An Adjoint-Based Algorithm for Mixed-Integer Dynamic Optimization

摘要

This paper presents an adjoint-based algorithm for the solution of mixed-integer dynamic optimization problems. The algorithm is based on Generalized Benders' Decomposition, where the solution alternates between a series of primal dynamic optimization problems (which give upper bounds on the solution) and master, mixed-integer linear programming problems (which give lower bounds). In contrast to earlier adjoint - based works [1,2], the algorithm is independent of the type of method used for integrating the differential-algebraic equation system in the primal problems and allows easier construction of the master problems. The application of the algorithm is illustrated with a process example taken from the literature.