based on the data collected by the supervisory control and data acquisition system， the random forest feature selection algorithm was used to extract the key feature variables related to equipments failure. By cointegration analysis of the key feature series， the cointegration coefficient was calculated and the cointegration model was established to obtain the optimal residual series， which might reflect the changes of equipments status. Probability plot was applied to analyze the optimal residual series and obtain the interval division of multiple operating situations. The residual warning thresholds corresponding to each operating conditions were determined so as to achieve state monitoring and fault warning. The research results of a certain type of direct-drive wind turbine show that the proposed method may analyze non-stationary time series effectively， monitor the faults of motors and identify the operating situations of the wind turbines by residual threshold， improve the accuracy of fault diagnosis.