Aiming at the serious problem of emission pollution of agricultural tractors, especially limiting the emission of NOx and Soot, taking a certain model of agricultural diesel engine of YITO as the research object, relevant research on engine emission optimization was carried out by using a combination of system modeling and simulation, bench test verification and simulation analysis. Firstly, a three-dimensional model of the combustion chamber of the agricultural tractor was constructed and imported into CONVERGE for combustion emission simulation and emulation, and through the comparison of the test values of the model cylinder pressure and heat release rate with the simulation values, it was proved that the model could describe the internal combustion and emission process of the engine better and had a high degree of accuracy. After that, an artificial neural network was established as an agent model with the shrinkage rate, camber depth, and combustion chamber depth as inputs, and NOx and Soot emitted from the engine as outputs. The goodness of fit R2 and the mean relative error (MRE) were calculated to verify the accuracy of the artificial neural network. Then an improved particle swarm optimization algorithm was proposed on this basis to obtain the optimal parameter combinations of the combustion chamber shrinkage rate, camber depth, and combustion chamber depth to form combustion chamber structure and imported into CONVERGE software to perform emission simulation calculations and compare the emissions with the original combustion chamber. It was found that the combustion chamber structure can reduce the engine NOx and Soot emissions, which provided reference and ideas for the design and development of the combustion chamber system of the related agricultural tractors.