Beruflich Dokumente
Kultur Dokumente
Cenk Temizel, Area Energy; Mehdi Nabizadeh and Nematollah Kadkhodaei, International Petro Asmari Company;
Rahul Ranjith, Anuj Suhag, and Karthik Balaji, University of Southern California; Diyar Dhannoon, Texas A&M
University
This paper was prepared for presentation at the SPE Intelligent Oil and Gas Symposium held in Abu Dhabi, UAE, 9-10 May 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
Decision making in waterflooding operations is a crucial process in petroleum oilfield activities where
numerous attributes and uncertainties exist in the complete process. This study investigates the reservoir
management of waterfloods in terms of injection/production practices. A well-organized historical database
that also collects real-time data is especially important in utilization of data-driven methods in the process of
determination of optimum injection/production practices for waterfloods that will result in better recovery
and sweep, which is illustrated in this paper.
Statistics is a strong tool to turn information or data into knowledge when used with care and physical
understanding of the cause-effect relation between attributes and the outcome. Unfortunately, historical
data and learnings from the past cannot be used in an efficient way in oilfield decisions due to the lack of
systematically organized historical data where there is a huge potential of turning terrabytes of data into
knowledge and understanding for improved decisions and results. Historical injection and production data
at pattern level is utilized to determine the optimum injection levels in light of significant factors that affect
the success of a waterflooding displacement process with commercial data analysis tools.
Analysis of injection/production data at associated injectors and producers reveals the optimum injection
levels depending on the significant factors including but not limited to subsurface conformance, number and
location of producers, vintage of wells, completion practices and injection history. The optimum injection
levels change depending on the changing variables that affect the displacement and injection processes,
thus, a real-time data flow from producers and injectors is required to capture and maintain the optimum
operating levels.
The significance of each parameter in this process is obtained in a dynamic manner with real-time feed
of field data and efficiently used to determine the optimum levels of injection at a specified time. Change
of important factors in the process in time is also important by means of adding another dimension on the
relative significance of parameters in the process, thereby shedding light on future decisions.
Introduction
The first step in any engineering design is cost effectiveness. Unless an application is cost effective and
economically favorable, designing levels does not kick in. Any oil reservoir should meet some minimum
2 SPE-187468-MS
technical and economic criteria to be a candidate for a waterflooding technique for oil recovery. During oil
recovery, water flooding is a method that is crucial in maintaining the pressure through oil reservoirs. As
oil is drawn from any reservoir, the pressure of the wells drop. To compensate for this pressure diminution,
water or any other commercially, physically and chemically compatible substance is to be pumped into the
well. However, the original reservoir water composition is expected to be significantly dissimilar from the
composition of accessible water for injection. This alteration in water form results in new chemical reactions
mostly due to the availability of various chemical compositions in them.
Figure 1The cross-section is illustrating how carbon dioxide and water can
be used to flush residual oil from a subsurface rock formation between wells.
Figure 2Different injection and producer pattern designs to maximize oil recovery
deeper. This difference is extremely smaller at 500 m depth and hardly noticeable at 5000m depth. It is
important to mention that salinity difference though measurable, is less that 10% even at the surface. The
scale for salinity measurement is called practical salinity scale (PSS), which is a measure of conductivity
of water at a given temperature. The chemistry of seawater as measured by Karl in part per million (ppm)
are shown in Table 1.
Magnesium Mg 1,290
Sulfur S 904
Potassium K 392
Calcium Ca 411
Bromine Br 67.3
Boron B 4.45
Carbon C 28.0
Fluorine F 13
Silicon Si 2.9
Irrespective of the source of inserted water, treatment is a necessity. The type of injected water governs the
essential treatment, to diminish the effect of suspended matter, microbiological fouling and corrosiveness
of water scale deposition.
6 SPE-187468-MS
Figure 6Pressure changes which could produce scale at different locations during water injection
Obtaining water resource is a grave concern for oil and gas companies. Transporting water to the
production site, separation, treatment and then disposal are costs in addition to scaling corrosion and
emulation value.
