Shape-anisotropic granules and their surrounding interphase networks are significant constituents in granular materials. Structural and physical configurations of these constituents significantly affect the overall thermal performance of granular materials, specifically microstructure-dependent thermal conductive properties. It has been a key but unresolved issue how to quantitatively understand the microstructure evolution of conductive interphase interacted by densely packed non-spherical particles triggering the change of thermal conductivity of granular materials. In this work, we devise a powerful scheme by using the discrete element method (DEM) and the dual-probability-Brownian motion simulation (DP-BMS) to accurately and efficiently predict the effective thermal conductivity of granular materials composed of homogeneous matrix, conductive (soft) interphase around randomly-dispersed elliptical particles over a broad range of aspect ratios with widespread applications, such as cracks, pores, fibers, cellulose whiskers, silicate nanorods, and aggregates. Comparison against extensive numerical and theoretical data validates that such the scheme can well predict the effective thermal conductivity of multiphase granular materials with densely packed non-spherical particles that is just the intrinsic limitation of the classical micromechanical homogenization theories. In this scheme, the DEM provides a direct means of investigating the time-dependent microstructure evolution of granular materials with elliptical particles from a loose parking state to a dense packing state. The DP-BMS provides an effective technique for predicting the effective thermal conductive transport properties of multiphase granular materials. By comparing with traditional numerical strategies like the finite element method (FEM) and random walk model (RWM), the DP-BMS is more user-friendly and efficient to accurately predict the effective thermal conductivity. This scheme can be regarded as a general procedure that is readily applicable to predictions of other transport properties of two-dimensional or three-dimensional multiphase granular materials. Furthermore, we use the scheme to probe the influences of the shape and high packing density of particles and the thickness and fraction of soft interphase on the effective thermal conductivity of granular materials. The results elucidate rigorous component-structure–property relations, which can provide sound guidance for composite design and microstructure optimization.