Study on Influence of Input Parameters by Back Propagation Neural Network on Performance Prediction of Air-type Photovoltaic Thermal System

Enough heat may be obtained by collecting the heat from the PV/T (photovoltaic thermal system), and the degradation of PV cell efficiency caused by overheating can be successfully avoided. Air cooling and liquid cooling are two common solar PV/T cooling solutions. The air-cooled PV/T system has numerous variables, and the heat transfer mathematical model is quite complicated. The BP (Back Propagation) neural network method can be used to create a simulation prediction model, which can be used to simulate and predict the thermal and electrical performance of a solar PV/T system. The simulation is based on data from a single-day experiment conducted during the heating season, and it records the BP neural network's six types of temperature difference prediction and electric power prediction under various input parameters (groups 2-group 4-group 6). The appropriate level of training has been attained. The electrical and thermal efficiency of PV/T systems can be calculated using these BP neural networks. The fitting degree of the anticipated value and the actual value of the BP neural network model fulfils the requirements when compared to the experimental data gathered over two days and the prediction results of the BP neural network model. On the two projected days, the prediction accuracy R2 of electric power value were 0.97009 and 0.95538, respectively. The temperature difference numerical fitting degree is somewhat worse than the electric power value. The R2 values were 0.90114 and 0.93547, respectively, for forecast accuracy. This research will assist in further predicting and analyzing the application benefit of PV/T systems in different regions and climate conditions， and will provide valuable information for PV/T system application in building integration.