Enhancing Short-Term Load Forecasting in the Electric Power Industry: A Hybrid Approach Integrating Machine Learning and Advanced Optimization Techniques

Abstract

This work introduces a hybrid methodology that combines machine learning with sophisticated optimization approaches to improve short-term load forecasting in the electric power sector. The model uses Support Vector Regression and Radial Basis Function together and then improves it with Aquila Optimizer, Particle Swarm Optimization, and Harris Hawks Optimization. The method was evaluated on a dataset that includes daily electricity usage data and important weather variables like temperature and humidity. We used important measures, including the Root Mean Squared Error, Mean Absolute Error, and the Coefficient of Determination, to see how well the model worked. The results show that the hybrid Radial Basis Function-Harris Hawks Optimization model works better than the standalone Support Vector Regression model. It has a 20% lower Root Mean Squared Error and a 15% higher accuracy. These gains show that the hybrid model makes more accurate predictions, especially for real-time operational planning. The hybrid optimization techniques help make load forecasting models much more accurate and efficient. They could be very useful for managing energy systems. This study underscores the significance of amalgamating machine learning with optimization algorithms to augment forecasting abilities and furnish critical insights for power system operators and academics seeking to increase grid stability and planning.