Artificial neural-network based model to forecast the electrical and thermal efficiencies of PVT air collector systems

Abstract In the recent decade, Machine Learning techniques have been widely deployed in solar systems due their high accuracy in predicting the performances without going through the physical modelling. In this work, the Artificial Neural Network (ANN) method is adopted to forecast the electrical and thermal efficiencies of a photovoltaic/thermal (PVT) air collector system. Indeed, two accurate modelling techniques have been used to generate the output results for training and validation. Both deployed electrical and thermal models have been validated experimentally and demonstrated high accuracy. Then, real climatic samples of one year with a 10 minute step of the Jordan valley location have been adopted to generate the electrical and thermal efficiencies. These latter are used in the training and validation of the developed ANN model under various combinations of the weather variables. The solar irradiance and the module temperature are the most important variables to consider as input in a NN-based model respectively. The developed ANN model shows MAE of 0.0078% and 3.3607% in predicting the electrical and thermal efficiency respectively. The electrical efficiency can be predicted with higher accuracy than the thermal efficiency. Further, the results demonstrate that the ANN outperforms the LS-SVM in forecasting the PVT air collector performances.