2021 LiveStream Test – Oil and Gas Machine Learning Symposium

An innovative Fault Pattern Detection Methodology has been carried out using a combination of Machine Learning Techniques to produce a seismic volume suitable for fault interpretation in a structurally and stratigraphic complex field.

Through theory and results, the main objective was to demonstrate that a combination of ML tools can generate superior results in comparison with traditional attribute extraction and data manipulation through conventional algorithms.

The ML technologies applied are a supervised , deep learning, fault classification followed by an unsupervised, multi-attribute classification combining fault probability and instantaneous attributes.

The results are encouraging showing a higher level of structural details when compared with other interpretation techniques. Furthermore, visualization enhancement to better define stratigraphic relationships has also been partially achieved by the combination of fault probability volumes obtained from ML CNN fault detection procedure and multi-attribute classification of seismic features or SOM (Self Organizing Maps).

Machine Learning Technology was applied to a reprocessed seismic dataset in depth domain to generate a detailed, robust and reliable seismic fault attribute volume. The results are being used to constructing a more confident structural framework in the area and better understand stratigraphic trends and relationships to serve as input for static modeling.

Seismic fault detection is one of the critical steps in seismic interpretation. Identifying faults is crucial for characterizing and finding the potential oil and gas reservoirs. Machine learning holds promise for eliminating some of the tedious and repetitive steps in fault interpretation. Seismic amplitude data serves as input for automatic fault detection and deep learning Convolutional Neural Networks (CNN) perform well on fault detection without any human interactive work. This presentation shows an integrated CNN-based fault detection workflow which enhances the final fault detection volume by applying pre and post processing and an unsupervised seismic classification to ultimately isolate faults within a 3D volume. The pre and post processing objectives were to suppress noise or stratigraphic anomalies subparallel to reflector dip and to sharpen fault and other discontinuities that cut reflectors. To suppress cross-cutting noise as well as sharpen fault edges, a principal component edge-preserving structure-oriented filter is first applied. The conditioned amplitude volume is then fed to a pre-trained 3D synthetic CNN model to compute fault probability. Finally, a 3D Laplacian of Gaussian filter is applied to the CNN fault probability to enhance fault images. The resulting fault detection volumes (fault probability, fault dip magnitude and fault dip azimuth) compare favorably with traditional human interpretation and in complex structural settings, provide a more complete and unbiased image of faults. Finally, the fault volume is input into an unsupervised machine learning seismic classification (SOM) to generate a 3D volume in which the faults volumes are classified into discrete neurons with known values. This provides superior final results which can subsequently be used to generate geobodies of individual faults or used directly as input to other fault surface extraction tools.

As Chief Technology Officer – Energy, David is responsible for Dell Technologies’ strategy for the Energy industry. He works with partners and clients to identify business needs and leads teams to develop strategies and architectures to support these requirements. David oversees the roll out of Dell Technologies’ Energy Industry strategy and solutions within the client base globally.

Prior to working for Dell Technologies, David established a highly respected position in the Petro-technical Computing and Information Management community where he was described as “a leader in the global Information Management and Infrastructure community” where he was “highly respected for his ability to engage with clients, develop new solutions and master complex technical/business problems.”

As Information Management Practice Manager for Halliburton-Landmark, David was responsible for architecting, building and operating some of the world’s largest outsourced geotechnical information management and application hosting solutions including for Shell and PGS as well as the National Hydrocarbon Databanks for the UK, Norway and Oman.
Prior to joining Dell Technologies, David was Director of Operations for FUSE IM where he built a team to bring to market FUSE IM’s cloud-based workflow, collaboration and data management tools for petro-technical data. FUSE IM was acquired by Target Energy Solutions in 2014.

David is Dell Technologies’ representative on the OpenEarth Community Executive Committee and was recently elected to the Open Group’s Open Subsurface Data Universe Management Committee. He has previously served on a number of industry committees including the European ECIM Management Committee and SPE’s “Petabytes in Asset Management.” He has delivered numerous technical papers at conferences around the world and holds a patent for his work on the remote visualization of geotechnical applications.

