Abstract: As Through-Silicon-Vias (TSVs) in three-Dimensional Network-on-Chip (3D NoC) accompany some overhead such as the cost and the area, in order to optimize the number of TSVs of 3D NoC in test mode and reduce the test time, a new method using evolution algorithm based on cloud model is proposed to research on the test scheduling of 3D NoC and the impact of TSVs number and their allocation in each layer on 3D NoC test. This method combines the cloud evolution algorithm with niche technology and hybridization technique in genetic algorithm. It uses effectively the concepts of heredity, natural selection and community diversity to improve the quality of the algorithm on optimizing speed and precision. Experimental results demonstrate that the proposed method can not only effectively prevent from running into local optimization solution, but also improve the ability and speed of searching the best solution, and that TSVs number of 3D NoC can be optimized to improve the TSVs utilization.