Home // SIMUL 2025, The Seventeenth International Conference on Advances in System Modeling and Simulation // View article
Authors:
Navid Shirzadi
Meli Stylianou
Keywords: smart building retrofitting; energy efficiency; greenhouse gas emissions; machine learning
Abstract:
Achieving energy efficiency and reducing greenhouse gas (GHG) emissions are critical goals for building retrofitting. This study tackles challenges such as limited data and scenario generalizability by adapting the U.S. ComStock database for Canadian buildings using a Euclidean distance-based matching algorithm, achieving a 92% success rate for matches below a 2.43 threshold. Machine learning models, Random Forest (RF) and Extreme Gradient Boosting (XGBoost), were selected due to their effectiveness in handling high-dimensional, non-linear datasets and were applied to predict Energy Use Intensity (EUI) and GHG emissions. XGBoost, with optimized hyperparameters, outperformed RF, achieving R² values of 0.91 for EUI and 0.86 for GHG emissions, with lower RMSE and MAE values, showcasing its capability in handling complex, high-dimensional data. A comparative analysis highlighted significant environmental benefits of transitioning Heating, Ventilation, and Air Conditioning (HVAC) systems to cleaner fuels, such as air-source heat pumps. The proposed distribution-based method, leveraging 100 buildings across diverse climates and types, offers a robust framework for policymakers to guide energy-efficient retrofitting decisions.
Pages: 1 to 8
Copyright: Copyright (c) The Government of Canada, 2025. Used by permission to IARIA.
Publication date: September 28, 2025
Published in: conference
ISSN: 2308-4537
ISBN: 978-1-68558-300-2
Location: Lisbon, Portugal
Dates: from September 28, 2025 to October 2, 2025