APPLICATION OF UNSUPERVISED MACHINE LEARNING FOR 3D SEISMIC, PLIOCENE TURBIDITIES, OFFSHORE NILE DELTA.

Sr Field is a gas field located 100 km offshore west Nile Delta, within the West Delta Deep Marine (WDDM) concession of Shell (Rashid/ El-Burullus joint venture). The water depth ranging from 500-1000 meters.

The Field was penetrated by 4 wells, two of them are exploration wells (So-1 & Sr-1) and the other two are development wells (Sr-a & Sir-c). It is a stratigraphic Pliocene channel of 35-40 km length and 500- 100-meter width, with a clastic sand reservoir of 100-200 meter thickness. The filed is covered by 3D seismic of good quality. One segment of this channel has about 0.65 TCF of gases and three wells from the four wells are included at this segment.

The channel was conventionally mapped using Hampson Russel Spectral decomposition. Unsupervised machine learning is recently used for clustering the data and isolate the channel using the Paradise® software from Geophysical Insights.

Spectral magnitude, spectral voice, spectral phase, sweetness, energy ratio similarity, and Envelop are were identified by Principal Component Analysis (PCA) as having the highest energy in the region of interest. These attributes were then classified simultaneously over the reservoir zone using the Self Organizing Map (SOM) application, a form of machine learning.

Application of the unsupervised Machine learning using SOM clearly demonstrate the strike and geomorphology of the Pliocene marine turbidities. The southern segment of the channel penetrated with the three wells are very well defined after posting the wells. No significant difference on the neurons (hexagons) at the locations of the three wells which reflects the similarity of reservoir nature, thickness, sand content and pay thickness. A significant other channel is resolved to the east of the main channel that is not detected using the conventional spectral decomposition.

Seismic fault detection is one of the critical steps in seismic interpretation. Identifying faults is crucial for characterizing and finding the potential oil and gas reservoirs. Machine learning holds promise for eliminating some of the tedious and repetitive steps in fault interpretation. Seismic amplitude data serves as input for automatic fault detection and deep learning Convolutional Neural Networks (CNN) perform well on fault detection without any human interactive work. This presentation shows an integrated CNN-based fault detection workflow which enhances the final fault detection volume by applying pre and post processing and an unsupervised seismic classification to ultimately isolate faults within a 3D volume. The pre and post processing objectives were to suppress noise or stratigraphic anomalies subparallel to reflector dip and to sharpen fault and other discontinuities that cut reflectors. To suppress cross-cutting noise as well as sharpen fault edges, a principal component edge-preserving structure-oriented filter is first applied. The conditioned amplitude volume is then fed to a pre-trained 3D synthetic CNN model to compute fault probability. Finally, a 3D Laplacian of Gaussian filter is applied to the CNN fault probability to enhance fault images. The resulting fault detection volumes (fault probability, fault dip magnitude and fault dip azimuth) compare favorably with traditional human interpretation and in complex structural settings, provide a more complete and unbiased image of faults. Finally, the fault volume is input into an unsupervised machine learning seismic classification (SOM) to generate a 3D volume in which the faults volumes are classified into discrete neurons with known values. This provides superior final results which can subsequently be used to generate geobodies of individual faults or used directly as input to other fault surface extraction tools.

Randall (Randy) W Gentry (Ph.D., Civil Engineering, University of Memphis, 1998) has over 20 years of experience leading research teams and organizations across academia and federal research organizations within federal agencies focused on energy and environmental issues. Dr. Gentry led many of these organizations through strategic planning and on-going organizational change while targeting mission critical programmatic research needs at the U.S. Department of Energy laboratories (National Energy Technology Laboratory and Argonne National Laboratory), U.S. EPA and served as a tenured faculty member at the University of Tennessee while also leading several research initiatives. Dr. Gentry joined the Petrolern team as R&D Director in early October 2021 and is excited to work with such a strong entrepreneurial group.

