Using machine learning to identify remaining hydrocarbon potential

Transcription

1 Using machine learning to identify remaining hydrocarbon potential The Oil & Gas Technology Centre Open Innovation Programme Call for Ideas Technical Documentation A Call for Ideas, part of the OGTC Open Innovation Programme.

2 2 Our Call for Ideas Grab a slice of c. 1 million to develop and bring your idea to reality. The Oil & Gas Technology Centre is calling for innovative ideas and concepts that will have a material impact on the identification of potential remaining hydrocarbon zones in North Sea oil and gas wells. We have a total fund of c. 1 million which we are looking to distribute to form a portfolio of projects that can be delivered in a 1 12-month timeframe. The Challengex Many oil and gas fields in the Northern North Sea (NNS) are nearing the end of their expected operating lives. If new sources of hydrocarbons are not found, then much of the production infrastructure (e.g. wells, pipelines, risers) may be decommissioned in the near future. There are large volumes of data from NNS oil and gas wells that could lead to new discoveries, but it is often held in different formats, of variable quality and takes time to assess. Traditionally, organising, distilling, screening and assessing data would require a large team of specialists, and there is always a risk of a person missing something or introducing unconscious bias into the evaluation. We re looking for digital analytics approaches, such as machine learning, that can deliver assessments of structured and unstructured well data quickly and accurately. The aim is to use the available data to identify and classify intervals which may indicate the presence of previously unrecognized or overlooked petroleum accumulations. This would allow the industry to fully evaluate these opportunities and prioritise investment, potentially extending field life and maximising economic recovery. We would like to hear your ideas to help with this challenge. Available Data If your idea is successful, over 175,000 items of industry data associated with wells will be made available to you by Common Access Data (CDA) and the Norwegian Petroleum Directorate (NPD), through the Oil and Gas Authority (OGA). The main types of data are summarised below (the chart

3 3 shows data available from CDA only this covers around half of the wells in the area). Much of it is in image format. The data sets may also be supplemented with unreleased data contributed by oil and gas companies in the area. Free samples of the data can be downloaded from with the creation of an online account. Once an account is created, data sets can be found under the OGA tab. Note that these sample data are from the entire UK offshore area and therefore only some will be from within the area of interest for this project. The samples are, however, representative of the type and format of data that will be made available to the project. The following tabs contain pertinent data:-

4 4 The 2015 Seismic Programme: Representative data from over 100 wells, including a number within the Northern North Sea The 30 th Round Undeveloped Discoveries Programme: Small Pool Discovery Montages and seismic and well data for over 900 wells (data from which could be used to train models) The 2016 Seismic Programme: Further data from 236 wells from across the UKCS, including composited log files. Similar data will be made available for Norwegian wells but the type and format will differ. Information can be found on the Diskos website ( where metadata can be downloaded. Additionally, published data can be accessed directly from the NPD website under the Factpages section ( An example data montage is illustrated below: Project Themesx Theme 1: Data conditioning Problem statement There is a wealth of structured and unstructured data within the Northern North Sea. However, the formats and quality of data varies considerably. The challenge is to use or develop techniques that can extract, compile and assemble integrated data sets that are suitable for further analysis. We would like to receive project proposals that address these challenges and: Provide access to and transfer of or merging of relevant datasets.

5 5 Scrape data sources and extract data (i.e. apply OCR or similar techniques to scanned images, identify data, e.g. lithology descriptions, core sample points, analytical data such as porosity measurements and parse into useable formats). Cleanse the data identify and rationalise duplicate data, identify erroneous or contradictory data and metadata, standardise units, identify (and potentially fill) gaps. (Industry participants may be able to donate datasets that are clean and have had a certain amount of prepreparation.) Catalogue the data while the data provided will already be organised by well and by type, the analytical approach may vary based on available data so re-organising available data into relevant categories could be useful. Ensure clarity of data provenance and conditions of use are maintained. Theme 2: Data analysis Problem statement There are multiple data types available resulting from drilling activities, the logging and ongoing testing of wells. The challenge is to assimilate and evaluate these in a consistent manner across large areas and to identify and predict remaining hydrocarbon potential. This usually takes a large team of specialists working together in an integrated way to identify this remaining potential. However, there is always a risk of the human missing something or introducing bias into the evaluation. We would like to receive project proposals that: Automate petrophysical characterization of log and other responses related to hydrocarbonfilled reservoir units of differing lithologies and stratigraphies ( pay intervals or zones) Apply appropriate solutions, such as machine learning techniques, to statistically extract geological intervals of interest, pay zones, and rank according to confidence of response match Populate databases in an industry workstation-readable format including depth and location characteristics Project Deliverablesx We envisage that successful projects will deliver results in some of the following forms but not limited to those: Brief technical report to describe the workflow, inputs, outputs, etc Summary technical presentation Delivery of any cleaned or conditioned well database in digital format List of remaining hydrocarbon pay opportunities, ideally in the form of a target database, ranked according to statistical confidence. Project proposals should be no more than 8 sides.

6 6 Proposed area of interestx The proposed area of interest is the Northern North Sea and it is envisaged that the initial projects will focus on using all available well data within this area. Technical Detailsx The identification of potential hydrocarbon bearing units ( pay ) should be based on the analysis of all available subsurface information from wells. The area of interest includes several thousand wells (including all exploration, appraisal and development wells) and the intent is to use these initially. Carrying out such an analysis with humans would take considerable time and effort, hence the requirement to use established and emerging machine learning and big data techniques to accelerate the analytical process. The project should use algorithms and software to build a database of targets that are ranked according to their statistical confidence. This database can then be published to industry to allow further maturation of opportunities supplemented by other techniques. The input data should include, but may not be limited to, released well data which includes structured, semi-structured and unstructured sources such as electric well logs, scanned pdf images and digital tabular data. An automated data extraction, transformation and loading phase will be required, and once a database has been populated, rules-based statistical methods can be applied to identify hydrocarbon pay and reservoir characteristics. If a training data set is required, there are numerous proven and producing discoveries within the NNS that can be used to train and test algorithms. Data may require conditioning prior to analysis. It is hoped that technological solutions may be offered to make this process as efficient and effective as possible. Our Open Innovation Programme Our Open Innovation Programme helps us to identify, accelerate and deploy innovative technologies to unlock the full potential of the UK North Sea. We work with the oil and gas industry, academic institutions and the wider technology community to deliver theme-based technology showcases, Tech Talks and workshops that inspire new thinking and collaborative innovation.

7 The Call for Ideas process is a key part of our Open Innovation Programme. It s our primary means of reaching out to the technology community to identify, support and fund innovative solutions. We issue a diverse range of calls throughout the year to support our Solution Centres, attract new ideas to our Innovation Hub, and identify small and medium-sized enterprises to benefit from our Technology Accelerator. All our Calls are run through our Ideas Portal. 7