Identifying Reservoir Opportunities Using Automated Well Selection and Ranking

® Originally appeared in World Oil OCTOBER 2020 issue, pgs 37-41. Posted with permission.

RESERVOIR MANAGEMENT Identifying reservoir opportunities using automated well selection and ranking

Effective reservoir The technology automates steps, typical- components, including: management requires ly performed during the selection of field • Engineering analytics - combining multiple development candidates, by applying ad- Automatically performs decline engineering/geoscience vanced algorithms and AI/ML to multi- curve analysis, using AI-based disciplines to identify disciplinary datasets, enabling teams to event detection and type rapidly review alternatives by leveraging curve generation; allocates improvement opportunities. cloud computing.1-3 This cloud-native production per zone for each To optimize analysis, an platform enables all members of an asset well, using a series of data- AI-powered management team to collaboratively assess field devel- driven techniques, integrating platform was developed that opment possibilities. production, completion, rock and delivers faster, more accurate It also integrates disparate sources fluid properties, and PLT/ILT results to maximize asset of data, including well logs, 3D reser- information. voir models, production history, and • Remaining pay identification - value. completion data, breaking down existing Identifies uncontacted pay with silos and opening new cross-functional many distinguishing features, workflows. Many steps are involved in such as pay connectivity analysis, ŝ HAMED DARABI, WASSIM BENHALLAM, generating a field development catalog, automatic baffle identification, JOHANNA SMITH, AMIR SALEHI and DAVID including identification of bypassed pay, perforation strategy and standoff CASTINEIRA, Quantum Reservoir Impact; drainage analysis, mapping of geo-en- constraints. EMMANUEL GRINGARTEN, Emerson gineering attributes to assess geological • Drainage analysis - Estimates Automation Solutions risks, and estimation of initial produc- areas that have been drained by tion potential. The process is an auto- existing wells, by offering both mated, systematic approach to assure a volumetric and facies-based reliable ranking of recompletion, infill drainage areas. Successful reservoir management re- and sidetrack options. • Wellbore accessibility - quires an up-to-date inventory of candi- Workflow automation allows asset Automatically evaluates the date opportunities that an engineering teams to run many more scenarios, ulti- mechanical feasibility via AI- team can pursue to meet expected pro- mately leading to better risk management based deciphering of wellbore duction targets. They include re-perfora- and shorter decision cycles. The synergy diagrams (automated wellbore tions, drilling a new well, or a sidetrack. that is formed between the domain ex- diagram digitization and feasibility Finding the best solutions requires com- perts and AI during the process ensures classification). bining geological knowledge, reservoir that the selected candidates are aligned • Geological risk assessment - behavior, production history, comple- with overall reservoir management strat- Automatically assesses structural tion information, and multi-disciplinary egy and that they satisfy pertinent engi- and mapping risks. know-how. The conventional process of neering and economic constraints. • Production rate forecast - gathering, vetting and analyzing data is Estimates production gains by labor-intensive and generally takes sev- METHODOLOGY selecting from many available eral months. It requires the use of mul- The framework consists of several options containing statistical, tiple software applications, making col- laboration difficult and often leading to an unreliable set of candidates. Also, it is Fig. 1. Summary of recompletion finder workflow. complicated to look at multiple scenari- os under different hypotheses and leaves little time to rigorously vet options un- der review.

SpeedWise Reservoir Opportunity (SRO) is an AI-based technology, the re- sult of a collaboration between Emerson and Quantum Reservoir Impact (QRI).

World Oil® / OCTOBER 2020 37 RESERVOIR MANAGEMENT

available to accelerate the vetting pro- Fig. 2. An example dashboard displaying summary of identified opportunities, location map, distribution by production zone and ranking based on production gain. cess and promote collaboration between asset team members. An example dash- board highlighting selected options, lo- cation map, distribution by production zone and ranking based on production gains is shown in Fig. 2.

CASE STUDIES The technology has been applied suc- cessfully to a diverse portfolio of over 100 onshore and offshore oil and gas fields producing from carbonates and sandstones as primary production and waterflooding. Significant value has been generated in these fields by increasing production, reserves, and capital efficacy while enabling more robust decision- making and increasing organizational agility. Four case studies are highlighted below as a reference. Table 1 exhibits or- ders of magnitude of efficiency attained for these case examples in terms of time, person-months spent, and the number of scenarios explored.

