INTERNATIONAL JOURNAL OF A publication of IIETA

HEAT AND TECHNOLOGY

ISSN: 0392-8764
Vol. 34, No. 3, September 2016, pp. 423-428
DOI: 10.18280/ijht.340311
Licensed under CC BY-NC 4.0 http://www.iieta.org/Journals/IJHT

Improved Aziz Prediction Model of Pressure Gradient for Multiphase Flow

in Wells
Mengxia Li 1, 2, 3, Ruiquan Liao 2, 3, Wei Luo 3, 4 and Yong Dong 1*
*1 School
of Information and Mathematics, Yangtze University, Jingzhou Hubei 434023, China
Petroleum Engineering College, Yangtze University, Wuhan Hubei 430100, China
2
3 The Branch of Key Laboratory of CNPC for Oil and Gas Production, Yangtze University, Wuhan

Hubei 430100, China

4 School of Geosciences, Yangtze University, Wuhan Hubei 430100, China

Email: dongyong80@126.com

ABSTRACT

It is crucial for the completion parameter design and production performance detection interpretation to have
an accurate pressure gradient. The Aziz prediction model of pressure gradient is a common calculation model
in oil-gas field development. The laboratory experiment results of multiphase flow show that the average
prediction relative error is 29.62% and the maximum relative error reaches 70.1%. By comparing the
prediction residual of the Aziz model with the experiment condition parameters, as the volume flow rate of
the liquid phase is constant, this paper considers that the prediction residual of the Aziz model is closely
related to the gas liquid ratio and has no clear correlation with the water fraction of fluid phase. Based on
unifying the orders of magnitude of the prediction residual and the gas liquid ratio, this paper uses a cubic
function of the gas liquid ratio to fit the prediction residual of the model to obtain a new pressure prediction
method which is called Aziz-I model. The results for ninety groups of experimental data show that the
average prediction relative error of the Aziz-I model is reduced to 10.82%. Hence, the Aziz-I model improves
the prediction accuracy of pressure gradient for multiphase flow.

Keywords: Pressure gradient, Multiphase flow, Prediction, Aziz model, Function fitting.

1. INTRODUCTION currently. On one hand, because of the complexity of the

multiphase flow, it is difficult to find an accurate model, and
Having an accurate prediction of the pressure gradient of the model used is only approximate for describing some
the multiphase flow in oil and gas wells is essential for the parameters, resulting in model error. On the other hand, there
completion parameter design and production performance are many variable factors in collecting data, resulting in
detection interpretation. The pressure gradient prediction is observation error. Hence, these errors affect the prediction
fundamental in the process of petroleum reservoir results.
exploitation. It determines many aspects in the dynamically In recent years, artificial intelligence methods (neural
changing process of petroleum reservoir exploitation. Many network, fuzzy logic, genetic algorithm, etc.) have become an
researchers have done much work [1-5] in this field. increasingly powerful and reliable technological means for
Internationally, academics have long been working on the petroleum engineers to analyze and interpret the problems in
problem of multiphase flow in wells, having proposed several oil and gas production areas. In 2004, Ayoub [6] proposed
prediction models. Without considering the models that have the common pressure drop model in pipes by using the neural
not received wide acceptance, among the proposed models network method. In 2010, Mohsen Ebrahimi [7] introduced
are the common models of the Hagedorn-Brown model, the fuzzy logic and neural network for the research of the
Orkiszewski model, Aziz model, Beggs-Brill model, pressure drop prediction models of multiphase flow in
Mukherjee-Brill model, Hasan model, Ansari model, Kaya horizontal pipes. In 2011, Ahmed Al-Shammari [8]
model, Cheng Jialiang model, Wang qisheng model, Liao researched the pressure distribution of gas and liquid two-
model. These models are empirical or semi-empirical phase flow in vertical wells by using the fuzzy logic method.
correlations. Although there are many continued In 2012, W Al-Mudhafer [9] introduced the hybrid genetic
improvements on the pressure gradient prediction models by algorithm and fuzzy logic method to the research of
some researchers, statistical calculation shows that they are multiphase flow. In 2015, aiming at the pressure gradient
not very satisfactory in obtaining the pressure gradient prediction of multiphase flow, M. Attia [10] analyzed the

