Sei sulla pagina 1di 14

SPE-193776-MS

The Role of Big Data Analytics in Exploration and Production: A Review of


Benefits and Applications

Christine I. Noshi, Ahmed I. Assem, and Jerome J. Schubert, Texas A&M University

Copyright 2018, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE International Heavy Oil Conference and Exhibition held in Kuwait City, Kuwait, 10-12 December 2018.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Due to the decrease in commodity prices in a constantly dynamic environment, there has been a constant
urge to maximize benefits and attain value from limited resources. Traditional empirical and numerical
simulation techniques have failed to provide comprehensive optimized solutions in little time. Coupled with
the immense volumes of data generated on a daily basis, a solution to tackle industry challenges became
imminent. Various expert opinion fraught with bias has posed extra challenges to obtain timely cost-effective
solutions. Data Analytics has provided substantial contributions in several sectors. However, its value has
not been captured in the Oil and Gas industry.
This paper presents a review of various Machine Learning applications in exploration, completions,
production operations to date. An overview of data-driven workflows in the fields of electric submersible
pump (ESP) failure and shutdown prediction, reservoir databases’ analysis, reduction of subsurface
uncertainty, EOR decisions using scarce data, improved oil recovery estimation, production impact
assessment, horizontal completion, fracturing techniques, production optimization in unconventional
reservoirs, production management, and field surveillance, is presented.
The review attempts to shed light on the benefits and applications of multiple challenges faced on a daily
basis by scientists, field personnel, and engineers to help solve and optimize the industry's multi-faceted
data-intense challenges.

Data Analytics and Artificial Intelligence


Computational intelligence (CI) or Machine Learning are two sides of a coin notorious for Artificial
Intelligence (AI). AI comprises of the employment of complex algorithms and networking tools to unravel
multidimensional problems mimicking human brainpower. AI uses science and engineering to allow
machines and computers to perform demanding computational tasks that require arduous human thinking
abilities. The designed models learn through continuous repetitive operations without depending on "rules-
based programming" thus constantly refining the models’ computing abilities with the intake of data
feed. Despite its basis on classical statistical methods from the early 1700s, its robustness originates from
contemporary computational facilities and massive data bases attained from present technology (Pyle and
San Jose 2015). ML has been gaining more acceptance and popularity for tackling case-based challenges
2 SPE-193776-MS

and complex problems in numerous applications. The produced models seize the ambiguity in real life
cause-and-effect situations, thus, integrating obtainable information with probability inferred computations.
(Xia et al. 2013). Especially, in the sectors of health care, economics, finance, banking, engineering, tech,
advertising, military, and marine (Mellit 2008). However, the petroleum industry has fallen behind these
disciplines in the adoption of this discourse. According to Bravo et al. 2014, only 13% of participants were
familiar with and actively involved with the technology in a current SPE survey. In spite of this lagged
adoption, there has been a noteworthy upsurge in ML research within the industry in the last few years.
Gharbi and Mansoori, 2005, attributed this rapid increase in the quantity of AI applications can be attributed
to the increase of human expert knowledge and plentiful publications of effective case studies.
It is undeniable that Big Data performs a progressively vital role in the E&P value chain. With the
continuous acquisition of real-time data, huge volumes of data are being assembled from drilling, logging
operations, seismic surveys, and production. Accordingly, valuable information is currently accessible
from high frequency surface and downhole sensors. Web-based monitoring pipelines and platforms for
instantaneous surveillance of producing assets are being deployed. These assembled workflows support
field personnel and engineers with their day to day monitoring and surveillance operations for reservoirs,
production systems, wells, and various fields. Not to mention, helps faster team collaborations for faster
decision making. The entirety of collecting, processing, and analyzing metadata is known as analytics
(Bravo et al. 2014). Data Analytics allows the extraction of maximal value from data through trend
identification that permit for real-time discovery of imminent failures, early prevention of tool failure,
production forecasting, and risk quantification of assets. Nevertheless, with these benefits comes an
associated problem of mining actionable information to enhance oil and gas production. The overwhelming
data can be an impediment for engineers when it comes to model optimization.
Noshi and Schubert 2018 discussed some of the most common ML algorithms used in the oil and gas
industry. The most prominent benefits of AI techniques is emphasized below (Medsker 1996; Mellit 2008;
Tu 1996; Benghanem 2012):
1. The capability of creating non-linear and complicated correlations and models between input and
output features without prior assumptions or simplifications.
2. AI leverages several human cognitive traits, in a variety of disciplines, for instance, identification and
diagnosis, monitoring and selection, prediction, and inspection.
3. Using several different training algorithms, a variety of ML models can be established.
4. AI handles extremely large and variable databases with great reliability to produce valuable and
relevant correlations and patterns with the presence of noisy and incomplete data.
5. The models can be trained to handle repetitive and cumbersome tasks faster with fewer errors than
human intervention.
6. For linear and/or non-linear multiple regression predictions, AI models display more accuracy
compared to other physical and empirical models.
7. Owing to their execution speed, ML tools are cost effective in the short and long-term. Efficient
system training saves time on long coding programs. Not to mention, flexibility during change to
accommodate various problem types.
8. AI approaches can implicitly discover complex nonlinear correlations between dependent and
independent features.

Machine Learning Applications


A sizeable quantity of trials and research studies were executed using ML tools in the petroleum industry.
Data driven techniques enables an intermediate pathway to the engineers searching for a quick robust
verification of well performance in a matter of seconds. The technology can address and deal with
complicated technical conundrums on a level that was formerly considered far-fetched to implement on
SPE-193776-MS 3

