Deep learning with regularized robust long- and short-term memory network for probabilistic short-term load forecasting

Short-term electricity load forecasting plays an essential role in power system operation and energy trading. However, the strong volatility, seasonality, and uncertainly of the load series pose challenges to accurate short-term load forecasting using existing forecasting models. To tackle these challenges, in this paper, we investigate a novel deep learning-based probabilistic short-term load forecasting model to predict the hourly ahead electricity load accurately. The proposed deep learning forecasting model embeds nonconvex minimax concave penalty in the robust long- and short-term memory (LSTM) network to simplify its complex structure while extracting the important information. The model parameters in the proposed regularized and robust LSTM network are fine-tuned using whale optimization algorithm which is a meta-heuristic approach. For empirical applications, electricity load data collected from New York Independent System Operator and ISO New England are considered, and numerical results show that the proposed method outperforms other state-of-the-art competitors in terms of both point and interval forecasting accuracy.