Petroleum Research xxx (xxxx) xxx

Contents lists available at ScienceDirect

Petroleum Research
journal homepage: http://www.keaipublishing.com/en/journals/
petroleum-research/

Artificial intelligence: New age of transformation in petroleum

upstream
Parth Solanki a, Dhruv Baldaniya a, Dhruvikkumar Jogani a, Bhavesh Chaudhary a,
Manan Shah b, *, Ameya Kshirsagar c
a
School of Petroleum Technology, Pandit Deendayal Petroleum University, Gandhinagar, 382007, Gujarat, India
b
Department of Chemical Engineering, School of Technology, Pandit Deendayal Petroleum University, Gandhinagar, 382007, Gujarat, India
c
Department of Information Technology, Symbiosis Institute of Technology, Pune, Maharashtra, India

a r t i c l e i n f o a b s t r a c t

Article history: In the Oil and Gas industry, the implementation of artificial intelligence techniques gives advantages of
Received 27 October 2020 better use of existing infrastructure. It provides better future outcomes, which makes it an essential
Received in revised form technology in the operations of the industry. In this paper, the implementation of artificial intelligence
24 June 2021
techniques for good planning, determination of reservoir rock properties, drilling optimization and
Accepted 5 July 2021
Available online xxx
production facilities are reviewed. With precise knowledge of permeability and porosity, reservoir en-
gineers can develop robust reservoir development plan and manage hydrocarbon recovery effectively.
For porosity prediction, wire-line log data and seismic attributes are used in artificial intelligence-based
Keywords:
Artificial neural network
models. Permeability prediction has been discussed by considering the different types of well-log data in
Enhanced oil recovery the modeling of artificial intelligence techniques to predict single-phase fluid. Solution of specific
Fuzzy logic system developed problems during the determination of permeability in the actual experiment is also
Artificial lift mentioned. Artificial intelligence techniques are used in the design and accurate analysis of the drilling
Permeability fluid. It assists in the selection of drill bit and drilling-related problems. If a reservoir cannot produce
Porosity naturally, we can select a suitable artificial lift and design the production system using Artificial intel-
ligence techniques. For EOR methods, SADA and NAVA knowledge base systems are discussed. Artificial
intelligence can help in optimizing production, improve production efficiency and recovery of
hydrocarbons.
© 2021 Chinese Petroleum Society. Publishing services provided by Elsevier B.V. on behalf of KeAi
Communication Co. Ltd. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction categories: (ANNs) and classical Artificial intelligence (Gharbi and

Mansoori, 2005). Artificial intelligence is one type of system
Artificial intelligence is a debatable subject because it involves which makes the computer and machine respond the same as
topics like brain architecture and human intelligence, which we human intelligence (Dodiya and Shah, 2021). Artificial intelligence
humans are oblivious to yet. Artificial intelligence is commonly techniques follow human-like capabilities e.g.: reasoning and
used in computer systems using programs with proper computer learning. The reasoning is very much practical as the repetition and
hardware. Therefore, it generally looks like a science illusion story complexity of data increases. Artificial intelligence is a complex
to unknowns. Artificial intelligence is sometimes called machine subject. It involves human intelligence and things that humans do
intelligence. Security and optimal performances can be improved not comprehend.
and physical assets or oilfield equipment readily available through Artificial intelligence has many definitions, but each definition
mobile asset tagging (Murray et al., 2006). LISP (List and Symbol interprets only one thing. Artificial intelligence is tolerated from
Processing) and PROLOG (Logic Programming) are examples of that. two conditions: small number and more number of definitions
Artificial intelligence studies are divided into two main (Agwu et al., 2018). Artificial intelligence can be described as the
ability to understand and retain it as knowledge within the envi-
ronment for context. Artificial intelligence is imitating human in-
* Corresponding author. telligence by machines and computer systems (Desai and Shah,
E-mail address: [email protected] (M. Shah). 2020; Panchiwala and Shah, 2020; Thakkar et al., 2020).

https://doi.org/10.1016/j.ptlrs.2021.07.002
2096-2495/© 2021 Chinese Petroleum Society. Publishing services provided by Elsevier B.V. on behalf of KeAi Communication Co. Ltd. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: P. Solanki, D. Baldaniya, D. Jogani et al., Artificial intelligence: New age of transformation in petroleum upstream,
Petroleum Research, https://doi.org/10.1016/j.ptlrs.2021.07.002
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

