ML-Driven Gas Lift Optimization Boosts Unconventional Yields
A new automated workflow uses Bayesian optimization to increase production in unconventional fields without requiring expensive downhole hardware.
Machine learning can now optimize gas injection rates in unconventional oil fields using only historical production time series data. This approach removes the need for costly downhole gauges or multi-rate well tests to forecast Gas Lift Performance Curves. By integrating an ML model with a Bayesian Optimization Framework, operators can solve for optimal injection rates while staying within the physical constraints of facility capacity.
Field pilots across 30 wells in the Bakken region demonstrated an average production uplift of more than 5%. This result has led to a full-scale deployment across more than 200 gas lift and plunger-assisted gas lift wells.
The economic value lies in the removal of hardware dependencies. Traditional optimization requires expensive sensors or disruptive testing cycles that often prove unfeasible due to cost or facility constraints. Shifting this process to a data-driven software layer allows for the optimization of legacy assets that were previously too expensive to monitor at a granular level.
This shift transforms gas lift management from a periodic manual adjustment into a continuous automated process. When applied across thousands of wells in unconventional basins, a 5% uplift represents a significant change in the aggregate recovery rate of a field. It moves the needle on the marginal cost of production by extracting more volume from existing infrastructure.
Other unconventional assets lacking downhole data are now viable candidates for similar optimization. The ability to derive performance curves from surface-level historical data means that the barrier to entry for precision production management is no longer capital expenditure on hardware, but the quality of the existing data archive.
Future scalability depends on the integration of these ML workflows into real-time facility control systems. The transition from offline optimization to closed-loop automation will likely define the next phase of efficiency gains in unconventional resource extraction.