Scale formations
Incompatibility of the formation water and injected water results in scale formation inside pipes. Scales are
a source of damage for both production and injection wells due to their contributions to equipment wear,
corrosion and flow restrictions. This damage has a great impact on oil production rates. Scales form from
supersaturation of the minerals dissolved in water if some minerals present in the solution are higher than
the determined solubility of those minerals at the given condition (temperature, pressure, and pH). Scales
form due to precipitation of the excessive amount of salts in the water. In conclusion, scales are formed due
SPE-187468-MS 7
to the mixing of two incompatible water types, yet not the only reason. Scales can typically form as the
water conditions change such as the temperature or pressure change of water.
Figure 8Flow Chart for Optimization of KPIs in Waterflood Operation Khan et al 2013
Scenario 1
Bottom hole pressure constrained well by well using bubble point pressure.
Scenario 2:
8 SPE-187468-MS
Additionally, as studied by Temizel et al (2016), using the production and injection data for the field
it is possible to use data mining operations for obtaining various trends in the sequences. By integrating
this process with Artificial Neural Networks and (or) genetic algorithms, we can change key performance
indicators (KPI) for optimizing the procedure. A full-physics model is derived regards to this optimizing
parameter. By integrating the geological model on simulation software to visualize and characterize the
reservoir, and by running soft constraints on the KPI, optimization is possible. First, we integrate both the
full physics model and the geological model and evaluate the changes with time in the KPI and other the
optimizing parameter. Once this result is obtained we integrate the simulation results with an optimization
software to understand the significance (weighted) of each parameter with respect to the objective function
thus understanding how each parameter would affect the eventual profitability of the wells.
Data-Driven Modeling
With constant surveilliance and improvement of technology, the amount of data being generated from
subsurface reservoirs as well as production facilities is on a rise. Major players are adopting waterflood
optimization using data-driven models.
There are multiple methodologies involved in the generation of these data-driven models, which mainly
include - data gathering, data selection, identification of key attributes, history matching and forecasting.
Data gathering can vary from simple frequent gathering of data to storage of data, reporting, documentation
and integration. It includes the use of smart well technology such as ICVs and regular rate recoding.
Acquisition of data includes data such as well trajectories, well interventions, production/injection data, co-
ordinated maps, well logs & petrophysical models and geological reservoir models. Storage of this data
includes various methods - SQL server integration, volume analysis, IAM, pumping & drilling databases.
The next step in the process involves generation of reports and well status.
Analysis of data is the most integral part of data-driven modeling, as the accuracy of identification
of trends and attributes determines its accuracy. Various methodologies are used in data-driven models
including artificial neural networks, fuzzy logic and capacitance-resistance models. We will look into these
methods in details in the next section.
Attributes are the building blocks of data which should be deliberated on and be chosen carefully as they
will form the instances that will characterize the result conceptually, depending on the nature of the result.
Instances are formed of different combinations of attributes, which are chosen from pre-processed data sets
of the initial data from the field to make informed decisions. The quantity (whole numbers or fractions)
or specification (labels or co-ordinates) is the measurement/identity of the instance. Attribute selection is
wholly dependent on the nature of the result, and can vary from case to case and is subject to the pre-
processed data. The attributes which come close to a more wholesome contribution is given preference over
other attributes of the dataset. Data cleaning and dimensionality reduction can be a solution to handle sparse
data if it were the case to represent the instance. Data cleaning is also a very important step to take care of
missing or faulty data that may be a part of the pre-processed data in order to get the most accurate result.
In case of large amounts of attributes, dimensionality reduction algorithms help to represent the attribute
as part of the instance for getting accurate results. If the data is not handled well, faulty attribute signatures
could result in inaccurate results hindering machine learning predictive algorithms. Decisions are based on
the descriptive, predictive outcome and prescriptive attributes of the data.
SPE-187468-MS 9
Where,
10 SPE-187468-MS
I(tk) is the effective injection at kth time from injector to producer in control volume and ew is the aquifer
influx. Oil production rate is obtained from the Buckley leverett fraction flow model by,
Where, M is the mobility ratio. Average saturation can be defined from material balance as,
The effective water saturation and water production data are used to get standardized saturation values.
The process can be calculated intrinsically and extrinsically. Runs can be randomized using CFD estimation
of key parameter ranges and then used to input them to generate different runs. By crossplots of the various
output parameters, the top 10% (according to user convenience) of data can be selected by cross-examining
and matching with field generated data. From these predictions, it possible to optimize the process. This
method can also be used for data forecasting.
Runo et al. (2015) demonstrated the use of CRM in a field case in the Salym gas field in Russia.
