Abstract Gas lift is one of the most important methods of artificial lift for oil wells across the globe. However, the productivity of gas lift wells hinges on the fluctuating availability of compressed gas, presenting a complex optimization challenge. Traditionally, this issue has been tackled either manually, with significant delays, or automatically, with limiting assumptions that hinder optimal results. We introduce a groundbreaking closed-loop software solution that revolutionizes gas lift allocation by combining speed and precision. This innovative software generates gas lift performance curves for each well using a hybrid approach incorporating both data-driven methods and physics-informed machine learning. These performance curves, along with critical physical constraints like minimum stable rates and maximum injection rates, enable the determination of optimal gas allocation strategies across varying gas availability scenarios. We also present a method for producing performance surfaces, which estimate production by considering choke position and flowline pressure in addition to gas lift injection rate, and thus allow optimization based on both gas availability and "take-away" constraints. The implemented solution continuously estimates the available gas from the compressor system, based on real-time discharge pressure readings, and selects the most suitable strategy every 15 minutes for over 600 wells. Additionally, it provides insightful data on optimization success, offering a clear understanding of the system's performance. Furthermore, the highly responsive nature of the process has been shown to inherently maximize the utilization of the compression capacity installed. Remarkably, this approach has achieved up to a 4% production uplift in gas lifted wells, even before the implementation of the multi-dimensional production surfaces, which we expect to provide significant incremental value.
Burmaster et al. (Mon,) studied this question.