Yang, Yuhao (University of Kansas) | Fu, Qinwen (University of Kansas) | Li, Xiaoli (University of Kansas) | Tsau, Jyun-Syung (University of Kansas) | Barati, Reza (University of Kansas) | Negahban, Shahin (University of Kansas)
The visualization and quantification of CO2 and oil interactions give insight into the multiple mechanisms controlling CO2 enhanced oil recovery processes. In this work, a high pressure high temperature full visual Pressure-Volume-Temperature (PVT) system is used to measure the equilibrium parameters including oil volume, gas volume and equilibrium pressure. In consequence, equilibrium properties including CO2 solubility, oil swelling factor and extraction pressure can be calculated and observed. In non-equilibrium condition, CO2 condensation and light oil component extractions are observed, as well as pressure decay data due CO2 dissolution is recoded. Furthermore, diffusion coefficient is calculated based on pressure data. Hence, the mechanisms of CO2 EOR process are identified and analyzed.
Firstly, excessive gaseous CO2 is charged into the piston-equipped view cell coexisting with the pre-loaded Bakken oil. Three types of phase behavior can be captured under certain conditions including liquid oil-Vapor CO2 (LV), liquid oil-liquid CO2-Vapor CO2 (L1L2V), and liquid oil-liquid CO2 (L1L2), which are common fluid types in the formation during CO2 EOR process. In equilibrium process, the cell is pressurized stepwise. The volume of the swollen oil caused by the dissolution of CO2 is recorded at each equilibrium pressure step, at which gas solubility, swelling factor as well as extraction pressure can be calculated and determined. Once the light components in Bakken oil start to be extracted into the liquid CO2 phase, the volume of oil-rich phase decreases and a couple of extraction columns can be observed. The heavy components in oil are harder to be extracted so that the oil volume eventually reaches a plateau. During the non-equilibrium process, the pressure increases continuously by moving the piston at the various rates until the pressure reaches the desired pressure at different temperatures. Finally, the pressure decay method is used to determine the diffusion coefficient between CO2 and Bakken oil.
It has been found that oil can be swollen by dissolving CO2 at high pressures. The swelling factor increases with pressure during LV condition, where the EOR mechanism is mainly oil swelling effect. However, the swelling factor decreases with pressure during L1L2V and L1L2 conditions, indicating a change in the main controlling mechanism to CO2. As for the non-equilibrium process, the extraction is found to be closely related to the CO2 physical state. The condensing flow from CO2 rich liquid phase to oil phase and the extracting flow from oil phase to CO2 rich liquid phase have been filmed to demonstrate the EOR mechanisms. The effective diffusion coefficient in Period I, which is dominated by natural convection, is found to be three orders larger than that in Period II, which is mainly driven by molecular diffusion.
In this work, both equilibrium and non-equilibrium properties have been measured and observed by using a piston-equipped visual cell. The mechanisms of CO2 EOR for Bakken oil have been comprehensively identified and analyzed at the different stages for the first time. This work sheds new light on the design of CO2-EOR application in unconventional oil reservoirs.
Hakimov, Nijat (University of Kansas) | Zolfaghari, Arsalan (University of Kansas) | Kalantari-Dahaghi, Amirmasoud (University of Kansas) | Negahban, Shahin (University of Kansas) | Gunter, Gary (Schlumberger-NExT)
Archie's law is commonly used for the estimation of petrophysical properties of porous media linking electrical resistivity to water saturation. Therefore, low resistivity formation is expected to have high water saturation and hence, high water production. This, however, is not the case for many reservoirs around the world, for which low resistivity pay zones have been reported with very low or none water cut.
One of the main causes of Low Resistivity Pay (LRP) phenomena, especially in carbonates, is microporosity. Due to their small pore sizes, micro-pores have much higher threshold capillary pressures than macro-pores during drainage in the water-wet samples which resembles the original state of reservoirs before oil migrations. Because of that, we often find formations in which micro pores and macro pores are saturated with brine and oil, respectively. This indicates that there is a correlation between pore fluid occupancies and pore size distributions. The existence of connected pathways through micro-pores that are fully saturated with a conductive phase (i.e., brine) creates ‘shortcuts’ for the electrical current which causes short circuit and, ultimately, lowers rock resistivity measured from log analysis.
