Sunteți pe pagina 1din 14

IADC/SPE-180573-MS

Effective Calculation Model of Drilling Fluids Density and ESD for HTHP
Well While Drilling
H. Zhou, X. Niu, H. Fan, and G. Wang, SINOPEC Research Institute of Petroleum Engineering

Copyright 2016, IADC/SPE Asia Pacific Drilling Technology Conference

This paper was prepared for presentation at the IADC/SPE Asia Pacific Drilling Technology Conference held in Singapore, 22–24 August 2016.

This paper was selected for presentation by an IADC/SPE program committee following review of information contained in an abstract submitted by the author(s).
Contents of the paper have not been reviewed by the International Association of Drilling Contractors or the Society of Petroleum Engineers and are subject to
correction by the author(s). The material does not necessarily reflect any position of the International Association of Drilling Contractors or the Society of Petroleum
Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the International Association
of Drilling Contractors or the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words;
illustrations may not be copied. The abstract must contain conspicuous acknowledgment of IADC/SPE copyright.

Abstract
Drilling wells for oil/gas has been increasingly challenging with the companies moving towards difficult
environments, such as in Tarim basin of China, some reservoirs buried so deeply (⬎7,000m) that we
experience high temperature and pressure. The problems faced in these locations range from very narrow
margin between pore (or collapse) and fracture pressure. The density of drilling fluid is often affected by
HTHP, the careful research on the drilling fluids density at HTHP is very important for precisely
predicting ESD as well as controlling the downhole pressure. A utility calculation model of drilling fluids
density and ESD was proposed, which can predict the HTHP density and ESD.
Firstly, a new utility artificial neural network HTHP drilling fluid density prediction model was
established based on the traditional BP neural network and PSO (Particle Swarm Optimization) optimi-
zation method. Then PSO-BP neural network HTHP drilling fluid density prediction model was proposed,
in which the influence of drilling fluid component (oil phase, water phase volume fraction) was taken into
account. Available experimental measurements of water-based and oil-based drilling fluids at pressure
ranging from 0-96MPa and temperatures up to 183°C were used to develop and the PSO-BP network
model and then the network weights, threshold parameters. Through this model the high-precision HTHP
drilling fluids density can be obtained easily with the knowledge of the drilling fluids component data (oil
phase and water phase volume fractions) and its density at standard conditions(0MPa, 20°C) based on the
basic principle of PSO-BP network. Moreover, an new comprehensive ESD calculation model of HTHP
well was established, which is applicable for all common type drilling fluids and through we can obtained
the ESD profile of the well easily.
The prediction of this model has been compared with an extensive set of data from literature, the
comparisons of different fluids density in HTHP show very good agreement, the prediction accuracy was
improved, and in which the maximum average absolute error of predictions is less than 0.005sg. Finally,
the proposed model has been applied for HTHP drilling fluids density and ESD prediction in several wells
of the Tarim basin in China, the results show that the proposed model can exactly provide the HTHP
drilling fluids density and ESD profile.
This study proposed an utility calculation of HTHP drilling fluids density and ESD profile while
drilling, based on the new PSO-BP neural network, the optimal network weights, threshold parameters of
2 IADC/SPE-180573-MS

