<!DOCTYPE html><html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<p>-------- Forwarded Message --------</p>
<div class="moz-forward-container">
<table class="moz-email-headers-table" cellspacing="0" cellpadding="0" border="0">
<tbody>
<tr>
<th valign="BASELINE" nowrap="nowrap" align="RIGHT">Subject:
</th>
<td>[DMANET] Post-doc position (INRIA Edge, Inocs and EDF
R&D)</td>
</tr>
<tr>
<th valign="BASELINE" nowrap="nowrap" align="RIGHT">Date: </th>
<td>Tue, 19 Sep 2023 10:05:44 +0000</td>
</tr>
<tr>
<th valign="BASELINE" nowrap="nowrap" align="RIGHT">From: </th>
<td>ROTTNER Cecile via dmanet
<a class="moz-txt-link-rfc2396E" href="mailto:dmanet@zpr.uni-koeln.de"><dmanet@zpr.uni-koeln.de></a></td>
</tr>
<tr>
<th valign="BASELINE" nowrap="nowrap" align="RIGHT">Reply-To:
</th>
<td>ROTTNER Cecile <a class="moz-txt-link-rfc2396E" href="mailto:cecile.rottner@edf.fr"><cecile.rottner@edf.fr></a></td>
</tr>
<tr>
<th valign="BASELINE" nowrap="nowrap" align="RIGHT">To: </th>
<td><a class="moz-txt-link-abbreviated" href="mailto:dmanet@zpr.uni-koeln.de">dmanet@zpr.uni-koeln.de</a> <a class="moz-txt-link-rfc2396E" href="mailto:dmanet@zpr.uni-koeln.de"><dmanet@zpr.uni-koeln.de></a></td>
</tr>
</tbody>
</table>
<br>
<br>
Dear colleagues,<br>
<br>
Please find below a full-time post-doc position offer on
decomposition algorithms for solving generation expansion planning
problems.<br>
The post-doc will be co-supervised by Inria Edge & Inocs teams
(Bordeaux, France) and EDF R&D (Palaiseau, France).<br>
<br>
Starting date: From 2023/10 to 2024/02<br>
<br>
Duration: 24 months<br>
<br>
Supervision:<br>
- Inria: Ayse Arslan, Boris Detienne and Aurélien Froger (Edge),
Bernard Fortz (Inocs), Frédéric Semet (Inocs)<br>
- EDF: Cécile Rottner and Rodolphe Griset (R&D/OSIRIS)<br>
<br>
Contact:
<a class="moz-txt-link-abbreviated" href="mailto:ayse-nur.arslan@inria.fr">ayse-nur.arslan@inria.fr</a><a class="moz-txt-link-rfc2396E" href="mailto:ayse-nur.arslan@inria.fr"><mailto:ayse-nur.arslan@inria.fr></a>,
<a class="moz-txt-link-abbreviated" href="mailto:boris.detienne@inria.fr">boris.detienne@inria.fr</a><a class="moz-txt-link-rfc2396E" href="mailto:boris.detienne@inria.fr"><mailto:boris.detienne@inria.fr></a>,
<a class="moz-txt-link-abbreviated" href="mailto:aurelien.froger@inria.fr">aurelien.froger@inria.fr</a><a class="moz-txt-link-rfc2396E" href="mailto:aurelien.froger@inria.fr"><mailto:aurelien.froger@inria.fr></a><br>
<br>
Scientific context: Multiple problems studied by the ”long-term
planning” team in the OSIRIS department<br>
of EDF R&D correspond to making investment decisions over a
multi-year planning horizon while taking<br>
into account economic, physical or legal restrictions and grid
balancing.<br>
By their nature, these problems can be modeled as multi-stage
stochastic optimization problems in the<br>
sense that it is possible to adapt the investment decisions to the
realization of uncertainty over the course<br>
of the planning horizon. The requirement to balance the
electricity grid results in Unit Commitment (UC)<br>
problems appearing as subproblems. The presence of numerous
discrete decisions is a notable characteristic<br>
of these problems.<br>
</div>
<div class="moz-forward-container"><br>
</div>
<div class="moz-forward-container">The complexity and the scale of
these problems lead EDF to resort to simplifying assumptions in
their<br>
solution (for instance, multi-stage problems are approximated by
two-stage ones or mixed-integer linear<br>
problems are approximated by linear problems). Currently, a tool
named TiLT (”Long-Term Investment<br>
Trajectories”) is used to optimize investment decisions with a
multi-year vision where ”trajectories” are<br>
used to formulate very large-scale linear models (with millions of
variables and constraints). A heuristic<br>
”Investment Loop” is used to fix the investment decisions based on
a ”Unit Commitment” (UC) solver<br>
called Continental. The team has started working on the solution
of the two-stage (meaning that all investment<br>
decisions are taken before the realization of uncertainty)
approximation of these problems exploiting<br>
the fact that when investment decisions are fixed the constraints
ensuring the electricity grid balance can<br>
be decomposed into independent subproblems by trajectory.
