Back to Top
Skip to main content
Subtask 3.1 - Bakken Rich Gas Enhanced Oil Recovery
Project Number
Last Reviewed Dated

The project goal is to determine the feasibility of reinjecting captured rich gas into a Bakken reservoir to enhance oil recovery. Specific research objectives related to this goal are as follows:

  • Determine the ability of various rich gas mixtures to mobilize oil in Bakken petroleum system reservoir rocks and shales.
  • Determine the changes in gas and fluid compositions over time in both the tight oil reservoir and surface infrastructure environments, and assess how those changes affect reservoir and process facility performance.
  • Optimize future commercial-scale tight oil EOR design and operations through the use of iterative modeling of surface infrastructure and reservoir performance using data generated by the field- and laboratory-based activities.
  • Establish the effectiveness of selected monitoring techniques as a means of reservoir surveillance and injection conformance monitoring in the Bakken petroleum system.
  • Determine the sorptive capacity of Bakken shales for rich gas components and the effects of sorption in the shales on gas utilization rates in samples representing areas of low, medium, and high thermal maturity. 
  • Use laboratory-based rock characterization activities to evaluate fluid flow pathways by determining wettability and relative permeability of rich gas components in Bakken rocks from different thermal maturity areas, and investigate the use of an emerging technique to predict fracture occurrence across a horizontal wellbore by correlating cuttings’ mineralogy to brittleness.
  • Model alternative commercial-scale injection scenarios based on models developed with the embedded discrete fracture model (EDFM) technique.
  • Apply machine learning (ML) and big data analytics (BDA) to develop virtual learning methods for Bakken rich gas EOR. Develop a methodology to integrate ML and BDA with advanced modeling (including coupled geomechanics and fluid flow) to examine the potential impact that multiple operational scenarios may have on conformance and sweep efficiency (i.e., a multiwell huff ‘n’ puff operation conducted over a township scale as opposed to a drill spacing unit [DSU] scale). Additionally, in these ML and BDA activities, include where to place or space infill wells to increase total recovery within multiple DSUs, when to start the injection of fluids, and which type of fluids to manage reservoir pressure to improve total recovery.

Energy & Environmental Research Center (EERC) – University of North Dakota, Grand Forks, ND 58202-9018


North Dakota is well-situated to demonstrate the implementation of rich gas-based EOR for tight oil formations. Although flaring associated with Bakken oil production has been reduced significantly in recent years, as of March 2019, approximately 19 percent of the rich gas produced in association with Bakken oil production continues to be flared. The associated gas from Bakken oil production operations is typically a mixture dominated by methane with a significant amount of ethane and other hydrocarbons. The results of recent preliminary laboratory investigations at the EERC suggest that pure ethane and mixtures of methane and ethane may be used to mobilize oil from Bakken rocks and thus could be viable injection fluids for EOR operations. The EERC is working with Liberty Resources (LR) to design and conduct an EOR pilot test using rich gas. The project is a joint initiative between the EERC, North Dakota Industrial Commission (NDIC) through the Bakken Production Optimization Program (BPOP), LR, and the U.S. Department of Energy (DOE). Project activities will be coordinated, managed, and evaluated by the EERC. LR will be responsible for providing the wells and rich gas necessary for the test and will operate the injection, production, and monitoring activities. An improved understanding of wettability, relative permeability, and fracture network distribution across the Bakken will be developed using advanced reservoir characterization techniques. Alternative injection strategies to optimize EOR strategies at scales larger than a DSU will be investigated using advanced reservoir-modeling methods. ML and BDA will be used to streamline pilot performance assessments. 


