A Revised Decomposition Method for MILP Problems and Its Application to Operational Planning of Thermal Storage Systems

Abstract

A revised decomposition method for solving large-scale mixed-integer linear programming (MILP) problems with block angular structure is presented to efficiently conduct the operational planning of thermal storage systems. The fundamental algorithm adopted here is composed of solving large-scale linear programming (LP) master problems by the Dantzig-Wolfe decomposition method and small-scale MILP subproblems by the branch and bound method, and these problems are solved repeatedly until an optimality or suboptimality criterion is satisfied. As one of the revision strategies to improve computation efficiency, a two-phase approach is introduced, by which a next LP master problem can be solved efficiently by utilizing the results of a previous one. An illustrative example on a heat supply system for district heating and cooling is given to show the effectiveness of the above revision strategy. A practical example on a heat supply system with multiple thermal storage tanks for brewing is also presented.