Dynamic modeling of photovoltaic–thermal systems using polynomial regression

This paper presents an autoregressive moving-average dynamic model approach for photovoltaic–thermal (PVT) systems using a machine-learning (ML) technique based on polynomial regression. The main contribution is the development and application of a global dynamic modeling method designed for accurate time-series prediction of the electrical and thermal power outputs of a PVT system. The performance of the proposed method is compared with traditional analytical and another ML-based method known as long short-term memory (LSTM) models to demonstrate its effectiveness and simplicity. The research includes two main parts: implementing the proposed polynomial regression-based modeling method in MATLAB and applying it to a real-world PVT system in Granges, Switzerland. A thorough frequency-domain analysis is performed to validate the model’s accuracy and reliability. The results show that the polynomial regression-based dynamic model offers clear advantages for system analysis, providing a more user-friendly alternative to the complex and often cumbersome analytical methods. Compared to analytical models, ML-based methods achieve more accurate results with lower modeling complexity and greater practical applicability. As a result, the proposed method streamlines the modeling process while improving the efficiency and accuracy of PVT system performance prediction and optimization.