Inherent to the sinusoidal nonlinearities dictating AC power flow is the fact that grid stability margins tend to decrease as the transmission system is more fully utilized. Motivated by these stability concerns, among the recommendations to emerge following the 2003 northeastern U.S. blackout was a call for improved
measurement technology, to enhance "situational awareness" for the grid. Major investment in Smart Grid technologies has followed, with large numbers of dramatically improved sensors now widely deployed across North America. Among the opportunities today is that of using this grid "big data" to improve real time stability monitoring.
This work will argue that for big data algorithms to be successfully applied to problems in power grid stability, one must exploit the structural features giving rise to the data, while seeking to avoid dependence on detailed parameter values to populate system models. Of interest here will be the problem of characterizing vulnerability to voltage instability, a phenomena that has contributed to a number of major blackouts and near miss events. We will examine a singular value decomposition (SVD) based algorithm, well established in the context of "full model based" assessment of voltage instability, and demonstrate its “model free” adaptation to use only real-time measurement data, without need for explicit identification of system model parameters.