a traditional simulation scope (Mohaghegh and Khazaeni 2011). This can be attributed to the unique
characteristics exhibited by AI methods which showed supremacy in comparison to traditional methods.
This includes comparable human cognitive and problem-solving abilities characterized in continual model
training and generalization for effective pattern recognition. As well as, function modelling, classification,
and filtering noise from data. Supervised ML or regression can encode data iteratively. Besides for
complicated dynamic systems, processing of oncoming data can be performed in conjunction with data
distribution. However, ML models such as Artificial Neural Networks (ANN) prediction is certainly not a
substitute for production forecasting using numerical or empirical simulation. Rather a hybrid approach of
ANN modelling can be exploited to deliver further confidence to these traditional forecasting approaches.
Fuzzy logic, ANN, and GA are amongst the most widely implemented algorithms in the oil and gas upstream
segment for over sixteen years to date, with the pioneering application of ANN implemented back in 1989
for the intelligent advancement of well-log interpretation, a smart interface for reservoir simulator, and drill
bit analysis. Other techniques include Support Vector Machine (SVR), cased based reasoning, and functional
networks (Bello et al. 2016; Braswell 2013).
Softwares with foundations from the aforementioned AI tools have been anticipated to shorten the gaps
in technology hampering intelligent fully automated monitoring and execution of key industry applications
such as recovery factor prediction, interpretation of well logs, failure prediction and diagnosis of pumps such
as ESP and sucker rod pumps, simulator interface development, intelligent HSE, reservoir characterization,
location selection of reservoir development, optimization and analysis of hydraulic fracturing design, reduce
subsurface uncertainties, well testing interpretation and selection of models, estimation and management
of IOR and EOR risk, optimization of gas-lift, PVT correlations, identification of best industry practices,
well design and drilling operations optimization, analysis of formation damage, pipe fluid flow analysis,
optimizing production in unconventional reservoirs, maintenance of production facilities and pipelines,
permeability reduction and sealing in oil and gas reservoirs.
ML techniques has been successfully executed on completion and production-associated issues. Field
surveillance and production management were amongst the two pioneering domains utilizing Big Data.
In their research, Nande, 2018, developed a novel methodology using ANN to forecast fracturing closure
pressure using mini-frac data for a given well to reduce the subjectivity concerns associated with analytical
models. Kellogg et al., 2018, developed a Logistic Regression model for the removal of wellbore damage
in the Wilmington Field using more than 700 producers and 400 injectors to examine and screen the
representative characteristics linked with small acid volume maintenance success at a pre-determined oil
price/bbl. The model achieved a 71% prediction success rate for the producer jobs and 77% for the injector
treatments to successfully predict whether a producer acid maintenance job would be a cost-effective
candidate. The study used acid volume and type, mechanism of delivery, formation damage, reservoir and
surface facility properties, completion design, and production/injection history. An ML model was used to
predict ESP lifespan in Devon wells across the Delaware Basin. They used several models and found that
high performance Random Forests (HP Forest) was the best based on average square error. It predicted
ESP lifespans which were within five days of the true ESP lifespan. Using electrical and frequency data
as input features, Guo et al., 2015, developed an SVR workflow to forecast failures in ESP (Sneed 2017).
Gupta et al., 2016, deployed a hybrid mathematical method consisting of an intelligent predictive monitoring
KPI that automatically identifies imminent glitches, diagnose root causes, and prescribe corrective actions
to abnormal ESP operational situations in real time raising alarms through predictive, diagnostic, and
prescriptive analytics. The pattern was captured before and during the trip or failure ranking twenty-two
attributes of discharge and intake pressures, and pump head. As well as motor features such as characterized
in motor temperature, leakage current, and vibration. The output resulted in a visual representation to
authenticate the success or failure the of remedial action taken. Moreover, identify the parameters’ safe
normal operating ranges and consequently develop event scenarios for autocorrection without manual
input using Hotelling T-square and Principal Component Analysis; counter to the features obtained in
4 SPE-193776-MS

unstable intervals calculated at each time step for the normal data. Raghavenda et al., 2013, described SVR
models’ execution in the prediction of failure in sucker rod pumps. Al-Fattah et al., 2001, used ANN to
forecast US natural gas production till 2020 and predict the supply of natural gas from 1999 throughout
the year 2020. Kalam et al. 1996 implemented backpropagation ANN to anticipate core sample saturations
and produce relative-permeability curves mimicking experimentally-generated outcomes. Nikravesh et al.,
1996, implemented ANN to forecast formation damage in tight fractured reservoirs. GA was employed for
the characterization of reservoir heterogeneity by Guerreiro et al., 1998. Sen et al., 1995, employed heat-
bath and genetic algorithms for forecasting the production of stochastic permeability fields. The authors
investigated formation damage caused by drilling fluid invasion. History matching was employed from
offset wells with similar lithology in combination with ANN to estimate production of an oil well along with
a performance prediction model of an un-drilled well using, geological maps, pressure and production data
taking into account operational limitations (Cao et al. 2016). Pankaj et al., 2018, aimed to develop a real-time
predictive model on wells from the Eagle Ford to accomplish sensitivity analysis on optimized completion
design parameters, including proppant loading, cluster, and well spacing, in a matter of hours. As well as
forecast production and Net Present Value (NPV) using the gradient boost algorithm. The authors created
"synthetic big data" from simulations based on rock physics and fluid dynamics to generate a substitute
model for the performance of hydraulic fracturing jobs in unconventionals with 90% accuracy compared to
actual outcomes. Vertical seismic data was used to predict density logs (Artun et al. 2005). Bowie, 2018,
presented a multiple Linear Regression and ANN workflow to optimize well design and completions on
262 wells in the Duvernay shale extending from volatile oil to dry gas using several completion variables.
The model successfully projected 78% of well performance variation on 52 blind test wells. Via parameter
optimization, possible performance improvements reached from 19% to 97%. The model suggested against
using ceramic-resin coasted proppants along with simplifying pad design which allowed cost savings of
over a million dollars per well and a 50% enhancement in the performance of the well. The model suggested
that total fracture tonnage and fracture pump rate were crucial to well performance optimization. Shadravan
et al., 2015, designed drilling fluids using Machine Learning algorithms. The use of Decision Trees, Logistic
Regression, and supervised Hierarchal Clustering was studied in the prediction of casing failures (Noshi
et al. 2018). Anderson et al., 2016, matched physics-based models with Data Mining practices using high
frequency data in the Marcellus shale to classify hydraulic fractures. Bangert et al., 2017, conducted smart
condition monitoring. The authors’ methods helped prevent unnecessary generation of false alarms while
firing alarms during abnormal equipment failure states or anomalous operating ranges. Mohaghegh et al.,
2011, analyzed shale production with the aid of Data Mining practices. Alvarado et al. 2002 used Machine
Learning algorithms to screen the best candidates for EOR/IOR enhancements on several mature fields in
Venezuela. An optimized velocity auto-picking approach using semblance-based and AVO auto-picking
was employed on large seismic data volumes using seismic attributes with an unsupervised clustering
algorithm to reduce velocity picking time and consequently the associated processing costs. This modified
attribute grouping created a greater separation power. The approach proved successful on two very different
marine and land datasets (Smith 2017). Zhang et al.,1991, implemented ML models in rock mechanics to
predict the different behaviors of rocks and pinpoint the chief parameters ruling its behavior. Hoeink and
Zambrano, 2017, successfully used Logistic Regression to classify shale vs. non-shale formations in the
Huron formation using density, gamma ray, resistivity, compressional and shear slowness logs. To mitigate
time-consuming and subjective local calibrations, the authors attempted to automate the process of picking
shale.
Andrianova et al., 2018, used algorithms based on ML methods to process large volumes of data and
restore the missing values. They applied this approach in well logging, PVT data, well interventions, and
predicting hydraulic fracturing performance. In, 1991, Brown used ML approach as a predictive tool in
primary production for 675 reservoirs. He also predicted that the technique can be used in all types of
databases throughout the O&G industry. Akinnikawe et al., 2018, utilized different ML methods including,
SPE-193776-MS 5

