Short-term wind speed interval prediction by convolutional long- and short-term memory networks based on attention mechanism

Abstract

Large-scale integration of wind energy is limited by the strong volatility and stochastic nature of wind speeds. Therefore, obtaining reliable and high quality wind speed forecasts is of great importance for the planning and application of wind energy. The objective of this study is to develop a hybrid model for short-term wind speed prediction and to quantify its uncertainty. In this study, a hybrid method, CLSTMA-GPR, which combines the Gaussian process regression (GPR) interval prediction features and the advantages of attention mechanism (AM)-based space-time feature fusion (CLSTM) point prediction, is proposed to perform interval prediction analysis of wind speed. Finally, the performance of the proposed model is verified in four aspects: point prediction accuracy, interval prediction suitability, and prediction reliability by using the actual wind speed data cases in Inner Mongolia region. The experimental results show that for the wind speed prediction problem, the model obtains higher accuracy, suitable prediction interval, and more reliable point prediction results and probability distribution than the traditional structural deep learning and Gaussian process regression models alone.