Prior to joining Petrolern, Dr. Gentry served as the Deputy Director and Chief Research Officer at the National Energy Technology Laboratory, one of the U.S. DOE’s National Laboratories. At the National Energy Technology Laboratory Dr. Gentry commissioned the strategic planning and inauguration of the Science-based Artificial Intelligence and Machine Learning Institute (SAMI) and served on its inaugural advisory board.

Dr. Gentry has pursued interdisciplinary research during his career and built teams around complex subsurface behavior and phenomena to better understand aquifer mixing behavior and characterize temporal scale response to various stress induced states on those systems as predictors of risk for better management tools, and for better understanding fundamental conceptual model development of reservoir systems. Dr. Gentry has co-authored papers with many of his team members in top tier international journals and is a recognized subject matter expert in multi-layer aquifer mixing behavior. Dr. Gentry was on two teams recognized in January 2021 by U.S. DOE Secretary’s Achievement Awards, the National Virtual Biotechnology Laboratory Team and the Science and Technology Risk Matrix Team for their work performed in 2020 across the National Laboratory system.

Despite its time-honored use of technologies, the upstream oil and gas sector has been slower than other industries to embrace the breakthrough solutions that have transformed several sectors over the past decades. That may be about to change. Much has happened and COVID-19 has had multiple impacts on the oil and gas industry including plunged prices and government revenues, significantly lowered demand, and inflated stockpiles of crude oil. As the industry seeks ways to return to profitability, Next-Generation technologies are emerging as a part of the answer, presenting the possibility of a radically more efficient new reality.

The development of Next-Generation technologies has the potential to transform the upstream oil and gas industry’s fortune by sustainably improving exploration portfolio, capital efficiency, lowering exploration and development costs, and reducing time to first oil. These new technologies bring significant changes to an upstream oil industry’s ecosystem, disrupting the traditional value chain and redefining business models. We took a Next-Generation Technology Stack Lens approach to explore how these breakthrough solutions will alter the upstream industry eco-system by examining Digital Programs, Workflow efficiency and Innovative Projects. This breakthrough approach offers the structure that the upstream industry needs to understand Next-Generation Technologies.

To this end, this presentation will dive deep into PETRONAS’s NexTGEN Technology plan for exploration that is converting data to insights and taking actions based on those insights. We believe that the process of NexTGEN will be unique to each step because upstream exploration contains a discrete set of technologies such as Machine Learning, Data Analytics, and Cloud-based Solutions for a distinct flow of data, a particular set of challenges, different digital maturity, and nature of data generated and utilized in each step differs from that others. NexTGEN is set to have a profound impact on PETRONAS’ exploration delivery mandates. A wave of Next-Generation Technologies is bringing greater technological innovation that powers value creation in Oil and Gas.

Automated, Unmanned Detection and Quantification of Methane Fugitive Emissions

Compressor stations used to move natural gas are one of the largest sources of fugitive methane emissions in the midstream sector, accounting for approximately 50% of all fugitive emissions(Zimmerle et al., 2015). This problem is most widespread at reciprocating compressors (Subramanian et al., 2015) where faulty seals are a key contributor to methane emissions (Johnson et al., 2015). As such, there is a significant need for a robust technology that could provide an early indication of an unexpected emission. Equally important, the technology needs to be able to account for biogenic versus anthropogenic sources of methane. One means of indirectly making this determination is to leverage optical technologies that can autonomously pinpoint the source of such leaks.

This presentation discusses recent work funded by the U.S. Department of Energy (DOE) National Energy Technology Laboratory (NETL), focused on the development of an innovative remote sensing technology that can reliably and autonomously detect fugitive methane emissions in near real-time, using computer vision and deep learning. The technology called the Smart Methane Leak Detection (SLED/M) system was initially developed to monitor facilities such as compressor stations in a stationary, pan-tilt-zoom configuration.

