一类复杂变拓扑动态网络系统的故障诊断方法与应用研究

The dynamic network system with variable topologies is a type of complicated system related to many practical engineering systems. It features the dynamic multi-state nodes, uncertain changes occurring in both the topological structure of the network and the connectivity among nodes caused by the changing operation conditions of the network, and that the faults inside of the network may result in changes of both the dynamic behaviors of the nodes and the topological structure of the network. The emerging Microgrid system possessing such properties is an example of this complicated dynamic network system with variable topologies. Very a little of fault diagnosis techniques have been put forward for such type of system in the case of simple changes occurring in the topological structure in the way of adding or reducing nodes. Meanwhile little efforts have been made to develop fault diagnosis methods while uncertain changes will occur either in the dynamic behaviors of the nodes or in the topological structure and the connectivity of the network. In this project we will focus on theoretical research of fault diagnosis techniques for this type of complicated dynamic network system with variable topologies in order to propose the robust diagnostic method based on nonlinear filtering algorithm for the dynamic multi-state nodes, to set up the self-organizing diagnostic knowledge representation framework in which both the quantitative and the qualitative models are integrated, to realize identification of the topological structure and the connectivity of the network based on Multi Agent System, and to achieve diagnosis of the network faults according to on-line adaptive tuning the diagnostic template, therefore to create the novel fault diagnosis techniques for this type of complicated dynamic network system with variable topologies. Moreover, we will also make efforts to solve fault diagnosis of the Microgrid system by utilizing the proposed techniques in order to achieve better diagnostic performances.

变拓扑动态网络系统是一类具有实际应用背景的复杂系统，其动态节点具有多模态特性，其拓扑结构和节点间连接特性可随工况发生不确定性变化，故障可同时影响节点动态行为和网络拓扑结构。正在新兴发展的微电网系统就具有这些属性。这类系统的故障诊断研究目前仅在节点加入或脱离网络的拓扑简单变化情况下取得了若干研究成果，综合考虑节点动态行为、拓扑结构和连接特性不确定性变化情况的故障诊断方法研究还是一个空白。 本课题拟从理论方法上研究这类复杂变拓扑动态网络系统的故障诊断问题，建立多模态动态节点基于非线性滤波器的鲁棒故障诊断方法，引入自组织机制建立定量模型与定性模型相混合的诊断知识表示框架，应用多Agent技术实现网络拓扑结构和连接特性识别，基于诊断模板自适应在线修正实现网络故障诊断，从而形成这类复杂变拓扑动态网络系统故障诊断新方法。把研究结果应用于微电网系统，提高微电网故障诊断技术水平。

一、项目背景.本项目研究一类变拓扑动态网络系统的故障诊断方法，对象特点是：网络节点是多模态动态系统，网络拓扑随工况变化，且故障因素也可引起拓扑结构变化。拓扑结构变化不限于节点增减，还包括节点间连接特性的改变。微电网是这类变拓扑动态网络系统的代表，也是本课题主要应用研究领域。.近年来网络系统故障诊断技术已取得许多重要研究成果，微电网故障诊断也有许多成功案例，但现有成果还不足以完全解决这类系统故障诊断问题，尤其当涉及拓扑变化型故障时。因此，本项目有助于拓展被诊断对象研究范围，并形成新的故障诊断方法。.二、主要研究内容.1. 多模态动态节点的故障诊断。以微电网为原型，建立包含多种不同工况的动态节点模型，开发相应的故障诊断新方法。.2.网络系统拓扑结构变化型故障适应性诊断。建立变拓扑网络系统的诊断知识表示方法，研究基于MAS的网络拓扑识别和自适应故障诊断方法。.3. 微电网系统的故障诊断技术。建立微电网实验系统，开展故障诊断性能实验测试分析，开发故障诊断包。 .三、重要研究结果.1. 多模态动态节点的故障诊断。将特征提取与局部统计相融合，提出了一种新的多模态工况故障检测方法。提出了一种基于模糊Petri网的故障诊断方法。.2. 变拓扑故障的适应性诊断。重点研究节点和连接关系同时变化时的拓扑结构识别问题。基于自组织诊断知识表示，提出了基于多Agent系统的拓扑识别算法，并提出了一种MAS协调优化策略。提出了基于小波奇异熵和SOM神经网络的拓扑故障诊断方法。应用于微电网系统，实现了拓扑结构频繁变化下的拓扑结构识别。.3．微电网系统的故障诊断技术。设计开发了微电网实验系统（含风光互补发电），可模拟常见故障，测试故障诊断算法性能，研究微电网运行、控制和效能评估。.四、关键数据.经测试，基于MAS的拓扑故障诊断方法可在故障发生后40ms内完成诊断，基于小波奇异熵和SOM网络的诊断方法对各类故障的诊断准确率达到92%。