Linear Regression, ANNs, Decision Trees, Gradient Boosting and Random Forest to generate synthetic
photoelectric (PE) and unconfined compressive strength (UCS) log. Bestagini et al., 2017, designed an
algorithm to classify facies based on wireline logging measurements using a small training set. It uses
a random forest classifier fed with a set of augmented features. They have validated it on a set of ten
wells and with a blind test performed on a previous publication by (Hall and Hall 2017) to confirm
the capability of the algorithm to generalize to new data. Putcha and Ertekin, 2018, developed a hybrid
approach that integrated ML with computational algorithms to expedite the numerical reservoir simulation
and wellbore hydraulic processes. The authors validated their approach by comparing their ML results
with the complete numerical model using single and two-phase sets. They also did a comparative gas
lift case study and found that their hybrid simulation protocol showed only approximately a 5% mean
deviation in comparison to the comprehensive numeric model for the full array of gas lift injection rates in
a much shorter time. Amr et al., 2018, used ML algorithms for to forecast monthly production horizontal
wells in an unconventional reservoir. The models attempted to predict both non-producing (NPLs) and
producing well locations (PLs). They compared outcomes with conventional Arps's decline curve analysis
where the model outperformed Arps's-based prediction for both NPLs and PLs. Li et al., 2014, devised
a model with vector machine for early warning detection of production decline. The model was applied
to twelve reservoirs in Shengli oilfield. Variations in abnormal production were successfully spotted via
production history analysis. Good accuracy was achieved paving possibility for future for oil production
monitoring. Belozerov et al., 2018, used ML for automatic well logs interpretation testing (predicting
net pays) where data was divided into three samples: training, validation, and test. They concluded the
efficient applicability of ML algorithms and that the use of several variants of interpretation of well log
data allowed using all the accumulated experience of well logs interpretation. Klenner et al., 2018, used ML
approach is presented to characterize the root causes of frac hits (infill-parent-well communication). Data
used for analysis include: microseismic, completion, surface and bottomhole pressure, tracers, production,
and petrophysical logs. The results showed that strong communication occurred when the infill wells
were completed in close proximity to the parent wells exhibit. So, visualization and real-time analysis
could be used to mitigate communication during these completions. Pandya et al., 2018, presented an ML
approach to determining anomalous behavior and predict potential unintended loss of uptime (trips) in
gas compression systems and probable root cause with sufficient warning to allow for intervention. The
detection time was improved which could make investigation and remedial action possible. Pennel et al.,
2018, used ML algorithms to diagnose and predict artificial lift problems including tubing and pump failure
from operating wells in the Bakken. The failure models were highly accurate in diagnostics which could
reduce downtime and improve pump effectiveness. Tarrahi et al., 2015, proposed an algorithm that could
accurately decide the appropriate enhanced oil recovery technique (steam injection, combustion, miscible
injection of CO2 and N2) to be used based on reservoir and fluid characteristics (permeability, depth, API,
and viscosity). It also predicted the probability of success of each EOR method. Fulford et al., 2015,
proposed an ML method to predict production in unconventional, liquids-rich and gas shale wells. The
implementation of both supervised ML and well-calibrated bias improved the estimation of uncertainty of
the distribution of forecasts. Alkamil et al., 2018, designed an unsupervised learning software improving
the drilling crew ability to swiftly detect H2S by training an H2S monitoring simulator which can analyze
and then adjust drilling fluid pH to neutralize free H2S present in it. Gaganis et al., 2012, proposed using
ML methods to solve the phase equilibrium problem. Using proxy models provided direct answers of the
phase equilibrium problem which reduced simulation's total CPU time significantly. Tian and Horne, 2015a;
2015b, applied three different ML techniques to flow rate and pressure data interpretation. Even with a
complex actual physical model and less than perfect knowledge of it. The author showed that Kernel-
Ridge Regression successfully construed pressure transients from Permanent downhole gauges (PDG).
Pressure analysis was also extended from single-well to multiwell systems and the model was capable to
6 SPE-193776-MS

capturing these interactions accurately. Moreover, flow-rate reconstruction was tested on both synthetic and
real data showing good performance. Nwachukwu et al., 2018, developed an ML-based proxy to quickly
predict reservoir responses to changes in location and control parameters during WAG injection. The model
was very accurate, and results were comparable to those produced with full reservoir simulator. Voleti et
al., 2017, showed that training with bigger dataset can confuse ML algorithms and that using selective
intelligent inputs (quantitative factors influencing permeability) with proper segregation provided better
permeability prediction. Sidaoui et al., 2018, used an ML model to predict the pore-volume-to-breakthrough
(PVBT) and the optimal injection rate used in successful acid job design with great accuracy. They claimed
the model could significantly reduce the laboratory experiments needed to obtain the PVBT curve and
that it can also be applied to field scale. Ramirez et al., 2017, used ANN model to forecast PVT fluids
characteristics such as bubble point pressure and oil formation volume factor as a function of oil specific
gravity, solution gas-oil ratio, gas specific gravity, and temperature. The results from their model had a
better accuracy estimation compared to the published empirical correlations. Honorio et al., 2015, presented
an ML method that learns former geological information, and then rebuilds a model to avoid distortion
due to reduction in the number of parameters used in Assisted-History-Matching (AHM). Gaurav, 2017,
used ML combined with local geological information to detect the well parameters pertaining to production
enhancement. Well data, for a certain area, were used in a manner that identify the type-curve and EUR/
well identifying its true economic potential. Khan et al., 2018, used multiple ML techniques to precisely
determine water saturation in carbonate reservoirs using traditional wireline logs as input and dean-stark
parameters as output attributes. Meek et al., 2017, used a model based on prestack inversion and supervised
neural networks that proved effective in creating rock property volumes such as Young's modulus, shear and
acoustic impedance, p-wave and s-wave velocities, Poisson ratio, critical strain, and brittleness. Depending
on lateral facies change, a multi-well low frequency model was proposed instead of single well model.
Xiao and Sun, 2017, integrated and established relationships between various data sources such as down
hole pumps, reservoir, wellbore, and production data, along with injection, surface gathering, and geo-
systems in the Daqing oilfield to optimize input data parameters for use in reservoir simulation of a polymer
surfactant flooding treatment. Injection variables were attuned accordingly which included the optimal
artificial lift method, well pattern and space, choke and chemical slug size. The following ML approach
reduced the prediction error in excess of 46%. The model provided continuous reservoir monitoring and
storage in a chronological manner. As a result of more accurate simulation and injection parameters, water
cut, and oil production rates considerably improved. Wang et al., 2011, used GA to solve a reservoir/
wellbore flow model to get the global optimal on discrete functions. To optimize perforation parameters,
a low perforation density in high-permeability sections and high perforation density in low-permeability
intervals was achieved. The strategy helped in homogenizing and idealizing the inflow-velocity profile
along horizontal intervals which helped prevent early water breakthrough and rapid water-cut increase.
Optimization algorithms were implemented for defining an optimized reservoir performance to increase
oil recovery in the Zubair formation in South Rumaila Oil Field. The location and count optimization
the of infill wells possessing maximum cumulative oil production and NPV was determined by Adaptive
Genetic Algorithm (AGA) using Fuzzy Logic (FL) for successful prediction. Cumulative oil production
from reservoir simulation was used as an input parameter (Al-Mudhafer and Alabbas 2012). Te Stroet et al.,
2017; Schuetter et al., 2015, both used ML to choose locations for reservoir development. Roy et al., 2013,
have concluded that PCA in conjunction with Kohonen Self organizing maps can deliver classification of
accurate facies and variations in lithology in 3D seismic data. Moreover, unsupervised learning techniques
has the potential to reveal features that could otherwise go unnoticed. Mukhanov et al., 2018, presented
various kernels to optimize SVM to automatically survey and detect water control Chan diagnostic plots
and patterns for timely discovery and control of excessive water production across several fields. Santos
et al., 2015, introduced production pipeline and facility maintenance based on various constraints using
ML. Romero et al., 2000, performed reservoir characterization coupling the merits of the pilot point
SPE-193776-MS 7

