Journal of Petroleum Science and Engineering 192 (2020) 107227

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: http://www.elsevier.com/locate/petrol

A fast method of waterflooding performance forecast for large-scale thick

carbonate reservoirs
Yong Li a, *, Qi Zhang a, Daigang Wang b, **, Benbiao Song a, Pengcheng Liu c
a
Research Institute of Petroleum Exploration and Development, PetroChina, Beijing, 100083, China
b
Beijing International Center for Gas Hydrate, Peking University, Beijing, 100871, China
c
School of Energy Resources, China University of Geosciences, Beijing, 100083, China

A R T I C L E I N F O A B S T R A C T

Keywords: Due to the severe heterogeneity of carbonate reservoirs, the traditional reservoir engineering methods for
Carbonate reservoir waterflooding performance forecast, mainly including the material balance equation, waterflooding character­
Waterflooding istic curve and production decline curve, suffer from less reliability. The fine-scale reservoir numerical simula­
Performance forecast
tion is extremely time-consuming, because petrophysical properties in carbonate geological models change so
Stacking pattern
Type curve
drastically. To predict the development performance rapidly, a new method for large-scale thick carbonate
reservoirs is proposed. By combining the geological characteristics and dynamic performance of each well,
vertical stacking pattern of the reservoir are classified, and also good insights into the lateral distribution of
typical stacking patterns are also gained. The stacking pattern distributions are further projected onto the
waterflooding well pattern, followed by rapid evaluation of production and injection rate of every block. Based
on the simulation results for typical stacking pattern blocks, type curves of water flooding on different stacking
pattern reservoirs are workout. Then the waterflooding performance for the whole carbonate reservoir is step-by-
step quickly predicted at a constant liquid rate so as to keep the reservoir injection-production balance with
voidage replacement ratio 1. Finally, a large-scale thick porous carbonate reservoir in the Middle East is taken as
an example to show the workflow of the proposed method. The accuracy is further validated by comparing the
predicted results from numerical simulation, indicating that the proposed method can save a lot of calculation
time while ensuring the accuracy of performance forecast.

1. Introduction distributions as well as strong heterogeneity. The severe contradiction

between carbonate reservoirs of good and poor petrophysical property
According to statistics of C&C Reservoirs in 2014, the global oil and decreases the efficiency of water flooding and makes the production
gas resource in carbonate reservoirs is abundant, with 1.8-trillion-barrel prediction further difficult.
recoverable reserves taking 48% of the global recoverable reserves. The Reservoirs are the result of different sedimentary, diagenetic, and
oil and gas production from carbonate reservoirs all over the world tectonic processes. that control the reservoir heterogeneity, thus control
accounts for 60% of the total global hydrocarbon production. 80% of the petrophysical properties, such as porosity and permeability. Het­
pay zones in the Middle East are carbonate rock, which contributes two- erogeneity is mainly observed in lithology, porosity and electricity. The
thirds of the total global production. In recent years, international oil petrophysical heterogeneity can be hierarchically ordered according to
companies have achieved rapid expansion in the oil and gas recovery deposits and petrophysical properties (Xi et al., 2017; El-Deek et al.,
business from Middle East carbonate reservoirs, which brings a signifi­ 2017; Sharawy and Nabawy, 2015). Stacking pattern is more studied in
cant growth in hydrocarbon reserves and production (Song and Li, deposition facies and stratigraphy sequence. The heterogeneity that
2018). However, the carbonate reservoirs in the Middle East are mainly zones with varying porosity and permeability are distributed in different
controlled by tectonic, sedimentary and diagenetic environments, usu­ parts of reservoir reflects the stacking pattern of formation cycles
ally leading to multiple rock types, complicated storage space (Schauer and Aigner, 1997; Pomoni-Papaioannou and Kostopoulou,

* Corresponding author.
** Corresponding author.
E-mail addresses: liyongph@petrochina.com.cn (Y. Li), dgwang@pku.edu.cn (D. Wang).

https://doi.org/10.1016/j.petrol.2020.107227
Received 4 October 2019; Received in revised form 30 November 2019; Accepted 24 March 2020
Available online 2 April 2020
0920-4105/© 2020 Elsevier B.V. All rights reserved.
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 1. Workflow for waterflooding performance forecast in large-scale carbonate reservoir.

