Abstract:It was demonstrated in previous related experimental and modelling studies that an ASHP unit having a novel dual-fan outdoor coil (ASHP / DFOC) can help achieve evener frosting along airflow direction in the outdoor coil, leading to its better operating performances. To enable an ASHP / DFOC unit to efficiently operate at varying ambient conditions, an optimal control strategy for the experimental ASHP / DFOC unit was developed using a Generalized Regression Neural Network (GRNN) modelling approach, which is a variation to radial basis neural networks. A database for developing the GRNN model was established first, and a new coefficient of performance related index, which was the change rate of averaged COP (CRAC) was proposed after analysing the database. Then, based on the developed database and CRAC, a GRNN model was developed and validated, and then used to predict and search the optimal dual-fan operating modes when the experimental ASHP / DFOC unit was operated at different ambient air conditions. Finally, an optimal control strategy for the experimental ASHP / DFOC unit based on those identified optimal dual-fan operating modes was developed and experimentally validated. The experimental results demonstrated that the use of the developed optimal control strategy led to a higher heating efficiency, a longer frosting - defrosting duration and a fewer number of dual-fan switching for the experimental ASHP / DFOC unit, with its averaged COP improved by 9.9 % and 6.8 %, CRAC decreased by 17.1 % and 10.5 %, defrosting frequency reduced by 22.2 % and 12.5 %, and averaged frosting - defrosting duration extended by 32.8 % and 19.4 %, respectively, when compared to the use of traditional single-fan operating mode and a fixed dual-fan operating mode at varying ambient conditions. The study results can help achieve evener frosting of ASHPs with a higher operating efficiency and better feasibility at varying ambient conditions.

An optimal control strategy for ASHP units with a novel dual-fan outdoor coil for evener frosting along airflow direction based on GRNN modelling