Scholarly Article
ASFO-SVM-Based Intelligent Load Prediction for Microgrid Energy Optimization in Renewable Energy Systems
Sundaramurthy, Arivoli, Jagadeesan, Ganesh Moorthy, Ramasamy, Karthikeyan, Vaithiyalingam, Chitra
2025-10-10 · Journal of Computational Systems and Applications · Cultech Publishing Sdn. Bhd.
Abstract
Accurate load forecasting is essential for efficient microgrid operation, stability, and planning. Traditional forecasting methods often suffer from limitations in handling nonlinear, complex, and dynamic load patterns, which can reduce accuracy and reliability. To overcome these challenges, this study proposes a hybrid framework that integrates the adaptive sunflower optimization (ASFO) algorithm with support vector machines (SVM) for short-term microgrid load forecasting. The ASFO algorithm is employed to optimize SVM hyperparameters, enhancing generalization capability and improving prediction accuracy. The baseline forecasting model recorded a mean absolute percentage error (MAPE) of 34.59%, indicating the need for optimization. After applying the proposed ASFO-SVM framework, forecasting errors were significantly reduced, achieving a minimum MAPE of 0.44% for summer weekdays and consistently remaining below 2% across all seasonal and daily horizons. Comparative analysis with existing models__including PSO-SVM, Random Forest, and standalone SVM__demonstrates that ASFO-SVM achieves up to a 62.30% improvement in accuracy while maintaining computational efficiency. The findings confirm that the ASFO-SVM approach provides a robust, accurate, and adaptable solution for microgrid load forecasting. This makes it a promising candidate for practical deployment in intelligent energy management systems, especially in scenarios requiring reliable decision-making under uncertainty.
Keywords
Adaptive sunflower optimization, Support vector machine, Electric load forecasting, Micro-grid, Intelligent forecasting system
Citation Details
Journal of Computational Systems and Applications, pp. 29-47