3D braided composites reinforced with graphene nanoplatelets （GNPs） contribute to improving the thermal conductivity coefficients of the composites. However， the microstructures of 3D braided composites are so complicated that the interior and surface unit cells have different thermal conductivity. Meanwhile， the synergistic reaction of GNPs further improves the challenge of predicting the thermal conductivity of the composites. This paper investigates the thermal conductivity of 3D braided composites reinforced with GNPs by finite element modeling based on multi-unit cell models. Firstly， the thermal conductivity of epoxy resin modified by GNPs is analyzed by 2D periodical unit cell models. Then a detailed multi-unit cell model is introduced to establish the finite element model for predicting the thermal conductivity. Meanwhile， the periodical thermal boundary conditions are also applied to the finite element models. Finally， the effectiveness of the multi-unit cell modeling is validated through two typical cases. The thermal conductivity for the interior， surface and integral unit cell models are analyzed， respectively. The results indicate the thermal conductivity of 3D braided composites is improved obviously with increasing GNPs mass fraction. With the braiding angle increasing， the transverse thermal conductivity of 3D braided composites gradually increases and the longitudinal thermal conductivity decreases almost linearly. Meanwhile， the thermal conductivity of the composites gradually increases with fiber volume fraction increasing. The results provide important theoretical basis for the design of the thermal conductivity of 3D braided composites reinforced with GNPs.