Case study 1 - Latin American sand- stone. The asset is a mature field with over 70 years of production history from 200 wells (mainly vertical well develop- ment). After a long period of primary analytical and ML-based models. engineering properties; what should be production, waterflooding has started • Target search - Identifies optimal minimum/maximum spacing between over the last ten years. The reservoir is locations for placing target wells, the wells; what are the cut-offs for net a tertiary fluvio-deltaic to alluvial silici- using robust and fast optimization pay estimation; and how to define the clastic in a faulted anticlinal trap, with techniques, featuring relative probability of success parameters. By an- low resistivity pay and 18 distinct layers. probability of success mapping and alyzing historical field performance and There was a need to identify recomple- comprehensive geo-engineering benchmarking against analog assets, the tion and infill drilling candidates that are constraints. technology intelligently picks the best aligned with overall reservoir manage- These individual components are as- settings, tailored to address the unique ment strategy, satisfy both geological and sembled to build larger distinct work- challenges posed by the field. engineering constraints, provide reliable flows, namely recompletion finder, ver- The assumptions and settings can be forecasts and are mechanically viable. tical sweet spots, and horizontal target altered to create many more scenarios The asset team’s existing workflow to up- search. Figure 1 illustrates the workflow effortlessly and account for uncertain- date the inventory was deterministic and for the recompletion finder, as many of ties in the key parameters. After the job would take over six months. One of the the mentioned elements are connected. is executed in a scalable and secure cloud key challenges was the poor understand- After uploading the required multi- environment, typically in a matter of ing of reservoir compartmentalization disciplinary data for a field, it usually minutes, the results are accessible via a and delineation of both original and cur- takes only a few minutes/hours to gener- set of interactive browser-based dash- rent remaining oil. This jeopardized the ate a catalog of field development candi- boards, ready for the asset team’s review reliability of the results generated by the dates. First, the dataset is automatically and vetting. Multiple views integrating existing deterministic workflow. QCed and validated by leveraging AI- multi-disciplinary data and analysis are The opportunity identification based techniques. Once the integrity of presented. framework presented was leveraged by the data is verified, the user is ready to These include an executive summary, the asset team, aiming to increase the experiment. Typically, several decisions historical well card and individual op- speed and accuracy of results. Fifteen need to be made for each experiment, portunity cards summarizing the criti- scenarios (three different geomodels controlled through more than 200 set- cal information related to each prospect. with five distinct setting groups varying tings that provide ultimate flexibility Additionally, petrophysical characteriza- most influential uncertain parameters) and customization. For example, which tion, geo-model assessment, production were generated to achieve a range of method to use for mapping the geo- forecast, and mechanical feasibility are potential outcomes and account for the

38 OCTOBER 2020 / WorldOil.com RESERVOIR MANAGEMENT uncertainty in crucial parameters. Af- powered solution to overcome the ex- over 50 years of production history and ter screening and evaluating more than isting obstacles and detect any remain- 90 active producers (mainly vertical well 10,000 recompletion and 17,000 vertical ing behind-pipe potential that could be development), under an active water- well possibilities, dozens of recomple- tapped into by recompleting existing flood. Previous reservoir management tions/reactivations (with potential oil wells. More than 30 scenarios were gen- and recent drilling strategy resulted in a gains between 4,600 to 7,000 bopd) and erated to account for uncertainty in sev- rapid decline in oil production over the vertical wells (with potential oil gains eral key parameters. The post hoc analy- past three years, well below peak produc- between 7,700 to 9,000 bopd) were se- sis revealed that the current water-oil tion (recorded near 200,000 bopd). In lected that satisfied the geo-engineering contact, net-pay identification criteria, that period, the average oil rate per well constraints and passed the scrutiny of a and drainage estimation method were dropped 40%, and the field water cut in- team of domain experts. The full cycle the most influential parameters for the creased from 30% to 50%. It was impos- of updating the inventory catalog took inventory outcome. Thousands of con- sible to achieve the expected ultimate six weeks, which resulted in signifi- tenders for each scenario were examined, recovery factor (54%) with the existing cant savings in time and workforce cost and the final collection consisted of 155 development plan, and the asset team (eight person-weeks compared to 24 feasible options, after applying a series of was looking into horizontal drilling to staff-months), while empowering the as- filters, which included minimum initial increase recovery per well and improve set team with more robust probabilistic rate, minimum oil thickness, maximum well economics. models for decision making. acceptable uncertainty, and the inspec- Although the team had recently tion of geologists, reservoir engineers, drilled a few horizontal pilot wells to Case study 2 - Middle East sandstone. and production engineers. test the idea and gather information, The asset is a giant sandstone reservoir, After performing a multi-faceted risk there was no systematic approach to cre- with 55 years of production history and assessment, the candidates were grouped ating a robust development plan with more than 1,000 producers. Oil produc- into 25 high-confidence selections with horizontal wells. The asset team initially tion peaked near 2 MMbopd, but had average initial production of 400 bopd, used the traditional reservoir simulation declined to 0.8 MMbopd by the time of and 130 mid-confidence prospects with approach. However, they recognized the study. The asset team was looking to an average initial production of 600 that building and calibrating the simu- identify recompletion prospects in exist- bopd. The base scenario included 116 lation model would take at least a year ing wells for a short-term increase in pro- recompletions with an estimated total to complete, and the robustness of the duction. The long history, commingled potential oil gain of 63,000 bopd. The final model was not guaranteed, due to production, and geological complex- study was completed in five weeks and inherent uncertainty in the geomodel ity (120 producing zones with different involved ten person-weeks (compared and complexity linked to carbonate res- properties) made it very challenging to to 24 staff-months), which significantly ervoirs. update the inventory effectively. The boosted the efficiency compared to the In pursuit of an alternative approach existing workflow took more than a year traditional approach. that could deliver faster and more accu- and was unable to account for uncertain- rate results, the asset team utilized the ty in several vital parameters, due to its Case study 3 - Middle East carbon- AI-based technology to systematically deterministic nature. ate. The asset is a well-connected, highly identify and inspect horizontal targets. The asset team utilized the new AI- faulted, mature carbonate oil field, with The search covered: a) shorter laterals

