Sustainable performance enhancement of a heat recovery ground source heat pump system using field data and machine learning

To mitigate soil heat accumulation that reduces the energy efficiency of ground source heat pump systems in cooling-dominated regions, a novel heat recovery ground source heat pump (HRGSHP) system is designed and a three-year field test is performed in this paper. The system parallels a conventional condenser and heat recovery condenser within a single heat pump, recovering excess condenser heat to provide 50 °C hot water during summer operation. Results demonstrate that the HRGSHP effectively limits soil temperature rise to ∼0.45 ℃ while meeting the air conditioning demand. To further enhance efficiency, a multi-objective optimization framework combining a genetic algorithm and backpropagation neural network (GA-BPNN) model with a technique for order preference by similarity to ideal solution (TOPSIS) is developed. This enables accurate energy performance prediction and optimal operational parameter setpoint determination. The optimized system achieved improvements, with average coefficients of performance for system (COPs) and heat pump (COPu) increased by 27% and 11% in winter, respectively, the energy efficiency ratios for system (EERs) and heat pump (EERu) increased by 21% and 11% in summer, respectively, and operational costs were reduced by 19%. This work provides experimental evidence and optimization guidelines for implementing HRGSHP systems in building applications.