Print Email Facebook Twitter Upscaling polymer flooding Title Upscaling polymer flooding: Developing a workflow based on particle swarm optimization Author De Vries, H.C. Contributor Van Kruijsdijk, C.P.J.W. (mentor) Faculty Civil Engineering and Geosciences Department Geoscience & Engineering Date 2016-09-02 Abstract In this research a workflow for upscaling polymer flooding is developed. The workflow generates pseudorelative permeabilities (pseudos) to represent flow behaviour affected by fine scale heterogeneities in a coarse grid. With this workflow, a better estimate of the difference in production from a polymer flood compared to a water flood can be made. In case history-matched pseudos for a water flood are readily available, the workflow provides insight in whether and how the pseudos generated for water should be adjusted to correctly model a polymer flood. In case no production data and no pseudos are available for a water flood, the workflow can be used to generate pseudos for both a water and a polymer flood. Using pseudos in a coarse model is needed because heterogeneity affecting connectivity and thereby flow behaviour is not (fully) represented as a result of upscaling. Therefore an error may be made in cumulative oil and water production when simulating a water or a polymer flood in a coarse model. Pseudos can be used to represent flow behaviour from a fine grid in a coarse grid. Polymer flooding improves sweep around heterogeneities compared to water flooding. This means flow behaviour differs for water and polymer flooding. As a result, pseudos for water and polymer flooding may differ (depending on circumstances). The need for the use of pseudos when using an upscaled grid is assessed by comparing cumulative production from a coarse grid simulated with optimized pseudos to cumulative production from a coarse grid simulated with rock relative permeabilities. The difference in upscaling a water flood from a polymer flood and the need for adjustment of pseudos generated for a water flood when simulating a polymer flood is assessed by comparing simulating a polymer flood in a coarse grid with pseudos optimized for polymer flooding to simulating a polymer flood in a coarse grid with pseudos optimized for a water flood. A sectormodel is used as a basis for upscaling. The upscaling procedure is based on adjustment of relative permeabilities using Particle SwarmOptimization, which is an exploring optimization algorithm. Endpoint water relative permeability, kr,we , and the Brooks-Corey coefficient for oil, no, are selected as optimizing parameters.The least square difference between cumulative oil production from the fine grid and from the coarse grid is taken as the objective function to be minimized. Results are post-processed by converting non-unique pseudos resulting from the optimization in unique fractional flow curves. A novel way of characterizing the fractional flow curve is developed, which allows representation of the entire curve through vshock . This provides insight in how pseudos change in sensitivity studies. Generally, the match in cumulative production is very good. The models investigated in this work show that the improvement of generating pseudos compared to simulating a coarse grid with fine grid relative permeabilities is significant. In 3D channelized turbidite models, shale draping causes a too late breakthrough and an overestimation in recovery when pseudos are not adjusted, e.g. when rock relative permeabilities are used in the coarse grid. Errors in breakthrough time are reduced from 0.3 PV injected when not adjusting relative permeabilities to 0.05 PV injected. Errors in late time recovery are reduced from a 0.2 to 0 difference in recovery (fraction of OIIP). The required adjustment in rock relative permeabilities to match cumulative production depends on heterogeneity and method and degree of upscaling. Sensitivity of pseudos to polymer concentration, and thus the need for adjustment of water pseudos when simulating a polymer flood, differs for surface-based upscaled models and voxelized models and depends on heterogeneity. Deriving a general trend for the difference between water and polymer pseudos fromthiswork is therefore not possible. In surface-based upscaled models, vshock increases, e.g. the fractional flow curve shifts left, with increasing polymer concentration. As a result, breakthrough time is estimated too late and recovery is overestimated when simulating a polymer flood with water pseudos. In voxelized models, vshock decreases, e.g. the fractional flowcurve shifts right,with increasing polymer concentration. As a result, breakthrough time is estimated too early and recovery is underestimated when simulating a polymer flood with water pseudos. Robustness of generated pseudos to changes in injection rate, amount of PV injected and streamlines due to different perforated intervals or well-placement is tested and found to be good for the simulated cases. Subject polymer floodingupscalingrelative permeabilityfractional flowturbiditesParticle Swarm Optimization To reference this document use: http://resolver.tudelft.nl/uuid:a93509ca-477a-48aa-9ee2-83adcb494dd8 Part of collection Student theses Document type master thesis Rights (c) 2016 Vries, H.C. de Files PDF Thesis_final_version_Henn ... _Vries.pdf 14.23 MB Close viewer /islandora/object/uuid:a93509ca-477a-48aa-9ee2-83adcb494dd8/datastream/OBJ/view