BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.bham.ac.uk//v3//EN
BEGIN:VEVENT
CATEGORIES:Optimisation and Numerical Analysis Seminars
SUMMARY:Polynomial optimisation in power systems at IBM Re
search - Jakub Marecek (IBM Research\, Dublin\, Ir
eland)
DTSTART:20180125T120000Z
DTEND:20180125T130000Z
UID:TALK3008AT
URL:/talk/index/3008
DESCRIPTION:In power systems\, problems that model alternating
-current transmission constraints are of a great a
nd growing importance\, but notoriously difficult
due to their non-convexity. We have shown\n-- a va
riety of hierarchies of convexifications\, whose o
ptima converge to the global optimum of the non-co
nvex problem \n-- custom first- and second-order m
ethods for solving the convexifications\, which ar
e competitive with leading heuristics\n-- methods
for switching from solving the convexification usi
ng the first-order methods (with trivial per-itera
tion time and memory requirements\, but poor rates
of convergence)\nto any second-order methods on t
he non-convex problem (with local quadratic conver
gence) based on Smale's work in algebraic geometry
. \nThis allows one to tackle large-scale instance
s in practice and to guarantee global convergence
in theory. This summarises recent papers in IEEE T
. Power Systems [31(1): 539–546]\, IEEE T. Smart G
rid [8(6): 2988-2998]\, and Optimization Methods a
nd Software [32(4): 849-871]\, which are joint wor
k with Bissan Ghaddar\, Alan Liddell\, Jie Liu\, M
artin Mevissen\, and Martin Takac.\n
LOCATION:Lecture Theatre 2\, Strathcona Building
CONTACT:Sergey Sergeev
END:VEVENT
END:VCALENDAR