Artificial Intelligence can be classified as weak and strong. A might be able to process data in more efficient ways to develop
weak Artificial intelligence system is which is trained and designed recommendations on whether to explore & develop or walk away.
for a particular task. Strong artificial intelligence is the system with Mine Portal Use for analysing the geosciences data. The petroleum
generalized human awareness when presented with an unfamiliar sector always has a safety issue. It is a primary concern both
solution without human intercession. The artificial intelligence offshore & onshore platform. Offshore has heavy equipment, plat-
process goes in a direction; one is pure science, which is related to form & transportation via helicopters. Onshore has toxic gases.
understanding and developing the true mechanisms of natural Safety is a success in day to day life. The data always capture due to
intelligence. That Interpretation gives information about the ma- safety issues that occur, from that analysed, prevent the data drive
chine that can emulate the people's performance, which humans to the root of the causing to make the preventive decision. AI can
consider as intelligence. quickly locate component and equipment parts in the event of
Artificial intelligence methods are classified into four main theft. Adopting AI meets the challenges in the oilfield operation,
categories: particle swarm optimization, evolutionary intelligence intervention and production optimization (Akanji and Ofi, 2016).
algorithms, fuzzy logic, and artificial neural networks (Rahmanifard In the petroleum has historically used the large warehouse,
and Plaksina, 2019). Artificial Neural networks, fuzzy logic, storing materials and equipment on hands for maintenance needs
Knowledge-based systems, expert systems and genetic algorithms, the potential to strategically predict the demand, making stocking
and others are new artificial intelligence techniques that are very types and levels a source of value. These advancements could help
useful in data analysis and applied in the prediction process, reduce working capital and ensure customers' needs are more
diagnostic and no lineal complex transforms with a high degree of satisfied.
uncertainty. (Tapias et al., 2001). After getting information about Porosity, permeability, water saturation, wellbore stability are
artificial intelligence, the research has been started in the direction predicted by using Artificial intelligence techniques. Porosity and
of how it is applied in human life. Permeability are two important petro-physical properties that can
The recent basis of development and studies prove that artificial be predicted by using the Adaptive Neuro-Fuzzy Inference System
intelligence can improve the working capacity and helpful in all the (ANFIS), artificial neural network (ANN), support vector machine
petroleum sectors. In past years the Artificial Neural Network (SVM), etc. The prediction of porosity and Permeability using AI
developed very quickly. Its approach is based on the human bio- doesn't require any mathematical equation to solve manually. ANN
logical system. Artificial neural networks used to predict bubble is one of the fast and efficient techniques for the prediction of
point pressure of crude oils, determine water-oil relative perme- reservoir properties and characterization. Artificial intelligence is
ability of reservoirs, and produce hydrocarbons from the shale used by the offshore oil and gas industry in data science to make
reservoir. Fuzzy logic has been used in several petroleum the complex data used for oil and gas exploration and production
engineering-related applications. These include reservoir charac- more accessible, so companies discover new exploration opportu-
terization, enhanced oil recovery, infill drilling, well stimulation, nities to make more use of existing infrastructure.
etc. These other non-linear equations and equations are recognized
The application of Artificial intelligence tools such as fuzzy logic as response surface models (RSM), and also, the simple mathe-
and neural networks is evolving as oilfield technology. In 1996 matical structure is well known for more straightforward imple-
PRRC researchers used fuzzy logics to compute the significance of mentation. Artificial intelligence is connecting input data to output
mini-permeameter permeability measurement. One of the goals of data by applying its techniques which improved the concern of
artificial intelligence in upstream operations is personnel strategy developers. Many researchers also improve analytical, numerical,
optimization to minimize risk. Artificial intelligence can affect and other processes to comprehend the concept implemented in
drilling space positively. There are many advantages of artificial the generation. These are complex systems so that these methods
intelligence in the petroleum industry. were inappropriate for forecast hydrocarbon production from
shale. So many appeals of LSSVM, which contains some geological
2. Impact of AI in petroleum industry parameters, are included by some useful parameters used as
completion parameters, rock-fluid properties, and fluid properties
The oil & gas sector hired thousands of employ every year. For collected from past studies where studies divulge these parameters
that, develop the AI technique to analyse the resume of the to be very important. The drilling process can be enhanced with the
candidate. This will find the clues in the resume which department use of AI. AI used in drilling dysfunction differential sticking, stuck
best for it as per the company's requirement, which saves them pipe, excessive down-hole vibration, hole cleaning, pressure tran-
time and money of the company's HR. The oil & gas sector has sients, managed pressure drilling. This situation is solved by ana-
complex accounting, which is required highly qualified accountants lysing the down-hole data intelligent drilling drill the smooth
& expenses that are time-consuming & cumber. Also, the cash al- wellbore, which will optimize the reservoir (Nunoo, 2018).
locations are very rule-based. To perform all work, AI trained for
that, which gives more valuable information & redeploys the ac- 3. A novel approach of artificial intelligence in drilling
countants. The pressure derivative 9e11 is defined as the most
beneficial tool for well-test data carried by interpretation proced- (G Lea, 2015) represented that ‘artificial’ is very easy to represent
ure (Allain and Horne, 1990). Drilling, Oil/gas exploration, reservoir anything. (Shane Legg and Marcus Hutter, 2007) represented that
management, and production have challenges because most oil and intelligence definition is vast and disparate. Rable (2017) States that
gas reserves are discovered and produced for many years. So the the drilling process in the oil and gas industries was a challenging
petroleum industry started the research and development potential term that involves such processes. In the oil well drilling process,
of using artificial intelligence techniques such as ANNs, which helps drilling fluid plays an important role.
to make decisions to reduce cost, time and save money. Equipment Drilling fluid has overtly or incidentally solution compared to
such as truck engines used in oil and sand sectors. Companies the above difficulties (Agwu et al., 2018). represented that drilling
spend a hefty amount on maintenance. However, AI reduces the fluid makes the drilling process successful (Ben Bloys et al., 1994).
cost drive to better performance. Oil and gas companies have stated that the drilling cost of the drilling mud puts around 5%e
skilled geologists, petrophysicsts & geophysicists who analyse the 15%, and it may occur 100% during the drilling process. So there is
geological formation predict where the hydrocarbon. Using AI, it only one way to decrease the drilling problem to select the best
2
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