2008; Zhu et al., 2013). difficult to fully capture the reservoir dynamics. In general, the tradi­
At present, the commonly used performance forecast methods can be tional reservoir engineering methods adopt empirical equations to pre­
categorized into two types: reservoir numerical simulation and reservoir dict the change trend of reservoir production by analyzing the available
engineering analysis (Wang et al., 2016). For reservoir simulation of production and pressure data, with only few geological factors of the
large-scale ultra-thick carbonate reservoirs, tremendous number of grids reservoir considered. It is also impossible to capture the dynamic change
are needed in order to accurately characterize the lateral and vertical in actual production of single well, block and layer with the traditional
reservoir heterogeneity, usually leading to a slow speed of fine-scale reservoir engineering methods if the development strategies are
reservoir numerical simulation which is impossible to guide the opti­ adjusted. However, the fine-scale reservoir numerical simulation based
mization and adjustment of development scheme in a real-time manner on the understanding of geological characteristics can provide a higher
(Wang et al., 2017a,b; Tueros et al., 2018). The reservoir engineering prediction accuracy at a smaller scale, because a variety of dynamic and
method mainly includes the material balance equation, waterflooding static influential parameters can be taken into consideration. Percola­
characteristic curve, production decline curve analysis and specific tion approach method is well adapted to estimate field scale flow and
prediction models. The material balance equation (Pletcher, 2002; Yil­ large scale properties and their uncertainty (Percolation Theory in
diz and Khosravi, 2007; Orozco et al, 2018) is a zero-dimension equation Reservoir Engineering, 2018).
and ignores the effect of reservoir heterogeneity, which assumes a uni­ In order to improve the accuracy of performance forecast in large-
form distribution of reservoir pressure. It is often used to forecast the scale carbonate reservoirs while reducing the calculation time as much
whole reservoir production performance in oil and gas field develop­ as possible, a quick method for waterflooding performance forecast,
ment, especially in gas reservoirs. The waterflooding characteristic based on the classification of reservoir stacking pattern, is developed by
curve (Han and Zhao, 2015; Tillero and Mogollon, 2017; Feng et al., combining the advantages of the traditional reservoir engineering
2017; Wang et al., 2017a,b; Dou et al., 2019) is adapted to the case of method and fine-scale reservoir numerical simulation. Fine-scale nu­
waterflooding or natural water drive when water cut is above 50%–60%. merical simulation and waterflooding characteristic type curves of
The semi-logarithmic relationship between cumulative water produc­ typical stacking pattern blocks are only required, and then step-by-step
tion and cumulative oil production should be modified at the early performance prediction of every block can be achieved using the nodal
stages of water breakthrough. The production decline analysis (Fanchi analysis of each producer and injector. Finally, case study is performed
et al., 2013; Gong et al., 2014; Percolation Theory, 2018; Tian et al., to show the accuracy of the proposed workflow.
2018) concentrates on the performance prediction in the mid-late
development stages. The Arps decline curve model is commonly used 2. Fast waterflooding performance forecast methodology
to predict the production data in the early stage of field development,
inducing severe system errors when water cut is higher than 90%. The To overcome the limitations of reservoir numerical simulation
other commonly used prediction models (Suleymanov et al., 2012; Table 1, including non-convergence or long-time calculation caused by
Chakra et al., 2013; Li et al., 2018; Mohammadi and Ameli, 2019) severe heterogeneity of fine-scale geological model, a new quick per­
include the time series analysis, gray prediction, neural network, and formance forecast workflow is developed by combining the advantages
functional simulation, with successful applications in other areas of of traditional waterflooding characteristic type curve and reservoir nu­
economy, control and artificial intelligence. However, those models are merical simulation, as shown in Fig. 1.
essentially specific mathematical models, where variables are found too The reservoir stacking pattern of wells is firstly identified according

2
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

to well testing, logging, coring and dynamic data. The lateral distribu­ stacking pattern classification of each well and group are further
tion of reservoir stacking patterns is determined according to the prior modified using production data. When the stacking patterns of each well
knowledge on the well stacking patterns. The whole reservoir is further are achieved, the lateral distribution of vertical stacking patterns across
divided into some stacking pattern blocks whose boundaries are the reservoir is gained. Each stacking pattern would be assigned a value,
consistent with the waterflooding well pattern. Typical sector model for so every well point would have a value standing for its stacking pattern.
each typical vertical stacking pattern is selected from the whole reser­ Areal interpolation can be done if there are enough well point identified,
voir. The waterflooding characteristic type curves for specified typical leading to the lateral distribution of stacking patterns in a carbonate
stacking pattern blocks are obtained by fine-scale reservoir numerical reservoir. For a specified well pattern, the adjacent wells with the similar
simulation of the selected typical sector models. As a matter of fact, a stacking pattern are grouped into the same block. There is at least one
detailed analysis of various static and dynamic data is essential to injector-producer well group for each block, and the boundary coincides
classification of typical stacking patterns. Reasonable stacking pattern with that of well group in order to calculate the reasonable injection and
classification and well pattern block division are critical to reliable liquid production rate of the blocks under the condition of volume
performance prediction. balance.
Based on nodal analysis for typical producers and injectors, the
correlation between the liquid production or injection rate and the
2.2. Waterflooding characteristic type curve
formation factor Kh can be determined. The Kh values at well points are
gained from the fine-scale geological model or well log interpretation to
During waterflooding, the reservoir is developed by rigid water drive
calculate the reasonable liquid production rate or water injection rate.
when the change of formation temperature as well as capillary pressure
The production rate of all producers and injection rate of all injectors
and gravity can be ignored. The ratio of water-phase and oil-phase
across a specified stacking pattern block are summed up to determine
relative permeability can be described as (Wang et al., 2013):
the liquid production rate and water injection rate of the block. The
lower value between liquid production rate and water injection rate of Krw
¼ d ​ ecSw
the block under reservoir condition is assigned as the reasonable for­ Kro
mation liquid production rate of this block. Based on Darcy’s law, the correlation between water-oil ratio and
For a specified well pattern, the waterflooding characteristic type water cut is stated as follows:
curves (i.e., water cut versus oil recovery efficiency) obtained from the
simulation results of typical stacking pattern blocks are regarded as the Qw μo Bo Krw
¼ ¼
fw
type curve for all blocks with the same stacking pattern. Before water Qo μw Bw Kro 1 f w
breakthrough, no water is produced, and oil production rate of a block The material balance equation for rigid water drive can be expressed
equals to the actual liquid production rate. After water breakthrough, as:
the water production rate is defined as the product between the actual
liquid production rate and the corresponding water cut in the type Sw ¼
1 S
NP þ SWC ¼ ð1 Swc ÞR þ SWC
curve, while the oil production rate equals to the difference between N
liquid production rate and water production rate. The whole-reservoir By considering the above three equations, we have:
performance prediction for liquid production rate, water cut, and oil � �
recovery factor, is achieved when we sum up oil and water production lg
fw
¼ lg
Qw μ Bo
¼ lg o d þ clg½ð1 Swc ÞR þ Swc �
rate of all stacking pattern blocks at the same time step. 1 fw Qo μw B w