The system has recently been adapted to monitor facilities from an unmanned aerial system (UAS). The speed and maneuverability of UAS platforms are attractive to leak detection and repair program operators but introduce several challenges. Many existing methane detection algorithms rely on mostly static backgrounds becoming unusable with motion. In addition, top-down views of fugitive methane emissions present differently in Optical Gas Imagers (OGI) compared to looking across the plume. Our work has focused on overcoming these challenges, enhancing the operators' ability to detect methane emissions and pinpoint their sources. Another recent adaptation to SLED/M is the ability to quantify methane emissions using passive sensors (OGI, thermal camera), environmental conditions, plume modeling, and deep learning.

SLED/M advances the state-of-the-art for methane emission detection and quantification by focusing on three key critical criteria for effective methane emission mitigation: (1) autonomy (no need for a human to be in the loop), (2) high reliability (low false alarm rates), and (3) real-time performance. Results from this work will be presented.

Time-lapse (4D) seismic analysis plays a vital role in reservoir management and reservoir simulation model updates. However, 4D seismic data are subject to interference and tuning effects. Being able to resolve and monitor thin reservoirs of different quality can aid in optimizing infill drilling or locating bypassed hydrocarbons. Using 4D seismic data from the Maui field in the offshore Taranaki basin of New Zealand, we generate typical seismic attributes sensitive to reservoir thickness and rock properties. We find that spectral instantaneous attributes extracted from time-lapse seismic data illuminate more detailed reservoir features compared to those same attributes computed on broadband seismic data. We develop an unsupervised machine learning workflow that enables us to combine eight spectral instantaneous seismic attributes into single classification volumes for the baseline and monitor surveys using self-organizing maps (SOM). Changes in the SOM natural clusters between the baseline and monitor surveys suggest production-related changes that are caused primarily by water replacing gas as the reservoir is being swept under a strong water drive. The classification volumes also facilitate monitoring water saturation changes within thin reservoirs (ranging from very good to poor quality) as well as illuminating thin baffles. Thus, these SOM classification volumes show internal reservoir heterogeneity that can be incorporated into reservoir simulation models. Using meaningful SOM clusters, geobodies are generated for the baseline and monitor SOM classifications. The recoverable gas reserves for those geobodies are then computed and compared to production data. The SOM classifications of the Maui 4D seismic data seems to be sensitive to water saturation change and subtle pressure depletions due to gas production under a strong water drive.

Nancy House, a member of SEG for nearly 40 years, joining in 1978 as a graduate student at CSM, has worked as a geophysicist for multinational corporations and small independent oil companies primarily as an interpreter on and offshore US, South America and Africa (West and East), and other areas. She is a second-generation geoscientist, growing up in South America and Singapore. She has a BA in Geology/Geophysics from the University of Wyoming, (1976), an MSc in Geophysics from Colorado School of Mines (1979), and did additional postgraduate work at Colorado School of Mines in Reservoir Characterization, Economics and Geophysics (2000-2002).

From the first SEG Annual Meeting Nancy attended in San Francisco in 1978 as a student, she knew that SEG would play an essential part in her career. Early on, SEG provided valuable training, networking opportunities and guidance in professional standards and ethics. Nancy has served on numerous committees including, GAC, Women’s Network Committee, Finance Committee, Membership Committees in SEG. She served as Denver Geophysical Society President/Past President from 2008-2010, General Chairman for SEG AM 2010, and Secretary-Treasurer 2011-2012, Chairman of SEG Women’s Network Committee 2012-2013, and the Finance Committee 2012-2014. Nancy has been a regular contributor to TLE, presenter at meetings (Best Poster 1995), a reviewer for Geophysics and a session chair for various meetings. She also served on several task forces to understand critical business issues around SEGs global activities. She has been a member of AAPG, Dallas GS, Den GS, RMAG, DivEnvirGeol(AAPG), AGU, AWG, and EAEG.