was obtained, through which the high-precision drilling fluids density and ESD can be obtained easily.
Moreover, the model has been verified and applied for field monitoring. Therefore, this model can be
applied to provide more accurate predictions of HTHP drilling fluids density and ESD while drilling.
Introduction
As deeper well are being drilled in searching for new crude oil and natural gas reservoir, the downhole
environment is very severe and tough. Such as high temperature high pressure (HTHP), narrow safe mud
window drilling problems are widespread during drilling in these wells (Yang et al., 2010). The prediction
and control Equivalent Circulating Density (ECD) while drilling become increasingly important. The
successful prediction ECD is crucially dependent on an accurate representation of the drilling fluid density
and Equivalent Static Density (ESD) under the HTHP environment. The drilling fluid density under
downhole conditions may be very different from those measured under ambient pressure and temperature
at the surface. Drilling fluid could be compressed under high pressure as well as be expanded under the
high temperature conditions. Therefore, pivotal of the ECD prediction and control is the accurate HTHP
drilling fluid density and ESD prediction.
There are many published literatures focusing on the drilling fluid density and ESD calculation under
HTHP conditions. Earliest, Methven et al. (1972) studied the oil-based mud density at HTHP based on the
HTHP experimental of two types drilling fluids and obtained mass PVT data of drilling fluid density, and
then proposed an the drilling fluid density curves under three geothermal gradient conditions, through
which the OBM density can be obtained easily. Hoberock et al. (1982) proposed a density prediction
model according to the drilling fluid composition based on the material balance principle, which is known
as the original component model for predicting drilling fluid density. McMordie et al. (1982) studied the
effect of temperature and pressure on both of water-based mud and oil-based mud, the muds density was
tested under elevated temperature (21.1°C-204.4°C) and pressure (0MPa-96.5MPa), and then proposed
the initial empirical model based on the experimental data. Thomas et al. (1984) provided a CaCl2 drilling
fluid density prediction model based on the component model, through which the drilling fluid density of
when the temperature lower than 122°C and pressure less than 150MPa can be calculated easily. Kemp
et al. (1987) established an improved component model, in which the more influence factors were taken
into account. Perts et al. (1988) analyzed the OBM density at different temperature and pressure based on
several set experiment and provided several set experimental data. Yan et al. (1992) found the law that
the density of OBM changing with temperature and pressure is comforted to an exponential form equation
based on the HTHP density experiment, and then a empirical HTHP density prediction model based on
the experimental data using multivariate non-linear regression analysis method. Karstad et al. (1998)
established a more complex empirical model for predicting HTHP drilling fluid density, the prediction
precision is more accuracy than the traditional empirical model because this model took the quadratic term
and cross term of pressure and temperature were into account. But there are also drawback that there are
more model coefficients and these coefficients need to be calculated based on the more complex
regression method. Wang and Liu (2000) established an exponential form model for predicting HTHP
drilling fluid density based on the theoretical analysis and introduced the theory of critical temperature and
critical depth. Based on the Peter’s and McMordie’s experimental data, Guan et al. (2003) proposed a
HTHP prediction for OBM by introducing thermal expansivity and compressibility. Moreover, there were
several researchers (Wang et al. 2000; Zhang et al. 2006; Wang et al. 2008) studied the HTHP drilling
fluid density based on the experimental data and provided the improved model based on the initial
empirical model, all most of these model are similar and just the form of the model function were
improved based on the more complex regression method. But all of above prediction model were
developed based on the initial component model either or the component model. The each component
HTHP density of drilling fluid such as water, oil and solid are all difficult to calculate for the component
model which limit the application of this model. Furthermore, the empirical model established based on
IADC/SPE-180573-MS 3

the experimental data of the particular drilling fluids, which is only suitable for the tested drilling fluid.
Recently, Osman et al. (2003) researched the PVT properties of crude oil and established a simple
back-propagation artificial neural network (BP ANN) model for predicting the HTHP drilling fluids
density, but in which the drilling fluid components didn’t be took into account.
Therefore, it is necessary to do further research on how to accurately calculate the drilling fluid density
under HTHP conditions. In this article, rather than use the traditional BP ANN model, a PSO-BP HTHP
drilling fluid density prediction model was established, in which the drilling fluid components were took
into account and its applicable for common type drilling fluid.

PSO-BP Model for HTHP Drilling Fluid Density Prediction


An ANN is a computer model that attempts to mimic simple biological learning processes and simulate
functions of human nervous system for developing associations, transformations or mapping between
objects or data, which also has a wide range of applications in oil industry. Gharbi and Elsharkawy (1997)
estimated the bubble point pressure and oil formation volume factor (FVF) for Middle East crude-oil
reservoirs used the ANN system. Osman et al. (2001) estimated an ANN model for calculating FVF at the
bubble point pressure. As mentioned before, Osman et al. (2003) proposed a BP ANN for prediction
HTHP drilling fluid density. But all previous studies used traditional BP ANN, and the BP ANN has some
inherent flaws as it does easy fall into local minimum rather than the global optimum network parameters.
Eberhart & Kennedy (1995) introduced the Particle Swarm Optimization (PSO) algorithm, which is a
computer algorithm evolved based on the swarm intelligence and overcomes the shortcoming of the
traditional BP algorithm. There are few works related to application PSO-BP networks in petroleum
problems, and non above about the HTHP drilling density predictions. And so in this section a POS-BP
ANN model for predicting HTHP drilling fluid density was be established.

