Comprehensive short-term analysis of building energy consumption patterns based on a robust and data-driven computational framework

This study introduces a hybrid optimization approach for short-term load forecasting utilizing an artificial neural network that integrates particle swarm optimization and grey wolf optimizer approaches within the radial basis function neural network. Conventional optimization techniques, such as particle swarm optimization and grey wolf optimizer, often encounter challenges including local optima, slow convergence, and ineffective exploration, which limit their effectiveness in optimizing artificial neural network parameters. The suggested hybrid method addresses these constraints by leveraging the advantages of both techniques, aiming to enhance convergence speed and prevent premature convergence. The proposed methodology is assessed via a series of experiments, comparing various optimization algorithms, including particle swarm optimization, grey wolf optimizer, accelerated grey wolf optimization, multi-verse optimizer, and moth-flame optimization, based on diverse performance metrics. The experiments adhere to a systematic procedure encompassing data preprocessing, optimization, training, and performance evaluation. The optimal parameters determined in the studies, comprising the most effective combination of particle swarm optimization and grey wolf optimizer values, produce a coefficient of determination of 0.9668 for cooling loads and 0.9411 for heating loads, surpassing all alternative methods. This study illustrates that hybrid optimization methodologies can markedly improve the efficacy of artificial neural networks, offering a more effective solution for short-term load forecasting.