Research On Emergency Resource System Deployment Optimization Under Uncertainty

Large-scale emergencies often result in shortages of critical resources in affected areas.It is a difficult and complicated process to carry out emergency rescue in a time-critical environment with supply shortages and uncertainties.Factors such as fairness,reliability,and operational costs need to be considered.As a part of the emergency rescue network,an emergency resource system is the link between supply and demand.The emergency resource system involves emergency facilities,emergency materials,emergency equipment,ambulance personnel and so on.The system should provide emergency resources for the uninjured victims and patients.Therefore,emergency management decision makers not only need to reasonably allocate emergency resources through the system to satisfy the demands of the victims as much as possible,but also need to reasonably dispatch and allocate emergency resources to satisfy the demands of patients through the system.Based on the functional role of the emergency resource system,this thesis studies emergency resource system deployment problem under uncertainties.Firstly,the issue of multi-centers and single-area emergency resource deployment problem with uncertain travel time,imprecise transportation information,and fuzzy supply/demand is investigated.Expert human knowledge using fuzzy control is employed to select reliable rescue paths.This thesis comprehensively considers the location of emergency centers,reliable path selection,and emergency vehicle scheduling,and formulates a fuzzy chance constrained model to ensure that the emergency resources can be delivered to the disaster area on time within a desired probability.To deal with high dimensions and fuzzy parameter information,a new method is presented to transform fuzzy chance constrained programming into a mixed integer programming.This thesis further verifies the reliability and economy of the fuzzy chance constrained model with practical data,and proves the robustness of the model in the case of lack of information.Secondly,this thesis studies an emergency resource deployment problem with multi-centers and multi-areas in disaster management.where the travel time for every pair of center-area is uncertain.This thesis proposes a joint chance constrained model to handle this problem,which simultaneously determines the selection of emergency logistics centers(ELCs).the amount of relief transported to these ELCs,the allocation and routing of vehicles,and the equitable distribution of emergency resources from the ELCs to the disaster areas.In this thesis,a distributionally robust optimization(DRO)model is developed to reformulate the joint chance constrainted model into a nonconvex second-order cone programming,which can be efficiently solved by a proposed fixed point iterative algorithm.This thesis applies the model and algorithm to practical data.The results show that the distributionally robust optimization method outperforms the scenario-based model in reducing cost and improving reliability,and evenly distributes vehicles on paths to curtail the rescue time.Then,this thesis studies the emergency resource deployment problem under emergencies,aiming to dynamically transfer patients among different hospitals,admit patients to,and discharge patients from different hospitals at the lowest number of deaths.In this thesis,the problem is first formulated as a stochastic dynamic programming in which the patients’ health conditions are represented by Markov Chains.The optimal decisions under different conditions and the priorities of different types of patients are given,and a forward dynamic programming algorithm is proposed.To overcome the intractability of the stochastic programming and the uncertainty of probability distribution,a hindsight-optimal method based on distributionally robust optimization is proposed.The model is approximated as a mixed integer second-order cone programming based on the second-order cone uncertainty set.Numerical experiments show that the DRO method is computational scalable and results in high values of transferring with relatively high transferring frequencies,that is.reducing the number of deaths.Finally,this thesis investigates an emergency resource deployment problem under random resource consumption,which considers emergency resource schedule,emergency resource allocation,and patient schedule.A two-stage optimization model is constructed for the problem.In the first stage,the number of emergency resources transported by emergency logistics centers to each hospital is determined.In the second stage,a Markov decision process(MDP)model is used to determine the number of patients transferred between hospitals and the allocation of emergency resources.The optimal strategies under different conditions are given.Considering the unavailability of parameter information and the correlation between demands,a secondorder cone uncertainty set considering partial cross moment information is constructed,the distributionally robust optimization model is transformed into a mixed integer second-order cone programming model,and a rolling horizon algorithm is proposed to solve the model.Through numerical experiments,this thesis shows the advantages of the models considering patient transferring and resource re-transportation in reducing transportation costs and the number of unserved patients compared with the sample average approximation method.In order to effectively improve the response speed and ability of emergency rescue,this thesis focuses on the emergency resource system deployment optimization problem under uncertainties.This thesis uses different decision-making methods to optimization emergency resource system deployment.Four research problems are considered:multi-centers single-disaster-area emergency resource deployment,multi-centers multi-disaster-areas emergency resource deployment,emergency resource deployment under uncertain demand,emergency resource deployment under random resource consumption.This thesis can further enrich and expand relevant theories and methods in the field of emergency resource deployment.