where Qw and Qo are the surface oil and water production rates,
2.1. Stacking pattern classification
respectively; μo and μw are the viscosity of oil-phase and water-phase,
respectively; Bo and Bw are the formation volume factors of oil phase
Due to the lateral and vertical heterogeneity in carbonate reservoir,
and water phase, respectively; Kro and Krw are the relative permeability
petrophysical properties suffer from drastic changes. The existence of
of oil phase and water phase, respectively; Swc is the irreducible water
different vertical stacking patterns usually causes the interlayer inter­
saturation; Sw is the water saturation; fw is the water cut; R is the oil
ference and the varying waterflooding performance. Therefore, classi­
recovery efficiency; d and c are constants which are affected by rock and
fying typical vertical stacking patterns based on the vertical distribution
fluid properties of the waterflooding reservoir.
of different types of reservoir is of great importance to dynamic sur­
It can be seen that, except for water cut and the oil recovery factor,
veillance and development optimization. Moreover, the lateral distri­
all the other parameters are known. In cases of the blocks under the same
bution of typical stacking patterns is mainly controlled by sedimentary,
stacking pattern, it is reasonable to assume that there is a uniform
structural and diagenetic environments. The whole reservoir is further
function between the water cut and the oil recovery factor, because the
categorized into some small well-pattern blocks so that the vertical
rock and fluid properties are similar in those blocks:
stacking pattern within each block is similar as much as possible. It is
favorable to understand the inherent geological characteristics distri­ f w ¼ yðRÞ
bution and development law in strongly heterogeneous carbonate res­
Typical sectors with representative geological characteristic of each
ervoirs, and gain powerful insights into the production performance and
stacking pattern are firstly identified and selected. Their corresponding
dynamic surveillance.
fine-scale reservoir numerical simulations are performed under the
Firstly, the reservoir type is categorized by investigating the rock and
specified waterflooding well pattern, and then the typical curve between
fluid properties with various geological data from well testing, logging
oil recovery efficiency and water cut is obtained. For this step, it is very
and coring. The classification to rock types is the base of stacking pattern
fast compare to the whole field reservoir simulation because of small
identification. Reservoir rocks are controlled by sedimentary facies,
sector model reservoir simulation. By type curve fitting, the statistical
diagenesis and structure, showing different spacing and flowing char­
formulas are regarded as the typical waterflooding characteristic curves
acteristics. For carbonate reservoirs, wells located in different positions
of each stacking pattern to carry on performance forecast.
performs differently in waterflooding, caused by various vertical dis­
tribution of each type of rock. The vertical assembly and combination of
different reservoir rocks are summarized into several stacking patterns 2.3. Injectivity and productivity evaluation
to describe the vertical heterogeneity in terms of the thick percentage of
each type of rocks and the way of their blocking or layering. The For a producer located in the center of a circular formation with

3
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 2. Inflow and outflow performance relationship curves.

constant pressures at the supply boundary (Li and Chen, 2012), the To keep the volume balance in each stacking pattern block, the lower
production rate can be calculated as below: value between the reasonable injection and production rate at reservoir
� condition is used as the actual liquid production rate.
2πKro h pr pwf � �
Qo ¼ � �a
Qprod sec Qinj sec
μo Bo ln rrwe 12 þ s Ql sec ¼ Min ;
Bo Bw
�
2πKrw h pr pwf where Qi prod and Qi inj are the surface liquid production and injection
Qw ¼ � �a rates for a specific well, respectively; Qprod sec and Qinj sec are the
μw Bw ln rrwe 12 þ s proportionally-summated surface liquid production and injection rates
of wells across a block, respectively; Q1 sec is the reasonable surface
where h is the pay zone thickness; re and rw are the drainage radius and injection/production rate per block; Bo and Bw are the formation volume
the wellbore radius, respectively; pwf and pr are the flowing bottomhole factors of oil phase and water phase, respectively; a is the liquid
pressure and the average reservoir pressure, respectively; S is the skin contribution proportion, which is 1/4 for an edge well and 1/2 for a
factor; a is the unit conversion coefficient. corner well in an inverted nine-spot well pattern; n is the total number of
The shape of the inflow performance relationship (IPR) curve de­ injectors or producers across the block.
pends on the drive mechanism. If the drive mechanism is the same, e.g.,
rigid water drive, the IPR curve is influenced by the reservoir pressure,
2.4. Time-step predictive calculation
pay zone thickness, permeability, fluid property and so on. Therefore,
with no changes in the reservoir pressure and fluid property, the IPR
For a specific stacking pattern block, the production performance
curve is mostly affected by the formation factor (Kh). Higher Kh results
prediction can be achieved on the basis of typical waterflooding char­
in the higher productivity, and the IPR curve climbs up more steeply and
acteristic curves for the corresponding stacking pattern as well as the
extends further away, as is shown in Fig. 2 with subscripts of Kh from 1
OOIP reserve and reasonable liquid production of the block.
to 5 representing Kh values from high to low. With the wellbore sketch
known, the intersection of the VLP (vertical lift performance) curve of
1 Set a time step Δt, for example, one month. The smaller the time step
the multi-phase wellbore flow and the IPR curve refers to the reasonable
is, the more the number of time steps, and more accurate the pre­
surface liquid production rate of a producer or surface water injection
diction result is. When the time step is relatively long, iterative
rate of an injector at a specified value of Kh.
calculation is required to meet the prediction accuracy. By
By changing Kh values, the reasonable water injection rate and liquid
substituting the oil recovery efficiency and water cut at the (k-1)th
production rate are computed using the inflow and outflow performance
time step, the indicators at the kth time step are predicted. If the
relationship analysis (Clarkson, 2013). The curve between well pro­
relative error is small enough, the calculation will continue to the
ductivity/injectivity and formation factor Kh is further fitted, which
(kþ1)th time step. Otherwise, the predicted indicators at the kth time
takes the form of:
step should be used for another iteration until the constraint for
Qi ¼ fi ðKhÞ relative error is satisfied.
2 If water production rate is zero, the liquid production rate equals to
where Qi indicates the liquid production or injection rate of a well; Kh is the oil production rate. The ratio of cumulative oil production to the
the formation factor. OOIP reserve of the block is treated as the oil recovery efficiency. If
The Kh value for each well point is obtained from the geological the predicted recovery efficiency is lower than that of water break­
model, and the injection/production rate Qi is calculated by the above through time in typical characteristic curve (the cross-sectional
function, which is then summed up to determine the reasonable pro­ point), the oil production rate at next time step is the same and
duction and injection rates of each stacking pattern block. The liquid water production rate remains zero, followed by the oil recovery
production rate of edge and corner wells are summed up in appropriate efficiency updated.
proportions allocated by well pattern.
P
k
Ql
X
n k
sec
Δt
Nprod Bo
Qprod sec ¼ a Qi prod Rksec ¼
sec
¼ i¼1
i¼1 Nsec Nsec

