Study on the methods for predicting the performance of a hybrid solar-assisted ground-source heat pump system

Abstract It is critical to find suitable setting parameters for designing a hybrid solar-assisted ground-source heat pump system in the practical engineering application, but the heat pump performance is unpredictable after many years of operation. This paper used 2000 sets of performance data collected from solar-assisted GSHP systems that keep operating over 20 years to simulate long term used heat pump with a professional software called GeoStar. Adopted the classification and regression tree (CART) method, the design of solar energy collector areas can be predicted. The multi-linear regression is also utilized to predict average monthly per meter borehole heat exchange. Seasonal factor decomposition and exponential smoothing are used to analyze the average monthly temperature of the circulating fluid, circulating fluid inlet and outlet temperatures of the heat pump after 20 years when we perform the time series prediction. Experimental results demonstrate that CART, multi-linear regression, seasonal factor decomposition and exponential smoothing are promising for practical applications.