Neural Network Modeling of the Ground Thermal Conductivity for Ground Source Heat Pump Applications

Abstract

The ground thermal conductivity is a key parameter for the analysis of heat transfer between the soil and borehole heat exchangers in a ground source heat pump (GSHP) system. At present, besides in-situ thermal response tests (TRTs), the laboratory analysis for geological samples is another major method to determine the ground thermal conductivity. In the present work, the thermal conductivities of ground samples from the Quaternary stratum in Tianjin were measured at laboratory using the thermal probes. Then, based on the experimental results, a generalized regression neural network (GRNN) model was presented to predict the ground thermal conductivity. Results showed that compared with the conventional regression model, the present GRNN model had better prediction accuracy, and can be used for the comparison and validation of in-situ TRT results during the GSHP applications.