A new model for stochastic optimization of weekly generation schedules

Because of the uncertainty on the demand forecast in France, the optimization of weekly generation schedules is actually a stochastic optimization problem. To deal with this kind of problem, a decomposition strategy has been proposed by P. Carpentier et al. (see IEEE Trans. Power Syst., vol.11-2, p.1067-73, 1996). This decomposition method, which may be considered as a generalization of a Lagrangian relaxation technique to a stochastic framework, has been chosen by EDF to develop a new model, MOSAR. This paper is devoted to the presentation of this model. After a brief outline of the decomposition strategy, the modeling of the thermal units, of the hydro-valleys, of the exchanges and of the demand side management are described. The algorithms applied to solve the subproblems and the coordination method are then presented. Finally, numerical tests over the whole French generation mix are presented and discussed