Upscaling and History Matching of Fractured Reservoirs Pål Næverlid Sævik Department of Mathematics University of Bergen Modeling and Inversion of Geophysical Data (Uni CIPR) March 26, 2015
Outline 2 EnKFSimulation
Fractured Rocks What makes fractures different from other heterogeneities? 3
What makes fractures special? Dual porosity behaviour Scale separation issues Heterogeneities are larger than lab scale Prior information on fracture geometry may be available 4
Dual porosity behaviour 5
Scale separation issues Large faults and fractures may be impossible to upscale 6
Large and small fractures The distinction between «large» and «small» fractures is determined by the size of the computational cell 7
Prior fracture information Core samples Well logs Outcrop analogues Well testing Seismic data EM data (?) 8
Fracture parameters 9 Roughness Aperture (thickness) Filler material Connectivity Fracture density Clustering Shape Size
Common assumptions 10 Fisher distribution of orientations Power-law size distribution
Numerical upscaling Flexible formulation Accurate solution Slow Gridding difficulties May not have sufficient data to utilize the flexible formulation 11
Analytical upscaling Idealized geometry Fast solution Easy to obtain derivatives Requires statistical homogeneity Difficult to link idealized and true fracture geometry 12
Effective permeability 13 p in p out K1K1 K2K2 φ 1 K 1 + φ 2 K 2 p in p out φ 1 K 1 + φ 2 K 2 = A·a·K 1 + K 2 = A·τ + K 2
Several fracture sets 14
Partially connected fractures 15
Percolation theory 16
Connectivity prediction 17
Connectivity and spacing 18
Transfer coefficient 19
Summary: Input and output parameters 20 Connectivity f Density A Transmissitivity τ Density A Aperture a Filler material Permeability K Porosity φ Transfer coefficient σ Orientation Shape Size Clustering Roughness
History matching of fractured reservoirs 21 EnKF Simulation
Integrated upscaling and history matching 22
Ensemble Kalman Filter update 23
Test problem: Permeability measurement Single grid cell Measured permeability: 200 mD ± 20 mD Expected aperture: 0.2 mm ± 0.02 mm Expected density: 1 m -1 ± 0.2 m -1 Randomly oriented, infinitely extending fractures Cubic law for transmissitivity 24
Test problem: Permeability measurement 25
Predicted fracture porosity 26
Predicted transfer coefficient 27
Inverse relation and connectivity 28
Predicted connectivity 29
Linear fracture upscaling 30
Predicted connectivity 31
Partially connected fractures 32
Predicted connectivity 33
Field case: PUNQ-S3 Three-phase reservoir 6 production wells 0 injection wells (but strong aquifer support) Dual continuum extension with capillary pressure Constant production rate 34
Field case: PUNQ-S3 2 years of production 2 years of prediction Data sampling every 100 days Data used – GOR – WCT – BHP Assimilation using LM- EnRML 35
Data match summary Number of LM-EnRML iterations Fracture parameters as primary variables BHP GOR WCT Total Upscaled parameters as primary variables BHP GOR WCT Total
37 BHP, PRO-1 GOR, PRO-12 WCT, PRO-11 Initial ensembleTraditional approachOur approach
38 Permeability Sigma factor Initial ensembleTraditional approach Our approachTrue case
39 Permeability Connectivity Initial ensembleTraditional approach Our approachTrue case
Conclusion Fracture upscaling creates nonlinear relations between the upscaled parameters These relations may be lost during history matching, if upscaled parameters are used as primary variables The problem can be avoided by history matching fracture parameters directly 40