No abstract

The characteristic curve of wind speed and power reflects the output state of wind turbine, and its characteristics are helpful to the accurate prediction of wind power. With the improvement of wind turbine power generation technology and other engineering applications, the data collected by SCADA system contains a large number of outliers, which makes it difficult to accurately fit the wind speed power curve. Firstly, this paper analyzes the types and causes of outliers in the actual data of wind turbines. Then, an abnormal data cleaning method based on Tukey’s method considering the operation parameters of wind turbines is proposed to clean the data of 12 wind turbines. Finally, the new data are clustered separately in the wind speed range by K-means, and the wind speed power characteristic curve is fitted. Compared with the uncleaned data, this method can significantly improve the fitting accuracy of wind speed power characteristic curve.

No abstract