Machine learning based multi-objective optimisation of energy consumption, thermal comfort and CO2 concentration in energy-efficient naturally ventilated residential dwellings

The complex correlation between energy consumption, Indoor Environmental Quality (IEQ) and occupancy is significant for residential buildings but often overlooked in design and operation phases. While it is easier to set standards for energy and IEQ individually, accounting for the influence of occupants on both simultaneously presents a significant challenge. This complexity affects the accuracy of prediction models and the effectiveness of multi-objective optimisation. This research proposes a low-computational methodology based on a metamodel approach tailored for rapid prediction and optimisation of heating energy consumption (kWh), thermal discomfort (hours), and elevated CO₂ levels (hours) under the influence of occupancy. The framework evaluates occupancy's impact on the Pareto optimal front generated through metamodel-based multi-objective optimisation. The optimisation process reduced computation time by 80% compared to traditional models, with over 99% accuracy. The study highlights that variables like occupancy density, metabolic rate, and window operations significantly influence heating consumption, thermal discomfort, and CO₂ levels. Higher occupancy and metabolic rates increase internal heat gains, reducing heating demand but risking overheating without adequate ventilation. Window operations balance air quality and thermal comfort; however, prolonged ventilation may cause heat loss in colder conditions. Including occupancy-related variables ensures predicted results and optimised parameters are resilient and within WHO and CIBSE TM59 limits, while aligning heating consumption with energy-efficient standards.