Intelligent completions typically include permanent downhole gauges (PDGs)
for continuous, real-time pressure and temperature monitoring. If applied
adequately, such new technologies should allow anticipation of oil production
and an increase of final recovery with respect to traditional completions. In
fact, pressure data collected from PDGs represent essential information for
understanding the dynamic behavior of the field and for reservoir surveillance.
The potential drawback is that the number of data collected by PDGs can grow
enormously, making it very difficult, if not impossible, to handle the entire
pressure history as it was recorded. As a consequence, it might often be
necessary to reduce the pressure measurements to a manageable size, though
without losing any potential information contained in the recorded data.
As reported extensively in the literature, long-term data might be subject
to different kinds of errors and noise and not be representative of the real
system response. Before the data can be used for interpretation purposes,
especially if pressure derivatives are to be calculated (for instance, in
well-test analysis), an adequate filtering process should be applied.
Multistep procedures based on the wavelet analysis were presented in the
literature for processing and interpreting long-term pressure data from PDGs.
In this paper, an improved approach largely based on the wavelet algorithms is
proposed and discussed for the treatment of pressure data.
All the steps of the procedure, namely outlier removal, denoising, transient
identification, and data reduction, were applied to both synthetic and real
pressure recordings. Results indicated that the application of the proposed
approach allows identification of the actual reservoir response and subsequent
interpretation of pressure data for an effective characterization of the
reservoir behavior, even from very disturbed signals.
Usually, the pressure data is acquired when production tests are performed,
during which the well should be produced at a constant rate to allow for
analytical interpretation. Because typical test durations range from a few
hours to a few days, pressure data are collected over short periods of time and
thus only allow the description of limited portions of the reservoir.
Permanent pressure monitoring represents a different and much more effective
approach for reservoir characterization and surveillance because both the
reservoir and well behavior are overseen continuously in real time by means of
PDGs (Baker et al. 1995). On the other hand, for a number of reasons (such as
workover, stimulation, and malfunction of the acquisition system), pressure
data collected by PDGs can contain extraneous pieces of information that are
not representative of the real dynamic behavior of the reservoir. Therefore,
the possibility to effectively use the collected pressure data hinges on the
application of an efficient treatment and interpretation process, so as to
capitalize on the information available for best exploiting the reservoir
The procedures proposed in the literature for processing and interpreting
long-term pressure data from PDGs are based on the wavelet analysis
(Athichanagorn et al. 1999; Khong 2001; Ouyang and Kikani 2002). The work
presented in this paper outlines an improved procedure for pressure-data
treatment and analysis. In chronological order, this procedure consists of
outlier removal, denoising, transient identification, and data reduction. The
main steps of the procedure, outlier removal and denoising, are still based on
the wavelet analysis, but the applied algorithms were selected on the basis of
a rigorous mathematical review (Mallat 1998; Goswami and Chan 1999). New
criteria were developed for the transient-identification process because the
method proposed in the literature, which was based on wavelet analysis, did not
seem to provide satisfactory results (Athichanagorn et al. 1999; Khong 2001;
Ouyang and Kikani 2002).
© 2007. Society of Petroleum Engineers
View full textPDF
- Original manuscript received:
21 July 2005
- Meeting paper published:
9 October 2005
- Revised manuscript received:
26 June 2006
- Manuscript approved:
8 April 2007
- Version of record:
20 August 2007