They developed a data gathering and processing unit for active monitoring of mature waterfloods in the
reservoir and developed criteria for decision making in terms of well selection, KPI selection and so on.
Jahangiri (2014) also generated a data-driven method for improving waterflood in the North Sea using
CRM methodology and IAM. Similarly, Fraguiro (2017) developed a multilayer waterflood optimization
tool based on data driven modeling using CRM technology. Another famous tool in data-driven waterflood
optimization is the interwell numerical simulation model (INSIM) that can be be used for calculation of
approximate performance of waterflood for the control volume. The method has been shown by C. Carpenter
(2016).
Overall recovery should be maximized through maximizing each component in the recovery process.
SPE-187468-MS 11
Vertical Conformance
Areal conformance and vertical injection conformance are important factors in the waterflood displacement
process. Figures 10 14 illustrate the differences that a bad and a good vertical injection conformance may
result in.
Figure 11Difference between good injection profile vs bad injection profile in final
saturation distribution for the reservoir simulation model - cross sectional view
Figure 13Injector subsurface conformance and areal conformance are important parameters of successful recovery
Figure 14Evenly distributed injection among the producers leads to better ultimate
recovery in reservoirs with homogeneous rock properties and saturation distribution
SPE-187468-MS 13
Figure 15Workflow: Areal conformance through inter-well connectivity provides better adjustment of injection rates
Prevent over injecting into a single well (that may cause high WCUT, flowing wells, etc.)
Better knowledge of how to take action when high GROSS/WCUT wells exist
Tap more oil in patterns where higher injection rates are possible
Better pattern (and thus reservoir management) with better knowledge of inter-well connectivity
This method works on creating pulses or disturbances on the injectors either by decreasing or increasing
the injection rates and observing the responses on the producers in the same pattern by looking at their run
time pump capacities or gross rates. There might be some complications in uses of these so the engineers
need to be careful in using this method.
Disturbances on injectors may not be measured right away at the producers due to the insufficient
frequency of well tests or runtime pump capacities not reflecting the true responses due to pump or well
conditions. Also, the method is applied such that the disturbance is done one-pattern-at-a-time so that the
signals do not overlap. This can be automated so that the disturbances at patterns at distant locations can
be carried out and then the analysis at pattern-level can be done and the injector and producer connectivity
can be understood. The magnitude and response time can change at producers depending on the degree of
connectivity between the injector and the producers. The scope of this study is to illustrate the methodology
rather than to investigate the details of a specific pattern. The data and information provided is synthetic
and for the sole purpose of illustration, no field data is utilized.
Steps
Magnitude of Injection
This is another important factor that needs to be optimized. It is not easy to find the appropriate amount
of injection on pattern level. However, robust use of historical and real-time data can lead to efficient
injection management in the patterns. It is known that injecting more than the matrix intake rate along with
the heterogeneity and mobility factors can lead to premature breakthrhough and/or recycling of water. To
minimize or prevent these it is important to have an understanding of how the producer is performing real-
time and how much injection from the injectors can the producers handle or produce efficiently without
recycling. Y-function introduced by Yang (2012) is one of these methods that can be utilized to understand
if the producer is performing normally without any recycle.
Figure 24Schematic Diagram for Steamflod and Waterflood Profiles (Yang, 2012)
Details of the Y-function methodology is beyond the scope of this paper and can be further read in the
references below. Brief explanation for the method is provided below and some examples are provided to
illustrate the method. The real-time data coming from the field can be effectively fed to the Y-function
to increase or decrease the injection rates at the pattern levels to prevent or minimize the recycling, thus,
increasing the efficiency of waterfloods.
Fig 25. shows the way Y-function works. It shows the waterflood performance and indicates if the well is
experiencing premature breakthrough, primary production or whether the well is having an expected normal
waterflood performance or not. This is important in understanding the level and amount of waterflood
support the producer is receiving and the status of it. Premature breakthrough happens when the injection
SPE-187468-MS 19
rates are more than required. As reservoirs have high heterogeneity with channels and thief zones, it is hard
to know and optimize the injection levels in the beginning of the waterflood even though engineers have a
feeling of the right, or required amount of injection. Y-function is very useful in identifying the status of
the producer in order to understand the condition of the producer and thus helps the reservoir engineers to
adjust the injection and/or fix other mechanical or near-wellbore factors that may be negatively affecting
the production.
