You are on page 1of 12

SPE-183688-MS

Multilateral Wells Evaluation Utilizing Artificial Intelligence

Ahmed Buhulaigah and Ali S. Al-Mashhad, Saudi Aramco; Sulaiman A. Al-Arifi, KFUPM; Mohammed S. Al-Kadem
and Mohammed S. Al-Dabbous, Saudi Aramco

Copyright 2017, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Middle East Oil & Gas Show and Conference held in Manama, Kingdom of Bahrain, 6-9 March 2017.

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
Multilateral wells are considered to be an advancement revolution in the petroleum industry. The
employment of multilateral wells ensured higher drainage and productivity of reservoirs through the
utilization of diverse configurations. Achieving higher productivity and maximizing the reach from a
multilateral well has highly improved inflow performance relationship (IPR) compared to that of a
conventional horizontal well under certain conditions. Several analytical models have been developed to
estimate the average oil flow rate of multilateral wells by utilizing reservoir parameters to come up with
decent correlations for better accuracy. These models are accompanied with uncertainties and limitations
due to the complexity of multilateral wells. Artificial Intelligence (AI) techniques have been proven to
predict various parameters associated with high uncertainties in the oil industry. One of these methodologies
is Artificial Neural Networks (ANN) which was utilized in this paper as new approach to predict the average
oil flow rate of multilateral wells though the use of some reservoir parameters along with flowing wellhead
data.
As a comparable method, an analytical model was used to calculate the flow rate from several multilateral
wells to quantify the value of utilizing ANN against other methods or correlations. Borisov's correlation
that was developed for estimating the productivity of multilateral wells of planar configuration was used
to calculate the oil flow rate of the multilateral wells and compared the results against actual average oil
flow rates. Additionally, PROSPER software was utilized to estimate some wells’ parameters including
Productivity Index (PI) and flowing bottomhole pressure (FBHP) for oil rate calculations.
Rigorous statistical error analyses have been obtained from ANN method and Borisov's correlation. The
overall regression correlation coefficient was calculated to be 0.97 for ANN which shows a strong matching
between predicted and actual field values with an overall absolute error of 7.85%. High divergence was
found between oil rate calculated from Borisov's correlation and the actual average oil rate with an error
greater than 50%. This indicates the actual advantage of the ANN method against other correlations.
This paper discussed a new method for predicting average oil flow rates for multilateral wells using
surface and reservoir parameters obtained from field data via the employment of Artificial Intelligence
modeling. A model was constructed for enhancing the prediction of oil flow rate for multilateral wells and
resulted in a great prediction accuracy proved by field data comparison.
2 SPE-183688-MS

Introduction
A multilateral well is a revolutionary transformation of horizontal wells. It is a horizontal well with several
branches radiated from the motherbore. Multilateral wells are considered to be magnificent contribution
to the oil industry since the 1990's. Technology Advancement for Multi-Laterals (TAML) is a well-known
methodology to categorize multilateral wells by their complexity and functionality. The categories of
multilateral wells include 7 different levels which are: level-1 openhole junction, level-2 cased-hole exit,
level-3 junction with connection but no seal, level-4 sealed junction, level-5 mechanical sealed junction
with reduced inside diameter, level-6 mechanical sealed junction with full ID and level-7 downhole splitter.
The implementation of this technique allowed higher drainage of reservoirs utilizing diverse
configurations. Also, it helps in increasing production of productable reserves by combining multiple
targets which are separately uneconomic, achieving high productivity in a confined space which would
limit horizontal well length, improve areal and vertical sweep, and reduce gas and water handling costs
through minimizing gas and water coning. Inflow performance relationship (IPR) for a multilateral well
is influenced mainly by wellbore's rock and fluid characteristics. There are various methods to evaluate
multilateral well performance such as: cross sectional area, formation anisotropic properties (heterogeneity)
and formation thickness. Multilateral wells are superior to horizontal ones in reservoirs with geological
obstacles influencing horizontal drilling. Reservoirs with large conductive faults where the water cut is
relatively high will have higher productivity indices if drilled as multilateral wells. This will reduce the
waterflooding effects as well.
Artificial intelligence (AI) is considered to be a powerful technique by which algorithms are manipulated
to intelligently resolve sophisticated dilemmas. AI techniques have been implemented to enhance key
parameters’ prediction affiliated with high uncertainties. One of the well-known techniques is the
Artificial Neural Networks (ANN) which consists of simple components which function in parallel. This
extraordinary methodology was inspired by the human being nervous system in which neural network is
achieved by adjusting the values of the connections between its components. Likely, ANN is operated
relying on intelligent adjustment between input data with the outcome by minimizing errors associated with
the tested parameters. In this paper, ANN tools are employed to predict the oil flow rates for 29 different
multilateral wells. The paper shows an outstanding correlation between predicted results and actual field data
with an average error less than 10%. Furthermore, for comparison with well-known correlation, productivity
indices (PI) were estimated for intelligent wells equipped with Permanent Downhole Monitoring System
(PDHMS); applying Borisov's correlation (Equation-1):