X
n
f kw ¼0
sec
Qinj sec ¼ a Qi inj
i¼1

4
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Table 1
Positive and weak points of reservoir performance forecast methods.
Advantages Disadvantages Computation time

Reservoir numerical ① heterogeneity is taken into consideration ① large amount of grids and severely variable properties Long several hours even
simulation ② Prediction can be applied in all lifetime of a reservoir at make computation time long. tens of hours
scales from single wells, single layers to block, field.
Reservoir ① Many empirical formulas and mathematical models are ① variable property are regarded as homogeneous. Short a few minutes
engineering convenient. ② Most methods are only adaptable to the certain
analysis ② Prediction results are more accurate at field scale. development stages of a typical carbonate reservoir.

3 As the oil recovery efficiency reaches the breakthrough point, the

water cut is computed using the fitting relationship of the typical Table 2
Classification of types of Mishrif reservoir with petrophyscial properties.
waterflooding characteristic curve for specific stacking patterns.
Once the reasonable liquid production rate is given, the oil produc­ Reservoir Type I II III Ⅳ
tion rate, water production rate, water injection rate and oil recovery Porosity (%) >27 17–27 11–17 <11
efficiency at specified time step can be estimated. Afterwards, the Permeability (mD) >100 10–100 0.1–10 <0.1
production performance at next time step is determined based on the
result of cumulative oil production at this time step.
recovery efficiency of the reservoir at the kth time step; Δt is the time
P
k � Ql step; n is the total number of stacking pattern blocks across the reservoir.
k 1 f iw 1sec sec
Δt
Nprod sec
Bo
Rksec ¼ ¼ i¼1
Nsec Nsec 3. Case study for fast waterflooding performance forecast
�
f kw sec ¼ y Rksec The Mishrif reservoir in the Middle East is a Cretaceous porous car­
bonate reservoir with depth over 2000 m. The northern and southern
k k
where Nprod sec and Winj sec are the cumulative surface liquid production parts belong to the same long-axis anticline structure and are connected
and water injection of the block at kth time step, respectively; f kw sec is by low saddle. The reservoir is deposited in continental slope of shallow
sea, open sea, reef, reef front and lagoon settings, and goes to coastal
the water cut of the block at kth time step; Rksec is the oil recovery effi­
facies towards the southwest, dominated by rudist reef-beach facies and
ciency of the block at kth time step; Nsec is the OOIP reserve of the block;
inter-reef clastic shallow beach facies. The formation spreads steadily
Δt is the time step.
over the field with an average thickness of 134 m. The thickness of
northern reservoir is greater than that of the south. The two layers, MA
4 By summing up the oil production rate and water production rate for
and MB, are separated by a thin limestone barrier. The MA is relatively
all stacking pattern blocks at the same time step, the production and
tight, while the more-permeable MB is regarded as the main payzone
injection rate across the whole reservoir are obtained. The oil re­
with the presence of high-permeability zones. The sub-layer MB3
covery efficiency and water cut for the whole reservoir are then
located at the bottom of MB has the greatest potential due to the
determined. The step-by-step predictive calculation is continued
favorable distribution of physical properties. When the primary recov­
until the endpoint.
ery rate reaches 5.0%, water injection is started to recover the reservoir
X
n n X
X k
� energy. In general, the Mishrif reservoir mainly suffers from medium
Npk ¼ k
Nprod sec;i ¼ 1 f jw sec;i Qjl sec;i Δt porosity, low permeability and strong heterogeneity, with micro­
fractures distributed regionally. It can be categorized into four different
i¼1 i¼1 j¼1

types of reservoir, and four typical stacking patterns are further classi­
X
n n X
X k
Wpk ¼ k
Wprod sec;i ¼ f jw j
sec;i Ql sec;i Δt
fied based on various vertical combinations for each type of reservoir.
i¼1 i¼1 j¼1 Pilot practices demonstrate that the vertical stacking patterns have
significant impacts upon the waterflooding development performance.
X
n n X
X k
Wik ¼ k
Winj sec ¼ Qjl sec;i Δt
i¼1 i¼1 j¼1 3.1. Stacking pattern block division

n � �
P
1 f kw Qkl The geological distribution of Mishrif reservoir is controlled by
sequence structure, sedimentary facies and diagenesis. The dominant,
sec;i sec;i
i¼1
f kw ¼ P regionally-distributed granular limestone of the biological debris beach
n
Qkl sec;i
i¼1 facies is mainly found at the middle and bottom of sub-layer MB, while
the remaining layers are either relatively tight or occasionally seen with
Npk a few thin zones with high porosity and permeability. Guided by the core
Rk ¼ P
analysis data, the type of Mishrif reservoir is further classified into four
n
Nsec;i
i¼1 classes in terms of porosity and permeability, as displayed in Table 2.

