SPE 112246

Rapid Model Updating with Right-Time Data - Ensuring Models Remain Evergreen for Improved Reservoir Management Stephen J. Webb, David E. Revus, Angela M. Myhre, Roxar, Nigel H. Goodwin, K. Neil B. Dunlop, John R. Heritage, Energy Scitech Ltd.

Copyright 2008, Society of Petroleum Engineers evergreen and providing the most up-to-date basis for the This paper was prepared for presentation at the 2008 SPE Intelligent Energy Conference making of important reservoir management decisions. and Exhibition held in Amsterdam, The Netherlands, 25–27 February 2008.

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 Introduction have not been reviewed by the Society of Petroleum Engineers and are subject to Since the early days of reservoir simulation, history correction by the author(s). The material does not necessarily reflect any position of the 1 Society of Petroleum Engineers, its officers, or members. Electronic reproduction, matching has been identified as one of the best methods 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 of validating a reservoir model’s predictive capabilities. abstract of not more than 300 words; illustrations may not be copied. The abstract must Often long periods of time have been spent adjusting the contain conspicuous acknowledgment of SPE copyright. reservoir description so that the reservoir simulator’s calculated results match the observed data from the reservoir. Unfortunately, due to limitations of data Abstract availability, computer hardware performance and History matching reservoir models to production data has mantime, history matching has been a discrete process been a challenge for asset teams since the early days of which is performed only at certain stages in a reservoir’s reservoir simulation. Keeping these models evergreen as life cycle rather than as a continuous process that updates production data continues to arrive, knowing when a re- the reservoir model as new data arrives. Although recent history match is required and being able to re-history advances in assisted matching technology2 have shown match easily and efficiently is also a major challenge there is potential to cut the time required to achieve which is often not addressed in a timely manner. This history matches, often the reservoir model is out of date need is becoming even more pressing as real-time before the history matching process has been completed reservoir performance data is increasingly available. because the production data used in history matching is Decisions can now be made with the support of the good frozen in time when the process is started and is not quality real-time data from the reservoir usually in the updated over the often many months that history matching form of pressure data from downhole gauges and rate data can take. from multiphase meters. With the advent of digital oil field technology3,4,5, data With the recent integration of existing technologies, a is now available from the reservoir in real-time. In rapid model updating workflow is now possible. The particular, downhole pressures6 and multiphase surface history matching workflow that was once a discrete flow rates7 can now be measured directly and process can now be a computer assisted continuous continuously. This real-time data is now routinely used process. Using statistically-based assisted history- for reservoir monitoring, flow assurance calculations, matching technology in conjunction with real-time data production optimization8,9,10 and for reservoir engineering acquisition, data monitoring, and reservoir simulation analysis11. However, despite many good intentions12,13, software, new production data can be quickly assimilated reservoir modeling has been slow to take up the use of into the reservoir model. Real-time data from the field real-time production data to condition its models. While measurement devices is filtered for consumption by the there have been some promising projects using Ensemble reservoir modeling software and compared with the Kalman Filtering to update reservoir models14,15, these forecasts from the reservoir simulator to determine if re- have mostly been research oriented and have not yet made history matching is required. The new data can be added it into mainstream use. to the history file and the model (revised if necessary) can be used in operational