Support vector machines for an efficient representation of voltage band constraints

Future Smart Grids emphasize the (at least) partial coordination of a large number of stochastic consumers and producers to balance consumption and generation of electrical energy. However, this coordinated and often time-synchronous activation and deactivation of demand and supply may result in the violation of the grid's feasibility constraints, since it may increase statistical simultaneity above the level assumed in the initial design of such networks. The detection and avoidance of such operational infeasibilities is essential for secure grid operations. In order to handle the complexity of the optimization problem at hand, autonomous software agents will play a vital role in future power management systems. The system's complexity is reduced down to several less complex sub-problems, which may then be solved in parallel, based on locally available information. In this paper we present ongoing work on the development of autonomous grid management under stability constraints imposed by the power grid.