Short-term Wind Power Prediction Based on Error Estimation

Abstract

Highly reliable wind power prediction is a crucial assurance for the large-scale integration of renewable energy into the power system. In order to enhance the accuracy of wind power prediction, this paper proposes a short-term wind power prediction method based on error estimation. Firstly, using historical data from numerical weather prediction (NWP) as input, a bi-directional long short-term memory (BiLSTM) network model is employed for preliminary wind power prediction. Preliminary prediction errors are calculated based on the prediction results. Subsequently, an Extreme Gradient Boosting (XGBoost) algorithm is utilized to build a learning model capable of efficiently estimating BiLSTM's prediction errors. This facilitates rapid estimation of BiLSTM's preliminary prediction errors when provided with future NWP data for a certain period. Furthermore, the error estimation is incorporated with the preliminary prediction results to yield the improved wind power predication. Finally, the proposed method is tested with real data collected from a wind farm, and comparative results with various mainstream wind power prediction methods demonstrate the effectiveness of the proposed approach in enhancing the performance of short-term wind power predictions.