ML-Based Gas Lift Optimization Workflow for Unconventional Fields
An innovative automated workflow leveraging Machine Learning (ML) for optimizing gas lift in unconventional fields has been created. This system combines an ML model that predicts the Gas Lift Performance Curve with a Bayesian Optimization Framework to identify the best gas injection rates while adhering to facility capacity limits. By utilizing historical production time series data, the ML model eliminates the necessity for downhole gauges or multi-rate well tests. Tested on 30 wells across 5 well pads in the Bakken formation, the workflow resulted in an average production increase of over 5%. Following the pilot's positive outcomes, it has been fully implemented in more than 200 gas lift and plunger-assisted gas lift (PAGL) wells in Bakken, offering a cost-effective solution for assets lacking downhole data or multi-rate testing capabilities.
Key facts
- Workflow uses ML to forecast Gas Lift Performance Curve
- Bayesian Optimization Framework solves for optimal gas injection rates
- ML model uses historical production time series data
- No downhole gauges or multi-rate well tests required
- Piloted on 30 wells across 5 well pads in Bakken
- Achieved over 5% production uplift on average
- Fully deployed across 200+ gas lift and PAGL wells in Bakken
- Presented as effective and economic solution for other assets
Entities
Locations
- Bakken