Speeding-up Benders
decomposition method
Abstract: Benders decomposition technique (Benders,
1962) is based on the idea of exploiting decomposable
structure present in the formulation of a given problem so that its solution can
be converted into the solution of several smaller sub-problems. One is the slave problem which is obtained by fixing a number of decision variables
of the initial problem to a feasible value and the second one is the
restricted master problem which is expected to provide the optimal solution
after the addition of a number of cuts. The cuts are
deduced from the resolution of the slave problem in each iteration
of the algorithm. In each iteration, cuts are appended to restricted master problem
which is solved again to optimality. Benders method is an algorithm that has been applied successfully to a variety of applications
and different fields of mathematical programming as for example stochastic
programming, global optimization, hierarchical optimization etc. In some cases the straightforward application of the classical
Benders algorithm does not lead to fast convergence. Over the years various techniques have been proposed to speed-up the
classical Benders decomposition algorithm. The work presented in the literature
has mainly focused on either reducing the number of iterations of the algorithm
or on restricting the solution space of the decomposed problems.
This work presents two new strategies for the
speeding-up of the Benders algorithm which are applied
to two case studies in order to evaluate their efficiency. These strategies
referred to as: covering cut bundle (CCB) generation
[saharidis et al. 2009] and maximum feasible
sub-system (MFS) cut generation [Saharidis and Ierapetritou 2009]. Both strategies implement in a novel way
the multiple constraints generation idea. The CCB and MFS generation are applied
to mixed integer problems arising from two types of applications the scheduling
of crude oil [Saharidis, 2006] and the scheduling
problem for multi-product, multi-purpose batch plants [Ierapetritou and Floudas, 1998].
In both cases CCB and MFS significantly decrease the
number of iterations of the Benders method leading to improved resolution
times.
Benders, J.F. Partitioning Procedures for Solving
Mixed-Variables Programming Problems. Numerische Mathematic. 1962, 4, p. 238-252.
Saharidis,
G.K.; M.Minoux; Ierapetritou, M.G. Accelerating Benders decomposition using
covering cut bundle generation. Accepted in International Transactions in Operational
Research. 2009.
Saharidis
G.K.; Ierapetritou, M.G. "Improving Benders
decomposition using Maximum Feasible Sub-system (MFS) Cut Generation Strategy"
Industrial Engineering and Chemistry Research 2009
Saharidis,
G.K. Pilotage de production a moyen terme et a court terme: contribution aux
problematique d'optimisation
globale vc locale et a l'ordonnancement dans les raffineries. Genie Industriel. Vol. PhD. 2006, Paris:
Ecole Centrale Paris.
150.
Ierapetritou, M.G.; Floudas, C. Effective Continuous-time Formulation for
Short-Term Scheduling: I Multipurpose Batch Processes Industrial &
Engineering Chemical Research. 1998
37(11), p. 4341-4359.