Sie sind auf Seite 1von 14

SPE 107521

Analyzing Transient Pressure From Permanent Downhole Gauges (PDG) Using


Wavelet Method
Zheng Shi-Yi, SPE, and Li Xiao-Gang, Heriot-Watt U.
Copyright 2007, Society of Petroleum Engineers
This paper was prepared for presentation at the SPE Europec/EAGE Annual Conference and
Exhibition held in London, United Kingdom, 1114 June 2007.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of 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 where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435.

Abstract
Reservoir pressure monitoring during its production life is to
evaluate its performance to ensure the effective extraction of
hydrocarbon from the reservoir. Continuous reservoir
monitoring offers a window of prospects to increase well
productivity, while reducing operating costs through an
improved and more accurate well performance. These are
achieved through analyzing the long term, real time, dynamic
transient pressure obtained from Permanent Down-hole
Gauges (PDG), which will provide input and vital information
for improving the existing reservoir model for flow
simulation.
Reservoir monitoring during field development can amplify
the
understanding
of
the
reservoir
depletion,
compartmentalization, and efficiency of water injection, also
the presence of any flow barriers (such as activated fault). On
this basis, reservoir management can be more accurate and
realistic.
In contrast to the data obtained from the traditional well test
such as DST, pressure data from PDG is large in quantity
(long term up to six years), noisy, multiple phase flow and
multiple rate. The key for the analysis of such data set to
extract reservoir information is to develop a technique in
handling the data, i.e. to make the data analyzable but not
losing the true information contained in the data.
This paper presents techniques using wavelet transform
method for such purpose. These have been developed recently
through an industry sponsored research Project. The
procedures include PDG data visualization, de-noising,
smoothing and break point (due to drop of rate and well
shut-in) identification through analyzing field examples
provided by the sponsors. After this process, transient pressure

data (extracted drawn-down and build up) can then be


analyzed using current traditional methods and software. The
analysis results such as skin factor and reservoir effective
permeability can be evaluated in a time lapse fashion to reveal
information about the dynamic changes of the reservoir.
Introduction
With the increased reliability and technology improvement,
more and more Permanent Down hole Gauges (PDG) have
been installed in oilfields around the world, especially in more
challenging environments such as the deepwater reservoirs
offshore. With the innovative and more robust PDG design,
this has gradually replaced the conventional production tests
and well tests to provide engineers with a cost effective means
for long term, real time reservoir monitoring and management
(Rossi et al., 2000; Chiriti et al., 2001; Ballinas et al., 2002;
Mason et al.; Weaver et al., 2005).
The published leading research in handling PDG data is from
Chevron and Stanford University (Kikani and He, 1998). By
examine simulated and field examples, the wavelet method
has been recognized as a useful tool in dealing with the long
term transient pressure data from PDG. Athichanagorn and
Horne (1999) from Stanford developed a multi-step procedure
for the processing and interpretation of long-term pressure
data through the analysis of both simulated and actual field
data. Their study found that the use of long-term data requires
special handling and interpretation techniques due to the
instability of in-situ permanent data acquisition systems,
extremely large volume of data, incomplete flow rate history
caused by unmeasured and uncertain rate changes, and
dynamic changes in reservoir conditions and properties
throughout the life of the reservoir.
Soliman et al. (2001) demonstrated the importance of using
wavelet transform for well test analysis through field
examples. Since this method is capable not only to smooth the
basic signal, but also has the ability of retention or even
enhancement of the details of the signal, while the logarithmic
pressure derivative diagnostic techniques currently used in
well testing tend to smooth data and conceal certain events,
the use of the wavelet transform to analyze the raw data could
prove to be very valuable as a major step in enhancing the
techniques of modern well test analysis.
Ouyang and Kikani (2002) proposed to apply nonlinear
regression via Polytope method for best-fitting PDG data to

