Abstract: Aiming at the problems of high NOx emission and high smoke exhaust temperature during the operation of the offset combustion boiler of ultra-supercritical unit, a low nitrogen combustion optimization study based on the improved greedy algorithm was carried out. A combustion model with mass conservation, energy conservation and momentum conservation was established to describe the characteristics of fuel combustion, heat transfer and pollutant generation in the combustion process, providing a theoretical basis for optimization. Based on the improved greedy algorithm, the key operating parameters of the burner such as swirl intensity, fractional air supply ratio and burn-out air volume were optimized, and the optimal solution was gradually approached through the neighborhood search strategy and iterative optimization process to achieve NOx emission optimization. Through optimization and comparison experiments, it is proved that the improved greedy algorithm can effectively reduce NOx emission concentration, from 450 mg/m³ before optimization to 320 mg/m³, with a decrease of 28.9%, and the exhaust temperature is reduced from 136° to 125°. The results show that this method has high engineering application value in the low nitrogen combustion optimization of ultra-supercritical units.