Several transient stability software packages are widely used for power grid planning and operations. Prior research and software documentations have shown that packages can vary in the implementations of dynamic models, and hence could potentially yield different results for the same system simulation. This paper presents a systematic methodology to determine the sources of the discrepancies seen in results obtained from different transient stability packages. This methodology can be applied to various types of dynamic models and is illustrated with generator-associated dynamic models in this paper. Consistent metrics, data mining techniques, and engineering heuristics are used to determine the discrepancy sources. This hybrid top-down+bottom-up method has been implemented on a practical, large-system case to demonstrate its scalability and transferability to real-world system studies.Index Terms-Commercial software packages, data mining, dynamic models, power system transient stability, simulations.
0885-8950