technique of petrophysical property description with a modified Genetic Algorithm for global optimization
using a complicated genome to achieve seven chromosomes for various reservoir parameters using unique
crossover and mutation operators. The author also compared this method against a Simulated Annealing
method. Tang and Spikes, 2017, quantified the uncertainty in the identification of six various mineral phases
using a shale specimen. A quick segmentation practice was developed using digitized Scanning Electron
Microscopy (SEM) images as input to ANN. The model then identified and chose the highest probability of
a pixel belonging to a specific class. Zhao et al., 2017, utilized a semi-supervised learning (SSL) technique
and seeded k-means algorithm to sort four different types of oil remaining oil over a period of high
water cut. The input was collected from micro-modelling of 2D etched glass quantifying micro-residual
oil and pore structure using PCA for dimensionality reduction. Alimkhanov and Samoylova, 2014, used
eleven approaches based on six algorithms such as ANN and Decision Tree to select criteria for hydraulic
fracturing candidates on reservoirs with complex geology in the Povkh Oil Field. Both classification and
regression models were used to predict the performance of post-frac jobs as well as estimate the influence
of different geological and operational constraints hydraulic treatment efficiency. Ani et al., 2016, reviewed
the developing technologies and present practices characterizing the modelling workflows of reservoir
uncertainty and investigated the confidence of AI applications such as GA, ANN, and Bayesian networks
reservoir parameter analysis. Zhong et al. 2017, proposed a workflow simulating a reservoir with no need
for simulation. The authors used partial differential equations for training the Least Square Support Vector
Machines (LS-SVM) model. ANN was employed to adjust the LS-SVM algorithm to problem of single-
phase flow. Martin et al., 2017, developed a two-step automated workflow for predicting production of a
thermal EOR field using ML methods. The first step forecasted supplementary field measurements which
then utilized these predicted field parameters to estimate production. The second step helped automate
data interpretation. Liu et al., 2015, developed a generalized framework applicable to an entire field to
remotely detect early failures and anomalies in rod pumps using fourteen features from historical data.
AdaBNet used boosting to learn multiple Bayesian Network models, then combined them using dissimilar
weights on 426 wells. The algorithm detected failures with a 90% accuracy. Guitton et al., 2017, emphasized
faults in 3D seismic volumes using an automatic fault-picking method to label faults. Feature vectors
for training and classification were executed using Histograms of Oriented Gradients (HOG) and Scale
Invariant Feature Transforms (SIFT). Seismic data was trained and classified using SVR with Gaussian
kernels. The models helped decrease the rate of false positive, thus generating clearer fault images. The
approach successfully predicted faults in synthetic and field data even with using mislabeled data. Cheung
et al., 2017, developed a kernel regression model to choose the optimal candidates for steam treatments
using 3200 and 800 steam jobs for training and testing respectively. The model outperformed the wells
selected by engineers and realized 176% production increase in comparison to 124% attained by experienced
engineers. Subrahmanya et al., 2014, reviewed several pattern recognition semi-supervised and active
learning algorithms to observe their effectiveness in large data set labeling and asset management scenarios.
Gaganis and Varotsis, 2012, reduced substantial CPU time when solving a two-phase equilibria problem.
Equilibrium coefficients were determined by regression models using temperature, feed composition, and
pressure. Model development is automated with quick and offline execution. Proxy models provides
succinct solver answers thus saving iterative CPU charge and time. El-Sebakhy et al., 2007, described
an SVM modelling framework for accurate PVT correlation prediction. The authors also performed a
comparative study between the performance of the novel approach and nonlinear regression, ANN, and
the empirical correlations equations. Initial prediction results showed SVR outperforming the rest of the
algorithms. Xuefei and Mohanty, 2003, estimated capillary pressure and relative permeability using GA
which possess better convergence in optimization compared to the traditional Newton method. A novel
coding technique was established to perform piece-wise spline interpolation. To test the algorithm reliability,
experimental and simulation data were compared. For lower drainage injection rates, an analogous result
between the AI model and both the simulation and experimental results was obtained. Ma et al., 2016,
8 SPE-193776-MS

proposed a framework to recognize shale heterogeneities in SAGD reservoir founded on production time-
series data with decline patterns. ANN was created to train an association between input production pattern
attributes and the resultant output geologic parameters characterizing the shale heterogeneities. The ANN
models were validated with the aid of several synthetic reservoirs with known shale distributions. The model
was found to reliably estimate the desired stochastic shale parameters and its accompanying uncertainties,
while successfully forecasting the subsequent production responses. Odi and Nguyen, 2018, used ML/deep
learning to forecast classifications of geological facies using computed tomography (CT). The machine
was able to capture the correlation between the CT scan physics and geological facies to predict successive
facies with the help of CT scan parameters. Additionally, unsupervised ML methods display promising
geological facies clustering results. Luo et al., 2018, built an ANN model to identify optimal well location,
completion strategy, and predict production using 2,000 horizontal wells in the Bakken shale. The authors
used the model to scrutinize the comparative significance of both the completion and geological variables
on first-year well production results. Onwuchekwa, 2018, investigated the performance of Collaborative
Filtering, Kernel Ridge Regression, KNN, Random Forest, SVR, and Adaboost in the prediction of reservoir
fluid properties from a dataset composed of 296 oil and 72 gas reservoirs from the Niger Delta. The
models gave accurate results for the bubble and dew point pressure, oil formation volume factor, and oil
viscosity prediction. Li and Han, 2017, forecasted single well production against field production data
in conventional and unconventional reservoirs using PCA and a sensitivity study to select key attributes.
Input parameters included formation pressure and temperature, permeability and porosity, in addition to,
fracture conductivity, half length, and width. An inversion method is combined with a decline curve model,
Logistic Growth Model (LGM), to achieve some decline curve parameters. ANN was employed to inspect
the correlations and patterns of the above-mentioned attributes with a Mean Squared Error (MSE) of 0.013
Mscf/D and an overall R value of 0.917. Al-Mudhafer and Shaheed, 2011, used two different approaches
of GA used to optimize the location and number for determining the optimal future reservoir performance
of infill wells for increasing oil recovery. The model mainly relied on cumulative oil production generated
from reservoir simulation using NPV for economic investigation. Miller et al., 2017, developed a successive
sampling pipeline for History Matching with a unique metric to improve sequential sampling solutions and
validate its supremacy to the traditional L2-norm metric.

Conclusions
The aim of this paper is to present a review concerning the myriad of different applications and benefits of
Data Analytics, Artificial Intelligence, and Machine Learning techniques in the disciplines of Exploration
and Production. The paper compiles the major workflows and achievements of the industry on a higher-level
overview with a focus on its leverage over other traditional modelling techniques. A summary of several
case studies and reports developed in the upstream sector has been demonstrated. The technology's benefits
and characteristics is also summarized. Data Mining techniques will continue to enhance solution reliability
and producing robust forward forecasting. With the great monitoring, diagnostic capabilities, predictive
capacities, and optimizing flexibility AI techniques have established, it is evident that the industry will reap
great future benefits in the overall operational efficiency and hence achieve greater profitability.

