Data Driven Model Discovery - Petroleum application
dc.contributor.advisor | Hiorth, Aksel | |
dc.contributor.author | Søndeland, Sander André | |
dc.date.accessioned | 2023-09-15T15:51:17Z | |
dc.date.available | 2023-09-15T15:51:17Z | |
dc.date.issued | 2023 | |
dc.identifier | no.uis:inspera:129730556:37060453 | |
dc.identifier.uri | https://hdl.handle.net/11250/3089786 | |
dc.description.abstract | ||
dc.description.abstract | The SINDy algorithm is a data driven algorithm that discovers dynamical system in data that evolves over time. The method can be utilized for every dataset that evolves over time. In this study we have looked the Lorenz system, covid-19 data and production data from two different oil fields on the Norwegian shelf. The aim of the study was to investigate if SINDy can be used on the well data to extract sparse and suitable well models. The complexity of the models are decided by the user when using prior knowledge to choose the candidate function. If you have limited knowledge about the system a handful of different models are tested and parameters are optimized to fit the data. Noisy and spiky data are an issue for the SINDy method due to its use of the differentiated data. Therefor filtering is needed on production data to minimize the large spikes and smooth out the data. The SINDy algorithm gives good results to the production data using polynomials to describe the data. The results are good for data from Draugen and Statfjord Øst. And the results from the covid-19 data are promising. | |
dc.language | eng | |
dc.publisher | uis | |
dc.title | Data Driven Model Discovery - Petroleum application | |
dc.type | Master thesis |
Tilhørende fil(er)
Denne innførselen finnes i følgende samling(er)
-
Studentoppgaver (TN-IDE) [823]
Studentoppgaver i informasjonsteknologi, datateknikk / kybernetikk, signalbehandling