Você está na página 1de 25

yyyy

;;;;
yyyy
;;;;
yyyy
;;;;
yyyy
;;;;

Model Solutions to Examination

O
N
t
N
O
no
TI
o
E: RA
td
M
T
bu ion
A
n
N GIS SE:
at
io
in
ct
RE UR
E:
se
am
:
R
s
ex
d
hi
CO AR TU
e t l the ishe
E
A
et
N
pl unti
fin
G
m
is
l
Co sea

SI

.:

8 Pages

Date:
Subject:

Reservoir Simulation
INSTRUCTIONS TO CANDIDATES
No.

Mk.

1. Complete the sections above but do not seal until the examination is finished.
2. Insert in box on right the numbers of the questions attempted.
3. Start each question on a new page.
4. Rough working should be confined to left hand pages.
5. This book must be handed in entire with the top corner sealed.
6. Additional books must bear the name of the candidate, be sealed and be affixed to
the first book by means of a tag provided

PLEASE READ EXAMINATION REGULATIONS ON BACK COVER

Answer Notes
#

=>

indicates one of several possible answers which are equally acceptable.

[] =>

extra information good but not essential for full marks - may get bonus.

Model Solutions to Examination

Q1

(i)

1. To perform broad scooping calculations which examine different


development options e.g. waterflooding, gas flooding etc.

2. To extend initial material balance calculations by examining


some other spatial factor such as well-placement or aquifer effect.

(ii)
#

1. To assess additional field management options such as infill drilling,


pressure blowdown etc.

2. To take the improved history match model which can be developed


after same development twice and to use this to assess various IOR
strategies e.g. gas injection, WAG or chemical flooding.

Q2
(i)
#

Because of its inherent simplicity you would virtually always apply


single material balance to assess your field performance - to see if DP
decline tallies with estimated field size, sources of influx and
production.

(ii)
Would be used when a more complex development strategy requiring
spatial information is essential e.g. well placement, assessment of shale
effects, gravity segregation etc.

Q3
(#)

Because it is the only tool we have to tackle complex reservoir


development/flow problems which extends material balance. Clearly, it
is much better than simple material balance alone.

Q4
(i)
The shale continuity strongly affects the hi/lo permeability layer
vertical communication (both pressure and fluid flow). Thus, it will
affect the effective kv/kh (lower or zero for continuous shales) and
will strongly influence gravity slumping of water in a waterflood. In
the situation above with high k on top, some vertical communication will
help recovery.

(ii)
Set up a simple 2D cross sectional model with , say, 50 blocks in the x direction and 10 vertical grid blocks - 5 in each layer. Run waterflood
cases with and without shales - and some in-between cases with
transmissibility modifiers set beween Tz = 0.0 0.01 0.1 0.5 1.0.
Compare water saturation fronts and recoveries as fraction of pv
water throughput. Result will allow us to assess the effects of the
shales in the waterflood.

(iii)
Different

Model Solutions to Examination

The high perm massive sand would have a small scale kv/kh ~1 which
would result in a similar larger scale value. In the laminated sands, the
small scale (say core plug scale) would have a low kv/kh of say 0.1 to
0.01 and this would result in a correspondingly lower kv/kh at the grid
block scale.

(iv)

Well

Gas

Oil
and
Water

Gas

Oil
and
Water

= perforations
Gas Coning It is the drawdown of the highly mobile (low mg) gas into
the perforations. Pattern is shown here in figure. Causes high GOR
production at a level well above the solution gas value.

Set up a near-well r/z geometry fine grid - possibly 50 layers and


set reservoir near-well rock properties e.g. Layering, Tz modifiers,
Rel. perms. etc.

Perform simulations to look at issues such as effect of rate, vertical


communication, gas/oil/water Rel. perms. etc
Generally needs a fine Dr, Dz grid, often finer near the well where
most rapid changes of Sg and pressure with time occur.

(v)
1. The geometry would be different: r/z for coning and cartesian or
corner point for full field.

