Study On Optimization And Prediction Of Floor Heating System With Solar Coupled Ground Source Heat Pump

In this paper,from the perspective of building energy conservation,based on the combination of ground source heat pump and solar energy in commercial buildings,the solar coupled ground source heat pump seasonal storage floor heating system is established.In the non-heating season,the system collects solar energy collection panels and stores them in the soil through buried pipes.In the heating season,the heat in the soil is transferred to the load end of the building through buried pipes,so as to make full use of solar energy and improve the recovery of soil temperature.At the same time,in order to improve the indoor thermal comfort of the building floor heating system,neural network prediction algorithm was used for the system to predict the indoor temperature of the building,and the algorithm was optimized to obtain a more accurate prediction algorithm.Finally,a new control strategy is proposed based on the predicted temperature values and real-time weather API data.The research content and results of this paper are as follows:First,TRNBuild was used to establish the dynamic model of building load,and the equipment was selected according to the design parameters calculated according to the specifications.Two operating modes based on different control strategies were proposed for the solar coupled ground source heat pump seasonal storage low-temperature radiant floor heating system.Dynamic simulation model was built by using TRNSYS software,and the energy efficiency,impact on soil and indoor thermal comfort of the two operating modes were compared.Secondly,due to the thermal hysteresis of floor heating,in order to improve indoor thermal comfort,12 factors of 2880h in heating season simulated by TRNSYS were selected as input and indoor temperature as output.MATLAB was used to build the indoor temperature prediction model of single hidden layer lm-LM-BP neural network.Thirdly,aiming at the disadvantages of random selection of initial weight and threshold of LM-BP neural network,genetic algorithm was used to optimize LM-BP neural network,and GA-LM-BP neural network indoor temperature prediction model was obtained.The optimized LM-BP neural network R~2 was 0.952218,R was 0.96296,and MAE was 0.71693.LM-BP neural network R~2 was 0.93267,R was 0.97217,and MAE was 0.82964.Therefore,compared with the traditional LM-BP neural network,GA-LM-BP neural network has higher prediction accuracy and can predict the indoor temperature value of buildings more accurately.Finally,after has been optimized GA-LM-BP neural network,on the basis of traditional control method is difficult to solve the floor heating of the characteristics of time-varying and hysteresis in the neural network predictive control based on feedforward compensation meteorological API data,using GA-LM-BP neural network model predictive control,feedforward compensation and feedback correction and rolling optimization steps using DMC step data to data processing,put forward a kind of feedforward-feedback structure of the neural network predictive control algorithm.