performance optimization. The approach taken in this paper is different from those reported previously in that it builds upon assisted This rapid model updating workflow can be run semi- history matching technology that is already in the automatically on a continuous basis as new production commercial domain and is in widespread use. First, a data is gathered, thereby keeping reservoir models process framework is established in which a history match of a reservoir model can be obtained. This process, which 2 SPE 112246 will be summarized in this paper and is described in detail and Bayesian response surface elsewhere2,16, allows the engineer to chose parameters in modeling2,16,17,18,19,20,21,22,23,24,25,26,27,28. the reservoir model that the assisted history matching software can sensitize so that it can explore which The technology used in this work is the advanced parameters need to be changed and how they need to be linear Bayesian tool, EnABLE™. It can be used with a changed to achieve a history match. The process also number of different reservoir simulators and is designed produces an estimate of the uncertainty in the history to assist in the process of identifying multiple acceptable match. While it is obtaining history matches, the process history matches by using a structured workflow process can be used to forecast reservoir performance and define which allows many parameters to be automatically confidence intervals in the forecast performance. These modified simultaneously. Advanced experimental design confidence intervals can then be used as a measure of the methods and linear Bayesian statistical routines are used forecast’s accuracy as new reservoir performance data which permit the use of objective data, subjective opinion arrives. If the new data lies within the confidence and other indirect information in specifying a prior intervals, the model is forecasting field performance to distribution29. A statistical estimator based on a response within the uncertainty of the modeling system. If the new surface model (the Estimator) is created as a proxy for the data lies outside the confidence intervals, either the field full reservoir simulation model and is updated after each is not being operated as forecast (and the simulator’s simulation run. By using the proxy to approximate the operating constraints need updating) or the model itself simulator, extensive exploration of the simulated requires updating. The process framework provides an responses across the solution space may be performed environment where this can be done in an almost without the computational expense of making a complete automatic fashion. simulator run in each evaluation. Starting with a set of scoping runs designed to broadly explore the entire This process will be illustrated by an example based solution space followed by additional “most informative on a large mature North Sea reservoir where the rapid runs” (whereby the regions of the parametric space where model updating process is illustrated by repeatedly the proxy model has the greatest uncertainty are explored) assimilating new production data into the history a reasonably predictive estimator is built. A number of matching process, thereby keeping the model “evergreen” history match solutions are then obtained more rapidly and as up to date as possible for field operating and than would be possible by simply running a large number development decisions. of simulation cases. Fit for purpose history matching and confidence interval estimation using the workflow Rapid Model Updating Process Components described here typically is accomplished with the use of a The rapid model updating process is comprised of the number of reservoir model runs of the order of 100 to 300 runs. Methods that do not employ an estimator are following components: 30 variously reported to require a thousand or more runs . • Statistically-based assisted history-matching The workflow process is shown in Figure 1: technology 1. Identify study objectives. • Robust full field black oil and compositional 2. Setup simulation model as usual. reservoir flow simulator 3. Identify parameters (modifiers) for investigation in the history match (and predictions and optimization31, if • Real-time field monitoring and production anticipated) and their ranges of data management system investigation.