drilling fluid, which depends upon the data analysis used in pre- design, maintenance facilitation.
vious drilling wells. Distinguished drilling mud gives mud proper- Artificial intelligence machine works itself without any human
ties and mud maintenance requirements. input. The basic principle is neural networks. In the sense of the
(Abdollahi, 2007) stated that the knowledge which is gained human brain, the eyes are the primary sensors that are controlled
from experience is easily left when a laboratory analyst or engineer by our brain neurons. In this engine, digital webcam work as human
who has essential experience left the company. That is why we eyes controlled by the image processor, which acts as a human
need a system that analysed and used all the data and give brain. Here, a computer works as image processing used to perceive
knowledge about drilling mud design (Tarrahi and Shadravan, the body on the machine bed on which the drilling procedure is to
2016). On wellbore drilling and drilling fluid engineering, several be held. The result generated from the MATLAB units decided to
AI kinds of research had been done. switch on the three pieces of equipment simultaneously (Table 1).
There are so many artificial intelligence techniques related to Surface roughness can be measured using the direct contact
drilling mud engineering. Some of our ANN, FUZZY LOGIC, Hybrid method (DM) and indirect contact methods. The first method is
Intelligence System, Genetic Algorithms, Support Vector Machine, based on the stylus machine, which must be directly connected to
Case-Based Reasoning, etc. So many AI techniques exhibit good the surface. The indirect method is based on the optical machine,
potential in generating accurate analysis. which is mostly not in contact with a surface and is easy to operate.
According to some enthusiasts and researchers, to solve many For the investigation of water state in drilling, took the help of a
problems related to scientific and engineering field problems, they fuzzy pattern machine (Fig. 1).
applied artificial intelligence. As a good technique, knowledge and (Bello et al., 2016) used artificial intelligence techniques in
popular in digenesis, geology, monitoring, forecasting fields. AI has drilling processes and designing purposes and examined artificial
good potential to get accurate and perfect databases and results. In intelligence techniques as concerned with one of the significant
the oil and gas industries, the drilling operation section is the most aspects in the oil and gas industry. He gave the idea of the profits,
crucial element. applications, and limitations of artificial intelligence and drilling
Drilling industries are dependent entirely on techniques. It is a system design and applications in the oil and gas industry. The
big operation around the world in which a large amount of money result of this may be predicted to accommodate the additional
is spent daily. So there are some applications of artificial intelli- development of this technique. As per the result, ANN, BPNN, CBR
gence in the drilling sector: well planning, pattern reorganization, AI methods were mostly used in drilling processes. Today oil and
drilling optimization, procedure decision-taking, and well integrity, gas industries have made huge development in drilling processing.
operational troubleshooting. AI also played an essential role in the (Chironjeet Kumar Chaki, 2011) discussed the discovery of
selection of good drill bits. Based on the formation's characteristics, automatically controlled drilling machines in artificial intelligence
the selection of the drilling bits is the most prospective sector. An and image processing using artificial neural networks (ANN). It is
artificial neural network has different tools for decoding data, ge- needful to increase and decrease the production cost, leading to a
ology, and format characteristics corresponding to the rocks. completely automatic machine station. The inaccurate drilling
(Agu and Lie, 2014) used some technical equations and rectan- process must give the tool deformation type of result. He used three
gular channel to determine the flow rate of fluid in stationary and disparate types of drill bit in prelim drilling. This was because the
dimension condition. The results of this works indicate that the rate main aim was to develop an intelligence machine so, tool defor-
of the flow of the fluid can only be measured in a steady-state mation was not properly considered but producing artificial intel-
condition, and it can be measured with a Kalman filter. ligence drilling machines on a large scale has been taken carefully.
(Aliouane et al., 2015) took the help of two artificial intelligence
methods, such as ANN and fuzzy logic, to define pore pressure in 4. Artificial intelligence approach in reservoir department
two horizontal wells. They used some input parameters. With
Eaton's model, the output parameters used to estimate porosity A study mentioned about data used of core samples from asmari
pressure. The results they got from a particular process, on that they formation in the southwest of Iran. In this paper, six key points
made report stated that fuzzy logic makes the good result more considered are cross-point saturation, Relative Permeability at
than ANN to predict the porosity of shale gas reservoir. Cross Point Saturation, water relative permeability at maximum
(BA et al., 2009) used small particles to determine the faults. water saturation, and maximum oil relative permeability predicted
Some researchers had studied how to initiate the time regarding by the ANN. Maximum water saturation, Initial water saturation are
porosity prediction of formations, and he got the result after known parameters used for prediction other key points. Predicted
combining two techniques: Eaton's model and logging while dril- data and actual data curve analysed for each key point proved that
ling. As a result, they stated that the Kalman filter is very effectively the predicted value is the same as the actual value. By considering
used in porosity prediction zones. They also stated that developed this key point, the constructed full relative permeability curve is
Algorithms utilized for design purpose in onshore-offshore drilling straightforward. The output showcased that ANN efficiently pre-
operations with the help of the present LWD result. dicted the whole relative permeability curve.
(Keshavarzi and Jahanbakhshi, 2013) used the ANN technique to An excerpt described that finding relative permeability through
measured porosity pressure in the iron formation. A drilling ma- mathematical modelling has limitations. ANN has the potential to
chine was in place in artificial intelligence (AI), Image Processing give effective results of the relative permeability prediction by
(IP) and neural network. This has been only hypothesized. In the controlling the mechanism of modal which uses in the prediction of
first step, a general model was discovered and adapted to image relative permeability. It is a two-phase (oil-water) relative perme-
processing software so that any shape which comes automatically ability prediction network discussed. This is developed in five
could be tested. So, by the appropriate calibration of the machine, stages; a separate model developed for two phases for relative
whatever result is come, sent to the drilling control unit. permeability prediction. The sub-model developed for the main
The mechanical system contains two things for the body's pace ANN network gives prediction at only water saturation and
according to the x, y, and z-axis. Screw collar used to get linear endpoint saturation input parameter. By increasing the developing
motion. In the drilling process, the work statement is the most stages, it gives high accuracy in the prediction. It confirms that the
important thing to drill, depth of the hole and drilling speed. Other network structure is suitable for determining relative permeability
characters are also crucial for reflection are the simplicity of the (Table 1).
3
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

Table 1
Summary of literature on the application of various Artificial intelligence techniques used in petroleum engineering.

Author Title of work Important aspect of the paper Used Artificial

Intelligence Technique

Mohaghegh et al. Prudhoe Bay oil production optimization: using Maximizes total oil field production by Neural model building
(2002) virtual intelligence techniques, stage one: optimizing the gas discharge rates and pressure
neural model building at the separation
Ilkhchi et al. (2006) A fuzzy logic approach for estimation of Assortment of rock type by using porosity and fuzzy c-means
permeability and rock type from conventional permeability data into TakagieSugeno fuzzy clustering
well log data: an example from the kangan inference system.
reservoir in the Iran offshore gas field
Elkatatny et al. (2018) New insights into porosity determination using Predicted the porosity for carbonate reservoir artificial neural
artificial intelligence techniques for Carbonate by use of artificial neural network network
reservoirs
Asoodeh and Core Porosity Estimation through Different Estimated porosity of the core plug by use of BPNN (Back
Bagheripour (2013) Training Approaches for Neural Network: Back- Back Propagation neural network Propagation neural
Propagation Learning vs. Genetic Algorithm network)
GONN (Genetic
optimized neural
network)
(A Hosseini et al., 2011) Artificial Intelligence for prediction of porosity Seismic attributes are utilize for prediction of radial basic function
from Seismic hydrocarbon reservoir porosity network, multi-layer
Attributes: Case study in the Persian Gulf feed
forward neural
network and
probabilistic neural
network
AlBinHassan and Wang Porosity prediction using the group method of Predicted the porosity of reservoir by use of General regression
(2011) data handling GMDH (Group method of data handling) neural networks and
Support vector machine
Ansari (2014) Use seismic colored inversion and power law Predicted porosity in a heterogeneous reservoir Seismic colored
committee machine based on imperial by implementation of Power Law Committee inversion and power-
competitive algorithm for improving porosity Machine constructed based on empirical law committee
prediction in a heterogeneous reservoir competitive algorithm machine

(Nooruddin et al., 2013) explained that the prediction of developed for the actual permeability measurement in this paper.
permeability from the MICP (mercury injection capillary pressure) Authors described mathematical models for predicting the
data of capillary pressure produces a high error to compare the permeability, which results compared to the feed-forward ANN
actual measurement. To solve this problem, ANN network was network for every two hidden layers of 15 neurons with a

Fig. 1. Structure of the Artificial intelligence drilling machine.

