Collaborative Research: High-Dimensional Spatio-Temporal Data Science for a Resilient Power Grid: Towards Real-Time Integration of Synchrophasor Data

合作研究：弹性电网的高维时空数据科学：同步相量数据的实时集成

• 批准号: 1934675
• 负责人: Le Xie
• 金额: $ 18.6万
• 依托单位: Texas A&M Engineering Experiment Station
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1934675&HistoricalAwards=false
• 关键词: Collaborative; Research; Dimensional; Spatio; Temporal

The project will establish an Institute at Arizona State University (ASU) with Texas A&M (TAMU) that considers the electric power grid and examines critical real-time decision-making by developing core data-driven science methods and applications. This is motivated by the modern electric power system which is experiencing heightened unpredictability from increasing demand for renewable energy, efficiency, and resilience. To address this, industry stakeholders are deploying GPS-synchronized phasor measurement units (PMUs), or synchrophasors, that provide direct measurements of voltage and current phasors with high temporal granularity. However, the potential real-time situational awareness enabled by these measurements has been impeded by the massive scale of the time-series PMU data and have limited its use to passive, post-event forensics. The Institute meets this need for PMU-based real-time decision-making by examining five critical problems: (i) ensure data quality against bad, missing, or stale data; (ii) exploit the fine granularity of PMU data to track real-time changes in network parameters; (iii) detect, identify, localize, and visualize oscillation and failure events; (iv) assess and visualize cybersecurity threats and countermeasures specific to PMUs; and (v) create synthetic PMU datasets for testing and validation. The Institute leverages the PIs' synergistic multidisciplinary background in information sciences and statistics, machine learning, data visualization, cybersecurity, and power systems. The team will apply state-of-the-art techniques including hidden Markov models, LSTM neural networks, graphical models, errors-in-variables models, graph signal processing, adversarial examples, low-dimensional feature extraction, and constrained GANs. Another key research focus is the development of visual analytics for high-granularity spatio-temporal PMU data to enable improved operator review and decision-making. These innovations will be fueled by massive PMU datasets accessible to the PIs.This Phase I institute has the potential to tip PMUs from a promising-but-mostly-underused resource into an essential part of power system best practices. The data science outcomes will impact application domains such as transportation networks, smart buildings, and manufacturing, each of which increasingly faces high-dimensional streaming data challenges. The PIs will disseminate their research to both academic and industry stakeholders and will continue their outreach on teaching AI and machine learning (ML) modules to underrepresented high school students. Finally, the multi-disciplinary strength of this institute lends itself naturally to a larger, integrated, and comprehensive Phase II institute focused on data-intensive research for critical infrastructure networks.This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity. This effort is co-funded by the Division of Electrical, Communications and Cyber Systems within the Directorate for Engineering.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.

该项目将在亚利桑那州立大学（ASU）与德克萨斯A&amp；M （TAMU）建立一个研究所，该研究所将考虑电网，并通过开发核心数据驱动的科学方法和应用来检查关键的实时决策。这是由于对可再生能源、效率和弹性的需求不断增加，现代电力系统的不可预测性正在增强。为了解决这个问题，行业利益相关者正在部署gps同步相量测量单元（pmu），或同步相量，提供高时间粒度的电压和电流相量的直接测量。然而，通过这些测量实现的潜在实时态势感知受到了大量时间序列PMU数据的阻碍，并限制了其用于被动的事后取证。该研究所通过审查五个关键问题来满足这种基于管理单元的实时决策的需求：(i)确保数据质量，防止数据损坏、丢失或过时；（ii）利用PMU数据的细粒度来跟踪网络参数的实时变化；（iii）检测、识别、定位和可视化振荡和故障事件；（iv）评估和可视化针对pmu的网络安全威胁和对策；(v)创建用于测试和验证的综合PMU数据集。该研究所利用pi在信息科学和统计学、机器学习、数据可视化、网络安全和电力系统方面的协同多学科背景。该团队将应用最先进的技术，包括隐马尔可夫模型、LSTM神经网络、图形模型、变量误差模型、图形信号处理、对抗示例、低维特征提取和约束gan。另一个关键的研究重点是开发高粒度时空PMU数据的可视化分析，以改进操作员的审查和决策。这些创新将由pi访问的大量PMU数据集推动。这个第一阶段的研究所有可能将pmu从一个有前途但大多未充分利用的资源转变为电力系统最佳实践的重要组成部分。数据科学的成果将影响交通网络、智能建筑和制造业等应用领域，每个领域都日益面临高维流数据的挑战。pi将向学术和行业利益相关者传播他们的研究成果，并将继续向代表性不足的高中生教授人工智能和机器学习（ML）模块。最后，该研究所的多学科优势自然使其成为一个更大、综合、全面的第二阶段研究所，专注于关键基础设施网络的数据密集型研究。该项目是美国国家科学基金会“利用数据革命（HDR）大创意”活动的一部分。这项工作是由工程局的电气、通信和网络系统司共同资助的。该奖项反映了美国国家科学基金会的法定使命，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。

项目成果

Generative Adversarial Networks-Based Synthetic PMU Data Creation for Improved Event Classification

基于生成对抗网络的综合 PMU 数据创建以改进事件分类

• DOI: 10.1109/oajpe.2021.3061648
• 发表时间: 2021
• 期刊: IEEE Open Access Journal of Power and Energy
• 影响因子: 3.8
• 作者: Zheng, Xiangtian;Wang, Bin;Kalathil, Dileep;Xie, Le
• 通讯作者: Xie, Le

Synthetic Dynamic PMU Data Generation: A Generative Adversarial Network Approach

• DOI: 10.1109/sgsma.2019.8784681
• 发表时间: 2018-12
• 期刊: 2019 International Conference on Smart Grid Synchronized Measurements and Analytics (SGSMA)
• 作者: Xiangtian Zheng;Bin Wang-;Le Xie