Tutorial on PMU and Time Series Data Analysis
Date: Monday, May 24 Time: –
Add to calendar: Tutorial3.ics
Abstract: This tutorial is designed to train researchers and practitioners on data analysis methods and applications for synchrophasor and point on wave data. The course is divided into three sections, covering (1) fundamentals of synchrophasor measurement, (2) data analysis tools and techniques, and (3) practical applications in industry. The course will help attendees put concepts from power engineering and data science into practice to establish efficient workflows for analyzing and visualizing time series data at scale.
Part 1: Fundamentals synchrophasor measurement (15 min)
Alexandra von Meier, PhD
The first part of the course describes foundational concepts for phasor measurement units (PMUs) and synchrophasors. This portion will cover concepts from physics and power engineering to provide intuition about how PMU measurements relate to physical phenomena on the grid. We’ll offer a quick refresher of relevant concepts for attendees familiar with power engineering and PMU data, and provide a brief introduction for attendees who are new to the field.
Part 2: Data analysis and machine learning workflows (60 min)
Mohini Bariya and Miles Rusch
The second part of the course will demonstrate analytics performed on synchronized measurement data from distribution systems. The focus is on showing methods that participants can reproduce on open access data, as well as applying and adapting algorithms to their specific purposes and needs. We discuss methods for unsupervised event detection and classification with time series clustering of voltage magnitude and frequency; these algorithms prioritize transparency and communication to the human user. Panellists will show how they use the PredictiveGrid platform in their work. PredictiveGrid (detailed in Part 4) is a state-of-the-art data analytics platform designed to increase the efficiency of data analysis workflows including data exploration, algorithm development, and data visualization.
Part 3: PMU applications in industry (60 min)
Chetan Mishra, Phd and Kevin Jones, Phd
The third part of the course will cover case studies from a recent initiative at Dominion Energy to uncover dynamics observed under ambient conditions using synchrophasor data. Traditional methods for integrating new resources onto the grid rely on generic models which fail to account for the reality of the dynamics that can occur in operation. Utilities often lack transparent models for characterizing the dynamic behaviors of certain devices — including solar inverters, FACTS devices, DERs, and industrial loads. This motivated an initiative to use continuous monitoring data to improve visibility and inform advancements in grid integration and controller design. Findings reveal that devices are often deployed with poorly set controllers which may become unstable. Panellists will describe the use of signal processing techniques to alert the utility to possible issues occurring under ambient conditions which would typically be brushed off as “noise”. The tutorial will cover the challenges faced when working with actual measurements as opposed to simulation data where often, the existing analysis techniques need to be adapted due to violation of underlying assumptions.
Part 4: Putting it to practice with open access data (15 min)
Laurel Dunn, Phd
NI4AI is a 3-year ARPA-E funded project led by PingThings’ which provides open access to a cloud-based data visualization and analysis platform to support rapid data exploration and application development. This talk will give an overview of open access PMU and point on wave data sets which attendees can use to advance their work. The session will provide an overview of resources available to attendees via the project, and will describe how attendees can leverage these resources to streamline their workflows and build partnerships to advance practical applications for time series data.
Name and title of the organizer: Laurel Dunn, PhD
Organisation: UC Berkeley, USA
Short biography of the chair: Laurel recently received her PhD in Civil Engineering from UC Berkeley on risk-aware decision making. She has worked closely with industry and with regulators about putting state-of-the-art methods from statistics and data science into practice. Currently, she is working to advance the use of next generation sensors that can remove guess-work from decision making processes. She is working with a startup called PingThings to lead an initiative called NI4AI. NI4AI is an ARPA-E funded project which provides data, computational tools, and educational content to advance the use of time series data for grid applications.
Alexandra “Sascha” von Meier (University of California, Berkeley) is an Adjunct Professor in the Department of Electrical Engineering and Computer Science, where she teaches a course on Electric Power Systems. She is also Director in CIEE’s Electric Grid program area focusing on power distribution systems, Smart Grid issues, and the integration of distributed and intermittent generation. Her current research projects center on the use of high-precision micro-synchrophasor measurements for situational awareness, diagnostics and control applications in distribution grids.
Kevin Jones (Dominion Energy) received his Ph.D. in Electrical Engineering from Virginia Tech as a Harry Lynde Bradley Fellow in 2013. He created and helped commercialize the open source linear state estimator and has led the development and integration of a variety of innovative technology solutions in his role at Dominion Energy including the cloud deployment of PredictiveGrid for synchrophasor analytics in Electric Transmission. He currently serves as Manager of ET Operations Engineering Support leading the Fault Analysis, Sensor Data Communication/Engineering/Analytics, Special Studies, and Web Development teams.
Chetan Mishra (Dominion Energy) is an electric transmission engineer with research interests in nonlinear dynamics and control, synchrophasors and renewable energy. He earned his PhD in Electrical Engineering from Virginia Tech in 2017 and has been with Dominion for over five years.
Mohini Bariya (University of California, Berkeley) is a PhD candidate focusing on the use of novel, high-resolution measurements for improved situational awareness in the electric grid. She has worked extensively with real PMU datasets. She has experience teaching science and engineering concepts to different audiences, including as a graduate student instructor for the Electric Power Systems course at Berkeley.
Miles Rusch (University of California, Berkeley) is a PhD student in Electrical Engineering at UC Berkeley. His research involves using unsupervised learning algorithms to perform data analysis on electric power systems.