where Nkprod sec;i ,

k
Wprod sec;i
k
and Winj sec;i are the surface cumulative oil 3.1.1. Stacking pattern classification and identification
production, water production and water injection of the stacking pattern Based on the vertical combinations of Type I, II, III and IV reservoirs,
block i at the kth time step, respectively; Npk , Wpk and Wik are the overall the formation architecture of the Mishrif reservoir can be grouped into
surface oil production, water production and water injection of the four stacking patterns, named as A, B, C and D, respectively, with pet­
whole reservoir, respectively; Nsec;i is the OOIP reserve of the stacking rophysical properties varying from good to bad, as shown in Fig. 3.
pattern block i; Qkl sec;i is the reasonable liquid rate of the block i at the The stacking pattern A primarily consists of Type I and II reservoirs,
having thick net pay and high permeability layers distributed in block.
kth time step; f kw sec;i and f kw are the water cuts of stacking pattern block i The Type I reservoir is mostly found within the sub-layer MB in the form
and whole reservoir at the kth time step, respectively; Rk is the oil of thick block. Seen from well logs, it has great OOIP, typically

5
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 3. Well section profiles for the four typical stacking patterns in Mishrif reservoir.

represented by high neutron porosity and large true resistivity. The stacking patterns are illustrated in Fig. 4.
formation pressure data from RFT (Repeat Formation Test) change
similarly as the depth increases, indicating a good vertical connectivity. 3.1.2. Distribution and block division of stacking patterns
In stacking pattern B, the proportion of Type I reservoir is lower than Based on the response of well logs, the stacking patterns for 246 wells
that in the stacking pattern A. Type I and II reservoirs are distributed with complete well logging curve in the Mishrif reservoir are identified.
alternatively in thin zones instead of blocks. Reservoirs with good Accordingly, 105 wells are sorted into the stacking pattern A. 65 wells
properties and poor properties are stacked in neighbor. From well log, are sorted into the stacking pattern B. 50 wells are sorted into the
high neutron porosity and large true resistivity are only observed at a stacking pattern C. 26 wells are sorted into the stacking pattern D. Their
few intervals, indicating a relatively great OOIP. The change trend for locations are marked in Fig. 5(a), where the wells for stacking patterns
formation pressure data is inconsistent with that of depth, which shows A, B, C and D are denoted by red, yellow, blue and gray dots, respec­
poor vertical formation connectivity and possible presence of barrier. tively. The majority of wells available to the classification of stacking
The stacking pattern C is mainly composed of Type II and III reser­ pattern is located at the north Mishrif reservoir. There are few wells in
voirs. The Type III reservoirs are dominantly distributed with low the south Mishrif reservoir, none of which are identified as the stacking
porosity and permeability. From well logs, low neutron porosity and low pattern A. The lateral distribution of stacking patterns in the whole
true resistivity are observed, indicating a relatively small OOIP. The Mishrif reservoir can be mapped by interpolation of the 246 identified
formation pressure data fluctuates greatly with the increase of depth, well points. The map shows that the stacking pattern A is mainly
and the vertical formation connectivity is poor. distributed nearby the north Mishrif, while the stacking pattern C is
In stacking pattern D, only extremely thin layers of Type I reservoir dominant at the south. Meanwhile, the distribution of the stacking
may exist. A majority of Type III and IV reservoirs with less net pay and pattern A lies in strip nearby the northern saddle, which is consistent
poorer petrophysical properties is widely found. Regional barrier are with the microfacies distribution.
distributed within Type- II and III reservoirs. Typical well logs of this The existing 900-m well pattern in the Mishrif reservoir is adjusted to
pattern show extremely low neutron porosity and true resistivity, indi­ an inverted nine-spot well pattern with 900-m spacing. Then, the in­
cating a very small OOIP. The formation pressure data usually suffers jectors and producers with the same stacking pattern are grouped into
from a drastic fluctuation as the depth changes. The vertical formation the same block along the boundary of well group based on the lateral
connectivity is extremely poor. Fig. 3 displays the typical well section distribution of different typical stacking patterns in the north Mishrif
profiles of four typical stacking patterns. The well log curves for typical reservoir, as shown in Fig. 5(b). The well stacking pattern occupied

6
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 4. Well logging responses for the four typical stacking patterns.

Fig. 5. Distribution and block division of typical stacking patterns: (a) Well points and interpolation distribution of stacking patterns; (b) Inverted nine-spot well
pattern; (c) Block division of stacking patterns.

7
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 6. Porosity distribution in the selected blocks for typical stacking patterns.

dominantly in a block is treated as the block stacking pattern. The

northern Mishrif reservoir is ultimately divided into 21 stacking pattern
blocks, as shown in Fig. 5(c). There are 5 blocks of Pattern A, 8 blocks of
Pattern B, 6 blocks of Pattern C and 2 blocks of Pattern D.

