Machine learning-assisted multi-objective optimization of cost and energy tradeoff in building-integrated photovoltaic envelopes

Solar energy is a clean, inexhaustible resource that plays a key role in reducing dependence on fossil fuels amid increasing climate change pressures. Building-integrated photovoltaic (BIPV) systems offer significant potential to improve building energy performance; however, existing studies lack efficient approaches for jointly optimizing energy efficiency and investment costs at the early design stage. A machine learning-assisted multi-objective optimization framework is proposed for the BIPV glass façade design. A comprehensive simulation dataset is generated using DesignBuilder, and multiple machine learning models are developed as surrogate models, with CatBoost achieving the best performance (R2 = 0.978 for net energy consumption (NetE) and 0.973 for investment cost (IC)). A multi-objective cheetah optimizer is employed to minimize both objectives NetE and IC simultaneously. The Pareto front reveals clear cost–energy trade-offs, with energy savings of up to 24.9%. The proposed framework integrates simulation, machine learning, and optimization, advancing current BIPV design methodologies by enabling efficient multi-objective decision-making. The approach provides practical design solutions and serves as an effective decision-support tool for improving both energy efficiency and economic feasibility.