Abstract: In view of the low efficiency of main grid maintenance plan compilation and difficult risk management and control under the access of high-proportion new energy, taking 166 monthly maintenance tasks of a provincial power grid as the research object, this paper proposes a maintenance scheduling optimization method based on Benders decomposition. A three-dimensional rigid constraint system covering safety guarantee, power supply guarantee and power quality guarantee is constructed. The maintenance scheduling is decoupled and solved by taking it as the master problem and the AC power flow N-1 security check as the subproblem. Benders cuts are generated iteratively based on parallel computing to realize fast optimal solution of large-scale maintenance tasks. Practical application results show that compared with the traditional manual compilation mode, the proposed method shortens the plan compilation cycle from 15 days to 7 days, reduces the proportion of overloaded equipment by 43.48%, completely eliminates hidden dangers of voltage limit violation, increases the monthly new energy accommodation capacity by 20 GW·h, and raises the maintenance intention satisfaction rate from 68% to 92%. The research results can provide a practical reference for the intelligent compilation of power grid maintenance plans.