Equation – 1

Where FBHP is taken from PDHMS gauges, F is 4, 2, 1.86, 1.78 for n=1,2,3,4, respectively.
Multilateral wells with PDHMS gauges were considered since these gauges can provide flowing
bottomhole pressure (FBHP) in psi.

Literature Review
In literature, there are many studies conducted on multilateral wells that can be categorized into two groups
which are analytical flow & productivity models and field applications. Some of the presented flow and
productivity models that are discussing the evaluation of horizontal and/or multilateral well performance are
Giger, Borisov, Raghavan, Economides, Retnanto, Larsen, Salas, Smith, Guo and Zarea. Other published
papers were conducted to assess the field application of multilateral are like Hall, Longbottom, Stalder, Lee
and Yaliz.
SPE-183688-MS 3

Analytical Flow & Productivity Models


For analytical flow and productivity models, Borisov (1984) introduced one of the earliest models for oil
production using horizontal and deviation wells through assuming a constant drainage pressure. Giger et
al. (1984) presented an analytical formula for estimating the initial productivity index and water coning
in horizontal wells. Raghavan et al (1993) introduced a method to evaluate the productivity of multiple
drainholes or fractured horizontal wells for homogenous reservoirs assuming steady state flow condition.
Their method is function of reservoir variables and number of radials or fractures. Raghavan et al (1995)
developed a mathematical model to evaluate the productivity of multilateral completions in transient
pressure behavior. Incorporated with numerical model which helps in evaluating the changes in reservoir
properties in addition to the influence of interference and boundaries, they found that the productivity of
multilateral could be improving by selecting the proper distance between the laterals. Economides at al.
(1994) came up with an analytical formula to assess the well performance of multiple horizontal wells of
arbitrary direction which is applicable for transient, mixed and no flow boundary conditions. Retnanto et
al (1996) delivered a study to estimate the multilateral well performance with using simulation model for
several lateral configurations including planar, multi-planar, dual opposing laterals and stacked laterals.
They calculated the optimum spacing, lateral lengths, and number of sidetracks for different reservoir
conditions. Larsen (1996) presented a model to calculate the productivity and skin factor for multilateral
wells regardless of fractures availability with the bottomhole pressure. Salas et al (1996) studied the
challenges of predicting multilateral well performance with different configuration and introduced analytical
and numeric simulation methods. The analytical method was developed to evaluate the productivity of the
multilateral wells and the results showed that the productivity is dependent on wellbore geometry. Numeric
simulation method was used to assess multilateral well performance and the issues of gas and water coning
in heterogeneous reservoir. The result showed productivity of multilateral lateral sustain higher from long
term oil production than single horizontal wells. Also, adding more laterals to the existing wellbore will
increase the production even in the heterogonous reservoir since more contact to the unswept oil area will
be reached with the laterals. Another analytical method to predict the productivity and drainage efficiency
of multi-lateral well arrangements has been presented by Smith et al. (1997). They found out it would be
more confidently approached when calibration data from new field are made available and more reliable
model must be developed to assess multiphase flow in the wellbore. Guo et al. (2006) presented an analytical
model with utilizing a computer stimulator to analyze the multilateral wells’ deliverability by linking the
production section, hydraulics and curve in wellbore sections with the inflow performance in each lateral.
Zarea (2011) presented an analytical model and procedure to predict and optimize the well performance
of advanced multilateral wells that are equipped with intelligent completions. They evaluated horizontal
wells performance and pressure decrease in the lateral across inflow control valves in single and two phase
flow conditions.