Geographical and temporal decoupling is<br>
also used to reduce the size of sub-problems. Decomposition
methods - notably Benders or Lagrangian<br>
decomposition - can be used to exploit this type of structure. In
a Benders decomposition approach, the<br>
master problem consists of fixing the investment decisions
(related to the electricity generation plants)<br>
and subproblems reduce to UC problems.<br>
</div>
<div class="moz-forward-container"><br>
</div>
<div class="moz-forward-container">Solving the targeted investment
problems with finer approximations is an important problematic for<br>
obtaining more judicious investment decisions. Taking into account
the discrete nature of certain decisions<br>
(for instance switch-on costs of plants, ramping constraints,
minimum on/off times, minimum power etc.)<br>
and the stochastic and multi-stage nature of these problems
represent technical barriers. In order to<br>
overcome this challenge it is necessary to develop advanced
algorithmic techniques based on reformulation<br>
and decomposition algorithms. In order to be efficient, these
techniques will also have to undergo a number<br>
of improvements.<br>
<br>
Work description: The first goal of this study will be to gather
the state-of-the-art of the latest algorithmic<br>
developments for solving two- and multi-stage stochastic problems
with continuous and integer recourse.<br>
For two-stage problems, we will be interested in recent advances
in methods based on Benders decomposition<br>
(Rahmaniani et al., 2017), especially stabilization, strategies
aiming to balance the information<br>
between the master and subproblems, batching of subproblems
(Blanchot et al., 2023), and cut strengthening<br>
(Chen and Luedtke, 2022). We will also study algorithms based on
Lagrangian dual decomposition<br>
such as progressive hedging or bundle methods for which the
application of strategies like branch-andbound<br>
allow to handle the integer recourse case (Atakan and Sen, 2018;
Kim and Dandurand, 2022). For<br>
multi-stage problems, we will be interested in stochastic dual
dynamic integer programming (SDDiP) (Zou<br>
et al., 2019) which has been applied to investment problems (Lara
et al., 2020) sharing similarities with<br>
the problem studied by EDF.<br>
</div>
<div class="moz-forward-container"><br>
</div>
<div class="moz-forward-container">Using publicly available
open-source implementations of the aforementioned techniques (for
instance (Kim<br>
and Zavala, 2018)) and implementations that will be developed by
the postdoctoral researcher, a numerical<br>
study will be conducted in order to quantify the loss of
optimality due to approximations (multi/two-stage<br>
and continuous/integer recourse) on smaller-scale problems having
the same structure as the application<br>
at hand. This part of the study will aim to identify the
limitations of the existing approaches as well as<br>
determine the most relevant approximation of the target problem.<br>
We will then work on trying to solve the scientific roadblocks
determined in the first phase for the most<br>
relevant version of the problem. Improving existing methodologies
based on the special structure of the<br>
problem at hand, or developing heuristic version of existing
approaches are possible avenues that might<br>
be explored.<br>
<br>
Profile: PhD in Operations research or Mathematical optimization.<br>
<br>
Required skills:<br>
- Mixed-integer linear programming<br>
- Significant experience with programming languages (C++
preferred)<br>
<br>
Skills/knowledge that would be appreciated:<br>
- Decomposition methods in mixed-integer linear programming
(Benders, Dantzig-Wolfe, Lagrangian)<br>
- Stochastic programming, robust optimization<br>
- Dynamic programming<br>
- Modeling or solving industrial problems<br>
<br>
Salary: ~2,250€/month (net)<br>
<br>
Environment: The work will primarily be conducted in the Inria
project-team Edge, in the building of the<br>
Mathematics Institute of Bordeaux located at Talence.<br>
There will also be a number of trips to EDF offices (Massy): one
to two weeks at the start of the position<br>
and approximately 2 days per month. There may also be one or two
trips to Lille. Transport and<br>
accommodation costs for these trips are covered.<br>
<br>
References<br>