3.2. Fine-scale sector model simulation for typical stacking pattern

Four well groups, where wells are almost identified as the same
stacking pattern, are cropped from the blocks A2, B2, C3 and D1,
respectively, denoting to the four typical stacking patterns, shown in
Fig. 6. Using the fine-scale geological models, waterflooding numerical
simulation is performed with the 900-m inverted nine-spot well pattern
to study the impact of stacking pattern on the development effect.
The fine scale geological models are parts of the whole field fine
model and selected corresponding to the geological features of the
classified stacking pattern. The typical refined model of stacking pattern
A has mass grids with better porosity and permeability than models of
other stacking patterns, for interpolation from most of wells in the block
model A sorted into stacking pattern A according to their well logs. As a
result, there are good reservoirs stacked in thick block in the model of Fig. 7. Waterflooding characteristic curves for typical stacking patterns in
stacking pattern A. Shown by the top right part of Fig. 6 framed by red inverted nine-spot well pattern.
rectangle, many grids in red and yellow represent the porous thick
zones. The top left part of Fig. 6 framed by yellow rectangle stands for which are favorable for modeling multiphase fluid flow in carbonate
typical model of stacking pattern B. It has the same color table as model reservoir.
A but red grids are seen in thinner layers penetrating the yellow grids. As can be seen from Fig. 7, the oil recovery efficiency gradually
More green grids mean poor porosity and are alternated by gray tight decreases from Patterns A to B, C and D, indicating a worse development
grids. The C type model at the bottom left of Fig. 6 framed by blue effect correspondingly. Due to the favorable distribution of thick block
rectangle shows that most grids in green and gray are distributed across reservoirs of Type I and II, the stacking pattern A has the greatest oil
the whole model in thick zones with red grids spreading regionally and recovery efficiency and the slowest water cut rising rate. The stacking
stratifiedly. Almost all of the grids in model of stacking pattern D in pattern A is regarded as the most preferable stacking pattern. Affected
bottom right part of Fig. 6 framed by the gray rectangle are green and by the decrease in reservoir Type I and large difference in petrophysical
gray on behalf of poor reservoir and tight rock, only few grids are yellow

8
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Table 3
Fitting parameters for waterflooding characteristic curves of typical stacking patterns.
a b c d e f g R2 Horizontal intercept

A 406.23 694.31 463.75 174.26 44.985 4.5713 0.1493 1 0.064

B 9098.7 9163 3504.5 683.27 87.855 5.1838 0.1034 0.9999 0.055
C 58301 68865 31431 6947 735.68 26.502 0.3053 0.9999 0.029
D 249642 184779 52667 7255 508.78 21.841 0.2471 0.9998 0.015

Fig. 8. Reasonable water injection or liquid production rate with Kh.

properties between layers, early water breakthrough through high

Table 4
permeable zones and low sweep efficiency are observed in the stacking
Calculation of reasonable liquid production and water injection rates in typical
pattern B. The few thin permeable zones existed in the stacking pattern C
block A1.
can improve the oil production rate, but it is usually accompanied with
earlier water breakthrough time and rapider rise of water cut. The wide Well Well Kh Surface liquid Ratio of liquid
name type (mD*m) production, m3 contribution
distribution of low permeability layers in stacking pattern D leads to the
worst development effect. R-15 Injector 111166.8 3328.3 1
R-17 Injector 115250.1 3320.5 1
The curve fitting for the waterflooding characteristic curves obtained
R-1 Producer 213633.7 630.9 1
from simulation of typical stacking pattern blocks is performed with the R-14 Producer 67134.51 631.9 1
polynomial formula y ¼ ax6þbx5þcx4þdx3þex2þfx þ g. The four R-16 Producer 38365.33 627.1 1
polynomial formulas are regarded as a tool to predict the development R-2 Producer 71446.94 627.8 1
performance of typical stacking pattern blocks, followed by the quick R-3 Producer 184275 618.9 1
R-4 Producer 73587.37 615.8 1
waterflooding performance forecast for the whole reservoir. The fitting R-18 Producer 68078.48 619.7 1/2
parameters are summarized in Table 3. R-27 Producer 71185.26 630.4 1/2
R-28 Producer 60482.48 634.5 1/2
3.3. Productivity evaluation for typical stacking pattern blocks R-29 Producer 181734.6 630.6 1/2
R-30 Producer 50785.48 590.8 1/2
R-5 Producer 35574.09 592.6 1/2
With regard to the target carbonate reservoir, the formation pressure R-31 Producer 18764.6 615.0 1/4
is 21.4 MPa. The casing depth of well configuration is 2500 m. The
corresponding tubing depth is 2325 m with a tubing diameter of 2.875 in
(73.025 mm) for producers and 3.5 in (88.9 mm) for injectors. The 3345:6674*Kh
wellhead pressure is set at 2.0 MPa for producers and 10.3 MPa for in­ Qi ¼ (2)
1194:8058 þ Kh
jectors. Based on the inflow and outflow performance analysis, the
correlation between the surface flow rate of a producer or injector and where K is the payzone permeability, mD; h is the payzone thickness, m;
the formation factor Kh is determined by assigning different values to Qp and Qi are the initial liquid production rate and water injection rate,
the permeability K and thickness h. All premise parameters are input respectively m3/d.
into the PROSPER module of IPM software while the value of kh is By substituting the Kh values exported from the full-reservoir
varying from 1 to 3 � 105 and every value of Qp and Qi related to a geological model into equations (1) and (2), the reasonable liquid pro­
certain kh are output to draw the curves in Fig. 8. As the formation factor duction rate of each producers and water injection rate of each injector
Kh increases, both the liquid production rate and water injection rate are calculated. The liquid production rate of each block is the sum of
increase gradually until a constant. The correlation curves between liquid rates contributed from all producers belonging to the same block,
liquid rate and the value of Kh for producer and injector are shown in following the rule that the corner well contributes a quarter of its liquid
Fig. 8. rate while the edge well contributes one-half of its liquid rate. The lower
The fitted curves are described as follows: value between the liquid production rate and water injection rate of the
block under the reservoir condition is chosen as the formation liquid
639:2526*Kh
Qp ¼ (1) production rate to keep reservoir volume balance, and the actual
2800:4348 þ Kh

