SELF-LEARNING RESERVOIR MANAGEMENT

SPE 84064
L. Saputelli1,2,3, M. Nikolaou2, and M. J. Economides2,3
1PDVSA, 2University of Houston, 3SPE member

PRH 034 - Formao de Engenheiros nas reas de Automao, Controle, e Instrumentao do Petrleo e Gs aciPG, participante do Programa de Recursos Humanos da ANP - Agncia Nacional do Petrleo - para o setor
Petrleo e Gs - PRH - ANP/MME/MCT
October 21 2004; Florianopolis

Agenda
Motivation: The reservoir management challenge
What is the Problem?,
What have been done?
What are the challenges?

Problem Formulation
The specific objectives and scope of this research
Reservoir modeling and identification
Model Predictive Control
Self Learning Reservoir Management
Conclusions
The Way Forward
2

Objective of this presentation

To review current petroleum production
issues regarding real time decision making
and,
To present the results of a continuous selflearning optimization strategy to optimize
field-wide productivity while satisfying
reservoir physics, production and business
constraints.

Reservoir Management is about a careful orchestration of technology, people & resources

The Reservoir Management Challenge

Exploitation Plan
Well location & number
Recovery mechanism
Surface facilities
Well intervention

Drill,
Drill, build
build &
& Operate
Operate
Production
Well & Facilities

Monitor
Monitor

Compression &
Treatment Plants
Injection
Facilities

Establish
Establish or
or revise
revise
Optimum
Optimum Plan
Plan
Control
Control

Drainage
Area

Subsurface
Subsurface
Characterization
Characterization

Update
Update Model
Model

Hydrocarbon production system suffering major technical problems

Motivation
Traditional Problems
Complex & risky operations
(Drilling, Workover, Prod.)
Poor reservoir prediction &
production forecasting
Limited resources: injection
volumes, facilities, people.
Unpredictability of events:
gas or water, well damage.
Poor decision making ability
to tune systems, thus, not
optimized operations

Current Approach
More front-end engineering
and knowledge sharing
Integrated Characterization &
Modern visualization tools
Multivariable optimization,
reengineering.
Monitoring & control devices,
Beyond well measurements
Isolated optimization trials
with limited success.

Challenges
More data for analysis and
integration limitations.
Long-term studies, Ill-posed
tools, simple or incomplete.
Models are not linked among
different layers
Poor Justification, real time
analysis in early stage.
Decisions made only on few
pieces. Lack of Integration
between subsurface-surface
5

To develop a field-wide continuous self-learning optimization decision engine

Research Specific Objectives

Model based control system used to continuously
optimize three-phase fluid migration in a multi-layered
reservoir
A data-driven model that is continuously updated
with collected production data.
A self-learning and self-adaptive engine predicts the
best operating points of a hydrocarbon-producing
field, while integrating subsurface elements surface
facilities and constraints (business, safety, quality,
operability).
6

Combination of petroleum reservoir physics and process control technologies

Research Framework
Data
Data
Handling
Handling

Model
Model
Building
Building

System
System
Identification
Identification

Reservoir
Reservoir
Performance
Performance

Bi-layer
Bi-layer
Optimization
Optimization

Close-loop
Close-loop
Control
Control

Data handling
Data acquisition, filtering, de-trending, outliers detection

Model building and identification

Gray box modeling: empirical reservoir modeling
Partial least square impulse response, neural network and sub-space

Reservoir performance prediction

Real time Inflow performance and well restrictions
Havlena-Odeh Material Balance

Bi-layer optimization of operating parameters

Reservoir best operating point based on the net present value optimization
Regulatory downhole sleeves and wellhead choke controls

Closed-loop control with history-matched numerical reservoir model

Study of the system behavior in closed-loop
7

Attempt to solve two significant reservoir management challenges

Problem Definition
Injector
Injector -- Producer
Producer Profile
Profile Mngt.
Mngt.