Nomenclature
AI Artificial Intelligence
AGA Adaptive Genetic Algorithm
AHM Assisted-History-Matching
ANN Artificial Neural Network
BPNN Back Propagation Neural Network
CT Computed Tomography
SPE-193776-MS 9

CI Computational Intelligence
EOR Enhanced Oil Recovery
E&P Exploration and Production
FL Fuzzy Logic
FR Fuzzy Reasoning
GA Genetic Algorithms
GRNN General Regression Neural Network
HOG Histograms of Oriented Gradients
KNN K Nearest Neighbors
LGM Logistic Growth Model
LS-SVM Least Square Support Vector Machines
MLR Multiple Linear Regression
ML Machine Learning
MSE Mean Squared Estimation
NPV Net Present Value
Npls Non-Producing Well Locations
PDG Permanent Downhole Gauges
PVT Pressure-Volume-Temperature
PE Synthetic Photoelectric
PCA Principal Component Analysis
Pls Producing Locations
PVBT Pore-Volume-To Breakthrough
SVM Support Vector Machine
SSL Semi-Supervised Learning
SEM Scanning Electron Microscopy
SIFT Scale Invariant Feature Transforms
SAGD Steam-assisted gravity drainage
UCS Unconfined Compressive Strength

References
Akinnikawe, O., Lyne, S., and Roberts, J. 2018. Synthetic Well Log Generation Using Machine Learning Techniques.
Presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Houston, Texas, USA, 23–25
July. URTEC-2877021-MS. https://doi.org/10.15530/URTEC-2018-2877021.
Al-Fattah, S. M. and Startzman, R. A. 2001. Predicting Natural Gas Production Using Artificial Neural Network. Presented
at the SPE Hydrocarbon Economics and Evaluation Symposium, Dallas, Texas, USA, 2-3 April. SPE-68593-MS.
https://doi.org/10.2118/68593-MS.
Al-Mudhafer, W. J. M. and Alabbas, M. 2012. Application of a Hybrid System of Genetic Algorithm & Fuzzy Logic as
Optimization Techniques for Improving Oil Recovery in a Sandstone Reservoir in Iraq. Presented at the SPE Latin
America and Caribbean Petroleum Engineering Conference, Mexico City, Mexico, 16–18 April. SPE-149982-MS.
https://doi.org/10.2118/149982-MS.
Al-Mudhafer, W. J. M. and Shaheed, M. 2011. Adopting Simple & Advanced Genetic Algorithms as Optimization Tools
for Increasing Oil Recovery & NPV in an Iraq Oil Field. Presented at the SPE Middle East Oil and Gas Show and
Conference, Manama, Bahrain, 25–28 September. SPE-140538-MS. https://doi.org/10.2118/140538-MS.
Alimkhanov, R. and Samoylova, I. 2014. Application of Data Mining Tools for Analysis and Prediction of Hydraulic
Fracturing Efficiency for the BV8 Reservoir of the Povkh Oil Field. Presented at the SPE Russian Oil and Gas
Exploration & Production Technical Conference and Exhibition, Moscow, Russia, 14–16 October. SPE-171332-MS.
https://doi.org/10.2118/171332-MS.
Alkamil, E. H. K., Abbas, A. K., Flori, R. et al. 2018. Learning from Experience: Real-Time H2S Monitoring System
Using Fuzzy ART Unsupervised Learning. Presented at the IADC/SPE Asia Pacific Drilling Technology Conference
and Exhibition, Bangkok, Thailand, 27–29 August. SPE-191097-MS. https://doi.org/10.2118/191097-MS.
10 SPE-193776-MS

Alvarado, V., Ranson, A., Hernandez, K. et al. 2002. Selection of EOR/IOR Opportunities Based on Machine Learning.
Presented at the European Petroleum Conference, Aberdeen, Scotland, United Kingdom, 29–31 October. SPE-78332-
MS. https://doi.org/10.2118/78332-MS.
Amr, S., El Ashhab, H., El-Saban, M. et al. 2018. A Large-Scale Study for a Multi-Basin Machine Learning Model
Predicting Horizontal Well Production. Presented at the SPE Annual Technical Conference and Exhibition, Dallas,
Texas, USA, 24–26 September. SPE-191538-MS. https://doi.org/10.2118/191538-MS.
Anderson, R. N., Xie, B., Wu, L. et al. 2016. Using Machine Learning to Identify the Highest Wet Gas Producing
Mix of Hydraulic Fracture Classes and Technology Improvements in the Marcellus Shale. Presented at the
Unconventional Resources Technology Conference, San Antonio, Texas, USA, 1–3 August. URTEC-2430481-MS.
https://doi.org/10.15530/URTEC-2016-2430481.
Andrianova, A., Simonov, M., Perets, D. et al. 2018. Application of Machine Learning for Oilfield Data Quality
Improvement. Presented at the SPE Russian Petroleum Technology Conference, Moscow, Russia, 15–17 October.
SPE-191601-18RPTC-MS. https://doi.org/10.2118/191601-18RPTC-MS.
Ani, M., Oluyemi, G., Petrovski, A. et al. 2016. Reservoir Uncertainty Analysis: The Trends from Probability to
Algorithms and Machine Learning. Presented at the SPE Intelligent Energy International Conference and Exhibition,
Aberdeen, Scotland, United Kingdom, 6–8 September. SPE-181049-MS. https://doi.org/10.2118/181049-MS.
Artun, E., Mohaghegh, S. D., Toro, J. et al. 2005. Reservoir Characterization Using Intelligent Seismic Inversion.
Presented at the SPE Eastern Regional Meeting. Morgantown, West Virginia, USA, 14–16 September. SPE-98012-
MS. https://doi.org/10.2118/98012-MS.
Bangert, P. 2017. Predicting and Detecting Equipment Malfunctions Using Machine Learning. Presented at the OTC
Brasil, Rio de Janeiro, Brazil, 24–26 October. OTC-28109-MS. https://doi.org/10.4043/28109-MS.
Belozerov, B., Bukhanov, N., Egorov, D. et al. 2018. Automatic Well Log Analysis Across Priobskoe Field Using Machine
Learning Methods. Presented at the SPE Russian Petroleum Technology Conference, Moscow, Russia, 15–17 October.
SPE-191604-18RPTC-MS. https://doi.org/10.2118/191604-18RPTC-MS.
Bello, O., Teodoriu, C., Yaqoob, T. et al. 2016. Application of Artificial Intelligence Techniques in Drilling System
Design and Operations: A State of the Art Review and Future Research Pathways. Presented at the SPE
Nigeria Annual International Conference and Exhibition, Lagos, Nigeria, 2–4 August. SPE-184320-MS. https://
doi.org/10.2118/184320-MS.
Benghanem, M. 2012. Artificial Intelligence Techniques for Prediction of Solar Radiation Data: A Review. International
Journal of Renewable Energy Technology 3 (2): 189–220. https://doi.org/10.1504/IJRET.2012.045626.
Bestagini, P., Lipari, V., and Tubaro, S. 2017. A Machine Learning Approach to Facies Classification Using Well Logs.
Presented at the 2017 SEG International Exposition and Annual Meeting, Houston, Texas, USA, 24–29 September.
SEG-2017-17729805.
Bowie, B. 2018. Machine Learning Applied to Optimize Duvernay Well Performance. Presented at the SPE
Canada Unconventional Resources Conference, Calgary, Alberta, Canada, 13–14 March. SPE-189823-MS. https://
doi.org/10.2118/189823-MS.
Braswell, G. 2013. Artificial Intelligence Comes of Age in Oil and Gas. J Pet Technol 65 (1): 50–57. SPE-0113-0050-
JPT. https://doi.org/10.2118/0113-0050-JPT.
Bravo, C. E., Saputelli, L., Rivas, F. et al. 2014. State of the Art of Artificial Intelligence and Predictive Analytics in the
E&P Industry: A Technology Survey. SPE J. 19 (4): 547–563. SPE-150314-PA. https://doi.org/10.2118/150314-PA.
Brown, J. P. 1991. A Machine Learning Approach to Studies of Recovery Efficiency. Presented at the Petroleum Computer
Conference, Dallas, Texas, USA, 17–20 June. SPE-22304-MS. https://doi.org/10.2118/22304-MS.
Cao, R., Banerjee, S., Gupta, J. et al. 2016. Data Driven Production Forecasting Using Machine Learning. Presented at
the SPE Argentina Exploration and Production of Unconventional Resources Symposium, Buenos Aires, Argentina,
1–3 June. SPE-180984-MS. https://doi.org/10.2118/180984-MS.
Cheung, C. M., Goyal, P., Tehrani, A. S. et al. 2017. Deep Learning for Steam Job Candidate Selection. Presented at the
SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 9–11 October. SPE-187339-MS. https://
doi.org/10.2118/187339-MS.
El-Sebakhy, E. A., Sheltami, T., Al-Bokhitan, S. Y. et al. 2007. Support Vector Machines Framework for Predicting the
PVT Properties of Crude Oil Systems. SPE Middle East Oil and Gas Show and Conference, Manama, Bahrain, 11–
14 March. SPE-105698-MS. https://doi.org/10.2118/105698-MS.
Fulford, D. S., Bowie, B., Berry, M. E. et al. 2015. Machine Learning as a Reliable Technology for Evaluating Time-Rate
Performance of Unconventional Wells. Presented at the SPE Annual Technical Conference and Exhibition, Houston,
Texas, USA, 28–30 September. SPE-174784-MS. https://doi.org/10.2118/174784-MS.
Gaganis, V. and Varotsis, N. 2012. Machine Learning Methods to Speed up Compositional Reservoir Simulation.
Presented at the SPE Europec/EAGE Annual Conference, Copenhagen, Denmark, 4–7 June. SPE-154505-MS. https://
doi.org/10.2118/154505-MS.
SPE-193776-MS 11