9
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

Fig. 9. Comparison of predicted production performance with those from numerical simulation.

reasonable liquid production rate and water injection rate of the block The daily oil production rate drops from 164,000 m3 to 62,568 m3, while
are then determined. the daily water production rate increases from zero to 101,432 m3.
Table 4 displays the predicted data of block A1 located in upper left As can be seen from Fig. 9, the predicted data are very close to the
corner of Fig. 6. By summing up products of the reasonable surface results obtained from reservoir simulation. Our proposed method en­
liquid production rate and the rate contribution ratio of all producers in ables to consider the vertical and lateral heterogeneity of carbonate
block A1, the surface oil production rate of Block A1 is 5756 m3. When reservoir with stacking pattern classification and stacking pattern block
the oil-phase formation volume factor is set to 1.3, the formation oil division, respectively. The computation time is less than 3 min.
production rate equals to 7540 m3. Similarly, the formation water in­ Compared with the fine-scale reservoir numerical simulation that re­
jection rate of block A1 will be calculated by the product of surface quires tens of hours, the proposed method saves a lot of calculation time
water injection rate and water-phase formation volume factor 1.0, while ensuring the prediction accuracy of production data for each well,
equaling to 6649 m3. The lower value, 6649 m3, is then regarded as the block and layer.
reasonable liquid production rate under the reservoir condition in order
to keep the injection-production volume balance. Ultimately, the surface 4. Conclusions
oil production rate, 5075 m3, is selected as the constant liquid produc­
tion rate of Block A1 at each time step for further performance forecast. (1) By combining the advantages of both the traditional reservoir
engineering method and fine-scale reservoir simulation, a quick
3.4. Performance forecast and validation performance forecast method for large-scale thick carbonate
reservoir is proposed on the basis of stacking pattern classifica­
Taking the northern Mishrif reservoir in the Middle East as an tion of the reservoir. It can be easily used for accurate forecast of
example, the proposed method is used for 20-year waterflooding per­ waterflooding production performance within a short time.
formance forecast when the inverted nine-spot well pattern with well (2) Using the prior knowledge of geological characteristics and fine-
spacing of 900 m is considered. The results denoted by the solid lines are scale geological models, the vertical stacking pattern classifica­
shown in Fig. 9. The ultimate oil recovery efficiency is 10.9% with water tion and lateral stacking pattern block division for the whole
cut of 61.27%. The daily oil production rate drops from 163,872 m3 to reservoir are carried out. Based on the simulation results for
63,581 m3, while the daily water production increases from zero to typical blocks of each stacking pattern, the production data for
100,291 m3. The production data obtained from fine-scale reservoir the whole carbonate reservoir will be step-by-step predicted
simulation is also illustrated in Fig. 9, denoted by the dotted lines. The rapidly at a constant liquid rate under the condition of reservoir
ultimate oil recovery efficiency is 10.6%, with the water cut of 61.84%. volume balance.

