Prediction of thermal performance of evacuated tube solar water heater based on deep belief network

As the largest country of solar water heater producer and installation, China has leading position in solar heat utilization in the world. The thermal performance of solar water heater is obtained by heat collection tests and heat loss tests. Because of low solar energy density, random meteorological conditions, and intermittent season nature, the strict test conditions is required in performance assessment. Limited by these conditions, the solar water heaters inspection efficiency is low. To improve the efficiency, a prediction model based on deep belief neural network (DBN) is presented. Based on the evacuated tube heat transfer theory, the key parameters were determined. According to different daily tests, include the local solar irradiance(G), ambient air temperature(ta) and cold water temperature(tb), the measured data(ΔT) obtained under standard or non-standard test conditions. Two-thirds of the data were put in the trained neural network. The rest were put in the tested neural network. The predicted result is close to measured data. It concluded that the DBN can accurately predict the thermal performance of solar water heater by hourly measured data which obtained under standard test conditions or fall short of the standard conditions. Compared to the genetic algorithm and back propagation (GA-BP) model, DBN model is superior in performance predication.