Data is the New Oil, even in the Oil and Gas Industry

Data is the New Oil, even in the Oil and Gas IndustryInsights from the SPE Data Science Convention 2019Shubham TiwariBlockedUnblockFollowFollowingMay 10Data science and machine learning have been the force behind major shifts in industries like the internet, finance, marketing, among others.

The oil and gas industry has been no exception, quickly getting acquainted with the concept in the early 2010s, and applying them to extract valuable insights in the upstream, midstream and downstream sectors.

Refined data has thus remained a valuable asset to companies within the industry.

Clive Humby had famously compared data to the new oil, noting its inherent value shining through after being refined.

“Data is the new oil.

It’s valuable, but if unrefined it cannot really be used.

It has to be changed into gas, plastic, chemicals, etc to create a valuable entity that drives profitable activity; so must data be broken down, analyzed for it to have value.

”— Clive HumbyComparisons between refined data and refined oil took on a new meaning during the Data Science Convention 2019 organized by the Data Analytics group of the Society of Petroleum Engineers (SPE) in Houston, TX.

Held on April 4, 2019 to a sold-out audience of 425 professionals, the Data Science Convention captured the excitement brought on by the spread of AI and organized data analysis throughout the O&G industry.

Suri Bhat, Head of the SPE Data Analytics group, introduced the theme of the second annual DSC: “Transforming the Upstream O&G Industry with Advanced Data Science Solutions,” and emphasized the importance of incorporating AI into the oil and gas industry, given the petabytes of data each oil producer processes.

As per the 2017 World Economic Forum report, Digital Transformation would generate USD 1.

7 trillion worth of value from 2016 through to 2025, with a consequent reduction in carbon emissions by 1.

2 million tonnes.

“The estimated value for the industry due to digital transformation is about $1.

7 trillion, with a consequent reduction in carbon emission by 1.

2 million tonnes.

”The sold-out convention had a good mix of technical and business presentations, with a student/industry poster session, company booths showcasing their services, and plenty of networking opportunities.

Deep Learning in Fault and Log InterpretationsEach speaker individually presented a taste of the breadth of projects where AI is being used in currently.

Operations associated with the upstream oil industry have to do with exploration and appraisal of resources, development of fields, drilling and completions, reservoir management, and production and facility operations.

The geosciences and exploration domain are actively working with automated well logs interpretation and exploration using machine learning.

Typical Log Data (obtained from source)One particular industry project dealt with real-time formation interpretation using Bayesian State Space Models and Monte Carlo simulations.

Another project utilized AI in fault interpretation using neural networks.

Unsupervised learning has been used as well previously in well log interpretation.

Missing 3-dimensional data in seismic logs presents unique challenges in modeling synthetic data for further interpretations.

Both real and synthetic data are utilized thus for reducing the amount of time spent in realizing geoscientific conclusions.

Grain size predictions are also being analyzed using convolutional neural networks with data from micro-resistivity logs and multiple other inputs.

Real-Time Torque and Drag Calculations using Neural NetworksIn drilling, both contextual data (in the form of daily drilling log reports) and structured visual data (obtained through logging and captured in the electronic drilling recorder), need to be analyzed.

Due to the time-bound nature of drilling operations, real-time decisions need to be made.

To this end, companies are using neural networks, and analyzing rig states for real-time data visualization, and for predicting drilling key performance indicators.

Torque and drag on the drill strings in any given well can be calculated in real-time now using AI, by estimating coefficient of friction and normal contact forces between the string and the wellbore.

Another operator is utilizing historical data of pump washouts, and implementing a smartphone alert system to the rig operators if and when a washout will occur.

Bottom hole assemblies (BHAs) undergo varied stresses, especially in horizontal well sections and having the right assembly is important given the operational parameters (angle of drilling, directional and depth targets, the expected rate of penetration), formation properties (abrasiveness and competency of the rock, pressure regime in the hole) and drilling parameters (drill string weight, applied RPM range, torque and anticipated shock pattern).

AI is currently being used to select the right BHA in real-time.

Feature Extraction Models to Predict Well Production ProfilesOil and gas production optimization analysis involves time series forecasting and recurring neural networks.

Major KPIs include prediction of oil rates, and gas-to-oil ratios.

One particular project used feature extraction models to use calculated bottom-hole pressure, choke, well head temperature and neighboring wells’ data to predict daily oil rate.

Fracture parameters are also being utilized for production decline curve predictions.

Another project utilized deep learning and neural networks for pattern recognition on sucker rod dynamometer cards.

Production changes due to well interruptions can also be modeled using interrupted time-series modeling (ITS).

Dynamometer Cards (obtained from Sage Technologies, Inc.

)Tackling the Two-Headed Monster: Proof of Sound Investment Returns using AI and Democratization of DataOf the major themes that major oil producers are tackling, convincing the upper management of the long-term benefits of AI proves to be the most challenging.

This is partly because of the expansive scale of operations in the O&G industry that slows adoption of new technologies.

The amount of data produced within each department is staggering, and this is just the upstream industry!.Data types also vary greatly in terms of flat files, unstructured data, spatial and temporal data, visual seismic and logging data, data with four and five dimensions.

Introduction of data science techniques would enhance its value through bench-marking and statistical analyses, artificial intelligence, machine learning, optimization, spatial and temporal analyses.

However getting everyone on board is challenging.

Upper management needs more solid proof-of-concept studies that increase business value.

One executive remarked that 80% of all data science and machine learning projects never end up being implemented because of the state of the industry.

Of the remaining 20% that do provide business value, 85% of the time and resources are spent in properly defining the problem, and the rest is pure data science.

Management and dissemination of corporate standards in use and analysis of data have begun, and conferences enable sharing of best practices.

Despite these challenges of turning the juggernaut that is the oil and gas industry, around, numerous strategic initiatives have been taken that are streamlining and standardizing the data science process.

One specific operating company has begun to enable and support citizen data scientists through prescriptive guidance and technical oversight, and also to increase the focus on growth of AI skills broadly within the company.

Management and dissemination of corporate standards have begun, and conferences enable sharing of best practices.

However, collaboration and innovation are key, and an open dialogue from the larger industry and outside needs to be facilitated.

Sammy Haroon, CEO of AlphaX Decision Sciences, drew comparisons between the AI technology and the O&G industry.

He emphasized open source as the key to AI’s success and speed of adoption, and remarked about a greater need for democratization of data in this sector.

Opportunities are ExpansiveIn addition to the rich content of the speakers and panelists, few data companies in the upstream industry showcased their services.

An interactive poster presentation session attracted students and industry professionals not just from across the USA, but from Pakistan and Norway as well!.Some of the salient applications of AI in the industry were highlighted: caving detection analysis using computer vision and neural networks, permeability prediction using machine learning, workflow description and data wrangling procedures for application of data analytics in improving drilling operations, time series frac data for prediction of completion events, and using AI to extract and structure data from exploration and production documents, among many others.

Growth of AI in the oil and gas sector is inevitable, and hiring the right people for the jobs is paramount to company successes.

The SPE Data Science Convention 2019 provided one such valuable opportunity for the attendees to successfully network and learn about these issues that concern the industry, and develop solutions accordingly.

Cross-functional data scientists would need to be trained thus, and opportunities are expansive.

To paraphrase a senior executive at the convention: “Data in this sector is a valuable asset, and data scientists and machine learning specialists are here to stay.


. More details

Leave a Reply