Abstract: In order to reduce the damage and loss of power systems caused by the extreme natural disasters such as typhoons, a two-stage optimal planning method of the all-stage anti-disaster backbone network considering differential reinforcement is proposed in this paper. Firstly, the differential reinforcement indexes (before the disaster), the invulnerability indexes (during the disaster) and the recovery indexes (after the disaster) for the transmission network that spans multiple typhoon disaster areas are proposed to evaluate the all-stage anti-disaster performances of the backbone network. The scenario analysis is utilized to address the randomness of the typhoon paths. On this basis, an optimal planning model of the all-stage anti-disaster backbone network is established to maximize the integrated index of the anti-disaster performance of the backbone network, and the proposed model is equivalently converted into a two-stage optimization model to improve the solution efficiency. Then, an improved non-dominated sorting genetic algorithm Ⅱ (NSGA-Ⅱ) based on the graph theory repair strategy and the genetic algorithm is proposed to solve the proposed two-stage model. Finally, a case study based on a regional transmission network in Guangdong province is used to verify the effectiveness and advantages of the proposed model.