New approach for power network clustering using a representative model

A new power network model that is suitable for network clustering techniques is presented. The method is based on representing the system buses as modules and the connecting branches as nets. The number of nets connecting each pair of modules is chosen to be a function of either a single electrical property or a normalised combination of several properties of the equivalent branch connecting the pair of buses. A heuristic approach, based on the approximation of the constrained network partitioning problem into a linear transportation problem, is used to cluster the network into S clusters each containing m buses. The use of a linear-time bus interchange algorithm provides the best possible cluster configuration. The solution technique provides complete freedom in choosing the electrical properties of the greatest importance to the clustering application. Consequently, the selectivity of the proposed solution technique in the aspect of choosing the electrical connectivity measure is evident