Data-driven modeling of residential air source heat pump system for space heating

Air source heat pump systems must operate efficiently during the winter to ensure that energy-saving targets are met and occupants of residential and commercial buildings have acceptable thermal comfort. Investigation on the behavior of air source heat pump systems based on the analysis of real data is important to improve the system performance and attain optimum performance during winter. This will assist governments and industry sectors in formulating energy policies and improving this technology efficiently. Currently, there is no available studies relating the use of data-driven models for improving air source heat pump system. Thus, a large experimental dataset was obtained from four real projects in Beijing, and a hybrid and general model (ambient temperature –20 °C to + 17 °C) coupling various data-driven models was established. A methodology was developed to assess and detect outliers in the dataset that cause incorrect results for the developed model. After removing the outliers, a regression analysis revealed an excellent agreement between almost 8000 experimental data samples and the predictions, with a coefficient of determination of 0.87. Two scenarios were considered for improving and optimizing the thermal performance of air source heat pump systems using the developed predictive model. A growth of 8.18% was observed in the coefficient of performance compared with that of the real operation when the indoor temperature was 18–20 °C. The total system performance increased by an average of 15.93% when the indoor temperature was set between 18 and 20 °C and the rated capacity increased from 11.5 to 12 kW.