Experimental Data
Basic Experimental Data The experimental measurements presented by McMordie et al. (1972) have
been used in developing the presented PSO-BP ANN model. McMordie et al gave several sets of the
drilling fluid density under different temperature and pressure (21.1°C-204.4°C, 0MPa-96.5MPa) about
both of water-based mud and oil-based mud. The basic formulation of water-based drilling fluid consisted
of 350g water, distilled water, 20g bentonite, 3g chrome lignosulfonate, 3g lignite and the PH value
regulator. The oil-based mud was composed of 298cm3 of No.2 diesel oil, 52cm3 of 30% calcium chloride
brine, 5g organophilic bentonite, 5g emulsifier and 2g hydroxide. Both water-based mud and oil-based
mud were made into three different densities, the initial densities of water-based mud were 1.28g/cm3,
1.63g/cm3, 2.15g/cm3 and the oil-based mud densities were 1.31g/cm3, 1.71g/cm3, 2.15g/cm3. Based on
the detail formula data of the drilling fluid, the volume fraction of water and oil about different density
drilling fluid can be obtained easily. And the Table 1 listed the detail data of different drilling fluid, the
initial temperature T0 was 21.1°C, and the initial pressure P0 is 0MPa.

Table 1—The data of drilling fluids composition


Water-based Mud Oil-based Mud

Mud Type and No. No.1 No.2 No.3 No.4 No.5 No.6

␳ (P0,T0) 1.28 1.63 2.15 1.31 1.71 2.15


fo(%) 0 0 0 10.57 9.14 7.33
fw(%) 88.92 78.33 62.54 72.63 62.8 50.36
4 IADC/SPE-180573-MS

Data Preprocessing Based on the experimental data, we can obtain several sets HTHP fluid density that
related to the corresponding temperature and pressure, viz. ␳(P , Ti) related to Pi and Ti. We used No.1, No.
i

4 and No.6 drilling fluid experimental data in training the presented PSO-BP ANN model. These data
were selected to represent the entire range of temperature and pressure. Before the training procedure, we
should preprocess the initial experimental data. For obtained the temperature difference and pressure
difference we can use the following equation:
(1)

where Pi and Ti were the corresponding temperature and pressure of ␳(P , Ti), ⌬P and ⌬T is pressure
i

difference and temperature difference corresponded with ␳(P , Ti). i

One important procedure before training the ANN was to uniformization the input and output data of
the ANN, which should be uniformed to the range [-1, 1]. And we could use the following equation
obtained the uniformed data for the input and output data:
(2)

where xi,max is the max value of the corresponding parameter and the xi,min is the min value of the
corresponding parameter; xi is the initial experimental data, such as ␳(P ,Ti), ␳0, ⌬T, ⌬P, fw, fo and ␳(P,T);
i

the xi was the uniformed data for the corresponding parameter xi, the value of which is between -1 to 1.
And the Table 2 listed the maximum and minimum value of the input and output parameters.

Table 2—The Maximum and Minimum values of the input and output parameters
Parameter ⌬P ⌬T ␳0 fw fo ␳ (⌬p,⌬T)

Max 96.46 183.435 2.1545 0.8892 0.7264 2.212


Min 0 0 1.2809 0.0733 0 1.149

Based on the Table 2 and the Eq. 2 the parameters for training the PSO-BP ANN can be obtained easily.
PSO-BP ANN Model of Drilling Fluid Density Prediction
ANN Basic Structure Determination In order to improve the efficiency and prevent divergence of
model, the classical three-layer network structure was selected to establish the PSO-BP ANN, which has
one input layer, one output layer and one hidden layer. Drilling fluid is composed of liquid-phase and
solid-phase, while the liquid-phase could be easier compressed with the pressure increases as well as be
expanded when temperature increase rather than solid-phase. And so we mainly considered the impact of
thermal expansion and pressure compression on liquid-phase during the model establishing procedure.
The liquid-phase of drilling mainly composed of water and oil and there are different variation rules of
thermal expansion and pressure compression between water phase and oil phase. As a consequence, the
variety of water and oil should be taken into account separately. Therefore, we set five neurons for the first
layer representing the values of the input data. These are the initial density at initial pressure and
temperature conditions (P0, T0), the corresponding pressure difference (⌬P), the corresponding temper-
ature difference (⌬T), the water volume fraction (fw) and the oil volume fraction (fo). The output layer
contains one neuron representing the desired output which is the density at the input temperature and
pressure difference. The neuron number of the hidden layer in the network should be obtained based on
IADC/SPE-180573-MS 5

