[en] In this Phd, we focus on the problem of weekly risk management in electric production. In the first part of this work, we investigate how to take into account stochastic inflows in the optimal management of a hydraulic valley. Our model is based on robust optimization and linear decision rules. A validation procedure based on simulation over random scenarios shows that we are able to postpone constraints violations of volume at very low cost. The second part deals with the problem of active management of electrical power margin, defined as the difference between the total offer and the total demand, considering the different random parameters which affect the electrical system. The objective is to determine optimal solutions to be taken in order to satisfy the demand in 99% of the cases. In that purpose, we propose a new open-looped formulation, based on the stochastic process of power margin and on the use of probabilistic constraints. To be able to solve this problem, we generate power margin's scenarios using more realistic methods than those used in exploitation. At last, a closed-loop approach, based on the heuristic 'Stochastic Programming with Step Decision Rules', introduced by Thenie and Vial, is studied. First results are quite promising in comparison with the opened-loop ones. (author)