Optimization of airfoil fin PCHE for the power conversion system of lead-based reactor based on reinforcement learning

As the Gen IV nuclear reactors, the lead -based reactor employs the S-CO2 Brayton cycle. As the intermediate heat exchangers, high or low temperature regenerators and precoolers, the thermal hydraulic characteristics of the PCHEs directly affect the efficiency of S-CO2 power cycle efficiency. In this work, two reinforcement learning algorithms, namely the Q -learning and DQN are used to optimize the key geometric parameters that affects the thermal hydraulics performance of the airfoil fin PCHE (AF-PCHE) with the S-CO2 as the working fluid. Furthermore, the thermal hydraulic performance of the optimal designs derived from these two reinforcement learning algorithms is compared with that of the original design. The results demonstrate that the two reinforcement learning algorithms significantly improve the optimization efficiency of AF-PCHE design. Specifically, the Delta Q/Delta P of the optimized AF-PCHEs are improved by 173.89% and 189.38%, respectively. In addition, the Nu/ Eu of optimized AF-PCHEs are increased by 129.15% and 141.94%, respectively.