This paper proposes a hybrid genetic algorithm (GA) for addressing service restoration problems in power distribution systems. The authors have already proposed a GA-based algorithm, called the two-stage GA, and this algorithm has been shown to perform well for small systems. However, it is difficult to apply to large systems from the viewpoint of computation time. To improve the time performance of the algorithm, the authors introduce three kinds of speedup strategy: a local search procedure, greedy algorithm, and efficient maximum flow algorithm. Computational results with several test systems show that the proposed hybrid algorithm can dramatically reduce the computation time compared with the two-stage GA, and can be applied to real-scale systems.