Simultaneously Structured Models with Application to Analyze Large Amount of Power System Synchrophasor Data

Abstract

The recent deployment of phasor measurement units (PMUs) allows synchronized phasor measurements of remote points in the power system with a much faster sampling rate than that in the traditional supervisory control and data acquisition (SCADA) system. These synchrophasor measurements can significantly improve the accuracy of power system state estimation and system identification. Data losses, however, often happen in an unpredictable way during the communication between PMUs and the central operator. In this paper, we propose a novel low complexity data recovery algorithm that can capture the temporal correlations in time series therefore, it can efficiently recover the data in both random and temporal missing data cases. A key property of these PMU data matrices is that they are simultaneously structured that is simultaneously low-rank and slowly vary, i.e., the signal can be approximated by a piecewise constant function. By augmenting the nuclear norm to encourage low-rank property with the total variation norm to promote the piecewise constant structure, the missing data recovery can be processed using convex programming. The results are illustrated using some historical data from Lawrence Berkeley National Laboratory (LBNL) PMU dataset.