an iterative training optimal number, and the iterative range can be calculated based on the following
formula:
(3)

where m is the number of hidden layer neuron, dimensionless; n is the number of input layer neuron
and l is neuron number for the output layer, dimensionless; ␣ is trial calculation parameter, the range of
which is 1-10. Based on the values of n and l, we can calculate the range of hidden layer neuron number
m is 4-13.
PSO-BP ANN Model As mentioned before, the traditional BP training method of ANN have the
inherent flaws as it does easy fall into local minimum rather than the global optimum network parameters.
And so in this paper for obtaining the optimal ANN parameters, the PSO optimization algorithm was
selected for the ANN training.
Before starting the training, we set the initial value ␣ as 1, and then the number of the hidden layer can
be obtained used the Eq.3. Based on the ANN topology architecture, the dimension of particle swarm D
can be calculated used the following formula:
(4)

After the dimension of particle swarm was calculated, the particle population should be initialized and
we set the size of the particles as o, and then initialed all the particle’s position xi and velocity vi, viz.
xi⫽(xi1,xi2,. . ., xiD), i⫽1,2,. . ., o; and vi⫽(vi1,vi2,. . ., viD). Based on the basic parameters of the particle
population, the fitness value of every particle can be obtained. And then the optimal fitness value for
particle individuals Pbest and the optimal fitness value for particle group Gbest can be calculated (Zhou,
2014). Estimated whether the iterations have achieved the maximum permissible iterations or the
precision has achieved the requirements, if not, the following equation should be used to evolve the
particles:
(5)

where o is the size of the particles, dimensionless; D is the dimension of the particle swarm,
dimensionless; k is the iterations, dimensionless; xidk is d-dimensional position vector component of the
k iteration, dimensionless; vidk is d-dimensional velocity vector component of the k iteration, dimension-
less; pbidk is d-dimensional position vector component of the optimal fitness value of the k iteration,
dimensionless; gbdk is d-dimensional position vector component of the optimal fitness value of all the
iterations, dimensionless; c1 and c2 is the learning factor for searching, dimensionless; R1* , R2* is the
random number between 0 to 1; w is inertia factor, through which we can adjust the searching range.
Fig.1 show the detail training procedure, through which the best ANN parameter for the setting a can
be obtained. And then we should renew the value of ␣ until it reached to 10. And then we can obtain the
highest accuracy parameter and established the optimal PSO-BP ANN. Based on the training procedure
and the training data, we calculated the optimal number of hidden layer neuron is 5, and the Fig. 2 shows
the ANN structure, and also the corresponded ANN parameter were obtained, which listed in the Table
3.
HTHP Fluid Density Calculation The drilling fluid density of different temperature and pressure can
be calculated based on the ANN parameters of the PSO-BP ANN model listed in the Table 3. Before
starting the calculation, the basic input data such as ␳0, ⌬T, ⌬P, fo, fw of the drilling fluid should be
obtained, then all of them should be uniformed to the range of [-1, 1] through Eq.2, and the max and min
6 IADC/SPE-180573-MS

Figure 1—The flow chart of PSO-BP ANN arithmetic

Figure 2—The basic schematic of PSO-BP ANN

Table 3—The weight values between Input layer and Hide layer of PSO-BP ANN
j

wij No. 1 2 3 4 5

i 1 1.910 15 0.044 60 -0.736 17 1.905 03 0.567 42


2 -0.392 01 -0.147 12 -1.222 94 -0.115 29 -0.138 34
3 -3.414 16 0.155 86 -0.818 74 -8.054 67 4.395 24
4 0.825 88 -0.957 05 -3.074 13 5.352 23 -0.769 70
5 -1.572 15 -0.859 94 -0.799 84 -0.112 74 0.009 34
Threshold value of Hidden layer j 1 2 3 4 5
␪j -2.452 17 -0.625 97 -1.195 98 -0.162 99 0.679 19
vjl j 1 2 3 4 5
-0.135 32 5.947 43 -0.278 48 0.129 15 0.367 34
Threshold value of Output layer ␪l -2.53 85
IADC/SPE-180573-MS 7

value of the input data listed in Table 2. During the calculating, the input data should be calculated using
the following equation:
(6)