Field Applications of Multilateral Wells


Regarding the field applications of multilateral wells, Hall et al. (1996) presented a study to optimize a 5-
spot waterflooding in multilateral horizontal wells. They reviewed the design and application of multilateral
wells and found out a short radius have a positive effect in increasing the production. Longbottom (1996)
discussed a development project of multilateral wells with different completions. The objective of his project
was to develop a multilateral system that connects the lateral liner to the casing in the motherbore, permits
lateral intervention, control the flow between the laterals and motherbore with sealing the junction. He
concluded that the system is feasible technically an operational with more effort required to develop a
system that could overcome some limitations like establishing multilateral junction. Stalder et al. (2001)
discussed the effect of multilateral designing on the productivity, ultimate recovery and operation cost in
field development with different completion design including Gullwing, Triple, Crow's Foot Triple and
Pitchfork dual lateral wells. They concluded with the fishbone well design is expected to increase the
4 SPE-183688-MS

productivity and ultimate recovery comparing with other completions. Lee et al. (2001) presented a study to
optimize the well deign a cost through utilizing numerical simulation to improve the productivity by 60% at
5-10 incremental cost comparing with single horizontal wells. Yaliz et al. (2005) conducted an experimental
study through utilizing an electrical analog apparatus and the results with the published analytical models
of the multilateral well and results showed a good matching between them.

Recent Works using Artificial Intelligence


In regard to artificial intelligence application in rate and productivity index prediction, Alajmi et al. (2015)
conducted a study to predict choke performance and well testing validation using Artificial Neural Network
(ANN). From 595 actual data sets for 31 wells, they calculated critical and subcritical multiphase choke
performance with using five existing correlations and model the choke performance with ANN to predict
the rate. Comparing the estimated and predicted results against actual values, ANN has the least absolute
error percentage which proves its reliability and accuracy in predicting the choke performance in order
to estimate the oil flow rate as a function of choke size and operational condition. Alarifi et al. (2015)
presented a study to predict the productivity index for a hundred oil horizontal wells using different
artificial intelligence techniques including ANN, fuzzy logic and functional network. They also calculate
the productivity index by using five well known correlations. Artificial Intelligence models have fewer
errors than existing correlations in estimating the productivity index which proves its reliability as powerful
predicting tool.

Artificial Intelligence: A Modern Approach for Oil Industry


AI is the state-of-art system which has lots of definitions by scientists and engineers started back in
1945. Two definitions of which AI are clearly defined at the desired interest. Haugeland (1985) defined
AI as "The exciting new effort to make computers think, machine with minds, in the full and literal
sense". Partridge (1991) has defined AI as "a collection of algorithms that are computationally tractable,
adequate approximations of intractably specified problems". AI can sometimes simulate human intelligence
through solving problems by observing other parameters utilizing new methods or approaches. One of the
approaches to perform and design AI is to define certain agents for a specific problem. In which specific
inputs that are related to one single aspect can be integrated to solve the problem.
In order to design AI model, four components to define a specific agent shall be quantified, percepts,
actions, goals and the environment. For example, a refinery controller is considered to be an agent in which
it has pressure and temperature readings as percepts and should have the functionality to be open or closed
as actions in order to maximize safety as a goal in refinery environment.

Artificial Neural Network (ANN)