10
Y. Li et al. Journal of Petroleum Science and Engineering 192 (2020) 107227

(3) Case study indicates that, compared with the fine-scale reservoir Han, S., Zhao, C., 2015. A new waterflooding characteristic curve for naturally fractured
sandstone reservoirs. Adv. Petrol. Explor. Dev. 9 (1), 9–12.
numerical simulation that requires 3 h and 50 min, the proposed
Li, K., Chen, Z., 2012. Computation of productivity index with capillary pressure
method only calculates for 2 min. The proposed method saves a included and its application in interpreting production data from low-permeability
lot of calculation time while ensuring the prediction accuracy of oil reservoirs. SPE J. 17 (4), 1041–1046.
performance forecast. It enables a rapid and accurate way to Li, Z., Zhang, S., Xu, J., Liu, Y., Li, W., 2018. Improved modeling of channel prediction
based on gray relational analysis and a support vector machine: a case study on the X
predict the production performance of large thick carbonate pilot area in the Daqing oilfield in China. J. Geophys. Eng. 15 (4), 1407–1418.
reservoir with severe heterogeneity. Mohammadi, K., Ameli, F., 2019. Toward mechanistic understanding of fast SAGD
process in naturally fractured heavy oil reservoirs: application of response surface
methodology and genetic algorithm. Fuel 253, 840–856.
CRediT authorship contribution statement Orozco, D., Aguilera, R., Selvan, K., 13-14 March 2018. Material balance forecast of huff-
and-puff gas injection in multiporosity shale oil reservoirs. In: Proceedings of the SPE
Yong Li: Conceptualization, Methodology, Resources, Writing - Canada Unconventional Resources Conference. Calgary, Alberta, Canada. SPE-
189783-MS.
original draft, Project administration, Funding acquisition. Qi Zhang: Percolation Theory in Reservoir Engineering, Published by World Scientific, pp 300, 978-
Methodology, Validation, Investigation, Writing - original draft, Visu­ 1-78634-523-3, 2018).
alization. Daigang Wang: Software, Validation, Formal analysis, Pletcher, J.L., 2002. Improvements to reservoir material balance methods. SPE Reservoir
Eval. Eng. 51 (1), 49–59.
Writing - review & editing. Benbiao Song: Validation. Pengcheng Liu: Pomoni-Papaioannou, F., Kostopoulou, V., 2008. Microfacies and cycle stacking pattern
Data curation, Supervision. in Liassic peritidal carbonate platform strata, Gavrovo-Tripolitza platform,
Peloponnesus, Greece. Facies 54 (3), 417–431.
Schauer, M., Aigner, T., 1997. Cycle stacking pattern, diagenesis and reservoir geology of
Acknowledgements
peritidal dolostones, Trigonodus-Dolomite, Upper Muschelkalk (Middle Triassic, SW-
Germany). Facies 37 (1), 99–113.
The authors appreciate Research Institute of Petroleum Exploration Sharawy, M.S.E., Nabawy, B.S., 2015. Geological and petrophysical characterization of
& Development (RIPED), PetroChina for the permission to publish the the lower senonian matulla formation in southern and central gulf of suez, Egypt.
Arabian J. Sci. Eng. 41 (1), 281–300.
paper and the financial support by National Natural Science Foundation Song, X., Li, Y., 2018. Optimum development options and strategies for water injection
of China (NSFC) (Grant No. 51874346). We also thank all the anony­ development of carbonate reservoirs in the Middle East. Petrol. Explor. Dev. 45 (4),
mous reviewers for their constructive comments which have greatly 679–689.
Suleymanov, A.A., Abbasov, A.A., Malikov, H.K., 2012. Analysis of “reservoir-well”
improve the quality of this paper. system conditions using functional diagnostic methods. J. Petrol. Sci. Eng. 100,
30–36.
Appendix A. Supplementary data Tian, L., Zuo, L., Wang, B., 2018. Methods of decline curve analysis for shale gas
reservoirs. Energies 11 (3). https://doi.org/10.3390/en11030552.
Tillero, E., Mogollon, J.L., 17-19 May 2017. An appropriate waterflood or water drive
Supplementary data to this article can be found online at https://doi. performance analysis for any oil/water viscosity ratio. In: Proceedings of the SPE
org/10.1016/j.petrol.2020.107227. Latin America and Caribbean Petroleum Engineering Conference. Buenos Aires,
Argentina. SPE-185465-MS.
Tueros, J.A.R., Horowitz, B., Willmersdorf, R.B., Oliveira, D.F.B., 2018. Non-distance-
References based localization techniques for ensemble-based waterflooding optimization.
J. Petrol. Sci. Eng. 170, 440–452.
Chakra, N.C., Song, K.Y., Gupta, M.M., Saraf, D.N., 2013. An innovative neural forecast Wang, R., Hou, J., Li, Z., 2013. A new water displacement curve for the high water-cut
of cumulative oil production from a petroleum reservoir employing higher-order stage. Petrol. Sci. Technol. 31, 1327–1334.
neural networks (HONNs). J. Petrol. Sci. Eng. 106, 18–33. Wang, D., Li, Y., Hu, Y., Li, B., Deng, X., Liu, Z., 2016. Integrated dynamic evaluation of
Clarkson, C.R., 2013. Production data analysis of unconventional gas wells: review of depletion-drive performance in naturally fractured-vuggy carbonate reservoirs using
theory and best practices. Int. J. Coal Geol. 109, 101–146. DPSO-FCM clustering. Fuel 181, 996–1010.
Dou, H., Zhang, H., Shen, S., 2019. Correct understanding and application of Wang, D., Li, Y., Chen, B., Hu, Y., Li, B., Gao, D., Fu, B., 2017a. Ensemble-based
waterflooding characteristic curve. Petrol. Explor. Dev. 46 (3), 1–8. optimization of interwell connectivity in heterogeneous waterflooding reservoirs.
El-Deek, I., Abdullatif, O., Korvin, G., 2017. Heterogeneity analysis of reservoir porosity J. Nat. Gas Sci. Eng. 38, 245–256.
and permeability in the Late Ordovician glacio-fluvial Sarah Formation paleovalleys, Wang, J., Shi, C., Ji, S., Li, G., Chen, Y., 2017b. New water drive characteristic curves at
central Saudi Arabia. Arabian J. Geosci. 10 (18), 400. ultra-high water cut stage. Petrol. Explor. Dev. 44 (6), 955–960.
Fanchi, J.R., Cooksey, M.J., Lehman, K.M., Smith, A., Fanchi, A.C., Fanchi, C.J., 2013. Xi, Z., Wei, S., Xiaojing, Y., 2017. Quantitative calculation of tight sandstone reservoir
Probabilistic decline curve analysis of Barnett, Fayetteville, Haynesville, and porosity evolution based on different diagenesis:A case study of Chang 6_3 reservoir
Woodford gas shales. J. Petrol. Sci. Eng. 109, 308–311. in Huaqing area,Ordos Basin. Petrol. Geol. Exp. 39 (1).
Feng, Q., Zhang, J., Wang, S., Wang, X., Cui, R.H., Wang, D.P., Bing, S.X., Rui, Z.H., Yildiz, T., Khosravi, A., 2007. An analytical bottom water drive aquifer model for
2017. Unified relative permeability model and waterflooding type curves under material balance analysis. SPE Reservoir Eval. Eng. 10 (6) https://doi.org/10.2118/
different levels of water cut. J. Petrol. Sci. Eng. 154, 204–216. 103283-PA.
Gong, X., Gonzalez, R., McVay, D.A., et al., 2014. Bayesian probabilistic decline-curve Zhu, H., Zeng, H., Liu, K., 2013. A quantitative simulation study of asymmetrical tectonic
analysis reliably quantifies uncertainty in shale-well-production forecasts. SPE J. 19 subsidence control on non-synchronous sequence stacking patterns of Eocene
(6), 1,047–1,057. lacustrine sediments in Bohai Bay Basin, China. Sediment. Geol. 294, 328–341.

11