Using the coherency function in measurement based small-signal analysis of large power systems

Load switching and other random phenomena in a loosely connected power system produce ambient process noise that contains much useful information about oscillatory dynamics. The estimated coherency between two signals, in particular, has long been taken as a heuristic indicator for the frequency and strength of modal interactions along specific network paths. The conjecture is that a sharp peak in estimated coherency corresponds to a mode that is common to both signals. Such information is useful in mode shape estimation, and as a guide to signal selection for multi-output Prony analysis. These results follow from known bias properties of the estimator, and require that the signals contain substantial levels of uncorrelated process noise in the frequency band of interest. As yet, this "orthogonal" noise has neither been modeled nor well understood. Observations suggest that it is a basic property of large power systems, and that it tends to increase with geographical separation of the signal sources. Recent growth of wide area measurement facilities in the Western North America power system have encouraged system tests that permit such phenomena to be examined more closely. This presentation provides a summary of recent results.