Deep-Learning Forecasting Method for Electric Power Load via Attention-Based Encoder-Decoder with Bayesian Optimization

• DOI: 10.3390/en14061596
• Publication Date: 2021-03-15T02:13:10Z
• Authors: Xue-Bo Jin; Wei-Zhen Zheng; Jian-Lei Kong; Xiao-Yi Wang; Yu-Ting Bai; Ting-Li Su; Seng Lin
• Keywords: Robustness; Electrical load; Hyperparameter

Short-term electrical load forecasting plays an important role in the safety, stability, and sustainability of power production scheduling process. An accurate prediction can provide a reliable decision for system management. To solve limitation existing methods dealing with time-series data, causing poor stability non-ideal accuracy, this paper proposed attention-based encoder-decoder network Bayesian optimization to do short-term forecasting. Proposed model is based on architecture gated recurrent units (GRU) neural high robustness data modeling. The temporal attention layer focuses key features input that play vital promoting accuracy Finally, method used confirm model’s hyperparameters achieve optimal predictions. verification experiments 24 h real from American Electric Power (AEP) show outperforms other models terms algorithm providing effective approach migrating time-serial by deep-learning technology.