支持非常规突发事件应急决策的模型形式化表示和模型组合方法研究

The key issues of model application in emergency decision support system are researched due to the unconventional emergency's characteristics such as complexity, diversity, uncertainty, secondary and derivative feature.The research contents are summarized as follows: 1)research the ontology -based formalized representation method for the model field knowledge, characterize the relationship between the knowledge of model fields scientifically and reasonably, and achieve the fuzzy extension and probability extension of the knowledge ontology in model fields; 2)The relationship among the scene features, model parameters and model performance index is evaluated quantitatively and the model selection that supports the multi-objective decision-making tasks under different of scenarios of unconventional emergencies is proposed;3)research how to integrate human-ware interactive evolution into emergency decision that aims at dealing with the unknown fields and uncertain factors and realize the interactive evolution of human-ware and ontology knowledge on the basis of Genetic Algorithm;4)In respect to the complexity and secondary and derivative feature of the unconventional emergency, the hierarchical task network planning based decision global task decomposing method and model integration method in human-ware embedded dynamic credibility network is researched to combine the models in dynamic ways. The study results can provide new modes and methods for the construction of Chinese emergency decision support system, and have a significantly scientific value and broad application prospects in enhancing the ability to cope with the unconventional emergency.

面对非常规突发事件的突发性、复杂性、多样性、不确定性和次生衍生性等特征，研究应急决策系统中模型应用的关键问题。主要包括：1）研究基于本体的模型领域知识的形式化表示方法，科学合理地表征各模型领域知识之间的关联关系，实现对模型领域知识本体的模糊扩展和概率扩展；2）研究基于事件情景特征和模型性能指标的模型评价方法，定量评价情景特征、模型属性、模型参量和模型性能指标的关联关系，支持针对不同突发事件情景中多目标决策任务的模型选择；3）研究人件交互进化式融入应急决策的方法，用以处理未知领域和不确定因素，基于遗传算法实现人件和本体知识的交互进化；4）针对非常规突发事件复杂情景和次生衍生特征，研究基于层级任务网络规划的决策任务分解方法和嵌入式人件的动态信任网络的模型组合方法，实现模型的动态组合。课题成果将为我国应急决策系统的建设提供新模式和方法，对提升非常规突发事件的应对能力具有重要的科学意义和应用前景。

非常规突发事件的突发性、复杂性、多样性、不确定性和次生衍生性等特征限制了单一突发事件模型在突发事件辅助决策过程中的应用效果。本项目针对非常规突发事件应急决策中模型应用方法进行理论方法研究和实例应用验证，提出了基于本体的模型领域知识的形式化表示方法，科学合理地表征各模型领域知识之间的关联关系，实现对模型领域知识本体的模糊扩展和概率扩展，研究了基于事件情景特征和模型性能指标的模型评价方法，定量评价情景特征、模型属性、模型参量和模型性能指标的关联关系，支持针对不同突发事件情景中多目标决策任务的模型选择，提出了人件交互进化式融入应急决策的方法，用以处理未知领域和不确定因素。针对非常规突发事件复杂情景和次生衍生特征，提出基于层级任务网络规划的决策任务分解方法和嵌入式人件的动态信任网络的模型组合方法，实现模型的动态组合。本项目运用这些理论成果建立了民用航空突发事件应急决策原型系统，针对民用航空典型的非常规突发事件进行良好的实验验证，实验表明项目组提出的这些方法能够很好的将模型组合用于突发事件应急决策中，应对非常规突发事件表现出来的突发性、复杂性、多样性、不确定性和次生衍生性等特征。