Fig. 26 illustrates a premature breakthrough case as understood from the legend in Fig 25. Injector that
supports this producer can be adjusted to reduce injection to have a better waterflood performance as it
seems that the current injection levels are yielding some premature breakthrough.
Fig 27. Also illustrates a case of premature breakhtough where injection on the injector side should be
adjusted and the curve here should be observed until oil cut improves.
Fig 28. illustrates a case of a successful waterflood where the producer is responding the water injection
in an efficient way after primary production. The injection can be kept at this level for this producer.
Acknowledgement
Authors thank the University of Southern California, Aera Energy (a Shell-ExxonMobil JV), International
Petro Asmari Company (IPAC) and VaalbaraSoft.
References
1. Yang, Z.M.: "A New Diagnostic Analysis Method for Waterflood Performance," SPE Reservoir
Evaluation & Engineering (April, 2009) 341351.
2. Yang, Z.M., "Analysis of Production Decline in Waterflood Reservoirs," SPE 124613 (2009a).
3. Yang, Z.M.: "Production Performance Diagnostics Using Field Production Data and Analytical
Models: Method and Case Study for the Hydraulically Fractured South Belridge Diatomite," SPE
Reservoir Evaluation & Engineering (Dec, 2012) 712724.
4. Yang, Z.M., and Urdaneta, A.: "A Practical Approach to History Matching Premature Water
Breakthrough in Waterflood Reservoir Simulation: Method and Case Studies in South Belridge
Diatomite Waterflood," SPE Reservoir Evaluation & Engineering (2016).
5. Yang, Z.M.: "Clarifying and Improving the Application of Waterflood Analytical Methods in X-
plot Conditions - from Empirical Approach to Analytical Approach,"
6. Merdhah, M. and A. Badr, The study of scale formation in oil reservoir during water injection at
high-barium and high-salinity formation water. 2008, Universiti Teknologi Malaysia, Faculty of
Chemical and Natural Resources Engineering.
22 SPE-187468-MS
7. Olajire, A.A., A review of oilfield scale management technology for oil and gas production.
Journal of Petroleum Science and Engineering, 2015. 135: p. 723737.
8. Morrow, N. and J. Buckley, Improved Oil Recovery by Low-Salinity Waterflooding.
9. Technologies, T. norganic scale formation, prediction and inhibition continues to be a Tomson
Technologies focus area. 2014 [cited 2016 10/29].
10. labratory, n.e.t., Carbon Dioxide Enhanced Oil Recovery. 2010 p. 5.
11. Rose, S.C., J.F. Buckwalter, and R.J. Woodhall, The design engineering aspects of waterflooding.
Stephen C. Rose, John F. Buckwalter, Robert J. Woodhall. Monograph / SPE Henry L. Doherty
series: vol. 11. 1989: Richardson, TX: Society of Petroleum Engineers, 1989.
12. Buckley, S.E. and M.C. Leverett, Mechanism of Fluid Displacement in Sands.
13. Kleppe, P.J., Reservoir Recovery Techniques. 2016.
14. Antonov, J., et al., World Ocean Atlas 2009, vol. 2, Salinity, edited by S. Levitus, 184 pp. US
Gov. Print. Off., Washington, DC, 2010.
15. Turekian, K.K., Oceans. [by] Karl K. Turekian. Foundations of earth science series. 1968:
Englewood, Cliffs, N.J.: Prentice-Hall, [1968].
16. Nasr-El-Din, H.A., et al., Injection of Incompatible Water as a Means of Water Shut-Off. Society
of Petroleum Engineers.
17. Serkan Dursun, Kaan Duman, Tayfun Tuna; "A Workflow for Intelligent Data-Driven Analytics
Software Development in Oil and Gas Industry;" SPE-170859; SPE ATCE, Amsterdam October
2014
18. M. Sayarpour, C.S. Kabir, L.W. Lake, "Field Applications of Capacitance-Resistive Models in
Waterfloods" Published in SPE Reservoir Evaluation and Engineering Journal, December 2009.
