“21st PSCC 2020 papers submission and review platform

Full Program »

Roundtable Discussion: Machine Learning for Power Systems: Present & Future

Monday, 29 June 2020
16:00 - 17:30

Chair: Spyros Chatzivasileiadis, Misha Chertkov, Pascal Van Hentenryck, Daniel Molzahn, Louis Wehenkel

Mediator: Pascal Van Hentenryck (Georgia Tech)

Review of the Field: Louis Wehenkel (U of Liege) – 15 min

Provocateurs: (7 min each)
Misha Chertkov (U of Arizona)
Eugene Litvinov (ISO NE)
Damien Ernst (U of Liege)
Jessica Harrison (MISO)
Spyros Chatzivasileiadis (DTU)
Antoine Marot (RTE)

30 min open discussion


Non-exhaustive list of subjects suggested by the organizers to provocateurs and to the audience:

1. How do you understand “machine learning for power systems”?
a) What type of questions can one answer with machine learning?
b) What types of machine learning methods can be useful in power systems?
c) What types of power system problems can one resolve with machine learning?

2. What are the (computational and other) needs for current and future power systems?
a) What are the strengths of machine learning methods that can be beneficial for power systems?
b) In what power systems applications can machine learning methods thrive?
c) What are the barriers (challenges) for machine learning methods in power systems?

3. When are ML methods inapplicable to power systems?
a) For which power system applications would it not make sense to use machine learning methods?
b) For which power systems applications would the use of machine learning make sense but encounter strong resistance? Why?
c) Which of those barriers can/should we overcome (if any, i.e. if it makes sense)? If yes, how?

4. Power systems are (i) safety-critical systems, and (ii) often well described by existing mathematical/computational models.
a) Does it make sense to apply machine learning methods to these systems in the first place?
b) How are power systems different from other fields of machine learning applications, e.g., computer vision, statistics and inference with big data?
c) What types of machine learning methods would be appropriate for power systems applications? (e.g. neural network verification? Physics-informed machine learning? What else?)

5. How can machine learning methods be used in energy markets?
a) What problems related to energy markets can potentially be addressed using machine learning methods?
b) What kind of machine learning methods have a potential in resolving these problems?

6. Where are data coming from?
a) What types of data are required? How do we assess and guarantee data quality? What is the value and appropriateness of observational versus simulation data? (How should simulated data be generated?)
b) When and how is it appropriate to share data? How can we improve and promote sharing? What are the barriers to making data open source?

7. How can machine learning be used as a tool for model reduction?
a) Does it make sense to use high-fidelity models to train reduced-order models based on machine learning?
b) What kind of applications can it apply to? Static? Dynamic?

8. Are there meaningful ways to apply machine learning in power systems while exploiting existing tools and knowledge? Alternatively, shall we strike everything and start from scratch?
a) What are the power systems problems where we should completely rethink how we address them/solve them? Can machine learning help provide these new solutions?
b) What are the power system problems/applications where machine learning can exploit existing knowledge (models/solution methods/insights) and revolutionize the way we address them? (e.g. either by offering better/faster solutions, or generating additional knowledge, or requiring less communication, or some kind of “model-free” approach(?), etc.)

9. Imagine a different world where we have available all the technology and knowledge we currently have with the complete freedom to design the systems for power supply, distribution, and demand, as we wish – i.e., in whatever way we think is optimal.
a) Would the use of machine learning methods make sense in such a system?
b) If so, for which applications? (this can be all the way from the design, to planning, operation, etc.)
c) Can you see some of these machine learning uses applicable in today’s or future power systems? If not, are there ways we can make them applicable?

10. From your experience, what are the three suggestions you would give to a researcher wishing to apply machine learning methods in power systems?
a) This assumes that you are in favor of applying machine learning in power systems.
b) If you are not, what are three reasons that researchers should avoid applying machine learning in power systems?

Biographies:

Dr. Pascal Van Hentenryck is the A. Russell Chandler III Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology and the Associate Chair for Innovation and Entrepreneurship. Prior to that, he was a professor of computer science at Brown University for about 20 years and led the Optimization Research Group at NICTA. Van Hentenryck is an INFORMS Fellow and a Fellow of the Association for the Advancement of Artificial Intelligence (AAAI), and the recipient of two honorary doctoral degrees. Several of his optimization systems, including the CHIP and OPL systems, have been in commercial use for more than 20 years. His current research focuses on machine learning, optimization, and privacy with applications in energy, mobility, and resilience.

