Neural network prediction methods of power consumption for GSHP system with bilateral variable flow

In order to realize the optimal control of variable frequency circulating pumps for ground source heat pump (GSHP) system it is necessary to build the prediction model of the total power consumption of GSHP system based on running data. Firstly the power consumption analysis of GSHP system with bilateral variable flow is presented. Then a Hyberball Cerebellar Model Articulation Controller (HCMAC) prediction model of power consumption for GSHP system is established. The inputs of model include circulating pump frequency at user-side circulating pump frequency at ground-source side and current air conditioning load. Finally according to the characteristics of the learning data of the GSHP system a differential parallel HCMAC learning method is proposed to improve the learning accuracy. The simulation experiments are performed according to the learning data provided by TRNSYS simulation platform. The experimental results show that the accuracy of differential parallel CMAC prediction model is better than that of the general HCMAC prediction model.