This study introduces a novel architecture for electricity price forecasting, comprising four modules designed for prediction and information fusion. Three modules are dedicated to preliminary price prediction, while the fourth integrates information from prior predictions to generate final forecasts. Experimental evaluations demonstrate the effectiveness of the proposed model, showcasing superior performance compared to models from classical ones to cutting-edge ones in time-series modeling. Specifically, results show improvements of 3.51%–53.09% on RMSE, and 4.77%–59.19% on MAE. Additionally, we conduct an ablation study to analyze the robustness of the proposed model and the distinct contributions of its modules. The findings highlight the different roles of each component and provide valuable insights for future research in electricity price prediction. Given the critical role of accurate electricity price prediction in promoting the efficiency of electricity trading and market health, this method offers a promising avenue for advancing prediction techniques in this domain.