Mixed-integer vs. real-valued formulations of battery scheduling problems

Publikationstyp
Conference Paper
Publikationsdatum
2018
Sprache
English
Author
Murray, Alexander
Faulwasser, Timm 
Hagenmeyer, Veit 
TORE-URI
https://hdl.handle.net/11420/46288
Enthalten in
IFAC-PapersOnLine 
Volume
51
Issue
28
Start Page
350
End Page
355
Citation
IFAC-PapersOnLine 51 (28): 350-355 (2018)
Publisher DOI
10.1016/j.ifacol.2018.11.727
Scopus ID
2-s2.0-85058212832
Publisher
IFAC
Coordinating the charging and discharging of a population of batteries with respect to load and generation forecasts decreases operating costs. However, in the case of asymmetric charging and discharging efficiencies the mathematical formulation of this problem is not straightforward. This paper compares the scheduling problem for a population of batteries in two formulations: mixed-integer (MINLP) and non-convex real-valued (NLP). In the case of the latter, the resulting separable NLP is solved using the Augmented Lagrangian based Alternating Direction Inexact Newton (ALADIN). In the case of the latter, the resulting MINLP is solved in hierarchical distributed fashion, we propose a mixed-integer extension of ALADIN. Real load and generation data is used to create an example of 54 batteries demonstrating improved performance for the MINLP formulation.
Schlagworte
ALADIN
Battery Scheduling
Distributed Optimization
MINLP
Power Systems
DDC Class
333.7: Natural Resources, Energy and Environment
004: Computer Sciences
510: Mathematics