Machine intelligence for power systems
Combine breaker, relay, meter and sensor data to identify abnormal equipment behavior before it becomes an outage. Four applications, each with a clear input, a clear output and a human decision at the end.
Four applications
Move from calendar-based to condition- and prediction-based maintenance for breakers and switchgear.
Analyzes
Output
Ranked maintenance recommendations with the evidence behind each one, and a confidence level tied to the size of the data record.
Detect unusual heating before it becomes an insulation failure or a bus fault.
Analyzes
Output
Phase-to-phase and load-normalized ΔT with trend rate; alarms explain why (loose joint signature vs. load-driven heating).
Turn meter waveforms and events into causes and consequences.
Analyzes
Output
Which loads drive distortion, whether limits at the point of common coupling are at risk, and what changed after a mitigation.
One defensible number per asset, built from evidence a maintenance planner can inspect.
Analyzes
Output
Health score with contributing factors, status and the next recommended action — e.g., Asset Health 87/100, Healthy — Monitor breaker operating time trend.
Worked example
52-LM1 from the demonstration system: a slow opening-time trend, a bus-joint thermal anomaly on its bus, and the harmonic spectrum at its meter.
Trend from timed operations over 24 months. Still inside the band; rate suggests lubrication at the next planned outage. Confidence: medium (14 records).
| Period | Opening time |
|---|---|
| M-23 | 32 |
| M-21 | 32 |
| M-19 | 33 |
| M-17 | 33 |
| M-15 | 33 |
| M-13 | 34 |
| M-11 | 34 |
| M-9 | 35 |
| M-7 | 35 |
| M-5 | 36 |
| M-3 | 37 |
| M-1 | 37 |
| Now | 38 |
A single phase rising while B and C track load and ambient is the signature of a degraded connection, not a load change. Recommendation: de-energized inspection and torque check at the next window.
| Period | Phase A | Phase B | Phase C |
|---|---|---|---|
| D-13 | 50 | 49 | 48 |
| D-12 | 51 | 50 | 49 |
| D-11 | 52 | 50 | 50 |
| D-10 | 54 | 51 | 50 |
| D-9 | 56 | 51 | 51 |
| D-8 | 58 | 52 | 51 |
| D-7 | 61 | 52 | 52 |
| D-6 | 64 | 53 | 52 |
| D-5 | 67 | 53 | 53 |
| D-4 | 70 | 54 | 53 |
| D-3 | 74 | 54 | 54 |
| D-2 | 77 | 55 | 54 |
| D-1 | 80 | 55 | 55 |
| D-0 | 83 | 56 | 55 |
5th and 7th dominate — consistent with six-pulse drives on Mechanical MCC A. Compare against the IEEE 519 limits that apply at the point of common coupling for this service.
| 3 | 0.9 |
| 5 | 4.6 |
| 7 | 3.1 |
| 9 | 0.4 |
| 11 | 2.2 |
| 13 | 1.4 |
| 15 | 0.2 |
| 17 | 0.9 |
| 19 | 0.7 |
| 21 | 0.1 |
| 23 | 0.5 |
| 25 | 0.4 |
How this stays safe
Asset health
Asset health — 52-LM1 LV Main 1
Healthy — Monitor breaker operating time trend
The score combines weighted evidence from six sources. It is a prioritization aid for the maintenance planner — it does not change protection settings or operate equipment.
Contributing factors (illustrative)
Weights and scoring are configured per asset class and site. Confidence is reported alongside the score when the data record is thin.
Apex Power Distribution · ApexPower.ai
Send a one-line, a specification or a photo of what is installed today. An engineer reviews it and tells you what fits — equipment, protection, controls and monitoring — before anyone talks price.