where yj is the input data of No.j neurons in hidden layer; wij is the weight values between input and
hidden layer; xi is the input data which obtained based on the Eq.2 and Table 2; ␪j is the threshold value
of hidden layer; n is the neurons number of input layer and m is for the hidden layer neurons.
In order to obtain the input data for the output layer, we should switch the data of every neuron in
hidden layer through the following formula:
(7)

where yi,s is the transformed data of the hidden layer neuron. And then the data of output layer neuron
can be obtained using the following equation:
(8)

where c is the output data of the PSO-ANN; vjl is the weight values between hidden layer and output
layer; ␪l is the threshold value of output layer. According to the output data c, we can calculate the HTHP
drilling fluid density easily through the anti-normalization based on the Eq. (2). It should be noted that the
max and min of output drilling density is listed in the Table 2, the data c should be introduced into the
Eq. (2) as the xi=.

HTHP ESD Calculation Model


The equivalent static density of drilling fluid is an expression of the hydrostatic pressure exerted by the
fluid. Hydrostatic pressure can be defined as the pressure exerted at any point by a column of liquid, and
is a function of the density of the liquid and the height of the liquid column. Because of the density of
drilling fluid is no longer equal to the surficial measured value, but a function of the condition temperature
and pressure. Therefore, in order to obtain the accurate hydrostatic pressure, we set a vertical depth
interval according to the actual situation and then segment the well hole into several segments from
surface to bottom hole. And we record the vertical and measured length of each section during the
segmenting process and the detail section results are shown in the Fig. 3.
8 IADC/SPE-180573-MS

Figure 3—The Detail of Wellbore Section

From the Fig. 3 we can see that, if we want to calculate the bottom density of section i and we know
the bottom density of section i-1, we can use the density of section i-1 bottom to calculate the hydrostatic
pressure of section i, and the pressure at the bottom of section i can be described as the following equation:
(9)

where Pi is the bottom pressure at the section i; Pi-1 is the bottom pressure at the section i-1; ␳i-1 is the
drilling fluid density at the section i-1 bottom. After calculated the bottom pressure of the section i,
combined the temperature at the section i bottom, we can calculated the fluid density at the section i easily
through the PSO-BP ANN density prediction model. It should be pointed out that we used the model was
proposed by Raymond (1969) to calculate the temperature profile in the well hole. Based on the density
value of each section and the Equivalent Static Density (ESD) can be expressed by the following:
(10)

Result and Discussion


Density Calculation Model Validation
Experimental data from literature The utility drilling fluids density model was tested using HTHP
drilling fluids density data these came from two different sources. The first experiment data were used
IADC/SPE-180573-MS 9

from work by McMordie et al (1982), which has been described in detail in previous sections. The second
source of experiment data is published work by Peters et al. (1988), who measured six types drilling fluid
density. The drilling fluids were formulated using three different based oil, and which were formulated
into two different densities, and the details information of the drilling fluid can be found in the paper that
published by Peters et al.
Results of HTHP drilling densities calculation Firstly, the No.2, No.3 and No.5 drilling fluid listed in
the Table 1 was calculated used the proposed calculation model, and the predictions were shown in the
Fig. 4 through Fig. 6. From the cross-plot we can see that there were very good match for both water-based
drilling fluid and oil-based drilling fluid between predictions and the measured density.

Figure 4 —Cross-plot of predicted versus measured density of water based drilling fluid (No.2)

Figure 5—Cross-plot of predicted versus measured density of water based drilling fluid (No.3)
10 IADC/SPE-180573-MS

Figure 6 —Cross-plot of predicted versus measured density of water based drilling fluid (No.5)

And then the prediction and measured density had been analyzed in detailed, and we found that the
maximum absolute error of the prediction is about 0.018g/cm3 and the maximum relative error is only
0.9%. Furthermore, the maximum average absolute error of predictions is less than 0.005g/cm3, and we
can see that the proposed PSO-BP ANN can predict the HTHP drilling fluids density very well.
Fig. 7 through Fig. 9 compared the predicted and experimental HTHP drilling fluid density from the
Peter’s study. The agreement is satisfactory between predicted and experimental values of all six types
drilling fluid.

