Experimental and machine learning evaluation of a solar PVT system with water and Al2O3 nanofluids for improved electrical and thermal efficiency

This study addresses the need for efficient and sustainable energy systems by investigating a solar photovoltaic-thermal (PVT) hybrid system for simultaneous electricity generation and hot water production, an approach crucial for residential applications in solar-rich regions. The novelty lies in combining experimental analysis with machine learning (ML) techniques to enhance and predict the system s performance, particularly using an Al2O3 nanofluid coolant, which has been rarely explored in prior literature. Experiments were conducted in Chennai, India, across winter months (December February) under varying flow rates (0.5 2.0 LPM) and two coolant types: water and a 0.1% vol. Al2O3 nanofluid. Key performance indicators, such as electrical efficiency, surface and tedlar temperatures, thermal output, and environmental conditions, were measured. To model and predict system behavior, four ML algorithms (Linear Regression, Random Forest, XGBoost, AdaBoost) were trained and evaluated using R2, mean absolute error (MAE), and root mean square error (RMSE). The XGBoost model outperformed others, achieving an R2 of 0.9726, MAE of 0.2411, and RMSE of 0.5437. Experimentally, the use of Al2O3 nanofluid improved electrical efficiency by up to 12.4% and thermal output by 18.7% compared to water. These findings demonstrate that integrating nanofluids and ML-based predictive analytics can significantly boost the efficiency of PVT systems, making them a viable option for sustainable domestic energy solutions in climates similar to that of Chennai.