Exploiting Low-dimensional Structures in Data Management of High-dimensional Synchrophasor Measurements for Power System Monitoring

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

Phasor measurements units (PMUs) in North America provide terabytes of synchronized phasor measurements of key operational parameters of power systems on a daily basis. In current practice of power system operation, the large amounts of PMU measurements are stored for past event analysis, and the in-situ data processing for real-time decision is beyond the capability of current technologies used in power systems. This proposal will develop a framework of collective processing of measurements in multiple PMUs across multiple time instants for various tasks in power system monitoring. The research goal is to develop efficient data management and information extraction methods that are suitable for real-time processing of large volumes of PMU data. The outcome of this project will positively impact the reliable operation of future power systems. The generic techniques for high-dimensional data analysis developed in this project can potentially find applications in other areas beyond power systems, e.g., Internet monitoring, social network analysis, image and video processing, etc. This proposal also contains an integrated educational agenda for K-12 students, undergraduates and graduate students. This proposal for the first time bridges the areas of power system monitoring and high-dimensional analysis based on low-dimensional models. By developing new generic tools that are motivated by tasks in power system monitoring, this proposal will contribute to the development of the field of PMU-based power system monitoring and the field of high-dimensional data analysis. Focusing on improving data integrity and data accuracy of PMU measurements, this proposal will address the following challenges and open questions by: 1. Developing new computationally efficient missing data recovery methods to fill in the measurements that are lost during communication. 2. Developing new methods to detect events in power systems by collectively processing PMU measurements in multiple channels. 3. Developing new convex-optimization-based methods to detect cyber data attacks to PMU measurements. 4. Analyzing the likelihood and frequency of cyber data attacks to power systems. All the developed methods will be numerically evaluated on actual PMU data in Central New York Power System. This proposal will establish a framework of data-challenged power system monitoring by exploiting low-dimensional structures. It will extend the current understanding of low rank methods to the field of PMU data analysis. It will connect high-dimensional data analysis with graph theory through the research on the detection of cyber data attacks to power systems. This project will provide new insights to optimization-based high-dimensional data analysis by investigating the theoretical limits of proposed methods.