A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs
Identifying subsurface resource analogs from mature subsurface datasets is vital for developing new prospects due to often initial limited or absent information. Traditional methods for selecting these analogs, executed by domain experts, face challenges due to subsurface dataset's high complex...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SAGE Publishing
2024-03-01
|
| Series: | Energy Exploration & Exploitation |
| Online Access: | https://doi.org/10.1177/01445987231210966 |
| _version_ | 1827311285077803008 |
|---|---|
| author | Ademide O. Mabadeje Jose J. Salazar Jesus Ochoa Lean Garland Michael J. Pyrcz |
| author_facet | Ademide O. Mabadeje Jose J. Salazar Jesus Ochoa Lean Garland Michael J. Pyrcz |
| author_sort | Ademide O. Mabadeje |
| collection | DOAJ |
| description | Identifying subsurface resource analogs from mature subsurface datasets is vital for developing new prospects due to often initial limited or absent information. Traditional methods for selecting these analogs, executed by domain experts, face challenges due to subsurface dataset's high complexity, noise, and dimensionality. This article aims to simplify this process by introducing an objective geostatistics-based machine learning workflow for analog selection. Our innovative workflow offers a systematic and unbiased solution, incorporating a new dissimilarity metric and scoring metrics, group consistency, and pairwise similarity scores. These elements effectively account for spatial and multivariate data relationships, measuring similarities within and between groups in reduced dimensional spaces. Our workflow begins with multidimensional scaling from inferential machine learning, utilizing our dissimilarity metric to obtain data representations in a reduced dimensional space. Following this, density-based spatial clustering of applications with noise identifies analog clusters and spatial analogs in the reduced space. Then, our scoring metrics assist in quantifying and identifying analogous data samples, while providing useful diagnostics for resource exploration. We demonstrate the efficacy of this workflow with wells from the Duvernay Formation and a test scenario incorporating various well types common in unconventional reservoirs, including infill, outlier, sparse, and centered wells. Through this application, we successfully identified and grouped analog clusters of test well samples based on geological properties and cumulative gas production, showcasing the potential of our proposed workflow for practical use in the field. |
| first_indexed | 2024-04-24T20:16:13Z |
| format | Article |
| id | doaj.art-476391cc3de24647afb4a848186468b8 |
| institution | Directory Open Access Journal |
| issn | 0144-5987 2048-4054 |
| language | English |
| last_indexed | 2024-04-24T20:16:13Z |
| publishDate | 2024-03-01 |
| publisher | SAGE Publishing |
| record_format | Article |
| series | Energy Exploration & Exploitation |
| spelling | doaj.art-476391cc3de24647afb4a848186468b82024-03-22T17:03:54ZengSAGE PublishingEnergy Exploration & Exploitation0144-59872048-40542024-03-014210.1177/01445987231210966A Machine Learning Workflow to Support the Identification of Subsurface Resource AnalogsAdemide O. Mabadeje0Jose J. Salazar1Jesus Ochoa2Lean Garland3Michael J. Pyrcz4 Hildebrand Department of Petroleum & Geosystems Engineering, , Austin, TX, USA Facultad de Ingeniería en Ciencias de la Tierra, , Guayaquil, Ecuador Technology, Digital & Innovation, , Houston, TX, USA Exploration & Production international, , Oslo, Norway Department of Geological Sciences, , Austin, TX, USAIdentifying subsurface resource analogs from mature subsurface datasets is vital for developing new prospects due to often initial limited or absent information. Traditional methods for selecting these analogs, executed by domain experts, face challenges due to subsurface dataset's high complexity, noise, and dimensionality. This article aims to simplify this process by introducing an objective geostatistics-based machine learning workflow for analog selection. Our innovative workflow offers a systematic and unbiased solution, incorporating a new dissimilarity metric and scoring metrics, group consistency, and pairwise similarity scores. These elements effectively account for spatial and multivariate data relationships, measuring similarities within and between groups in reduced dimensional spaces. Our workflow begins with multidimensional scaling from inferential machine learning, utilizing our dissimilarity metric to obtain data representations in a reduced dimensional space. Following this, density-based spatial clustering of applications with noise identifies analog clusters and spatial analogs in the reduced space. Then, our scoring metrics assist in quantifying and identifying analogous data samples, while providing useful diagnostics for resource exploration. We demonstrate the efficacy of this workflow with wells from the Duvernay Formation and a test scenario incorporating various well types common in unconventional reservoirs, including infill, outlier, sparse, and centered wells. Through this application, we successfully identified and grouped analog clusters of test well samples based on geological properties and cumulative gas production, showcasing the potential of our proposed workflow for practical use in the field.https://doi.org/10.1177/01445987231210966 |
| spellingShingle | Ademide O. Mabadeje Jose J. Salazar Jesus Ochoa Lean Garland Michael J. Pyrcz A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs Energy Exploration & Exploitation |
| title | A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs |
| title_full | A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs |
| title_fullStr | A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs |
| title_full_unstemmed | A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs |
| title_short | A Machine Learning Workflow to Support the Identification of Subsurface Resource Analogs |
| title_sort | machine learning workflow to support the identification of subsurface resource analogs |
| url | https://doi.org/10.1177/01445987231210966 |
| work_keys_str_mv | AT ademideomabadeje amachinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT josejsalazar amachinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT jesusochoa amachinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT leangarland amachinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT michaeljpyrcz amachinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT ademideomabadeje machinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT josejsalazar machinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT jesusochoa machinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT leangarland machinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs AT michaeljpyrcz machinelearningworkflowtosupporttheidentificationofsubsurfaceresourceanalogs |