19. M. Sayarpour, C.S. Kabir, K. Sepehrnoori, L.W. Lake. "Probabilistic History Matching with the
Capacitance-Resistance Model in Waterfloods: A Precursor to Numerical Modeling." Presented
at the SPE Improved Oil Recovery Symposium, Tulsa, Oklahoma. 24-28th April 2010. SPE
129604
20. C. Temizel, S. Aktas, H. Kirmaci, O. Susuz, Y. Zhu, K. Balaji, R. Ranjith, S. Tahir, F.
Aminzadeh, C. Yegin. "Turning Data into Knowledge: data-Driven Survelliance and
Optimization in Mature Field." Presented at SPE ATCE, Dubai, UAE. 26-28th September 2016.
SPE-181881-MS
21. F. Cao, H. Luo, L.W. Lake. "Development of Fully Coupled two-phase flow based capacitance
Resistance Model (CRM)." Presented at SPE Improved Oil Recovery Symposium, Tulsa,
Oklahoma, 12-16th April 2016. SPE-169485-MS.
22. A. Albertoni, L.W. Lake, "Inferring Connectivity only from well-rate fluctutaions in
waterfloods." Published in Reservoir Evaluation and Engineering Journal, 6(1):616. 2003.
23. A.A. Yousef, P.H. Gentil, J.L. Jensen, L.W. Lake. "A Capacitance Model to inder Interwell
connectivity from production and Injection rate Fluctuations." Published in SPE Reservoir
Evaluation and Engineering, 9(5): 630646. 2006.
24. D.B. Weber, T.F. Edgar, L.W. Lake, L.S. Lasdon, S. Kawas, M. Sayarpour. "Improvements in
Capacitance-Resistive Modeling and Optimization of large scale reservoirs." Presented at SPE
WRM, San Jose, CA. 24-26th March 2009. SPE 121299.
25. H. Zhao, Z. Kang, X. Zhang, H. Sun, L. Cao, A.C. Reynolds. "INSIM: A data-driven Model for
History Matching and Prediction for waterflooding monitoring and management with a field
application." Presented at SPERSS, woodlands, TX. 23-25th February 2016. SPE 173213.
26. R. Mijnarends, A. Frolov, F. Grishko, S. Kryanev, E. Mikhaylenko, E. Nizamutdinov, O.
Plokhotnichenko, I. Surovets, E. Ulyanov, Y. Volokitin, A. Gladkov, M. Belyanushkina, A.
SPE-187468-MS 23
Lvov. "Advanced Data-Driven Performance Analysis for Mature Waterfloods." Presented at SPE
ATCE, Houston, TX. 28-30th September 2015. SPE-174872-MS.
27. H.R. Jahangiri, C. Adler, S. Shirzadi, R. Bailey, E. Ziegel, J. Chesher, M. White. "A data-driven
approach enhances Conventional Reservoir Surveillance Methods for Waterflood Performance
Management in the North Sea." Presented at SPE Intelligent Energy Conference & Exhibition,
Utrecht, Netherlands. 1-3rd April 2014. SPE-167849-MS.
28. M. Fragio, A. Lacivita, J. Valle, M>Marzano, M. Storti. "Integrating a data driven model into a
Multilayer Pattern Waterflood Simulator." Presented at SPE Latin America & Caribbean Mature
Fields Symposium, Salvador, Bahia, Brazil. 15-16th March 2017. SPE-184908-MS.
29. Khan, H., Saputelli, L. A., Carvajal, G. A., Ranjan, P., Wang, F., & Knabe, S. P. (2013, September
16). Multi-Objectives Constrained Waterflood Optimization in Tight Carbonates. Society of
Petroleum Engineers. doi:10.2118/166051-MS
30. Schoeling, L. G., Barnett, G. B., Michnick, M. J., Walton, A. W., Green, D. W., & Willhite, G.
P. (1996, January 1). Development of an Improved Waterflood Optimization Program in the
Northeast Savonburg Waterflood. Society of Petroleum Engineers. doi:10.2118/35367-MS
31. Spath, J., & McCants, S. (1997, January 1). Waterflood Optimization Using a Combined
Geostatistical - 3D Streamline Simulation Approach: A Field Example. Society of Petroleum
Engineers. doi:10.2118/38355-MS
32. Temizel, Cenk, et al. "Production Optimization through Voidage Replacement using Triggers for
Production Rate" Presented at SPE Latin America and Caribbean Heavy and Extra Heavy Oil
Conference, Lima, Peru, 2016. SPE 184131 MS.