Power outages can cause significant interruption to business, infrastructure and interdependent lifeline systems, and daily lives. Having a good understanding of power resilience, such as the likelihood and duration of power outages caused by natural hazards, can help prevent or reduce the economic loss induced by power interruption. In this paper, a multi-hazard power recovery model framework using synthetically generated power distribution networks is proposed. A Monte-Carlo simulation framework is used to sample the damage state of the components and evaluate the functionality of the damaged network. Given the damage situation from each sample, an efficient prioritybased recovery model is proposed to simulate the power restoration process and estimate building-level downtime and recovery using synthetically generated distribution networks. The priority-based recovery model considers substation repairs, emergency responders recovery, and large and small service area recovery to best represent and estimate the real power restoration process. It considers crew scheduling and component-level repair time to drive an optimal recovery of the network functionality. This framework can simulate different types of natural disasters. A model validation study against the 2017 Hurricane Irma for seven counties in Southern Florida is presented. The application to analyse power resilience in the Miami region under hurricane wind hazards, and in the Los Angeles region under seismic hazards for multiple return periods events is presented. The results show that our model can provide an accurate estimation of the recovery process to the ground truth, as well as a building-level power resilience estimation under multiple hazards.