We show conditions for the convergence of a sequence of relaxed stochastic optimization problems • We show examples satisfying theses conditions • We are motivated by a lagrangian decomposition algorithm • We give consistency result of the approximation scheme