Quadratic programming-based nonlinear model predictive control of coal-fired power plant integrated with molten salt thermal storage system for its flexibility improvement

To address the challenge of simultaneously enhancing fast load response and operational stability of coal-fired power plants (CFPPs) coupled with molten salt energy storage during deep peak shaving, this study proposes a bi-level optimal control framework based on load command decomposition. The framework consists of an outer-layer parameter optimization and an inner-layer nonlinear model predictive control (NMPC). In the outer layer, sequential quadratic programming (SQP) is employed to tune the key structural parameters of the NMPC adaptively, providing a reliable foundation for the inner-layer control. The inner-layer NMPC handles multivariable coupling and nonlinear constraints, ensuring both dynamic and steady-state performance under rapid load variations. On this basis, a load decomposition factor is introduced within the NMPC, and the optimal power allocation is determined through hierarchical search. Simulation results indicate that, under medium-to-high load conditions, the optimal load decomposition scheme assigns 2.10% Pe/min to the CFPP and 2.90% Pe/min to the steam generation system (SGS), effectively balancing the load-ramping rate and dynamic deviation. Under extreme load-varying conditions, the proposed SQP-NMPC bi-level optimization control scheme achieves a maximum load ramp rate of 4.61% Pe/min, representing improvements of 7.45%, 11.8%, and 2.2% compared with PID, FF-PID, and conventional NMPC approaches, respectively. In addition, under disturbance scenarios involving molten salt temperature and feed-water temperature, the proposed method demonstrates significantly enhanced dynamic disturbance rejection capability. The bi-level control strategy fully exploits the transient power support capability of the SGS, thereby facilitating flexible peak shaving of the CFPP-SGS integrated system.