基于多率采样数据和动态拓扑的网络化模糊系统分布式协同滤波

Distributed filtering systems over wireless sensor networks (WSN) are receiving increasing attention due to their powerful ability in the aspects of information collection and collaboration processing. However, some factors including limited energy in nodes, constrained network bandwidth and data sharing among multiple sensors make distributed filtering analysis challenging. Thus there has arisen some new problems such as different network-induced imperfections in different channels, dynamic topology, and multi-rate sampled-data fusion in WSN-based distributed filtering systems. Traditional filtering theory cannot meet the design requirements of these systems. Therefore, it is an urgent topic to study new models and theories to solve the design problems of distributed filtering for WSN-based networked systems. By introducing a new classification strategy of multi-rate sampling to determine a sampled-data fusion method, constructing an adaptive event-triggered transmission scheme to save the resources of WSN, and considering various network-induced imperfections in a unified framework, this project will investigate WSN-based distributed H∞ filtering for a class of nonlinear networked systems which can be described by T-S fuzzy models. More specifically, the following issues are investigated in this project: (1) the H∞ filtering of networked fuzzy systems with multi-variable multi-rate sampling and the event-triggered transmission scheme, (2) the distributed H∞ filtering of networked fuzzy systems with multi-group identical-sensor multi-rate sampling, (3) the distributed H∞ filtering of wireless sensor networked systems with multi-rate sampling and dynamic topology. For these problems, some new criteria are derived for system performance analysis, and some co-design methods of the distributed filter and the event-triggered conditions are proposed. Finally, the proposed distributed filtering methods will be further improved by studying the state estimation of security situation for an overhead transmission line in a WSN environment. This project will not only enrich theories in the field of distributed filtering systems, but also provide some effective design approaches for their practical applications.

无线传感网络化分布式滤波系统以突出的信息获取和协作处理能力受到广泛关注。节点能量和网络带宽受限以及多传感器数据共享使得分布式滤波研究仍面临诸多挑战，这会带来多通道互异网络诱导特征、动态拓扑和多率采样数据融合等新问题。传统滤波理论无法满足这些系统的设计要求，亟需研究新的模型和理论来解决其滤波器设计问题。本项目针对T-S模糊模型表示的非线性系统，引入多率采样分类策略来确定数据融合方式，构造分布式自适应事件触发传输机制，并考虑多种不理想信道特征共存，分别研究：(1)具有多传感器多率采样和事件触发传输机制的网络化系统H∞滤波；(2)具有多组相同传感器多率采样的网络化系统分布式H∞滤波；(3)具有多率采样和动态拓扑的无线传感网络化系统分布式H∞滤波。进一步给出满足某些系统性能的分布式滤波器和触发条件的协同设计方法。通过应用于架空输电线路安全状况估计来改进上述滤波方案，为其实际应用提供有效的理论支撑。

分布式滤波\控制系统以突出的信息获取和协作处理能力受到广泛关注。考虑节点能量和网络带宽受限、恶意网络攻击及多传感器数据共享等挑战，本项目分别研究多率采样的线性或非线性系统的网络化事件触发滤波\控制、分布式滤波及分布式一致性控制等问题。主要研究成果包括：(1)首次设计一种连续的多传感器多率采样数据融合机制，其刻画了多传感器的分组采样策略，及多率采样数据和滤波器采样状态的匹配机制。考虑到多率采样数据的多个更新模态对系统性能的影响，提出基于多率采样数据的网络化跳变模糊滤波器和事件触发传输条件的协同设计方法；(2)构建一种带有固定切换率的双率采样切换观测器来估计被控对象的状态，提出一种新颖的能补偿双侧网络诱导时延和丢包的网络化事件触发预测控制的协同设计方法。进一步引入云端模拟攻击机制来补偿观测器和预测控制器所用的控制输入不一致情形，给出带有多组相同传感器多率采样的复合网络攻击下多智能体系统的云端预测控制的设计方法；(3)针对带有两类采样机制的多智能体系统，构建仅依赖于采样位置数据的两类时变增益观测器，分别给出相应的分布式一致性控制器的设计方法。所提方法允许不同智能体异步采样，且采样周期可在允许的期间内独立配置。(4)针对基于两类典型多率采样的分层结构的异构多智能体系统，考虑到邻居间动态补偿器的异步采样数据融合问题，引入部分数据开环预测机制，建立基于时滞输入的动态补偿器设计方法来匹配不同智能体控制输入，进而给出异构多智能体系统的输出一致性控制的设计方法。(5)提出一种非线性系统的状态估计器设计的分层判据。证明所提的估计判据的保守性随着B-L不等式中N的增大而降低。(6)从电力系统典型设备安全状况监测以及分布式电力信息网络易受恶意攻击等热点问题角度出发，分别提出拒绝服务攻击下多区域互联电力系统负荷频率模糊控制、及电厂锅炉三维重建和安全监测等方法。这些成果将服务电力生产过程的管控需求。