Energy demand forecasting using ridgelet neural networks Boosted Beluga whale optimization

Abstract Deep learning has become increasingly popular in energy demand forecasting, due to its ability to efficiently process large amounts of data with improved accuracy. By utilizing a deep learning approach, energy demand forecasting models can capture and learn the complex patterns and relationships that exist within energy demand data. This enables more accurate predictions of future energy consumption, allowing for better decision-making and efficiency. This research introduces an advanced Ridgelet Neural Network (RNN) optimized through metaheuristic approaches as a powerful tool for predicting energy demand. A refined Beluga Whale Optimization (BBWO) algorithm is employed to enhance the RNN’s structure, resulting in more accurate predictions. When tested on a standard case study featuring short-, medium-, and long-term energy demand forecasts, the proposed RNN/BBWO model surpasses the performance of previously published methods. The MSE of RNN/BBWO method in short, medium and long term forecasting is 0.18, 0.16 and 0.14 lower than LASSO; 0.08, 0.11 and 0.1 lower than Bi-LSTM; 0.02, 0.06 and 0.05 lower than DRL. The findings highlight the effectiveness of this technique as a superior solution for forecasting energy consumption across different timeframes.