A multi-stage predicting methodology based on data decomposition and error correction for ultra-short-term wind energy prediction

Abstract

As an emerging clean energy, wind energy has become an important part of energy development all over the world. One of the major ways to use wind energy is wind power. Accurate wind power forecasting is significant to the wind energy development and utilization, and the power systems safe and stable operation. Due to the fluctuation and randomness of wind energy, improving the accuracy of ultra-short-term wind energy prediction has become the key to wind energy development and utilization, and it is also the focus of wind energy development research in various countries. Therefore, this paper proposes a new combination model based on complementary empirical mode decomposition (CEEMD), T-S fuzzy neural network (FNN) optimized by improved genetic algorithm (IGA) and Markov error correction to improve the accuracy of ultra-short-term wind power prediction. First, the CEEMD is used to decompose the wind data into several components; then, the trained IGA-FNN model is used to individually predict each modal component to improve accuracy and stability; finally, the prediction results of all modal components are superimposed and the Markov process is used for error correction to obtain the final prediction result. The empirical results show that the mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) of the proposed model is 15.59%, 17.95% and 6.94%, respectively. The empirical result proves that compared with the BPNN, Elman NN, and FNN, the prediction results MAE of the proposed method is reduced by 68 0.6%, 61.7%, 59.2%, the RMSE is reduced by 70.7%, 65.0%, 63.9%, the MAPE is reduced by 75.5%, 67.6%, 60.4%. The prediction accuracy of the proposed method is significantly higher, and it is available for wind power development and utilization.