2. The fineness of the grid would be different. Very fine for nearwell; much coarser for full field.

(Dimensionality too 2D vs. 3D)

Model Solutions to Examination

Q5
(i)

(ii)
Numerical dispersion is the artificial spreading of saturation fronts
due to the numerical grid block structure in the simulation. It arises
because we take large grids to represent moving fronts. It can be
improved by refining the grid (globally or locally) or by using improved
numerical methods.

(iii)
Wells same distance
apart in Figs A and B
L
I

Fluid tends to
flow along (parallel)
to the grids

Fig A

Fig B
I = injector ;

P = producer

(iv)
The injected fluid tends to flow parallel with the grid from the
injector (I) to the producer (P) - see previous page. This means that
early breakthrough and poorer recoveries are seen in A then in B
above. i.e.
Fig B
Actual Recovery
Fig A
%00IP
Producer

Pv or Time

Q6
(i)

Model Solutions to Examination

(ii)
2 P
In this scheme the spatial term in Eq. 1 i.e.
would be specified at
x 2
the new time level n+1

A set of linear equations is the following type of equation (e.g.3x3system)

a11 X1 + a12 X2 + a13 X3 = b1


a21 X1 + a22 X2 + a23 X3 = b2
a31 X1 + a32 X2 + a33 X3 = b3

where X1, X2, X3 are unknown - the as are a matrix of known


coefficients and bs are a known right-hand side.

A direct solution method (e.g. Gaussian Elimination


Elimination) is an algorithm
with a fixed number of steps which will solve these linear equations
(under certain conditions). [Typically forward elimination is applied to
get an upper triangular A* matrix and back substitution is then easily
applied to get the X solution]

In contrast, an iterative technique starts with a first estimate of the


unknown vector X (0) where the (o) denotes 0th iteration: This is then
improved by some algorithm to a better and better solution of the
original linear equations i.e. X(1) X(2) X(V) until the method
converges e.g.

/ X(V+1) - X(V)/ < small number TOL. [Methods such as the Jacobi, LSOR,
etc. are examples of this].
Q7
In block (i, j), then material balance
can be applied for each phase (e.g.
oil and water) for 2-phase flow.
mo
mw

mo
i, j

Mass
Accumulation of

Amount that

Amount that

oil over time

flows in over

flows out over

Dt

Dt

Dt

But amount that flows in/out is given by the pressure differences


between blocks i.e.

A.k.kro ( So )
Qoil ( i -1, j ) ( i , j )
Po - Poi -1 j

mo

i - 1 ij
2

Thus the two phase Darcy Law supplies the relation for volumetric flow
rate and pressure in the grid block. These volumetric flows can be
converted to MASS flows (x by density) and then put into the material

10

Model Solutions to Examination

balance equation to obtain a conservation equation and in pressure


equation for oil and water.

\ Material Balance + Darcys Law => 2-phase Flow Equation.

Q8
(i)

(ii)

(a)

Vo = Dx Dy Dz f So

(b)

11

(iii)

Q9

1. The black oil model essentially treats a phase (o,w,g) as the basic
conserved unit or pseudo component

2. Compositional models are based more correctly on the conservation


of components (CH4, C10, H2O etc.) - the black oil model simply treats
gas dissolution in oil through Rso - gas solubility

3. The compositional models incorporate a full PVT description of the


oil whereas the black oil model relies on the simple Rso type treatment.

12

Model Solutions to Examination

(iii)

(iv)
#

(a) Waterflood calculations in a low GOR - say 30 API - oil reservoir


with pressure maintenance.

(b) CO2 injection in a - say 36 API - light oil system [Condensate


system - gas recycling etc]

Q10
(i)
Upscaling in a waterflood essentially means getting the correct
(effective) parameters (-e.g. rel. perm.) for the larger scale grid blocks
which will reproduce a correct fine grid model.

(ii)
Rock relative permeabilities are meant to be the intrinsic
representative properties of a representative piece of reservoir rock
at the small (i.e. core plug) scale.

Pseudo rel. perms. are effective properties at the larger (usually


gridblock) scale which incorporate other effects and artefacts (e.g.

13

numerical dispersion, heterogeneity etc..) in addition to the intrinsic


rock rel. perms.

