Data quality monitoring and improvement of synchronized phasor measurements

Thumbnail Image
Hao, Yingshuai
Issue Date
Electronic thesis
Electrical engineering
Research Projects
Organizational Units
Journal Issue
Alternative Title
Phasor measurement units (PMU) provide time-synchronized high-resolution measurements at rates from 30 up to 120 samples per second. Since the provided synchrophasor measurements facilitate the wide-area monitoring, protection, and control of power systems, the devices have been growingly deployed in North America. Data quality issue, however, limits the incorporation of the synchrophasor measurements into practical applications. In this thesis, we study the data quality improvement in several aspects and develop efficient algorithms to improve measurement quality through exploiting the intrinsic low dimensional property of data blocks, which results from the strong spatial-temporal correlations among different channels.
We finally study the missing data estimation during power system disturbances. Considering the nonlinearity of power systems under disturbances, existing algorithms based on the analysis of linear dynamical system models may result in a large deviation from the ground-truth data. We analyze the transformation of a nonlinear system model to a linear model by lifting the measurements and study how we can first estimate the lifted measurements and then compute the missing points. To reduce the computational complexity, we employ kernel trick and propose an efficient method to estimate the missing points with the adopted complex Gaussian kernel.
We also consider how to estimate the missing measurements in a batch data processing mode. Exploiting the low-rankness of the Hankel structure, we propose an efficient algorithm to estimate the missing points with a linear convergence rate and a low per-iteration complexity. The theoretical guarantee of the algorithm's convergence to the ground-truth data is provided. We further employ the heavy-ball method to accelerate the convergence of the developed algorithm theoretically. Numerical experiments on practical measurements show that the algorithm with the heavy-ball method not only enjoys a faster convergence rate, it can also tolerate a higher portion of data losses.
We next study how we can efficiently improve the quality of streaming synchrophasor measurements with the existence of missing data points and corrupted measurements. Based on the intrinsic low-dimensionality of data blocks, we further study the low-rank property of the Hankel structure constructed with consecutive synchrophasor data. Different from the original data block, the Hankel structure can capture the spatial-temporal correlations in time series measurements explicitly. This property facilitates us to differentiate measurements under system disturbances from consecutive corrupted measurements effectively and efficiently. Our proposed approach has low memory cost and low computational complexity and can pre-process streaming PMU data for subsequent real-time applications.
We first consider the cyber data attack, which represents one type of bad data injected by a malicious adversary to mislead system operators about the operating state of power systems. From the perspective of attackers, we study how attackers act in a dynamical environment and how we can model their attack decision across time. One contribution of this work is the development of a general model of an intruder's attack process. The problem is studied within the framework of Markov Decision Processes (MDPs). The solution to the formulated MDP is a mapping from power systems states to the intruder's optimal actions, including which buses to intrude and how to manipulate the measurements. Then the attack likelihood and the system vulnerability to attacks are analyzed from the optimal attack policy, which can benefit system operators on formulating rules to protect the devices and measurement channels from data attacks.
December 2018
School of Engineering
Full Citation
Rensselaer Polytechnic Institute, Troy, NY
PubMed ID