Numerical analysis and deep learning algorithm for photovoltaic-thermal systems using various nanofluids and volume fractions at Riyadh, Saudi Arabia

Abstract

Conventional photovoltaic (PV) systems have elevated temperatures in the hot climate of Riyadh, resulting in reduced electrical power generation. Therefore, nanofluids are employed in photovoltaic-thermal (PV-T) systems to absorb the self-generated heat that limits efficient operation. In addition, developing deep learning models would help improve the optimization and control of the proposed system. This study's primary goal is to numerically examine a PV-T system under the influence of using various nanofluids with varying volume fractions in the hot climate of Riyadh and to develop time-series deep learning algorithms based on the findings in this examination to predict the PV-T system's potential. A mathematical model to investigate the PV-T panel performance under various volume concentrations and nanofluids is proposed. The generated data from the different nanofluids is deployed to train a deep learning model, which is convolutional neural networks integrated with two layers of long short-term memory (CNN-LSTM), in order to predict the PV-T panel temperature. According to the investigation's findings, the best PV-T coolant is CuO nanofluid at 4 % volume concentration. Utilizing the aforementioned nanofluid can deliver an enhancement in the average daytime PV-T panel temperature, electrical, thermal, and total exergy efficiency by 34.5 °C, 16.7 %, 79.2 %, and 18.07 %, respectively. The developed deep learning algorithms were evaluated using the mean absolute error (MAE) and coefficient of determination (R2) and scored 0.18–0.35 and 97.5–98.75 %, respectively.