In order to effectively define the influencing factors of power outage events, improve the accuracy of emergency scenario dynamic deduction models, and avoid the occurrence of power emergencies, a large-scale power outage emergency scenario dynamic deduction model based on improved Bayesian networks is designed.Firstly, five network levels of initial scenario, triggering scenario,outbreak scenario, recovery scenario, and disappearance scenario are analyzed to construct a large-scale power outage scenario network. Then, emergency decision- making subject, object, target,plan, and decision- making environment are considered, the emergency decision-making process of large-scale power outage events is simulated, improved Bayesian networks is used to calculate the probability of movement between multi-level scenario networks, and the dynamic evolution law of events is determined. Finally, based on the input results of large-scale power outage event data, the dynamic inference path and optimal emergency plan including large-scale power outage events are obtained. The experimental results show that after applying the design model, the reduction rate of power outage area is 48.57% , and the actual power outage area of power outage events is significantly reduced. The dynamic deduction process of emergency scenarios has been improved, and the emergency response plan is matched with the actual power outage situation, which can efficiently respond to large-scale power outage events and has high application value.