Efficient Estimation of the Convective Cooling Rate of Photovoltaic Arrays with Various Geometric Configurations: a Physics-Informed Machine Learning Approach

Convective heat transfer is crucial for photovoltaic (PV) systems, as the power generation of PV is sensitive to temperature. The configuration of PV arrays have a significant impact on convective heat transfer by influencing turbulent characteristics. Conventional methods of quantifying the configuration effects are either through Computational Fluid Dynamics (CFD) simulations or empirical methods, which face the challenge of either high computational demand or low accuracy, especially when complex array configurations are considered. This work introduces a novel methodology to quantify the impact of geometric configurations of PV arrays on their convective heat transfer rate in wind field. The methodology combines Physics Informed Machine Learning (PIML) and Deep Convolution Neural Network (DCNN) to construct a robust PIML-DCNN model to predict convective heat transfer rates. In addition, an innovative loss function, termed Pocket Loss is proposed to enhance the interpretability of the PIML-DCNN model. The model exhibits promising performance, with a relative error of 1.9\% and overall $R^2$ of 0.99 over all CFD cases in estimating the coefficient of convective heat transfer, when compared with full CFD simulations. Therefore, the proposed model has the potential to efficiently guide the configuration design of PV arrays for power generation enhancement in real-world operations.