【作者】 叶永；

【导师】 刘南；

【作者基本信息】 浙江大学 ， 管理科学与工程， 2014， 博士

【摘要】 地震灾害具有突发性、成纵性、续发性等特点,不仅可以直接造成工程设施破坏和人员伤亡,且往往引发一系列次生灾害和衍生灾害,以灾害链的形式造成更大的破坏。应急资源配置是地震灾害应急救援的重要保障,应急决策者必须在诸多不确定性条件下做出如何将有限资源在有限时间内公平高效地从出救点运送到受灾点的决策。同时,地震灾害对运输道路的损坏使得应急资源配置决策更加困难。因此,地震灾害应急资源配置是一个供应不确定、需求不确定、时间紧迫环境下的多出救点、多受灾点、多目标、多约束、多运输方式、多运输工具,同时具有后续共享性和信息更新性的复杂问题。本文在分析前人研究的基础上,分两条主线：从后续共享性出发和信息更新性出发,分别对震后应急资源配置决策方法进行建模求解,其中基于后续共享模型可以协调前后阶段的应急资源供需情况,而基于信息更新的模型则可以根据当前阶段的实时信息做出及时有效的具体配置策略。首先,地震灾害应急资源配置具有时间连续性,前后阶段物资需求和供应具有极大的不确定性,使得地震灾害应急资源配置决策非常复杂。考虑到在实际应急资源配置过程中,后续阶段可以共享部分前面阶段的资源,而前面阶段则不可共享后续阶段资源的特性(即应急资源配置的后续共享性),以资源配置效果和运输效率为目标,综合考虑纵向配置(时间纵向上的多阶段配置)和横向配置(某阶段物理空间上多对多的物资配置)方案,建立基于后续共享的震后应急资源一体化配置决策模型。该模型运用确定的方法描述不确定情景,并能获取纵向配置部分的解析解；横向配置部分则可通过运输模型快速求得,求解及其方便快捷且精确度高,非常符合应急环境。最后,数值仿真分析验证了一体化配置模型获取应急物资配送方案的可行性。其次,考虑到应急情况下的信息不完备性和可更新性,以及地震灾害应急资源配置复杂性、动态性和序贯性等特点,灾害应急资源配置决策需要综合运用灾害历史信息和样本信息,是一个“观测—决策—配置”的多阶段序贯决策过程。以随机变量的形式记录道路损毁率的历史信息和样本信息,并在此基础上计算应急情况下的资源运输时间。在根据公平原则确定各受灾点资源配置量的基础上,通过应用贝叶斯分析、最优化理论等对基于道路损坏率信息更新的应急资源“观测—决策—配置”序贯决策问题进行系统建模,建立基于单维信息更新的震后应急资源配置决策模型,并设计基于矩阵编码的遗传算法进行求解。最后,通过数值仿真验证模型和算法的有效性。再次,地震灾害带来的诸多不确定因素导致应急救援物资需求激增、供给不确定、应急物流时间限制等紧迫问题,亟需一套能够根据地震灾害信息、物资需求与供给信息和交通网络信息等来实现应急物资的实时有效配置。应用贝叶斯分析理论将灾害对需求和应急物流时间影响体现在决策失误损失和物流失误损失上。考虑到应急资源需求是与受灾人口转移安置情况相关的、交通网络畅通度是与道路损毁程度相关的,应用两个多维随机变量来描述地震灾害应急资源配置中的应急资源需求和交通网络通畅度。在建立无信息更新模型的基础上,再应用贝叶斯理论建立多供应点、多需求点、多阶段、多运输方式、多运输工具的基于多维信息更新的震后应急资源综合配置模型,并改进基于整数矩阵编码的遗传算法对其进行求解以获取各阶段的应急物资配送方案。最后,将模型和算法应用于汶川地震应急物资配置中,进行实例分析,以验证模型和算法的有效性。更多还原

