A Physics-Based Data-Driven Model for History Matching, Prediction and Characterization of Unconventional Reservoirs*

Similar documents
PET467E-Analysis of Well Pressure Tests 2008 Spring/İTÜ HW No. 5 Solutions

Chapter Seven. For ideal gases, the ideal gas law provides a precise relationship between density and pressure:

Reservoir Flow Properties Fundamentals COPYRIGHT. Introduction

Pressure Transient Analysis COPYRIGHT. Introduction to Pressure Transient Analysis. This section will cover the following learning objectives:

Rate Transient Analysis COPYRIGHT. Introduction. This section will cover the following learning objectives:

Parameter Estimation in Reservoir Engineering Models via Data Assimilation Techniques

SPE Comparison of Numerical vs Analytical Models for EUR Calculation and Optimization in Unconventional Reservoirs

Petroleum Engineering 324 Reservoir Performance. Objectives of Well Tests Review of Petrophysics Review of Fluid Properties 19 January 2007

Dimensionless Wellbore Storage Coefficient: Skin Factor: Notes:

XYZ COMPANY LTD. Prepared For: JOHN DOE. XYZ et al Knopcik 100/ W5/06 PAS-TRG. Dinosaur Park Formation

A NEW SERIES OF RATE DECLINE RELATIONS BASED ON THE DIAGNOSIS OF RATE-TIME DATA

SPE Uncertainty in rock and fluid properties.

Petroleum Engineering 324 Reservoir Performance. Objectives of Well Tests Review of Petrophysics Review of Fluid Properties 29 January 2007

The SPE Foundation through member donations and a contribution from Offshore Europe

Figure 1 - Gauges Overlay & Difference Plot

Petroleum Engineering 324 Well Performance PRACTICE Final Examination (Well "B") 05 May 2003 (08:00-10:00 a.m. RICH 302)

Petroleum Engineering 613 Natural Gas Engineering. Texas A&M University. Lecture 07: Wellbore Phenomena

National Exams May 2016

Before beginning, I would like to acknowledge the amazing contributions of Ken Nolte. I suspect that the origins of most of our discussion during

Petroleum Engineering 324 Well Performance Daily Summary Sheet Spring 2009 Blasingame/Ilk. Date: Materials Covered in Class Today: Comment(s):

and a contribution from Offshore Europe

Shale Gas Reservoir Simulation in Eclipse

WATER INFLUX. Hassan S. Naji, Professor,

Workflow for Applying Simple Decline Models to Forecast Production in Unconventional Reservoirs

A NOVEL APPROACH FOR THE RAPID ESTIMATION OF DRAINAGE VOLUME, PRESSURE AND WELL RATES. A Thesis NEHA GUPTA

Coalbed Methane Properties

If your model can t do this, why run it?

Technology of Production from Shale

The role of capillary pressure curves in reservoir simulation studies.

A Better Modeling Approach for Hydraulic Fractures in Unconventional Reservoirs

Inflow Performance 1

Propagation of Radius of Investigation from Producing Well

Flow of Non-Newtonian Fluids within a Double Porosity Reservoir under Pseudosteady State Interporosity Transfer Conditions

Petroleum Engineering 324 Well Performance Daily Summary Sheet Spring 2009 Blasingame/Ilk. Date: Materials Covered in Class Today: Comment(s):

Multiscale Investigation of Fluid Transport in Gas Shales. Rob Heller and Mark Zoback

Presentation of MSc s Thesis

Far East Journal of Applied Mathematics

Oil and Gas Well Performance

Imperial College London

SPE Copyright 2003, Society of Petroleum Engineers Inc.

Novel Approaches for the Simulation of Unconventional Reservoirs Bicheng Yan*, John E. Killough*, Yuhe Wang*, Yang Cao*; Texas A&M University

Derivation of the fractional flow equation for a one-dimensional oil-water system. Consider displacement of oil by water in a system of dip angle α

Simulation of Naturally Fractured Reservoirs with Dual Porosity Models

Evaluation and Forecasting Performance of Naturally Fractured Reservoir Using Production Data Inversion.

A Better Modeling Approach for Hydraulic Fractures in Unconventional Reservoirs

Faculty of Science and Technology MASTER S THESIS

GENERALIZED PSEUDOPRESSURE WELL TREATMENT

APPLICATIONS OF LEVEL SET AND FAST MARCHING METHODS IN RESERVOIR CHARACTERIZATION

SPE MS. Copyright 2014, Society of Petroleum Engineers

Chapter 3 Permeability

Pressure-Transient Behavior of DoublePorosity Reservoirs with Transient Interporosity Transfer with Fractal Matrix Blocks

Measure Twice Frac Once

Reservoir Simulator Compaction Modelling: A Predictor for Accelerated Coupled Rock Mechanics -- Reservoir Simulation

