A forward maximum principle algorithm with decision horizon results

摘要

We present an algorithm with convergence results for an optimal control problem with a multidimensional control vector and a one dimensional constrained state vector. The algorithm detects and exploits decision horizon properties. For the case of a nonconvex problem we give optimality conditions for a perturbed problem. The algorithm has been applied to a case of optimal scheduling over a week of production of heat for district heating from a cogeneration unit in connection with a heat storage.