Supervised model predictive control of large‐scale electricity networks via clustering methods

This article describes a control approach for large-scale electricity networks, with the goal of efficiently coordinating distributed generators to balance unexpected load variations respect nominal forecasts. To mitigate difficulties due size problem, proposed methodology is divided in two steps. First, network partitioned into clusters, composed several dispatchable and nondispatchable generators, storage systems, loads. A clustering algorithm designed aim obtaining clusters following characteristics: (i) they must be compact, keeping distance between loads as small possible; (ii) able internally maximum possible extent. Once has been completed, layer system designed. At lower layer, local model predictive controller associated each cluster managing available generation elements compensate variations. If sources are not sufficient cluster's variations, power request sent supervisory which optimally distributes additional resources from other network. enhance scalability approach, supervisor implemented relying on fully optimization algorithm. The IEEE 118-bus used test design procedure nontrivial scenario.