4. Make modifications to the simulator input These technology components are directly interfaced with data to sensitize these parameters. each other to provide a comprehensive and integrated 5. Generate a set of ‘scoping runs’. These total system which is described in the subsequent sections runs, made using an experimental design of this paper. procedure, provide the initial basis for the

Estimator model. Typically 25 runs are used Assisted History Matching for this step, although less may be sufficient The availability of powerful and inexpensive computing for small numbers of modifiers. Multiple hardware over the last decade has led to the development runs can be submitted for simultaneous of software tools that can greatly help the history computation, if multiple processors are matching process. Additionally these tools allow for the available. inherent uncertainty in the reservoir description and 6. Import the historical data on well indeed the production data, and reflect that uncertainty in performance and production. the reservoir performance forecasts used for operational 7. Validate: decide by inspection whether the and field development decision-making. The technologies model parameterization has the potential to that have been employed are many, and include gradient lead to a history match of sufficient methods, simulated annealing, evolutionary algorithms accuracy to meet project objectives. SPE 112246 3

8. Identify the observed data that is considered 14. Select forecast points. to be most diagnostic of reservoir and flow 15. Generate a series of most informative and physics and pick points for history best match runs (e.g. 12 most informative matching. runs followed by 3 best match runs). The 9. Initialize the Estimator models using results most informative runs reduce the Estimator from the scoping runs. The tool will select uncertainty in the simulator response at the most important modifiers for each match forecast points. The best matches ensure point and construct an initial statistical that there are forecasts that are associated response surface model. with good history match runs. The 10. Generate additional runs to improve the Estimator is updated with the run modifiers Estimator model with a set of ‘most and results at the end of each simulation run. informative’ runs. This is a sequential experimental design approach. These runs explore where there is the most uncertainty A further step that can be added is to optimize the about the simulation model. Each of these operational plan of the reservoir. By adding controllable runs causes the tool to perform a Bayes parameters that can be subjected to optimization to the set update of the Estimator model using the new of modifiers that is being used (e.g. choke settings, results. workover scheduling, well placement, compression 11. Conduct a sequence of ‘best match’ timing, etc.) recommendations for field operational and simulation runs. Since each run updates the development optimization can be derived. Estimator model, a sequence of best match runs is needed, where each run is using an Regarding confidence intervals and estimated improved Estimator model. The parameter prediction uncertainty, the technology estimates the values are chosen by the tool which confidence associated with the set of results for the match performs an optimization that is informed by point. This is a measure of the uncertainty in the value of the Estimator model. the production at a future point in time. The measure of 12. Evaluate the runs graphically using the confidence takes the form of percentile values, and the tool’s built-in capabilities. Modify confidence interval (CI) is the probability that a value is weighting of match objective function to between an upper limit and a lower limit. The default steer match, ensuring that the phenomena confidence intervals used in this work were 80% expected to have the most impact on intervals. predictions are targeted. Repeat from 9 (or 10) until good matches are seen to be Real-time Production Monitoring emerging. The real-time production system used in this workflow, and shown in Figure 2, consists of a specialized field Forecasting Under Uncertainty monitoring system (at the local field level), and a Once a set of acceptable history match runs has been comprehensive production data management system (at obtained, uncertainty in the production forecasts can be the office level). The real-time production system explored. Simply by extending the simulated time period acquires, collects, and stores real-time and historical to include the chosen production forecast scenario, the intelligent instrumentation data, and provides a common range of expected production behavior can be explored. desktop for visualization, field monitoring, analysis, and By repeating steps in the workflow above for a time interpretation. period that includes the original history match period and the forecast period, confidence intervals in the production It features “life of the field” storage capability at the forecast can be generated. Rather than the Estimator just office level with flexible sampling rates as low as per providing information about the quality of the history second with high precision. A broad range of specialized match, it is now providing details of the uncertainty in the interpretation, analysis, monitoring, and diagnostics tools production forecasts given the known quality of the are available to streamline analyses, processes, daily associated history match. For simplicity, Figure 1 shows operations, production optimization, and reservoir this as “Start forecasting”. The details of the workflow management. are: Available configurations for the real-time production 13. Generate a number of scoping runs system include direct interface with intelligent (typically 15) to scope the newly added instruments (downhole, subsea, and topside/surface), forecast period. These runs scope the new and/or instrument interface via the operator’s automated solution space so that the new results in the process control infrastructure of Distributed Control forecast period following on from the Systems (DCS), Supervisory Control and Data history will be taken into account in the Acquisition (SCADA), Subsea Master Control Systems Estimator before starting the refinement (MCS), Information Management Systems (IMS), runs. Automated Systems, and third-party historians. 4 SPE 112246

process for repeatedly including new production data in The real-time production system utilizes the industry the history matching process. Because history matches standard network communications protocol, TCP/IP to have often taken so long, engineers are reluctant to repeat connect the operator’s communications network, and to the process even though the models they are using for connect remote locations via a telecommunications decision making purposes can be out of date. The next provider, satellite services provider, over the internet, or section addresses that challenge. via a wireless network for near proximity locations. Rapid Model Updating Workflow This availability of real-time pressure, temperature A process has been established whereby “right-time” data (topside, subsea or downhole) and multiphase rate data can be extracted from a real-time monitoring application (topside or subsea) means that high quality reservoir in a suitably aggregated format and, employing the performance data are now available to the reservoir workflow described in the sections above, used to update engineer. While these data have been enthusiastically reservoir flow simulation models that are interfaced to taken up and used by operations and production engineers assisted history matching technology and potentially for field monitoring, field surveillance, flow assurance geological reservoir models. The key components of this and production optimization, the reservoir engineering workflow are illustrated in Figure 4. It should be noted community has been somewhat slower to react. Take up that the rapid model update process and the individual has been inhibited by the lack of tools to quickly component applications are capable of handling any assimilate this real-time data into reservoir models. desired time period, including intra-day production data.