The purpose of this work is to investigate the impact of microporosity on the electrical properties of porous media through pore-scale network modeling techniques. To achieve this, a tortuous pore network is constructed on 2D rectangular regular lattice to represent macro pores and throats in the network. Next, the macro network is modified to include micro-pores. This has been done by adding a small rectangular lattice network of micro pores and throats. Radii and lengths of each element are chosen from the pre-specified ranges. This is done carefully to ensure that all networks of different scales fit geometrically within the lattice of a given size. We are specifically interested to investigate flow and electrical properties as a function of the locations, dimensions, and orientations of the micro-porosity regions. To achieve this goal, a comprehensive set of sensitivity analyses are done to assess the impact of various parameters including number of pores, tortuosity, geometry and location of microporosity (i.e., parallel or in series, continuous or non-continuous). The results are compared against Archie's equation, which is commonly used in the industry for log interpretation. This work helps to further expand the use of this equation for field applications, specifically, for formations containing rocks with wide pore size distribution.
This paper uses case studies to introduce a Ten-Step Integrated Petrophysical Rock Type (PRT) Verification Process that Combines Deterministic Methods, Saturation Height Modeling (SHM), Advanced Flow Units and Independent Probability Self-Organizing Mapping (IPSOM) neural networks. This method was tested in carbonates, sandstones and un-conventional shales. The input data to the method is a core-log integrated porosity, permeability and a pore throat radius indicator based on deterministic methods and mercury injection porosimetry capillary pressure.
Challenges remain for investigators and teams in applying PRT techniques in field studies. The integrated verification 10-Step Method combines several processes into a PRT workflow. These results provide confidence that PRT can be successfully applied for populating 3-D grids in non-cored wells and inter-well areas. Results from this 10-Step method reduces uncertainty and provides a step by step workflow process, that starts with determistic PRT and then applies a IPSOM method verifying the solution.
This workflow process combines deterministic techniques with neural networks using the following steps after core-log integration and deterministic petrophysical rock types are determined ( The number of PRTs (based on cutoffs) is selected using common statistical analysis approaches (such as Cumulative Distribution Functions, histograms or probability plots). The best PRT grouping is determined from the shape of log computed Sw and Swirr compared to the shape of PRT and pore throat radius indicators in depth space. Further validation of PRT includes comparing results to geological facies and mercury injection capillary pressure (MICP or HPMI), special core analysis results and apply a saturation height model (SHM) to verify the definitions of the Deterministic Petrophysical Rock Types (DPRT). Then repeat the SHM process after the probabilistic PRT are determined in Step 10. Core and Log based thickness-weighted averages are computed and compared for each DPRT. Core-Log X-Y cross plots are prepared for each method. Select a limited number of "PRT training points 1-3" for each of the PRTs as determined in steps 1-4. Apply an IPSOM neural network and Heterogeneous Reservoir Analysis (HRA) then compare predicted probabilistic PRT to initial deterministic PRT in depth space and cross-plot space and repeat until the best statistical results are obtained. Repeat IPSOM neural network analysis using "no training points" and evaluate PRT results. Individual wells can be further verified using Multi-Component Advanced Flow Unit Plots and confirm reservoir flow and storage capacities relate to PRT. Completing a final verification of identifying the "best PRT" includes comparing core-log based saturations with SHM model predicted Sw, free water level and honoring the geological column height based on DPRT.
The number of PRTs (based on cutoffs) is selected using common statistical analysis approaches (such as Cumulative Distribution Functions, histograms or probability plots).
The best PRT grouping is determined from the shape of log computed Sw and Swirr compared to the shape of PRT and pore throat radius indicators in depth space.
Further validation of PRT includes comparing results to geological facies and mercury injection capillary pressure (MICP or HPMI), special core analysis results and apply a saturation height model (SHM) to verify the definitions of the Deterministic Petrophysical Rock Types (DPRT). Then repeat the SHM process after the probabilistic PRT are determined in Step 10.
Core and Log based thickness-weighted averages are computed and compared for each DPRT.
Core-Log X-Y cross plots are prepared for each method.
Select a limited number of "PRT training points 1-3" for each of the PRTs as determined in steps 1-4.
Apply an IPSOM neural network and Heterogeneous Reservoir Analysis (HRA) then compare predicted probabilistic PRT to initial deterministic PRT in depth space and cross-plot space and repeat until the best statistical results are obtained.