Gaurav, A. 2017. Horizontal Shale Well EUR Determination Integrating Geology, Machine Learning, Pattern Recognition
and MultiVariate Statistics Focused on the Permian Basin. Presented at the SPE Liquids-Rich Basins Conference -
North America, Midland, Texas, USA, 13–14 September. SPE-187494-MS. https://doi.org/10.2118/187494-MS.
Gharbi, R. B. and Mansoori, G. A. 2005. An Introduction to Artificial Intelligence Applications in Petroleum Exploration
and Production. J Petrol Sci Eng. 49 (3-4): 93–96. https://doi.org/10.1016/j.petrol.2005.09.001.
Guerreiro, J. N. C., Barbosa, H. J. C., Garcia, E. L. M. et al. 1998. Identification of Reservoir Heterogeneities Using
Tracer Breakthrough Profiles and Genetic Algorithms. SPE Res Eval & Eng 1 (3): 218–223. SPE-39066-PA. https://
doi.org/10.2118/39066-PA.
Guitton, A., Wang, H., and Trainor-Guitton, W. 2017. Statistical Imaging of Faults in 3D Seismic Volumes Using a
Machine Learning Approach. Presented at the 2017 SEG International Exposition and Annual Meeting, Houston,
Texas, USA, 24–29 September. SEG-2017-17589633.
Guo, D., Raghavendra, C. S., Yao, K.-T. et al. 2015. Data Driven Approach to Failure Prediction for Electrical Submersible
Pump Systems. Presented at the SPE Western Regional Meeting, Garden Grove, California, USA, 27–30 April.
SPE-174062-MS. https://doi.org/10.2118/174062-MS.
Gupta, S., Saputelli, L., and Nikolaou, M. 2016. Applying Big Data Analytics to Detect, Diagnose, and Prevent Impending
Failures in Electric Submersible Pumps. Presented at the SPE Annual Technical Conference and Exhibition, Dubai,
UAE, 26–28 September. SPE-181510-MS. https://doi.org/10.2118/181510-MS.
Hall, M. and Hall, B. 2017. Distributed Collaborative Prediction: Results of the Machine Learning Contest. The Leading
Edge 36 (3): 267–269. https://doi.org/10.1190/tle36030267.1.
Hoeink, T. and Zambrano, C. 2017. Shale Discrimination with Machine Learning Methods. Presented at the 51st U.S.
Rock Mechanics/Geomechanics Symposium, San Francisco, California, USA, 25–28 June. ARMA-2017-0769.
Honorio, J., Chen, C., Gao, G. et al. 2015. Integration of PCA with a Novel Machine Learning Method for
Reparameterization and Assisted History Matching Geologically Complex Reservoirs. Presented at the SPE
Annual Technical Conference and Exhibition, Houston, Texas, USA, 28–30 September. SPE-175038-MS. https://
doi.org/10.2118/175038-MS.
Kalam, M. Z., Al-Alawi, S. M., and Al-Mukheini, M. 1996. Assessment of Formation Damage Using Artificial Neural
Networks. Presented at the SPE Formation Damage Control Symposium, Lafayette, Louisiana, USA, 14–15 February.
SPE-31100-MS. https://doi.org/10.2118/31100-MS.
Kellogg, R. P., Chessum, W., and Kwong, R. 2018. Machine Learning Application for Wellbore Damage Removal in the
Wilmington Field. Presented at the SPE Western Regional Meeting, Garden Grove, California, USA, 22–26 April.
SPE-190037-MS. https://doi.org/10.2118/190037-MS.
Khan, M. R., Tariq, Z., and Abdulraheem, A. 2018. Machine Learning Derived Correlation to Determine Water Saturation
in Complex Lithologies. Presented at the SPE Kingdom of Saudi Arabia Annual Technical Symposium and Exhibition,
Dammam, Saudi Arabia, 23–26 April. SPE-192307-MS. https://doi.org/10.2118/192307-MS.
Klenner, R., Liu, G., Stephenson, H. et al. 2018. Characterization of Fracture-Driven Interference and the Application of
Machine Learning to Improve Operational Efficiency. Presented at the SPE Liquids-Rich Basins Conference - North
America, Midland, Texas, USA, 5–6 September. SPE-191789-MS. https://doi.org/10.2118/191789-MS.
Li, Y. and Han, Y. 2017. Decline Curve Analysis for Production Forecasting Based on Machine Learning. Presented
at the SPE Symposium: Production Enhancement and Cost Optimisation, Kuala Lumpur, Malaysia, 7–8 November.
SPE-189205-MS. https://doi.org/10.2118/189205-MS.
Li, J., Jichen, Z., Hanqiao, J. et al. 2014. A New Early Warning Model with Vector Machine for Abnormal Production
Prediction. Presented at the SPE Annual Caspian Technical Conference and Exhibition, Astana, Kazakhstan, 12–14
November. SPE-172307-MS. https://doi.org/10.2118/172307-MS.
Liu, J., Jaiswal, A., Yao, K.-T. et al. 2015. Autoencoder-derived Features as Inputs to Classification Algorithms for
Predicting Well Failures. Presented at the SPE Western Regional Meeting, Garden Grove, California, USA, 27–30
April. SPE-174015-MS. https://doi.org/10.2118/174015-MS.
Luo, G., Tian, Y., Bychina, M. et al. 2018. Production Optimization Using Machine Learning in Bakken Shale. Presented
at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Houston, Texas, USA, 23–25 July.
URTEC-2902505-MS. https://doi.org/10.15530/URTEC-2018-2902505.
Ma, Z., Leung, J. Y., and Zanon, S. 2016. Integration of Artificial Intelligence and Production Data Analysis for Shale
Heterogeneity Characterization in SAGD Reservoirs. Presented at the SPE Canada Heavy Oil Technical Conference,
Calgary, Alberta, Canada, 7–9 June. SPE-180716-MS. https://doi.org/10.2118/180716-MS.
Martin, E., Wills, P., Hohl, D. et al. 2017. Using Machine Learning to Predict Production at a Peace River Thermal EOR
Site. Presented at the SPE Reservoir Simulation Conference, Montgomery, Texas, USA, 20–22 February. SPE-182696-
MS. https://doi.org/10.2118/182696-MS.
Medsker, L. R. 1996. Microcomputer Applications of Hybrid Intelligent Systems. Journal of Network and Computer
Applications 19 (2): 213–234. https://doi.org/10.1006/jnca.1996.0015.
12 SPE-193776-MS