423
effects the four artificial intelligence methods which are the
fuzzy logic method, neural network method, support vector
machine, and decision-making tree, and the prediction
accuracy is relatively high. But all these artificial intelligence
methods use genetic algorithms to determine the parameters
of the models. Because of a large number of parameters for
the genetic algorithms, it is complicated to program. In
addition, the convergence rate of genetic algorithms is
relatively slow and the search efficiency is very slow so that it
easily gets trapped into local optimal solution.
In previous research, based on the experimental data of oil-
gas-water three phases flow in vertical wells which are from
the branch of key laboratory of CNPC for oil and gas
production, the research team of this paper compared and
analyzed the prediction results of the above-mentioned
models, and found that the prediction accuracy is generally Figure 1. Comparison results between the Aziz pressure
low, and there is a great difference in prediction accuracy for gradient and experimental pressure gradient
the same model under different flow conditions. After
analyzing the prediction residuals of some models, a With regard to various liquid flow rates, a curve similar to
mathematical regularity is obtained. The Aziz prediction the abscissa GLR and the ordinate predicted residual is
model of pressure gradient [11] is a common calculation shown in Figure 2. In Figure 2, three results can be seen: (1)
model and it has been widely applied. The reference [12] For the residual curves with water content 30%, 60% and
proposed a method based on the residual model to improve 90%, there is no residual curve between the other two
the existing prediction model of pressure gradient and residual curves. This shows that there is no significant
achieved good results. Because of the particularity of the regularity between the pressure gradient and the water
results of reference [12], it is difficult to generalize to other fraction; (2) As the liquid flow rate increases, the value of gas
prediction models. The method proposed in the reference [13] liquid ratio corresponding to the maximum prediction
can be generalized, but the realization process is very residual has a decreasing trend; (3) For the constant liquid
complex. Based on the experimental data obtained from oil- flow rate, the prediction residual curves of gas liquid ratio
gas-water three phases flow experiment in a vertical well with have certain similarities in their geometry shapes. Hence,
diameter 75 mm, this paper uses the Aziz model to calculate with respect to the constant liquid flow rate, this paper
the average relative error between the calculation pressure establishes the correlation between the prediction residual
gradient and experimental pressure gradient exceeding 29%. and the gas liquid ratio.
Hence, the Aziz model must be further researched to build the
improved calculation model.
This paper firstly analyzes the prediction residual of the
Aziz model. Here, the prediction residual, referred to as the
Aziz residual, is defined by subtracting the experimental
pressure gradient from the pressure gradient calculated by the
Aziz model. Then, the relationships between the Aziz
residual and the experiment parameters is analyzed,
discovering that the gas liquid ratio and the Aziz residual
have good correlation. Hence, this paper builds a new
pressure gradient prediction model, referred to as the Aziz-I
model, by using the cubic polynomial function of the gas
liquid ratio to fit the Aziz residual. The results show that
Aziz-I model more coincides with the experimental results.

2. PREDICTION EFFECT OF AZIZ MODEL

The basic experiment conditions are given below. The

diameter of the vertical well is 75 mm. The water fractions
are 30%, 60% and 90% respectively. The liquid flow rates
are 10 m3/d, 15 m3/d, 20 m3/d, 30 m3/d, 40 m3/d and 50
m3/d. The gas liquid ratios are 50, 100, 150, 200 and 300.
The experiment temperature is between 14℃and 17℃. The
flow media is air, 5# white oil and water. There are 90 groups
of experimental data. The roughness is taken as 0.0002 mm.
The comparison result between the pressure gradient
calculated by the Aziz model and the experimental pressure
gradient is shown in Figure 1. The average residual of the
Aziz residual is 0.3528 kPa/m and the average relative error
reaches 29.62%.

424
 3. CUBIC FUNCTION FITTING MODEL OF AZIZ
RESIDUAL

Bases on the foregoing analysis, for six different liquid

flow rates, the cubic polynomial function must be used to fit
the prediction residual with respect to the gas liquid ratio.
Because of the large difference in the orders of magnitude
between the Aziz residual and the gas liquid ratio, before
building the cubic function fitting, the gas liquid ratio is
divided by 300.
This paper applies the cubic polynomial function to obtain
the improved Aziz residual model, denoted as the Aziz-I
model. It is shown in Eq(1).

3 2
 zj   zj   zj 
rˆj  Aj    Bj    Cj    Fj ,
 300   300   300 
j  1, 2, ,6 (1)

In Eq(1), z j represents the gas liquid ratio with unit

m3/m3; rˆj is the fitting value of the prediction residual of the
Aziz model with unit kPa/m; Aj , B j , C j and Fj are
undetermined fitting coefficients.
The values of undetermined coefficients Aj , B j , C j , and
Fj for j  1, 2, ,6 can be determined by the least square
method [14-16] or some optimization algorithms [17-19].
In this paper, we use an improved particle swarm
optimization [19], which is described in the next section, to
determine the coefficients Aj , B j , C j , and Fj for
j  1, 2, ,6 .