Repeat IPSOM neural network analysis using "no training points" and evaluate PRT results.
Individual wells can be further verified using Multi-Component Advanced Flow Unit Plots and confirm reservoir flow and storage capacities relate to PRT.
Completing a final verification of identifying the "best PRT" includes comparing core-log based saturations with SHM model predicted Sw, free water level and honoring the geological column height based on DPRT.
Results of applying this new method are it improves and refines the PRT process, reduces uncertainty and subjective interpretations. Reducing uncertainity is important, especially when petrophysical rock types are the basis to compute the initial fluid saturations at each grid node and assign dynamic properties such as relative permeability curves in reservoir simulations. This new method provides a verification process that uses both deterministic and probabilistic techniques. These final PRTs are coupled with a saturation height model can be extended to fill 3D volumes and fluid distributions.
Al-Farisi, Omar (Abu Dhabi Marine Operating Co.) | Belhaj, Hadi (The Petroleum Inst.) | Ghedan, Shawket (The Petroleum Inst.) | Negahban, Shahin (Abu Dhabi Co. Onshore Oil Operation) | Gomes, Jorge (The Petroleum Inst.) | Yammahi, Fatmeh (The Petroleum Inst.) | Amr, Mohamed (The Petroleum Inst.) | Khamissa, Hocine (The Petroleum Inst.) | Ibrahim, Khalil (The Petroleum Inst.) | Al-Shamsi, Abdullah (The Petroleum Inst.)
The heterogeneity in carbonate rocks, made it hard for Geoscientists andReservoir Engineers to define a universal classification methodology that isable to honour the critical reservoir static properties. Most classifications,like, lithofacies, capillarity and textural methods have based their rocktyping concept on one or two static properties, then tried to find an analog toother static properties to cluster or group them, then worked to populate therock types across the whole field. However, from field observations andexperiences of utilizing these conventional techniques, it was obvious thatthey suffered from several gaps, like inability to have the properties analogconsistent throughout the whole reservoir. Moreover, the groups or clustershave big dispersion that produced overlaps, and then theoretically they couldnot fully honour the physics and rock properties links.
Therefore, in this study, rock typing is made to honour static properties alltogether through changing the classification concept to resolve the gaps of thetraditional methodologies. The ultimate objective of all reservoircharacterization and rock classification is to enable building geological andsimulation models, with optimum honouring of rock properties. To achieve thisobjective, the established framework in this research is based on analyzing theeffects of each of the rock properties on another and the value and impact thateach can add to the models most critical parameters. By this technique, the gapof pore and pore-throat network is resolved through Multiple PropertiesIntersection.
This Integrated Carbonate Rock Typing technique starts with capturing theheterogeneity of carbonate rock by generating matrix of core permeability,capillary pressure (end point, threshold pressure and Plateau), pore-throatsize distribution and porosity. Then intersecting this matrix to constructweighted links between these properties and identify unique groups. Resultedclasses are novel carbonate rock type classes that entered to feedback analysisnode to explore and validate the logic of linked physics to tune the classes'thresholds and assure no overlap between any of classification properties.Finally for utilizing this technique in non-cored wells, an analog with loggingdata is structured through novel permeability, capillary pressure andsaturation function called the C-Function to be the replacement of theJ-Function in Carbonate.
Al-basry, Ali Hassan (Abu Dhabi Co. Onshore Oil Opn.) | Al-Hajeri, Salma Khalfan (ADCO) | Saadawi, Hisham N.H. (Abu Dhabi Co. Onshore Oil Opn.) | Al Aryani, Fatema Mohamed (ADCO) | Al Obeidi, Adel (ADCO) | Negahban, Shahin (ADCO) | Bin-dhaaer, Ghaniya Salim
The paper presents compositional data and PVT data for a Middle East reservoir fluid with a reservoir temperature of 394 K and reservoir pressure of 287 bar. The PVT data was selected and designed to provide the best possible starting point for developing an EOS model that would accurately reproduce the phase behavior of a reservoir fluid subject to injection of either CO2 or a hydrocarbon gas.