Meek, R., Davis, B., and Bello, H. 2017. High Resolution Seismic Data Derived From Prestack Inversion and Machine
Learning to Accurately Position Horizontal Wells in the Midland Basin, Texas. Presented at the SPE/AAPG/SEG
Unconventional Resources Technology Conference, Austin, Texas, USA, 24–26 July. URTEC-2695422-MS. https://
doi.org/10.15530/URTEC-2017-2695422.
Mellit, A. 2008. Artificial Intelligence Technique for Modelling and Forecasting of Solar Radiation Data: A
Review. International Journal of Artificial Intelligence and Soft Computing 1 (1): 52–76. https://doi.org/10.1504/
IJAISC.2008.021264.
Miller, J., Thiagarajan, J. J., Bremer, P.-T. et al. 2017. Data-Driven Metric Learning for History Matching. Presented
at the SPE Reservoir Simulation Conference, Montgomery, Texas, USA, 20–22 February. SPE-182683-MS. https://
doi.org/10.2118/182683-MS.
Mohaghegh, S. D., Grujic, O. S., Zargari, S. et al. 2011. Modeling, History Matching, Forecasting and Analysis of Shale
Reservoirs performance Using Artificial Intelligence. Presented at the SPE Digital Energy Conference and Exhibition,
The Woodlands, Texas, USA, 19–21 April. SPE-143875-MS. https://doi.org/10.2118/143875-MS.
Mohaghegh, S. D. and Khazaeni, Y. 2011. Application of Artificial Intelligence in the Upstream Oil and Gas Industry.
International Journal of Computer Research 18 (3/4): 231-267.
Mukhanov, A., Arturo Garcia, C., and Torres, H. 2018. Water Control Diagnostic Plot Pattern Recognition Using Support
Vector Machine. Presented at the SPE Russian Petroleum Technology Conference, Moscow, Russia, 15–17 October.
SPE-191600-18RPTC-MS. https://doi.org/10.2118/191600-18RPTC-MS.
Nande, S. 2018. Application of Machine Learning for Closure Pressure Determination. Presented at the SPE
Annual Technical Conference and Exhibition, Dallas, Texas, USA, 24–26 September. SPE-194042-STU. https://
doi.org/10.2118/194042-STU.
Nikravesh, M., Kovscek, A. R., Johnston, R. M. et al. 1996. Prediction of Formation Damage During Fluid Injection
into Fractured, Low Permeability Reservoirs via Neural Networks. Presented at the SPE Formation Damage Control
Symposium, Lafayette, Louisiana, USA, 14–15 February. SPE-31103-MS. https://doi.org/10.2118/31103-MS.
Noshi, C. I., Noynaert, S. F., and Schubert, J. J. 2018. Casing Failure Data Analytics: A Novel Data Mining Approach
in Predicting Casing Failures for Improved Drilling Performance and Production Optimization. Presented at the
SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA, 24–26 September. SPE-191570-MS. https://
doi.org/10.2118/191570-MS.
Noshi, C. I. and Schubert, J. J. 2018. The Role of Machine Learning in Drilling Operations; A Review. Presented at
the SPE/AAPG Eastern Regional Meeting, Pittsburgh, Pennsylvania, USA, 7–11 October. SPE-191823-18ERM-MS.
https://doi.org/10.2118/191823-18ERM-MS.
Nwachukwu, A., Jeong, H., Sun, A. et al. 2018. Machine Learning-Based Optimization of Well Locations and WAG
Parameters under Geologic Uncertainty. Presented at the SPE Improved Oil Recovery Conference, Tulsa, Oklahoma,
USA, 14–18 April. SPE-190239-MS. https://doi.org/10.2118/190239-MS.
Odi, U. and Nguyen, T. 2018. Geological Facies Prediction Using Computed Tomography in a Machine Learning and
Deep Learning Environment. Presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference,
Houston, Texas, USA, 23–25 July. URTEC-2901881-MS. https://doi.org/10.15530/URTEC-2018-2901881.
Onwuchekwa, C. 2018. Application of Machine Learning Ideas to Reservoir Fluid Properties Estimation. Presented at the
SPE Nigeria Annual International Conference and Exhibition, Lagos, Nigeria, 6–8 August. SPE-193461-MS. https://
doi.org/10.2118/193461-MS.
Pandya, D., Srivastava, A., Doherty, A. et al. 2018. Increasing Production Efficiency via Compressor Failure Predictive
Analytics Using Machine Learning. Presented at the Offshore Technology Conference, Houston, Texas, USA, 30 April
– 3 May. OTC-28990-MS. https://doi.org/10.4043/28990-MS.
Pankaj, P., Geetan, S., MacDonald, R. et al. 2017. Application of Data Science and Machine Learning for Well Completion
Optimization. Presented at the Offshore Technology Conference, Houston, Texas, USA, 30 April – 3 May. OTC-28632-
MS. https://doi.org/10.4043/28632-MS.
Pennel, M., Hsiung, J., and Putcha, V. B. 2018. Detecting Failures and Optimizing Performance in Artificial Lift Using
Machine Learning Models. Presented at the SPE Western Regional Meeting, Garden Grove, California, USA, 22–26
April. SPE-190090-MS. https://doi.org/10.2118/190090-MS.
Putcha, V. B. and Ertekin, T. 2018. A Hybrid Integrated Compositional Reservoir Simulator Coupling Machine Learning
and Hard Computing Protocols. Presented at the SPE Kingdom of Saudi Arabia Annual Technical Symposium and
Exhibition, Dammam, Saudi Arabia, 23–26 April. SPE-192368-MS. https://doi.org/10.2118/192368-MS.
Pyle, D. and San Jose, C. 2015. An Executive's Guide to Machine Learning. Mckinsey Quarterly (3): 44–53.
https://www.mckinsey.com/industries/high-tech/our-insights/an-executives-guide-to-machine-learning (accessed 23
September 2018).
SPE-193776-MS 13

