Nos tutelles

CNRS

Nos partenaires


Accueil > Projets

PGMO


PGMO Project 2013-2017


Coordinator : Prof Philippe Mahey


Main lab : LIMOS, Université Clermont Auvergne, UMR 6158

Collaborating teams and staff
LIMOS : P. Mahey, J. Koko
EDF : A. Lenoir, K. Barty
LJLL (Paris VI) : P. Combettes
U. Sherbrooke, Canada : J.P. Dussault, L. Marchand (Doc)
IMCA Lima, Peru : E. Ocaña, E. Oré Albornoz (Doc)
U. Edinburgh, UK : K. Mac Kinnon, M. Lémery (M.Sc.)

Project Spec
In [1] a mathematical model for a stochastic multizonal production planning problem is given with a potential application to a long-term simulation of the electricity European market. The huge size of the model requires a decomposition between the zones and a decomposition over time of the multi-stage model. We analyze the coupling of zonal decomposition based on some monotone operator splitting techniques [3] and the use of Dynamic Programming in each local subproblem. Technically, this is computationally tractable if some information relaxation is introduced in the model [6]. We propose to compare the decomposed and relaxed solution with the application of Stochastic Dual Dynamic Programming [2] on the scenario tree. Numerical comparisons with other existing Dynamic Programming methods are expected.

Recommended reading


  1. A stochastic zonal decomposition algorithm for network energy management problems, A. Dallagi and A. Lenoir, 2013 EDF Research Report
  2. Multi-stage stochastic optimization applied to energy planning, M.V. Pereira and L.M.V. Pinto, Math. Programming 52, pp. 359-375, 1991 Paper
  3. Survey on operator splitting methods, A. Lenoir and P. Mahey, RAIRO OR 51, 2017
  4. Splitting methods for a dynamic spatial model for long-term energy pricing problem, P. Mahey, J. Koko, A. Lenoir, Research Report LIMOS 2015
  5. A unified splitting algorithm for composite monotone inclusions, E. Oré Albornoz, P. Mahey, Research Report LIMOS 2016
  6. Decomposition of Large-scale Stochastic Optimal Control Problems, K. Barty, P. Carpentier and P. Girardeau Paper
  7. Asynchronous block-iterative primal-dual decomposition methods for composite monotone inclusions, P.L. Combettes and J. Eckstein, Math. Programming 2017
  8. Solving composite monotone inclusions in reflexive Banach spaces by constructing best Bregman approximations from their Kuhn-Tucker set, P.L. Combettes and Nguyen V.H., J. Convex Analysis 2016
  9. Data for EDF’s model Report
  10. Coupling decomposition with dynamic programming for a stochastic spatial model for long-term energy pricing problem, P. Mahey, J. Koko ; A. Lenoir, L. Marchand, J.P. Dussault, Research Report LIMOS

Slides


  1. Dual Approximate Dynamic Programming for Large Scale Hydro Valleys, Pierre Carpentier and Jean-Philippe Chancelier Slides
  2. Core 50th Aniversary Meeting. 4th talk : Recent Applictation of multistage Stochastic Optimization to Power System Planning and Operations, Mario Pereira. Video link
  3. CIRRELT Montréal : séminaire Modèle stochastique Mai 2014 (P. Mahey)
  4. Optimization & Big Data, Edinburgh : Séminaire Méthodes de ‘Splitting’ Mai 2015 (P.L. Combettes)
  5. EUROPT Edinburgh, Congrès EURO Juillet 2015 : Décomposition du modèle dynamique déterministe (P. Mahey)
  6. Journées PGMO, ENSTA Octobre 2015 : Eclatement d’opérateurs pour inclusions composites (E. Chouzenoux)
  7. ICSP Conference on Stochastic Programming, Buzios, Brésil, Juillet 2016 : Information relaxation and spatial decomposition (P. Mahey)
  8. Séminaire SPOT, Toulouse, décembre 2016 : Generalized splitting (P. Mahey)

Poster
- Poster PGMO Oct 14