4. NEW CHAOS PARTICLE SWARM OPTIMIZATION

COMBINED WITH CHAOTIC PERTURBATION

The specific steps of the new chaos particle swarm

optimisation algorithm combined with the chaotic
perturbation are described as follows.
Step 1 Initialize the inertia weight wmax and wmin , the
learning factor c1 and c2 , the group size N , the maximum
iterations MaxDT , the problem dimension D , and the
accuracy control eps  106 . Give the optimal space [lb, ub]
and the speed limit vmax . Give the maximum iterations of
chaos search HDT .
Step 2 Randomly produce a D -dimensional particle over
the interval [0,1) . According to the statement in section 2.2,
obtain N s D -dimensional particles which are denoted as
xi , i  1, 2, , N . A random D -dimensional space vector is
selected from [0,1) . Use chaos mapping [20] to obtain
N 1 s particles, and obtain the N s D -dimensional vector
which is the initial particle speed. Let the iterations be 0, and
continue to step 3.
Step 3 Insert xi into the objective function to calculate the
fitness fi , and determine the global optimal positions of the
Figure 2. Correlation between the prediction residual and particle swarm gbest , the experienced optimal positions of
the gas liquid ratio
particles pbesti ， i  1,2, , N . Continue to step 4.
Step 4 w decreases according to Eq(2). The positions and
speed of the particles are updated according to Eq(3) and

425
 MaxDT  t  1
Eq(4). The iterations increase by 1. Take k1  ,
MaxDT
and determine the chaotic mutation and the chaos search
interval [lb(t ), ub(t )] according to Eqs(5) and (6). Perform
chaotic mutation to update gbest and pbesti . Perform
chaotic search to gbest , and update gbest and pbesti .
Compute the fitness variance of the current particle swarm. If
the absolute value of the difference between the fitness
variance of the current particle swarm and that of the pre-
iteration is less than eps . Go to step 5, or step 6.

w  wmax  t  (wmax  wmin ) / MaxDT (2)

v(t  1)  w  v(t )  c1  rand1 ( pbest  x(t ))

c2  rand 2  ( gbest  x(t)) (3)

x(t  1)  x(t )  v(t  1) (4)

CD(t )  k1  k2   kt  CD(0) (5)

ub(t )  (1  k1k2 kt )  x(t )  k1k2 kt  ub (6)

Step 5 Compute the numbers of the particles s that need

to be replaced. Produce s s new particles using a similar
manipulation to produce the initial particle positions. Use the
new particles to replace the current s s particles that have the
worst fitness. Then go to step 4.
Step 6 If the iterations are less than MaxDT , go to step 4.
Otherwise, go to step 7.
Step 7 Output the final results: gbest and fbest .
For different liquid flow rates, the values of Aj , B j , C j ,
and Fj for j  1, 2, ,6 are shown in Table 1.

Table 1. Values of undetermined coefficients for different

liquid flow rates

Liquid flow Aj Bj Cj Fj
rate（
m3/d）
10 -0.9011 -1.1512 3.6104 -1.1542
Figure 3. Comparison between Aziz residual and the
15 10.2898 -21.2416 12.7851 -1.9154
improved Aziz residual
20 11.6489 -23.4322 13.4969 -2.0891
30 6.6078 -8.0887 0.6005 0.6505
40 -7.3502 16.3762 -11.0033 1.8583 5. IMPROVED AZIZ MODEL
50 -18.8094 35.8775 -19.9684 2.7399
As indicated in the introduction, the Aziz residual is
obtained by subtracting the experimental pressure gradient
When the liquid flow rates are taken as 10 m3/d, 30 m3/d from the Aziz pressure gradient. By combining the Aziz
and 50 m3/d respectively, the Aziz residual values and the model and the improved Aziz residual model of residual,
improved Aziz residual values are shown in Figure 3. denoted as Aziz-RF model, the pressure gradient prediction
In Figure 3, the black curve represents the improved Aziz model can be established, denoted as the Aziz-I model, which
residual curve by fitting. Figure 3 shows that the improved is obtained by subtracting the Aziz-RF model from the Aziz
Aziz residual value and the Aziz residual are very close at the model.
measure point set by the experimental scheme. Hence, it is The prediction pressure gradient of the Aziz-I model is
feasible to use the improved Aziz residual curve to
approximate the Aziz residual. The improved Aziz residual
Aziz  I Value  Aziz Value  Aziz  RF Value (7)
model can be taken as a reliable estimation model of the Aziz
residual.

