Methanol injection is a hydrate prevention tool. Inject too little and you risk a plug. Inject too much and you're wasting chemical and venting unnecessary emissions. Most operations default to the conservative end — which means most wells are over-injecting every day.
EcoInject replaces that fixed-rate approach with a model trained on thousands of simulation cases. Feed it your well conditions and it tells you the minimum safe injection rate — continuously, without a reservoir engineer in the loop.
This page focuses on the backend model and data pipeline; the Analytics page shows how operators see, compare, and act on those predictions in the field.
Stop guessing on methanol. Start optimizing it with a model built from physics-based simulation data and trained for Alberta gas wells.
We integrate with your existing SCADA system — including AVEVA and OSIsoft PI — or install sensors directly at the wellhead. No rip-and-replace. We work with what's there, pulling live field data and pushing recommendations or status updates back to your control system.
Our ML model — trained on extensive simulation data across thousands of well conditions — predicts the optimal methanol rate and hydrate risk score for your well in real time.
Get an advisory recommendation your operators can act on manually — or let EcoInject close the loop and adjust your pump automatically.
Every injection event is logged. For the first time, you have a complete record of what was injected, when, and why — across your entire well fleet.
Before the model specifics, here’s the end-to-end flow that both Advisory and Control run on.
Tap the diagram to enlarge it and explore the high-level data, prediction, and control flow.
Hydrate plugs shut in wells, damage equipment, and cost days of deferred production. The standard fix — methanol injection — works, but most operators are either over-injecting or under-injecting because real-time dose optimization requires expensive, batch-only workflows.
EcoInject replaces those workflows with a machine learning surrogate model. Using gas composition, brine salinity, water cut, and line pressure, it predicts the optimal methanol dose and the hydrate risk threshold that matters for every well.
Per-well dose recommendations based on actual conditions, not conservative fleet-wide rules of thumb.
Hydrate risk scores update on each production cycle so you see trouble before it happens.
Supports lean gas, rich gas, condensate, and sour wells with automated incompatibility checks.
Scores hundreds of wells in the time traditional workflows take to set up one case.
Accuracy: 1.4°F RMSE against held-out simulation cases. R² = 0.994. Trained on 885,000 simulation cases covering the full Alberta gas composition space, 50–1,200 psi, sweet to moderately sour. Runs where your data already lives — CSV batch workflow, Databricks/cloud connectivity.
The model is trained on hundreds of thousands of pressure-step observations spanning the full Alberta gas composition space, and it uses physics-based fallback when a well falls outside the calibrated training envelope.
1.37°F RMSE, 0.43°F MAE, R² = 0.994 on held-out test cases.
Automated domain checks and physics fallback keep the model from making blind predictions on incompatible stream conditions.
XGBoost gradient-boosted trees trained on hundreds of thousands of pressure-step observations.
van der Waals–Platteeuw thermodynamics cover cases outside the model training envelope.
Optuna Bayesian optimization tunes model hyperparameters with efficiency and repeatability.
MLflow keeps training runs, metrics, and model versions auditable.
Scores an entire well fleet in seconds instead of hours.
Applies the same logic to every well for audit-grade chemical spend recommendations.
Provides HDT output in °F so your team compares predictions directly to ambient conditions.
Flags incompatible wells before scoring and uses physics fallback instead of blind predictions.
Pneumatic pump venting is regulated and increasingly scrutinized by AER. Most operators don't have a systematic way to track it. EcoInject does.
Each stroke of a gas-driven pneumatic pump vents methane. Across a well fleet, that adds up — and regulators are asking for it. Because EcoInject logs every injection event, you get a complete record of pump activity that forms the data foundation for your pneumatic device emissions inventory.
EcoInject captures injection event data at the pump level. That means you can quantify vented emissions per well rather than relying on emission factors and estimates.
Every recommendation, setpoint change, and injection event is logged with a timestamp. For the first time, you have a defensible record of what was injected, when, and why — across your entire fleet.
Optimizing injection rates directly cuts unnecessary pump strokes — which means less vented gas per well, lower chemical spend, and a smaller emissions footprint without any additional reporting burden.
Aggregate injection data across all wells so your environmental team has the numbers they need for TIER reporting and AER emissions inventories — without chasing spreadsheets from the field.
EcoInject is designed to work with the equipment and processes operators already use, while adding a layer of data-driven control that reduces waste and keeps hydrate risk in check.