Figure 7—Cross-plot of predicted versus measured density of oil based drilling fluid (No.2 diesel)
IADC/SPE-180573-MS 11

Figure 8 —Cross-plot of predicted versus measured density of oil based drilling fluid (Mineral A)

Figure 9 —Cross-plot of predicted versus measured density of oil based drilling fluid (Mineral B)

ESD calculation model application


The ESD calculation model has been applied in several wells Tarim basin in China while drilling. Take
the K_C well as an example, K_C well is an exploratory well and located in the C block of Keshen
structural in Tarim Basin. A oil-based mud was selected while drilling the 6885-7070m well section, the
drilling fluid density was 1.93g/cm3 and the ratio of O/W was between 85/15-90/10. Based on the
information of the drilling fluid formula data, we calculate the volume fractions of oil and water, the oil
phase volume fraction was about 57.4% and the water phase volume fraction is about 8.95%. As
mentioned before, the well hole temperature profile were calculated using the model which proposed by
Raymond (1969). Based on the PSO-BP ANN drilling fluid density calculation model and the proposed
ESD calculation model, we calculated the drilling density profile and ESD profile in the well hole, which
is shown in the Fig. 10.
12 IADC/SPE-180573-MS

Figure 10 —Drilling fluid density and ESD profile of K_C well

From the Fig. 10 we can see that the drilling fluid density at the surface is 1.93g/cm3, which would be
compressed in the well hole and the density reached to 1.98g/cm3 at the bottom hole. At the same time
the ESD at the bottom hole is about 1.947g/cm3, which was also greater than the drilling fluid density at
the surface.
In order to protect the reservoir, OBM was selected for drilling most of the well’s target formations in
block C, the densities of OBM were between 1.80g/cm3-2.58g/cm3. Therefore, the ESD delta and pressure
delta caused by the drilling fluids variation were shown in Fig. 11. From the Fig. 11 we can see that the
downhole ESD is higher than the drilling fluid density which measured at the ground. The delta ESD
increased with the drilling fluid density increasing while the drilling density less than 2.3g/cm3. With the
drilling densities keep increasing which were bigger than 2.3g/cm3, the compressible component such as
water and oil phase of drilling fluid declining quickly. Therefore, the delta ESD of whose density bigger
than 2.3g/cm3 had a declining trend.

Figure 11—Delta ESD and delta BHP of caused by the drilling density variation of different drilling fluid
IADC/SPE-180573-MS 13

Conclusion
HTHP drilling fluid density and ESD calculation is very important for precisely calculating and
controlling the downhole pressure. Based on the artificial neural network optimization algorithm,
combined with the traditional components HTHP density prediction model, a utility PSO-BP ANN HTHP
density calculation model was established, in which the drilling fluid components such as the water phase
and oil phase volume fraction. And the PSO-BP ANN parameters has been proposed based on nearly 120
sets HTHP drilling fluid density data, through which the HTHP drilling density can be calculated easily.
Moreover, the ESD calculation was described too, through which we can obtained the ESD profile of the
well easily. Then the utility HTHP drilling fluid density model which predicts the HTHP density of
different type drilling fluid has been presented. The excellent application results show that the model can
precisely calculate HTHP drilling fluid density and ESD while drilling.

Nomenclature
fo ⫽ oil phase volume fraction, dimensionless
fw ⫽ water phase volume fraction, dimensionless
k ⫽ the number of iterations, dimensionless
l ⫽ the number of output layer neuron, dimensionless
m ⫽ the number of hidden layer neuron, dimensionless
n ⫽ the number of input layer neuron, dimensionless
P ⫽ pressure, MPa
P0 ⫽ initial pressure, MPa
⌬P ⫽ pressure difference, MPa
T ⫽ temperature, °C
T0 ⫽ initial temperature, °C
⌬T ⫽ temperature difference, °C
vjl ⫽ the weight values between hidden layer and output layer, dimensionless
wij ⫽ the weight values between input and hidden layer, dimensionless
WHP ⫽ well head pressure, MPa
yj ⫽ the input data of No.j neurons in hidden layer, dimensionless
␪j ⫽ the threshold value of hidden layer, dimensionless
␪l ⫽ the threshold value of output layer, dimensionless
␳0 ⫽ the initial drilling fluid density that measured at P0 and T0, g/cm3
␳(P,T) ⫽ the drilling fluid density that calculated at P and T, g/cm3

Acknowledgments
The paper is funded by the Research project of SINOPEC Science Technology Ministry (No.P12075 and
No.P15173) and ⬙China Post doctoral international exchange program⬙.

