CASE 004 / APPLIED INTELLIGENCE

FAULT INTELLIGENCE · ELECTRICITY DISTRIBUTION

ONE PROBLEM, DECODED.

Know which faults
will persist.

A classification system that predicts whether a feeder fault will be short-lived, so field resources can be directed where intervention matters.

CURRENT READINGS · OUTAGE REPORTS · LOCAL WEATHER
CURRENTOUTAGEWEATHER
CLASSIFY?
SHORT-LIVEDPERSISTENT

ONE EVENT / TWO DIFFERENT RESPONSES

Not every outage needs
the same response.

Reliability incidents make the maintenance and management of large electricity networks more complex. When a fault occurs, the operation must decide where limited field resources should go first.

A meaningful share of feeder faults are temporary and resolve quickly. Distinguishing those events from persistent outages changes the response: unnecessary dispatches can be avoided while durable faults receive attention sooner.

Classify duration. Allocate resources.

The prediction is made at the point of action, using the context available when an outage enters the operation.

We connected feeder current measurements with outage reports describing time, location, reason, and duration. Local temperature and wind speed added the environmental context around each event.

A classification model predicts whether the fault will be short-lived. Deployed as a web service and integrated with SCADA, it returns the prediction as soon as ground personnel enter the outage information.

MODEL INPUTS
  • Feeder current
  • Fault time and location
  • Outage reason and duration
  • Temperature and wind speed
  1. 01

    Connect

    Join electrical measurements, outage history, and local weather.

  2. 02

    Classify

    Estimate whether the event will be short-lived or persistent.

  3. 03

    Integrate

    Expose the model through the existing SCADA workflow.

  4. 04

    Allocate

    Use the prediction to direct operational resources intelligently.

A faster answer at the moment of dispatch.

PROJECTED

70%+

classification accuracy

expected under live validation
DEPLOYED

SCADA

integrated workflow

prediction inside the operational environment
DEPLOYED

NOW

decision timing

returned when outage information is entered

The value is not prediction in isolation. It is a better dispatch decision: distinguishing temporary faults from those that require scarce field capacity.

What looks fixed
may only be familiar.

Bring us the problem