BIO: Halvor Jahre
Geologist with a strong interest for geophysics who grew up in Saga Petroleum in the 1990s. Explored for oil in Yemen and Mozambique in the early 2000s with DNO. Founding member of Genesis Petroleum from 2004-2009. Joined Lundin Norway in January 2010 and headed up one of the most successful Exploration depts. in Europe from 2015 until January 2020. R&D, technology development and smart solutions regarding use of data is also close to my heart. Now, since November 2020, doing exiting new ventures in Harbour Energy.

SUMMARY:
The initiative to make all historic exploration and appraisal wells started as rebellion which landed well. The trigger was the 2018 Konkraft report, a joint effort from the industry associations and the labor unions. This time, the exploration part of the report suffered from good ideas and concluded with drilling one or more stratigraphic well(s) in underexplored basins as the best joint effort. A time-out and a creative exploration department in Lundin ended up in proposing to make all exploration and appraisal wells drilled on the Norwegian shelf since 1965 a part of the digital journey. The purpose was to prepare drilling and exploration for the digital future. To provide an unbiased complete digital library for the whole well, not only the reservoir, for cooperation on the Norwegian shelf (NOCS). The alternative to stratigraphic well(s) ended up with analyzing the cuttings data stored for all >1900 exploration and appraisal wells in Norway using an established workflow. The workflow starts with washing, photo in UV and white light before a workflow with XRD (X-Ray diffraction), XRF (X-Ray fluorescence), TOC (Total Organic Carbon), SPECAM (Infrared Spectroscopy) and QEMSCAN (Energy-dispersive X-Ray Spectroscopy) was applied on 5% of the wells. The 5% wells (80 in total) were simply the selection of the most recently released wells. The rest received photo and XRF. There is a good correlation, in the order of 95%, between the sample sets. The operators/owners of the wells supplied cuttings from their archives and NPD made their archive available where the companies had holes in their database. The total number of wells for the project ended with 1933, more than 700 000 samples were analyzed and stored in 3.8 million files filling 60 terabytes into the national DISKOS database. 95% of the samples have undergone non-destructive analysis, is archived and readily available for the extended workflow. Photo and XRF is now establish as compulsory for all future exploration and appraisal wells on the NOCS. Hence, the database will continue to grow as an unbiased standardized dataset.
The more detailed list of project goals is ambitious and can be significantly expanded depending on digitalization of other elements of the wells, in particular drilling parameters. There are a number of AI/ML related drilling initiatives and workflows in the industry, and any element influenced by geological parameters can bolt on to this database. Project goals involves:
* Improved correlation of log responses to cuttings in uncored intervals including grainsize, porosity, permeability estimations and improved mineralogical prediction from XRF.
* Improved geological resolution where petrophysical and seismic data is non-conclusive.
* Identification of permeable and non-permeable zones leading to improved well design and better reservoir predictions.

Dismiss

BIO: Doug Langton
Douglas has a degree in Geology from Kingston University and a Master’s degree in Engineering Geology from University of Durham. He has been a Director of Rockwash since 2011, responsible for the business development of the company. He has more than 30 years’ experience in oil exploration and geotechnical fields, working at the wellsite and developing geological services and datasets globally.

SUMMARY:
One of the leading principles of the Norwegian Released Wells Initiative, an industry-supported cuttings digitalisation project, was to analyse every cuttings sample available for each well to generate a continuous and consistent dataset that could be compared from well-to-well. Cuttings images and XRF (X-ray Fluorescence) data were collected for every well and, for a selection of 80 of the most recently drilled exploration wells, TOC (Total Organic Carbon), XRD (X-ray Diffraction), SpecCam Infrared Spectroscopy and QEMSCAN (Energy-dispersive X-ray Spectroscopy) analyses were performed on every cuttings sample.
For this presentation, we have selected seven of the recently drilled wells along the southern edge of the Halten Terrace, bordering the northwestern flank of the Frøya High. Full-well profiles have been prepared displaying processed XRF, XRD, Spec Cam and TOC data to demonstrate the elemental composition, mineralogical and lithological features of each geological formation through the wells. The Jurassic and Triassic targets for each well will be compared to highlight new insights that can be gained from the presentation and interpretation of cuttings data with respect to lithology, reservoir properties, sediment provenance, and indications of source rock potential. We will present how the integration of the cuttings data with wireline log data enhances the modern post-well evaluation process for both exploration and development applications.

