The current upheaval in the electricity sector demands distributed generation
schemes that take into account individually configured energy units and new
grid structures. At the same time, this change is heading for a paradigm
shift in controlling these energy resources within the grid. Pro-active
scheduling of active power within a (from a controlling perspective) loosely
coupled group of distributed energy resources demands for distributed
planning and optimization methods that take into account the individual
feasible region in local search spaces modeled by surrogate models. We
propose a method that uses support vector based black-box models for
re-constructing feasible regions for automated, local solution repair during
scheduling and combine it with a distributed greedy approach for finding an
appropriate partition of a desired target schedule into operable schedules
for each participating energy unit.