Field-Wide
Field-Wide Management
Management

Control undesired fluid production

Optimization fluid production (< bottle-necks)

Exploit efficiently multilayer horizons

Commingle multilayer reservoirs

Characterize inter-well relationship

Minimize production costs

Maximize reserves and production

Maximize reserves and production

Control from surface measurement

Control from surface measurement

Gas
Crudo

h1

k1

hn

kn

Agua

By collecting data, a digital image is used to make decisions

Traditional (Ideal) Integrated Management Approach

1

9
Implementation

System:
Reservoir, Wells &
Surface Facilities

Data
Collection

Business
Constraints

Database
Parameters
Check Condition
Applications

Reservoir
Model

7
8

Decision
Making

Optimization
Well
Model

Surface
Model

Business
Model
Exploitation
Options
5
9

Spatial distribution of pressure as a time function of saturation

Reservoir Modeling: Fluid Transport in Porous media

Multiphase Darcys Law

vp =

This realization is not used in this

research, since it requires the
knowledge of parameters that cannot
be directly measured

krp K

p
Z
p

p
p
gc

Pressure Laplacian as a function of the saturation change

Continuity
Equation

c=

c
+ ( cv ) = 0
t

M W Ax S p p S p
=
=
p
VM
Ax

krp K

S p
g
c p p p Z =

gc
t p

yi

zi+1

yi+1
zi

Molar density in terms of Porous Volumes

yi-1

zi-1
10

Oil, water and gas flow as linear functions of the drawdown pressure

Reservoir Modeling: Flow through Wellbore

Radial Diffusivity Equation

2 p 1 p p
K
2 +
=
(c f + c) r
r r t

re

General Solution given by Exponential Integral

ct r 2
q

p (r , t ) = pi
Ei

4kh 4kt

rs

rw

pe

Wellbore flow given by logarithmic approximation

pwf = pi

q
4kt
ln
4kh ct rw2

pwf
Proposed IPR for continuous monitoring

Steady-state Equation for the Undersaturated Oil-Flow

qo ,b =

kkro h ( pe pwf )

141.2 Bo o ln (re rw ) + s

Inflow Performance (IPR) for Saturated reservoirs

2
pwf
pwf
pb J
1 0.2
qo = qo ,b +
0.8

1.8
pb
pb

qok = a0 + a1 pek + a2 pwfk + a3 ( pwfk )

qwk = b0 + b1 pek + b2 pwfk + b3 ( pwfk )

q = c0 + c1 p + c2 p + c3 ( p
k
g

k
e

k
wf

k
wf

11

Average Reservoir Pressure is a function of net mass production

Reservoir Modeling: Average Pressure Modeling

Expansion of Oil and Original dissolved gas, Eo

Material Balance Equation

Net Underground =
Withdrawal, F

+ Expansion of Gas Caps, Eg

+ Reduction of Hydrocarbon Pore Volume, E fw
+ Natural Water Influx, We

Simplification

f p ( t ) = g ( N p , G p , W p , We )
p = a0 + a1 qo + a2 qw + a3 qg + a4 qwi

dp
= b1qo + b2 qw + b3 qg + b4 qwi
dt

1
t

p p

k 1

k
k
k
k
c0 + c1 p + c2 pwf
1 + c3 ( pwf 1 ) + c5 pwf 2 + c6 ( pwf 2 )
2

Proposed Pressure Modeling for continuous monitoring

( p) = ( p)
k

k 1

+ c1 + c2 p

k
wf 1

+ c3 ( p

2
k
wf 1

+ c4 p

k
wf 2

+ c5 ( p

2
k
wf 2

12

Vertical flow as non-linear functions of flow rates

Reservoir Modeling: Flow Through Pipes

Mechanical Energy Equation
2

2 f f u dL
udu g
+
+ dz +
+ dWs = 0

gc gc
gc D

dp

1000

Pressure (psia)
2000
3000

5000

0
1,000 STB/D @ 500 SCF/STB; WOR=1
1,000 STB/D @ 1000 SCF/STB; WOR=1
1,000 STB/D @ 500 SCF/STB; WOR=0
1,000 STB/D @ 1000 SCF/STB; WOR=0

