Advanced forecasting of building energy loads with XGBoost and metaheuristic algorithms integration

Abstract

Accurate forecasting of cooling and heating loads in buildings is vital for effective energy management, cost efficiency, and environmental sustainability. However, traditional forecasting models often face limitations in capturing the complex and non-linear characteristics of building energy consumption patterns. To address this challenge, this study proposes a hybrid predictive approach by integrating eXtreme gradient boosting (XGBoost) with eight metaheuristic optimization algorithms. The selected algorithms are ant colony optimization (ACO), barnacles mating optimizer (BMO), genetic algorithm (GA), gradient-based optimizer (GBO), hippopotamus optimization (HO), Kepler optimization algorithm (KOA), particle swarm optimization (PSO), and teaching-learning-based optimization (TLBO). Each metaheuristic was used to optimize the hyperparameters of the XGBoost model, resulting in the following hybrid models: ACO-XGBoost, BMO-XGBoost, GA-XGBoost, GBO-XGBoost, HO-XGBoost, KOA-XGBoost, PSO-XGBoost, and TLBO-XGBoost. The models were evaluated based on the root mean square error (RMSE) metric to determine prediction accuracy. Among the tested combinations, the GA-XGBoost model produced the lowest RMSE for both cooling and heating load forecasting, indicating superior performance. These findings suggest that hybridizing XGBoost with metaheuristic algorithms can substantially improve forecasting accuracy. The consistent effectiveness of GA highlights its continued relevance in solving complex optimization tasks, aligning with the no free lunch theorem which states that no single algorithm performs best across all problems.

Topics & Concepts