The power grid of megacities like Guangzhou mainly adopts a 3T wiring configuration for its 110 kV network, and its operation modes are highly variable, significantly impacting power grid planning and operation. Given the presence of numerous critical users in the megacity power grid, their power supply reliability is a crucial factor in determining the operational approach. If the arranged operation mode concentrates the power supply paths of certain essential users on the same component, it can drastically reduce the reliability of power supply for these users. An optimized decision model for power supply paths of critical users in megacity power grids is hence established. The objective is to minimize the average load factor of 220 kV lines in the grid, with the requirement that each essential user must have multiple power supply paths originating from at least two different 220 kV substations. To solve this mixed-integer nonlinear programming model quickly, a multivariate decision tree to transform the model into an integer nonlinear programming problem is introduced; subsequently, it is converted into an integer linear programming problem through variable substitution, enabling fast and accurate solutions. Finally, the feasibility and effectiveness of the proposed optimized decision model and solution method are validated using actual data from the Guangzhou power grid.