determine the noise level. It is found that the new approach is


superior to the least square error (LSE) linear regression as
used by Khong1, because the bottom-hole well bore pressure
response in a well should be treated as a nonlinear function of
time over majority of the well production/injection/shut-in
period.
Guan et al. (2004) reviewed the application of wavelet method
in the oil industry. A wavelet is a waveform of effectively
limited duration that has an average value of zero and
wavelets are a family of basis functions, which can separate a
signal into distinct frequency packets that are localized in the
time domain. Thus, wavelets are well suited to analyze nonstationary data. They can smooth the basic signals and keep
the details of basic signals. Therefore, they provide a multiresolution framework for data representation.
Olsen and Nordtvedt (2005) conducted a comprehensive
analysis examining a series of possible wavelets for PDG data,
and demonstrate that spline algorithm has excellent properties
for cleansing of pressure and rate data.
The applications of the data obtained from PDG include
pressure monitoring, interference testing, water and gas
injection monitoring, transient well testing, evaluation of well
performance, reducing the flowback time of new wells,
monitoring hydraulic fracturing operations, and monitoring
pump inlet and outlet pressures for pumping wells (Haddad et
al., 2004; Frota and Destro, 2006; Chorneyko, 2006).
However, the high density data acquisition at an interval down
to one second over several years with changes in production
rate and reservoir pressure plus noise has imposed a great
challenge for data analysis and interpretation. Companies are
faced with stewardship of the investment, communicating
learnings across the organization, developing or acquiring
software tools for data management and analysis, proposing
and supporting industry-wide standards for consistent data
capture and communication, and integrating data into work
processes and into applications such as smart wells and smart
fields (Chorneyko, 2006).

SPE 107521

This paper presents study results on the analysis of transient


pressure from PDG using software developed within the
PRIME Project, called PRIME Tool Box. The paper will be
focused on long term, large quantity and noisy PDG data
processing using selected wavelet method coded in the Tool
Box. Then pressure draw downs and build ups distilled from
the software are analyzed using the traditional approach in a
time lapsed fashion.
This paper presents study results on the analysis of transient
pressure from PDG using software developed within the
PRIME Project, called PRIME Tool Box. The paper will be
focused on long term, large quantity and noisy PDG data
processing using selected wavelet method coded in the Tool
Box. Then pressure draw downs and build ups distilled from
the software are analyzed using the traditional approach in a
time lapsed fashion.
PDG data characteristics and analysis workflow
There are three main characteristics for PDG data: a) noisy
due to rate fluctuations during production, b) large quantity
due to high density sampling, and c) long term. As shown in
Fig.1, a six years transient pressure recorded by a permanent
down hole gauge. It is difficult, even to visualize this kind of
data without suitable computer software, let alone distilling
information out of it.
There are two approaches developed in the PRIME Tool Box
for PDG data analysis: a) Traditional Approach, which
includes data visualization, outlier removal, denoising,
smoothing, data compression, DD and BU evaluation and
analysis using traditional well test analysis methods. b)
Advanced Approach, which will analyze transient pressure
data as a continuous signal or reservoir system response. This
includes recovering flow rate history from measured transient
pressure, allocating the production rate, numerical well testing
assisted by neural network modeling and simulation, guided
pressure history matching. These together can achieve the
objectives of real time reservoir management through model
calibration and future reservoir performance and production
forecasting. This work flow is shown in Fig.2.

Under the current oil industry situations, where 95% of


hydrocarbon reserves were found in the world and 90% of
these have been under development. This means that the
future of the oil industry will depend on how efficient
engineers can manage the mature field to maximize the
ultimate recovery to satisfy the energy need from our society
in the next 30 years and beyond. Real time reservoir
monitoring and testing through the reliable permanent downhole gauges will provide the key information to achieve this
objective.

In this paper, only the Traditional Approach will be presented


due to the time limit.

Following this trend, Petroleum ReservoIr Monitoring and


tEsting (PRIME) project was launched in year 2004 at HeriotWatt University, with the support from ConocoPhillips, BG
group and Wintershall. One key research area of the project is
focused on developing techniques for PDG data analysis and
interpretation.

