(Some) Directions for Impactful Research at the Intersection of Data Science and the Power Grid

Wednesday, April 27, 2016
7:00 AM to 8:00 AM
MEZ 1.306
Open to ECE graduate students

It is arguable that the term "big data" is not applicable to power grid control, operations, and planning - particularly in comparison to remote sensing and cyber applications. However, power grid applications should still be of significant interest to the data science community, due to the potential and largely differentiating linkages between diverse spatio-temporal data sources (e.g., forecasted and realized renewables output) and control algorithms used to operate power grids worldwide. In particular, the outputs of data science analyses can both directly impact existing power grid operational paradigms and inform the design of novel control architectures and investment plans, ultimately improving cost-effectiveness, reliability, and resiliency.

In this talk, I will introduce various exemplars of existing, on-going research projects that lie firmly at the data-science/power-grid intersection. These projects range from being loosely to tightly coupled with power grid control and planning paradigms, and range from (physics) "model-free" approaches to more complex applications that directly couple data science outputs with control models and algorithms. Typically, such projects are staffed almost exclusively by engineers and operations researchers - including the speaker, who have leveraged and developed a range of heterogeneous and often ad-hoc approaches to dealing with the requisite data analyses. Direct engagement with the data science community has the potential to dramatically improve the fidelity and rigor of data modeling tasks (e.g., projecting hurricane impacts) and the effectiveness of control architectures (e.g., operations models based on probabilistic forecasts of renewables generation). Further, I argue that major impediments to deploying advanced control technologies in the power grid often center around data science questions - for which we often have poor or inadequate answers. 

x x


Dr. Jean-Paul Watson

Distinguished Member of Technical Staff
Discrete Math and Optimization Department Center for Computing Research, Sandia National Laboratories

Dr. Jean-Paul Watson is a Distinguished Member of Technical Staff in the Discrete Math and Optimization Department at Sandia National Laboratories, in Albuquerque, New Mexico. He has over 12 years of experience applying and analyzing algorithms for solving difficult combinatorial optimization and informatics problems, in fields ranging from logistics and infrastructure security to power systems and computational chemistry. His research currently focuses on methods for approximating the solution of large-scale deterministic and stochastic mixed-integer and non-linear programs, with applications in the domain of electricity grid operations, planning, and resiliency. Previously, he developed solutions for real-world stochastic optimization problems in logistics (Lockheed Martin and the US Army) and sensor placement (US Environmental Protection Agency). Additionally, he led the development of programs involving the use of semantic graph technologies for performing geospatial imagery analysis. He is a co-developer of Sandia's Pyomo open-source software package for modeling and solving optimization problems, and has published over 40 journal articles in the areas of optimization algorithms and their application.