Enhanced Probabilistic Spatiotemporal Wind Speed Forecasting Based on Deep Learning, Quantile Regression, and Error Correction
Shuang Zhu, Xudong Chen, Xiangang Luo, Kai Luo, Jianan Wei, Li Jiang, Yanping Xiong
Abstract
Developing wind speed forecasting is a prerequisite for the safe and effective utilization of wind power. In this study, an enhanced spatiotemporal wind speed forecasting model is proposed for short-term wind speed prediction, which consists of convolutional long short-term memory network, quantile regression, and error correction modules. The model makes use of the powerful time-series mining ability of long short-term memory (LSTM) and the measurement of variable uncertainty by quantile regression so that the model has the advantages of advanced certainty and uncertainty prediction at the same time. In addition, the error correction module is added to further improve the forecast accuracy. The proposed model has been validated in three large-scale regions in the United States and compared with three other state-of-the-art models. In the deterministic prediction, compared with the best-performing LSTM among the baseline models, the mean absolute error and root mean square error are reduced by 30.71% and 26.99%, respectively. In probabilistic prediction, the proposed model performs better than Gaussian process regression with higher reliability. The results of statistical testing demonstrate that the proposed model can obtain both accurate deterministic prediction and reliable probabilistic prediction. This indicates that the model has advantages in the spatiotemporal prediction of large-scale regions.