The HAAR wavelet was found to be capable of identifying the


event in PDG data because the transient pressure signal is step
change, particularly for draw downs and build ups. This is
used to decompose the original data into signals at different
frequencies (higher or lower). The values of the high
frequency signals (positive or negative) are used to detect
event.

PDG data processing and evaluation


No matter what approaches will be used for well test analysis,
the PDG data needs to be loaded and visualized for evaluation
before it can be analyzed. This requires a standard data format,
database and computer programme to handle the data set. In
PRIME Tool Box, HAAR wavelet was chosen for data
filtering, processing.

SPE 107521

As shown in Fig.3 and Fig.4 respectively, data sets from two


filed examples are selected to demonstrate the capability of the
developed software: a) to demonstrate the capability of the
developed software on processing long term transient pressure
data, b) to demonstrate the power of the developed software
on data processing such as compression and signal high
frequency analysis.
Flow events detection through wavelet filtering
Since the PDG data is long term, this can only be processed on
a period by period basis or time window basis. This process
will be carried out forward from the beginning until the time
required.
Fig.5 shows the decomposition analysis of the filed example
B. where S is the original signal, a3 is low frequency signal,
while d3, d2, and d1 are frequencies at three levels (lower to
higher). Fig.6 shows that the HAAR wavelet was used for
processing the PDG data from the field example A. It shows
the decomposed signals after the wavelet treatment, together
with the original PDG data, where S is the original signal, a3
is low frequency signal, while d3, d2, and d1 are frequencies
at three levels (lower to higher).
Fig.7 shows the original PDG data from field example A,
together with the chosen high frequency component at a
proper level after processing. The blue line is the original
transient pressure and, the red line is the high frequency of the
original transient pressure data at the chosen level.
In the process of wavelet filtering, the value of threshold
controls the efficiency and quality of data denoising,
smoothing and event detection. The algorithm for selecting the
threshold needs lots of practice and experiment. The PRIME
Tool Box has designed with well organized threshold
definition according to the PDG data resolution and sampling
density. Using this software, the event can be identified
accurately and the threshold for processing the data can be
selected automatically.
Quality DD and BU identification
The developed software can identify any changes in transient
pressure including DDs and BUs. However, for well test
analysis, only those BUs, which were due to well shut in with
a proceeding stable flowing history having a duration at about
half of that of BU. Based on the wavelet frequency analysis,
this can be resolved by no mistake.
By examine the high frequency; any events during the
production history can be detected as a sharp change. The
value of the high frequency at the point, where a sharp change
occurred is used to distinguish different types of events, such
as pressure draw downs and build ups. When the value of the
frequency is positive, the pressure DD happens, while the
values are negative, the pressure BU happens.
Fig.8 shows the DDs and BUs identified using this method
from field example A (part of the data set). Fig.9 shows
outlier, the DDs and BUs due to drop of rate as well as well
shut-in identified using HAAR wavelet processing method for