One issue is the challenge of timescales. Whereas This right-time data is then made available to the rapid real-time data arrives and can be stored on a second by model updating workflow to append to the existing second basis, for this workflow, the reservoir engineer is history that has been used in the assisted history matching typically interested in daily, weekly, or monthly averaged process. The controls on the existing forecast runs are production data (Figure 3). So for real-time production replaced with the newly observed production data data to be ready for consumption by reservoir models, controls and the confidence intervals around dependent some form of data aggregation is generally needed, which aspects of the modeled production behavior re-calculated. may include filtering, interpolation, averaging, and other This is achieved by rerunning steps 13 to 15 in the summarization methods, depending upon the workflow. The model set up determines whether a characteristics of the data. This aggregated data may be production value is assigned or dependent; dependent provided to reservoir models in the form of ASCII, values are typically phase rates and pressures but may spreadsheet, or XML, using the emerging PRODML include compositions and temperatures. New actual format32. This is often referred to as “in-time” or “right- recorded values are compared with model generated time” data rather than real-time data. values to determine where they lie with respect to the model’s confidence intervals. If the new data lies within Therefore, prior to consumption by the reservoir the confidence intervals, then the field is operating within engineer, the real-time data used in this work were filtered the limits of the model’s ability to forecast and the using a custom formula defined by the user, or a default reservoir model is still capable of predicting field moving average smoothing algorithm supplied with the performance within the uncertainty of the model. application. These formulae enable customised noise filtering to smooth extreme outliers without modifying the If, however, the new data lies outside the confidence original real-time data. The calculation is performed intervals, then the reservoir model is no longer capable of automatically on the selected time span of real-time data. forecasting the performance of the field and it needs Once smoothed, the resulting filtered data are averaged updating. This can be achieved by rerunning steps 9 (or and interpolated to produce data at any desired time 10) to 12. If this automatic step does not lead to an period, and may then be exported into an appropriate acceptable match then human intervention is required. format. Currently the export is a manual process whereby Action may be required of an engineer to steer the history the averaged and interpolated data are exported, with the match differently as in step 12. Or if the mismatch is help of templates, in specialized ASCII format files that, sufficiently severe the process may need to start at steps 2 may be directly inserted as Include Files to the reservoir or 3 where new modifiers are introduced. flow simulator data deck, or imported as well history into the assisted history matching application. However, in the Once these steps have been performed the forecast future it is anticipated that this data transfer will be in an workflow is repeated (steps 13 to 15) so that a new XML format (PRODML) and that the rapid model update forecast and new confidence intervals are generated with applications will receive the data when it becomes the new model. This new model can also be used for available using a publish and subscribe mechanism optimization runs to re-evaluate and revise, if necessary, similar to that available with WITSML33 data. the field operating and development plans.

