Integrated ANN–NSGA-II framework for photovoltaic–green roof (PV-GR) design optimization in Dubai’s climate: energy performance, thermal comfort, and roof cooling

Abstract Given Dubai’s hot climate, rapid urbanization, and high energy demand, improving rooftop performance is essential for sustainable development. This study proposes an integrated Artificial Neural Network (ANN) and NSGA-II optimization framework to enhance photovoltaic–green roof (PV-GR) systems for the top floor of a hypothetical residential building. The objective is to reduce energy use while improving thermal comfort and rooftop thermal conditions. A total of 1,000 design scenarios were simulated using Ladybug Tools, and a PV Energy Adjustment Post-Processing (PEAP) method was applied to account for the potential cooling influence of green roofs on PV performance. The ANN surrogate models, developed using 5-fold cross-validation, achieved high predictive accuracy (R² up to 0.99), enabling rapid multi-objective evaluation. Compared with the base case, the Pareto-optimal solutions improved annual thermal comfort by 15.07–15.78%, while reducing annual load intensity by 1.08–1.24% and rooftop outside-face temperature by 3.69–4.61%. The estimated adjusted PV energy generation reached a maximum of approximately 6,594 kWh, while the overall results indicate the potential of optimized PV-GR systems to improve building performance under hot-arid climatic conditions.