<p>There is high load density and multiple power supply demand of the urban power grids. Improving the emergency capacity of urban power grids in extreme condition can assist the quick recovery of the system, and guarantee continuous consumption of the critical consumer to enhance the anti-vulnerability of power grid. In this paper, the functional resonance analysis model (FRAM) and system dynamics (SD) were integrated to construct a novel anti-vulnerability improvement method for large urban power system, which considered the impact of emergency input factors on anti-vulnerability and the relationship between them. Firstly, identify the vulnerability factor of urban power grids in extreme events by functional resonance analysis model, and determine the critical factor based on the Monte Carlo simulation. Then put the critical factor into the system dynamic model as vulnerability subsystem, emergency capacity as anti-vulnerability subsystems. Finally, by changing the anti-vulnerability input ratio and total inputs, determine the optimal anti-vulnerability investment strategy, realize the anti-vulnerability income maximize. The simulation results show that the optimal investment ratio is 45% for material resource costs, 30% for human resource costs, 20% for opportunity loss costs, and 5% for financial resources. And in this case, anti-vulnerability benefits would be improved by 4%. If the total input is raised to 3.4 times, the power grid benefit will be increased by 62%. However, there is also a limiting effect between total inputs and benefits. As total inputs increase by 5 × 10⁵, the rate of improvement in benefit decreases from 7.9 to 4.29%, resulting in a decline of 3.61%. The proposed anti-vulnerability model can balance the economic benefit and demand of extreme viability, provide the technical support of effective anti-vulnerability strategy for power grid enterprise.</p>