To reasonably sectionalize the large system into multiple sub-regions and restore them in parallel is an effective strategy to rapidly recover the system. In this paper, a new system optimal partition model and its solving algorithm are proposed, considering the influence of the parallel recovery process of the subsystems. The modal takes into account the influence of recovery time approximately, the objective function of the model is to minimize the sum of the loss of power outage and the number of tie lines between subsystems, and the constraints are the uniqueness of unit attribution, the power balance of subsystems and the connectivity of subsystems, which makes the partition scheme gained by the model more reasonable and scientific. Since the model is a complex mixed integer nonlinear programming issue, which is difficult to be solved directly, the paper further proposes a decomposition solution strategy, that is, the model is decomposed into a simple partition sub-problem with known recovery time of each load node and a set of subsystems recovery sub-problem with given partition schemes. The two sub-problems are mixed integer linear programming problems and can be easily solved by CPLEX solver. The approximate solution of the original model can be obtained by iteratively solving the two sub-problems. The effectiveness of the model and its solving algorithm is verified by several examples.