Home / Case Studies / HOOTS

Case Study · Energy infrastructure

Catching transformer faults before they cut the power

puniq worked alongside HOOTS sensorsystems to turn raw transformer telemetry into early, actionable warnings, so a fault is found while it is still a trend, not yet a blackout.

Sector Energy infrastructureRegion GermanyPartner HOOTS sensorsystems

The challenge

The data was there. No one could read it together.

Distribution transformers are the quiet backbone of the grid. They run for decades with almost no attention, and then they fail expensively. When a transformer overheats, sheds insulation oil, or drifts out of its normal load envelope, the first sign operators usually get is the outage itself. By then the cost is already paid: lost supply, emergency call-outs, and energy bled away as heat in the weeks before anyone noticed.

HOOTS sensorsystems builds the sensing hardware that reads these machines: temperature, load, vibration and electrical signatures, sampled continuously in the field. The data was there. The problem was that it sat in streams that were hard to read together. A rising trend on one transformer looked the same as normal seasonal load until it was too late.

HOOTS needed the raw telemetry to become a clear, ranked picture of which assets were drifting toward trouble.

That is exactly the problem puniq is built for. We treat a fragmented sensor fleet the same way we treat a fragmented building: many signals, many protocols, no single readable layer. The fix is not more hardware. It is making the existing measurements speak one language.

The approach

Make the measurements speak the same language

We started by inventorying every signal HOOTS captures, normalising units and timestamps, and building one vendor-neutral data model every downstream step could trust. No sensor was replaced. Nothing was ripped out. From that clean layer we modelled what "healthy" looks like for each transformer under its own load and weather, then flagged the deviations that actually predict faults rather than the ones that are just noise.

How the layer was built

  • Normalise every sensor stream into one timestamped, vendor-neutral model
  • Learn a per-asset baseline that accounts for load and seasonality
  • Surface early-warning anomalies, ranked by how strongly they predict failure
  • Keep a human in the loop: the platform advises, the engineer decides

The principle throughout was advisory and human-in-the-loop. The system ranks risk and explains why, and a HOOTS engineer decides what to do. No automatic switching, no black-box calls on live grid hardware.

The outcome

From reactive repairs to a ranked watchlist

0

Earlier fault warning, indicative (2 to 4 weeks)

0

Fewer unplanned call-outs, indicative (30 to 45%)

0

Vendor-neutral data layer, fully owned by HOOTS

Figures shown are indicative ranges that reflect the design intent of the monitoring layer, not audited results from a single site.

What we built

A short, ranked list of what needs attention this week

Engineers stop scrolling through dashboards of green and start their day with the three transformers that are actually drifting. Each flag carries its evidence: which signal moved, by how much, against which baseline, and how confident the model is.

OUTCOME 01

Faults seen as trends

Developing faults now show up as ranked anomalies weeks before they would have tripped an outage. Indicative early-warning lead time of two to four weeks across the monitored fleet.

OUTCOME 02

Less energy wasted

Transformers running hot or out of envelope are caught early, cutting the slow energy losses that build up between failure and discovery.

OUTCOME 03

Crews sent where it matters

Condition-based scheduling replaces blanket calendar maintenance, with an indicative 30 to 45 percent drop in unplanned call-outs.

Why it held up

Respect the hardware, make it readable

The work succeeded because it respected the hardware already in the field and made it readable rather than redundant. German engineering rigor applied to messy, multi-source telemetry: documented, vendor-neutral, and handed over in full. Because the model is documented end to end, every input, every threshold, every assumption written down, HOOTS owns it and can extend it without us in the room.

Delivered to HOOTS

  • One normalised, vendor-neutral telemetry model
  • Per-asset health baselines tuned for load and season
  • Ranked, explained anomaly watchlist
  • Condition-based maintenance scheduling logic
  • The Documented Data Map, fully owned by HOOTS

We always had the sensor data. What puniq gave us was the ability to read it: to know which transformer needs an engineer this week, and why. That changes how we run the fleet.

Dr. Henry Kutz
CEO, HOOTS sensorsystems

More work

Related case studies

Have telemetry you cannot read?

If your sensors are talking but no one is listening together, we will map the path to one intelligence layer.

العربية