An Intelligent Control Simulation Platform for Nuclear Power Plants Using TCP/IP Real-Time Communication Framework

The current intelligent control simulation for Nuclear Power Plants (NPPs) heavily depends on offline training parameters, which prevents the simulation from presenting a real-time scenario of the actual NPP operation. This paper decouples the existing intelligent control optimization simulation platform into three major parts: Simulink control simulation platform, a real-time communication framework based on TCP/IP, and a Long Short-Term Memory (LSTM) model running independently. The Simulink platform provides real-time operating condition simulation as the basis for prediction models, and extracts real-time prediction results as the feedback compensation for the Simulink control loop. The real-time data exchange between the two is provided by the real-time communication framework. In validation experiments, intelligent PID control achieved performance improvements by obtaining optimized control parameters in real-time. The overshoot reduction rate reached 63.60%, while the maximum adjustment time was shortened by up to 280 s, with an optimization ratio of 73.35% compared to traditional PID control. This study demonstrates the feasibility of implementing online, decoupled training-simulation logic using a real-time communication framework, providing valuable insights for further optimizing the structure of NPP intelligent control simulation systems and a beneficial exploration of the intelligent control in NPPs.