4
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

developed tan-sigmoid transfer function result. The same input to predict the permeability. The trained network predicts actual
parameter is considered in above the mathematical model. Ann permeability than the actual measurement.
model outperformed all other models with much fewer errors. Several researchers proposed the neural network to determine
(Anifowose et al., 2013) studied about applying advanced arti- the permeability of porous media, the resistivity of the deep lateral
ficial intelligence in seismic and log data sets to predict the accurate log, the resistivity of micro gradient log, the interval transit time of
reservoir permeability predictions. There is three artificial tech- acoustic log, the resistivity of the shallow log, resistivity of standard
nique applied to study: SVM (Support Vector Machine), ANN micro log and potential spontaneous log used as the input
(Artificial Neural Network) & t2fls (type-2 fuzzy logic system) on parameter and permeability use as an output parameter. There are
seismic and log database. For evaluation of ANN network, three 35 and 40 units used in the two hidden layers. Through the testing
parameters consider: correlation coefficient (CC), root mean and analysing the structure of the neural network is finally built.
squared error (RMSE) and mean absolute error (MAE) used with an The study concluded that the backward propagation network of
execution time (ET). The result shows that combining seismic and ANN successfully predicted the permeability. Three-factor has a
log data improved the prediction accuracy of permeability. significant influence on that ANN network.
Applying the technique to the broader field gives more compre- (Al-Fattah, 2007) presented six developed networks of ANN was
hension detail & the possibility to save more time, effort and to predict accurate oil/water relative permeability by using three
improve hydrocarbon recovery. groups of data sets of the carbonate reservoir of Saudi Arabians oil
(Ahmadi et al., 2016)described that relative permeability is fields. Genetic algorithm, forward and backward stepwise algo-
essential in a reservoir filled with one homogenous fluid. It is rithm, sensitivity analysis are used to determine the significant
dependent on the effectiveness and relative permeability. In 1944, contribution of the variable in ANN as consider the input variable
the research started to determine relative permeability from the developing the model. Data is divided into three parts used in
laboratory method. But that experimental determination is very trainingdVerifying & testing of ANN by GRNN (Generalized
costly and time-consuming. After that authors introduced the Regression Neural Network). ANN network gave excellent perfor-
correlation to predict the relative permeability. By that data, the mance compared to all published equations. In the future, it pro-
application of particle swarm optimization, artificial neural vides the best estimation of relative oil/water permeability for
network & genetic algorithms to determine the oil-water relative newly drilled wells.
permeability. The genetic algorithm is time-consuming to improve (Kaydani et al., 2014) proposed to depend on Multi-Gene Ge-
the accuracy. The hybrid genetic algorithm and particle swarm netic Programming (MGGP) to predict permeability in heteroge-
optimization are combined and create the hybrid genetic algorithm neous oil reservoirs in Iran. ANN (artificial neural network),
and particle swarm optimization. That mean square error mini- adaptive neuron-fuzzy inference system (ANFIS) and genetic pro-
mises the use to compare the output with experimental data. It is gramming used in prediction. The equation used to calculate the
excellent compared to any other intelligent method. ANOVA tech- permeability. It is obtained from MGGP. The result of comparison
nique concluded that water saturation formation type and forma- MGGP model and core permeability measurements which showed
tion type on the oil relative permeability are more than any other the superiority of MGGP in the prediction of permeability in the
factor. porous medium.
A group of enthusiasts and researchers collected data on 3 wells (Ahmadi, 2015) used a high rated branch Least Square Support
of carbonate reservoir located in southern Iran. In this paper, pre- Vector Machine (LSSVM) to improve the accurate permeability
diction of permeability of three wells discussed by applying RVR prediction. For developing and testing massive experimental rela-
(Relevance Vector Regression) & SVR (Support Vector Regression). tive permeability data taken from literature for the approached
For doing that as an optimizer, the genetic algorithm was used to model. Prediction of Oil/gas relative permeability has high perfor-
find the best log for prediction of permeability. Two wells data are mance and low uncertainty of the addressed model it has certified.
considered for training & testing, one well data used for validation It gives a more precise curve for relative permeability.
of the network. The accuracy of RVR is greater than SVR. But for SVR (Elkatatny et al., 2018) explain how the AI technique is used to
(Support Vector Regression) RMSE value is slightly less than RVR. It determine porosity in a carbonate reservoir. Porosity is determined
was the second option after the SVR (Support Vector Regression). in the laboratory, which is an expensive method. It is also deter-
(Olatunji et al., 2011) collected the five different actual industrial mined from different types of well loges like bulk density, sonic
well log data from middle-eastern petroleum reservoirs to predict compression time, neutron porosity, but there is a lot of uncertainty
the permeability by type 2 fuzzy logic system. It is developed to to find porosity from wire-line logs. Building a mathematical model
handle all form uncertainties in the well log data. The paper's that can predict porosity based on log data using the AI technique is
objective is to develop intelligence for permeability prediction from the main objective of this research. There are three different AI
well log data, compare its results with the ANN, SVM (Support technique Support Vector Machine (SVM), Artificial Neural network
Vector Machine), and type-1 fuzzy logic gives detailed information (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS). For
about how to type
2 fuzzy logic system generate without extra this AI model, 1700 actual measurements data were used for
computational cost. The results show that type-2 fuzzy logic per- training and testing. For this tested data, ANN and ANFIS types of
formed outstanding compared to others to handle the data in un- techniques can be used to estimate porosity for low AAPE (average
certain situations. The paper motioned that this framework used to absolute percentage error) and high correlation coefficient (R).
build the porosity, lithofacies, PVT properties through future work. Predict the reservoir porosity based on log data using the ANN
(Mohaghegh et al., 1994) designed an artificial neural network model of a high correlation coefficient of 0.98 and AAPE less than
that predicts the accurately predicted permeability by using the 8%.
well log data. A three-layer network of the ANN network with 15 (Singh et al., 2016) did a case study for the Kansas gas field in the
neurons in the hidden layer.well log and core data used for the USA and obtained how to predict the porosity from other well log
three-layer network of feed-forward, back-propagation network. data using the ANN technique. There are three steps in ANN process
Twenty-three examples were separated out from 151 examples. testing, training, and validation of the network training step.
Gamma-ray, deep induction log responses, bulk density, perme- Porosity generated from neutron log data and combined density is
ability measurements use in each example to training the ANN. compared with the porosity generated from ANN. The high degree
Results portrayed that neuro estimation is a feasible methodology of correlation between these two shows the potential of the
5
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

