Wind Farm Anomaly Data Detection Algorithm Based on Error Stratification of Power Curve
Due to sensor anomalies,transmission interference,and wind power curtailment,wind farm power curve data often contains significant amounts of abnormal data,which not only interferes with wind power prediction accuracy but also impedes further application research on measured data.To address this,a method is proposed that first eliminates wind curtailment and power restriction data based on the time-varying nature of wind speed and power sequences using a volatility method.For isola-ted and outlier data,a regularized adaptive learning algorithm is applied to establish an autonomous mapping relationship be-tween wind speed and power.Subsequently,the difference between the power curve and measured data is computed to obtain the power error distribution,and the generalized error distribution method is used to fit the probability density function.By setting different confidence levels,the measured data is stratified,and abnormal data in different layers is identified and corrected using appropriate methods.Finally,a predictive algorithm is employed to evaluate the accuracy of abnormal data detection.Simulation results demonstrate that the proposed power curve error stratification algorithm offers superior performance and recognition ac-curacy compared to the traditional 3-sigma algorithm.
anomaly data detectionerror stratificationpower predictionregularized adaptive learning