As SEG President 2017-2018, she focused on increasing diversity and inclusion in the profession of geophysics, continued strategies implemented by Rd. Bradford and Bill Abriel and recognized the social contribution of Geophysics and applied geophysics in areas additional to oil and Gas

A multi-disciplinary approach to establish a workflow for the application of machine learning for detailed reservoir description - Wisting case study

A multidisciplinary approach that is maximizing information extraction from seismic to predict lithofacies and reservoir properties, based on the following steps is presented:
Multi-attribute seismic analysis was applied based on an unsupervised machine learning process called Self- Organizing Maps (SOMs) in Paradise software. The selection of input attributes was thoroughly tested and optimized, based on close co-operation between geophysicists and geologists to extract more extensive and detailed geological features from seismic.

Using the information from nearby wells and knowledge of rock physics, the individual neural classes were quantified and validated and then reorganized and translated to formation properties such as lithofacies, porosity, and clay content.

The study focused on the benefits and additional information that can be gained with this new approach compared to traditional quantitative interpretation approaches (i.e., a prediction from acoustic impedance alone). Multi-attribute classification using machine learning, SOM, gave a better representation of seismic characters, detecting the geologic trends in the field. A detailed quantitative interpretation of SOM neural classes was established to validate and translate formation-related classes optimally for reservoir prediction, and to eliminate classes irrelevant to the formations (i.e., seismic noise).
The result from the Wisting case study shows that the new method gives the best match to the well data and extracts more reservoir related information from seismic compared to the conventional quantitative interpretation (QI) approach. In the upper part of the Triassic, with fluvial sediments assigned to the RG2 unit (Fruholmen Fm.), the reservoir quality and extent of mud clast rich channel intervals are debated. In two of the wells (E and B), a thicker mud unit was observed, and it was debated if this could act as a barrier towards the overlying good reservoir. The result shows that the mud intervals are deposited locally and do not represent a regional mud layer. The additional information from seismic seems to be valuable when used as input and refinement to the digital geological model.

LiveStream Test

Welcome! Please see the schedule of talks with links to join below.

Day 1 - Tuesday, 26 Oct

08:15 – 08:45 U.S. CDT UTC-5

Five Trends in Machine Learning for Geoscience Interpretation

09:00 – 09:45 U.S. CDT UTC-5

Innovation and Diversity Impact for Artificial Intelligence Success in Applied Geosciences

10:00 – 10:45 U.S. CDT UTC-5

A Combined Deep Learning and Unsupervised Machine Learning Fault Detection Workflow

11:00 – 11:45 U.S. CDT UTC-5

Challenges and Opportunities for Machine Learning in Upstream Applications

Day 2 - Wednesday, 27 Oct

09:00 – 09:45 U.S. CDT UTC-5

Next-Generation Technologies Powers Value Creation in Oil and Gas

10:00 – 10:45 U.S. CDT UTC-5

Building Robust Data-Driven Machine Learning Models for Geologic Carbon Storage (GCS) Applications: Are We There Yet?

11:00 – 11:45 U.S. CDT UTC-5

Unsupervised Machine Learning for Time-Lapse Seismic Studies and Reservoir Monitoring

Day 3 - Thursday, 28 Oct

09:00 – 09:45 U.S. CDT UTC-5

From Subsurface to Surface in 2 AI Products

10:00 – 10:45 U.S. CDT UTC-5

Fifth Wave of AI and Fusion of Innovation-centric Energy Industry Future

11:00 – 11:45 U.S. CDT UTC-5

Machine Learning Fault Detection: A Case Study

Day 4 - Friday, 29 Oct

09:00 – 09:45 U.S. CDT UTC-5

RoundTable: Machine Learning Applications Today

10:00 – 10:45 U.S. CDT UTC-5

RoundTable: Machine Learning Applications Tomorrow

Completion Analytics, Utilizing Descriptive and Predictive Analytics for Optimizing Shut in Strategies of Adjacent Wells During Multi-Stage Hydraulic Fracturing Operations

