Predictive ANN models of ground heat exchanger for the control of hybrid ground source heat pump systems

Hybrid ground source heat pump (HGSHP) system coupled with supplemental heat rejection equipment in parallel configuration is suitable for buildings where the cooling load is much higher than the heating load. Appropriate control is expected to improve its energy efficiency. A control strategy is proposed for cooling season as to compare the temperatures of the water exiting the ground heat exchanger (GHE) and the cooling tower (CT) directly. The logic is to choose the GHE or the CT to reject the condensing heat whenever the corresponding exit cooling water temperature is lower. During operation such a HGSHP system requires the knowledge of both the exit temperatures to take the shift control action. In many situations, however, only one of the two exit temperatures is measurable because only one is working at one moment. This paper develops artificial neural network (ANN) models for predicting the temperature of the water exiting the GHE. A numerical simulation package of a HGSHP system is adopted for training and testing the model. These models are also optimized regarding inputs, learning algorithms and neurons in the hidden layers. Results show that the ANN model can predict the GHE exit temperature with an absolute error less than 0.2 °C.