Water leak detection from data.

What is water leak detection?

Water leak detection from metering data means treating every meter as a sensor rather than a billing device. A meter that reports hourly already knows whether water is moving at three in the morning, whether consumption stepped up overnight and stayed there, and whether flow has reversed. Those three patterns catch the majority of leaks that matter, long before a wet patch appears on a road or a customer opens a surprising invoice.

It works at every scale. In an apartment building, continuous flow through one unit’s meter is usually a running toilet. At a venue like Unibet Arena, 15-minute readings surfaced potential leaks within two weeks of the first meters going live. At network scale, the same logic becomes minimum night flow per zone.

How it works on Divako

1. Meter. Any meter that reports a series will do – LoRaWAN, NB-IoT, wM-Bus, or drive-by collection. The device profile knows the meter’s register format, its plausible range and which alarm flags the meter itself raises.
2. Network. Readings arrive on their own schedule. Daily is enough to find a running toilet; hourly or 15-minute data is what you need to see a burst while it is still small.
3. Platform. Readings are validated on arrival – gaps, duplicates, implausible steps – so an alarm is not triggered by a missing message. Consumption is then compared with the connection’s own history rather than with a global threshold.
4. Analysis. Continuous flow, sudden sustained steps, reverse flow, tamper and frost are evaluated per meter type. A DN15 apartment meter and a DN200 trunk main do not share thresholds.
5. Integration. Alarms route to email, SMS, push, webhook or your incident system, and the underlying series is available over REST and MQTT for SCADA, GIS or a customer-service screen.

What to watch out for

• Continuous flow is a symptom, not a diagnosis. A running toilet, a garden tap left open, a dripping heat exchanger and a cracked service pipe all look identical in the data. The alarm’s job is to send someone to look; be honest about that in the message, or field crews will stop trusting it.
• The baseline has a season. Irrigation, pool filling, flushing for water quality and holiday homes all move consumption without anything being wrong. Build the baseline from the connection’s own history across a year, and expect to suppress a predictable period rather than to explain it every time.
• A customer-side leak needs an owner before it needs an algorithm. Detection is the easy part; the awkward part is who calls the household, what they are allowed to say and who pays for the plumber. Agree the script with customer service before switching the alarms on.
• The reading interval decides which leaks you can see. Daily readings find things that run for days. A burst that empties a service pipe in an afternoon needs hourly or finer data – which is also a battery decision, so set it per meter type rather than everywhere.