Another issue that has inhibited the use of real-time This whole process is repeated every time new data data in reservoir models is the lack of a robust automated becomes available and is illustrated in Figure 5. In the SPE 112246 5 reservoir management workflow this would normally not data were prepared from the actual field performance data be more frequent than every month (see Figure 2) but on a monthly basis for this example case. could, if necessary be repeated on a more frequent basis if the field performance was changing quickly. The objective of the example was to illustrate how a reservoir model that is kept current can be an excellent In summary, as new history data (recorded data and reservoir surveillance tool, and thus improve the quality operational controls) are added they are checked to see if of reservoir management36. Water production was the they are consistent with the forecast based on the existing focus of attention, because it was anticipated that this history match (i.e. within the confidence intervals). If they parameter would provide the greatest disparity between are, new forecasts are made based on the augmented simulated and observed data. For the purposes of this history. If not, a re-history match is required which example it was assumed that today’s date (“current date”) includes all the previous history and the new data. This is the end of September 1989 and that everything prior to may be possible using the existing modifiers or new that date is “past” and everything after that date is modifiers may have to be introduced. The new forecasts “future”. The model was history matched up to September should be narrower than the previous forecasts from the 30, 1989 using the workflow shown in Figure 1 and the new forecast point but may be offset from previous cumulative water production results, for the well group of forecasts as operational controls have most likely been producers, are shown in Figure 6. Up until this point all changed. Also, the modifier ranges which generate good the wells in the models were controlled on oil rate and history matches should be reduced. Therefore as time water production was allowed to be predicted by the progresses and more and more production data is models. Forecasts were then run from October 1, 1989 assimilated, both the history match and forecast are until January 1991 with the wells on bottom-hole pressure refined. New information reduces uncertainty in the control and a field oil production rate target (Figure 6). geological and simulation models which results in the Confidence intervals for these forecasts were generated at forecast being refined (a consequence of refining the approximately 3 month intervals, which was the planned history match). The new information may allow the time period for reservoir surveillance. selection of a narrower range of possible geological models. The “current date” was then advanced to the end of December 1989. The confidence intervals were compared Using this rapid model updating process to update the with the actual water production observed (Figure 7). It history match as described above, the set of possible can be seen that the actual water production rate was still reservoir models is always kept up to date ensuring that within the confidence intervals set for the forecasts. the best possible set of models is used for operating and Therefore, the models were still able to forecast field development planning decision-making. performance of the field within the confidence intervals and no further adjustment of the models was required. Finally, there is no inherent requirement that this rapid Forecasts were then made from January 1990 onwards model updating procedure is performed only using the using the existing models. reservoir simulator. It has been shown that during the history matching process using assisted history matching The “current date” was then advanced to March 31, tools, geological models and reservoir simulation models 1990 and the process repeated. Now the actual cumulative can be kept consistent by applying modifiers to the water production was outside the confidence intervals geological model as well as the simulation model34. By (Figure 8). The deviation between actual and forecast is applying the rapid model updating workflow to the seen even more dramatically in the water production rate geological model as well as the simulation model, the (Figure 9). This meant that the models were no longer geological model can be updated as new production data able to forecast field performance and needed to be is assimilated thereby keeping both the simulation and adjusted to assimilate the new production data. The geological models “evergreen”. forecast field performance controls were replaced with the actual field performance controls and the models re- Example matched using the existing modifiers (as described above An example of this workflow is based on a publicly this may be achieved by re-running steps 9 to 12). The available version35 of a simulation model of the Gullfaks results of this re-match are shown in Figure 10. A field on the Norwegian Continental Shelf. The simulation comparison of the actual modifier values for the best model is of the eastern part of the Gullfaks field (Block match history runs to September 30, 1989 and the run 30/10) and models the field from startup in 1986 until the closest to the actual history to March 31, 1990, indicated end of 1996. Although there are some 60 wells defined in that the Kv/Kh ratio and aquifer strength were some of the the model only about a quarter are active early on in the key factors in achieving the re-history match. This match production of the field. This workflow was executed by is not perfect, however, and had a more accurate re- using the advanced linear Bayesian tool described above, history match have been required, the process could have together with the full physics reservoir flow simulator been re-started at step 2 and new modifiers introduced. Tempest-MORE and real-time production monitoring The re-history matching process (steps 2 through 12) software. The simulation recurrent data and historical would then have resulted in a modified set of models that 6 SPE 112246

