Learning-Enabled Modeling, Monitoring, and Decision Making for Distribution Grids

该NSF项目旨在通过开发一个整合电网边缘可再生资源和灵活资源的整体框架，推动电网基础设施的零碳排放过渡。该项目将为这些电网边缘资源的实时监测和协调带来变革性的变化，以支持其连接的配电网的效率和安全性。这将通过将机器学习的进步综合到算法开发中来实现，这些开发可以识别底层系统的主导物理，并解决配电网网络基础设施的限制。该项目的智力优势包括一套支持机器学习的解决方案，能够在由于模型知识有限和可观测性较低的信息约束下实现网格边缘资源的高效和安全运行。该项目的更广泛影响包括加速将可再生能源和低碳资源整合到电力基础设施中，以及一项全面的教育计划，该计划包括更新电力工程课程和为大学预科学生设计动手演示。这项提议的总体目标是建立一个学习框架，以高效、自适应和健壮的方式运营分布式能源资源(DER)。为了解决配电网中有限的传感和通信的现状，我们主张结合底层馈线模型和数据配置文件的独特功能。我们的研究包括三个方面：T1)在部分可观测性下设计数据驱动的分布式建模方法；T2)开发来自不同数据源的网格边缘资源监控算法；以及T3)使用基于图和风险感知的学习来开发可扩展的、安全的DER策略。这三项任务将进一步整合，以相互支持，形成一个整体框架，并得到现实世界馈线系统和数据集的验证。简而言之，我们的研究议程将实现双重目标，即通过充分利用多种数据源来实现分销系统的运营，同时实现及时和安全的行动，以应对信息有限和资源有限的情况。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This NSF project aims to propel the zero-carbon emission transition of the electric grid infrastructure by develop a holistic framework for integrating renewable and flexible resources at grid edge. The project will bring transformative changes to the real-time monitoring and coordination of these grid-edge resources in support of the efficiency and safety of their connected distribution grids. This will be achieved by synthesizing machine learning advances into the algorithmic developments that can recognize the governing physics of the underlying systems and address the limitations in cyber infrastructure in distribution grids. The intellectual merits of the project include a suite of machine learning enabled solutions to attain an efficient and safe operation of grid-edge resources under the information constraints due to limited model knowledge and low observability. The broader impacts of the project include the acceleration of integrating renewable energy and low-carbon resources into the electricity infrastructure, and a comprehensive education plan consisting of updating power engineering curriculum and designing hands-on demos for pre-college students. The overarching goal of this proposal is to establish a learning-enabled framework for operating distributed energy resources (DERs) with efficiency, adaptivity, and robustness. To address the status quo of limited sensing and communications in power distribution grids, we advocate to incorporate the unique features of the underlying feeder models and data profiles. Our proposed research consists of three cohesive thrusts: T1) Designing data-driven distribution modeling approaches under partial observability; T2) Developing monitoring algorithms of grid-edge resources from heterogeneous data sources; and T3) Developing scalable and safe DER policies using graph-based and risk-aware learning. These three tasks will be further integrated to support each other into a holistic framework as validated by real-world feeder systems and datasets. In a nutshell, our research agenda will fulfill the dual objectives of enabling distribution system operations by fully embracing a multitude of data sources, while attaining timely and safe DER actions to address the information-limited and resource-constrained scenarios.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.