According to the microcosmic configuration of the samples on cross-section, a three-dimensional particle random distribution model of the spherical Al2O3/SIR composites was established based on the concept of representative volume element (RVE). Through sample measurement (macro perspective) and model calculation (micro perspective), the influences of volume fraction, particle size and binary filling schemes of the Al2O3 fillers on the thermal conductivity of Al2O3/SIR composites were studied. Increasing the volume fraction of fillers can increase the contact probability of filler particles. When the volume fraction of Al2O3 fillers exceeds a certain threshold, the thermal conductivity of composites increases slightly first and then decreases with the increment of the particle size. For Al2O3/SIR composites filled with different particle size fillers, as the volume fraction of the reinforced fillers with small particle size increases, the thermal conductivity of composites rises. As the particle size of the reinforced fillers increases, the thermal conductivity rises first and then reduces. The optimal volume ratio of the main fillers and the reinforced fillers will be affected by the total filling content. When the total filling content of the Al2O3 fillers is 30vol%, the particle contact probability is relatively the largest at volume ratio of 2:3 and thus the thermal conductivity of Al2O3/SIR composites reaches a relatively optimal value.