Abstract: A particle system can often be required to analyze the interaction between rice seeds and agricultural machinery components in agricultural engineering, especially in precision rice sowing, efficient harvesting, and grain storage and transportation. The discrete element method (DEM) has been widely used to optimize the component parameters and operational performance. However, the long computation time of the DEM is confined to the rice seed systems, due to the irregular shape of the rice seeds in practical engineering. Ordinary computer hardware cannot fully meet the large-scale requirements of computation. Currently, the maximum number of particles in DEM can reach tens of millions, while the thousands of trillions are usually observed in actual agricultural engineering. Therefore, it cannot meet the simulation requirements of actual engineering, even with the computer accelerators. The significant challenge can remain to simulate the rice seed particle systems at the large-scale of agricultural engineering. Existing coarsening is also limited for the applicability of the DEM. This study aims to simulate the large-scale rice seed particle systems using precise scaling and coarsening in the DEM. The simulation efficiency was also improved to verify the effectiveness. The scaling relationship of the physical quantities was then derived for the precise scaling systems using dimensional analysis. The representative volume units were used to establish an approximate conservation relationship of the mass and momentum between coarse-grained and original systems at macroscopic and microscopic scales. The hybrid rice variety, Taixiang 812 (with a real grain size of 2.76 mm) was taken as the research object. The angle of repose was used as the experimental indicator. Three sets were applied as the target variables, including the internal collapse unloading and stacking, side wall collapse, and bottomless cylindrical stacking of the rice seeds. Seven scaling factors (1.0, 1.5, 2.0, 2.5, 3.0, 3.5, and 4.0) were selected for the simulation, corresponding to a grain size range of 2.76~11.04 mm. A comparison was also made with the real experiment. The results show that there was some increase in the relative errors between the simulated and the experimental values from the three sets of test angles of repose, as the particle size of the rice seeds increased. When the particle size of the rice seeds was 2.76 mm, the relative errors of the three sets were 5.35%, 3.01%, and 2.92%, respectively, indicating an acceptable range. When the grain size of the rice seeds was 2.76 and 5.52 mm (the scaling factors of 1 and 2), the computational efficiency was reduced from 35.00 and 24.70 h to 11.67 and 3.61 h, compared with the variable proportion generalized coarsening. The large-scale rice seed particle systems can be expected to improve the computational efficiency and the accuracy of the simulation using precise scaling and coarsening in the DEM. The finding can also provide a theoretical basis for the large-scale particle systems at the engineering scale.