Typical Patterns, Atypical Events, and Uncertainty in Complex Systems

The power grid is a complex system. Multiple quantities are measured from hundreds of locations, at rates up to 30 Hz. There are both correlated and uncorrelated variables. Powerful methods are needed to examine this large amount of data and better understand the complex system, and in the case of the power grid, identify imminent adverse events, such as blackouts, by uncovering the hidden “gems” within the data. These gems include identification of the uncertainty, characterization of the typical patterns, and the discovery of atypical events.

The Pacific Northwest National Laboratory (PNNL) has been conducting R&D for a project known as the “Power Grid Monitoring and Alerting System.” This project leveraged the methodologies of the R&D 100 award-winning “Morning Report: Advanced Proactive Safety and System Monitoring” technology for NASA by applying this new data-intensive analysis approach to the electrical power grid.

The Morning Report, as applied to aircraft safety data, uses statistical and mathematical methods to present the expert with a set of flights that are atypical and consequently much more likely to have noteworthy flight data to study. The tool identifies problems that the expert may have envisioned as possible problems and problems that the expert may never have envisioned. This can help the expert think outside the box and find more obscure but important insights that have been hidden in the data and thus facilitate improved aviation safety.

In this talk, we will present an application of the same types of analyses to newly available real-time data collected via phasor measurement units (PMUs) from multiple locations throughout the United States and parts of Canada. We will discuss the intricate methods used to explore the data, and the novel displays used to communicate the findings. We will also delve into the exploration of other complex systems using similar methods

Speaker Bio: Brett Amidan has been a Research Scientist with the Pacific Northwest National Laboratory since 1997, and has worked in multivariate analyses of large data sets for more than ten years. He has applied various multivariate techniques to help identify and display patterns and atypicalities within diverse domains, including the power grid, aircraft safety, climate, and finance. He has won R&D 100 and FLC awards for multivariate analyses performed on aircraft safety data. This work also resulted in the 2006 Outstanding Statistical Application Award from the American Statistical Association. His financial analyses resulted in a package called the Anomalator which is currently being licensed by V-Indicator. He has also been involved in projects using experimental design, statistical modeling, and cost analysis. Brett earned his B.S. and M.S. degrees both in statistics at the Brigham Young University.

Presentation Slides: PPT

Note: Visitors from outside NIST must contact Robin Bickel; (301) 975-3668; at least 24 hours in advance.