field example B, where the contrast between frequencies due


to rate changes and those during DDs and BUs makes it very
easy to distinguish the events.
PDG data compression
Another important development in PRIME Tool Box is its
capability for PDG data compression or re-sampling. This is
essential for PDG data processing in all the processes from
data visualization until event detection. Very high density data
sampling frequency in an interval down to one point per
second will result in hundreds of data points in short term such
as the example included in this paper from Field B. This large
quantity of data is not only slow down the analysis process,
but is also not necessary. Some time this will cause problems
such as stuck the computer in the process of the analysis.
So the data needs to be re-sampled or compressed. However,
there is a risk in losing the true information by doing so. An
algorithm, which can, not only compress the data set
significantly, but can also ensure not losing any information or
true response of the data, was developed and coded into the
PRIME Tool Box.
Fig.10 shows the power of the algorithm used in PRIME Tool
Box for PDG data compression, demonstrated by field
example B. The original signal has 111,702 points. The data
points after the compression are only 1033, while the true
response of the original data retained.
The compression coefficient for this example is 100. The used
algorithm also considered the non-uniform sampling
characteristics of the transient pressure data from the PDG.
PDG data analysis after processing
After the wavelet filtering and processing, quality pressure
DDs and Bus can be exported, then analyzed using well test
analysis software such as PanSystem.
However, further data smooth is required before the analysis
can proceed for the reasons below: a) the last flowing pressure
or the first pressure point of the BU need to be made right, b)
the compressed BU is still too noisy, which will be no good
for pressure derivative diagnostic.
A new data smooth function combining averaging and wavelet
methods was developed and coded into the software. Fig.11
shows three BU data sets after wavelet processing from field
example B: a) original data before compression (solid blue
dot), b) compressed signal (red circle) and, c) further
smoothed signal using the developed smooth function (dotted
green diamond). Apparently further smoothed signal can best
represent the measured transient pressure BU.
In order to prove the developed smooth function works,
original pressure BU data before the wavelet processing and
compression is further plotted in the logarithmic pressure
derivative diagnostic plot to check the quality of the data.
As shown in Fig.12, the pressure derivative log-log plot of the
original BU data from Fig.11, before wavelet processing and

SPE 107521

data compression, from which one can hardly identify any


flow regimes from the calculated derivatives, but just see a
heap of the crowded data points with no trend.
In contrast, Fig.13 shows the log-log diagnostic plot of the BU
from Fig.12, after wavelet processing, compression and
further smoothing, from which the well bore storage, reservoir
radial flow, possible some outer boundary effects are all
distinguishable. This further proved that the developed data
smoothing function used in PRIME Tool Box is working well.
After being through the data wavelet processing, compression,
and further smoothing for the selected quality DDs and BUs,
traditional well test analysis procedures can be conducted to
derive the reservoir effective permeabilities and skin factors in
a time lapsed fashion. These are then plotted against the
elapsed time to reflect the dynamic changes of these properties
during the selected time window.
Field examples B and A were analyzed in this way using the
developed Tool Box and PanSystem. The analysis results are
shown in Fig.14 and Fig.15 respectively.
Conclusions
A PC based computer software, PRIME Tool Box, using
wavelet method on handling PDG data was developed through
PRIME Project. Primary testing using field examples proved
its capability in handling long term and high sampling density
PDG data through HAAR wavelet signal frequency
processing, data compression and further smoothing. In
summary, these include:

PDG data processing and evaluation


Flow events detection through wavelet filtering
Quality DD and BU identification
PDG data compression
PDG data analysis after processing, including further
smoothing

In this paper, focuses are on the PDG data handling and


processing using wavelet method. Only functions using
traditional approach on well test analysis was presented.
Advanced approach, which is also available in the PRIME
Tool Box, will be presented on another conference in the near
future.
Further development of the PRIME Tool Box will include
capabilities on handling PDG data due to multi-phase flow,
well production and reservoir interferences as well as long
term production data handling and optimization using method
such as de-convolution and modified decline curve analysis.
These together will bridge the gap between long term PDG
data and reservoir flow simulation model. Through using
PRIME Tool Box, PDG data can be processed, analyzed to
provide input to the simulation model in real time. So the
ultimate goal of reservoir monitoring, testing through
permanent down hole gauges can be achieved, i.e. real time
reservoir management.

