Data-Driven Modeling of Photovoltaic Panels Using Neural Networks

Photovoltaic (PV) systems are essential for the shift towards sustainable energy. Accurate performance modeling of these systems is vital for optimizing their efficiency and adapting to changing environmental conditions. Traditional analytical models often struggle with real-world complexities. This paper uses a neural networks (NNs) method to model PV system performance, offering superior accuracy and adaptability. By leveraging NNs, the model adopts a black-box methodology, bypassing the need for detailed knowledge of PV system complexities. The study utilizes historical performance data from two panels located in Sion, Switzerland, and Eugene, Oregon, in the United States. This approach captures the interdependencies between input parameters such as solar irradiance, temperature, and applied load and output variables such as voltage and current. Additionally, it accounts for key PV metrics including open circuit voltage, short circuit current, maximum power point current, and maximum power point voltage. The NN model is trained with a comprehensive dataset featuring high persistent excitation along with real measurements. Performance metrics such as root mean square error demonstrate that the data-driven NN model surpasses traditional methods, offering a more practical solution for PV system performance modeling.