A Trust-Region Based Real-Time Optimization Scheme for Modular High-Temperature Gas-Cooled Reactor Nuclear Power Plant

The advanced fourth-generation modular high-temperature gas-cooled reactor (MHTGR) nuclear power plant is designed with a unique multi-module with one-turbine layout, which provides significant advantages for module coordination and flexible matching to the load demand of the power grid. However, a major challenge is to maintain stable and reliable operation during load changes under unfavorable conditions of limited process data and operational experience. This work proposes a model-based real-time optimization (RTO) scheme to assist in the operation and regulation of the nuclear system, ensuring the safety of wide-ranging load changes. A parameter mapping based on support vector regression (SVR) is integrated into the RTO scheme to adapt model modification in a derivative-free manner. To ensure the global convergence of RTO iterations, a trust-region policy is incorporated into the algorithm framework. The effectiveness of the proposed approach is validated through simulation and compared with the gradient-based trust-region optimization scheme.