Acknowledgement
Authors would like to take this opportunity to thank the
PRIME Project (Petroleum ReservoIr Monitoring and tEsting PRIME Project) sponsors: BG Group, ConocoPhillips and
Wintershall AG for their financial support.
Schlumberger and EPS, Weatherford is acknowledged for
providing softwares to conduct studies within the Project
including this study.
References
Ballinas, J., and Owen, J., Evaluation and Control of Drilling,
Completion and Workover Events with Permanent
Downhole Monitoring: Applications to Maximise Production
and Optimise Reservoir Management paper SPE 74395
presented at the 2002 SPE International Petroleum
Conference and Exhibition of Mexico, Villahermosa, Mexico
10-12th February 2002.
Chiriti, K.E. (PDVSA), Moffatt, T.G. (PROMORE
Engineering Inc) and Bussiere, C. (PROMORE Engineering
Inc), Permanent Downhole Monitoring for Extreme
Temperature and Pressure Environments paper SPE 71593
presented at the 2001 SPE Annual Technical Conference and
Exhibition, New Orleans, Louisiana, 30th September 3rd
October 2001.
Chorneyko D. M., ExxonMobil Upstream Research Co.,
Real-Time Reservoir Surveillance Utilizing Permanent
Downhole Pressures - An Operator's Experience, paper
SPE103213 presented at the SPE Annual Technical
Conference and Exhibition, 24-27 September, 2006, San
Antonio, Texas, USA
Frota H.M., SPE, and Destro W., Petrobras, Reliability
Evolution of Permanent Downhole Gauges for Campos
Basin Subsea Wells: A 10-Year Case Study, paper
SPE102700 presented at the SPE Annual Technical
Conference and Exhibition, 24-27 September, 2006, San
Antonio, Texas, USA.
Guan L., L.Li.U.: "Wavelets in petroleum Industry: Past,
Present and Future", Paper SPE 89952.
Haddad, S. (Schlumberger), Proano, E. (Schlumberger) and
Patel, Y. (BP), A Method to Diagnose Depletion, Skin , kh
and Drive Mechanism Effects Using Reservoir Monitoring
Data paper SPE 90032 presented at the 2004 SPE Annual
Technical Conference and Exhibition, Houston, Texas,
U.S.A., 26-29th September 2004.
Jitendra Kikani, "Multi-resolution Analysis of Long-term
Pressure Transient Data Using Wavelet Methods",
SPE48966.
Mason, J. (Armeda Hess), Laidlaw, J., and Pai, S.
(Schlumberger), Downhole Communications: North Sea
Platform proves Permanent Monitoring Reliability Journal,
Oil and Gas Journal.
Ouyang Liang-Biao, "Improving Permanent Downhole Gauge
(PDG) Data Processing via Wavelet Analysis", SPE 78290

SPE 107521

Rossi, D.J., Gurpinar, O., Nelson, R., Jacobsen, S.,


Discussion on Integrating Monitoring Data into Reservoir
Management Process paper SPE 65150 presented at the
2000 SPE European Petroleum Conference, Paris, France,
24-25th October 2000.
Soliman, M.Y., Ansah, and Manda, B.: "Application of
Wavelet Transform to Analysis of Pressure Tests", paper
SPE 71571.
Suwat Athichanagorn, Roland N.Horne,: "Processing and
Interpretation of Long-term Data from Permanent Downhole
Pressure Gauges", SPE56419,1999 SPE Annual Technical
Conference and Exhibition, 3-6 October, Houston, Texas
Weaver, M. (Andarko Petroleum), Kragas, T. (Weatherford
Intl. Inc), Burman, J. (Exploitation Technologies), Copeland,
D. (Andarko Petroleum), Phillips, B. (Weatherford Intl. Inc)
and Seagraves, R. (Intl. Lift Systems), Installation and
Application
of
Permanent
Downhole
Optical
Pressure/Temperature Gauges and Distributed Temperature
Sensing in Producing Deepwater Wells at Marco Polo paper
SPE 95798 prepared for presentation at the 2005 SPE Annual
Technical Conference and Exhibition, Dallas, Texas, U.S.A.,
9-12th October 2005.

SPE 107521

Fig.1 A six years transient pressure recorded by a Permanent Down-hole Gauge (PDG). Three main characteristics can be observed: a)
noisy due to rate fluctuations during production, b) large quantity due to high density sampling, and c) long term.

Fig.2 There are two approaches developed in the PRIME Tool Box for PDG data analysis: a) Traditional Approach and, b)
Advanced Approach.

SPE 107521

Fig.3 Field Example A: One month Permanent Down hole Gauge (PDG) data set (about 30,000 data points) from an oilfield in the
North Sea in Cartesian plot, where the vertical axis shows recorded pressure, while the horizontal axis shows the elapsed time. A
series of transient pressure draw downs and build ups are shown as continuous changes due to flowing rate changing or well shut-in.
This data set was chosen to demonstrate the capability of the developed software on processing long term transient pressure data.

