[Objective] As the accelerated transition toward modern power systems with high-penetration renewable energy integration continuously weakens the equivalent rotational inertia of the grid, 1000 MW ultra-supercritical once-through boiler thermal power units face increasingly strict dispatch performance criteria. Due to their inherent thermodynamic lag and strong non-linear characteristics, these megawatt-class units frequently encounter severe command coupling conflicts and destructive energy clashes when executing automatic generation control and primary frequency regulation instructions simultaneously. To fundamentally resolve these operational bottlenecks, this research introduces a model-driven, coordinated optimization control strategy. 

[Methods] First, adopting a hybrid modeling concept, the intricate steam-water cycle of the boiler is spatially lumped to construct a highly efficient 3rd-order affine non-linear dynamic control model. To guarantee wide-range robustness, an objective-adaptive non-dominated sorting genetic algorithm is employed to perform the online calibration of unknown dynamic parameter matrices, specifically targeting pressure conversion coefficients, energy conversion variables, and thermal time constants. Capitalizing on this reduced-order model's internal boundary constraints, state reconstruction technologies are embedded in the control layer. Specifically, the transient reheat steam flow is extracted through a frequency-domain complementary filtering network to effectively eliminate volume-induced phase delays. Concurrently, the actual equivalent coal feed rate is reverse-calculated via a high-order Butterworth band-pass inversion algorithm to successfully strip away coal transportation delays. Furthermore, aligning strictly with the latest grid auxiliary service assessment rules, a comprehensive coordinated control logic strategy matrix is configured. This matrix features active main steam pressure-raising for energy storage at deep peak-shaving loads, adaptive variable-gain compensation algorithms for small frequency deviations, cross-interlocking protection mechanisms to prevent reverse actions during large frequency deviations, and a discrete micro-step decomposition module for continuous automatic generation control ramp commands. 

[Results] Field test data gathered from an industrial unit thoroughly verify the proposed strategy's exceptional efficacy. The reduced-order mathematical model exhibits remarkable tracking fidelity under wide-range load-varying conditions, maintaining an average main steam pressure prediction error of only 3.15%. Following the control logic upgrade, the initial mechanical response delay for primary frequency regulation was compressed to 2.65 seconds. This drove the monthly primary frequency regulation qualification rate from 62.4% to a stable 81.5%, safely surpassing the penalty-exemption baseline. Simultaneously, the comprehensive automatic generation control performance index registered a leap from 0.51 to 1.18, systematically eradicating power oscillations and inverse adjustments traditionally caused by thermodynamic inertia.

 [Conclusion] Balancing computational efficiency and physical fidelity, the proposed scheme establishes a practical engineering paradigm for 1000 MW units to deeply engage in the ancillary service market through pure software logic reconfiguration. It fundamentally breaks through the response bottlenecks of boiler-turbine coordination and significantly strengthens the unit's support capability for next-generation power systems.