A5025050297- Image and Signal Denoising Methods
- Sparse and Compressive Sensing Techniques
- Smart Grid Security and Resilience
- Network Security and Intrusion Detection
- Power System Optimization and Stability
- Seismic Imaging and Inversion Techniques
- Blind Source Separation Techniques
- Information and Cyber Security
- NMR spectroscopy and applications
- Statistical and numerical algorithms
Rensselaer Polytechnic Institute
2015-2019
This paper presents a new framework to improve the quality of streaming synchrophasor measurements with existence missing data and bad data. The method exploits low-rank property Hankel structure identify correct data, as well estimate fill in is advantageous compared existing methods literature that only by leveraging observation matrix. proposed algorithm can efficiently differentiate event from even simultaneous consecutive has been verified through numerical experiments on recorded datasets.
This paper studies the multichannel missing data recovery problem when measurements are generated by a dynamical system. A new model, termed low-rank Hankel matrices, is proposed to characterize intrinsic low-dimensional structures in time series. The formulated as nonconvex optimization problem, and two fast algorithms (AM-FIHT RAM-FIHT), both with linear convergence rates, developed recover points provable performance guarantees. required number of observations significantly reduced,...
Cyber data attacks are the worst-case interacting bad to power system state estimation and cannot be detected by existing detectors. This paper for first time analyzes likelihood of cyber characterizing actions a malicious intruder in dynamic environment, where evolves with time, measurement devices could become inaccessible due attacks. analysis is important operator evaluate vulnerability systems A Markov decision process proposed model an intruder's strategy, objective maximize its...
This paper develops a model-free approach to recover the missing points in streaming synchrophasor measurements obtained nonlinear dynamical systems. It can accurately simultaneous and consecutive data losses across all channels for some time consecutively without modeling dynamics at all. The idea is lift system an infinite-dimensional linear exploit low-rank Hankel lifted dimension characterize dynamics. kernel technique employed handle implicit lifting function. Compared with existing...
This paper studies the low-rank matrix completion problem by exploiting temporal correlations in data. Connecting matrices with dynamical systems such as power systems, we propose a new model, termed multi-channel Hankel matrices, to characterize intrinsic low-dimensional structures collection of time series. An accelerated fast iterative hard thresholding (AM-FIHT) linear convergence rate is proposed recover missing points. The required number observed entries for successful recovery...
Cyber data attacks are the worst-case interacting bad to power system state estimation and cannot be detected by existing detectors. In this paper, we for first time analyze likelihood of cyber characterizing actions a malicious intruder. We propose use Markov decision process model an intruder's strategy, where objective is maximize cumulative reward across time. Linear programming method employed find optimal attack policy from perspective. Numerical experiments conducted study strategy in...
The large amounts of synchrophasor data obtained by Phasor Measurement Units (PMUs) provide dynamic visibility into power systems. Extracting reliable information from the can enhance system situational awareness. quality often suffers losses, bad data, and cyber attacks. Data privacy is also an increasing concern. In this paper, we discuss our recently proposed framework recovery, error correction, enhancement, event identification methods exploiting intrinsic low-dimensional structures in...