Parameter analysis and layout optimization of a rooftop photovoltaic power system using a data-driven approach

Rooftop photovoltaic (RPV) systems play an important role in supporting urban energy transition, yet their layout optimization remains challenging due to complex physical modeling and limited multi-objective capability. This paper proposes a data-driven surrogate-based optimization approach that integrates Grey Relational Analysis (GRA) and Response Surface Methodology (RSM) to improve the technical performance of RPV systems. Three key configuration parameters, including string size, tilt angle, and PV row spacing, are systematically evaluated through PVsyst-based simulation experiments. A 630 kW rooftop RPV installation in China is used as a case study to validate the proposed approach. The optimized configuration increases total energy generation from 786.29 MWh/year to 820.86 MWh/year and improves system efficiency from 81.45% to 84.85%. ANOVA results confirm that tilt angle and PV row spacing are the dominant contributors to performance improvement, while string size exhibits a marginal influence. Compared with conventional metaheuristic methods, the proposed approach achieves rapid performance enhancement by constructing a data-driven surrogate model from a limited number of simulation outputs, without requiring explicit formulation and coupling of intricate physical equations in the optimization process. The analysis of coupling effects between the dominant parameters offers practical guidance for RPV layout design, facilitating real-world decision-making for rooftop PV deployment.