Applying DEM or CFD-DEM to large-scale industrial systems such as blast furnaces or direct reduction shaft furnaces relentlessly reveals the computational limits of these simulation methods. To reduce the required computational resources it is inevitable to decrease the level of detail of the applied method. A potential strategy is the usage of a coarse-grain (CG) approach which lowers the computational demand by using coarser (pseudo) particles to represent a certain amount of original particles. However, due to the violation of geometric similarity, this approach fails to capture effects that inherently depend on particle size. We have proposed a multi-level coarse-grain (MLCG) model of the DEM [1] to alleviate the deficiencies and increase the applicability of DEM coarse-graining. In this model multiple concurrently simulated coarse-grain levels are coupled to adjust the resolution of the system as needed. The MLCG model can also be applied to fluid-particle systems using CFD-DEM. To fully picture industrial plants involving the reduction of iron ore, however, the ability to consider heat transfer and chemical processes is also indispensable. The reduction process of interest in this study is described via a three-layer unreacted shrinking core model considering the different iron oxides hematite, magnetite and wüstite as well as the surrounding gas properties. Regarding the MLCG model, this means additional information transfer between the differently resolved DEM CG-levels as well as between the DEM and CFD components, i.e. particle temperature and reduction state. Furthermore, to reach process relevant time scales, a time extrapolation method [2] is applied in this investigation. The applicability of this model set is demonstrated by the reduction of iron ore pellets at elevated temperatures in a silo.