Συντάχθηκε 03-06-2018 20:00
από Antonios Deligiannakis
Email συντάκτη: adeligiannakis<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: ΔΕΠ ΗΜΜΥ.
Ημερομηνία: Τρίτη 5/6/2018
Ώρα: 12:00 μ.μ.
Αίθουσα: 137.Π39 του Κτ. Επιστημών
Title: Towards the next generation of multiperiod optimal power flow solvers
Speaker: Δρόσος Κουρούνης (Drosos Kourounis)
Senior Researcher
Universita della Svizzera italiana
Institute of Computational Science, Lugano, Switzerland.
https://www.ics.usi.ch/index.php/people-detail-page/65-drosos-kourounis
Abstract:
Modern power grids will have to incorporate significant technological achievements towards their transition to the smart grid era. The processing of big data, the automated demand response, the installation of distributed energy storage devices, and the large scale integration of renewables are major engineering challenges requiring robust, efficient, and innovative software components. Single period optimal power flow (OPF) algorithms commonly employed for economic dispatch throughout a transmission network will have to adapt to incorporate multiperiodic OPF formulations, required for modeling energy storage devices. The latter however, introduces intertemporal couplings of the individual OPF problems defined at each subdivision of the time period of interest, and as a result the multiperiod optimal power flow (MPOPF) problem becomes intractable prohibiting, forecasting, and planning over long time periods. Interior point (IP) methods have been extensively employed for the solution of OPF and MPOPF problems defined over short time period. This talk proposes an efficient IP algorithm, BELTISTOS, particularly designed for MPOPF problems over a large number of time periods. The structure of the linear system associated with the Karush-Kuhn-Tucker conditions is revisited, and a Schur-complement-based approach tailored to its structure is proposed. Through benchmark cases involving power-grid models of increasing complexity, the BELTISTOS algorithm is demonstrated to provide several orders of magnitude faster solution times than standard optimization methods, such as IPOPT, MIPS, and KNITRO, using significantly less memory.