Petroleum Engineering 324 Well Performance Daily Summary Sheet Spring 2009 Blasingame/Ilk. Date: Materials Covered in Class Today: Comment(s):

Sand Control Rock Failure

Perspectives on the Interpretation of Flowback Data from Wells in Shale Reservoir Systems

Training Venue and Dates Ref # Reservoir Geophysics October, 2019 $ 6,500 London

Module for: Analysis of Reservoir Performance Introduction

Modeling of 1D Anomalous Diffusion In Fractured Nanoporous Media

Effect of Pressure-Dependent Natural-Fracture Permeability on Shale-Gas Well Production

A PSEUDO FUNCTION APPROACH IN RESERVOIR SIMULATION

Well Test Interpretation

SPADES: Swift Production Data Analysis and Diagnostics Engine for Shale Reservoirs

Production System Analysis

CO 2 Foam EOR Field Pilots

Geostatistical History Matching coupled with Adaptive Stochastic Sampling

Estimation of Relative Permeability Parameters in Reservoir Engineering Applications

Integrated Approach to Drilling Project in Unconventional Reservoir Using Reservoir Simulation

STUDY OF WATERFLOODING PROCESS IN NATURALLY FRACTURED RESERVOIRS FROM STATIC AND DYNAMIC IMBIBITION EXPERIMENTS

The SPE Foundation through member donations and a contribution from Offshore Europe

SPE Copyright 2008, Society of Petroleum Engineers

Petroleum Geomechanics for Shale Gas

Local Time Step for a Finite Volume Scheme I.Faille F.Nataf*, F.Willien, S.Wolf**

A BENCHMARK CALCULATION OF 3D HORIZONTAL WELL SIMULATIONS

Shale Gas Well Test Analysis

Exploration / Appraisal of Shales. Petrophysics Technical Manager Unconventional Resources

A Multi-Continuum Multi-Component Model for Simultaneous Enhanced Gas Recovery and CO 2 Storage in Stimulated Fractured Shale Gas Reservoirs Jiamin

Radial- Basis Function Network Applied in Mineral Composition Analysis

Excellence. Respect Openness. Trust. History matching and identifying infill targets using an ensemble based method

Optimized Recovery from Unconventional Reservoirs: How Nanophysics, the Micro-Crack Debate, and Complex Fracture Geometry Impact Operations

READ THIS PAGE COMPLETELY BEFORE STARTING

FLUID FLOW MODELING IN MULTI-STAGE HYDRAULIC FRACTURING PATTERNS FOR PRODUCTION OPTIMIZATION IN SHALE RESERVOIRS

Recent Work in Well Performance Analysis for Tight Gas Sands and Gas Shales

IMPERIAL COLLEGE LONDON. Department of Earth Science and Engineering. Centre for Petroleum Studies

Procedural Animation. D.A. Forsyth

Model Inversion for Induced Seismicity

Non-Darcy Skin Effect with a New Boundary Condition

Introduction to Well Stimulation

IMPERIAL COLLEGE LONDON

A New Method for Calculating Oil-Water Relative Permeabilities with Consideration of Capillary Pressure

MACHINE LEARNING FOR PRODUCTION FORECASTING: ACCURACY THROUGH UNCERTAINTY

What Can Microseismic Tell Us About Hydraulic Fracturing?

Fracture-matrix transfer function in fractured porous media

(Formation Evaluation and the Analysis of Reservoir Performance) Module for: Analysis of Reservoir Performance. Introduction

Flow equations The basic equation, on which all flow equations are based, is Darcy s Law for radial flow is given by: p

Jornadas de Producción, Tratamiento y Transporte de Gas El Desafío del Gas no Convencional

SPE Copyright 2008, Society of Petroleum Engineers

Subsurface Maps. K. W. Weissenburger. Isopach. Isochore. Conoco, Inc. Ponca City, Oklahoma, U.S.A.

(Page 2 of 7) Reservoir Petrophysics: Introduction to Geology (continued) Be familiar with Reservoir Petrophysics (continued)... Slides Reservoi

Machine Learning Applied to 3-D Reservoir Simulation

Transcription:

A Physics-Based Data-Driven Model for History Matching, Prediction and Characterization of Unconventional Reservoirs* Yanbin Zhang *This work has been submitted to SPEJ and under review for publication