Dr. Louis Wehenkel graduated in Electrical Engineering (Electronics) in 1986 and received the Ph.D. degree in 1990, both from the University of Liège (Belgium), where he is full Professor of Electrical Engineering and Computer Science. His research interests lie in the fields of stochastic methods for modeling, optimization, machine learning and data mining, with applications in complex systems, in particular large scale power systems planning, operation and control, industrial process control, bioinformatics and computer vision. Recently, he has been the Scientific Advisor of the GARPUR European FP7 project.

Antoine Marot is the lead AI scientist at RTE. He owns a double master degree in Engineering from Ecole Centrale Paris and Stanford University. After interning at Tesla Motors, he joined RTE R&D on the Apogee project 6 years ago with the long term goal to develop a personal assistant for control room operators with AI. Through collaboration with INRIA (the french AI research lab), he supervised several PHD students on augmented power system simulators with AI and on Human-Intelligent Machine interactions with a strong focus on interpretability. He recently co-authored several papers using AI for power systems and gave different talks on the topic such as IJCNN AI conference keynote. He advocates for a new "AI for power system community" bringing together researchers from both fields to accelerate the application of AI. The « Learning to Run a Power Network « challenge which will run along NeurIPS 2020, the largest AI conference, is a strong step forward towards it.​

Dr. Damien Ernst received the M.Sc. and Ph.D. degrees in engineering from the University of Liège, Belgium, in 1998 and 2003, respectively. He is currently Full Professor at the University of Liège, where he is affiliated with the Montefiore Research Unit. His research interests include electrical energy systems and reinforcement learning, a subfield of artificial intelligence. He is also the cofounder of Blacklight Analytics, a company developing intelligent software solutions for the energy sector. He has co-authored more than 300 research papers and two books. He has also won numerous awards for his research and, among which, the prestigious 2018 Blondel Medal. He is also regularly consulted by industries, governments, international agencies and the media for its deep understanding of the energy transition.

Dr. Eugene Litvinov is a chief technologist at the ISO New England. He is responsible for advanced System and Markets solutions, Smart Grid and Technology strategy and is a lead of the Research and Development activities in the organization. Dr. Litvinov has over 40 years of professional experience in the area of Power System modeling, analysis and operation; Electricity Markets design, implementation and operation; information technology. He has extensive expertise in management and technical leadership of large engineering and information technology projects; development of computational methods in power system analysis and operation; market clearing engines, electricity pricing; settlements. Dr. Litvinov holds BS, MS and PhD in Electrical Engineering. He is an IEEE Fellow and a member of the National Academy of Engineering.

Jessica Harrison is the Senior Director of Research and Development (R&D) in the Market and Grid Strategy division of MISO. In this role, she coordinates MISO's R&D policies, objectives and initiatives. Her focus includes pursuing R&D initiatives and partnerships, researching emerging technologies and supporting the integration of future grid resources. Ms. Harrison holds dual Master of Science degrees from the Massachusetts Institute of Technology in Technology and Policy and in Civil and Environmental Engineering, and a Bachelor of Science degree in Physics from the University of Michigan.

Dr. Spyros Chatzivasileiadis is an Associate Professor at the Technical University of Denmark (DTU). Before that he was a postdoctoral researcher at the Massachusetts Institute of Technology (MIT), USA and at Lawrence Berkeley National Laboratory, USA. Spyros holds a PhD from ETH Zurich, Switzerland (2013) and a Diploma in Electrical and Computer Engineering from the National Technical University of Athens (NTUA), Greece (2007). In March 2016 he joined the Center for Electric Power and Energy at DTU. He is currently working on power system optimization and control of AC and HVDC grids, and machine learning applications for power systems.

Dr. Michael (Misha) Chertkov is a Professor of Mathematics and Chair of the Graduate Interdisciplinary Program in Applied Mathematics at the University of Arizona. Dr. Chertkov area of focus is mathematics, including statistics and data science, applied to physical, engineered and other systems. Dr. Chertkov received his Ph.D. in physics from the Weizmann Institute of Science in 1996, spent three years at Princeton University as a R.H. Dicke Fellow in the Department of Physics, and joined Los Alamos NL in 1999, initially as a J.R. Oppenheimer Fellow and then as a Technical Staff Member. During his 20 years at LANL he led multiple LDRD/DR, DTRA and DOE/EERE projects, in particular on “physics of algorithms”, “optimization, inference and learning of energy systems” and “machine learning for turbulence”. Dr. Chertkov has moved to Tucson in 2019. He has published more than 200 papers, is a fellow of the American Physical Society and a senior member of IEEE.

 

Powered by OpenConf®
Copyright ©2002-2014 Zakon Group LLC