Artificial Neural Network method. ANN method designed so that if energy to produce liquids to the surface naturally. Bona-fide and
the lithology is not changed and real well log data is not available accurate calculations of down-hole pressure, intake pressure,
for the other well of the same field, the study of the ANN designed down-hole temperature, down-hole equipment temperature, and
to estimate the porosity of this well. free gas amount, etc (Table 1).
(Asoodeh and Bagheripour, 2013) use the different training ap- By (Elmahbes et al., 2016), simulating and predicting different
proaches for neural networks to estimate the core porosity. GONN conditions to select artificial lift technology is complicated. Thus,
(genetic optimized neural network) formulate conventional well AutographPC™ software simulates and sizes different forms of
log data into core porosity. There are some stages in this study; the artificial lift and selects artificial lift systems for production. This
first stage is to check the quality of datasets and remove bad hole software can be used to design a production system for different
intervals by processing conventional well log data. In the second forms of artificial lift technology, including ESP, ESPCP, RDPCP,
stage, to estimate the porosity from conventional well log data, the HPUMP, and gas lift system.
three-layer back-propagation neural network is constructed. In the It is a tool that helps to analyse and simulate the well data that
third stage, the unfilled skull of a neural network is introduced to can assist the reservoir in selecting the appropriate artificial lift
the genetic algorithm (GA), and it is filling this skull into its sto- technology and accurately predicting the future performance of
chastic search capability. Optimizing neural network using GR is selected artificial; lift technology. It helps to determine the flow
called GONN. Porosity from conventional well log data is precisely information, fluid viscosity, and free gas percentage in that with
estimated from BPNN. GONN also has enhanced accuracy to predict entered data.
porosity. The correlation coefficient for both is equal, so GONN is In the EASPE & RDPCP mode, the software provides different
also used instead of BPNN. pumps, gas separators, motor-head, cable, gearbox and controller
A study aimed to predict the porosity using a combination of the options that proper for different applications and challenges. In
Discrete Element Method (DEM) and Artificial neural network Gas-lift mode, the software provides different gas lift size valve
(ANN) technique of Gravel Bed River. The first 3D structure of the selection and different gas injection scenarios to evaluate the best
gravel bed and fine sediment infiltration simulate by DEM. economic scenarios to produce the targeted flow-rate.
This result verified with fine sediment distribution and experi- The AutographPC dynamic simulator can analyse well data and
mental data of porosity. Then one algorithm developed that ana- system performance dynamically. For example, the user can make
lysed DEM results and calculated the grain size distribution and the pump operate beyond its recommended operating range and
porosity. The Data, which is based on DEM results in design ANN, change well parameters like water cut, down-hole pressure, PI, etc.
finds the relationship between grain size distribution and porosity. Artificial lift system often has to be installed in wells that are not
This result proved that the combination of ANN and DEM suc- vertical. Mostly those are installed below the Dogleg at the tangent
cessfully predicts the gravel bed's porosity and fine sediment section. The bending stress must not exceed the equipment stress
infiltration process. limit, permanent damage will occur, and performance will be
(Li et al., 2009) uses distance and attribute weighted k-nearest affected. The autographPC stress analysis program (ASAP) allows
Neighbor to predict the reservoir porosity. This paper first reviewed the user to configure ESP & ESPCP and compute the bending stress
some regression algorithms and then presented KNNDAW (k- as the equipment undergoes dog leg during installation and dis-
nearest Neighbor), which the accuracy is higher than other played ref flag if stress is beyond recommended limits.
algorithms. (Ashkok and Behounek, 2018) introduced that during petroleum
(Al-Anazi and Gates, 2010), use the Support vector regression to production, the drilling failure, fluid influx/loss, and drilling
predict the porosity in the heterogeneous reservoir. Support vector dysfunction occur for that solution there are many alerting and
machines (SVM) are explored as an intelligent technique to corre- detection system applied. It includes the MWD pressure and mud
late porosity to well log data. Hear, for prediction of porosity in the motor data that may improve detection and high frequency.as a
heterogeneous reservoir the SVR (support vector regression) based result, and it does not recognize the failure because of poor sensor
model is developed. Porosity prediction in the heterogeneous and various factors affected by the flow. As the detection and
reservoir by using SVR modelling is more accurate than Radial basis alerting system, the Bayesian network system gathered data from
function neural networks and General Regression neural networks. sensors. That data and contextual data as input give a probabilistic
(AlBinHassan and Wang, 2011) predicted the porosity using the system that gives the belief between 0 and 1 values; this is based
group method of data handling. GMDH is a supervised machine upon how the human brain uses experience to react worldwide
learning tool, and it is also a self-organized model. GMDH algo- with its intelligence. There are three types of nodes used; nodes for
rithm is applied in seismic attributes to predict the reservoir the unplanned event that is programmed for associated with
porosity and the nonlinear relationship between the seismic probabilities or beliefs. Nodes for planned events for pumped start-
attribute and porosity by finding using the GMDH model. The stop routine and lastly, other nodes are made up for processed real-
advantage of GMDH is that is automatically determined the best time input. T nodes are connected by conditional probability tables
neural network structure and number of nodes without over- (CPT). Each CPT has assigned specific weight on its impacts by the
training the data. The result shows that the predicted porosity by outcome. Training for CPT is given by the learning data set. for
using the GMDH model has a better resolution than ANN. washout and the pump failure has two-stage alerting the first stage
is alert to the drilling engineer that pump failure and or washout
5. Optimization of production and selection of artificial lift maybe occur, in second stage drilling engineer analyse precisely the
by using artificial intelligence factor for pump failure or washout which one is occurring.
From (Al-Fattah and Startzman, 2001) Prediction of Natural gas
Artificial lift production system sizing and simulation software supply for the short term and the long term has only possible as a
are applied in different forms of artificial lift systems to optimize neural network model. Artificial intelligence is represented as to
production, improve efficiency and increase ultimate recovery. To the accuracy of this system. This process used to enhance the oil
determine the best application for different artificial lift systems, and to raise the production of petroleum. Drilling is defined as
this software ESP (Electric Submersible Pumps) is used. another process to minimize well spacing and in so much in a water
Selecting the appropriate artificial lift technology is a crucial fluid injection pattern. This is a typical process of estimating oil and
step to keep wells producing when reservoirs cannot provide gas.
6
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

