Collaborative Research: Factor-Graph Approach to Monitoring and Failure Assessment in Smart-Grid Networks

该提案的动机是需要引入概率概念来感知和管理不断变化的电力系统。许多新的分布式资源的本质，包括响应性需求，是高度可变的，很难预测。此外，系统可能容易受到代表低概率高影响事件的网络安全威胁。这挑战了今天的基本假设？S的操作和计划基本上是确定的。特别是需要一个整体的随机公式来说明传感和通信问题是正常情况下和故障情况下供需调度的一个组成部分。故障可能是由系统部分的强制中断或故意攻击引起的。威胁系统完整性和平衡供需能力的事件的概率通常取决于系统条件，而不是一次性确定的。为了开始填补这一空白，本项目建议将不断变化的电网视为一个具有许多高度可变的分布式资源的电网，特别是微型太阳能和微型风力发电厂。许多分布式负载也在变化，必须进行监测；此外，它们也对感知和传达的信息作出反应。该方案的关键思想是将未来网格表示为因子图。一旦这样做了，就有可能利用正式的通信理论方法来计算系统处于某些感兴趣状态的部分的概率。这反过来又为经济实惠的传感器和通信架构设计创造了基础，以支持新的操作实践。将概率推理引入供需平衡是关键，因为如果没有这种对系统状态的主动跟踪，确保可靠服务的最坏情况设计方法就会变得低效得令人无法接受，同时，也无法提供有关最坏情况服务场景可能性的信息。本项目提出的方法将基于感知和通信信息的概率估计结合在一起。然后使用这些估计来引入不同分布资源的自我调度，而系统操作员的协调最少，正如其中一个共同项目负责人已经提出的那样，在正常情况下平衡供需。在这个项目上与通信和安全合作项目负责人合作，将进一步推广自我调度概念，以概率地解释设备故障。这些故障可能是设备被迫中断，也可能是电网受到网络攻击引起的。目标是通过启用因子图的关于系统状态严重性的概率估计，提供概率自调度的概念证明说明。传感器和通信对整体性能的影响将以夏威夷瓦胡岛的网格为实验实例加以说明。智力上的优点：这个建议最基本的智力上的新颖之处在于将概率推理渗透到电网监测和控制领域。该项目将应用因子图和安全信念传播形式，为电网的分布式监测、评估、安全防护、控制和风险管理铺平道路。其目标是对夏威夷瓦胡岛部分电网的概率自我调度进行概念验证。更广泛的影响：该项目的研究活动将导致有效利用分散的可再生能源的方法。这将有助于加快向清洁/可再生能源经济转型的步伐。在夏威夷土著科学与工程指导计划（NHSEMP）的保护下，联合pi将继续致力于通过本科生研究经验项目，整合代表性不足的群体在工程领域的参与。

The proposal is motivated by the need to introduce probabilistic concepts for sensing and managing changing electric energy systems. The very nature of many new distributed resources, including responsive demand, is highly variable and hard to predict. Moreover, the system may be prone to cyber-security threats which represent low-probability high-impact events. This challenges the fundamental assumptions underlying today?s operations and planning which are by and large deterministic. In particular, a holistic stochastic formulation is needed to state the problem of sensing and communications as an integral part of supply/demand dispatch during normal conditions as well as during failures. Failures could be caused by either forced outages or intended attacks on portions of the system. The probabilities of events threatening the integrity of the system and the ability to balance supply and demand are generally dependent on system conditions and are not determined once for good. In order to begin to fill this gap, it is proposed in this project to view the changing electric power grid as an electric network with many highly variable distributed resources, micro-solar and micro-wind plants, in particular. Very many distributed loads are also varying and must be monitored; moreover, they are also responsive to the sensed and communicated information. The key idea in this proposal is to represent such future grid as a factor graph. Once this is done, it becomes possible to draw on formal communications theory methods to compute the probabilities of portions of the system being in certain states of interest. This, in turn, creates the basis for an affordable sensors and communications architecture design in support of novel operating practices. It is key to bring probabilistic reasoning into supply and demand balancing because without such proactive tracking of system state the worst-case design approach to ensuring reliable services becomes unacceptably inefficient, and, at the same time, does not provide information about the likelihood of the worst-case service scenario. The approach proposed in this project brings together the probability estimates based on the sensed and communicated information. These estimates are then used to introduce a self-dispatch by different distributed resources with minimal coordination by the system operators, as already proposed by one of the co-PIs for balancing supply and demand during normal conditions. In collaboration with the communications and security co-PIs on this project, the self-dispatch concept will be further generalized to probabilistically account for equipment failures. These failures could be either forced equipment outages, or initiated by cyber attacks on the power grid. The objective is to provide proof-of-concept illustrations of probabilistic self-dispatch facilitated by the factor-graph-enabled probability estimates about the severity of system state. The effects of sensors and communications on the overall performance will be illustrated using the grid of the Hawaiian island of Oahu as the experimental example.Intellectual merit: The fundamental intellectual novelty in this proposal is the penetration of probabilistic reasoning into the field of power-grid monitoring and control. The project will apply factor-graph and secure-belief-propagation formalisms to pave the way towards distributed monitoring, assessment, safe-guarding, control and risk management in power grids ? the goal being a proof-of-concept illustration of probabilistic self-dispatch on a portion of the power grid of the Hawaiian island of Oahu.Broader impact: The research activities in this project will lead to methods for efficient utilization of scattered renewable energy resources. This will contribute to the accelerated pace of transformation towards a clean/renewable energy economy. The co-PIs will continue to be engaged in integrating the participation of under-represented groups in engineering through research experience programs for undergraduates under the umbrella of the Native Hawaiian Science and Engineering Mentorship Program (NHSEMP).