Machine Learning-based Energy Demand Forecasting

Increasing energy demand necessitates efficient management strategies to balance the energy demand-supply grid effectively. The research focuses on integrating energy demand forecasting approaches using regression models into building management systems. The research uses a comprehensive dataset spanning 2016 to 2018, including numerous measurements and observations on power use in buildings and time-series meteorological factors. Several regression models, including the Support Vector Machine (SVM), Random Forest (RF), Ridge, Lasso, and Gradient Boosting regressors, are used and compared. Each model's effectiveness is evaluated using performance indicators such as the R-squared (R2) score and Mean Squared Error (MSE). SVM emerges as the best option, with the highest R2 score and lowest MSE. Altogether, by integrating these forecasting models, real-time operations can optimize energy consumption management and improve prediction accuracy.