AI- aided surrogate model for prediction of HVAC optimization strategies in future conditions in the face of climate change

Climate change significantly affects buildings' energy consumption, which accounts for a large portion of global energy use. Mitigating these impacts can result in substantial energy savings and CO 2 emissions reduction. This study investigates the use of machine learning (ML) algorithms to predict the performance of energy savings and thermal comfort improvements resulting from optimized HVAC systems in office buildings across the United States in the future in the face of climate change. This environmental issue has a significant impact on buildings' energy consumption, accounting for a large portion of global energy use. Future weather files for 16 cities were generated using a statistical downscaling method based on the HadCM3 climate model and A2 GHG scenario. Baseline data and these future weather files were used to simulate energy performance for a medium office building prototype developed by the Department of Energy. The model was modified by adjusting HVAC setbacks, and the resulting energy savings and thermal comfort improvements were analyzed. The accuracy of the predictions was evaluated based on the popular ML model assessment metrics and the difference between predicted values and the simulated values. The results show that the surrogate ML model (Light Gradient boost Model Regressor LGBMR) can predict energy savings and thermal comfort improvements resulting from optimized HVAC systems, with a high degree of accuracy 88 %. These findings highlight the potential of ML algorithms to predict the effectiveness of HVAC setback temperature strategies for climate change mitigation in buildings, which can be used as an easy way to assess different HVAC optimization strategies for future conditions.