The difficulty of hybrid refrigeration and heating system is how to switch the cold and heat source reasonably. An improved DBN network is used to predict short-term, medium-term and long-term cooling and heating loads, and the accuracy of the model is verified by experimental data. On the basis of load prediction, health assessment technology is used to evaluate the short-term, medium-term and long-term health status of ground sourceheat pump, and corresponding control is carried out according to the assessment results, so as to achieve long-term stable and efficient operation. Finally, the performance of the two different control methods is compared, and the coefficient of performance of the host is improved by 8.5% after adopting the new control method. In conclusion, the control method based on health assessment technology is beneficial to the efficient operation of hybrid system.