Enhancing operational flexibility and stability of coal-fired units under wide-load conditions: A thermo-chemical frequency-based modeling and control strategy

Bridging the gap between combustion mechanism and practical operation remains a critical challenge for improving the flexibility and stability of coal-fired power generation systems. This study proposes a frequency-based modeling and decision-support framework that integrates reduced-order combustion modeling with flame image analysis for combustion stability assessment and operational adjustment. A multi-physics particles group combustion model incorporating a dynamic equilibrium concept, is developed to characterize thermo-chemical low-frequency fluctuations (<50 Hz), and represent transient behavior via a unified physical quantity. The model predicts ignition time within ±5% under varying particle number densities. Feature importance analysis based on a Random Forest Regressor identifies the primary air/fuel ratio and initial coal particle radius as dominant factors, with relative importance of 69.34% and 58.39%, respectively. A reduced-order surrogate based on Response Surface Methodology achieves high predictive accuracy (R2 = 0.9924, predicted R2 = 0.9605, p < 0.0001), enabling rapid evaluation across operating conditions. Flame images from a pulverized-coal burner are processed to extract real fluctuation frequencies reflecting thermo-chemical timescales. Comparison with model predictions yields relative deviations that define a quantitative stability index in the operating-parameter space. Three representative conditions, with deviations of 13.8%, 17.3% and 4.4%, respectively, and exhibiting different local sensitivities, are analyzed, based on which directions for stable combustion regulation are identified. The proposed framework enables rapid identification of stability boundaries and supports trend-aware control strategies, offering scalable decision support for stable operation under load-following and deep turndown conditions, and facilitating integration into online monitoring systems.