Proven at compressor stations
Deployed random-forest and gradient-boosting fuel-optimization models across 30+ North American compressor sites delivered 3–5% savings — then drifted within weeks. Adaptive DRL per component holds the gains and goes further.
Where the fuel goes: recycle
The 6–9% comes off one mechanism more than any other. This is that mechanism.
Surge is a sudden flow reversal that strikes when a compressor's throughput falls too low for the pressure it is developing — flow breaks down, reverses, and oscillates roughly once a second, stressing and rapidly damaging the machine (the looping open-loop trace above). The surge line marks that stability limit on the compressor map. Your existing anti-surge control — a dedicated high-speed loop running at 20–50 ms — holds the machine clear of that line by opening a recycle valve whenever the operating point drifts too close.
Recycle is the cost. Every unit of recycled gas is compressed twice and delivered once. OmniPath does not move your surge control line and does not touch your anti-surge controller. It works one layer up: choosing speed, load split, and unit staging so the operating point never drifts toward the line in the first place, and anticipating the transients that would otherwise drive it there. Less time near the line, less recycle, less fuel — with the anti-surge controller unchanged and still holding the backstop.
Assumptions behind these figures
Stated explicitly so you can substitute your own. Every figure in the tables below is derived from this set and nothing else.
| Parameter | Value | Basis |
|---|---|---|
| Average station duty | 11 MW shaft | Stated |
| Annual operating hours | 8,400 h | Stated |
| Annual shaft work | 92,400 MWh | 11 MW × 8,400 h |
| Driver thermal efficiency | 30% | Simple-cycle mechanical drive; site-specific |
| Annual fuel input | ~1,109,000 GJ | 92,400 MWh ÷ 30% × 3.6 GJ/MWh (≈1.0 Bcf) |
| Fuel gas price | C$2.50/GJ | AECO-referenced; $2.00–$3.50 range applied below |
| Emission factor | 50 kg CO₂e/GJ | Fuel-based; ≈0.60 t CO₂e per MWh of shaft work |
| Annual emissions | ~55,450 t CO₂e | 1,109,000 GJ × 50 kg/GJ |
Fuel gas, per typical station, per year
A 7.5% fuel reduction — the mid-point of the 6–9% range — saves ~83,000 GJ/yr. At C$2.50/GJ that is ~$208k; across the $2.00–$3.50/GJ range, $166k–$291k.
| Scenario | Fuel input (GJ) | Fuel cost @ $2.50/GJ | Saving vs baseline |
|---|---|---|---|
| Baseline | 1,109,000 | $2,773k | — |
| DCS tuning (2%) | 1,087,000 | $2,718k | $55k |
| Traditional ML, pre-drift (4%) | 1,065,000 | $2,663k | $110k |
| OmniPath supervisory DRL (7.5%) | 1,026,000 | $2,565k | $208k |
Fuel input rounded to the nearest 1,000 GJ; cost and saving computed from the rounded figure in the same row, so each column subtracts cleanly. Fuel saving is a fuel saving only — it is not combined with any carbon figure.
Emissions avoided: ~4,150 t CO₂e/yr
What that's worth depends on your compliance position and the realized credit price. 83,000 GJ of fuel not burned at 50 kg CO₂e/GJ.
| Effective credit price | Annual compliance value |
|---|---|
| C$20/t | ~$83k |
| C$60/t | ~$249k |
| C$110/t | ~$457k |
| Your assumption — C$30/t | ~$125k |
Canadian industrial carbon pricing is output-based. Under Alberta's TIER regulation, compliance cost is assessed against a facility benchmark rather than on gross emissions, and the value of an avoided tonne is set by the realized credit price — which traded between roughly C$17 and C$39 during 2025, well below the headline schedule. We present emissions avoided in tonnes and let you apply your own compliance position. We do not model carbon as a guaranteed saving.