A Combined Algorithm for Cleaning Abnormal Data of Wind Turbine Power Curve Based on Change Point Grouping Algorithm and Quartile Algorithm

There exist plenty of outliers in power curve of wind turbines, which is not conducive to the follow-up information mining. Improving the wind power curve data quality of wind turbines has great engineering value. According to the spatial distribution characteristics and shapes, the outliers of wind power curve are divided into four types: the bottom curve, the mid curve, and the top curve stacked outliers, as well as scattered outliers around the curve. Based on the outlier distribution characteristics, this paper proposes a method and process of cleaning the outliers based on the change point grouping algorithm and the quartile algorithm, followed by a theoretical analysis of its feasibility. Case studies and its comparisons with the quartile-change point grouping algorithm and the local outlier factor algorithm show that the proposed change point grouping-quartile algorithm can effectively identify the four types of outliers, with good cleaning effect, high efficiency, and strong versatility.