In the modern power system

ultra-supercritical units (USC) face challenges in sustaining long-time stable ultra-low load operation and frequent start-up/shut-down to meet load demands. Therefore

the design of a coordinated optimization control system under wet state and state conversion becomes crucial for extending load boundaries and improving peak-shaving depth. This study first enhances existing the hybrid model to accommodate controller design requirements in wet state and state conversion. Subsequently

a dual-layer multi-model stair-like generalized predictive control strategy is proposed

which optimizes setpoint compensation to address unmodeled dynamic errors through a supervisory optimization layer that estimates these errors and calculates compensation values. Based on this strategy

a boiler-following-based coordinated control system is developed to achieve variable-load operation in wet state and automatic state conversion. Simulations demonstrate that the system has good control quality and robustness. Under load changing rates of 1.0%Pe/min and 2.0%Pe/min

the system achieves MAPE of 1.26% and 1.31% respectively for unit load

with maximum main steam pressure deviations consistently below 0.08 MPa. These outcomes significantly enhance control quality during wet state and state conversion

providing valuable references for designing ultra-low load control systems for USC units in modern power grids.