ConclusionsThe performance prediction of GSHP system by real-time mon-itoring data and data-driven models is studied in this paper. DMtechnologies are simultaneously used to process the monitoringdata and find the required inputs for data-driven models. BPNNalgorithm is used to establish the data-driven models, in which theoutputs are COPsys(EERsys). The long-term performance of GSHPsystem is obtained by the results of the data-driven models. Therelationship between the short-term and long-term performanceof GSHP system is investigated for the purpose of predicting thelong-term performance of GSHP system by a short-term monitoringdata.The important parameters used in data-driven models areselected out. And six classical sorting algorithms which are suit-able for numeric variables are used to extract the mapping betweeninputs and outputs. Four basic metrics, including MAE, Std AE,MAPE and Std APE are used to compare prediction accuracy andBPNN is the most accurate algorithm in the three models.