could better have matched the production performance of the Intelligent Energy Conference and the field and represented a better set of models to make Exhibition, Amsterdam, The Netherlands, April, forecasts for future performance of the field. However, for 2006. the purposes of this example, the re-matching workflow 5. H. Potters, P. Kapteijn, “Reservoir Surveillance had been demonstrated and the process stopped here. and Smart Fields”, IPTC11039 (2005) presented at the International Petroleum Technology Conclusions Conference, Doha, Qatar, November, 2005. A process has been described and illustrated whereby new 6. T. Unneland and T. Haugland, “Permanent production data can be quickly and continually Downhole Gauges Use in Reservoir assimilated into a set of history matched reservoir models. Management of Complex North Sea Oil Field”, By extending the framework of an assisted history SPE26781 (1994) SPE Production and Facilities, matching process, new production data can be added to August, 1994. the old production history and, if necessary, be used to 7. K.H. Frantzen, M. Brandt, K, Olsvik, update the existing history match. Confidence intervals “Multiphase Meters – Operational Experience in from the reservoir performance forecasts are used as a the Asia-Pacific” SPE80502 (2003) presented at guide to whether the existing reservoir models are capable the SPE Asia Pacific Oil and Gas Conference of modeling the new production data. If they are, the and Exhibition, Jakarta, Indonesia, September, performance forecasts are updated by simply introducing 2003 the new operating conditions. If not, the models are 8. D. Reeves, R. Harvey, T. Smith, “Gas Lift adjusted, repeating the history matching process which Automation: Real Time Data to Desktop for may, if necessary, include new history matching Optimizing an Offshore GOM Platform”, parameters. SPE84166 (2003) presented at the SPE Annual Technical Conference and Exhibition held in As data continues to arrive, it is monitored against the Denver, Colorado, October 2003. forecast performance and the associated confidence 9. W.E.J.J.Van Zandvoord, A.A. Maskery, T. intervals. At discrete time intervals, new operational data Schmid, “Application of Real Time Surveillance replaces the forecast operational control and the rapid and optimization Tools on a Large Asset”, model updating process is repeated. SPE100342 (2006) presented at the 2006 SPE Asia Pacific Oil & Gas Conference and In this way, the set of history match models can Exhibition, Adelaide, Australia, September 2006. always be kept up to date or “evergreen” providing the 10. L.A. Saputelli et al., “Promoting Real-Time most up to date models as a basis for important reservoir Optimization of Hydrocarbon Producing management and development decision making. Systems”, SPE83978 (2003) presented at Offshore Europe, Aberdeen, UK, September Acknowledgements 2003. The authors would like to thank the management of Roxar 11. A.G. Mezzatesta, D.C. Shaw, S.J. Webb, for permission to publish this work and for the use of the “Constant Reservoir Evaluation through Real- EnABLE™ software, the Tempest-MORE reservoir Time Data as a Service”, SPE109856 (2007) simulator and Tempest simulation suite, and the presented at the SPE ACTE Anaheim, CA, USA, Fieldwatch/Fieldmanager real-time monitoring and November 2007. production data management software. They would also 12. D.J. Rossi, O. Gurpinar, R. Nelson, S. Jacobsen, like to thank Bob Parish, Robert Frost and David Ponting “Discussion on Integrating Monitoring Data into for their useful contributions to this work. the Reservoir Management Process”, SPE65150 (2000) presented at the SPE European Petroleum References Conference in Paris, France, October, 2000. 1. K.H. Coats, “Use and Misuse of Reservoir 13. O. Nygaard, C. Kramer, R. Kulkarni, J-E. Simulation Models”, SPE2367 (1969), JPT Nordtvedt, “Development of a Marginal Gas- November 1969 Condensate Field Using a Novel Integrated 2. D.S. Bustamante, D.R. Keller and G.D. Monson, Reservoir and Production Management “Understanding Reservoir Performance and Approach”, SPE68734 (2001) presented at the Uncertainty Using a Multiple History Matching SPE Asia Pacific Oil and Gas Conference and Process” SPE95401 (2005) presented at SPE Exhibition in Jakarta, Indonesia, April 2001. ATCE Dallas, TX, USA, October 2005. 14. G. Naevdal, L. M. Johnsen, S. I. Aanonsen, E. H. 3. T. Unneland and M. Hauser, “Real-Time Asset Vefring, “Reservoir Monitoring and Continuous Management: From Vision to Engagement – An Model Updating Using Ensemble Kalman Operator’s Experience” SPE96390 (2005) Filtering”, SPE84372 (2003) presented at the presented at SPE ATCE Dallas, TX, USA, SPE ACTE, Denver, CO, USA, October 2003. October 2005. 15. X.-H Wen, W.H. Chen, “Real-Time Reservoir 4. C. Reddick, “Field of the Future: Making BP’s Model Updating Using Ensemble Kalman Vision a Reality”, SPE99777 (2006) presented at Filtering”, SPE92991 (2005) presented at the SPE 112246 7

