Theories And Methods Of Low Grade Energy Utilization For Ground Source (Heat Pump) Based On Load Prediction

As a renewable energy,shallow geothermal energy has been wildly used in the field of building energy-saving.However,the application of the shallow geothermal energy is always companying with the consumption of high grade energy.Changing the way of the shallow geothermal energy utilization and reducing the consumption of high grade energy will be conducive to improving building energy efficiency further.Due to the average temperature of underground soil approaches that of atmospheric environment,the low temperature water exchanged heat with underground soil,which is regarded as a natural cooling source,can be directly supplied to cool buildings.And this cooling mode is called free cooling in the professional field.It should be emphasized that the free cooling mode still needs to consume a small amount of power for circulating pumps.This paper will focus on performing theoretical analysis,experimental test and simulation study of low grade energy utilization for ground source(heat pump)based on the air-conditioning load prediction.And the research is supported by the National Key Research and Development Program of China during the Thirteenth Five-year Plan: The key technology research of low cost commissioning,operation and management of existing large scale public buildings(Grant No.2016 YFC 0700707).Firstly,four air-conditioning load prediction models based on ANN(artificial neural network)were analyzed theoretically in this study.And an ANN prediction model improved with the sample similarity was proposed.To quantitatively measure similarities between prediction time and historical time samples,the grey relational analysis improved with the information entropy was presented to calculate the synthetic similarity coefficient.Secondly,in order to improve efficiencies of GSHP(ground source heat pump)systems and solve the thermal imbalance problem of underground soil,the heat transfer model of the ground coupled heat exchanger was analyzed theoretically in this study,and then a heat pump system cooling with the ground source directly(i.e.low temperature water in ground-coupled heat exchangers was used to cool buildings directly)was proposed.Moreover,to optimize the operation of the heat pump system cooling with the ground source directly,a piecewise linear control strategy based on indoor and outdoor temperature difference distribution was put forward according to the internal relationship between indoor and outdoor temperature difference and air-conditioning load.Finally,an office building in Tianjin area was selected as a research project.Field investigation and test were conducted.The building physical model was built in the Design Builder software at the same time.Based on the measured data,the prediction accuracies of the building physical,the ANN and the improved ANN models were validated.In addition,a GSHP system model was established to perform the simulation analysis with different operation strategies through the TRNSYS 16.0 software.The results showed that prediction accuracies of BPNN(back progation neural network)and ELM(extreme learning machine)models were improved effectively,when the training samples of the prediction model were selected based on the principle of sample similarity.But this sample data selection method was not applicable for GABPNN(genetic algorithm-back progation neural network)and SVR(support vector regression)models.Compared with conventional strategies,when the piecewise linear control strategy was used to conduct the operation of the heat pump system cooling with the ground source directly,the cooling energy efficiency of the heat pump system in summer was improved effectively.And the running cost of the heat pump system was reduced greatly.Correspondingly,the underground soil thermal imbalance was improved.