Error correction of vitrinite reflectance in matured black shales: A machine learning approach
Vitrinite reflectance (Ro) analysis is a maturity indication parameter for oil or gas prone source rocks in evaluating hydrocarbon potentials. As a result of challenges in Ro calculations from pyrolysis results, finding a way to estimate the maturity of source rocks has been an interesting subject f...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
KeAi Communications Co., Ltd.
2022-01-01
|
Series: | Unconventional Resources |
Subjects: | |
Online Access: | http://www.sciencedirect.com/science/article/pii/S2666519022000048 |
_version_ | 1797861105479974912 |
---|---|
author | Esther Boateng Owusu George Mensah Tetteh Solomon Asante-Okyere Haylay Tsegab |
author_facet | Esther Boateng Owusu George Mensah Tetteh Solomon Asante-Okyere Haylay Tsegab |
author_sort | Esther Boateng Owusu |
collection | DOAJ |
description | Vitrinite reflectance (Ro) analysis is a maturity indication parameter for oil or gas prone source rocks in evaluating hydrocarbon potentials. As a result of challenges in Ro calculations from pyrolysis results, finding a way to estimate the maturity of source rocks has been an interesting subject for researchers. There is a current need to improve the Ro calculated from the temperature at which the maximum rate of hydrocarbon generation occurs (Tmax) during maturation of shale formations. As this will go a long way to accelerate decision making and help avoid excessive expenditure costs on maturity determination using measured Ro for source rock samples while saving time. After the application of the conventional multiple linear regression analysis, the present study employed machine learning methods of random forest (RF), decision tree (DT), gradient boosting machine (GBM), ensembles and bagger (EnB), and multivariate adaptive regression splines (MARS) models for improving the calculated Ro of matured black shales. Total organic carbon (TOC), oxygen index (OI), the amount of carbon dioxide produced during pyrolysis of kerogen (S3), and vitrinite reflectance measurement (Ro) were used as inputs to estimate the error margin between the calculated vitrinite reflectance and measured vitrinite reflectance. The predictions from the models were then summed with the calculated vitrinite reflectance to produce an improved vitrinite reflectance measurement. The model that generated the most improved vitrinite reflectance measurement, thus, having the least amount of statistical error was selected. EnB achieved the highest accuracy with a correlation coefficient (R) of 0.82 and coefficient of determination (R2) of 0.67 for the improved Ro model. Therefore, a conclusion can be drawn from the results that EnB can adequately improve the Ro of matured rocks through error correction. |
first_indexed | 2024-04-09T21:56:42Z |
format | Article |
id | doaj.art-b60bd0607158401d93df0dd2faec7ec4 |
institution | Directory Open Access Journal |
issn | 2666-5190 |
language | English |
last_indexed | 2024-04-09T21:56:42Z |
publishDate | 2022-01-01 |
publisher | KeAi Communications Co., Ltd. |
record_format | Article |
series | Unconventional Resources |
spelling | doaj.art-b60bd0607158401d93df0dd2faec7ec42023-03-24T04:23:25ZengKeAi Communications Co., Ltd.Unconventional Resources2666-51902022-01-0124150Error correction of vitrinite reflectance in matured black shales: A machine learning approachEsther Boateng Owusu0George Mensah Tetteh1Solomon Asante-Okyere2Haylay Tsegab3University of Mines and Technology, Department of Petroleum Geosciences and Engineering, School of Petroleum Studies, Tarkwa, Ghana; Corresponding author.University of Mines and Technology, Department of Geological Engineering, Faculty of Mineral Resources Technology, Tarkwa, GhanaUniversity of Mines and Technology, Department of Petroleum and Natural Gas Engineering, School of Petroleum Studies, Tarkwa, GhanaUniversiti Technologi Petronas, Department of Petroleum Geoscience, 32610, Bandar Seri Iskandar, Perak Darul Ridzuan, Malaysia; Universiti Teknologi PETRONAS, Southeast Asia Carbonate Research Laboratory, Department of Geoscience, Perak Darul Ridzuan, 32610, Bander Seri Iskander, Tronoh, Perak, MalaysiaVitrinite reflectance (Ro) analysis is a maturity indication parameter for oil or gas prone source rocks in evaluating hydrocarbon potentials. As a result of challenges in Ro calculations from pyrolysis results, finding a way to estimate the maturity of source rocks has been an interesting subject for researchers. There is a current need to improve the Ro calculated from the temperature at which the maximum rate of hydrocarbon generation occurs (Tmax) during maturation of shale formations. As this will go a long way to accelerate decision making and help avoid excessive expenditure costs on maturity determination using measured Ro for source rock samples while saving time. After the application of the conventional multiple linear regression analysis, the present study employed machine learning methods of random forest (RF), decision tree (DT), gradient boosting machine (GBM), ensembles and bagger (EnB), and multivariate adaptive regression splines (MARS) models for improving the calculated Ro of matured black shales. Total organic carbon (TOC), oxygen index (OI), the amount of carbon dioxide produced during pyrolysis of kerogen (S3), and vitrinite reflectance measurement (Ro) were used as inputs to estimate the error margin between the calculated vitrinite reflectance and measured vitrinite reflectance. The predictions from the models were then summed with the calculated vitrinite reflectance to produce an improved vitrinite reflectance measurement. The model that generated the most improved vitrinite reflectance measurement, thus, having the least amount of statistical error was selected. EnB achieved the highest accuracy with a correlation coefficient (R) of 0.82 and coefficient of determination (R2) of 0.67 for the improved Ro model. Therefore, a conclusion can be drawn from the results that EnB can adequately improve the Ro of matured rocks through error correction.http://www.sciencedirect.com/science/article/pii/S2666519022000048Rock-eval pyrolysisVitrinite reflectanceMatured black shalesMultiple linear regressionMachine learning |
spellingShingle | Esther Boateng Owusu George Mensah Tetteh Solomon Asante-Okyere Haylay Tsegab Error correction of vitrinite reflectance in matured black shales: A machine learning approach Unconventional Resources Rock-eval pyrolysis Vitrinite reflectance Matured black shales Multiple linear regression Machine learning |
title | Error correction of vitrinite reflectance in matured black shales: A machine learning approach |
title_full | Error correction of vitrinite reflectance in matured black shales: A machine learning approach |
title_fullStr | Error correction of vitrinite reflectance in matured black shales: A machine learning approach |
title_full_unstemmed | Error correction of vitrinite reflectance in matured black shales: A machine learning approach |
title_short | Error correction of vitrinite reflectance in matured black shales: A machine learning approach |
title_sort | error correction of vitrinite reflectance in matured black shales a machine learning approach |
topic | Rock-eval pyrolysis Vitrinite reflectance Matured black shales Multiple linear regression Machine learning |
url | http://www.sciencedirect.com/science/article/pii/S2666519022000048 |
work_keys_str_mv | AT estherboatengowusu errorcorrectionofvitrinitereflectanceinmaturedblackshalesamachinelearningapproach AT georgemensahtetteh errorcorrectionofvitrinitereflectanceinmaturedblackshalesamachinelearningapproach AT solomonasanteokyere errorcorrectionofvitrinitereflectanceinmaturedblackshalesamachinelearningapproach AT haylaytsegab errorcorrectionofvitrinitereflectanceinmaturedblackshalesamachinelearningapproach |