This study aims to maximize the efficiency of photovoltaic (PV) power systems by quantitatively analyzing how variations in structural design parameters—an area relatively underexplored in previous research—affect overall system performance. Focus...
This study aims to maximize the efficiency of photovoltaic (PV) power systems by quantitatively analyzing how variations in structural design parameters—an area relatively underexplored in previous research—affect overall system performance. Focusing on ground-mounted fixed-tilt PV structures, three key design variables were defined: tilt angle, inter-row spacing, and module height. A large-scale dataset consisting of over 10,000 design combinations was generated using PVsyst simulations. A Deep Neural Network (DNN) surrogate model was then developed to predict annual energy yield and initial installation cost with high accuracy
For the optimization stage, a multi-objective evolutionary algorithm, NSGA-II, was applied to derive a Pareto-optimal set of solutions that simultaneously maximize energy yield and minimize initial cost. Results show that shading loss increases nonlinearly below an inter-row spacing threshold while increasing structural height enhances rear-side ventilation and reduces module operating temperature, thus improving overall efficiency. The proposed balanced design (Case C) achieved a +2.7% increase in annual yield and a +1.8% cost reduction compared to the base case, with statistically significant improvements.
This study provides a new scientific framework for PV structural design by integrating AI-based surrogate modeling with multi-objective optimization. The proposed methodology offers practical guidelines for improving the economic performance and standardization of future photovoltaic power plants.