To eliminate the uncertainty from use of default molecular weights and densities for the C7+ hydrocarbon fractions the reservoir fluid composition was analyzed using a True Boiling Point (TBP) analysis. PVT experiments, both routine and gas injection (EOR) experiments, were carried out including solubility swelling, equilibrium and multi contact experiments and slim tube tests. With both injection gases the reservoir fluid shows a combined vaporizing/condensing drive mechanism.
A 9-component EOS model was developed for the volume corrected Peng-Robinson equation of state, which shows a good match of all available data. Two methods were used to predict the vaporizing/condensing MMP; (a) a multi-component tie-line MMP algorithm and (b) a compositional 1D simulator. The CO2 MMP is considerably lower than the reservoir pressure while the MMP seen with the hydrocarbon gas is close to the reservoir pressure.
Carbon Dioxide (CO2) Enhanced Oil Recovery (EOR) process is most likely to become the preferred hydrocarbon recovery process in future in Abu Dhabi. The complexity and cost of implementing large scale EOR projects require the development of a detailed EOR strategy, clearly defined targeted objectives, a visionary work-plan (roadmap) and staged evaluation prior to full field commercial implementation. Abu Dhabi Company for Onshore Oil Operations (ADCO) is in the process of conducting the first Middle-East CO2-EOR pilot in an onshore complex carbonate reservoir.
This paper discusses the design and implementation of the first ADCO CO2-EOR Pilot Project which addresses key technical and business uncertainties and risks associated with CO2 injection in ADCO reservoirs in the future. The pilot was considered the best approach to evaluate feasibility of this EOR approach on a field scale in addition to other verifications through laboratory and simulation studies.
The project started with a company-wide screening study which attempted to identify both the most appropriate EOR option for ADCO reservoirs and the best reservoir candidates. Subsequently, an in-house simulation study was conducted to confirm the best identified candidate reservoir. The Roadmap defined studies required to fill the required data gaps. Advanced CO2-PVT, Asphaltene and SCAL studies were conducted to reduce the uncertainties related to CO2 injection. The pilot objectives and constrains were clearly defined at the early stage of the project. Challenges of obtaining the required CO2 volumes for injection were overcome.
A history matched compositional reservoir simulation model was used as an effective tool to design and optimize the pilot. Sufficient time and effort were expended in the pilot design and optimization in order to meet the implementation objectives in a timely manner. These include drilling of the wells, construction of the surface facilities, development and execution of the surveillance and monitoring plan and finally the operation of the pilot.
This paper outlines the design and implementation of the first ever Middle-East CO2-EOR pilot in an On-shore complex carbonate field in Abu Dhabi, UAE.
Minimizing uncertainty associated with predicted log water saturation can be best achieved through the integration of log analysis (formation evaluation) and capillarity (core measured high pressure mercury and oil/water drainage capillary pressure data). Further reduction of uncertainty in predicted formation water saturation requires; first the most probable log porosity solution interpretation; second the most probable log porosity/permeability relationship (match core measured values); and third a quantitative capillary pressure model that represents measured high pressure mercury and oil/water drainage capillary pressure data.
This paper is a continuation of previous work done using the classical Leverett-J method (Leverett 1941) in integrating capillary pressure data with routine core analysis and petrophysical rock types to generate a robust model for predicting formation water saturation profiles. Two datasets were used in the application of this method in two phases. Phase 1 was accomplished by using the first dataset, which is for a crestal well, in characterizing reservoir rock types by the assignment of capillary pressure curves to their corresponding rock types. This process was accomplished by grouping rock types according to discrete ranges of the interval speed which is defined as the square root of core measured permeability divided by porosity ( k / f ) data. Each range of the interval speed will have a corresponding Leverett J function from which water saturations will be interpolated according to its respective height above the Free Water Level (FWL) (using the log predicted permeability-porosity ratios). In phase 2, the second dataset, which is from a flank well, was used to quality check the rock typing interpretation and to aerially map the change in reservoir rock types when going from crest to flank. One of the main findings from this work is the importance of quantifying the cementation process that took place in rocks' post-deposition.
In the absence of clear fluid contact, a trial and error method was followed to adjust the FWL until a good agreement between the log predicted water saturation and the predicted water saturation from the capillary pressure correlation was reached. The error in averaging was minimized by increasing the number of rock types (minimizing the ranges of interval speed) using a large number of measured high pressure mercury and oil-water drainage capillary pressure data. Furthermore, this paper presents a comprehensive comparison of the log derived and model predicted water saturations.