Table 1. Summary of efficiency gains (orders of magnitude) achieved by the technology (compared to existing asset team’s workflow) for the four case studies.

CASE STUDY 1: CASE STUDY 2: CASE STUDY 3: CASE STUDY 4: Latin American Sandstone Middle East Sandstone Middle East Carbonate Gulf-of-Mexico Carbonate (Recompletions/Vertical Wells) (Recompletions) (Horizontal Wells) (Deviated Wells)

200+ Wells 1,000+ Producers 90 Active Producers 8 Active Producers 18 Reservoir Zones 55 Years Production 50 Years Production 20 Years Production 70 Years Production 2 MMbopd Peak Production 200,000 bopd Peak Production 120 MMbbl Peak Production 10 years Waterflood

Existing Existing Existing Existing SRO SRO SRO SRO Workflow Workflow Workflow Workflow

Completion 6 6 5 12 4 12 8 12 Time Weeks Months Weeks Months Weeks Months Weeks Months

Person-Months 8 24 10 24 10 30 6 30 Spent Weeks Months Weeks Months Weeks Months Weeks Months

Number of 15 1 30 1 10 1 6 None Scenarios Cases Case Cases Case Cases Case Cases

World Oil® / OCTOBER 2020 39 RESERVOIR MANAGEMENT that could be accessible as sidetracks els accurately predicted water behavior, VALUE ADDED from existing wells to minimize water resulting from a poor understanding of The oil and gas markets in the post- encroachment; and b) longer laterals the fracture network and the limitations Covid-19 world represent a new eco- that could be drilled as new wells. The of conventional dual-porosity and dual- system that is governed by the need for overall search strategy was targeting the permeability finite difference models. As extreme efficiency in capex/opex. The top of the reservoir, increasing reservoir a result, the asset team had no choice but new AI-powered automation technol- contact, and maximizing standoff from to write off 1P reserves by 90%, and there ogy brings speed, accuracy, value creation, water-oil contact. More than ten scenar- was no planned activity beyond operat- and risk mitigation to field development ios were produced to account for uncer- ing the eight active producers. plans, specifically to the selection of the tainty in crucial parameters, namely the The asset team turned to the AI-based best wells—clearly the most capital- geomodel and water-oil contact surface. technology to maximize the value of the influential part of upstream activities. After filtering based upon geo-engineer- asset before abandonment. The focus Going forward, the conventional pro- ing attributes and rigorous vetting by do- was on looking for a) workover candi- cesses that usually take many months main experts, the final catalog consisted dates in existing wells to mitigate water to produce sub-optimal field development of 32 horizontal targets, ranging from encroachment, and b) sweet spots in candidates will be replaced by solutions 500 m to 1,000 m in lateral length, with a the attic oil structure between the water that usher in the age of digital transforma- minimum spacing of 20 acres. cones to maximize the stand-off from tion in reservoir development planning. After careful consideration, the top water-oil contact, accessed by highly de- The benefits of the technology pre- five options in the base scenario were viated wells to maximize reservoir con- sented are threefold: speed, accuracy/ selected for execution in the short tact. One of the key challenges was the risk mitigation, and value creation. As term, with an estimated total oil gain poor understanding of the current water- highlighted in the examples, more than of 40,000 bopd. To further support the oil contact. An ensemble of 25 dynamic 90% of speed-up is typically achieved development plan, minimize water pro- reservoir models was built and calibrated compared to conventional workflows. duction, maximize water injection ef- to estimate the current water saturation Also, due to enhanced data mining and ficiency, and improve pressure support, map, using QRI’s proprietary forecasting AI-based vetting algorithms in well se- a proprietary waterflood management technology; an efficient and rapid well/ lection, the resulting field development technology from QRI was utilized; this reservoir forecasting technology that catalog offers a much higher degree of is an innovative solution, based on criti- builds hybrid models (combining phys- confidence and accuracy than conven- cal physics and data-driven models that ics-based and data-driven approaches) tional manual work processes, leading have been deployed globally in many with a systematic top-down approach. to better risk management. Finally, the oil fields to manage active waterfloods This solution aims for optimum com- compelling list of ranked and vetted field and optimize well controls.4,5 Delivering plexity by addressing both bias and vari- development options—recompletions, feasible and actionable inventory only ance in the system.6-8 side-tracks, and new wells—typically took four weeks and involved ten per- For the present study, the robustness of produces thousands of incremental bar- son-weeks, awarding the asset team with models was verified through error analysis rels of oil/gas, both short-term and in much shorter decision cycles and smart- on both a cross-validation set (used only EURs. SRO as SaaS can be described er decisions than the labor-extensive, to adjust the hyper-parameters, and not most succinctly as a capital-efficiency traditional simulation approach with un- during the history matching) and a test set engine driven by AI, as demonstrated promised returns. (data not accessible during model build- by the case examples and is essential ing). Subsequently, different saturation in modern reservoir management.9,10 Case study 4 - Gulf of Mexico fractured maps were used as input of the opportu- Typical annual savings in capex/opex, at carbonate. The asset is a complex, highly nity identification framework to gener- equivalent production levels, range from fractured, mature oil field with over 20 ate several scenarios. After careful con- 20% to 50%. years of production history. Over the sideration, six scenarios were examined REFERENCES past two years, oil production dropped in more detail to uncover the remaining 1. Castineira, D., R. Toronyi and N. Saleri, “Automated almost 80% (from peak production of potential and manage risks. The base sce- identification of optimal deviated and horizontal well targets” SPE paper 192279, presented at the SPE Kingdom of Saudi 120 MMbbl of oil to 25 MMbbl of oil), nario consisted of 11 technically and eco- Arabia Annual Technical Symposium and Exhibition, 2018. due mostly to rapid water encroachment nomically feasible prospects, including DOI:10.2118/192279-MS 2. Kansao, R., W. Benhallam, A. Aronovitz, M. Ravuri, A. that resulted in well shutoffs. A typical eight workovers (plug-backs and choke Maqui, V. Suicmez, D., Castineira and H. Darabi, “Intelligent well was producing oil between 20,000 optimization in existing wells) and three and automated workflow for identification of behind-pipe recompletions and new infill location opportunities for an bopd and 30,000 bopd with no water, highly deviated targets. The incremental onshore Middle East field,” SPE paper 199090, presented at the many years after commission. However, recovery for the base scenarios was esti- SPE Latin American and Caribbean Petroleum Engineering Conference, 2020. DOI:10.2118/199090-MS 2020 after the water breakthrough, oil produc- mated between 9.4 MMbbl and 25 MMb- 3. Deng, L., Salehi, A., Benhallam, W., & Castiñeira, D. (2020). tion dropped by more than 90% in less bl of oil. The complete process of iden- Intelligent and Automated Workflow for Identification of Behind Pipe Recompletions and New Infill Locations than six months. It took the asset team tifying and ranking the candidates only Opportunities for an Onshore Middle East Field. In SPE Conference at Oman Petroleum & Energy Show. Society of more than a year to build and calibrate took eight weeks—which is considered Petroleum Engineers. a traditional reservoir simulation model an industry record for addressing such a 4. Zhai, X., T. Wen and S. Matringe “Production optimization in waterfloods with a new approach of inter-well connectivity to understand water behavior in the field complex problem—and the unlocked po- modeling,” SPE paper 182450, presented at the SPE Asia and optimize the development plan. tential gave new life to the field, where no Pacific Oil and Gas Conference and Exhibition, 201. Despite all the effort, none of the mod- alternative solution was available. DOI:10.2118/182450-MS 2016

