Performance investigation and machine learning modeling of PV panels equipped with PCM based passive cooling systems

Abstract

In this work performance of two PCM based passive cooling systems for PV panels was investigated. The cooling systems were in the form of rectangular PCM containers, one equipped with fins while the other with porous medium. Outdoor experiments were conducted in winter and summer to examine both daytime cooling and nighttime heat rejection of the two containers, and to explore the performance variation across the two seasons. Moreover, machine learning models were generated to predict the PV panel's temperature and power for the considered systems. Both systems effectively cooled the PV panels in winter during the day and achieved sufficient heat rejection during the night. However, the porous medium container had superior performance, achieving around 3.40 degrees C average temperature reduction and up to 3.56 % enhancement in efficiency. In summer where PV temperatures are much higher, the porous medium container underperformed while the finned container resulted in noticeable cooling achieving 3.42 degrees C temperature reduction and up to 6.58 % increase in efficiency. On other hand, the finned container encountered insufficient heat rejection at night. Machine learning models generated using ANN illustrated acceptable prediction accuracy with coefficient of determination (R2) of 0.98 for winter and 0.99 for summer. The mean square errors (MSE) in the predicted PV's temperature and power were 1.08 degrees C2 and 0.27 W2 for winter and 1.57 degrees C2 and 0.11 W2 for summer, respectively. The findings of the study highlights the season-dependent nature of different PCM based passive cooling systems.