面向DS证据理论的关联信息融合研究

项目名称： 面向DS证据理论的关联信息融合研究

项目编号： No.61503237

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 苏晓燕

作者单位： 上海电力大学

项目金额： 20万元

中文摘要： DS证据理论在信息融合领域应用广泛。如何有效地融合关联信息是证据理论应用中最具挑战性的关键问题之一。现有的关联证据融合方法主要集中在证据结构层次的相关性分析，忽略了对最根本的关联本质的探索，导致融合结果不尽合理，极大影响了融合系统性能。针对这一问题，本项目从证据理论的数学原理入手，首次将关联成因分为系统内部关联和外部感知器关联，并建立描述不同关联情形的数学模型，合理度量关联程度大小。提出从外部关联到内部关联的智能化分层处理融合算法，以全面处理各种关联关系，并确定系统的关联阈值作为判定条件：当关联程度小于设定阈值时，采用经典的融合规则；反之，才选用较为复杂的关联融合算法进行融合，这有利于降低系统的计算复杂度。最后以电力系统的可靠性分析及安全评估为背景，验证并应用所提出的关联证据融合方法。通过本项目研究，可以有效实现关联证据的融合，提高信息融合系统性能，对证据理论的实际应用具有重要意义。

中文关键词： 融合规则；DS证据理论；关联信息融合

英文摘要： DS Evidence Theory is widely used in the information fusion system. How to efficiently combine dependent evidence is one of the most challenging and important tasks. Previous methods for dependent evidence combination are mainly focus on the structure level, which ignore the basic nature of dependence. This may cause the unreasonable result and degrade the performance of information fusion system. To address the issue, this project analyzes the intrinsic logic and causes based on the mathematical principles of Evidence Theory, and first divides the causes of dependence into two categories (i.e., interior dependence and exterior dependence), and constructs mathematical models according to different dependence relationships and measures the dependence degree properly. Also, we establish an intelligent multi-level combination model to deal with general dependent situations, and define threshold value of the dependence degree: when dependence degree is less than the threshold, the classical combination rule is adapted; otherwise, the more complicated dependent evidence combination method is adapted. This contributes to the reduction of computation complexity. Finally, applies the proposed methods into the field of reliability analysis and risk evaluation of electric power systems. Through the research in this project, the dependent evidences can be fused effectively, and the performance of the information fusion system will be consumedly improved, which are of great importance to apply Evidence Theory in practical information fusion system.

英文关键词： Fusion Rule;DS Evidence Theory;Dependent Information Fusion