We describe a method for solving convex optimization problems which usually appear in Lagrangian decomposition approaches. The suggested cutting plane method is capable to handle infinite objective function values and demands an oracle to be given. Some questions of realizing the oracle are discussed.
The convergence proof of the level method uses a restrictive condition. We show the essentiality of this condition by example. Moreover, we present a strategy to adjust the problems such that they meet all requirements of the level method. This strategy involves perturbation of the original problem. Finally, we present numerical results, which demonstrate the effectivity of the perturbation approach.