Research on ammonia combustion is gradually attracting widespread interest. Developing a simple, widely applicable, and reasonably simplified reaction model for ammonia combustion is of great significance for computational fluid dynamics (CFD) research. This study simplifies the detailed reaction model using the Directed Relation Graph method with Error Propagation and Full Species Sensitivity Analysis method. Then, sensitivity analysis is used to identify crucial reactions that influence combustion characteristics, while genetic algorithms are employed to optimize the kinetic parameters of these reactions. In heat insulation steady perfectly stirred reactor simulation, and 1-D premixed laminar flame simulation, the simplified model is compared with the detailed model. It is found that the simplified model provides reasonable predictions for species concentration, gas temperature, and flame velocity across a wide range of premixed gas equivalence ratios (0.6 1.4), oxygen concentration (21 % 100 %), and temperature (300 K 500 K). In cases of elevated oxygen concentrations, the prediction error of the simplified model increases but remains within acceptable limits. The prediction of the simplified model in a CFD simulation is also close to that of the detailed model. Moreover, the simulation time cost is reduced by 48.11 %, demonstrating the efficiency of the simplified model in conserving computational resources. In an unsteady closed homogeneous (0-D) simulation, however, the simplified model could only accurately predict the ignition delay time when the oxygen concentration of the premixed gas is greater than 40 %.