The future smart energy grid demands for new control paradigms that are able to incorporate a huge number of rather small, distributed and individually configured energy resources. In order to allow for a transition of the current central market and network structure to a decentralized smart grid, with small units pooling together to jointly trade their electricity production on specialized markets, self-organization concepts will become indispensable as an efficient management approach. In order to enable ahead of time planning of electricity that incorporates global objectives and individually constrained distributed search spaces in such highly dynamic environment, mcta-models of constrained spaces of operable schedules are indispensable for efficient communication and uniform access. An essential prerequisite for building-up machine learning based domain models of individually constrained search spaces is a training set of operable example schedules. Drawing such a sample from an electricity unit's simulation model is a challenging task due to the high dimensionality of the problem. We present two computationally feasible sampling methods and analyze their complexity and appropriateness. Moreover, the embedding of these methods and the interplay of sampling and simulation in a multi agent simulation is presented.
展开▼
