{
  "name": "MAES glycol dehydrator emissions model",
  "url": "https://tetrasoftco.com/tools/glycol-dehydrator-emissions/",
  "evaluate": {
    "method": "POST",
    "url": "https://tetrasoftco.com/tools/dehydrator-emissions.json",
    "note": "Returns emissions for one operating point. The fitted coefficients are not published and are evaluated server side."
  },
  "modelReviewed": "2026-08-03",
  "license": "Free to use with attribution to TetraSoft.",
  "source": {
    "title": "Mechanistic Modeling of TEG Dehydrator Emissions in Oil and Gas Industry",
    "authors": "Mdigo, J.; Santos, A.; Duggan, G.; Vora, P.; Shonkwiler, K.; Zimmerle, D.",
    "journal": "Fuels",
    "year": 2026,
    "datePublished": "2026-04-07",
    "volume": 7,
    "issue": 2,
    "articleNumber": 21,
    "doi": "10.3390/fuels7020021",
    "url": "https://doi.org/10.3390/fuels7020021",
    "access": "Open access under CC BY.",
    "institution": "The Energy Institute, Colorado State University",
    "method": "Second-order polynomial regressions fitted to 117 ProMax scenarios: 13 glycol circulation ratios by 3 wet gas temperatures by 3 wet gas pressures, with the contactor modelled as three theoretical equilibrium stages."
  },
  "whatItModels": {
    "summary": "Hydrocarbons absorbed into triethylene glycol at the contactor and released at the flash tank and still vent of a TEG dehydration unit.",
    "outlets": [
      {
        "id": "contactor",
        "label": "Dry gas to sales",
        "maesStream": "gas_sales",
        "emitsToAtmosphere": false,
        "note": "The dehydrated product stream leaving the top of the contactor tower."
      },
      {
        "id": "separator",
        "label": "Flash tank",
        "maesStream": "flash_tank_flashes",
        "emitsToAtmosphere": true,
        "note": "Gas released from the rich glycol when it is depressurised, typically to 40 to 100 psig. A flash tank can recover up to 90 percent of the absorbed gas, which can be routed to a control device or reused as reboiler fuel instead of vented."
      },
      {
        "id": "still_vent",
        "label": "Still vent",
        "maesStream": "still_vent_emissions",
        "emitsToAtmosphere": true,
        "note": "Hydrocarbons released with the steam in the still column as the rich glycol is heated and the water is boiled off."
      }
    ],
    "species": [
      {
        "id": "c1",
        "label": "Methane",
        "formula": "CH4"
      },
      {
        "id": "c2",
        "label": "Ethane",
        "formula": "C2H6"
      },
      {
        "id": "c3",
        "label": "Propane",
        "formula": "C3H8"
      },
      {
        "id": "ic4",
        "label": "Isobutane",
        "formula": "i-C4H10"
      },
      {
        "id": "nc4",
        "label": "n-Butane",
        "formula": "n-C4H10"
      },
      {
        "id": "ic5",
        "label": "Isopentane",
        "formula": "i-C5H12"
      },
      {
        "id": "nc5",
        "label": "n-Pentane",
        "formula": "n-C5H12"
      },
      {
        "id": "c6",
        "label": "Hexane",
        "formula": "n-C6H14"
      },
      {
        "id": "c7",
        "label": "Heptane plus",
        "formula": "n-C7H16"
      },
      {
        "id": "c8",
        "label": "Octane plus",
        "formula": "C8"
      },
      {
        "id": "c9",
        "label": "Nonane plus",
        "formula": "C9"
      },
      {
        "id": "CO2",
        "label": "Carbon dioxide",
        "formula": "CO2"
      },
      {
        "id": "N2",
        "label": "Nitrogen",
        "formula": "N2"
      },
      {
        "id": "H2S",
        "label": "Hydrogen sulfide",
        "formula": "H2S"
      }
    ],
    "averageOutletFractions": {
      "note": "Mean volumetric split across all 117 simulated cases. Used as a fallback when the fitted polynomials return a physically impossible split outside the calibrated range.",
      "contactor": 0.9995568,
      "separator": 0.0002856,
      "still_vent": 0.000145
