A methodology for computing robust dynamic equivalents of large power systems

•A simple voltage sensitivity based 'coherency' classification and a medoid clustering method is presented.•It is shown that sufficient accuracy can be achieved with tuned second order synchronous machine model for clustered generators.•Issues related to network reduction are revisited and the related adopted methods are justified.•The concept of buffer zone, used for network equivalents is extended to dynamic equivalents. Sensitivities indices to identifying such zone are presented.•Overall, the paper presents a complete and relatively simple methodology for deriving dynamic equivalents for large power system models.

This paper presents a simple and computationally efficient methodology for computing robust dynamic equivalents. While simple to implement, the methodology involves many factors, and these are discussed, explained and justified throughout the paper. It classifies clusters of generators to be equivalenced by their electrical distances, and identifies second order reduced models for them. It maps them into classical synchronous machine models and aggregates them. Reduction of the network is based on the Ward PV method. The quality of the dynamic equivalent, in terms of flexibility and reduced dependence on disturbances, is automatically verified by performance indices. The methodology can easily be embedded in time domain simulation packages. Two large real power system networks are used for testing the performance of the approach.