Estimates for original oil in place  (OOIP) in the Bakken petroleum system range from 300 billion to 900 billion  barrels. Current resource recovery factors for Bakken wells are typically 10% or  less. If this trend continues, billions of barrels of oil will be left stranded  in the reservoir. Analysis conducted by the North Dakota Pipeline Authority  indicates that the current gas-gathering infrastructure in North Dakota  (including pipelines, compressor stations, and gas processing facilities) is  insufficient to accommodate all of the associated gas that is produced as part  of oil production from the Bakken. The geographically isolated location of the  Bakken oil play relative to large natural gas markets, combined with continued  low natural gas prices, has made it economically challenging for industry to invest  capital in expanding gas-gathering infrastructure in North Dakota. Therefore,  management of rich gas production from the Bakken is still a high priority for  government and industry stakeholders in North Dakota. This project will  demonstrate the viability of utilizing rich gas for EOR in the Bakken, which  will result in reduced flaring and an improvement in recovery factors. The  primary impacts of this project will be reductions in greenhouse gas emissions  associated with Bakken activities, and potentially the production of billions  of barrels of incremental oil.

Accomplishments (most recent listed first)

The project was initiated on September 1, 2017. A hearing of the NDIC Oil and Gas Division was held September 21, 2017, for the purpose of LR providing testimony for its application to obtain the necessary permits for the pilot injection test. Permits for injection activities in six wells have been granted to LR. LR has purchased a compression unit that is necessary for the operation of the pilot injection test. Specific accomplishments include the following:


  • Large-scale pilot tests were conducted in two wells in the Stomping Horse complex beginning on November 20, 2018, and continuing through May 2019.
  • A gas tracer was introduced to the injection well on November 21, 2018. A second tracer study — which included the injection of gas, oil, and water tracers — was initiated on January 27, 2019. Multiple sampling and analysis events for multiple wells were conducted to look for tracers as a means of identifying fast flow pathways for gas, oil, and water between the injector and various offset wells.
  • The maximum injection rate for the large-scale test is 2.0 MMscfd. For each injection cycle the pilot testing plan called for injection into each well  until one of three criteria are achieved: 1) total injection of 60 MMscf, 2) 30 days of injection, or 3) clear evidence of substantial breakthrough at an offset well.
  • As of May 1, 2019, over 130 MMscf of rich gas had been injected into four wells during six different injection periods.
  • Key observations from the pilot testing so far include:
    • The ability to effectively inject rich gas into Bakken and Three Forks reservoirs has been demonstrated.
    • Injectivity is readily established and has not been a constraint on operations.
    • Reservoir surveillance and monitoring demonstrate the injected gas can be controlled and has been contained within the drill spacing unit.
    • Pressure buildup is occurring and is showing a positive trend towards achieving minimum miscibility pressure (MMP).
    • The conceptual approach of using laboratory-based testing to inform modeling, which in turn guides injection scheme design and operations, has been effective.
  • Baseline reservoir characterization data collection has been completed for all wells within the Leon-Gohrick drill spacing units in the Stomping Horse complex. Parameters measured included analysis of produced oil, water, and gas as well as bottomhole pressure and temperature for wells permitted for injection and offset wells.
  • MMP studies have been conducted to determine the MMP of rich gas components and different rich gas mixtures in oil from the Stomping Horse complex. MMP data for methane, ethane, propane, and different relevant mixtures have shown that “richer” gas mixtures will result in lower MMP values (e.g. methane MMP > ethane MMP > propane MMP).
  • Rock extraction studies of the rich gas components on Bakken shale and nonshale samples have shown that when it comes to mobilizing hydrocarbons from Bakken rocks, methane is the least effective, propane is the most effective, and ethane has an intermediate effect. The rock extraction studies also showed that propane is effective at all pressures, ethane is effective at higher pressures, and methane is the least effective at any pressure.
  • Modeling-based studies of the potential effects of rich gas EOR operations on the surface infrastructure of the Stomping Horse complex predict that rich gas EOR will not adversely affect surface facility operations.
  • Reservoir modeling of selected injection/production scenarios predicts incremental oil recovery may exceed 25%.
  • Small-scale injectivity tests were conducted in two wells in the Stomping Horse complex during the summer of 2018. A total of 24.6 MMscf of rich gas was injected during three tests conducted in two wells between July 17 and September 10, 2018. The maximum injection rate achieved was 1.14 MMscfd. Downhole pressure and temperature data were collected before, during, and after the injection tests from six wells in the drill spacing wells, including the injection wells and the immediately adjacent offset wells. Data obtained from the small-scale injection tests were used to refine the design of the subsequent larger pilot tests.
  • Sorption isotherms for methane, ethane, propane, and a rich gas mixture were successfully measured on three shales representing areas of low, medium, and high thermal maturity in the Bakken Formation using the high-pressure magnetic balance.
  • Large-scale injection testing was conducted into the Gohrick 4-MBH well, which began on January 17, 2019. Multiple severe weather events in February 2019 caused interruptions to the injection activities. Consistent injection was reestablished in early March 2019 and continued until May 9, 2019. A total of 74.5 MMscf was injected into the Gohrick 4-MBH well. Injection into the Gohrick 6-TFH well was initiated May 15, 2019, and ceased in early June 2019, with a total of 17.4 MMscf injected. Injection was ceased because of a constraint on rich gas availability.
  • Reservoir surveillance data from the 2018–2019 pilot injection testing in the Leon and Gohrick wells in the Stomping Horse area were used to revise the reservoir model. The revised reservoir model also incorporated a new EDFM, which more accurately accounts for the complexity and heterogeneity of the fractured reservoir. The use of the reservoir surveillance data and revised model resulted in improved history matching of oil, gas, and water production.
  • Rich gas flow-through experiments were conducted by injecting a 70% methane, 20% ethane, and 10% propane mixture into two samples of Upper Bakken Shale from different wells. Samples were characterized using a suite of tests to investigate properties responsible for observed methane retention. The samples show different levels of porosity/permeability and organic matter content, with the tightest sample showing a capacity to capture and retain a significant quantity of injected hydrocarbon gas. In both experiments, methane appeared to be preferentially retained, while the rate and capture exhibited by each sample differed greatly. Nanoscale porosity associated with abundant organic matter in rock samples was suggested to cause the observed phenomena.
  • A tensiometer is being used to determine the IFT (interfacial tension) and contact angle of fluid–fluid and fluid–rock pairs. These measurements and mercury injection capillary pressure (MICP) analysis will provide an understanding of wettability and aid in the determination of relative permeability.
  • Elastic properties from minerals were used to model predictions for geomechanical properties. Four geomechanical properties, including Young’s modulus, Poisson’s ratio, bulk modulus, and shear modulus, were predicted from rock physics modeling. These were combined with x-ray diffraction (XRD) and x-ray fluorescence (XRF) analyses. Various ML algorithms (K-nearest neighbor regressor, random forest regressor) were applied to the prediction of mineral composition and brittleness index. The purpose is to optimize the prediction performance and avoid the nonunique algorithm effect.
  • A ten-component equation-of-state (EOS) was developed for a typical Bakken oil sample based on a detailed analysis of multiple pressure, volume, temperature (PVT) data sets. The EOS enables engineers to consider different gas-flooding scenarios flexibly since all gas components can be adjusted individually to mimic the field observations.
  • Two reservoir simulation models with different natural fracture and induced fracture settings have been developed using the EDFM approach. Test runs showed the models developed with the EDFM approach run faster than the traditional fracture models with the same reservoir size and number of fractures.
  • An initial Bakken–Three Forks geologic model (geomodel) was created for a DSU-sized area (approximately 2 miles × 1 mile) in northwestern North Dakota on the western flank of the Nesson Anticline. The reservoir database will provide the virtual learning environment for the ML activity.
  • A series of data acquisition/analysis and geologic modeling activities have been conducted to develop a large-scale geologic model for conformance control and EOR studies. Data of 77 wells in the Dunn–McKenzie area were collected from the NDIC database. The geologic model encompasses a 3-mile by 4-mile (12-square-mile) area. The formations of interest in the modeling effort included the Lodgepole, Bakken, and Three Forks Formations.
  • Based on the large-scale geologic model built, a simulation model with seven wells was developed to simulate the reservoir dynamics in the Bakken. Included in the model were 25% of the fractures in each well to make the simulation run efficiently while keeping the well interference effects considered. The distribution of wells and fractures in the model not only enables the project team to consider the pressure and flow interference between wells/fractures in the production/injection processes but also allows the project team to study conformance control strategies in the EOR operations. The nonintrusive EDFM technique was employed to generate fractures in the simulation model. This recently developed technique enables modeling of complex fracture geometry using structured grids, which significantly improves computational efficiency while maintaining the simulation accuracy.
  • Over 300 rich gas injection EOR cases were set up utilizing the seven-well simulation model. Key EOR design and operational parameters including injector location, gas injection rate and time, soaking time, and production time, were considered in the simulation cases. The results will be used as input data for the ML study.
  • Gas breakthrough behavior was studied using pure gas injection. Methane, ethane, and propane were used as the sole injection solvent in different simulation cases to identify the interference between the injector and neighboring producers. Premature gas breakthrough can be detected by observing the concentration change of the injected solvent in the produced gas from neighboring producers.
  • Based on the gas breakthrough results observed in the pure gas injection cases, conformance control using water injection was studied. Oil–water interfacial tension measurements indicated that the capillary pressure between oil and water does not change significantly with reservoir pressure in the rich gas injection EOR process. Therefore, water injection could confine the injected gas around the huff ‘n’ puff well for EOR purposes, but it may not improve the oil production in the wells that are used for water injection.
  • The virtual learning work is using two simulation case matrices to quantify the effect of DSU development and operational parameters on DSU production (oil, gas, and water production). The Set 1 cases were completed and evaluated 18 realizations that explored the effects of DSU well count, EOR development timeline, and EOR injectate on DSU production. Based on the results from Set 1, an additional case matrix of 274 realizations (Set 2) was designed to explore the parameter space more broadly for a seven-well DSU, early EOR development timeline, gas-only injectate, and the following operational variables: injection rate, injection time, soak time, and production time.
  • Different visualization techniques were used to assess pilot test field data acquired from propane gas injection targeting the Middle Bakken in Mountrail County, North Dakota; a pilot-scale field test of CO2 injection into a tight oil reservoir in northern Dunn County, North Dakota; and a rich gas EOR pilot in Williams County, North Dakota.
Current Status

