Page 116 - Intelligent Digital Oil And Gas Fields

84 Intelligent Digital Oil and Gas Fields

3.2.2 Advanced Validation Techniques
The basic validation described above is meant to keep bad data (regardless of
cause) from corrupting DOF databases. Advanced data validation is meant to
detect particular problems with data. The problems could be bad meters,
poor operation of surface facilities, abnormal operating conditions, etc.
These techniques make considerable use of fundamental production attri-
butes, statistics, and model-based methods.
After basic single-variable validation, it is common to look at multivar-
iable calculations with respect to fundamental laws of production. As the use
of flow rate meters is increasing dramatically in many assets, it is crucial that
these meters are calibrated for precision and accuracy. Just because they read
a number and are not frozen does not mean the results are good. Operators
spend many hours chasing suspect meter data. Many automated methods can
be used to check meter accuracy as well. For example, mass balances are
commonly used in calculations as checks. In addition, volume or flow ratios,
like gas–oil ratio or water cuts, should not shift drastically in a short time. If
they do, it is likely a metering issue. A metering issue could be due to the
meter itself needing calibration or it could be due to the separation equip-
ment not operating correctly; for example, water gets carried over with the
oil channel and is measured as oil. Again, these checks determine that a prob-
lem exists and approximately where it is but may not directly identify the
specific cause. The ratios described here necessarily use data for the well sta-
tus required. For example, these tests should not use data when the well is
down or not at steady state.
Another common practice is to use SPC practices. There are two rea-
sons to perform these checks: to detect a shift in (a) the process itself or
(b) bad data. Although it is beyond the scope of this book to present
SPC, many texts explain it. The basic rules make use of a long-term aver-
age or mean and the signal standard deviation. For example, you can cal-
culate a 7- to 30-day moving average and standard deviation. Note that
these calculations should use the status signals described above (the average
and standard deviation should only use data from when the well is up or in
steady state). The rules for abnormal process behavior are applied under
these conditions:
• if the current data (daily average) is beyond the average 3 standard
deviations;
• if the last 2 daily averages are beyond the long-term average 2 standard
deviations;