The development of novel materials and the refinement of existing ones demand accurate and efficient computational models. This study proposes improvements to the Lattice Discrete Particle Model (LDPM) and its application to polymer simulations. We present a modified LDPM model designed explicitly for polymers, called the LDPM-P, and enhancements to the generation structure. The research is divided into two interconnected parts, culminating in a cohesive approach that advances the understanding of polymer materials. The LDPM-P model addresses the limitations of the original LDPM model, primarily concerning Poisson's ratio and material parameters. To correctly capture the Poisson's ratio of polymers, we decompose the normal component of the deformation into its bulk and deviatoric parts, following a volumetric-deviatoric split. Using random field theory, we address the shortcomings of the spatial variability of material parameters in the LDPM-P facets and later link it to the internal structure via homogenization. To account for the lower bound of the gradation curve, we implement homogenization with higher scales, which are then mapped to the properties of the LDPM-P facets using surrogate models. These modifications result in a true multi-scale, multiphase LDPM-P model. Following the development of the LDPM-P model, we plan to verify the model on existing baseline tests. Ultimately, we aim to advance the field of polymer simulation by introducing a comprehensive, efficient, and accurate computational model that will pave the way for future material development and refinement.