The building sector is a significant contributor to greenhouse gas emissions, primarily due to its high electricity consumption, which is sourced mainly from fossil fuels. As sustainability becomes increasingly essential, accurate forecasting of building electricity demand is critical for efficient energy management and demand response strategies. This study employs machine learning techniques, Random Forest and CatBoost, to forecast the hourly electrical load of a commercial user in Bucciano, Benevento, Italy. Historical electricity consumption and meteorological data were used to train the models. Model performance was assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The Random Forest model, with an R² value of 0.8670, outperformed CatBoost, with an R² value of 0.8593, demonstrating a superior ability to model complex energy consumption patterns. The forecasting results were further utilized to optimize load scheduling and enable proactive energy management decisions. These findings underscore the potential of data-driven forecasting to improve operational efficiency and facilitate the transition toward low-carbon, sustainable buildings.
AI-Driven Load Forecasting for Energy Optimization in Commercial Buildings
Akram, Muhammad;Martone, Chiara;Sasso, Maurizio;
2026-01-01
Abstract
The building sector is a significant contributor to greenhouse gas emissions, primarily due to its high electricity consumption, which is sourced mainly from fossil fuels. As sustainability becomes increasingly essential, accurate forecasting of building electricity demand is critical for efficient energy management and demand response strategies. This study employs machine learning techniques, Random Forest and CatBoost, to forecast the hourly electrical load of a commercial user in Bucciano, Benevento, Italy. Historical electricity consumption and meteorological data were used to train the models. Model performance was assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The Random Forest model, with an R² value of 0.8670, outperformed CatBoost, with an R² value of 0.8593, demonstrating a superior ability to model complex energy consumption patterns. The forecasting results were further utilized to optimize load scheduling and enable proactive energy management decisions. These findings underscore the potential of data-driven forecasting to improve operational efficiency and facilitate the transition toward low-carbon, sustainable buildings.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