The production process is a direct process. This process is description, utility number.
determining the oil and gas quantities, data related to production,
etc. The numerical method used the real variability of different - NAVA Knowledge Base.
studies which was related to each well in the drilling process. But in
a real case, a production outline can not the same as the real pro- It is advisable to know the additional percentage of OOlP levi-
duction outline. So many elements which do not relate to the able by the selected EOR methods. The NAVA knowledge base
drilling process itself and effect the long term forecast, and some system is discovered using the expert system shell PC-PLUS to
challenging production and injection rate. provide this information upon guidance by the user. This knowl-
By (Boomer, 1995), The professional prediction has a small edge base is concocted of two parameters, two-goal parameters,
defect of 39BOPD and an average element of 99BOPD. The further and 22 rules. The two parameters of the N A V A knowledge base are
process gave the result that professionals predict with 20BOPD just “selected EOR method” and “percentage of OOlP recovered by
apparent of the time and 10BOPD some % of the time. primary and conventional oil recovery processes (PAC)". The former
As the network keeps increase, the test for this product will is a single-valued parameter used to immediate the user for the
illustrate that. When the network obtains the error in the test, it type of EOR methods previously selected by the SADA knowledge
will continue to training as it will only understand the result. base as most suitable for the candidate reservoir. NAVA knowledge
Neural Networks are so good at learning that they can memorize all base has 22 rules applied to set the values of the parameters and
of the training examples. This usually produces a network that does the goal parameters discussed above.
not understand; it will be poor at applying its new-found knowl- We explained the results of sample runs to test the credibility
edge. As training will continue, the performance of the model is and compatibility of the expert systems developed and explained
improved to these tests. above using real data from oil fields. U.S. and Canada were used for
During training the network, as the error increases after that, it the tests, four sets of data for reservoir from oil fields. These
suite to continue the network's training. An error of the network reservoir data are show application. The data and results have been
gives the measurement of the good performance of a network. It condensed for brevity.
does not matter to select the finding, examine and interpret data A group of researchers has discovered a tool to estimate this
sets to be taken lightly. In this process, the data must be taken from process for reservoir engineers. Any systematic method which
the actual problem. The point for which we are concerned about depends upon numerical evolution was used to discover the criteria
whatever model bullied it has to capture is used to solve the used in the expert system. They find the solution which was related
problem by professionals. to the technical selection of the EPR technique. The selection of
proper EOR methods depends upon technical and economic stan-
6. Selection of EOR method using artificial intelligence dards (Table 1).
A neural network system solved the problem related to esti-
Two knowledge-based expert system has been formed for select mating technically and economically EOR methods. However,
best EOR methods for petroleum engineers for quantities of how economic examination, which was not clear up by this system, now
much oil is conversable and economically possible to select the EOR clear up by the neural network approach. To design two separate
methods. Artificial intelligence field developed in all areas of en- networks, one which has aimed to get the technically successful
gineering. Artificial intelligence is now beneficial in many petro- project and can be applied under reservoir condition. Moreover,
leum engineering departments such as drilling, well testing, examine the project by applying a second neural network whether
evaluating economically, well logging and interpretation. This the project is beneficial.
expert system and interactive computer programs help select and Fuzzy logic, which acts as an artificial intelligence technique,
evaluate the capability of EOR methods. This knowledge-based was a help to develop the EOR screening criteria model. Around the
expert system has been examined with reservoir data and reser- world, the new screening methodology was developed from sta-
voir engineering information. tistical analysis of a database of 347 successful EOR projects. The
Two knowledge-based expert systems were formed in this case. new generated tool can be applied to screening for 14 EOR
For the reason of this discussion, we shall refer to these expert methods. The screening results for some of the Egyptian oil reser-
systems as SADA and NAVA. Fundamental components are the voirs are conferred and examined. In some of the Gulf of Suez
parameters in the development of expert systems, and they contain reservoirs, steam injection is feasible. In the western desert reser-
the information used to reach a logical conclusion. voirs, the method is not seen to be feasible. In-Situ Combustion is
the most appropriate method for the Gulf of Suez reservoirs. In
- SADA Knowledge Base. most western desert reservoirs, the chemical methods are feasible,
but they are not seen as feasible in most of the reservoirs in the Gulf
It has been selected to produce the most suitable EOR technique of Suez. Egyptian oil reservoirs are good candidates for immiscible
based on the information provided by the user Geophysical, eco- gas injection. However, miscible methods are not seen to feasible in
nomic, and operational parameters in the SADA knowledge base are most of these oil reservoirs.
single-valued; there are 29. The SADA knowledge base has two goal Simulation of EOR processes by the using of available software
parameters. The first one is the “selected EOR method”. Depends on programs being difficult and expensive laboratory work. The real
the user's response to the values of the parameters, the value of this comparison of several different EOR methods through software
goal parameter will be set to one or more of the 11 EOR methods. A simulation is ineffective because these applications want different
goal parameter value with a certainty factor (CF) is 99 indicates that parameters for different input methods. In contrast, the preparation
the EOR method is most appropriate for the application, as defined of these methods is a time-consuming and expensive procedure.
by the user. A value of CF that equals zero indicates that the EOR There are new tools that can be used as available data from past
method is least suitable, and a negative value of CF indicates that activities to estimate the status of new works. Intelligent systems
the EOR method is not suitable at all. The second goal parameter is are the most powerful tools in this area, which includes various
“and recoil.” It is a single-valued parameter. The knowledge base types. While using individual intelligent systems, a combination of
SADA has been made up of 40 rules. The rules were discovered on these methods could not achieve an acceptable model in the pre-
an “if and then” basis, and then some other rule properties such as sent study. The fuzzy neural model, ANFIS, was considered. The
7
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

result of this scattered data, model design procedure complete in and simulate well data and predict performance. By using Auto-
three stages and techniques such as normalization and clustering graphPC software, we can simulate and select a different type of
were used to decrease the order of error. Results indicate that these artificial lift system and production system. That provides different
techniques have a notably low error rate. To detain spending time gas lift size valve selection and different gas injection scenarios to
and high expenses, a model was designed with the help of past evaluate the best economic scenarios to produce the targeted flow
works and without any need for laboratory work. This model can rate. The oil and gas industry has done extensive development in
suggest an effective oil production rate for a field in five different drilling, reservoir, production and EOR sectors by using Artificial
EOR methods and predict the best performance for field intelligence. Artificial intelligence techniques are accurate, but the
application. cost is one of its barriers. Based on the available data, the Fuzzy
Artificial intelligence depends on ML-NN is applied for multi- logic technique has been used to build EOR models. Both SAVA and
dimensional interpolation of the relative permeability in EOR NADA expert systems can be helpful to find the best EOR process
processes. Hybrid LSS flooding is a promise to the EOR method as it and calculate the amount of recoverable oil by EOR methods.
gives a high recovery factor, up to 92% OOIP and less surfactant
retention at the laboratory tests. Authors contribution

