Investigations on optimal discharge pressure in CO2 heat pumps using the GMDH and PSO-BP type neural network—part B: Experimental study

•An innovative real-time control strategy of the transcritical CO2 system was achieved in the test prototype.•The precision of the PSO-BP neural network based control method was experimentally validated.•The control effect of the proposed method was found much better than that of current correlations.•The system performances were given under the supervision of the proposed control method. In this second part of a two-part article, the Particle Swarm Optimization (PSO) based Back-Propagation Neural-Network (BP) based algorithm for the discharge pressure controlling was experimentally achieved based on a subcooler-based transcritical CO2 rig, for further developing an acceptable real-time control approach. The detail of the control strategy in practice was clearly shown including the recirculating water PID control, the PSO-BP based discharge pressure optimization and the electronic expansion valve regulatory mechanism. Besides, the optimal discharge pressure sought by PSO-BP and corresponding system performances were compared with the results from Wang/Liao's predictions and the tested values, which validated the prominent effectiveness of the PSO-BP method due to its satisfactory consistency with the tested data. Additionally, the subcooler-based rig under the discharge pressure from PSO-BP control had more than 15 and 25% improvements over the baseline cycle under floor heating and radiator conditions, respectively, which provided an innovative and appropriate idea for developers and manufacturers.