Machine-learning multi-objective optimisation criteria for identifying synergies in comfort-carbon-energy towards integrated climate-resilient building designs in a typical cooling-dominated region (Ghana)

Abstract

Carbon emissions reduction, comfort of occupants and energy use (CCE) have become priority targets in a continually changing climate. However, a classic trade-off emerges where it becomes extremely challenging to achieve synergy due to trade-offs. Existing CCE optimisations suffer from conventionally strict preliminary boundary design conditions that over-constrain the optimisation space. The current study proposes a relatively flexible optimisation criteria for integrated climate-resilient building designs in a typical cooling-dominated region (Ghana). Seven surrogate models were trained using data from EnergyPlus via the Latin Hypercube Sampling method. Gradient boosting demonstrated superior performance, with accuracy exceeding 99%. NSGA-III revealed 91 pareto fronts, while TOPSIS analysis was utilised to identify an ideal solution. A keen examination of a more feasible solution from the 91 pareto fronts demonstrated a 31% reduction in operational carbon emission, 135% reduction in cooling loads, and a substantial reduction in discomfort hours to negligible values (10 hr). However, this feasibility is highly dependent on cooling setpoint of 29 °C, efficient external wall insulation (U value = 0.347 W/m2-K), moderate infiltration (0.849817 ac/hr), optimal cooling efficiency (4.4), local shading (2 m) and natural ventilation rate of 2.5 ac/hr (concurrent mixed-mode ventilation). As a novel contribution, while the ideal solution appears diagnostic and overly ambitious, it challenges conventional design boundaries and signals the possibility of achieving synergy in climate-resilient building designs beyond existing preliminary boundary conditions in a changing climate. The surrogate models presented in the current study constitute the very first for optimising the CCE dynamic in this climatic region (Ghana).