Electrical performance assessment of a curved composite photovoltaic panel using hybrid particle swarm optimization–artificial neural network modeling and a standalone mathematical formulation

Source abstract

Abstract Curved and conformal photovoltaic (PV) panels are increasingly considered for applications where conventional flat modules cannot be easily integrated. However, panel curvature introduces geometric effects that complicate accurate electrical performance assessment and prediction. This study presents an experimental and data-driven framework for evaluating the electrical performance of a curved composite PV panel under outdoor operating conditions. A glass fiber-reinforced curved PV panel with a center curvature angle of 140° was used in the experimental investigation. A geometric model was employed to account for curvature-related irradiance effects. Electrical power output was predicted using a hybrid modeling approach combining particle swarm optimization (PSO) and a feed-forward artificial neural network (ANN) with Bayesian regularization, based on measured solar irradiance and ambient temperature data. In addition, a standalone mathematical formulation was derived to enable rapid power estimation without the need for advanced computational tools. The hybrid PSO–ANN model demonstrated strong prediction performance on the independent testing subset, achieving an R2 value of 99.2%, root mean squared error of 0.698 W, mean absolute error of 0.446 W, and mean squared error of 0.4873 W2. The results show that the proposed framework can effectively support the electrical performance assessment of curved composite PV panels for conformal PV applications.