In this paper, the concept of AI will be applied into the desired outputs using ANN. An artificial neuron is a
computational model stimulated by the natural neurons. These neurons receive signals through an algorithm
or a function which activate them accordingly to produce the outputs. When the signals received are strong
enough to exceed a certain limit, the neuron is activated. This signal might be sent to another input, and
might activate other neurons. Artificial neurons consist of inputs, which are multiplied by weights and then
computed by a mathematical function. Accordingly, another function computes the output of the artificial
neuron will be applied to a certain limit similar to the inputs.
The inputs have specific weights in order to produce the desired outputs. The strength of the input will
be related to the weight of the artificial neuron. The strongest input will be given the highest weight of an
artificial neuron in which weights could be positive and negative and hence the weight of the neuron for a
specific input can be adjusted to produce the desired outputs through an activation algorithm or a function
as in Figure-1. Due to the large number of neurons, the weights are difficult to be set manually. Several
SPE-183688-MS 5

algorithms are utilized to ease this process. This process of adjusting the weights is called training followed
by testing process to qualify the optimized model. Since the first neural model by McCulloch and Pitts
(1943), hundreds of different models have been developed and considered as ANNs. The main difference
between models is the programmed function, inputs and training algorithms. In this work, one type of ANN
will be utilized that is being trained using the back propagation algorithm (Rumelhart and McClelland,
1986) for training the appropriate weights, since it is one of the most common models used in ANNs, and
many others are based on it. There are a wide variety of ANNs that are used for engineering purposes and
the oil industry specifically, such as, pattern recognition, validating, forecasting, and data compression. The
main objective in this paper to use the ANN technique to predict and validate one important parameter,
which is required in field simulation models and oil flow rate for multilateral wells. The below is one of the
training algorithms that has been utilized in this work, to produce a calculated oil flow rate, which is the
Gradient Descent with Adaptive Learning Rate Backpropagation Algorithm.

Figure 1—Simple Artificial Neuron Network (ANN)

Gradient Descent with Adaptive Learning Rate Backpropagation Algorithm


Backpropagation is a training technique for feed forward neural networks that consists of an iterative
optimization, called the error function, representing a measure of the performance of the network. It is the
algorithm used toward building the ANN network. The error function E used in this model is defined as
the mean square sum of differences between the values of the network output and the actual target values,
calculated for the complete set of data (Equation-2):

Equation – 2

tj and aj are defined as the target and actual values of output neuron j, NL is the number of output neurons
and L being the number of layers. During the training process, a training pattern has a certain set of samples
and each sample consists of a group of inputs and equivalent target output. The pattern is accessible to
the network in an iterative mode. The suitable weight adjustments are being performed during the training
process to adapt the network to get the desired outputs.
The optimum reduction in the error function is carried out using a gradient descent technique. The
necessary adjustments to the weights of the network for each iteration (n) are obtained by calculating the
partial derivative of the error function, according to each weight wij, which gives a direction of the steepest
and smallest descent.
6 SPE-183688-MS

Model Initialization and Data Processing


A group of 174 data sets from several multilateral wells was collected from one of the onshore fields in the
Middle East. The data set consists of surface measurements and downhole parameters for the reservoir. The
laterals are drilled in a fishbone shape and some are drilled on the same plane. Maximum lateral length for
single multilateral well exceeds 12000 feet. Multi-rate testing was conducted on every individual lateral and
flowing wellhead pressure (FWHP) was recorded. The data input parameters as in Figure-2 are: effective
length Le (in feet), open-hole size (in inch), choke size (in percent), reservoir pressure PR (in psi), flowing
wellhead pressure FWHP (in psi), average permeability K (in md) and number of laterals. Some assumptions
were made to proceed with the calculations, such as the average permeability was considered per well,
overall well productivity index value was assumed to be the same as for each lateral. The above mentioned
inputs were integrated into ANN model feeding with input data, and simulating neurons to result into the
desired output, which is the lateral average oil rate. This is what is called ANN.

Figure 2—ANN inputs and outputs

The algorithm used to calculate the average oil rate per lateral is the Gradient Descent with Adaptive
Learning Rate Backpropagation through utilizing Artificial Neural Network (ANN). Seventy percent (125
data sets) of the data was used for training, in which inputs and outputs were provided to build the
relationships into the network. The remaining 30% (49 data sets) was used for testing and final model
in which the input data were provided only with hiding the output, and then the prediction results were
compared with the output of the testing data. Moreover, the selection of training and testing data was purely
random. It is worth mentioning that all data used as an input can be known without any rate tests, just from
basic reservoir and well properties such as permeability, reservoir pressure.
A statistical analysis was done on the data sets in terms of min, max and standard deviation (S.D.) as
shown in Table-1.

