Page 109 - Intelligent Digital Oil And Gas Fields

78 Intelligent Digital Oil and Gas Fields

non-SCADA information like the producing type. A user can quickly look
through these pages every day for a few minutes and find problems quickly.
A special type of manual data check should occur when a well or equip-
ment is first brought into the system or needs some particular analysis. There
should be a special query available that lists all data for only that well/equip-
ment. Until a user has manually validated all of this data, the well stays in
quarantine and should not be added into the DOF system. The user who
does this analysis and makes this decision needs to be a field person with
domain expertise. Only when the data is confirmed to be accurate should
this well be added into the system.
The automated checks should not rely on poll times or communica-
tion checks. The best method is to look at data in ways that conform and
test against physical reality. For example, it would be very unlikely that a
route would have a rate totalizer that decreased during the day, or that
daily volumes should not change during the day. If some quantities fluc-
tuate, that might indicate either direct individual equipment values are
changing or that the numbers of wells in a route or other organization
are changing, which is not likely. In these situations, automated alerts
can be issued.
Once the data is found to be suspect either at an individual or aggregated
level, it needs to be flagged to the user or a ticket created for resolution. As
soon as one of the above data checks fails, the user interface should indicate
the issue, for example, stoplight-type indicators are common. As soon as the
data becomes suspect, a yellow color is displayed; after time has elapsed with-
out resolution, then the stoplight turns red. The time between yellow and
red is generally determined by the update time required by the fastest
workflow. For example, if that is a daily workflow, then it could be 4 to
8h between a yellow and a red alert.
Finally, each individual reading should be minimally conditioned at this
stage, which means two things: first, if the data fails the range or freeze check,
then the last good value is held, and second, a small filter is applied to remove
high-frequency noise. We recommend this be a digital implementation of an
exponential average filter and not a moving average.

3.1.1 Data Processing
Typical surface pressure, temperature, and flow rate sensors capture data
values at 1s or higher intervals. Data are transferred through the system using
an RTU or PLC to a SCADA environment, then on to a data historian