Performance improvement of a novel trans-critical CO2 air conditioner system with various cabin loads by using surrogate models and genetic algorithms

The efficiently integrated air conditioner systems are essential for enhancing vehicle thermal comfort and safety. Compared with the synthetic refrigerants, the air conditioner systems employing natural refrigerant CO2 offer advantages in thermal cycle performance and environment friendliness. However, it has higher throttling losses and lower system performance during trans-critical operation. To enhance the cooling performance of trans-critical CO2 air conditioner systems, this research integrates an ejector into the system and employs machine learning algorithm to optimize key operational parameters of the system with the aim of proposing a systematic and fast optimal method of air conditioner system to maximize the system performance. Firstly, the simulation model of the air conditioner system was developed and validated, in which the cabin load was varied according to operating conditions. Secondly, the simulation models were adopted to analyze how operating conditions and internal parameters impact the system COP, and a database was established. Subsequently, an artificial neural network surrogate model was employed to predict the performance of the air conditioner system. Finally, the genetic algorithm was applied to optimize the system to maximize its COP. The results indicate that the optimization methods by using genetic algorithm can maximize improve the system COP by 78.97 % and decrease the exergy destruction by 42.63 % compared with other optimization methods. The average COPs of the optimized system were 2.37 and 3.84 when the cabin temperatures were 18 °C and 24 °C, respectively, and the average COP improvement ratios were 18.97 % and 33.91 % compared with the baseline system. In addition, the sensitivity analysis demonstrates that the ambient temperature has the greatest impact on system COP, which is about 53.91 %, followed by relative humidity and cabin temperature, which are 23.78 % and 11.69 %, respectively. In a word, the optimization method by combining numerical simulation, artificial neural network and genetic algorithm provides a new solution for maximizing the performance of trans-critical CO2 air conditioner system, which offers important reference for engineering applications.