Research on thermal performance of ground source heat pump based on artificial neural network predictive model

Abstract

The long-term operation of ground source heat pump systems can affect the soil environment, causing a decrease in heat transfer performance of underground heat exchangers. Based on a real project, this study aims to determine the actual soil thermal property and propose optimization schemes. A combination of 1D-3D heat transfer model and artificial neural network model was proposed. This research confirms the possibility of using the outlet water temperature as input to predict the soil thermal property parameters. Through model calculations, the soil comprehensive thermal conductivity of 1.78 W/(m·K) and comprehensive heat capacity of 1909.8 J/(kg·K) was determined. On this basis, the impacts of burial depth, spacing, and initial ground temperature on the heat transfer of the underground heat exchangers was investigated. The results showed that the heat exchange is logarithmically related to the depth and spacing of the boreholes, but the heat exchange decreases as the initial ground temperature increases, following a quadratic function relationship. For the current hourly cooling demand, it is recommended to have the pile spacing of 5.5 m and the depth of 110 m. With the addition of a cooling tower that takes on 2/3 cooling load, the system's average coefficient of performance can reach 5.62, resulting in a 19.6 % improvement.