Research on Load Forecasting based on Optimized Support Vector Machine Model

Abstract

Accurate prediction of building energy consumption plays a vital role in energy scheduling, conservation, and emission reduction. However, conventional forecasting methods often struggle to capture nonlinear patterns effectively, leading to limited prediction accuracy. Focusing on rural residential buildings in Beijing, this study develops a Support Vector Machine (SVM) model based on measured data, and introduces Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) to optimize the model's hyperparameters. The results demonstrate that the traditional SVM model achieves a coefficient of determination (R²) of only 72.3%, with a mean absolute percentage error (MAPE) of 9.27% and a root mean square error (RMSE) of 0.92 kW. After optimization, the GA-SVM and PSO-SVM models significantly outperform the baseline: their R² values increase to 85.6% and 92.7%, respectively, while MAPE is reduced by 3.85 and 6.02 percentage points, and RMSE decreases by 0.39 kW and 0.53 kW, respectively. The findings underscore the effectiveness of optimization algorithms in enhancing SVM-based load forecasting and provide a reliable reference for energy management in similar building contexts.