Abstract: In recent years, particulate burnable poisons have attracted much attention due to their superior properties and are widely used in various reactor design. It is uniformly distributed in the fuel cell, does not occupy the position of the fuel element, and will not cause distortion of the power distribution. Dispersed particulate poisons have multiple dimension for adjustment and optimization, including sizes of particles, concentrations and type of poison particles. They can be adjusted according to actual needs, which can effectively and long-term control the reactive process, with strong flexibility. However, they have special spatial structure and strong spatial self-shielding effect, which leads to microscopic stratification during the burnup process. In numerical simulation, the direct refined solution will bring a huge amount of calculation and great grid density, which has certain challenges. Therefore, a new burnup calculation method based on multi-scale coupling was proposed in this paper. First of all, a new parameter named effective burnable poison proportion fv was defined and a microscopic fine spherical model was built according to the target cell model. Secondly, equivalent homogenization, which was based on the reactivity equivalence strategy, was performed and the fine spherical model was modified into a microscopic homogeneous spherical model. It could provide important parameters for macroscopic model. By coupling of the microscopic spherical model and the macroscopic cell model, the fine solution problem of the dispersed particulate media was simplified into a fast solution of a simple conventional media. It solves the problems of large amount of calculation and high grid density in the process of solving the fine burnup of the dispersed particulate poison in the global scope, accurately characterizes the burnup characteristics of the dispersed particulate poison, which provides a novel idea for the solution of the dispersed particulate media. Later preliminary verification was provided. Compared with current collision probability method, the new algorithm has an overall error in effective multiplication factor at about 200 pcm, while the initial error is 100 pcm. However, a relative high error in the middle and late stages burnup periods is detected. Luckily, this period lasts for a short time. When the burnable poison is burnt out, the error returns to normal level (about 200 pcm). The new algorithm has an error in thermal neutron flux level within 1.25%, and the average error level of the burnable poison number density is 1.41%. In general, the new algorithm is proved to have great performance in neutron physical parameters such as effective multiplication factor, neutron flux density and nuclide number density. Besides, it would be promising in future large-scale computation.