J Chen et al
2
yield and hence in improving the oil recovery ratio and saving
energy. In underground oil wells, as it is hard to separate oil
from water and measure their ow rates independently, the
total ow rate and water cut can be measured rst and then
the water and oil ow rates can be obtained. The water cut is
dened as the water ow rate divided by the total ow rate.
The total ow rate is generally measured by a turbine ow-
meter, due to the small difference between the densities of
oil and water. However, accurate measurement of water cut
is still a challenging task owing to complex ow regimes and
slippage between oil and water [4, 5].
Measurement of the water cut of multi-phase ows has
been widely studied for decades and a number of techniques
has been proposed such as rapidly closing valves [6], differ-
ential pressure, ultrasound, thermal sensors, electrical sensors
(probe [7–10], electrode [11–14], ring [15] and plate [16]),
microwave, radioactive rays [17], optical probes, etc. Rapidly
closing valves and radioactive ray-based techniques accurately
measure water cut. However, the rapidly closing valve is not
suitable for on-line measurement, and radioactive techniques
are complex, costly, and harmful to human health. Low-cost,
reliable, and safe on-line methods for water cut measurement
are still required.
A conductance probe is a kind of electrical sensor for the
measurement of water cut in multi-phase ows based on the
principle that the conductance of water is much greater than
that of oil. The conductance probe was proposed to measure
local void properties in gas–liquid ow by Neal et al in the early
1960s [7]. A dual-sensor probe was employed to measure water
cut in oil–water dispersed ow by Zhao et al [8], and in vertical
bubbly oil-in-water ows by Lucas and Panagiotopoulos [9].
A four-sensor conductance probe was developed to measure
water cut in vertical air–water ow by Revankar and Ishii [10].
The research outcomes validate that it is feasible to use the con-
ductance probe to measure water cut. However, the conduct-
ance probe method is far from mature. Treatment and usage of
response signals of the conductance probe still need deep stud-
ying to further improve the accuracy of water cut measurement.
Combining soft-measuring techniques with conventional
multi-phase ow sensors can greatly enrich the usage of
data that are measured from multi-phase ows, and hence
ultimately increase measurement accuracy. In general, a
soft-measuring technique includes the following steps: data
mining, feature extraction, data fusion, and parameter esti-
mation, etc. Nonlinear time series analysis was employed
to characterize oil–water ow patterns in combination with
four ring-type stainless-steel electrodes by Jin et al [15]. An
extended Kalman lter was employed to estimate ow rates
in combination with pressure gauges by Gryzlov et al [18].
The research results verify the feasibility and effectiveness of
soft-measuring techniques in multi-phase ow measurement
by using conventional sensors.
In a vertical well, oil and water distributions on a concen-
tric circle are statistically symmetrical. The response signal
of a conductance probe placed in the well can reect oil and
water distributions on the concentric circle where the probe
was placed, but only reect oil and water distributions on the
whole cross-section of the vertical well to some extent. Oil
and water distributions are dependent on the total ow rate
and water cut. If the total ow rate is added as a variable to
help characterize the oil and water distributions, the accu-
racy of water cut measurement is expected to be improved.
Therefore, in this paper, a conductance probe-based well log-
ging instrument was developed, and the total ow rate and
the response of the conductance probe are combined together
to estimate the water cut of the oil–water ow in a vertical
well. Apart from direct usage of the raw features extracted
from the probe response, pretreatment techniques including
principal component analysis (PCA) and partial least squares
analysis (PLSA) were employed to reduce data redundancy
within the extracted features. Models mapping the total ow
rate and the features extracted from the probe response onto
the water cut were established by using linear least squares
regression (LSR) and nonlinear support vector regression
(SVR), respectively.
2. Methodology
2.1. Well logging instrument and overview
A conductance probe-based well logging instrument was
developed, as illustrated in gure1. A retractable probe arm
is installed on the instrument shaft. It can be opened and
retracted by a motor driver. The conductance probe is installed
on the probe arm, and its tip is positioned towards the coming
ow. The diameter of the circle where the probe is located is
96 mm. The conductance probe comprises a stainless needle,
an insulation layer, and a stainless shell. The length of the
needle tip is 3 mm. The outer diameter of the stainless shell
is 3 mm, and the shell is grounded. The insulation layer is
used to separate the tip from the shell. When the probe arm
is opened, the conductance probe can record the time-varying
electrical characteristics of the oil–water ow by a conduc-
tance telemetry circuit. As the resistance of oil is much higher
than that of water, the conductance probe can distinguish con-
tinuous phases of oil and water, and distinguish oil and water
bubbles whose diameters are larger than 3 mm. The response
data of the conductance probe are transmitted to the ground
via Manchester code through a logging cable.
Figure 1. Conductance probe-based well logging instrument.
Cable connector
Circuit
Probe arm in open state
Motor
Conductance probe
Insulation layer
Needle tip
Stainless shell
Stainless needle
Meas. Sci. Technol. 26 (2015) 095306