Optimization Of Well Distance For Ground Source Heat Pump Based On Artificial Neural Network

In the actual construction of ground source heat pump projects,various reasons(such as equipment aging,improper design)can lead to ground source heat pump systems not meeting the operational requirements of the design.In particular,the location of buried pipes,or wells,may cause problems related to engineering heat recovery during system operation when the drilling locations are unevenly distributed,greatly reducing the heat exchange efficiency of the system.Therefore,only by scientifically and reasonably determining the distance of buried pipes can the heat exchange efficiency of buried pipes in a ground source heat pump system be controlled to ensure the service life of the ground source heat pump system.The traditional research method for ground source heat pump well distance is to simulate the site through regional experience or establishing mathematical models.However,regional experience sometimes differs greatly from the actual situation;There are also certain limitations in using numerical simulation methods to study.For example,when factors such as site conditions are too complex,the cost of establishing a model is high,and models from different regions and projects do not have generalization.The innovation of this study is to combine numerical simulation with artificial neural network,use the calculation results of numerical simulation as the database of artificial neural network,establish a feedforward neural network FNN model,and use this model to predict the temperature influence radius changes of buried pipes in aquifers under various formation conditions,thereby determining the engineering well distance.In summary,the research content of this article mainly includes the following points:(1)Based on different combinations of influencing factors for the temperature influence radius of buried pipes,1500 sets of three-dimensional numerical heat transfer models considering groundwater seepage are established to obtain data on the change of the temperature influence radius of the stratum around the buried pipes during winter heating and recovery periods.The effects of groundwater seepage velocity,thermal conductivity,porosity,permeability,and annual effective operating power of the system on the temperature field around the buried pipe are analyzed in turn.(2)Build an FNN artificial neural network model,iteratively train four FNN models with different number of hidden layers for 500 times,select the four hidden layers with the smallest average relative error and total relative error as the model structure,and divide the training set and verification set in a 9:1 ratio.The training model is established using Re LU activation function,Adam optimization algorithm,mean square error loss function(MSE),and early stop method.The evaluation indicators include mean absolute error MAE,root mean square error RMSE,and multiple decision coefficients.Through testing,the prediction performance of the model is good.(3)Using actual site monitoring data to verify the accuracy of FNN model prediction,the comparison proves that the prediction curve of the model has certain accuracy and generalization.Based on the prediction curve of the model,an attempt was made to design a Tangshan rural holiday water source heat pump system as a ground source heat pump system,with a buried pipe distance of 7 m.Comparing the four different well layout methods of buried pipes,it is found that compared with other well layout methods,when the buried pipes are arranged in a regular hexagonal shape,the cold accumulation area is the smallest.