2000
4000
Depth (ft)

Single-Phase Solution, Incompressible

2 f f u 2 dL

g
2
p = p1 p2 =
u +
z +
2 gc
gc
gc D

4000

6000
8000
10000

Two-Phase Solution, Hagerdorn & Brown (1965)

( u 2 gc )
dp
fm&
144
=+
+
10
5
dz
z
( 7.413 10 D )
2
m

12000
14000

Proposed Pressure Drop Modeling for Continuous Monitoring

(p

k
wf

pth

= b q + b q + b q + b4 ( q
k
1 o

k
2 w

k
3 g

k 2
o

+ b5 ( q

k 2
w

+ b6 ( q

k 2
g

13

Well operating point given by the intersection of reservoir and tubing performance

Reservoir Modeling: Well Deliverability

Tubing Head Pressure

pwf , [psia]

Choke
Tubing
Performance
pwf , [psia]
q, [BPD]
pwf , [psia]
Fractional Flow of Phases

Bottomhole Flowing Pressure

q, [BPD]
Reservoir
Performance

Reservoir Pressure
q, [BPD]

14

Knowing input-output relationships, reservoir could seen as a process plant

Reservoir as a Process Control System Structure

Backpressure
Ambient Temperature
Flow Restrictions
Injection Fluid Restriction

Measured
Disturbances

Unmeasured
Disturbances

Reservoir Rock Heterogeneity

Reservoir Fluid Distribution
Scheduling

Feed forward path

Manipulated Inputs
Flow Choke

Unmeasured Outputs

Zone Control

Well flowing Pressure: pwf

ESP Speed
Gas

Gas Lift

Controller

Reservoir Pressure: pres

Crudo

Solvent Injection

Agua

Water Injection

Reservoir Saturations: So, Sw

Flow Impairment: S, Krs

Heat Injection

Zone Multiphase Flow: qo, qw, gq

Gas Injection

Drainage Area: A
Tubing Head Pressure: pTHP

Feed back path

Tubing Head Temperature: TTHT

Multiphase Flow: qo, qw, gq
Solid Production, Water Analysis

Measured Outputs
15

Recursive self-adaptive identification

Reservoir Model Identification

Physical System

pk +1 , qk +1

Set point

Reservoir
Reservoir
Simulator
Simulator

sp
wf

qGsp

Measure

Control
Control
Implementation
Implementation

Control

Interpret

qo ,opt

Optimize
Set point

k
k
qok = a0 + a1 pek + a2 pwf
+ a3 ( p wf
)

)
2
k
k
q gk = c0 + c1 pek + c2 pwf
+ c3 ( pwf
)
q = b0 + b1 p + b2 p
k
w

k
e

k
wf

Model
Model
Identification
Identification

+ b3 ( p

k
wf

Reservoir
Reservoir Value
Value
Optimization
Optimization

k
o
k
w
k
g

a0

= b
0
c
0

a1 a2
b1 b2
c1 c2

y k + n + j|k =
Model

a3 p k
e
b3 p k
wf
c3 k 2
( pwf )

n+ N

hu

i = n +1

i k + n + j i

+ dk |k

LS Optimization Loop
= X + e
Y
-1
N
min ei2 ( X T X ) XT Y
a ,b
i =1

qo, g , w = f1 p k , p k 1 ... qTk , qTk 1 ,...

k
k 1
pres = f n p k , p k 1 ... qT , qT ,...

