Temperature prediction of the molten salt collector tube using BP neural network

The collector tubes in a receiver play a vital role in the solar power tower system, and directly influence the cost of the power generation. Fast forecast of the temperatures of the collector tubes from the limited number of the temperature measurement data is important. Different from the common computational fluid dynamics prediction method, in this study a back-propagation neural network method is developed to fast acquire the temperature of the receiver, such as the peak temperatures of the inner and outer surfaces, and the outlet mean temperature and the outlet highest temperature of the molten salt. The numerical simulations are implemented to validate the feasibility and effectiveness of the proposed method. Moreover, in the proposed method the temperatures of the tube wall and the molten salt can be fast forecasted without the thermal physical parameters of materials, the boundary conditions or the initial conditions, and the solution of the complicated governing equations.