Study On The Identification Of Soil Thermal Properties Based On Artificial Neural Network

The energy consumption and environmental pollution caused by building heating and air conditioning is an indisputable fact and a common concern.It is an important implementation area of the national energy conservation and emission reduction strategy.Using clean renewable energy to achieve building energy supply is the fundamental way to solve the above problems.Among many clean energy supply technologies,the ground source heat pump technology of shallow soil heat energy utilization has the characteristics of energy saving,high efficiency,stable performance,no pollutant emission,meeting the needs of cold and heat at the same time,and low operating cost.This technology has been widely concerned and applied in many projects at home and abroad.Although the ground source heat pump system has unique advantages,it is also limited by other factors.As the basic parameters of GSHP system design,the accuracy of the parameters will directly affect the design of GSHP system,and then affect the economic benefits and long-term operation performance of the system.Aiming at the problems of poor identification accuracy of underground soil heat exchange parameters and extensive design of buried pipe heat exchange system,a fast and accurate identification method of soil heat exchange parameters is proposed in this paper.Firstly,based on the principle of on-site thermal response test,the heat transfer process between the ground heat exchanger and the surrounding soil,which is the core component of the ground source heat pump system,is analyzed theoretically.The CFD theory is used to model the underground heat transfer process,and a high simulation experimental platform is established.Set up a model experiment platform with controllable experimental conditions for experiment,and verify or modify the model and simulation program through experimental data;Then,BP artificial neural network is constructed,and the established three-dimensional high-precision simulation platform is used to change the value of soil thermal physical parameters.The simulation thermal response experiment is carried out,and the change rule of the characteristic parameters of the measurable thermal response(the inlet temperature,the outlet temperature,the wall temperature,the core temperature of the borehole)is measured in real time.On this basis,the soil thermal physical parameters and different measurable thermal responses are established.The mapping relationship between the combination of characteristic parameters is used to train the BP neural network by using some of the data in the obtained mapping relationship,and the identification accuracy of the trained neural network is compared and analyzed with the rest data.The best combination of measurable input characteristic parameters isdetermined by comparison,and the requirements of different test conditions(thermal response test time interval,heating power,circulating water flow rate)are studied The training length of neural network data is required to determine the optimal thermal response test parameters and conditions;Finally,from the actual construction survey situation,this paper analyzes the influence of noise data generated by different types of measurement errors on the accuracy and propagation law of the identification method proposed in this paper,and puts forward the noise reduction method for the initial data preprocessing,in order to improve the practicality of the identification method in the project.