Show simple item record

dc.creatorNoureen, Subrina Sultana
dc.date.accessioned2020-06-24T00:26:29Z
dc.date.available2020-06-24T00:26:29Z
dc.date.created2019-08
dc.date.issued2019-08
dc.date.submittedAugust 2019
dc.identifier.urihttps://hdl.handle.net/2346/85995
dc.description.abstractA multitude of advancements in the field of smart grids has added new features like bidirectional communication protocols with the center control unit to the existing power grids, which begets numerous security concerns to the generation and distribution level of smart grids. Thus, physical and cyber security threats along with, the weaknesses in the smart grid demands identification beforehand. An unparalleled flexible management system with intelligent officiating capabilities has no alternatives to ensure highly efficient energy management system for vitiated energy depletion. To achieve this goal, predicting the future energy requirement (i.e. load) is considered as one of the key features. This dissertation primarily focuses on analyzing various anomalies in a smart grid system by implementing a few machine learning techniques. Additionally, it proposes some improvement techniques for a higher accuracy rate for the energy demand forecasting. To examine different scenarios of physical and cyber anomalies, this work adopted the Phasor Measurement Unit (PMU) data optimized with Binary Swarm algorithm from a simulated power grid system along with a real-time simulator system, which were added to the virtual PMUs to create the cyber-attacks. Furthermore, to detect malicious events, a novel approach was adopted by employing a Regression algorithm, along with, a Scikit-learn machine learning algorithm for clustering and capturing real-time data. Afterward, a synthetic data set was generated and fed for testing and verification purposes. A benchmark data set of the small-scale industrial load was taken into account for both “train and test” methods on monthly and yearly time-span data. Last, a comparative and elaborate study was employed between a time series analysis technique (ARIMA model) and a ML technique (Random Forest) for determining higher accuracy rate in regard to better forecasting energy demand.
dc.format.mimetypeapplication/pdf
dc.language.isoeng
dc.subjectSmart grid analysis
dc.subjectCyber-physical system
dc.subjectLogistic regression
dc.subjectARIMA model
dc.subjectRandom forest algorithm
dc.subjectLoad forecasting
dc.titleObservability, anomaly detection, and energy demand forecasting for smart grid analysis
dc.typeThesis
dc.date.updated2020-06-24T00:26:30Z
dc.type.materialtext
thesis.degree.nameDoctor of Philosophy
thesis.degree.levelDoctoral
thesis.degree.disciplineElectrical Engineering
thesis.degree.grantorTexas Tech University
thesis.degree.departmentElectrical and Computer Engineering
dc.contributor.committeeMemberGiesselmann, Michael
dc.contributor.committeeMemberHe, Miao
dc.contributor.committeeChairBayne, Stephen B.
dc.rights.availabilityRestricted until 2024-09.
local.embargo.terms2024-08-01
local.embargo.lift2024-08-01


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record