Table 1—Statistical analysis for the data set

K avg Le Hole size FWHP Choke size SBHP (or PR) Avg Oil rate

Unit mdarcy feet inch psi % psi bbl/day

Min 3 2794 5.5 276.8 5 1831 972

Max 253 12335 8.5 1621 100 2754 9350

Mean 22.3 6951 5.953 919.2 36.4 2244.2 4803.7

S.D 26.1 2327.3 0.4 242.5 19.4 209.3 2275.8

Range 250 9541 3 1344.1 95 923 9630


SPE-183688-MS 7

Borisov's Model for Multilateral Wells


Other than ANN method, different comparable method has been utilized in this work to quantify the value
of utilizing ANN against other methods or correlations. Borisov's correlation (Equation-1) for single phase
oil multilateral (planer) well has been used as another method. Multiple reservoir properties were inputted
in Borisov's model to calculate the rate for PDHMS wells. For wells with no or defective PDHMS installed,
the latest rate test data, such as flowing wellhead pressure (FWHP) was used along with Gas Oil Ratio
(GOR) and water cut to estimate PI. Productivity index was estimated after building and calibrating the well
models for the multilaterals, in which as mentioned above, the PI value for individual lateral was assumed
to be equal to the overall PI value for the well and same for the overall rate. This estimation was done using
PROSPER software. This estimated value was used to calculate the flowing bottom-hole pressure (FBHP)
in psi, hence utilize it in Borisov's correlation to calculate the rate through productivity index equation
(Equation-3). FBHP, at datum reference, was taken from PDHMS gauge was corrected at the middle of the
motherbore with the obtained oil gradient.

Equation – 3

Results and Discussion


The oil production rates from multilateral wells' data have predicted through utilizing the AI and the results
were evaluated using statistical error analysis through computing: average percentage relative error, average
absolute percentage relative error, standard deviation of the absolute errors as well as coefficient of the
correlation.
The percentage relative deviation
Equation – 4
i =1, 2, 3, …., n
The average percentage relative error (Er) is computed with the following equation:

Equation – 5

The average absolute percentage relative error (Ea) is calculated by:

Equation – 6

Spread of data around the mean is measured by the standard derivation (S), which is defined by following
equation:

Equation – 7

Where;

Equation – 8

The correlation coefficient (R) which represents the success of regression analysis to reduce the standard
derivation is obtained by:
8 SPE-183688-MS

Equation – 9

Table 2 and summarizes the results of the statistical error analysis obtained from AI and Borisov method
comparing with the actual average oil rate test. While the errors calculated from Borisov are high indicating
the divergent degree from actual rates, they are very low in the predicted ANN indicating the accuracy of this
method. Figure 3 – 6 comparing the results of the actual average rate Vs predicted data utilizing Artificial
Intelligence and Borisov's correlation. The overall regression correlation coefficient was calculated to be
0.97 for ANN, which shows a strong correlation between predicted and actual field values with an overall
error of 7.85%. Calculated data with Brasov's error was found to be greater than 50% with a coefficient
of 0.3, which shows weak relationship between the calculated rates from Borisov and actual rates. This
indicates the actual advantage of the ANN method against other correlations due to minor made assumptions.

Table 2—Statistical error analysis obtained from AI

Predicted from ANN Calculated


Training Test Overall from Borisov

Er -0.48 -2.93 -1.17 30

Ea 5.89 12.83 7.85 50.4

Min Ea 0.04 0.07 0.04 3.4

Max Ea 38 39 39 143

S 9.4 17.1 12.1 53.8

R 0.986 0.914 0.968 0.3

Figure 3—Actual Average Oil Flow Rate vs. Predicted Rate for Training Data Using ANN
SPE-183688-MS 9

Figure 4—Actual Average Oil Flow Rate vs. Predicted Rate for Test Data Using ANN

Figure 5—Actual Average Oil Flow Rate vs. Predicted Rate for Overall Data Using ANN
10 SPE-183688-MS

Figure 6—Actual Average Oil Rate vs. Calculated Rate with Borisov Model

Conclusion
• Predicting oil flow rate for multilateral wells is achieved utilizing ANN.

