In this article, an integrated algorithm for multi-criteria group decision-making is presented, based on the linguistic operator of ordered weighted averaging, for selecting the optimal evacuation strategy. The algorithm is multi-level, where the first level involves constructing a task model in the form of an evacuation network. The second level ranks evacuation zones to determine their optimal order based on various criteria, such as capacity, transport accessibility, zone safety, evacuation route safety, and distance from the source of danger. The third level implements macroscopic dynamic flow evacuation, considering the possibility of accommodating aggrieved at intermediate service points that are not shelters. The proposed algorithm is based on the linguistic operator of ordered weighted averaging for aggregating additive linguistic preference relations provided by all experts into a collective additive linguistic preference relation. Based on the developed algorithm, decision-making modeling in an evacuation network is conducted for mass evacuation scenarios, simulating the movement of the maximum possible human flow from dangerous zones to shelters, considering intermediate storage in other nodes. The study proposes transporting the evacuation flow first to shelter-sinks, followed by transporting people from intermediate nodes, taking into account their limited throughput capacities. The developed approach ensures the transportation of the maximum number of evacuees from dangerous zones, which is the sum of those evacuated to safe zones and the excess evacuation flow that does not reach the destination points. The excess flow is stored in intermediate nodes for a specified time and then transported to safe zones.