A Study On The Model-Based Predictive Control Of The Hybrid Ground Source Heat Pump System

The hybrid ground source heat pump system is applied in buildings where the coolingload is much larger than the heating load. The control strategy has a significant effect onthe performance of the hybrid ground source heat pump systems. In parallel connectedsystems, a new control method is proposed, that is to compare the water temperatureleaving the ground heat exchanger and cooling tower. The problem is only one of the twotemperature can be obtained in actual engineering projects because when one is workingthe other one is idle. It needs to build reliable predictive models to realize the controlmethod.Artificial neural networks (ANN) have been widely used in prediction and intelligentcontrol. This paper tried to build the predictive model of the cooling tower and groundheat exchanger with back propagation networks. Optimization is conducted on inputvariables, number of neurons in the hidden layer and learning algorithm. An officebuilding in Wuhan is selected and a hybrid ground source heat pump system is designed. Adynamic numerical model of the system is set via FLUENT. Data monitored from thenumerical system is used to train and test the ANN models. Then the predictive controlmethod is used in the hybrid ground source heat pump systems and compared with othertwo methods, time based control and temperature difference based control. Conclusionsare summarized after the performance analysis.Results show that ANN could be used to model the ground heat exchanger and thecooling tower. The predictive method could achieve the lowest energy consumption. Itcould make full use of the advantage of the outdoor air and the soil. Further study isrequired in order to determine which method is optimal. This paper gives some directionfor the further study on the energy efficiency of hybrid ground source heat pump systems.