Fig.4 Field Example B: This example data set shows the distinct differences from the one shown in Fig.1. The data duration is about
31 hours, which contains about 100,000 data points due to high density sampling from the PDG (down to one data point per second).
This data set was chosen to demonstrate the power of the developed software on data processing such as compression and signal high
frequency analysis.

SPE 107521

Fig.5 The HAAR wavelet was found to be capable of identifying the event in PDG data because the transient pressure signal is step
change, particularly for draw downs and build ups. This is used to decompose the original data into signals at different frequencies
(higher or lower). The values of the high frequency signals (positive or negative) are used to detect event. The figure shows the
decomposition analysis of the filed example B, where S is the original signal, a3 is low frequency signal, while d3, d2, and d1 are
frequencies at three levels (lower to higher).

SPE 107521

Fig.6 The HAAR wavelet was used for processing the PDG data from the field example A. This figure shows the decomposed signals
after the wavelet treatment, together with the original PDG data, where S is the original signal, a3 is low frequency signal, while d3, d2,
and d1 are frequencies at three levels (lower to higher).

10

SPE 107521

Fig.7 This figure shows the original PDG data from field example A, together with the chosen high frequency component at a proper
level after processing. The blue line is the original transient pressure and, the red line is the high frequency of the original transient
pressure data at the chosen level.

Fig.8 By examine the high frequency, any events during the production history can be detected as a sharp change. The value of the
high frequency at the point, where a sharp change occurred is used to distinguish different types of events, such as pressure draw
downs and build ups. When the value of high frequency is positive, the pressure DD happens, while the values are negative, the
pressure BU happens. This figure shows the DDs and BUs identified using this method from field example A (part of the data set).

SPE 107521

11

Fig.9 This figure shows outlier, the DDs and BUs due to drop of rate as well as well shut-in identified using HAAR wavelet
processing method for field example B, where the contrast between frequencies due to rate changes and those during DDs and BUs
makes it very easy to distinguish the events.

Fig.10 This figure shows the power of the algorithm used in PRIME Tool Box for PDG data compression, demonstrated by field
example B. The original signal has 111,702 points. The data points after the compression are only 1033, while the true response of the
original data retained. The compression coefficient for this example is 100. The used algorithm also considered the nonuniform
sampling characteristics, particularly for transient pressure data.

12

SPE 107521

Fig.11 This figure shows three BU data sets after wavelet processing from field example B: a) original data before compression (solid
blue dot), b) compressed signal (red circle) and, c) further smoothed signal (dotted green diamond). Apparently further smoothed
signal can best represent the measured transient pressure BU.

Fig.12 shows the pressure derivative log-log plot of the original BU data from Fig.11, before wavelet processing and data
compression, from which one can hardly identify any flow regimes from the calculated derivatives, but just see a heap of the crowded
data points with no trend.

SPE 107521

13

Fig.13 shows the log-log diagnostic plot of the BU from Fig.12, after wavelet processing, compression and further smoothing, from
which the well bore storage, reservoir radial flow, possible some outer boundary effects are all distinguishable. This further proved
that the developed data smoothing function used in PRIME Tool Box is working well.

Fig.14 After the wavelet data processing, compression and further smoothing, quality pressure DDs and BUs can be exported, then
analyzed using PanSystem to derive reservoir effective permeabilities and skin factors in a time lapsed fashion. These are then plotted
against the elapsed time to reflect the dynamic changes of these properties during the selected time window (field example B).

14

SPE 107521

Fig.15 After the wavelet data processing, compression and further smoothing, quality pressure DDs and BUs can be exported, then
analyzed using PanSystem to derive reservoir effective permeabilities and skin factors in a time lapsed fashion. These are then plotted
against the elapsed time to reflect the dynamic changes of these properties during the selected time window (field example A).