16

Example for Model Identification and Block Diagram

Empirical model whose
structure is determined
by first principles

Identification
Identification

Empirical
Empirical
Model
Model

d
+

Reservoir
Reservoir
(Simulator)
(Simulator)

pwf

qo
qw
qg

Outputs (Y)
Inputs (U)
Producer Flowing Pressure, pwf1 u1
Injector Flowing Pressure, pwf2 u2

qwinj

qo

y1
y2
y3
y4
y5
y6

Reservoir Pressure: P
Oil Rate: qo
Water Rate: qw
Water Fraction: fw
Gas Rate: qg
Water Injection Rate: qwi
17

Model Identification Experimental Set-up

Windows and Eclipse Environment

Generate
Data
File

Run
Eclipse

Reservoir
Numerical
Model

Run
Summary
File

Convert
Eclipse
To Excel

Run
Matlab

Matlab Environment Level

Read
Excel
File

Subspace
Identification
Plot Rsc
Calculated
& Measured

Neural
Network

FIR PLS

x,y

Split Data
Test & Pred

x=0
=1

Auto
Scale

U, Y

Select Input
and Outputs

Ac , Ad

A
cumsum(A)
diff(A)

Rescale
Parameters

Least Squares
Estimator

18

Predictions Using Empirical Structured models

19

Errors Using Empirical models

20

Coefficients Using Empirical models

21

MPC minimizes future prediction error while satisfying input constraints

Model Predictive Control

At time k future predictions of the output y can be made as

y k + n + j|k =

n+ N

i = n +1

hi uk + n + j i + dk |k

dk |k = yk

where

n+ N

hu

i = n +1

i k i

Minimization Problem to solve

p
SP
m in ( y k + n + j | k y
j =1

s .t .

y m in y k + n + j | k y m a x

u m in u k + j 1| k u m a x
u
= u k + m 1| k
k + i |k

+ R u k2 + j 1| k
j =1

j = 1, L , p

j = 1, L , m

i = m ,L , p 1

Controls operation while optimizing performance

Done over a receding or moving horizon
Requires a setpoint from an upper level

Set Point Tracking Example

All Variables normalized so that They have zero mean and Std. Dev = 1
22

Example for Control and Block Diagram

PLS
PLS Impulse
Impulse
Identification
Identification
Empirical
Empirical
Model
Model

qo , sp
qw, sp
qg , sp

qo

+
-

MPC
MPC
Controller
Controller

pwf

Reservoir
Reservoir
(Simulator)
(Simulator)

qo
qw
qg

Outputs (Y)
Inputs (U)

qwinj

qoT

Producer Flowing Pressure, pwf1 u1

Producer Flowing Pressure, pwf2 u2
Injection Flowing Rate, qwinj u3

Layer 1, kh1
Layer 2, kh2

y1
y2
y3
y4
y5
y6

Reservoir Pressure: P
Oil Rate Layer 1: qo1
Oil Rate Layer 2: qo2
Water Rate Layer 1: qw1
Water Rate Layer 2: qw2
Water Injection Rate: qwi
23

MPC minimizes future prediction error by satisfying input constraints

Model Predictive Control Response

24

Continuous self-learning optimization decision engine

New Self-learning Reservoir Management Technique

Physical System

pk +1 , qk +1

Set point

Reservoir
Reservoir
Simulator
Simulator

sp
wf

qGsp

Measure

Control
Control
Implementation
Implementation

qo ,opt

Control

Interpret

Model
Model
Identification
Identification

Optimize
Set point

Reservoir
Reservoir Value
Value
Optimization
Optimization

y k + n + j|k =
Model

n+ N

hu

i = n +1

i k + n + j i

+ dk |k

LP Optimization Loop
N

max NPV = f ( qo , qw , qg ,$, T )

qo ,qw ,qg 1

pmin pk + p , k pmax
s.t.
qmin qk + p qmax

{qo ,opt , q g ,opt , qw,opt

25

Upper optimization layer passes the best operating point to lower layer

Multilayer Reservoir Control Model

Linear
Linear Programming
Programming
Optimizer
Optimizer

Net
Net Present
Present Value
Value
Function
Function

Reservoir
Reservoir Forecasts
Forecasts

Optimization Layer
Regulatory Layer

Information
PLS
PLS Impulse
Impulse
Identification
Identification
Empirical
Empirical
Model
Model

qo , sp
qw, sp
qg , sp

qo

+
-

MPC
MPC
Controller
Controller

pwf

Reservoir
Reservoir
(Simulator)
(Simulator)

d
+

qo
qw
qg

26

Best operating point (LP) problem subject to well constraints

Linear Optimization Problem

N

max NPV = f ( qo , qw , qg ,$, T )

qo ,qw ,qg 1

NPV =

k
k
k
k
( qok Po + qgk Pg qwp
(1 r k )