    },
    "calibrationComposition": {
      "note": "Inlet wet gas the regressions were characterised on, in mole percent. Composition sensitivity is a stated limitation of the model.",
      "molePercent": {
        "c1": 68.978,
        "c2": 11.654,
        "c3": 5.694,
        "ic4": 0.823,
        "nc4": 1.916,
        "ic5": 0.441,
        "nc5": 0.423,
        "c6": 0.508,
        "c7": 1.183,
        "c8": 0.592,
        "c9": 0.197,
        "N2": 5.668,
        "CO2": 1.907,
        "H2S": 0.015
      }
    }
  },
  "operatingEnvelope": {
    "note": "The range the regressions were fitted across. Outside it the model extrapolates, and the authors report it underpredicts at high circulation ratios.",
    "circulationRatio": {
      "min": 1,
      "max": 7,
      "units": "gal glycol per lb water"
    },
    "temperatureF": {
      "min": 77,
      "max": 122,
      "units": "degrees F"
    },
    "pressurePsia": {
      "min": 400,
      "max": 1200,
      "units": "psia"
    },
    "recommendedCirculationRatio": {
      "min": 3,
      "max": 5,
      "source": "EPA Natural Gas STAR"
    }
  },
  "validation": {
    "methaneRmse": 0.00091,
    "worstLightHydrocarbonRmse": 0.02935,
    "dryGasFractionRmse": 0.002273,
    "flashTankFractionRmse": 0.001411,
    "stillVentFractionRmse": 0.002056,
    "note": "Root mean square error against held-out ProMax results. Thirty percent of the simulation set was reserved from fitting and used for validation."
  },
  "fieldFindings": {
    "note": "From the AMI 2024 deployment reported in the same paper. The denominator differs between the abstract and Section 3.4, so shares are quoted rather than a unit count.",
    "gasDrivenPumpUnits": 54,
    "gasDrivenPumpShare": 31,
    "uncontrolledFlashTankUnits": 6,
    "uncontrolledFlashTankShare": 11.1,
    "oversizedUnits": 22,
    "oversizedShare": 40.7,
    "pumpShareOfDehydratorEmissions": 90.25,
    "pumpShareOfSiteEmissions": 63.1,
    "pumpShareRangeLow": 83.2,
    "pumpShareRangeHigh": 92.4,
    "project": "AMI 2024"
  },
  "notCovered": [
    {
      "id": "pumps",
      "heading": "Gas-assisted glycol pumps are not in this model",
      "detail": "A gas-driven pump uses wet gas as its motive force and vents it. That gas is not absorbed in the contactor, so the absorption regressions do not see it. In the field study, pump emissions were 90.25 percent of total dehydrator emissions at units with uncontrolled flash tanks, and 63.10 percent of site-level emissions. If your unit has a gas-assisted pump, the number here is a floor and not a total."
    },
    {
      "id": "envelope",
      "heading": "Outside the calibrated range the model extrapolates",
      "detail": "The regressions were fitted for circulation ratios of 1 to 7 gal/lb, 77 to 122 degrees F and 400 to 1200 psia. The paper reports facilities running at 18, 31 and 71 gal/lb, well outside that range, where the model underpredicts because absorption becomes increasingly non-linear at high solvent flow."
    },
    {
      "id": "composition",
      "heading": "It was fitted on one gas composition",
      "detail": "The regressions were characterised against a single inlet composition, held constant so the effects of circulation, pressure and temperature could be isolated. Heavier hydrocarbon and acid gas content change absorption, and the authors name composition sensitivity as a limitation and a direction for further work."
    },
    {
      "id": "reboiler",
      "heading": "Reboiler combustion is excluded",
      "detail": "The model does not cover carbon dioxide from burning fuel in the reboiler, or methane slip from that burner. Those are real emissions from a dehydrator and are not counted here."
    },
    {
      "id": "upsets",
      "heading": "Steady state only",
      "detail": "Failure modes and upset conditions are not modelled. The 2024 field campaign observed a dehydrator failure emitting 138 kg an hour, which no steady-state model predicts."
    }
  ]
}