The pilot test by LR and activities surrounding the pilot test are complete. IFT and contact angle measurements of fluid–fluid and fluid–rock pairs are ongoing. MICP analyses are in progress. ML approaches are being explored for geomechanical properties estimation using XRD, XRF, and geomechanical results from core and cutting samples. Systematic simulation runs are being conducted to optimize the operational parameters in the huff ‘n’ puff process when a representative Bakken rich gas composition is used. Two simulation models are being developed to investigate multiple-well huff ‘n’ puff EOR operations in the Bakken. Hydraulic and natural fractures are considered in these models to capture the fluid flow behavior in the complex fractured reservoir. Advanced EOR strategies including high-pressure propane injection and surfactant injection are being studied based on the latest field data. Set 2 data will be analyzed using response surface methodology, regression, and/or ML techniques to quantify the effect of the operational factors on DSU production. The virtual learnings from Set 1 and Set 2 will provide the basis for optimization studies looking to maximize oil production and minimize water production or, equivalently, to maximize the net present value of the DSU. One or more of the simulations that generated data at monthly resolution will be rerun at higher time-series resolution (e.g., hourly) to explore real-time 

Project Start
Project End
DOE Contribution


Performer Contribution


Contact Information

NETL – Gary Covatch ( or 304-285-4589)
EERC – Steve Smith ( or 701-777-5108)