Agricultural non-point source pollution is a difficult and prior point in the field of environmental pollution control due to its dispersion, hysteresis, and uncertainty. In recent years, the process model of non-point source pollution migration and transformation is developing rapidly. In contrast, the research on the simulation evaluation model of agricultural non-point source pollution management policy is lagging behind. The existing evaluations and opinions on policy effects are often merely carried out on a regional scale, which is insufficient to simulate the policy effects heterogeneity caused by the spatial heterogeneity of resource endowments and natural processes. Therefore, realizing the progress of agricultural resources and the environment from macro-management stride forward to precise management, and continuously improving the efficiency of resource utilization is an important development trend in the future. Especially for agricultural non-point source pollution that is closely related to and affected by the spatial heterogeneity of resources and the environment, taking account into the characteristics of geographical environment elements, considering the heterogeneity of environmentally-constrained agricultural production behaviors, and realizing the spatially explicit effects of related management policies are hardly inevitable development trends that should be paid more attention in the upcoming future. This paper systematically sorts out and analyzes the relevant progress of the research on the simulation model of agricultural non-point source pollution management policies in three dimensions:top-down macroscopic target constraints, bottom-up microscopic behavior conduction, and microscopic up and down coupling. Meanwhile, the strengths and weaknesses of the basic theories and methods of each above model are researched and summarized. Furthermore, this paper proposes a simulation model of agricultural non-point source pollution policy that integrates the cognition of the process law of the water system in the basin and combines the top-down constraints of the macro policy objectives as well as the bottom-up conduction of the micro-subject's behavior. This model can realize the nested conduction, parameter inter-examination, and systematic prediction of policy simulation across the country, region, river basin, and grid across scales, to solve the complex systemic problems caused by the diversity of prevention and control of agricultural non-point source pollution and the extensive characteristics of the control object. Through policy simulation on the cross-scale parameter nesting, result inter-examination and quantitative evaluation in the country, region, river basin, and grid scales, the policy evaluation and assessment are quantified. As a result, these simulation results of policy effects present spatial explicit expression, and then realizing the policy management of agricultural non-point source pollution control from extensive to precise direction.