A hybrid SVR–NSGA-II approach for design-oriented multi-objective optimization of photovoltaic retrofits considering roof geometry and aesthetics

With the growing demand for renewable energy, photovoltaic (PV) panels are increasingly applied in both new constructions and building refurbishment projects. However, how to select and design a photovoltaic system while comprehensively considering the goals of cost, power generation, aesthetics, and environmental sustainability, as well as the constraints of building roof geometry, remains an unresolved issue. This study investigates Building-Integrated Photovoltaics (BIPV) and Building-Applied Photovoltaics (BAPV), as well as end-of-life PV panels, by analyzing the impact of multi-dimensional parameters on system performance, including panel type, tilt angle, orientation, color, and service life. A multi-objective optimization is performed by integrating the Support Vector Regression (SVR) and Non-dominated Sorting Genetic Algorithm II (NSGA-II). The non-dominated solutions are then ranked using the hypervolume indicator and Euclidean distance, followed by applying these solutions to building roofs to obtain a more matched and robust solution set. The framework is subsequently validated through application to a building project. The results indicate that the proposed approach enhances both solution accuracy and computational efficiency, thereby supporting the design and sustainable deployment of photovoltaic systems in buildings.