One of the most common techniques engineers use to estimate a quantity is by comparing readings from multiple sensors, or comparing before and after with the same sensor.
It’s a simple calculation of subtraction, yet full of nuances in real-life applications.
Multiple Sensors
Let’s start with a familiar scenario. Imagine Facility A and Facility B, connected by a 50 km pipeline.
- Meter A measures oil exported from Facility A.
- Meter B, downstream, measures the combined oil from both A and B.
To know how much oil Facility B produces, we calculate by difference:
Facility B Production = Meter B - Meter A
Straightforward, right? But remember error propagation: small uncertainties in each meter can amplify when you subtract them. And that’s only part of the story.
Production is typically reported daily, say from 7 AM to 7 AM. But oil molecules don’t teleport 50 km from A to B. The fluid leaving A at 7 AM won’t reach B at the same moment.
So why don’t we time-shift the readings? Because we assume steady-state flow, a condition where pressure, temperature and fluid properties remain constant with time at any location. Under steady state, what enters at A equals what exits at B, and our “difference” over the same period holds valid.
However, during transient periods, such as start-ups or shutdowns, or changing PVT properties, that assumption collapses. The pipeline becomes a storage element, which can accumulate or deplete. What B measures no longer corresponds instantaneously to what A produced.
So it’s no surprise to any production engineers to observe when Facility B’s reported production suddenly spikes when A shuts down or drops when A starts up.
Don’t panic. It’s not a metering or well fault.
Single Sensor
Now, consider the single sensor case, the classic well test by difference. In a commingled production system, isolating a single well for testing can be difficult and costly if shutting all other wells. Instead, we perform a test by difference procedure:
- Measure total flow with the well on,
- Then shut in the well
- Measure again with the well off,
- The difference gives the well’s contribution.
If the test separator is some distance away, the engineer must carefully determine when the second test can validly begin. The system needs time to stabilise and that delay depends on pipeline geometry, flow rate and few other factors.
There’s another challenge, too: pressure change. When a large well is turned on or off, it disturbs the network. Backpressure on other wells changes, making your “difference” unreliable. Test a very small well, and the difference might be smaller than your meter’s uncertainty, which is a textbook case of error propagation.
Measuring Production Losses
Measurement by difference also underpins production loss (or deferral) tracking. When one or more wells go offline, we estimate the loss by comparing total production before and after.
In some operations, engineers have the luxury of wellhead flow meters, which seems like it should simplify things.
If the sum of the offline wells’ production equals the observed loss at the receiving facility, we’re good, right?
Not quite. The metered volumes rarely reconcile perfectly. Network pressure changes, transient effects, and metering noise all play their parts.
“More meters” and “more data” don’t automatically mean better accuracy.
Respect the Assumptions & Understand the Physics
I might sound like a broken record when I talk about measurement, but that’s because the lesson bears repeating.
Every measurement, every device, every technique carries assumptions grounded in physics. Some are explicit, others are hidden. But knowing them is what separates routine data collection from true engineering insight.
Understanding those assumptions lets you transcend operations, facilities, and basins.
So when things don’t add up, don’t rush to fix the numbers or the wells.
First, fix the understanding.
💡 This article continues from myprevious poston “Error Propagation”.