A Graph Signal Processing Framework for Situational Awareness in Smart Grids

智能电网作为社区的重要基础设施，其可靠性和安全性至关重要。网络或物理压力，甚至更糟的是，智能电网中传输网络的联合网络和物理压力可能会产生广泛的破坏性影响，例如大停电。用于监测和分析系统的网络和物理状态的态势感知是智能电网中的一项基本功能，最终可以从意外事件中缓解和恢复。该项目将研究和开发新的方法，以通过图形信号处理（GSP）框架增强智能输电网的态势感知，适用于分析结构化能源数据和系统组件之间相互作用的动态数据。该项目的成果预计将在分析智能电网数据方面绘制出一个新的视角和技术范式，并有可能应用于其他联网系统和关键基础设施。该项目还将对教育产生广泛的影响。也就是说，综合教育计划包括通过课程项目向学生介绍能源数据分析主题，以及促进研究经验，特别是对代表性不足的学生。该项目的研究部分有两个连贯的重点。在第一个推力，图形频谱分析技术，滤波器设计，系统频率响应事件，和图形采样技术将用于网络压力检测，定位和状态估计的压力。机器学习方法还将用于学习各种GSP域中的应力签名，包括顶点，图形频率和联合顶点频率域，以及信号特性，包括图形信号平滑度，以改进此类技术。除了网络压力，对物理压力的态势感知也很关键，但由于与物理压力相关的独特属性而具有挑战性。例如，由于物理应力引起的能量信号振荡不是完全局部化的，并且由于电的物理性质而可能在一定距离处发生。 此外，包括故障的某些物理事件可以改变底层物理拓扑，并且因此改变图形信号的频率基础。因此，该项目的第二个重点将集中在提高身体压力的情景意识，通过解决这些挑战的检测和定位技术的身体压力在一个基于GSP的框架。此外，还将研究不确定性和缺失信息对分析物理压力的作用，这将有助于评估某些网络和物理联合攻击对系统的影响。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The reliability and security of smart grids, as critical infrastructures for communities, are of great importance. A cyber or physical stress, or even worse, a joint cyber and physical stress on transmission networks in smart grids can have widespread and devastating effects such as large blackouts. Situational awareness for monitoring and analyzing the cyber and physical states of the system is an essential function in smart grids that can ultimately enable mitigation and recovery from unexpected events. This project will investigate and develop new methodologies to enhance situational awareness in smart transmission grids through a Graph Signal Processing (GSP) framework, suitable for analyzing structured energy data and data on dynamics of interactions among system components. The outcomes of this project are expected to map out a new perspective and technical paradigm in terms of analyzing data for smart grids with the potential to be applied to other networked systems and critical infrastructures. This project will also have substantial broader impacts on education. Namely, the integrated education plan includes introducing energy data analytics topics to students through course projects as well as promoting research experiences, especially for underrepresented students.The research component of this project has two cohesive thrusts. In the first thrust, graph spectral analysis techniques, filter design, system frequency response to events, and graph sampling techniques will be used for cyber stress detection, localization and state estimation under stresses. Machine learning methods will also be used to learn the signatures of stresses in various GSP domains, including vertex, graph-frequency, and joint vertex-frequency domains, and in signal properties, including graph signal smoothness, to improve such techniques. In addition to cyber stresses, situational awareness towards physical stresses is also critical but challenging due to the unique properties associated with physical stresses. For instance, the energy signal oscillations due to physical stresses are not fully localized and can occur at a distance due to the physics of electricity. Moreover, certain physical events including failures can change the underlying physical topology, and consequently the frequency bases of the graph signals. Hence, the second thrust of this project will focus on improving situational awareness of physical stresses by addressing such challenges in the detection and localization techniques for physical stresses in a GSP-based framework. The role of uncertainties and missing information on analyzing physical stresses will also be investigated, which will enable evaluation of the effects of certain joint cyber and physical attacks on the system.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.