Vis enkel innførsel

dc.contributor.advisorNormann Anfinsen, Stian
dc.contributor.advisorRen, Huamin
dc.contributor.advisorCoello, Christoffer
dc.contributor.authorBarkost, Per Harald
dc.date.accessioned2020-09-09T09:19:32Z
dc.date.available2020-09-09T09:19:32Z
dc.date.issued2020-05-15
dc.description.abstractDetecting electrical vehicle (EV) charging from smart meter data (EV detection) is a highly relevant problem for the distribution system operators (DSOs), especially with the expected growth of EVs worldwide. There are several reasons why DSOs may want to detect EV charging. In the present day, the main motivation is to reduce the total load on the grid in high demand periods. This can be achieved by giving incentives to EV owners to charge their EVs in low demand periods. In the future, it is also anticipated that EVs can act as an energy reservoir, which can be a further motivation for EV detection. In this thesis, we explore two problems of EV detection. First, can we detect customers that charge an EV at home (EV load profiling)? Second, can we detect when an EV is charging (EV event detection)? To solve these problems, we analyze smart meter data provided by Eidsiva (a DSO from Norway). For the problem of load profiling, we propose, a feature-based Gaussian mixture modeling of weekly load profiles. The results are promising, showing that some EV owners have unique power consumption patterns. For the problem of event detection, we propose a modified version of UTime for EV event detection. UTime is a fully convolutional feed-forward neural network, initially proposed for sleep stage segmentation. The modified UTime is compared with previously proposed convolutional architectures for the problem of EV detection. Results show that UTime for EV detection outperforms the previous models on a generated labeled dataset. In order to solve the problem of EV detection, a labeled data set with ground truth is crucial. Unfortunately, this is lacking in this thesis. We resolve this issue by proposing a method of generating a labeled data set by combining two data sources. Even though the method show promise and models seem to generalize for an unlabeled dataset, more verification is needed to state conclusively that our proposed method is efficient.en_US
dc.identifier.urihttps://hdl.handle.net/10037/19274
dc.language.isoengen_US
dc.publisherUiT Norges arktiske universiteten_US
dc.publisherUiT The Arctic University of Norwayen_US
dc.rights.accessRightsopenAccessen_US
dc.rights.holderCopyright 2020 The Author(s)
dc.rights.urihttps://creativecommons.org/licenses/by-nc-sa/4.0en_US
dc.rightsAttribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)en_US
dc.subject.courseIDFYS-3900
dc.subjectVDP::Mathematics and natural science: 400::Information and communication science: 420::Simulation, visualization, signal processing, image processing: 429en_US
dc.subjectVDP::Matematikk og Naturvitenskap: 400::Informasjons- og kommunikasjonsvitenskap: 420::Simulering, visualisering, signalbehandling, bildeanalyse: 429en_US
dc.titleDetecting EV Charging From Hourly Smart Meter Dataen_US
dc.typeMaster thesisen_US
dc.typeMastergradsoppgaveen_US


Tilhørende fil(er)

Thumbnail
Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel

Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)
Med mindre det står noe annet, er denne innførselens lisens beskrevet som Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)