predicting electrical load demand using bagging ensemble of multi-layer perceptron and adjusted long short-term memory with metaheuristic methods

Effective prediction of electric power demand is critical for maintaining the stability and reliability of the energy supply in both residential and industrial sectors. accurate energy demand forecasting is essential for balancing consumption needs with grid stability. however, the complexity of energy consumption data, influenced by a variety of factors, makes this forecasting challenging. traditional methods often struggle to capture the intricacies of such complex data, highlighting the need for more advanced and adaptable approaches. in this research, we propose a novel solution based on a bagging ensemble of multi-layer perceptron (mlp) and long short-term memory (lstm) networks, combined through a voting mechanism to improve the accuracy and generalization ability of the model. metaheuristic methods, including particle swarm optimization (pso) and the genetic algorithm (ga), are employed for optimal hyperparameter tuning of the lstm. unlike many existing studies that rely on proprietary or limited datasets, this approach uses publicly available data from the electric power consumption dataset of tetouan city (01-01-2017 to 12-31-2017), making it more accessible and applicable to broader contexts. it also enhances prediction performance by combining the results of multiple models, allowing for a more robust and accurate prediction of energy consumption. experimental results demonstrate that the proposed approach significantly outperforms existing machine learning and deep learning methods.