案例推理作为一种重要的决策支持方法，近年来受到学者们的广泛关注。当将传统的案例推理方法应用于应急决策支持时，由于突发事件的复杂性，一个完整突发事件的应对过程实际是由若干面向不同决策目标的“情景—应对”所组成，因而相对于分析相似的历史案例，分析相似的历史情景更可能获取具有针对性的决策支持。为此，论文从应急决策的实际需求出发，对基于突发事件案例的情景推理方法进行了研究，提出了突发事件情景的多维量化表达方法和基于多维情景模型的案例结构化表达方法，构建了突发事件情景的相似性评估框架，实现了基于情景推理的应急决策支持和案例分析云平台的设计。论文对突发事件情景的多维量化表达方法和基于多维情景模型的案例结构化表达方法进行了研究。论文从实时应急决策的需求出发，将“突发事件情景”界定为“一个带有时空信息的、能够触发应急决策的承灾对象灾情片段”，并在此基础上提出了突发事件情景的划分准则，建立了由触发事件表达、承灾对象表达、对象灾情表达、救援方案表达和时空属性表达所组成的多维情景表达模型，构建了用以支持灾情严重度评估的多维情景空间，并给出相应的数学形式化描述，保证了模型的一般性和规范性。基于多维情景表达模型，论文提出了基于情景的突发事件案例结构化表达方法，建立了情景演化的时空框架，实现了灾情发展演化全过程的结构化呈现。论文基于多维情景表达模型，建立了融合概念结构相似性评估、混合属性相似性评估和文本相似性评估的突发事件情景相似性评估框架，并在此基础上提出了基于情景推理的应急决策支持范式，构建了基于主观效用评估与客观效用评估的备选方案综合效用评估模型，并以油罐火灾的应急决策支持为例对基于情景推理的应急决策支持范式进行了验证。基于上述研究成果，论文设计开发了基于情景推理的突发事件案例分析云平台，该云平台能够以“情景”为单元对灾情发展过程进行可视化呈现，实现基于云端的案例信息的共享分析，既能为实际应急决策提供支持，也能为开展相关研究工作提供技术手段和方法支持。
As an important decision support method, case-based reasoning has received extensive attention from researchers in recent years. However, due to the complexity of emergencies, the whole response process of an emergency case is actually composed of several ‘scenario-based response’, each of which is tailored for a certain decision objective. Therefore, compared to analyzing similar emergency cases, analyzing similar historical scenarios often provides more targeted emergency decision support. To this end, this dissertation started from the actual needs of emergency response to study the scenario reasoning method based on emergency cases. A multi-dimensional quantitative representation of emergency scenarios was proposed. Based on this, a scenario-based case representation, as well as a similarity assessment framework for emergency scenarios, was developed. Furthermore, a scenario-based reasoning decision support paradigm was proposed, and a case cloud was developed.The multi-dimensional quantitative representation of emergency scenarios and the scenario-based representation of emergency cases were studied in this dissertation. Starting from the needs of emergency decision-making, the concept of ‘emergency scenario’ was defined as a spatio-temporal episode of disaster that can trigger certain emergency decision-making. Based on this definition, a multi-dimensional scenario representation model, composed of trigger event representation, disaster object representation, disaster status representation, emergency response representation and spatio-temporal attributes representation was developed, as well as the formal mathematical description of each representation. Based on the multi-dimensional scenario representation, a scenario-based representation framework for emergency cases was proposed, which defined the spatio-temporal evolution relation on different scenarios to develop a more structured presentation for the whole disaster evolution process of emergency cases.Based on the developed multi-dimensional scenario representation, the similarity assessment framework of emergency scenarios was developed, which combines the concept similarity evaluation, multiple attributes similarity assessment, and text similarity assessments. Based on this, a scenario-based emergency decision support paradigm was developed and a comprehensive utility assessment method, including objective utility assessment and subjective utility assessment was proposed. To verify the efficiency of the scenario-based emergency decision support paradigm, this dissertation took storage tank accidents as an example.Based on the aforementioned study results of this dissertation, a scenario-based reasoning case cloud was developed. In the case cloud, the disaster evolution process can be visualized by several ‘scenarios’, and the online case editing function contributes to information sharing and updating of emergency cases. The developed case cloud can provide support for emergency decision-making and can contribute to scientific research of emergency cases.
