Missing data detection and estimation has been studied in many applications, such as image processing, computer network, and traffic information. In power and energy industry, it draws a lot of attentions as the number of Phasor Measurement Units (PMU) increases and many measurement-based applications are proposed. As accurate data is required to have better results in the applications, many bad data detection algorithms have been introduced to have reliable data. Motivated by low dimensional characteristics of PMU data, new bad PMU data detection and correction method is proposed and tested with synthetic PMU data. The proposed bad data detection and estimation method is based on subspace projection and lifting, and is strongly related to the least squares solution of overdetermined equations.

Electric power systems can display a range of undesirable dynamic phenomena by which acceptable, stable operation may be lost. Quasi-steady-state operational problems such as the voltage instability phenomena are among these. Ill-conditioning of the power flow, reflected in high sensitivity of bus voltage magnitudes to load variation is an often observed precursor to such quasi-steady state operational problems. Motivated by this insight, work here will propose a voltage stability and conditioning monitor that is model-free in real-time, based solely on phasor measurement unit (PMU) data, arguing that such an approach is well suited to near-real-time application. We review model-dependent singular value analysis in voltage stability assessment, and relate these existing approaches to our proposed model-free method in real-time application. The proposed algorithm is first applied under the idealized assumption of full measurement data at all buses. This work then extends the algorithm to apply in the more practical case for which only subset of buses have available measurement data. The proposed approach is illustrated in IEEE test cases, augmented to include heavy load conditions that stress voltage stability. Algorithms for efficient computation of small numbers of singular values, as well as means to exploit low-rank updates in data, are reviewed to demonstrate opportunities for fast computation that allow these SVD methods to operate in near-real-time in large systems.

Extensive researches have been done to partition power system network into clusters based on coherency. The needs for partitioning in Singular Value Decomposition (SVD) based information retrieval using Phasor Measurement Unit (PMU) data is similar to the exiting partitioning algorithm, but it is not identical. The SVD based method's primal concern is to distribute SVD computation burden to each partition. In this paper, one of existing partitioning algorithms is reviewed and apply it for SVD based information retrieval algorithm. Also, the existing algorithm is modified to meet the need of SVD based method without breaking coherency much.

Ill-conditioning of the power flow, reflected in high sensitivity of bus voltage magnitudes to load variation, often serves as a basic indicator of vulnerability to voltage instability. This sensitivity may be quantified by the induced 2-norm associated with the power flow Jacobian inverse, as may be computed from Singular Value Decomposition(SVD). While this conceptual picture is simple, near-real-time assembly of the power flow Jacobian under volatile operating conditions, and monitoring its singular values, presents severe computational and data management problems. As an alternative, work here will propose a “model-free” voltage stability and conditioning monitor, based solely on phasor measurement unit (PMU) data, arguing that such an approach is much better suited to near-real-time application. We first review singular value analysis in voltage stability assessment, and relate these existing approaches to our proposed method. We offer computational evidence that the proposed measure closely tracks the largest singular value of the inverse of the power flow Jacobian under continuous change in operating point, without the need for any network admittance data nor construction of the power flow model. This work goes on to examine characteristics of the PMU-based SVD measure in scenarios of discontinuous changes to network topology. We demonstrate that the new measure also provides a highly sensitive indicator of the occurrence of topology change, that may offer a valuable “consistency check” to supplement directly reported breaker status.

Electric power systems can display a range of undesirable dynamic phenomena by which acceptable, stable operation may be lost; among these is the “voltage instability” phenomena. Formal analyses in smooth models classify voltage instability as a bifurcation phenomena. Ill-conditioning of the power flow, reflected in high sensitivity of bus voltage magnitudes to load variation is an often observed precursor to voltage instability. Motivated by this insight, this work will propose a conditioning monitor based solely on phasor measurement unit (PMU) data, without need for network parameters or topology to construct the power flow model, and hence suitable for near-real-time implementation. As an initial proof of concept in computational test systems (for which complete power flow data is available), we will demonstrate that the proposed measure closely tracks the largest singular value of the inverse of the power flow Jacobian.