Acceleration of Benders decomposition for mixed integer linear programming

This paper presents a novel strategy for speeding-up the classical Benders decomposition for large-scale mixed integer linear programming problems. This method is particularly useful for the cases where the optimality cut is difficult to obtain. The distances between the selected feasible points and feasibility cutting planes, as a metric, determine the tighter constraint, thus improving the convergence rate. The application of this approach in a scheduling problem for multi-product, multi-purpose batch plants show substantial improvement both in the computational time and the number of Benders iterative steps.