Rapid load following of a pressurized water reactor with input delay and load-dependent parameters in nuclear-renewable integrated energy systems

Enhancing the operational flexibility and control precision of load following for nuclear power plants in nuclear-renewable integrated energy systems (NR-IES) is an important and challenging problem. This article proposes a novel rapid and actual control framework by integrating reinforcement learning with a fixed-time disturbance observer for load following of pressurized water reactors (PWRs) with load-dependent parameter variations, system uncertainties, external disturbances and input delay. Notably, this work presents the first systematic solution addressing input delay effects for load following of PWRs, effectively mitigating the induced instability risks that would otherwise compromise NR-IES dynamic equilibrium, energy dispatch efficiency and grid resilience. Furthermore, to fundamentally address the effect of uncertainty, this work pioneers a novel finite-time convergent reinforcement learning method that achieves faster convergence than conventional approaches, where the actor neural network is used to approximate the model uncertainties and the critic neural network is employed to evaluate the control performance. Meanwhile, to robustly estimate both compound disturbances and reinforcement learning approximation errors, a fixed-time disturbance observer is developed, ensuring convergence within a fixed time regardless of initial conditions. Based on the reinforcement learning and fixed-time disturbance observer, a novel controller is developed to make the reactor’s output power follow the desired power within a fixed time, achieving more rapid and actual control than conventional controllers. The stability of the control system is theoretically proven using Lyapunov stability method. Finally, simulation results are presented to illustrate the effectiveness of reinforcement learning approximation, disturbance observer estimation, and the load following performance.