SPE Reservoir Simulation Symposium, Houston, Software Techniques to the History Matching TX, February 2005. Process“ SPE25250, 1993 16. A.J. Little, H.A. Jutila, A. Fincham, Energy 26. A.J. Watkins, K.N.B. Dunlop, V.M. Alcobia, Scitech Ltd. “History Matching With Production "The Stochastic Revision of Knowledge in Uncertainty Eases Transition into Prediction” Reservoir History Matching", 22nd Petr. SPE100206 (2006) presented at the SPE Itinerary Congress of OMBKE, (Hungarian Europec/EAGE Annual Conference in Vienna, Mining & Mineral Society), Tihany Hungary, Austria, 12-15 June 2006. Oct. 6-9 1993. 17. C.E. Romero, J.N. Carter, R.W. Zimmerman, 27. Parish R.G. & Little A.J.H., 1994. “A Complete A.C. Gringarten, “Improved Reservoir Methodology for History Matching Reservoirs” Characterization Through Evolutionary 6th ADIPEC, October 1994. Computation”, SPE62942 (2000), presented at 28. Parish R.G. & Little A.J.H., 1997. “Statistical the SPE ACTE Dallas, TX, USA, October 2000. Tools to Improve the Process of History 18. Various Authors, “PUNQ. Production Matching Reservoirs” SPE37730, MEOS, Forecasting with Uncertainty Quantification”. A March 1997. research project funded in part by the European 29. I. Miller and J. Freund, “Probability and Commission under the Non-Nuclear Energy Statistics for Engineers”, second edition, Programme (JOULE III), contract F3-CT95- Prentice-Hall, Inc., Englewood Cliffs, NJ (1977). 0006, January 1996 – May 1999. 30. J. Lach, K. McMillen, R. Archer, J. Holland, R. 19. F. Bennett, T. Graf, “Use of Geostatistical DePauw, B.E. Ludvigsen, “Integration of Modeling and Automatic History Matching to Geologic and Dynamic Models for History Estimate Production Forecast Uncertainty - A Matching, Medusa Field”, SPE95930 (2005) Case Study” SPE74389 (2002) presented at the presented at the SPE ACTE, Dallas, TX, SPE International Petroleum Conference and October, 2005. Exhibition, Villahermosa, Mexico, February 31. H.A. Jutila, SPE, and N.H. Goodwin, Energy 2002. Scitech Ltd. “Schedule Optimization to 20. M. Feraille, F. Roggero, E. Manceau, L.Y. Hu, I. Complement Assisted History Matching and Zabalza-Mezghani, L. Costa Reis, “Application Prediction Under Uncertainty”, SPE100253 of Advanced History Matching Techniques to an (2006), Presented at the SPE Europec/EAGE Integrated Field Case Study”, SPE84463 (2003) Annual Conference in Vienna, Austria, 12-15 presented at the SPE ACTE Denver, CO, USA, June 2006. October 2003. 32. B. Weltevrede, A, Doniger, L. Ormerod, S, 21. G.J.J. Williams, M. Mansfield, D.G. MacDonald, DeVries, “Second-Stage Callenge for the M.D Bush, “Top-Down Reservoir Modelling”, PRODML Standard: Adaptive Production SPE89974 (2004) presented at the SPE ACTE, Optimization”, SPE110907 (2007) presented at Houston, TX, USA, September 2004. the SPE ACTE, Anaheim, CA, November, 2007. 22. M. Litvak, M. Christie, D. Johnson, J. Colbert, 33. M.A. Kirkman, M.E. Symmonds, S.W. M. Sambridge, “Uncertainty Estimation in harbinson, J.A. Shields, M.Will, A. Doniger, Production Predictions Constrained by “Wellsite Information Transfer Standard Markup Production History and Time-Lapse Seismic in a Language, WITSML, an Update”, SPE84066 GOM Oil Field” SPE93146 (2005) presented at (2003) presented at the SPE ACTE, Denver, CO, the SPE Reservoir Simulation Symposium, The October, 2003. Woodlands, TX, USA, February 2005. 34. S.J. Webb, J.S. Bayless, K.N.B. Dunlop, 23. Watkins A.J., Parish R.G., 1992. "A Stochastic “Enabling the ‘Big Loop’- Ensuring Consistency Role For Engineering Input to Reservoir History of Geological and reservoir Simulation Models” Matching." SPE23738, LAPEC II; Caracas, paper presented at the 2007 AAPG Annual Venezuela, March 1992. Convention and Exhibition, Long Beach, 24. Watkins A.J., Parish R.G., 1992. California, April, 2007. "Computational Aids to Reservoir History 35. Gullfaks Data released by Statoil is a selection of Matching." SPE24435. SPE Petroleum availiable petroleum technical data from the Computer Conference, Houston, Texas. July Gullfaks field, finalized before 01.01.1998 1992. 36. C. C. Mattax, R. L. Dalton, “Reservoir 25. Parish R.G., Watkins A. J., Muggeridge A., Simulation”, SPE Monograph Series Volume 13, Calderbank V. J. 1993. “Effective History SPE, 1990. Matching: The Application of Advanced

