Sponsor: IEEE PES FOOTHILL SECTION
Speaker: Professor Bragin from University of Connecticut

Meeting Date: May 26, 2021
Time: 5PM
Cost:
Reservations: IEEE

Summary:
This talk is on the efficient resolution of operation optimization problems arising in power systems. One of the major difficulties of these problems is the presence of discrete decision variables, such as unit commitment decisions within unit commitment problems. Because of the many possibilities in which units can be committed, the computational complexity involved in the search of optimal solutions increases exponentially with the increase of problem size. However, operation optimization problems such as unit commitment need to be solved accurately fast. Historically, the “price-based” decomposition and coordination Lagrangian Relaxation method was used following the “divide-and-conquer” principle by relaxing “coupling constraints,” decomposing relaxed problems into subproblems, and coordinating subproblem solution based on Lagrangian multipliers (“shadow prices”) following the “supply and demand” principle. The traditional version of the Lagrangian Relaxation method, however, suffers from major convergence difficulties such as zigzagging of Lagrangian multipliers leading to slow convergence. To resolve these difficulties, I will present my recently developed “Surrogate Lagrangian Relaxation” (SLR) method with major advancements to efficiently solve unit commitment problems. Within the method, efficient price-based coordination of units within the unit commitment problem is enabled by efficiently obtaining proper directions to update Lagrangian multipliers without having to optimally solve associated subproblems as long as the high-level “surrogate optimality” convergence condition is satisfied. As a result, the computational effort is much reduced and multiplier zigzagging is much alleviated. Also, unlike the standard Lagrangian relaxation method, the convergence of the SLR method does not require the knowledge of the generally unavailable “optimal dual value” for convergence proof as well as for practical implementations. The latest developments include further acceleration of convergence through piece-wise linear penalty terms, and asynchronous update of multipliers to coordinate distributed subproblems. The efficiency of the approach is also demonstrated for TSO-DSO coordination and for coordination of microgrids. The approach opens up new directions for the efficient solution of MIP problems to obtain near-optimal solutions with quantifiable quality in a computationally efficient manner.

Bio: Mikhail A. Bragin (Member: IEEE, PES, INFORMS) received the B.S. and M.S. degrees in Mathematics from Voronezh State University, Voronezh, Russia, in 2004, the M.S. degree in Physics and Astronomy from the University of Nebraska-Lincoln, Lincoln, NE, USA, in 2006, and the M.S. and Ph.D. degrees in Electrical and Computer Engineering from the University of Connecticut, Storrs, CT, USA, in 2014 and 2016, respectively. He is currently an Assistant Research Professor in Electrical and Computer Engineering at the University of Connecticut. His research interests include operations research, mathematical optimization, including power system optimization, grid integration of renewables (wind and solar), energy-based operation optimization of distributed energy systems, scheduling within manufacturing systems, and machine learning through deep neural networks.