426
where, Aziz  I Value denotes the calculating pressure ACKNOWLEDGMENT
gradient value from the Aziz-I model with unit kPa/m;
Aziz Value denotes the calculating pressure gradient value The authors wish to thank the Branch of Key Laboratory of
CNPC for Oil and Gas Production for their great help. This
from the Aziz model with unit kPa/m; Aziz  RF Value paper is supported by National Natural Science Foundation of
denotes the improved Aziz residual value with unit kPa/m. China (61572084 and 51504038).
The relative error between the prediction pressure gradient
of the Aziz model and the experimental pressure gradient is
defined in Eq(10). REFERENCES

Aziz  I Value-E Value [1] A. S. Kaya, C. Sarica and J. P. Bril, “Mechanistic

err  (8) modeling of two-phase flow in deviated wells,” Spe
E Value
Production and Facilities, vol. 16, no. 3, pp. 156–165,
Aug., 2011. DOI: 10.2118/72998-PA.
In Eq(8), E Value denotes the experimental pressure [2] Tian Xianglei, Jiang Haiyan, Yue Yanru and Zhang
gradient with unit kPa/m. Yongchang, “Research on production law for water
According to the 90 groups of experimental data shown in drainage and gas production by gas lift,” Fault-Block
Figure 2, the average value of the prediction relative error of Oil & Gas Field, vol. 19, no. 6, pp. 760–763, Feb.,
the Aziz-I model is 10.82% by comparing the relative error 2013. DOI: 10.6056/dkyqt201206020.
between the Aziz prediction pressure gradient and the [3] Bai Xiaohong, Tian Wei, Tian Shubao and Li Kang,
experimental pressure gradient. “Optimization on wellbore flow parameters of gas lift
The comparison between the calculating pressure gradient deliquification in low production and liquid loading
of the Aziz-I model and the experimental pressure gradient is gas wells,” Fault-Block Oil & Gas Field, vol. 21, no.
shown in Figure 4. 1, pp. 125–128, Jan. 2014. DOI:
10.6056/dkyqt201401031.
[4] Tian Yun, Wang Zhibin, Li Yingchuan, Bai Huifang
and Li Kezhi, “Evaluation and optimization of
wellbore pressure drop model for drainage and gas
recovery by velocity string,” Fault-Block Oil & Gas
Field, vol. 22, no. 1, pp. 130–133, Jan. 2015. DOI:
10.6056/dkyqt201501029.
[5] M. Attia, A. Abdulraheem and M. A. Mahmoud,
“Pressure drop due to multiphase flow using four
artificial intelligence methods,” in SPE North Africa
Technical Conference and Exhibition, 2015.
[6] J. A. Ayoub and A. D. Hill, “Production operations
R&D—Optimizing performance from the sandface to
the tanks,” Journal of Petroleum Technology, vol. 56,
no. 7, pp. 36–39, Jan. 2004.
[7] Mohsen Ebrahimi and Ali Sajedian, “Use of fuzzy
Figure 4. Comparison between the calculating pressure
logic for predicting two-phase inflow performance
gradient of the Aziz-I model and experimental pressure
relationship of horizontal oil wells,” in Trinidad and
gradient
Tobago Energy Resources Conference, 2010.
[8] Ahmed Al-Shammari. “Accurate prediction of pressure
By comparing the results in Figure 1, the distribution of drop in two-phase vertical flow systems using artificial
data points is closer to the diagonal. It shows that the new intelligence,” in SPE/DGS Saudi Arabia Section
model (Aziz-I model) has higher accuracy. Technical Symposium and Exhibition, 2011.
[9] W Al-Mudhafer, M Alabbas, “Application of a
hybrid system of genetic algorithm & fuzzy logic
6. CONCUSION as optimization techniques for improving oil recovery
in a sandstone reservoir in Iraq,” in SPE Latin
(1) This paper firstly establishes the cubic fitting function America and Caribbean Petroleum Engineering
model of the Aziz prediction residual with respect to the gas Conference, 2012.
liquid ratio. Then, it builds a new pressure gradient prediction [10] M. Attia, A. Abdulraheem, M. A. Mahmoud, “Pressure
model (Aziz-I model) by combining the cubic fitting function drop due to multiphase flow using four artificial
model and the Aziz model. intelligence methods,” in SPE North Africa
(2) For the pressure gradient under the experimental Technical Conference and Exhibition, 2015.
condition, the average prediction relative error value of the [11] K Aziz, GW Govier, “Pressure drop in wells
Aziz-I model is 10.82%. This is a significant improvement on producing oil and gas,” The Journal of Canadian
the prediction accuracy of the Aziz model. Petroleum Technology, vol. 11, no. 3, pp. 38–48, July
(3) The Aziz-I model proposed in this paper is more 1972. DOI: 10.2118/72-03-04.
concise compared to that in reference [8]. It also has good [12] Mengxia Li, Ruiquan Liao, Yong Dong and Wei Luo,
extendibility. “New prediction method of pressure drop for oil-gas-
water three-phase vertical flow,” J Balk Tribol AssoC,
vol. 22, no. 1A-I, pp. 680-695, 2016.

