Optimization of energy use and thermal comfort in underserved hot-dry climate schools using a machine learning framework
Year: 2025
Type: article
Abstract: Schools constructed in underserved areas with hot-arid climates of developing countries such as Iran are suffering from harsh conditions in terms of critical energy use and cooling challenges; this is mostly due to lack of infrastructure and using convenient mechanical HVAC systems which has caused high energy use and prolonged thermal discomfort in those areas; therefore, the current study tries to optimize the small-scale buildings utilizing a Machine Learning Framework approach. Regarding to this scope, a detailed model of a two-classroom school in city of Yazd in Iran hot climate was constructed, and over three million scenarios were generated via EnergyPlus and jEPlus simulations. The studied design parameters include as orientation, insulation thickness, wall type, shading types, heating and cooling setpoints and the study objectives are to minimize cooling energy and discomfort hours. Based on the study goal, six machine learning algorithms were evaluated for building energy simulation metrics prediction; ultimately, the Gradient Boosting model was fine-tuned as the most accurate predictor, achieving an R2 of 0.9971. The statistical analyses revealed that factors such as building orientation and cooling setpoint temperature significantly influenced both the metrics. In particular, rotating the classrooms approximately 172°—reorienting them from south-facing to north-facing—improved performance for both objectives. The optimal configuration comprised external insulation of approximately 10 cm thickness, shading extensions measuring 0.83 m in depth, and a cooling setpoint of 28.3 °C, resulting in a cooling load of 1898.4 kWh and thermal discomfort of 1547.3 h.
Source: Energy Reports
Authors: Sina Allahyari, Zahra Rahmany Khorram, Masoumeh Ahmadi, Javad Ahmadi, Farshid Aram
Institutions: Tabriz Islamic Arts University, Urmia University
Language: English
Subfield: Building and Construction
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,040
OpenAlex ID: W4416584918