• Surface measurements and reservoir properties were used in ANN to calculate average oil flow
rate, with an error of less than 10% (7.8%) against actual rates.
• ANN proved its accuracy within the boundary of the data set used against other correlations, such
as Borisov's correlation in which the correlation gave an error more than 50% between calculated
and actual rates.
• All data used for the model are accurate and taken from the reservoir, and well properties and no
well test or core analysis required retrieving them.
• Further work is to be carried out to predict the rate and contribution from each lateral, using
accessorized downhole gauges and smart well completion technology.

Acknowledgement
The authors would like to thank and acknowledge Saudi Aramco; Northern Area Production Engineering
and Well Services Department (NAPE&WSD) for their permission to publish this paper. Appreciation is
also extended to NAPE&WSD technical review committee for their effort and feedback in reviewing this
work.

Nomenclature
AI : Artificial Intelligence
aj : The actual response value of output neuron j
ANN : Artificial Neural Networks
B : Oil Formation Volume Factor, bbl/STB
Ea : The average absolute percentage relative error
Ei : The percentage relative deviation
Er : The average percentage relative error
SPE-183688-MS 11

FBHP : Flowing Bottom-hole Pressure, psi


FWHP : Flowing Wellhead Pressure, psi
GOR : Gas Oil Ratio, scf/STB
h : Reservoir thickness, feet
IPR : Inflow Performance Relationship
k : Permeability, md
L : The number of layers
Le; Lateral Effective Length, feet
MBOD : Thousand Barrels of Oil per Day
µ : Oil Viscosity, cP
NL : The number of output neurons
PDHMS : Permanent Downhole Monitoring System
PR : Reservoir Pressure, psi
Qactual : The Actual oil flow rate, MBOD
Qestimated : The Estimated or calculated oil flow rate, MBOD
Qaverage : The average actual oil flow rate, MBOD
R : The correlation coefficient
re : Drainage Radius, feet
rw : Wellbore Radius, feet
S : The Standard Deviation
TAML : Technology Advancement for Multi-Laterals
tj : The target response value of output neuron j

References
1. AlAjmi, M. D., Alarifi, S. A. and Mahsoon, A. H., (2015): "Improving Multiphase Choke
Performance Prediction and Well Production Test Validation Using Artificial Intelligence: A New
Milestone," Paper SPE 173394, presented at the SPE Digital Energy Conference and Exhibition
held in Texas, USA, 3-5 March.
2. Alarifi, S. A., AlNuaim, S. and Abdulraheem, A., (2015): "Productivity Index Prediction for
Oil Horizontal Wells Using Different Artificial Intelligence Techniques," Paper SPE 172729,
presented at the SPE Middle East Oil & Gas Show and Conference held in Manama, Bahrain,
8-11 March.
3. Choi, S. K., Ouyang, L. B. and Huang, W., (2008): "A Comprehensive Comparative Study on
Analytical PI/IPR Correlations," Paper SPE 116580, presented at the SPE Annual Technical
Conference and Exhibition held in Colorado, USA, 21-24 September.
4. Economides, M.J., Brand, C.W. and Frick, T.P., (1994): "Well Configurations in Anisotropic
Reservoir," Paper peer approved Paper SPE 27980, presented at the University of Tulsa
Centennial Petroleum Engineering Symposium held in Tulsa, 29- 31 August.
5. El-Sayed, A. H., Al-Awad, M. N., Al-Blehed, M. S., and Al-Saddiqui, M. A., (2000):
"Productivity and Economical Feasibility of Multilateral Wells in Comparison of a Horizontal
Well," Engineering Journal of the University of Qatar, Vol. 12, 2000, pp.
6. El-Sayed, A. H. and Amro, M. M., (1999): "Production Performance of Multilateral Wells,"
Paper SPE/IADC 57542, presented at the SPE/IADC Middle East Drilling Technology
Conference held in Abu Dhabi, UAE, 8-10 November.
7. Fiesler, E. and Moreira, M. (1995): "Neural Networks with Adaptive Learning Rate and
Momentum Terms," Retrieved April 2015, from http://infoscience.epfl.ch/record/82307/
files/95-04.pdf.
12 SPE-183688-MS

