The technique presented in this talk is the "feasible path approach", which discretizes the controls using suitable piecewise continuous functions over a number of moving finite elements in time. According to this the model is reduced to a standard nonlinear optimization problem.

Theoretical issues examined are the development of a general formulation for a class of multistage optimal control problems, implementation of control profiles, efficient modelling of state (path) constraints, and use of sensitivity equations for derivation of gradients used in the optimization procedure. Finally, some issues regarding the possibility of the introduction of additional nonconvexities by the discretization method are examined. A number of case studies are used to highlight the key features of the proposed methodology.