(iii)
Methodology
This is a geologically consistent approach to the task of upscaling. i.e.
data collection, sedimentological framework,

[The function of the methodology is to get the geologically + fluid]


mechanically right answer.

Techniques?
These refer to the actual mathematical algorithm to go from a fine
grid coarse grid. E.g. Kyte and Berry, Stones method, two phase
tensors etc

[N.B. This just needs to reproduce the fine grid result - even if it is
WRONG - at the coarse scale]

14

Model Solutions to Examination

Q11

(i)

(ii)
It takes a lag distance of about the range to see the field variability
(standard dev. - i.e. ~ 100mD) of the field.

15

(iii)

lag

Q12
(i)

(ii)
The effective permeability is clearly the harmonic (thickness weighted) average as follows:

16

Model Solutions to Examination

(iii)
The keff in the randomised model would be between the two answers
in (i) and (ii) above (the answer in (ii) being the lower).

e.g. Strictly in a randomised distribution of permeability the average


value tends to the geometric average (kg) in 2D
kg - is less than the arithmetic (along layer) answer.
kg - is greater than the harmonic (across layer) answer.

Q13
(i)
Note - we take the same contour values (c= 0.1, 0.5, 0.8) in all sketches
below.

17

(ii)

(iii)

(iv)

18

Model Solutions to Examination

Q14 With no maths

Consider only ACTUAL flows of Qw and Qo across this interface


interface only. Thus, if the fine grid water (say) flows have not
reached the coarse grid interface

then Qw = 0. => set krw= 0

When there is oil or water flow e.g. water flow.

19

Q15
(i)
Lamina The simplest unit within which we can assume (almost)
homogeneous k. (length mm m)
Lamina set A collection of the above (cm m) e.g. core.

hi k
lo k

Bedform How lamina sets are joined together geometrically to


form 3D beds.

20

Model Solutions to Examination

Eroded/? top sets

e.g.

Tabular cross-bedding

2m

~50cm

Bottom sets

or climbing ripples

Para-sequence/sequence-stacks of bedforms

(ii)

Para sequence - sequence scale

At parasequence - sequence - also bed form influence

Para sequence - bed form

Lamina set - bed form

Q16
(a) There is a double peak - the bimodality probably arises from the
lower perm plugs from deltaic sands, and the higher permeability plugs
from the fluvial channel.

21

(b)

laminated sand
pseudo

Tightly laminated
deltaic sands
A
Crossbeddes
fluvial
channel
-stacked crossbeds

2 Scale pseudo-isation - inclined cross bed pseudo A


- bedform pseudo B

Q17
(i)

(a)

hi

lo

hi

lo

hi

lo

Slow Flow
CAPILLARY DOMINATED

Water flow direction

High water Sw in
LOW perms in a
water-wet system

Sw

HIGH "remaining"
oil in hi k
Spontaneous water inhibition into the LOW k laminae occurs in
Pc-dominated flow. This traps oil in the HIGH k laminae behind the
front where it is well above "residual" but it can't move because the
Rel. Perm. to oil in the low k water-filled laminae is so low.

22

Model Solutions to Examination

(b)

hi

lo

hi

lo

hi

lo

VISCOUS DOMINATED
WATERFRONT

"Fast" Flow
of water

Water flow direction

High water Sw
LOW perm in a
water-cut system

Sw

Note at Viscous dominated conditions a water front goes through


which reduces the oil in all layers to its local "residual" level.
Recovery of oil is better in this case since it is not "stranded" by
downstream capillary imbibition.

(ii)
It is higher in case (a) for the reasons already explained.
[I give a slightly over-detailed answer to part (a) and (b) above].

(iii)
The central implications are twofold.

(a) The two phase pseudo relative perms. are highly anisotropic for
such laminar systems. Along and across layer water displacement in
laminar system gives widely different pseudos.

23

Along

Across

(b) The levels of remaining oil can be vastly different in laminar


systems which, in simulation/upscaling, moves the pseudo rel. perm. end
points. (see above).

24

Model Solutions to Examination

25

Você também pode gostar