7. Challenges and future scope All the authors make a substantial contribution to this manu-
script. PS, DB, DJ, BC, MS and AK participated in drafting the
Several artificial intelligence techniques were recognized that manuscript. PS, DB, DJ, BC, MS and AK wrote the main manuscript.
mostly are appropriate to forecast how to optimize drilling mud All the authors discussed the results and implication on the
parameters and some questions related to mud. In future work, the manuscript at all stages.
prediction of other models like PVT properties, history, porosity
matching, and other reservoir engineering properties, lithofacies Availability of data and material
will become successful by using the fuzzy-2 logic system. Artificial
intelligence helps in many prospects like Improved Exploration and All relevant data and material are presented in the main paper.
Efficiently, Reduced costs, Greater worker's safety, Predict and
Prevent Spills and Accidents. Funding
AI also has capabilities to enhance seep detection capabilities so
that it can be helpful to discover the reservoir. Many companies are Not Applicable.
also deploying the robot on offshore platforms for deep-sea oper-
ations. AI use used to making the offshore platform smart which Consent for publication
saved time to focus on different projects. On the offshore platform,
the AI-enhanced drone used to deliver the material to the platform Not applicable.
becomes transportation. Droned technology also developed to
detect methane, so chance also developed using AI to detect the oil Ethics approval and consent to participate
and gas reservoir.
Numerous offshore oil and gas industries use artificial intelli- Not applicable.
gence in data science to access the complex data used for oil and gas
exploration and production. Artificial intelligence for oil and gas is Declaration of competing interest
more practical in the field; predictive algorithms monitor the
critical safety systems to help prevent failure and keep workers The authors declare that they have no competing interests.
safe. Sensor hardware and software powered by enables us to use a
considerable amount of data to gain real-time response on the best Acknowledgements
future course of action.
The authors are grateful to School of Petroleum Technology and
8. Conclusion Department of Chemical Engineering, Pandit Deendayal Petroleum
University for the permission to publish this research.
It has been concluded that Artificial intelligence techniques are
implemented in various divisions of industry in the oil and gas References
industry. Its methodology significantly impacts discrete applica-
tions, leading to optimizing risk, optimizing cost, and solving Abdollahi, J., 2007. Analysing Complex Oil Well Problems through Case-Based
Reasoning, 206. Fakultet for ingeniørvitenskap og teknologi.
complex problems. Artificial intelligence techniques were used that Agu, C.E., Lie, B., 2014. SMART SENSORS FOR MEASURING FLUID FLOW USING A
predicts relative permeability curve same as relative permeability VENTURI CHANNEL.
curve determined from actual experiment. That provides accurate Agwu, O.E., Akpabio, J.U., Alabi, S.B., Dosunmu, A., 2018. Artificial intelligence
techniques and their applications in drilling fluid engineering: a review.
permeability data that is distorted during the actual experiment J. Petrol. Sci. Eng. 167, 300e315. https://doi.org/10.1016/j.petrol.2018.04.019.
and overcomes the difficulties involved in the conventional method Ahmadi, M.A., 2015. Connectionist approach estimates gas-oil relative permeability
of determining permeability data. Mentioned artificial intelligence in petroleum reservoirs: application to reservoir simulation. Fuel 140, 429e439.
https://doi.org/10.1016/j.fuel.2014.09.058.
techniques can be used with reservoir simulation software that Ahmadi, M.A., Zendehboudi, S., Dusseault, M.B., Chatzis, I., 2016. Evolving simple-
provides advancement in reservoir management. Artificial to-use method to determine watereoil relative permeability in petroleum
intelligence-based model is a cost-effective and time-saving tech- reservoirs. Petroleum 2, 67e78. https://doi.org/10.1016/j.petlm.2015.07.008.
Akanji, L.T., Ofi, I., 2016. Application of artificial intelligence in oilfield operation and
nology because it reduces the number of laboratory experiments to
intervention. Society of Petroleum Engineers - SPE Intelligent Energy Interna-
determine reservoir properties. Swarm vector optimization, Artifi- tional Conference and Exhibition. Society of Petroleum Engineers. https://
cial neural network, Discrete Element method and Genetic opti- doi.org/10.2118/181116-ms.
mized neural network techniques are used to predict porosity, Al-Anazi, A.F., Gates, I.D., 2010. Support vector regression for porosity prediction in
a heterogeneous reservoir: a comparative study. Comput. Geosci. 36,
which takes less time than the conventional method and gives 1494e1503. https://doi.org/10.1016/j.cageo.2010.03.022.
accurate reservoir data. From different digital tools, we can analyse Al-Fattah, S.M., 2007. Artificial Intelligence Technology Predicts Relative