【Abstract】 Earthquake has the characters of suddenness, pervasiveness, successiveness. A severe earthquake destroys the facilities and causes casualties at the same time, and induces secondary disasters and derivative disasters which will cause major destruction. The resource distribution for post-earthquake response is very important. Post-event response for earthquake is a complicated task with challenges of surging demand, uncertain supplies, and rough transportation in the face of infrastructure vulnerabilities. Therefore, resource distribution for post-event response after the earthquake is an uncertain planning problem to distribute many commodities to affected areas from distribution centers with multi-objectives, multi-constraints, multi-transportation mode and multi-transportation tool. Emergency planners need to make effective humanitarian logistics plans to efficiently allocate vehicles and relief resources under uncertain circumstances. Especially, they need to consider the follow-up sharing character (the subsequent phases should share part of former phases’ resources if previous phases’supplies were relatively surplus compared with the following phases’supplies) and group-information update. Based on the previous research, the resource distribution for post-earthquake response is studied by two different aspects:the follow-up sharing character, which coordinates resources between different phases; and group-information update, which makes the specific resource distribution plan just in time according the current information.Firstly, the resource distribution for post-earthquake response has the character of temporal continuity. It is necessary to coordinate the supplies between former phases and following phases to make the distribution plan more effective. To explain the effectiveness and efficiency of distribution plans, this paper addresses the resource allocation effectiveness losses (RAEL, the losses caused by the mismatch between supply and demand in impacted areas) and the emergency logistics time costs (ELTC, the time costs caused by logistics processes under emergency conditions). Considering the reality of resource distribution for post-earthquake response, the following phases can share parts of the former phases’resources, but the former phases cannot share any part of the following phases’resources (the follow-up sharing character). Therefore, based on FSC, this paper proposes an integrated model (IM) that aims to minimize RAEL and ELTC. The IM combines a time dimension model (TDM, which coordinates the phases of demands and supplies) and a space dimension model (SDM, which generates a specific distribution plan for the first phase). An analytical solution is proposed for TDM, whereas SDM is solved through a transportation programming model. The IM can be solved efficiently, which makes the proposed methodology fit the emergency circumstance well although the decision is time-constrained. Besides, a numerical simulation study is proposed to analyze the feasibility of this model.Secondly, unlike the normal logistics, emergency logistics in natural disasters is much more complicated. In emergency logistics planning, the information for decision-making is usually not complete and is updated every second. Besides, considering the dynamic and sequential process of post-earthquake resource distribution, the decision-making of emergency resources allocation in a natural disaster is a multi-phase process of "sampling-planning-dispatching". Therefore, both historical and current sample information are used to make effective plans in the proposed model by employing Bayesian information update approach. In this paper, a random variable is used to record historical and sample information for describing the road affected information. Besides, the allocation amount of each affected area is decided by the equity principle. On the base of this, a concept of transportation time cost due to logistics processes under emergency conditions is proposed. Therefore, by using Bayesian analysis theory and optimization theory, a sequential approach of "sampling-planning-dispatching" is proposed to guarantee the resources supply in the affected areas. In the solution approach, a matrix-coding-based genetic algorithm is developed to solve the model. Finally, a simulation study is conducted to verify the efficiency and effect of the proposed methodology.Finally, the uncertainty of earthquake brings the surging demand, uncertain supplies, and rough transportation time problems. There is absolutely a need to find a methodology to make effective and efficient distribution plans according to disaster, demand, supply and transportation information. Based on the Bayesian analysis theory, two losses are addressed to explain the influence of a disaster on supply, demand, and humanitarian logistics. The two losses include losses caused by the mismatch between supply and demand in affected areas and the time losses caused by logistics processes under emergency conditions. Considering the demand is related to population transfer status, and the transportation condition is related to the road damage level, a multi-period planning model with group information updates (GIU) is established based on the model without GIU using Bayesian theory. The established model describes a setting in which commodities are transported from dispatching centers to affected areas through multi-transportation modes of delivery in each emergency response period. Then, the model with GIU is revised into a single-objective model, and then a matrix-coding-based genetic algorithm is developed to solve the revised model. Finally, the proposed methodology is applied to the humanitarian logistics problems of emergency response encountered during the Wenchuan Earthquake in China. Computational results show that the proposed methodology can generate specific logistics plans for allocating relief resources according to updated information.更多还原