Power Curve Anomaly Identification Method Based on Improved Dbscan Algorithm

The power curve is one of the most important indicators to measure the operation of wind turbines. In order to evaluate the operating condition and power generation efficiency of wind turbines quickly and efficiently, it is of great significance to develop a method for identifying power curve anomalies with versatility and accuracy. In this paper, based on the SCADA historical operation data of wind turbines in a wind farm in North China, a power curve anomaly identification method is investigated, which improves DBSCAN algorithm by setting specific power range threshold values for different wind speed segments of the power curve. The anomalies detected by DBSCAN algorithm and exceeding the power range thresholds are defined as final identified anomalies. It has been verified that this method performs well and improves precision rate by 6 % and reduces the false alarm rate by 90 % compared to the original DBASCAN method. In this case, this method can provide a strong technical support for the analysis of wind turbine power generation efficiency and identification of power curve degradation.