C
q
C
T
I
C
)
wp
wi
wi
k
T
F

(1 + i )

Down hole Flowing Pressure , psia

k =1

k Tk
365

VLP 4: fw,max +pTHP,min

Reservoir

Performan
c

pthp

e
4

pwf,max

VLP 3: fw,max +pTHP,max

VLP 2: fw,min +pTHP,min

fw

3
2

VLP 1: fw,min +pTHP,max

pwf,min
2

ql,min

ql,max

Total Liquid Rate, BPD

27

The self-learning cased permitted less water and more oil produced

Injector-producer Management Problem Results

Experimental Base:

History-matched Model from El Furrial, HPHT, deep onshore, light oil

Base Case No control

Early water irruption reduced
High water cut reduced wells vertical lift
Further recovery possible at a greater cost

Self Learning Case

Water irruption detected and controlled
Zone shut off permitted better wells vertical lift
Recovery accelerated at a minimum cost

28

Clear benefits from extra little oil but with a lot less effort.

Field-wide life cycle comparison Results

Oil Cumulative

Oil rate

5%

Self-Learning

Np=500 Mbbls
Rev=$5 Million

Non-Controlled

Self-Learning

Non-Controlled

Water rate
Wp, Produced Water Cumulative
Non-Controlled

Wp= -18 MMbbls

Wi= -19 MMbbls
Rev= -$92.5 Million

Self-Learning

Winj, Injected Water Cumulative

Winj Non co

ntrolled

-55%
led
trol
con
n
o
N
Wp
Winj Controlled

-78%

Wp Controlled

29

Continuous self-learning optimization decision engine

New Self-learning Reservoir Management Technique

Physical System

pk +1 , qk +1

Set point

Reservoir
Reservoir
Simulator
Simulator

sp
wf

qGsp

Measure

Control
Control
Implementation
Implementation

qo ,opt

Control

Interpret

Model
Model
Identification
Identification

Optimize
Set point

Reservoir
Reservoir Value
Value
Optimization
Optimization

y k + n + j|k =
Model

n+ N

hu

i = n +1

i k + n + j i

+ dk |k

QP Optimization Loop

LP Optimization Loop

LS Optimization Loop

m
p
2

min ( y k + j y SP ) + R uk2+ j
u
j =1
j =1

s.t.

max NPV = f ( qo , qw , qg ,$, T )

1
qo ,qw ,qg

pmin pk + p , k pmax
s.t.
qmin qk + p qmax

= X + e
Y

ymin y k + j|k ymax ; j = [1, p ]

umin uk + j|k umax ; j = [1, m]
uk +i|k = uk + m 1|k ; i = [ m, p ]

{qo ,opt , q g ,opt , q w,opt

-1
2
min ei ( XT X ) XY
a ,b
i =1
qo , g , w = f1 p k , p k 1 ... qTk , qTk 1 ,...

k
k 1
pres = f n p k , p k 1 ... qT , qT ,...

30

Summary & Conclusions

Novel multilevel self adaptive reservoir performance
optimization architecture
Upper level calculates the optimum operating point
Based on NPV
Optimum set point passed to underlying level

Feasibility of the method demonstrated through a

case study
Reservoir performance continuously optimized by an
adaptive self-learning decision engine
Method capitalizes on available remotely actuated devices

Algorithm feasible for downhole implementation

Impart intelligent to downhole and surface actuation devices
31

Acknowledgement
Research work was done under the guidance of Dr.
Michael J. Economides and Dr. Michael Nikolaou at
the University of Houston
Research partially funded by PDVSA and Cullen
College of Engineering Research Foundation at the
University of Houston
Academic access to software technology: EPS,
Stonebond Technologies, KBRs Advanced Process
Control framework.

32