Dismiss

BIO: Dafydd Evans
Dafydd is a geologist with a degree and masters in geology from the University of South Wales and has since been working at Rockwash Geodata since 2022. At Rockwash Dafydd has been working on the development of machine learning models for the interpretation of drill cuttings images in close partnership with Earth Science Analytics, along with being the chair of a research consortium for the use of such models on the NOROG dataset.

SUMMARY:
Through the completion of the Norwegian Released Wells Initiative (RWI), cutting sample images have been collected from over 1,900 wells within the Norwegian continental shelf area, providing a total of over 715,000 images. Within these images is valuable geological data such as the modal lithology of a sample, sample mineralogy, sand grain size and cementation properties within reservoir sections, as well as the presence of any contaminants within a sample. However, as this data needs to be manually interpreted from the image, which requires an experienced wellsite geologist, it would be prohibitively time consuming and expensive to extract this data from every sample image. With this in mind Earth Science analytics teamed up with Rockwash and formed a research consortium with the goal of automating the interpretation of these samples using machine learning.



There are a number of machine learning methods that can be used for identifying features within images, of which we have used, image classification, semantic segmentation, and object detection, with future plans to include instance segmentation. The primary difficulty in producing working models using these methods is the requirement of having a label set for the model the train, which must meet the requirements of being both high quantity and high quality. To achieve this, we have utilized a number of methods, including active learning, model-assisted labelling and programmatic labelling using auxiliary data. Using these methods, we have been able to produce working models for the interpretation of the modal lithology of a sample using semantic segmentation, as well as sand grain size and cementation properties through image classification. In this talk we will give an overview of each computer vision method and labelling strategy used, along with examples of sample and well data collected using these methods.

Dismiss

BIO: Jenny Omma
Jenny is a Senior Technical Advisor at Stratum Reservoir, focussed on delivery of QEMSCAN data to oil & gas, mining and CCS industries. She was previously Geological Director of Rocktype, before Rocktype was acquired by Stratum Reservoir. Prior to founding Rocktype in 2014, she was a reservoir quality geologist with BP’s central services team, supporting global assets. She worked as a heavy mineral geosteering expert at wellsite on the BP Clair Field. Jenny completed a PhD on sediment provenance into Arctic Basins, based at CASP and Cambridge University.

SUMMARY:
External pressures on the oil & gas industry require a shift to more efficient working practices, with shorter project time scales. In response, the industry is becoming more collaborative, digital and data-centric and there is an increased call to create Big Data to analyse in algorithmic, next generation subsurface workflows. Big Data studies require standardised, digital data, which in turn requires data collection to be standardised and automated.
To date the impact of mineral data on many subsurface workflows has been limited by the time-consuming and expensive nature of collecting mineralogical data. High-volume data collection is made possible by automation decisions, strategic technological innovation and continuous process improvement.
For the Norwegian Cuttings Project, we delivered automated mineralogy (QEMSCAN) data from 25K cuttings samples, analysing all cuttings from first returns to total depth for 80 wells.
Full well cuttings QEMSCAN data has the potential to underpin next generation subsurface workflows, bringing large scale mineralogy data to the heart of subsurface workflows, including as input priors for rock physics models, for next generation seismic inversion workflows and as input data for reservoir simulations powered by machine learning.
In this talk we present QEMSCAN data from the Norwegian Cuttings Project and also from a 5000 cuttings dataset from eight Faroe Islands wells, with datasets including cuttings photographs, QEMSCAN images, modal mineralogy, per lithotype mineralogy separating results from sandstones, siltstones, shales, limestones per sample), cuttings size, grain size, calculated log properties and predicted mineralogical changes from CO2 injection into key reservoirs. Stratigraphic, sediment provenance and reservoir quality interpretations derived from the analysis of the data will also be presented.

Dismiss