427
[13] Yong Dong, Mengxia Li, Ruiquan Liao and Wei Luo, c2 learning factor
“Modification of Beggs-Brill pressure gradient CD length of the corresponding interval of
predicting model for multiphase flow in vertical the mutation
wells,” Journal of Oil and Gas Technology, vol. 38, D problem dimension
no. 1, pp. 40–47, Mar. 2016. DOI: eps accuracy control
10.12677/jogt.2016.11006. experimental pressure gradient, kPa/m
E Value
[14] O. Axelsson, “A generalized conjugate gradient, least
square method,” Numerische Mathematik, vol. 51, no. F dimensionless fitting coefficient
2, pp. 209–227, Mar. 1987. DOI: fi fitness
10.1007/BF01396750. fbest optimum value
[15] Y. Shi, C. H. Liang, “The finite-volume time-domain global optimal positions of the particle
gbest
algorithm using least square method in solving swarm
Maxwell’s equations,” Journal of Computational maximum iterations of chaos search
HDT
Physics, vol. 226, no. 2, pp. 1444-1457, Oct. 2007. adjustment coefficient of weight
k
DOI: 10.1016/j.jcp.2007.05.033.
lb lower limit of the initial search space
[16] W. Zeng and S. Feng, “Approximate reasoning
algorithm of interval-valued fuzzy sets based on least ub upper limit of the initial search space
square method,” Information Sciences, vol. 272, no. 8, MaxDT maximum iterations
pp. 73–83, July 2014. DOI: 10.1016/j.ins.2014.02.078. N group size
[17] Tao Lin, Peng Wu and Fengmei Gao, “Study on SVM pbesti experienced optimal positions of
temperature compensation of liquid ammonia particles
volumetric flowmeter based on variable weight PSO,” r fitting value of the prediction residual of
Int J Heat & Tech, vol. 33, no. 2, pp. 151–156, Jun. Aziz model, kPa/m
2015. DOI: 10.18280/ijht.330224. rand1 uniform random numbers over the
[18] C. H. Liang, S. Zeng, Z. X. Li, D. G. Yang and S. A. interval [0,1]
Sherif, “ Optimal design of plate-fin heat sink under rand 2 uniform random numbers over the
natural convection using a particle swarm optimization interval [0,1]
algorithm,” Int J Heat & Tech, vol. 34, no. 2, pp. 275- s numbers of the particles which is needed
280, Jun. 2016. DOI: 10.18280/ijht.340217. to be replaced
[19] Li Mengxia and Liao Ruiquan, “A new chaos particle current iterations
swarm optimization combining the chaotic t
perturbation,” IJSIP, vol. 8, no. 4, pp. 41-48, April vmax speed limit
2015. DOI: 10.14257/ijsip.2015.8.4.04. particle velocity
[20] R. Anderson, “Industrial cryptography,” Iee Review, v
vol. 42, no. 3, pp. 118-120, May 1996. DOI: w weight
10.1049/ir:19960314. maximum of the weight
wmax
minimum of the weight
NOMENCLATURE
wmin
A dimensionless fitting coefficient x particle position
Aziz Value calculating pressure gradient value by xi the i-th D -dimensional particle
Aziz model, kPa/m z gas liquid ratio, m3/m3
Aziz  I Value calculating pressure gradient value by
Aziz-I model, kPa/m Subscripts
Aziz  RF Value improved Aziz residual value, kPa/m
B dimensionless fitting coefficient i serial number of the particle
C dimensionless fitting coefficient j the number of liquid flow rates
c1 learning factor

428