Short-Term Wind Power Prediction Method Considering Wind Turbine Operation Status Clustering Under Low-Temperature Conditions

ObjectivesLow-temperature weather poses challenges to the operation of power systems with a high proportion of new energy, such as wind power. Improving the accuracy of short-term wind power prediction under low-temperature conditions will provide effective decision-making information for power system scheduling and operation. To address this, a wind power prediction method considering the clustering of unit operation status under low-temperature conditions is proposed.MethodsThe fuzzy C-means (FCM) clustering algorithm is used to cluster wind turbines based on their operation status and protection control information. Then, a prediction method based on support vector machine is proposed to predict whether the wind turbines are in normal operation status. The LightGBM algorithm in ensemble learning is employed to predict the power output of wind turbines under normal operation. Based on the prediction results of both operation status and power values, the overall wind power output of the wind farm is determined. Finally, a case study of a wind farm in northern Hebei is conducted to validate the effectiveness of the proposed method.ResultsBy fully utilizing the characteristics of wind turbine protection control behaviors under low temperatures, the proposed method accurately predicts the critical shutdown time of wind turbines and provides the shutdown capacity. It effectively fits the variation patterns of wind power curves,which improves the prediction accuracy of the wind power to more than 90%.ConclusionThe proposed method can provide reliable prediction information for power scheduling and control. Additionally, it can provide a reference for short-term wind power prediction under other extreme weather conditions, such as strong winds.