40 OCTOBER 2020 / WorldOil.com RESERVOIR MANAGEMENT

5. Maqui, A. F., X. Zhai, A. Suarez Negreira, S. F. Matringe WASSIM BENHALLAM is simulation with complex physics, along with and M. A. Lozada, “A comprehensive workflow for near-real product development lead for new approaches for compositional modeling time waterflood management and production optimization, QRI’s SpeedWise technology. and general-purpose simulation of multiphase using reduced-physics and data-driven technologies,” SPE With a bachelor’s degree in multicomponent flow in porous media. He also paper 185614, presented at the SPE Latin America and computer science and a modeled a CO sequestration processes and Caribbean Petroleum Engineering Conference, 2017. 2 DOI:10.2118/185614-MS 2017 master’s in geology, Mr. data-driven reservoir modeling with focus on 6. Saleri, N. G., “Re-engineering simulation: Managing Benhallam uses his machine learning applications. complexity and complexification in reservoir projects” interdisciplinary skills to develop fast, automated, SPE Reservoir Evaluation & Engineering, 1(01), 5-11. and data-driven solutions that incorporate DR. DAVID CASTINEIRA is chief 1998DOI:10.2118/36696-PA geological and engineering data to maximize oil technology officer for QRI. He is 7. Esmaeilzadeh, S., A. Salehi, G. Hetz, F. Olalotiti-lawal, H. Darabi production. As a computational geoscientist a recognized technology leader and D. Castineira, “A general spatio-temporal clustering- with QRI, he has developed patented with 12 years of experience in based non-local formulation for multi-scale modeling of technologies to automatically identify remaining compartmentalized reservoirs,” SPE paper 195329, presented at the oil and gas sector. Before the Society of Petroleum Engineers Western Regional Meeting, behind-pipe opportunities, sweet spots for new joining QRI, Dr. Castineira 2019. DOI:10.2118/195329-MS drills, and optimal targets for horizontal drilling. worked in R&D at Shell 8. Salehi, A., G. Hetz, F. Olalotiti, N. Sorek, H. Darabi and D. International Exploration and Production, where Castineira, “A comprehensive adaptive forecasting framework JOHANNA SMITH is a he developed and applied new quantitative for optimum field development planning,” SPE paper 193914, development geologist for QRI, analytical methods for reservoir management. presented at the SPE Reservoir Simulation Conference, 2019. working on some of the world’s He holds a PhD in chemical engineering from the DOI:10.2118/193914-MS largest reservoirs. She is University of Texas at Austin and a MS degree in 9. Saleri, N. G., “Learning reservoirs: Adapting to disruptive assigned to the Quantum computational science and engineering from technologies. Journal of Petroleum Technology, 54(03), 57-60, 2002. DOI:10.2118/73695-JPT Analytics Team as a Reservoir Harvard University. 10. Saleri, N. G., “Reservoir management tenets: Why they matter Characterization Analyst, where to sustainable supplies,” Journal of Petroleum Technology, 57(01), she has gained a strong technical background in DR. EMMANUEL GRINGARTEN 28-30, 2005. DOI:10.2118/0105-0028-JPT asset evaluation and reservoir characterization is senior director of product by utilizing varied available datasets in every management at Emerson project with her diverse software proficiencies. Automation Solutions. He has DR. HAMED DARABI is a Ms. Smith previously worked at ExxonMobil in 20 years of experience building product manager at Quantum both conventional and unconventional resource reservoir models and leading Reservoir Impact, responsible play exploration, production and operations. the development of geological for delivering the SpeedWise modeling, reservoir engineering and production reservoir management platform DR. AMIR SALEHI serves as a modeling software. Dr. Gringarten’s work has to the industry. Dr. Darabi has 12 technical team lead for QRI’s focused on facilitating multi-disciplinary years of industry experience SpeedWise Technologies. Dr. collaboration, data integration, workflow across 100+ reservoirs, including analysis in the Salehi received his PhD in automation and risk quantification He holds a Permian, Eagle Ford, Canada, Mexico, Kuwait, petroleum engineering from BSc degree in mathematics from Imperial Colombia, Iran, UAE, Oman and Iraq. He Stanford University, where as a College, London and received an MS degree received a PhD in petroleum engineering from research assistant, he worked and a PhD in petroleum engineering from the University of Texas at Austin. on upscaling modeling methods for reservoir Stanford University.

Article copyright © 2020 by Gulf Publishing Company. All rights reserved. World Oil® / OCTOBER 2020 41 Not to be distributed in electronic or printed form, or posted on a website, without express written permission of copyright holder.