Motivation

Reservoir Characterization with Play-Doh

Reservoir Characterization with Play-Doh Reservoir 4

Reservoir Characterization with Play-Doh Reservoir Wellbore 5

Reservoir Characterization with Play-Doh Reservoir Wellbore 6

Reservoir Characterization with Play-Doh 4 5 6 Reservoir Wellbore 7

Reservoir Characterization with Play-Doh 6 5 4 4 5 6 Reservoir Wellbore 8

Reservoir Characterization with Play-Doh y 6 5 4 4 5 6 Wellbore x 9

Reservoir Characterization with Play-Doh Fractures y fracture 4 4 4 x 0

Diffusive Diagnostic Function (DDF) D D Set of contours of the pressure solution in D well well PV PV PV PV 4 PV 5 PV PV PV PV 4 T 0 T T T T 4 Each ring is represented as a single cell PV 5 From Darcy s Law Qቚ = ර Σ Σ Tቚ = ර Σ Σ vds = ර Σ k ds dl ds ȁ Σ P dl A contour surface of the pressure solution in D ȁ Σ P+dP k dp dl ds = Tቚ dp Σ dpvቚ = ර Σ Σ φ dl ds In physical space, we need two functions PV x and T(x) PV = σδξ ቐ T = σ Δξ dξቚ Σ dpvȁ Σ = Σ Tȁ Σ Σ φ dl ds k ds dl = φdl Σ k dl Σ Σ ξȁ Σ = Σ0 dξ where Σ0 is the completion sand face σ(ξ) In dimensionless ξ space, we only need one function σ ξ which is called the DDF σቚ Σ dpvቚ Tቚ = Σ Σ ර Σ φ dl ds ර Σ k ds dl k = Sቚ φdl Σ Σ dl Σ

The D Simulation Model with DDF y 6 5 well unit: ft md / σ -Dimension N Grid blocks PV i = σ i + σ i (ξ i ξ i ) 4 σ 4 T i = σ i (ξ i+ ξ i )/ x Well P wf σ 0 σ σ σ ξ ξ ξ ξ 4 σ N ξ N ξ N J = σ 0 ξ / ξ unit: ft/md / We are doing the same, old, regular reservoir simulation except that we replace the D grid with DDF Complex fluid and rock model Changing well constraints Capillary pressure Adsorption Coupled with wellbore flow modeling and surface network Caveat Remember we reduce D reservoir into a D model and that is an approximation However, as we will show later, it is quite a good approximation in many cases

All this is good, but... How do I know if I should cut my reservoir this way or that way? 4 5 6 4 y 6 y 5 4 4 x x

All this is good, but... How do I know if I should cut my reservoir this way or that way? 4 5 6 The bad news: we don t know in general 4 y 6 y 5 4 4 x x 4

All this is good, but... How do I know if I should cut my reservoir this way or that way? 4 5 6 y 6 5 4 4 y 4 σ The bad news: we don t know in general The good news: () We can guess and we ll make a lot of guesses () We can adjust our guesses by history matching DDFs forward modeling ξ history matching q Production Data History Matching using ESMDA t x x Ensemble Smoother with Multiple Data Assimilation 5

We don t do wild guesses; we guess based on DDF Characteristics x f radial flow linear flow d f r w r x f L w d N f σ slope:πkh infinite acting σ σ 0 = 4x f h kφ infinite acting σ 0 ~4x f N f h kφ σ σ 0 = πr w h kφ boundary boundary SRV p slope~αkh boundary r w = r w e s 0 ξ = r φ/k ξ 0 ξ = d φ/k ξ σ~(4x f + L w + 4d f )h 0 kφ ξ SRV ~ d f φ/k ξ Take out your Play-Doh and construct this DDF! 6

DDF Characteristics - Summary Characteristics of the DDF Diagnostic Properties Approximate Equations Comments σ level Flow area and Reservoir quality σ~a kφ Fractures cause sharp increase in σ level. Interferences or boundaries cause drop in σ level. σ level keeps constant for linear flow Slope of linearly increasing σ ξ Reservoir permeability Δσ Δξ ~αkh α depends on flow pattern. α = π for radial flow. Generally, α > π for irregular flow pattern. Area under the DDF curve Pore volume Area = V p interferences. For unconventional reservoirs, SRV may A dramatic drop in σ level signifies boundary effect or be identified in this way. ξ at which σ behavior changes Distance ξ~d φ k The estimation of distance is difficult to be precise because the transition of σ behavior is usually not clear-cut and may span a wide range. 7

Synthetic Example: Vertical Well D Cartesian grid 0 0 DX = DY = 9.9 ft DZ = 00 ft Before HM r w = 0.ft r e = 500ft k = 0.00md φ = 0.05md h = 00ft After HM ξ (ft/md / ) (a) Time (days) (b) Black oil fluid model (P b = 800 psia) Initial reservoir pressure P i = 5000 psia Initial water saturation S wi = S wir = 0.5 Well producing at constant BHP = 000 psia ξ (ft/md / ) (c) Time (days) (d) 8

Synthetic Example: Multiple Fractures Before HM After HM ξ (ft/md / ) (a) Time (days) (b) 0 Infinite conductivity fractures Other parameters are the same as previous slide ξ (ft/md / ) (c) Time (days) (d) 9

Applications Physics Based Complex fluid, multi-phase flow Data Driven Extremely fast history match Reservoir Characterization Total fracture area and SRV Integrated Workflow Coupled with surface network Optimization / uncertainty analysis History matching and forecasting for a Gas Well Oil Well Examples 0

Summary. Physics-Based: DDF provides a general D simulation framework to approximate D reservoir. Data-Driven: DDF is probabilistically conditioned to production data Future Research First-principle-based computation Machine learning and big data DDF?

