Prediction of Electrical Power of Ag/Water-Based PVT System Using K-NN Machine Learning Technique

In this study, the effectiveness of the machine learning model for predicting the electrical power output of PVT system is evaluated. Specifically, a K-Nearest Neighbor (K-NN) method is explored, using various hyper-parameters and characteristics. In this study, machine learning techniques are used to simulate the electrical performance of PVT systems that are cooled by water-based nanofluids. In the proposed model, the mass flow rate, volume fraction and solar radiation have been considered as the input variables in order to predict the electrical powers versus time. Datasets has been extracted from previous experimental research for an Ag/water-based PVT system (laminar flow rate and turbulent with 2% and 4% of volume fractions). Results demonstrated the capability of the method developed to predict a new output database containing 116 values of the electrical power versus time with the absolute average relative deviation AARD = 1.022%, and root mean squared error (MSE) of 4.0577%.