References
Eberhart R. C., Kennedy J. A. A new optimizer using particle swarm theory. Proceedings of 6th International Symposium
on Micro Machine and Human Science, Nagoya, Japan: IEEE, 1995.
Gharbi R., Elsharkawy A. M. Universal neural network based model for estimating the PVT properties of crude oil
systems. Paper SPE 38099 was presented at SPE Asia Pacific Oil and Gas Conference and Exhibition, Kuala Lumpur,
Malaysia, 14-16 April, 1997.
Guan Z. C. 2003. Effect of temperature and pressure on fluid column pressure of wellbore drilling fluid in deep water
drilling with oil based drilling fluid. Journal of the China university of petroleum: Edition of Natural Science, 27 (4):
48 –52.
14 IADC/SPE-180573-MS

Hoberock L. L., Thomas D. C., Nickens H. V. Bottom-hole mud pressure variations due to compressibility and pressure
variations due to compressibility and temperature effects. Paper was presented at 1982 Drilling Technology Confer-
ence of the International Association of Drilling Contractors, 9-11 March, 1982.
Karstad E., Aadnoy S. Density Behavior of drilling fluids during high pressure high temperature drilling operations. Paper
SPE/IADC 47806 was presented at the IADC/SPE Asia Pacific Drilling Technology Conference, Jakarta, Indonesia,
7-9 September, 1998.
Kemp N. P., Thomas D. C. Density modeling for pure and mixed-salt brines as a function of composition, temperature,
and pressure. Paper SPE/IADC 16079 was presented at the SPE/IADC Drilling Conference, New Orleans, LA., 15-18
March, 1987.
McMordie W. C., Bland R. G., Hauser J. M. Effect of temperature and pressure on the Density of drilling fluids. Paper
SPE 11114 was presented at the 57th Annual Fall Technical Conference and Exhibition of the Society of Petroleum
Engineers of AIME, New Orleans, LA., 26-29 September, 1982.
Methven N. E., Baumann R. Performance of oil muds at high temperatures. Paper SPE 3743 was presented at the
SPE-European Spring Meeting of Society of Petroleum Engineers of AIME, Amsterdam, Netherlands, 16-18 May,
1972.
Osman E. A., Abdel-Wahhab O. A., Al-Marhoun M. A. Prediction of oil PVT properties using neural networks. Paper SPE
68233 was presented at SPE Middle East Oil Show and Conference, Bahrain, 17-20 March, 2001.
Osman E. A., Aggour M. A. Determination of drilling mud density change with pressure and temperature made simple and
accurate by ANN. Paper SPE 81422 was presented at the 13th Middle East Oil show & Conference, Bahrain, 5-8
April, 2003.
Peters E. J., Chenevert M. E., Zhang C. H. A model for predicting the density of oil-Based muds at high pressures and
temperatures. Paper SPE 18036 was presented at the SPE Annual Technical Conference and Exhibition, Houston, 2-5
October, 1988.
Thomas D. C., Atkinson G., Atkinson B. L. Pressure and temperature effects on Brine Completion Fluid Density. Paper
SPE 12489 was presented at the Formation Damage Control Symposium, Bakersfield, CA, 13-14 February, 1984.
Wang G., Pu X. L., Luo X. S. 2008. Research on density of water-base drilling fluid at HTHP. Oil drilling & production
technology, 30 (3) 38 –40.
Wang H. G., Hao M. H., Yang L. P. 2000. The P-␳-T (Pressure-density -temperature) behavior of HTHP drilling fluid and
its effect on wellbore pressure calculation. Oil drilling & production technology, 22 (1): 17–21.
Wang H. G., Liu Y. S., Yang L. P. 2000. Effect of temperature and pressure on drilling fluid density in HTHP wells.
Drilling & production technology, 23 (1): 56 –60.
Yan J. N., Li Y. 1992. Mathematical model of prediction mud density under high temperature and high pressure. Oil
drilling & production technology, 12 (5): 27–34.
Yang X. W., Zhou Y. C., Fang S. L., Wang K. 2010. Strategy analysis of narrow window drilling technology and practice.
Oil Field Equipment, 39 (8): 7–11.
Zhang J. B., Yan J. N. 2006. The construction of a new model prediction HTHP mud densities. Drilling fluid & completion
fluid, 23 (5): 1–3.
Zhou H. B. 2014. Theoretical Model Research on Improving the Bottom hole Pressure Calculation Accuracy of HTHP
Wells. PhD. Dissertation of Oil & Gas Well Engineering. China: China University of Petroleum, Beijing.

S-ar putea să vă placă și