Acknowledgement ETC/RPP Jincong He Jiang Xie Xian-huan Wen ETC/RPS Robert Fitzmorris Shusei Tanaka ETC/PEWP Jorge Acuna ETC/TRU Reza Banki MCBU Baosheng Liang Hannah Luk AMBU James Wing Richin Chhajlani Please reach out to me for any questions or to connect with me. You may contact me at: Yanbin.Zhang@chevron.com

Backup Slides

History Matching using ESMDA with DDF Ensemble Smoother with Multiple Data Assimilation Diffusive Diagnostic Function σ model update DDFs m i+ = m i + Δm forward modeling ξ history matching q Production Data data mismatch Δm = C MD C DD + α i C D (d obs d pred ) perturbed observations / d obs = d obs + α i C D zd where z d ~N(0, I Nd ) t Workflow. Come up with initial ensemble of DDFs. Perform forward modeling to obtain data prediction a) predictions way off, go back to b) if predictions follow the trend and cover the range of observed data, go to. Randomize the model to avoid ensemble collapse 4. For the ith (out of n) iteration, use α i = /n in the equation to update the ensemble of models considering the mismatch of all data points simultaneously 5. Model regularization by smoothing the DDF curves and eliminate negative values (set to 0) 6. Go to 4 for next iteration 4

Synthetic Example: Single Fracture Infinite Conductivity vs. Finite Conductivity Infinite conductivity fracture k f = 0 6 md σ starts at a much larger value σ 0 = 4x f h kφ = 5 ft md / ξ (ft/md / ) Finite conductivity fracture k f = md σ starts at a smaller value, but increases rapidly near ξ = 0 Infinite or finite conductivity single fracture Other parameters are the same as previous slide ξ (ft/md / ) T 5

Synthetic Example: Multiple Fractures DDF Diagnostics σ ~500 ft md / x f L w hφ = 6.5x0 5 ft SRV p ~ 6x0 5 ft x f d f σ ~000 ft md / Slope ~ ft md L w ξ i ~40 ft/md / N f L w = 450 ft N f = 0 d f = 50 ft σ σ ~.5 ξ (ft/md / ) ξ (ft/md / ) (a) (b) x f ~ 0 ft φ = 0.05 σ ~ 500 ft md / k ~ 0.00 md σ ~4x f N f h kφ σ ~(4x f + L w + 4d f )h σ 0 SRV p kφ slope~αkh ξ SRV ~ d f φ/k boundary ξ 6

Normalized Gas Potential (psi/cp/mscfd) Bottomhole Pressure (psia) Gas production Rate (MMSCF/D) Well length (L w ) 0 ft Number of hydraulic fractures (N f ) Reservoir thickness (h) 50 ft Reservoir porosity (φ) 0.065 Initial reservoir pressure (P i ) 5008.8 psia Reservoir temperature (T) 60 F Connate water saturation (S wc ) 0.9 Rock compressibility (c f ) 0-6 psi - Gas specific gravity (γ) 0.57 Square Root Time Plot Field Example: Marcellus Gas Well 5000 4500 4000 500 000 500 000 500 000.0 500 0 0.0 0 500 000 500 000 E+07 HM prediction Log-Log Plot Time (days) Integral of Normalized Gas Potential Bourdet Derivative 6.0 5.0 4.0.0.0 E+06 A straight line can be used to fit the data even though it is not linear flow Not half slope E+05 00 000 0000 00000 Material Balance Time (hr) 7

Field Example: HM and prediction with DDF Before HM ξ (ft/md / ) After HM ξ (ft/md / ) (a) Time (days) (b) ξ (ft/md / ) (c) Time (days) (d) 8

Field Example: DDF Diagnostics Characteristic of finite conductivity fractures x f d f L w N f ξ (ft/md / ) σ SRV p ~.6 0 7 ft σ ~4x f N f h kφ A total kφ~ 0 4 ft md / φ = 0.065 A total k~4 0 4 ft md / SRV p slope~αkh boundary Total fracture sandface area σ ~(4x f + L w + 4d f )h 0 kφ ξ SRV ~ d f φ/k ξ 9

Field Example: Probabilistic Nature of the DDF Method HM P50 SRV p decreases ξ (ft/md / ) Time (days) SRV p uncertainty range decreases HM ξ (ft/md / ) Time (days) 0