Strip the jargon away and the entire discipline is one sentence: measure the fuel and the air the machine is actually getting, and control the burn for those conditions — not the average conditions it was shipped for. This page is the instrumentation, the math and the code that make that sentence true.
Stoichiometric air/fuel ratio for pure methane, by mass — and every real gas moves it
Energy delivered through the same orifice by a routine field-gas Wobbe drift — uncommanded, unalarmed
Fuel energy saved per point of electrical efficiency recovered
Coldest ambient envelope our calibration methods are validated against
Every gaseous fuel is a mixture, and the mixture is never the datasheet. Pipeline gas drifts daily with supply blending; field gas carries ethane, propane, CO₂, N₂ and H₂S in proportions that change with the well; biogas is half diluent on a good day. Three numbers characterize what a fuel will actually do in a combustion system: heating value (LHV/HHV — how much energy is in it), the Wobbe Index (how much of that energy flows through a fixed orifice — the number the fuel system actually experiences), and the methane number (how hard you can push it before it knocks). A calibration built without measuring all three is a calibration built for a fuel the machine is not burning.
A wellhead machine commissioned on pipeline-spec gas sees the supply enrich as a new zone comes on. Nobody touches a control. Here is what physically happened:
| Gas A — pipeline spec | Gas B — enriched field gas | |
|---|---|---|
| CH₄ / C₂H₆ / C₃H₈ (mol %) | 100 / 0 / 0 | 85 / 10 / 5 |
| HHV (MJ/m³) | 39.7 | 45.7 |
| Specific gravity | 0.554 | 0.651 |
| Wobbe Index (MJ/m³) | 53.4 | 56.6 (+6.0%) |
| Methane number | 100 | ≈ 70 |
| Stoich AFR (mass) | 17.2 : 1 | 16.8 : 1 |
Through the same fuel orifice, the machine now receives ~6% more energy per unit of airflow. A lean-burn engine sitting at λ = 1.70 slides to λ ≈ 1.60: flame temperature climbs on the order of 50 °C, thermal NOx — exponential in temperature — rises steeply, and 30 points of knock margin are gone. No alarm sounds. The machine simply runs hotter, dirtier and closer to its limits until someone measures the fuel. Composition-aware feed-forward corrects all of this before the flame notices — that is the entire premise, in one table.
Combustion air is not a constant. Density falls with altitude and temperature and is displaced by humidity; a −30 °C prairie morning delivers meaningfully more oxygen per intake stroke or per compressor revolution than a +30 °C afternoon at the same site. Factory ambient-correction curves approximate this for a fleet-average climate. We replace approximation with measurement: the site's real seasonal envelope — temperature, barometric pressure, humidity — becomes a calibrated input, not a lookup table written somewhere else.
For any fuel there is an exact mass of air that burns it completely — the stoichiometric air/fuel ratio. Lambda (λ) expresses how far the actual mixture sits from that point, and it is the single most consequential number in the machine: it sets flame temperature, flame speed, stability margin, NOx and CO formation, and efficiency simultaneously. The entire calibration problem reduces to computing the correct λ target for the real fuel and real air at every operating point, and then holding it there. Rich of target wastes fuel and makes NOx; lean of target risks misfire, lean blowout and CO; the optimum moves whenever the fuel or the weather does.
Our control architecture has two layers. Feed-forward uses what we know — gas composition, ambient conditions, requested load — to compute the operating point before the flame sees the change. Feedback uses what the exhaust reports — stack O₂ for excess air, CO for incomplete combustion, NOx for excessive temperature — to hold the target while everything drifts. Feed-forward without feedback goes stale; feedback without feed-forward is always reacting late. Together, with the calibration defining the map between them, the machine runs at its computed optimum continuously rather than at commissioning day's best guess.
Every machine has hard boundaries — turbine inlet temperature, exhaust temperature, knock margin, lean blowout, surge margin, emissions permits, mechanical speed. Factory calibrations park far from all of them simultaneously, because they must survive every fuel and climate the machine might ever meet. Site calibration measures where the limits actually are for your fuel and your atmosphere, and operates confidently close to them — with the margins proven by data rather than assumed by fleet policy. Inside that measured envelope there is exactly one objective: minimum fuel energy per unit of useful output.
