This study identifies the characteristics of a photovoltaic system with a gray-box model. This model integrates the interpretability of a white-box model with the accuracy and flexibility of black-box models, offering the advantages of knowledge-driven and data-driven artificial intelligence approaches. This integration not only boosts the overall performance and reliability of the model but also enhances its adaptability to changing environments, making it an effective tool for simulation and prediction of photovoltaic systems. To validate the method, the paper employs a publicly available photovoltaic dataset to test the gray-box model against existing models, including the CNN, LSTM, BiLSTM, and CNN-BiLSTM approaches. The results demonstrate that the gray-box model surpasses those models in terms of prediction accuracy, interpretability, robustness, and generalization capabilities, effectively improving the optimization and management of photovoltaic systems.