8. Gershenson, Carlos.: "Artificial Neural Networks for Beginners," Retrieved April 2015, from
https://datajobs.com/data-science-repo/Neural-Net-[Carlos-Gershenson].pdf.
9. Giger, F.M., Reiss, L.H. and Jourdan, A.P., (1984): "The Reservoir Engineering Aspects of
Horizontal Drilling," Paper SPE 13024, presented at the 59th Annual Technical Conference and
Exhibition held in Houston, Texas, 16- 19 September.
10. Guo, B., Ling, K. and Ghalabor, A.L., (2006): "A Rigorous Composite-IPR Model for
Multilateral Wells," Paper SPE 100923, presented at the SPE Annual Technical Conference &
Exhibition held in San Antonio, TX, September 24 -27.
11. Hall, S.D., (1996): "Multi-Lateral Horizontal Wells Optimizing a 5-Spot Waterflood," SPE
35210, presented at the SPE Penmam Basin Oil & Gas Recovery held in Midland Tx, USA,
March. 27-29.
12. Lee S.S., Veeken, C.A. and Frens, A.M., (2001): "Multi-Lateral Well Modelling to Optimise Well
Design and Cost," SPE-68728, presented at the SPE Asia Pacific Oil and Gas Conference and
Exhibition held in Jakarta, Indonesia, 17-19 April 2001.
13. Longbottom, J.R., (1996): "Development and Testing of a Multi-Lateral System," Paper SPE
35545, presented at the European Productions Conference held in Stavanger, Norway, April
16-17.
14. Pasicznyk, A., 2001: "Evolution toward Simpler, Less Risky Multilateral Wells," Paper SPE/
IADC 67825, presented at the SPE/IADC Drilling Technology Conference held in Amsterdam,
The Netherlands, 27 February- 1 March.
15. PROSPER. Version 12.1. 2013. Edinburgh, UK: Petroleum Experts Ltd.
16. Raghavan, R. and Joshi, S. D., (1993): "Productivity of Multiple Drainholes or Fractured
Horizontal Wells," Paper SPE 21263 presented at SPE Eastern Regional Meeting held in
Columbus, OH, 31 October.
17. Retnanto, A., Frick, T.P., Brand, C.W. and Economides, M.J., 1996: "Optimal Configurations of
Multiple-Lateral Horizontal Wells" SPE 35712, presented at the SPE Western Regional Meeting
held in Anchorage, USA, May 22 - 24.
18. Russell, S. J., and Norvig, P., (1995): "Artificial Intelligence: A Modern Approach," Prentice-
Hall, Inc., New Jersey, USA, First Edition 1995.
19. Salas, J.R., Clifford, P.J. and Jenkins, D.P., (1996): "Multilateral Well Performance Prediction,"
Paper SPE 35711, presented at the Western Regional Meeting held in Anchorage, Alaska, 22-24
May.
20. Smith, S.J., Tweedie, A.A. and Gallivan, J.D., (1997): "Evaluating the Performance of Multi-
Lateral Producing Wells: Cost Benefits and Potential Risks," Paper SPE 38974, presented at the
5th Latin American and Caribbean Petroleum Engineering Conference and Exhibition held in Rio
de Janeiro, Brazil, August 30.
21. Stalder, J. L., York, G.D., Kopper, R. J., Curtis, C. M. and Cole, T. L., (2001): "Multilateral-
Horizontal Wells Increase Rate and Lower Cost per Barrel in the Zuata Field, Faja, Venezuela"
Paper SPE 69700, presented at the 2001 SPE international Thermal Operations and Heavy Oil
Symposium held in Porlamar, Margarita Island, Venezuela, March 12-14.
22. Yildiz, T., (2005): "Multilateral Horizontal Well Productivity," Paper SPE 94223, presented at the
SPE EUROPEC/EAGE Annual Conference and Exhibition held in Madrid, Spain, 13-16 June.
23. Zarea, Marwan and Zhu, Ding, 2011: "An Integrated Performance Model for Multilateral Wells
Equipped with Inflow Control Valves," Paper SPE 142373, presented at the SPE EUROPEC/
EAGE Annual Conference and Exhibition held in Vienna, Austria, 23-26 May.

You might also like