Robust Subspace Approaches for Analyzing Incomplete Synchrophasor Measurements

Abstract

Synchrophasor measurements can significantly enhance the monitorability of the power grid by revealing the dynamics of grid operation. However, due to high-rate samples collected in large volume, big data challenges emerge to efficiently process the data. The present work advocates robust subspace approaches including robust principal component analysis and subspace clustering, to identify low-dimensional structures in the synchrophasor data, even when portions of measurements are missing due to sensor or network issues, and outliers are present in the data. The outliers can model abnormal dynamics or cyber-attacks in the grid. Numerical tests using simulated synchrophasor data illustrate the utility of the approaches.

Online Robust Subspace Clustering With Application to Power Grid Monitoring

“…The PMU at each bus acquires measurements at a sampling rate of 40 samples per second and a signal-to-noise power ratio (SNR) of 92 dB. More detailed grid parameters are provided in [17]. To simulate events, the transmission line connecting buses 3001 and 3003 was tripped at t = 10 seconds and closed at t = 70 seconds.…”

“…Missing measurements in power grid data were reconstructed using low-rank matrix completion approaches in [15]. Subspace clustering approaches were devised and tested for synchrophasor data [16], [17]. Measurements deviating from the postulated models were detected as indicating disturbance events or cyber-attacks [17]- [19].…”

“…Subspace clustering approaches were devised and tested for synchrophasor data [16], [17]. Measurements deviating from the postulated models were detected as indicating disturbance events or cyber-attacks [17]- [19]. The classification problem to detect the attacks in the smart grid was tackled using supervised and semi-supervised methods [20], [21].…”

“…Some preliminary results were presented in conferences [17], [41]. Compared to these precursors, the current journal version contains slightly different formulations for the consistency of batch and online problems.…”