A dynamic network flow optimization for large-scale emergency evacuation | Zai shi qu yu xing shu san de dong tai wang luo liu zui you hua yan jiu | 災時區域性疏散的動態網絡流最優化研究

The cities in Mainland China and Hong Kong are densely populated. Serious natural or man-made disasters, such as nuclear plant accident, floods, tsunami, wildfire, leakage of toxic gases, tornadoes, earthquake or war may cause huge lost of lives and properties. Evacuation from the hazardous region(s) is commonly used to migrate the ill effects of such disasters. It serves as the last defensive line to protect people by evacuating them from the affected areas so as to reduce the casualties. Despite its importance, the study of evacuation process is still inadequate because it is impossible to perform drill exercise to evaluate the evacuation efficiency. With the recent development of digital computer, some computer programs, such as the MASSVAC, have been developed to evaluate the evacuation plan. However, such programs are mainly concentrated on transportation aspects and in particular the optimization of the traffic system. The evacuation process indeed involves the movement of people and vehicles. A comprehensive study on people and vehicle evacuation is of vital importance. Accordingly, the major objectives of the study are to establish an integrated system for managing large-scale evacuation of densely populated regions by integrating simulation models, visualization tools, GIS, and optimization algorithm. Firstly, a competitive random walker model has been established by taking into account the cooperative behavior of bi-direction pedestrian flow and uncooperative behavior of evacuees under high time stress. The model has been adopted to simulate the evacuation of a subway and an interesting and counter intuitive phenomenon, known as the Braess’s Paradox, has been observed in the people’s moving network. Secondly, a capacity constrained evacuation model (CCEM), based on the queue model has been established on a GIS platform to model a regional traffic evacuation. It provides a cost-effective way to evaluate the evacuation planning in real time application. An iterative assignment has been implemented in the model to allocate the evacuees to the quasi-optimal routes. Thirdly, a time-varying quickest flow problem (TVQFP), based on dynamic network flow, has been developed to optimize the evacuation plan. This algorithm can be adopted to optimize the evacuation shelters, evacuation routes and evacuation schedule simultaneously. It can effectively simulate the effect of unconstrained (unlimited capacity) and constrained (limited capacity) shelters – temporary safe places. Finally, a hybrid evacuation model, integrating the dynamic network flow optimization approach and simulation, has been established in this study. The optimization model was used to optimize the evacuation routes, schedules, and shelters. The simulation model took the optimized strategies and provided a description about the evacuation behavior under emergency situations. The hybrid model made full use of the advantages of both simulation-based model and dynamic network flow based optimization model.

Includes bibliographical references (leaves 238-252)

Thesis (Ph.D.)--City University of Hong Kong, 2006