Projected Alternating Direction Method of Multipliers for Hybrid Systems

Abstract : In large-scale control optimization problems, a decentralized control structure can offer great scalability and rapidity advantages over a centralized implantation. Alternating Direction Method of Multipliers (ADMM) is a decentralized optimization algorithm which has the important benefit of being quite general in its scope and applicability in continuous systems. In this paper, a new projected ADMM algorithm is defined and it can work in hybrid systems. The key point is to add a convexification and a projection process during each iteration of ADMM algorithm. We have applied it to a charging control problem of electric vehicles. Simulation results show that the proposed algorithm can converge to a similar result as a centralized control within a limited iteration time. Due to its availability, simplicity and scalability, the projected ADMM algorithm may be attractive in some practical engineering application.
Document type :
Conference papers
Complete list of metadatas

https://hal-centralesupelec.archives-ouvertes.fr/hal-01589791
Contributor : Romain Bourdais <>
Submitted on : Tuesday, September 19, 2017 - 9:52:42 AM
Last modification on : Friday, December 14, 2018 - 1:07:27 AM

Identifiers

  • HAL Id : hal-01589791, version 1

Citation

Chengang Feng, Romain Bourdais. Projected Alternating Direction Method of Multipliers for Hybrid Systems. IFAC World Congress, Jul 2017, Toulouse, France. pp.8753-8758. ⟨hal-01589791⟩

Share

Metrics

Record views

120