Deep Learning Approach Based on Feature Extraction for Short-Term Electricity Price Prediction

With the dynamic evolution of energy supply and demand structures and the influence of multi-source perturbing environmental factors, accurate day-ahead electricity price (EP) prediction has become a critical issue in the multi-agent decision-making process within the electricity market. In this paper, a deep learning method based on feature extraction is proposed for short-term EP prediction. First, external feature variables related to EP are introduced, and a feature extraction method based on the attention mechanism is applied to process these variables. Variational Mode Decomposition (VMD) is then employed to extract EP features from the EP sequence. Gated Recurrent Unit (GRU) is utilized to train the predicted EP, while the Snow Ablation Optimizer (SAO) algorithm is used to optimize the hyperparameters of both GRU and VMD. Finally, an empirical case study is conducted using data from the Spanish electricity market to validate the effectiveness of the proposed model. The results demonstrate that the proposed model achieves higher prediction accuracy compared to other models.