Predicting Energy Consumption in Buildings Using Various Artificial Intelligence Models

Authors

  • Yazeed Abushanab
  • Mohamed I Youssef
  • Omar Shaker
  • Mohamed Abdelaziz Sayed Youssef
  • Ryo Amano University of Wisconsin-Milwaukee

Abstract

Accurately predicting energy consumption in buildings is vital for optimizing energy efficiency, reducing costs, and supporting sustainability efforts. This study uses a dataset that spans and broadcasts hourly energy consumption for a specific building in Spain, using a dataset spanning an entire year. The dataset includes hourly energy usage in kilowatt-hours (kWh) and features representing environmental conditions, including temperature, humidity, and precipitation. alongside time-related variables, including the hour of the day, day of the week, and seasonal markers. These features provide a detailed view of how internal and external conditions influence energy usage patterns. Data preprocessing included handling missing values, feature selection, and engineering temporal variables such as Hour, Day of Year, and Is Weekend, which capture essential behavioral and operational dynamics. The building analyzed is a representative structure with typical heating, ventilation, and air conditioning (HVAC) systems. This model is well-suited for analyzing energy consumption patterns across different environmental and operational conditions. Various regression models were applied, including Linear Regression, Ridge and Lasso Regression, Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Random Forest, XGBoost, and Neural Networks. Model performance was assessed using Mean Absolute Error (MAE) and R-squared (R²) metrics. Random Forest emerged as the best-performing model, achieving an MAE of 8.33 and an R² of 0.954, highlighting its strong ability to capture the building’s energy consumption patterns. This research highlights the potential of regression models and artificial intelligence in improving energy forecasting, serving as a foundation for advancing building energy management systems.

Downloads

Published

2025-05-14