A study on efficiency estimation of water based PV/T systems with machine learning methods

With the inevitable increase in energy consumption, there has been a growing interest in renewable energy sources. Therefore, efforts to increase the efficiency of solar panels have gained significant importance. In experimental studies conducted on Photovoltaic/thermal (PV/T) systems in the literature, the obtained data have been used in theoretical and mathematical calculations to derive efficiency expressions. However, deviations in the calculated efficiency values occur when environmental conditions and input parameters change. Therefore, in this study, efficiency prediction has been performed using machine learning (ML) methods to eliminate deviations and uncertainties. The input parameters, including mass flow rate, ambient temperature, inlet fluid temperature, cell absorptance, PV/T surface area, and solar radiation, along with a dataset collected from the literature, were used with ML methods such as Support vector regression (SVR), Linear regression (LR), Gaussian process regression, Stepwise LR, and artificial neural network (ANN) algorithms. To prepare the data for the algorithms, a preprocessing stage was applied, during which various data preprocessing methods, including min-max normalization, z-score normalization, and logarithmic transform + standardization, were implemented. According to the results, the highest prediction performance was achieved when min-max normalization was applied during the preprocessing stage, with the support vector regression algorithm yielding an root mean square error (RMSE) value of 0.225 and an R 2 value of 0.98018. Furthermore, according to the results of the F-test, it was concluded that the most influential parameter on efficiency is the PV/T surface area.