The model reduction problem was investigated for singular systems in this paper. To solve the problem, the Silverman-Ho algorithm was given, and based on this algorithm, an optimal model reduction method for the fast subsystem was presented to obtain reduced-order stable models for singular systems. Sequentially, the H. suboptimal model reduction algorithm was obtained for singular systems. The advantage of the presented algorithm is that the impulsive nature of singular system is preserved in the reduced-order models. Then some necessary and sufficient conditions for the existence of a stable reduced-order system were given and these conditions can be verified numerically. Finally, illustrative examples were given to show the effectiveness of the proposed approach.