The fault reconfiguration of the annular network of the ship power system is an essential multi-objective nonlinear combinatorial optimization problem with constraints. To solve the power supply recovery problem in case the ship power system fails, an improved two-particle swarm optimization algorithm is proposed. One is the primary particle group and another is auxiliary particle group. The primary particle group improves the population initialization and adaptively adjusts the inertia weight and the learning factor, which improves the global optimization ability of the main particle swarm algorithm. At the same time, an improved chaotic local search strategy is adopted in the auxiliary particle swarm, which enhances the diversity of the population and the ability of local optimization and effectively solves the problem of easily to fall into local extremum. The performance of the proposed algorithm is compared against several different optimization algorithms by system simulation. The results show that the algorithm has the superiority in search efficiency and optimization. The algorithm can effectively improve the speed and accuracy of fault recovery, and it shows high performance in dealing with the network fault reconfiguration of ship power system. This work is supported by National Natural Science Foundation of China (No. 51579047), Basic Program of National Defense Technology (No. JSHS2015604C002), Open Projects of State Key Laboratory of Millimeter Waves (No. K201707), Natural Science Foundation of Heilongjiang Province (No. QC2017048), Harbin Natural Science Foundation (No. 2016RAQXJ077), and Fundamental Research Fund for Central Universities (No. HEUCF180407).