TRANSFER LEARNING FOR SUBSURFACE FAULT AND SALT INTERPRETATION

OVERCOME SILOED APPROACH FOR OPTIMIZATION TO IMPROVE END-TO-END PRODUCTION WITH A PREDICTION-OPTIMIZATION FRAMEWORK

TRENDS AND APPLICATIONS OF COGNITIVE COMPUTING IN GEOSCIENCE AND PRODUCTION OPTIMIZATION

DELIVERING IMPACTFUL AI INNOVATION IN UPSTREAM

REAL-TIME FLUID TRACKING: THE MISSING LINK FOR FRAC MODELING MACHINE LEARNING FRAMEWORKS

A TALE OF TWO RESERVOIRS: HOW MACHINE LEARNING CAN HELP DEFINE “SWEET SPOTS” IN CONVENTIONAL AND UNCONVENTIONAL RESERVOIRS

APPLICATION OF UNSUPERVISED MACHINE LEARNING FOR 3D SEISMIC, PLIOCENE TURBIDITIES, OFFSHORE NILE DELTA.

SUBSURFACE DIAGNOSTIC ANALYTICS: WHAT COULD MACHINE LEARNING TELL YOU ABOUT FRAC HITS?

TABIO: AN OPEN-SOURCE TOOLKIT FOR DETECTING AND SEGMENTING TABLES FROM UNSTRUCTURED DOCUMENTS WITH MACHINE LEARNING

AI TREND IN OIL & GAS

ACCELERATING SEISMIC INTERPRETATION TASK

FINDING THE BEST ATTRIBUTE COMBINATION FOR SEISMIC FACIES CLASSIFICATION

DESCRIBING THE RESERVOIR: SEISMIC MACHINE LEARNING AND DATA ANALYTICS

ARTIFICIAL INTELLIGENCE: A KNOWLEDGE-ENHANCED MACHINE LEARNING APPROACH

MAKE DEEP LEARNING ACCESSIBLE TO SEISMIC INTERPRETERS

STATISTICAL CALIBRATION OF SOM RESULTS WITH WELL LOG DATA (CASE STUDY)

ACCELERATING SHELL’S SUBSURFACE TRANSFORMATION THROUGH MACHINE LEARNING

MACHINE LEARNING IN OIL & GAS - MOVING FROM HYPE TO REALITY

Automated, Unmanned Detection and Quantification of Methane Fugitive Emissions

Gas Hydrates, Reefs, Channel Architecture, and Fizz Gas: SOM Applications in a Variety of Geologic Settings

DIGITAL DEMOCRACY TO SCALE DIGITAL TRANSFORMATION

ARE YOU ON THE EDGE?

THE IMPACT OF MACHINE LEARNING ON OIL AND GAS SERVICE INDUSTRIES

INTELLIGENT APPROACH FOR GOSP OIL RECOVERY ENHANCEMENT

THE UNCONVENTIONAL REVOLUTION IN EXPLORATION GEOPHYSICS

GO BEYOND DASHBOARDS: DIGITAL TWINS FOR OIL & GAS DRILLING OPERATIONS

MACHINE LEARNING APPLICATIONS AND USE CASES IN RESERVOIR ENGINEERING

A multi-disciplinary approach to establish a workflow for the application of machine learning for detailed reservoir description - Wisting case study

THE HUMAN FACTOR: KEY TO SUCCESS IN MACHINE LEARNING

08:15 – 08:45 U.S. CDT
UTC-5

09:00 – 09:45 U.S. CDT
UTC-5

10:00 – 10:45 U.S. CDT
UTC-5

11:00 – 11:45 U.S. CDT
UTC-5

09:00 – 09:45 U.S. CDT
UTC-5

10:00 – 10:45 U.S. CDT
UTC-5

11:00 – 11:45 U.S. CDT
UTC-5

09:00 – 09:45 U.S. CDT
UTC-5

10:00 – 10:45 U.S. CDT
UTC-5

11:00 – 11:45 U.S. CDT
UTC-5

09:00 – 09:45 U.S. CDT
UTC-5

10:00 – 10:45 U.S. CDT
UTC-5