Akademik Veri Yönetim Sistemi
Applied Thermal Engineering, cilt.288, 2026 (SCI-Expanded, Scopus)
The main objective of this study is to develop and compare three regression-based models, namely feed forward backpropagation neural network (FFBPNN), Gaussian process regression (GPR), and neural network regression model (NNRM), to predict the maximum thermal power, optimize thermal performance, and determine the optimal mass flow rate in Photovoltaic-thermal (PVT) systems integrated with natural zeolite plates. The analysis uses experimental data collected from two different configurations: one containing natural zeolite plates with enhanced thermal conductivity by graphite and the other without graphite. While meteorological parameters such as ambient temperature, wind speed, and solar radiation are known to significantly impact the thermal performance of PVT systems, this study addresses, for the first time, the simultaneous maximization of thermal energy output and optimization of mass flow rate, one of the most critical and controllable operating parameters. In constructing the models, ambient temperature, solar radiation, wind speed, inlet and outlet temperatures, inlet and outlet pressures, and mass flow rate were identified as the key parameters for PVT systems. The performance metrics indicate that all three regression models exhibit strong predictive capability, with the GPR model consistently achieving the highest accuracy across validation, test, and cross-validation stages. For the graphite-enhanced PVT system, GPR produced notably low RMSE and MAE values, whereas the NNRM model showed comparatively weaker performance. In contrast, in the system without graphite, all models yielded near-perfect accuracy, with R2 values exceeding 0.999 and substantially reduced error metrics. These findings demonstrate that the models—particularly GPR—provide highly reliable predictions, and system behavior without graphite is captured with exceptional precision. With the FFBPNN, GPR, and NNRM models, the optimum mass flow rates for the system with graphite were found to be 2.10, 1.93, and 2.10 kg/min, respectively, while for the system without graphite, they were found to be 3.39, 3.07, and 3.39 kg/min, respectively.