A COMPARATIVE PERFORMANCE ANALYSIS OF DIFFERENT MACHINE LEARNING ALGORITHMS IN EFFICIENCY OPTIMIZATION OF A NANOFLUID-COOLED PHOTOVOLTAIC THERMAL SYSTEM

Efficiency optimization of nanofluid-cooled photovoltaic thermal systems is of great importance in the advancement of renewable energy technologies. This study conducts a comparative analysis of various machine learning algorithms to determine their effectiveness in predicting the efficiency of nanofluid-cooled photovoltaic thermal (PVT) systems. The performance of artificial neural network models trained with Levenberg-Marquardt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SCG) algorithms in predicting cell efficiency was extensively studied. In network models with multilayer architecture, four input parameters were defined: volume fractions of TiO<sub>2</sub> and Ag nanoparticles, velocity of cylinder and Reynolds number, and cell efficiency was estimated. Analysis of performance parameters and comparative results show that machine learning algorithms achieve high success in optimizing the efficiency of nanofluid-cooled photovoltaic thermal systems. Every network model successfully predicted the cell efficiency value with average deviation rates below 0.01%. The models had mean square error values below 4.01E-07 and coefficient of determination values exceeding 0.95. In addition, it stands out among the research findings that the model with the Levenberg-Marquardt algorithm has the highest prediction accuracy.