A Month of Meter Freezes in 10 Minutes: What AI Makes Possible in Winter Measurement Season
Three proof-of-value studies. More than 10,000 orifice meters each. One result: every meter freeze hour identified in under 10 minutes. Here's what that means for your measurement team.
Winter is the most expensive season in natural gas measurement. Not because of equipment failures (though those happen) but because of what winter does to orifice meters, and how long it takes most measurement teams to find out.
A meter freeze is not a sudden event. It develops over hours as cold weather affects the meter tube, impulse lines, or meter body itself. During a freeze, differential pressure readings collapse or spike in ways that don’t reflect actual flow. Volume calculations go wrong. Data that looks plausible in FlowCal is silently incorrect.
Finding those hours and correcting them is one of the most time-consuming jobs in measurement analysis. Until recently, it was also almost entirely manual.
What Manual Meter Freeze Review Actually Looks Like
Ask a measurement manager what their analysts do during a cold snap, and you’ll get some version of the same answer: they work the freeze list.
That means pulling meter data for every orifice meter in the area, reviewing it for the signature behavior patterns of a freeze event: the characteristic drop in differential pressure, the flow mode shift, the anomalous volume. Then flagging the hours that need correction, editing those hours in FlowCal, documenting the changes, and moving to the next meter.
Across a large pipeline system, this process consumes most of an analyst’s work week during active cold weather. On a very large system, it takes multiple analysts working simultaneously, often with overtime, trying to close the review window before month-end creates a billing or reporting deadline.
The work is important. The method is slow. And the risk of a high-probability freeze event getting missed or deprioritized because there aren’t enough hours in the day is real.
Three Proof-of-Value Studies. One Consistent Result.
Over the past several years, we’ve run proof-of-value studies on meter freeze identification with measurement teams operating across multiple basins and regions throughout the United States. Each study involved pipeline systems with more than 10,000 orifice meters, large enough to represent the full complexity of real-world measurement operations.
The question we set out to answer was simple: how long does it take NDL (Meter Freeze) to identify every freeze event across a full month of meter data?
Across all three studies, the answer was the same: under 10 minutes.
Not hours. Not a work week. Ten minutes for a complete identification of every freeze event across 10,000-plus orifice meters, across a full month of data.
To be precise about what that means: every hour in which a meter freeze occurred was surfaced as an anomaly, ranked by confidence, and presented to the analyst for review. The analyst’s job became confirmation and acceptance. Not searching, not pattern-matching, not scrolling through columns of FlowCal data looking for the deviation.
Why the Speed Difference Is So Large
The gap between “most of a work week” and “10 minutes” isn’t a marginal improvement in efficiency. It’s a structural difference in how the problem is being approached.
Manual meter freeze review is sequential and dependent on analyst pattern recognition. An analyst reviews one meter, then the next, evaluating each against their mental model of what a freeze looks like. That model is good. Experienced analysts catch most freezes. But it’s slow, it’s fatigue-dependent, and it can’t process thousands of meters simultaneously.
NDL (Meter Freeze) runs a machine learning model trained specifically on orifice meter freeze behavior across diverse operating conditions, geographies, and meter types. It evaluates every meter simultaneously. It doesn’t get tired in the third hour of a freeze review. It doesn’t deprioritize a meter because the exception queue is already long.
The model was trained to recognize the patterns that characterize a meter freeze at the hourly level: not just the obvious events, but the borderline ones, the early-stage ones, and the freezes that manifest differently depending on equipment configuration and weather conditions. The result is identification that’s both faster and more complete than manual review.
What Happens When You Remove the Human from the Loop Entirely
For measurement teams using NDL (Analyst), there’s a further step available: NDL Edit.
With NDL Edit enabled, meter freeze anomalies don’t just get flagged for analyst review. They get corrected automatically on identification. The AI finds the freeze hours, makes the edit, and writes the correction back to FlowCal without an analyst accepting each one individually.
The practical result is a measurement workflow where winter freeze season stops being a staffing problem. Instead of pulling analysts from other work to run the freeze list during cold snaps, the list runs itself. Analysts see what was corrected, can review any flagged event they want to examine more closely, and spend their time on the measurement issues that actually require human judgment.
Zero search time. Zero manual editing. Corrections in the system of record before the analyst’s morning coffee.
What This Means for Your Team Going Into Next Winter
The 10-minute result isn’t a projection or a model estimate. It’s a measured outcome from proof-of-value work on real pipeline systems, across multiple basins, representing real-world operating conditions.
The measurement teams that participated in those studies didn’t change their FlowCal setup. They didn’t restructure their workflows or retrain their analysts. They pointed NDL (Meter Freeze) at their meter data and watched the freeze identification happen in a window of time that, frankly, took some convincing to believe the first time.
If your team spends significant analyst time on freeze identification every winter, or if you’re concerned about freezes being missed because the review process can’t keep pace with the weather, the gap between what your team is doing today and what’s possible is now measurable.
It’s about a work week. Per analyst. Per cold weather event.
NDL (Meter Freeze) is available as part of the NDL suite, with direct FlowCal integration. Schedule a demo to see how it works on your meter data before next winter.
Schedule your demo