Every engagement closes the loop the same way it started — with measurement. Performance is corrected to reference conditions so the calibration's effect can be separated from weather and load, compared against the instrumented baseline taken before any change, and documented with the full data logs. If the improvement can't be shown in corrected, time-aligned data, we don't claim it. That is the entire quality system: the exhaust tells the truth, and so do we.
fuel composition × Wobbe / LHV × inlet temp × baro pressure × humidity × requested load → optimum fuel mass flow → equivalence ratio → combustion temp → output
Six relationships carry the entire discipline. Everything we deploy is these, computed continuously for measured inputs instead of assumed ones.
AFRₛ = Σ xᵢ·AFRᵢThe exact air requirement of the mixture, computed component-by-component from the full gas analysis — methane through hexanes-plus, CO₂, N₂, H₂S. The number every other calculation stands on.
λ = AFR / AFRₛThe governing ratio. Sets flame temperature, flame speed, stability margin, NOx and CO formation and efficiency simultaneously. The calibration problem is computing the right λ target for real fuel and real air, then holding it.
WI = HHV / √SGEnergy delivered through a fixed orifice at fixed pressure. The number the fuel system actually experiences — when WI drifts, the machine re-tunes itself without permission.
ρ = P / (R·T)Oxygen per intake stroke or compressor revolution, set by temperature, barometric pressure and humidity. The atmosphere's daily edit to the calibration — measured here, not assumed.
T𝒂𝒅 = f(comp, λ, Tᵢₙ)Adiabatic flame temperature — what the hot section actually feels. Exponentially linked to thermal NOx and to hot-section creep life; managed by λ and composition-aware scheduling.
HR = E𝒇ᵤₑₗ / kWhₑFuel energy per unit of output — the objective function. Every equation above exists so this number can be minimized inside measured limits and proven against baseline.
Every fuel needs an exact quantity of air to burn completely. Too little and you waste fuel out the exhaust; too much and you heat air for nothing. Pure methane wants about 17.2:1 by mass — but ethane, propane, CO₂ and H₂S in real gas all move that number. Lambda (λ) is simply "how far from perfect" the mixture is, and it defines the safe window between misfire on the lean side and knock or NOx on the rich side. We compute it for the gas you actually burn.
Two different gases flowing through the same orifice can deliver different amounts of energy. The Wobbe Index predicts exactly how much — which means when the gas supply drifts, the machine silently re-tunes itself without anyone touching a control. A Wobbe swing from 48 to 54 MJ/m³ pushes a lean-burn engine from λ 2.0 to 1.78: hotter, dirtier, closer to knock. We correct for it before the flame notices.
Methane number tells you how hard you can push a fuel before it fights back. It works like octane: methane scores 100, hydrogen 0, and heavier hydrocarbons drag the score down. High-compression engines want MN ≥ 80; real field gas often can't deliver it. Below the design number the machine must knock, retard timing, or derate — unless the calibration was built for the actual fuel from the start.
Cold air simply contains more oxygen than hot air. Altitude thins it, humidity displaces it. The machine can only burn the air the atmosphere actually delivers — so a −30 °C prairie morning and a +30 °C afternoon are two different machines, tens of kilowatts apart. We calibrate to the measured seasonal envelope, not a one-size correction curve.
Whatever the intent was, the exhaust reports what actually happened. Stack O₂ says how much excess air went through. CO says where the flame went too lean or ran out of time. NOx says where it ran too hot. Feed-forward from gas data sets the operating point; feedback from the exhaust holds it there while fuel, load and weather drift.
Every machine has real limits — temperature, speed, surge, knock, emissions — and factory calibrations park a long way from all of them, everywhere, forever. Site calibration measures where the limits actually are for your fuel and your climate, and operates confidently close to them. Inside that space, one objective: minimum fuel per unit of useful output.
Send a gas analysis and a month of operating data — we'll compute what site-specific calibration recovers, with the reasoning shown.
Start with a Free Fuel-Data Review