<br>
Atakan, S. and Sen, S. (2018). A progressive hedging based
branch-and-bound algorithm for mixed-integer<br>
stochastic programs. Computational Management Science,
15(3-4):501–540.<br>
Blanchot, X., Clautiaux, F., Detienne, B., Froger, A., and Ruiz,
M. (2023). The benders by batch algorithm: design<br>
and stabilization of an enhanced algorithm to solve multicut
benders reformulation of two-stage stochastic<br>
programs. European Journal of Operational Research,
309(1):202–2016.<br>
Chen, R. and Luedtke, J. (2022). On generating lagrangian cuts for
two-stage stochastic integer programs.<br>
INFORMS Journal on Computing, 34(4):2332–2349.<br>
Kim, K. and Dandurand, B. (2022). Scalable branching on dual
decomposition of stochastic mixed-integer<br>
programming problems. Mathematical Programming Computation,
14(1):1–41.<br>
Kim, K. and Zavala, V. M. (2018). Algorithmic innovations and
software for the dual decomposition method<br>
applied to stochastic mixed-integer programs. Mathematical
Programming Computation, 10(2):225–266.<br>
Lara, C. L., Siirola, J. D., and Grossmann, I. E. (2020). Electric
power infrastructure planning under uncertainty:<br>
stochastic dual dynamic integer programming (sddip) and
parallelization scheme. Optimization and<br>
Engineering, 21(4):1243–1281.<br>
Rahmaniani, R., Crainic, T. G., Gendreau, M., and Rei, W. (2017).
The benders decomposition algorithm: A<br>
literature review. European Journal of Operational Research,
259(3):801–817.<br>
Zou, J., Ahmed, S., and Sun, X. A. (2019). Stochastic dual dynamic
integer programming. Mathematical<br>
Programming, 175:461–502.<br>
<br>
<br>
<br>
[<a class="moz-txt-link-freetext" href="cid:image001.png@01D9EAF0.93CE6D10">cid:image001.png@01D9EAF0.93CE6D10</a>]<br>
Cécile Rottner<br>
Ingénieure-chercheure<br>
EDF R&D département OSIRIS<br>
Groupe R36 « Méthodes, modèles et outils d’optimisation »<br>
7 boulevard Gaspard Monge, 91 120 Palaiseau, France<br>
<a class="moz-txt-link-abbreviated" href="mailto:cecile.rottner@edf.fr">cecile.rottner@edf.fr</a><a class="moz-txt-link-rfc2396E" href="mailto:cecile.rottner@edf.fr"><mailto:cecile.rottner@edf.fr></a><br>
<br>
<br>
<br>
<br>
<br>
Ce message et toutes les pièces jointes (ci-après le 'Message')
sont établis à l'intention exclusive des destinataires et les
informations qui y figurent sont strictement confidentielles.
Toute utilisation de ce Message non conforme à sa destination,
toute diffusion ou toute publication totale ou partielle, est
interdite sauf autorisation expresse.<br>
<br>
Si vous n'êtes pas le destinataire de ce Message, il vous est
interdit de le copier, de le faire suivre, de le divulguer ou d'en
utiliser tout ou partie. Si vous avez reçu ce Message par erreur,
merci de le supprimer de votre système, ainsi que toutes ses
copies, et de n'en garder aucune trace sur quelque support que ce
soit. Nous vous remercions également d'en avertir immédiatement
l'expéditeur par retour du message.<br>
<br>
Il est impossible de garantir que les communications par
messagerie électronique arrivent en temps utile, sont sécurisées
ou dénuées de toute erreur ou virus.<br>
____________________________________________________<br>
<br>
This message and any attachments (the 'Message') are intended
solely for the addressees. The information contained in this
Message is confidential. Any use of information contained in this
Message not in accord with its purpose, any dissemination or
disclosure, either whole or partial, is prohibited except formal
approval.<br>
<br>
If you are not the addressee, you may not copy, forward, disclose
or use any part of it. If you have received this message in error,
please delete it and all copies from your system and notify the
sender immediately by return message.<br>
<br>
E-mail communication cannot be guaranteed to be timely secure,
error or virus-free.<br>
<br>
</div>
<div class="moz-signature">-- <br>
Join <a href="https://www.informatics-europe.org/ecss/home.html">ECSS
2023</a> in Edinburgh, 23-25 October!
<a href="https://www.informatics-europe.org/ecss/registration/how-to-register.html">Registration</a>
open!
<br>
<br>
<a href="https://www.informatics-europe.org/ecss/home.html">
<img src="https://users.ics.aalto.fi/orponen/images/ie/ECSS2023_RegisterNow.png">
</a></div>
</body>
</html>