风机异常及缺失数据的填补方法研究

The identification and filling of wind turbine abnormal data and missing data is of great significance for the assessment of the operating status of the wind turbine and the prediction of future wind speed. This paper considers that some wind turbines in SCADA system may have abnormal data and a large amount of missing data. Firstly, the wrong data is identified and excluded, and then, classified the missing data. In the case of missing individual discontinuities, filling of the mean of adjacent data is carried out; In the case of continuous missing and side wind turbine data reference, based on the adjacent wind turbine data in the same time period, the wind direction filling model is firstly established, the continuous and complete wind direction data is drawn, and then, SVM method is adopted to establish the wind speed filling model in each wind direction interval respectively. For the missing data without the side wind turbine reference, the NAR neural network is used for point-by-point wind speed filling. In this paper, the measured data of a certain wind field is used for data verification, and compared with other traditional neural network filling methods. The test results show that the proposed method outperforms other models.