8
P. Solanki, D. Baldaniya, D. Jogani et al. Petroleum Research xxx (xxxx) xxx

Permeability of Giant Carbonate Reservoirs. Society of Petroleum Engineers Keshavarzi, R., Jahanbakhshi, R., 2013. Real-time prediction of pore pressure
(SPE). https://doi.org/10.2118/109018-ms. gradient through an artificial intelligence approach: a case study from one of
Al-Fattah, S.M., Startzman, R.A., 2001. Predicting Natural Gas Production Using middle east oil fields. Eur. J. Environ. Civ. Eng. 17, 675e686. https://doi.org/
Artificial Neural Network. Society of Petroleum Engineers (SPE). https://doi.org/ 10.1080/19648189.2013.811614.
10.2118/68593-ms. Kumar Chaki, Chironjeet, 2011. Making of an Artificial Intelligent Drill Machine
AlBinHassan, N.M., Wang, Y., 2011. Porosity prediction using the group method of [WWW Document]. URL. accessed 6.21.21. https://www.researchgate.net/
data handling. Geophysics 76. https://doi.org/10.1190/geo2010-0101.1. publication/327510739_Making_of_an_artificial_intelligent_drill_machine.
Aliouane, L., Ouadfeul, S.-A., Boudella, A., 2015. Pore Pressure prediction in shale gas Lea, G., 2015. The struggle to define what artificial Intelligence Actually Means.
reservoirs using neural network and fuzzy logic with an application to Barnett accessed 6.21.21. https://scholar.google.com/scholar?hl¼en&as_sdt¼0%
Shale. Geophysical Research Abstracts. 2C5&q¼Lea%2CþG.%2Cþ2015.
Allain, O.F., Horne, R.N., 1990. Use of artificial intelligence in well-test interpretation. þTheþstruggleþtoþdefineþwhatþartificialþintelligenceþactuallyþmeans.
JPT. J. Petrol. Technol. 42, 342e349. https://doi.org/10.2118/18160-PA. &btnG¼.
Anifowose, F.A., Abdulraheem, A., Al-Shuhail, A., Schmitt, D.P., 2013. Improved 2009 Li, Chaoqun, Jiang, Liangxiao, Wu, Jia, 2009, Distance and.. - Google Scholar
permeability prediction from seismic and log data using artificial intelligence [WWW Document]. J. Inf. &COMPUTATIONAL Sci. 6.2. URL. https://scholar.
techniques. SPE Middle East Oil and Gas Show and Conference, MEOS, Pro- google.com/scholar?hl¼en&q¼Chao%2CþQ.%2CþLiang%2CþJ.þandþWu%
ceedings. Society of Petroleum Engineers (SPE), pp. 2190e2196. https://doi.org/ 2CþJ.þ%282009%29þ%27Distanceþandþattributeþweightedþk-nearest-
10.2118/164465-ms. neighborþandþitsþapplicationþinþreservoirþporosityþprediction%27%
Ansari, H.R., 2014. Use seismic colored inversion and power law committee ma- 2CþJournalþofþInformationþ%26þComputationalþScience%2CþVol.þ6%
chine based on imperial competitive algorithm for improving porosity predic- 2CþNo.þ2%2Cþpp.23-31.#d¼gs_cit&u¼%2Fscholar%3Fq%3Dinfo%
tion in a heterogeneous reservoir. J. Appl. Geophys. 108, 61e68. https://doi.org/ 3As07BIJRM9m8J%3Ascholar.google.com%2F%26output%3Dcite%26scirp%3D0%
10.1016/j.jappgeo.2014.06.016. 26hl%3Den. accessed 6.21.21.
Ashkok, P., Behounek, M., 2018. An artificial intelligence belief system reduces Mohaghegh, S., Arefi, R., Ameri, S., Rose, D., 1994. Design and development of an
nonproductive time. J. Petrol. Technol. 70, 18e20. https://doi.org/10.2118/1018- artificial neural network for estimation of formation permeability. Proceedings
0018-jpt. - Petroleum Computer Conference. Publ by Society of Petroleum Engineers
Asoodeh, M., Bagheripour, P., 2013. Core porosity estimation through different (SPE), pp. 147e154. https://doi.org/10.2118/28237-pa.
training approaches for neural network: back-propagation learning vs. Genetic Mohaghegh, S.D., Hutchins, L.A., Sisk, C.D., 2002. Prudhoe bay oil production opti-
algorithm. Int. J. Comput. Appl. 63, 11e15. https://doi.org/10.5120/10461-5172. mization: using virtual intelligence techniques, stage one: neural model
Ba, A., Hbaieb, S., Mechbal, N., Verge , M., 2009. Faults detection by marginalized building. Proceedings - SPE Annual Technical Conference and Exhibition. So-
particle filters: application to a drilling process. IFAC Proc 42, 342e347. https:// ciety of Petroleum Engineers (SPE), pp. 2763e2777. https://doi.org/10.2118/
doi.org/10.3182/20090630-4-es-2003.00057. 77659-ms.
Bello, O., Teodoriu, C., Yaqoob, T., Oppelt, J., Holzmann, J., Obiwanne, A., 2016. Nooruddin, H.A., Anifowose, F., Abdulraheem, A., 2013. Applying artificial intelli-
Application of artificial intelligence techniques in drilling system design and gence techniques to develop permeability predictive models using mercury
operations: a state of the art review and future research pathways. Society of injection capillary-pressure data. Society of Petroleum Engineers - SPE Saudi
Petroleum Engineers - SPE Nigeria Annual International Conference and Exhi- Arabia Section Technical Symposium and Exhibition 2013. Society of Petroleum
bition. Society of Petroleum Engineers. https://doi.org/10.2118/184320-ms. Engineers, pp. 554e569. https://doi.org/10.2118/168109-ms.
Bloys, Ben, Davis, Neal, Brad Smolen, Bailey, Louise, Otto, Houwen, Reid, Paul, Nunoo, N.A., 2018. Guest editorial: how artificial intelligence will benefit drilling.
Sherwood, John, Fraser, Lindsay, Hodder, Mike, 1994. What influences the J. Petrol. Technol. 70, 14e15. https://doi.org/10.2118/0518-0014-jpt.
choice of fluid?. Olatunji, S.O., Selamat, A., Abdulraheem, A., 2011. Modeling the permeability of
Boomer, R.J., 1995. Predicting production using a neural network (artificial intelli- carbonate reservoir using type-2 fuzzy logic systems. Comput. Ind. 62, 147e163.
gence beats human intelligence). Proceedings - Petroleum Computer Confer- https://doi.org/10.1016/j.compind.2010.10.008.
ence. Society of Petroleum Engineers (SPE), pp. 195e204. https://doi.org/ Panchiwala, S., Shah, M., 2020. A comprehensive study on critical security issues
10.2118/30202-ms. and challenges of the IoT world. J. Data, Inf. Manag. 2, 257e278. https://doi.org/
Desai, M., Shah, M., 2020. An anatomization on breast cancer detection and diag- 10.1007/s42488-020-00030-2.
nosis employing multi-layer perceptron neural network (MLP) and convolu- Rahmanifard, H., Plaksina, T., 2019. Application of artificial intelligence techniques
tional neural network (CNN). Clin. eHealth. 4, 1e11. https://doi.org/10.1016/ in the petroleum industry: a review. Artif. Intell. Rev. 52, 2295e2318. https://
j.ceh.2020.11.002. doi.org/10.1007/s10462-018-9612-8.
Dodiya, M., Shah, M., 2021. A systematic study on shaping the future of solar pro- Shane, Legg, Hutter, Marcus, et al., 2007. Advances in Artificial General Intelligence:
sumage using deep learning. Int. J. Energy Water Resour. 1e11. https://doi.org/ Concepts, Architectures and [WWW Document]. IOP Press. URL. Google Books
10.1007/s42108-021-00114-8. (accessed 6.21.21). https://books.google.co.in/books?hl¼en&lr¼&
Elkatatny, S., Tariq, Z., Mahmoud, M., Abdulraheem, A., 2018. New insights into id¼t2G5srpFRhEC&oi¼fnd&pg¼PA17&dq¼Legg,þS.
porosity determination using artificial intelligence techniques for carbonate ,þHutter,þM.,þ2007.þAþcollectionþofþdefinitionsþofþartificialþintelligence.
reservoirs. Petroleum 4, 408e418. https://doi.org/10.1016/j.petlm.2018.04.002. þ17þ–þ24.&ots¼hBXPmVLCR1&
Elmahbes, R., Quintero, R., Larez, A., 2016. Artificial lift production systems sizing sig¼ohN3nNpbBJYraZPSzDqWzaPwcW0&redir_esc¼y#v¼onepage&
and simulation software. Society of Petroleum Engineers - Abu Dhabi Inter- q¼Legg%2C S.%2C Hutter%2C M.%2C 2007. A collection of definitions of artificial
national Petroleum Exhibition and Conference 2016. Society of Petroleum En- intelligence. 17 – 24.&f¼false.
gineers. https://doi.org/10.2118/183534-ms. Singh, S., Kanli, A.I., Sevgen, S., 2016. A general approach for porosity estimation
Gharbi, R.B.C., Mansoori, G.A., 2005. An introduction to artificial intelligence ap- using artificial neural network method: a case study from Kansas gas field.
plications in petroleum exploration and production. J. Petrol. Sci. Eng. 49, Studia Geophys. Geod. 60, 130e140. https://doi.org/10.1007/s11200-015-0820-
93e96. https://doi.org/10.1016/j.petrol.2005.09.001. 2.
Hosseini, A., Ziaii, Mansour, Rouhani, A Kamkar, , Amin Roshandel Kahoo, Tapias, O., Soto, C.P., Sandoval, J., Perez, H.H., Bejarano, A., 2001. Reservoir Engineer
Gholami, Raoof, Hanachi, J., 2011. Artificial intelligence for prediction of porosity and Artificial Intelligence Techniques for Data Analysis. Society of Petroleum
from seismic attributes: case study in the Persian Gulf | request PDF. Iran. Engineers (SPE). https://doi.org/10.2118/68743-ms.
J. Earth Sci. 3, 134e152. Tarrahi, M., Shadravan, A., 2016. Advanced big data analytics improves HSE man-
Ilkhchi, A.K., Rezaee, M., Moallemi, S.A., 2006. A fuzzy logic approach for estimation agement. Society of Petroleum Engineers - SPE Bergen One Day Seminar. So-
of permeability and rock type from conventional well log data: an example ciety of Petroleum Engineers. https://doi.org/10.2118/180032-ms.
from the Kangan reservoir in the Iran Offshore Gas Field. J. Geophys. Eng. 3, Thakkar, H., Shah, V., Yagnik, H., Shah, M., 2020. Comparative anatomization of data
356e369. https://doi.org/10.1088/1742-2132/3/4/007. mining and fuzzy logic techniques used in diabetes prognosis. Clin. eHealth. 4,
Kaydani, H., Mohebbi, A., Eftekhari, M., 2014. Permeability estimation in hetero- 12e23. https://doi.org/10.1016/j.ceh.2020.11.001.
geneous oil reservoirs by multi-gene genetic programming algorithm. J. Petrol.
Sci. Eng. 123, 201e206. https://doi.org/10.1016/j.petrol.2014.07.035.