Raghavenda, C. S., Liu, Y., Wu, A. et al. 2013. Global Model for Failure Prediction for Rod Pump Artificial Lift Systems.
Presented at the SPE Western Regional & AAPG Pacific Section Meeting 2013 Joint Technical Conference, Monterey,
California, USA, 19–25 April. SPE-165374-MS. https://doi.org/10.2118/165374-MS.
Ramirez, A. M., Valle, G. A., Romero, F. et al. 2017. Prediction of PVT Properties in Crude Oil Using Machine Learning
Techniques MLT. Presented at the SPE Latin America and Caribbean Petroleum Engineering Conference, Buenos
Aires, Argentina, 17–19 May. SPE-185536-MS. https://doi.org/10.2118/185536-MS.
Romero, C. E., Carter, J. N., Gringarten, A. C. et al. 2000. A Modified Genetic Algorithm for Reservoir Characterisation.
Presented at the International Oil and Gas Conference and Exhibition in China, Beijing, China, 7–10 November.
SPE-64765-MS. https://doi.org/10.2118/64765-MS.
Roy, A., Jayaram, V., and Marfurt, K. J. 2013. Active Learning Algorithms in Seismic Facies Classification. Presented at
the 2013 SEG Annual Meeting, Houston, Texas, USA, 22–27 September. SEG-2013-0769.
Santos, I. H. F., Machado, M. M., Russo, E. E. et al. 2015. Big Data Analytics for Predictive Maintenance Modeling:
Challenges and Opportunities. Presented at the OTC Brasil, Rio de Janeiro, Brazil, 27–29 October. OTC-26275-MS.
https://doi.org/10.4043/26275-MS.
Schuetter, J., Mishra, S., Zhong, M. et al. 2015. Data Analytics for Production Optimization in Unconventional Reservoirs.
Presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, San Antonio, Texas, USA, 20–
22 July. URTEC-2167005-MS. https://doi.org/10.15530/URTEC-2015-2167005.
Sen, M. K., Datta-Gupta, A., Stoffa, P. L. et al. 1995. Stochastic Reservoir Modeling Using Simulated Annealing and
Genetic Algorithm. SPE Form Eval 10 (1): 49–56. SPE-24754-PA. https://doi.org/10.2118/24754-PA.
Shadravan, A., Tarrahi, M., and Amani, M. 2015. Intelligent Tool to Design Fracturing, Drilling, Spacer and Cement
Slurry Fluids Using Machine Learning Algorithms. Presented at the SPE Kuwait Oil & Gas Show and Conference,
Mishref, Kuwait, 11–14 October. SPE-175238-MS. https://doi.org/10.2118/175238-MS.
Sidaoui, Z., Abdulraheem, A., and Abbad, M. 2018. Prediction of Optimum Injection Rate for Carbonate Acidizing Using
Machine Learning. Presented at the SPE Kingdom of Saudi Arabia Annual Technical Symposium and Exhibition,
Dammam, Saudi Arabia, 23–26 April. SPE-192344-MS. https://doi.org/10.2118/192344-MS.
Sneed, J. 2017. Predicting ESP Lifespan With Machine Learning. Presented at the SPE/AAPG/SEG Unconventional
Resources Technology Conference, Austin, Texas, USA, 24–26 July. URTEC-2669988-MS. https://doi.org/10.15530/
URTEC-2017-2669988.
Smith, K. 2017. Machine Learning Assisted Velocity Auto-Picking. Presented at the 2017 SEG International Exposition
and Annual Meeting, Houston, Texas, USA, 24–29 September. SEG-2017-17684719.
Subrahmanya, N., Xu, P., El-Bakry, A. et al. 2014. Advanced Machine Learning Methods for Production Data Pattern
Recognition. Presented at the SPE Intelligent Energy Conference & Exhibition, Utrecht, The Netherlands, 1–3 April.
SPE-167839-MS. https://doi.org/10.2118/167839-MS.
Tang, D. and Spikes, K. 2017. Segmentation of Shale SEM Images Using Machine Learning. Presented at the 2017 SEG
International Exposition and Annual Meeting, Houston, Texas, USA, 24–29 September. SEG-2017-17738502.
Tarrahi, M., Afra, S., and Surovets, I. 2015. A Novel Automated and Probabilistic EOR Screening Method to Integrate
Theoretical Screening Criteria and Real Field EOR Practices Using Machine Learning Algorithms. Presented at
the SPE Russian Petroleum Technology Conference, Moscow, Russia, 26–28 October. SPE-176725-MS. https://
doi.org/10.2118/176725-MS.
Te Stroet, C., Zwaan, J., de Jager, G. et al. 2017. Predicting Sweet Spots in Shale Plays by DNA Fingerprinting and Machine
Learning. Presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Austin, Texas, USA,
24–26 July. URTEC-2671117-MS. https://doi.org/10.15530/URTEC-2017-2671117.
Tian, C. and Horne, R. N. 2015a. Applying Machine Learning Techniques to Interpret Flow Rate, Pressure and
Temperature Data From Permanent Downhole Gauges. Presented at the SPE Western Regional Meeting, Garden
Grove, California, USA, 27–30 April. SPE-174034-MS. https://doi.org/10.2118/174034-MS.
Tian, C. and Horne, R. N. 2015b. Machine Learning Applied to Multiwell Test Analysis and Flow Rate Reconstruction.
Presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, USA, 28–30 September.
SPE-175059-MS. https://doi.org/10.2118/175059-MS.
Tu, J. V. 1996. Advantages and Disadvantages of Using Artificial Neural Networks versus Logistic Regression
for Predicting Medical Outcomes. Journal of Clinical Epidemiology 49 (11): 1225–1231. https://doi.org/10.1016/
S0895-4356(96)00002-9.
Voleti, D. K., Kundu, A., and Singh, M. 2017. Normalized Depths as Key Input and Detailed QC Steps for
Improved Permeability Predictions Using Existing Machine Learning Techniques. Presented at the Abu Dhabi
International Petroleum Exhibition & Conference, Abu Dhabi, UAE, 13–16 November. SPE-188493-MS. https://
doi.org/10.2118/188493-MS.
Wang, Z., Wei, J., and Jin, H. 2011. Partition Perforation Optimization for Horizontal Wells Based on Genetic Algorithms.
SPE Drill & Compl 26 (1): 52–59. SPE-119833-PA. https://doi.org/10.2118/119833-PA.
14 SPE-193776-MS

Xiao, J. and Sun, X. 2017. Big Data Analytics Drive EOR Projects. Presented at the SPE Offshore Europe Conference &
Exhibition, Aberdeen, United Kingdom, 5–8 September. SPE-186159-MS. https://doi.org/10.2118/186159-MS.
Zadeh, L. A. 1965. Information and Control. Fuzzy Sets 8 (3): 338–353.
Xia, J. C., Xie, F., Zhang, Y. et al. 2013. Artificial Intelligence and Data Mining: Algorithms and Applications. Abstract
and Applied Analysis 2013. http://dx.doi.org/10.1155/2013/524720.
Xuefei, S. and Mohanty, K. K. 2003. Estimation of Flow Functions During Drainage Using Genetic Algorithm. Presented
at the SPE Annual Technical Conference and Exhibition, Denver, Colorado, USA, 5–8 October. SPE-84548-MS.
https://doi.org/10.2118/84548-MS.
Zhang, Q., Song, J., and Nie, X. 1991. Application of Neural Network Models to Rock Mechanics and Rock
Engineering. International Journal of Rock Mechanics and Mining Sciences & Geomechanics 28 (6): 535–540. https://
doi.org/10.1016/0148-9062(91)91130-J.
Zhao, Y., Jiang, H., Li, J. et al. 2017. Study on the Classification and Formation Mechanism of Microscopic Remaining Oil
in High Water Cut Stage Based on Machine Learning. Presented at the Abu Dhabi International Petroleum Exhibition
& Conference, Abu Dhabi, UAE, 13–16 November. SPE-188228-MS. https://doi.org/10.2118/188228-MS.
Zhong, H., Wu, K., Ji, D. et al. 2017. Using Least Square Support Vector Machines to Approximate Single Phase
Flow. Presented at the SPE Europec featured at 79th EAGE Conference and Exhibition, Paris, France, 12–15 June.
SPE-185881-MS. https://doi.org/10.2118/185881-MS.

Potrebbero piacerti anche