[Objective] This study proposes a two-phase optimal scheduling strategy for the rapid repair of distribution grids that considers strong resource constraints to cope with the problems of heavy repair tasks， strong resource constraints， and low dispatch efficiency under complex fault scenarios of smart distribution grids in new power systems； to improve the emergency response speed to faults； and to reduce socioeconomic losses. [Methods] The first phase aims to rapidly restore power supply to non-failed areas， considering constraints such as the reconstruction of the distribution network in the power outage area and the availability of mobile emergency power generation vehicles. A distribution network rapid restoration repair model is constructed with the objective of minimizing the power outage duration. The second stage addresses the challenge of optimizing the scheduling of multiple repair tasks within a faulted area. An improved serial scheduling generation mechanism is designed， based primarily on the reliability level of the power supply. Furthermore， an improved moth-flame optimization （MFO） algorithm is employed to search for the optimal solution for the second-stage model， ensuring accuracy while improving the solution speed. [Results] Simulations based on a 44-node system indicated that nodes 3 and 12 in the first stage caused a loss of power to the primary loads with a significant weighted loss of the load. Priority allocation of emergency power generation vehicles was required. In the second stage， the improved MFO algorithm achieved an 18% increase in the convergence speed， a 14% reduction in the total repair time， 0.7% reduction in economic losses， and superior algorithm stability compared with traditional methods. [Conclusions] The two-stage strategy significantly improved repair efficiency by coordinating resources and tasks in stages. The improved MFO algorithm performed efficiently in strong-constraint multi-objective optimization， providing new ideas for intelligent distribution network fault dispatching. The proposed optimized dispatching strategy has engineering application value.