8 SPE 112246

1 Agree study objectives

2,3 Geomodeling, gridding Identify modifiers & ranges

Import model into EnABLE

4,5 Revise Scoping runs modifiers or (Experimental design) Yes No 6 Revise Import/Review Observed Geomodel Data

7 No Model validated?

Yes 8,9 Pick Estimator (match) Points

10 Informative Runs Review/Revise tolerances 11 & match points Best match runs

12 No Good matches emerging?

Yes

13,14,15 Start forecasting

No Match, optimisation, forecasts complete?

Yes

Recommend Decision

Figure 1 Assisted history matching and forecasting workflow diagram

SPE 112246 9

Topside

On-shore offices

Subsea Downhole

Figure 2 Offshore real-time production system

Reservoir Surveillance

Capacity Planning/ Business Design Operational Planning

Scheduling/Real-Time Optimization

Supervisory Control

PID/Regulatory Control

Seconds M Ho Days M Y inu urs on ear tes ths s

Data Collection Period

Figure 3 Application of real-time data by duration and granularity10

10 SPE 112246

Real-Time ASCII, Assisted History Production System PRODML Matching Workflow Data Historian

Real-Time Proxy Model for Production System Uncertainty Analysis Field Controls Qo, Qw, Qg Sand rate Small Loop

Reservoir Flow Simulation Model Qo, Qw, Qg Sand rate

Big Loop

P, T Qo, Qw, Qg Reservoir Geologic Sand rate Model

Figure 4 Rapid Model Updating Workflow

History match for (0, t0)

Predict for (t0, tend)

n = 0

Record data from (tn, tn+1)

No Production model for

(tn, tn+1) as expected?

Remove runs covering original prediction

Yes phase (t0, tend)

Redo runs

covering (t0, tend) using new schedule, treating

(t0, tn) as history

and (tn, tend) as prediction

Do (tn, tn+1) data fit predicted CIs ? No

Yes Change (t0, tn+1) to history and rematch, including prediction phase

(tn+1, tend)

n = n + 1

Figure 5 Rapid Model Updating Flowchart SPE 112246 11

Confidence Intervals P90 for forecast P50

End of history P10

Figure 6 History match of cumulative water production to October 1989 and forecast with confidence intervals

New history still within Confidence Intervals after 3 months

Figure 7 New cumulative water production history data to January 1990 showing new data still within confidence intervals

12 SPE 112246

After 6 months new history trending out of Confidence Intervals

Figure 8 At April 1990 water production data moving outside confidence intervals and re-history match is required

New history

Figure 9 Water production rate data shows an even more dramatic deviation from forecast SPE 112246 13

New re-matched run

Figure 10 Effect of re-history match to April 1990 and re-forecast on water production rate match