A Robust Method for Data-Driven Gas-Lift Optimization

Abstract

Abstract Traditional simulation-based approach for Gas-Lift Optimization depends heavily on the quality of reservoir and fluid data. Excessive OPEX and man-hours are needed to maintain data integrity and to ensure the models are suitably calibrated. Even then, pseudo-steady-state models do not consider losses due to multi-pointing condition and slugging behavior; and for dynamic multiphase flow simulation, the added complexity and man-hours required to assert accurate results cannot be sustained on a full field scale deployment. Gas-Lift Optimization essentially relies on the relationship between the Well Production Rate with the Gas-Lift Injection Rate. The objective of the proposed solution is to remove the need for well models, correlations and personnel from the optimization process and to implement a data-driven (model-free) approach that, by focusing just on the relationship of these variables over time is able to find the next best optimized Gas-Lift Injection Rate setpoint and to implement it directly at the wells via an automated local control loop. This data-driven approach has been compartmentalized and developed as an Edge Application, ran directly on site in an IIOT gateway device. This method has the advantage of providing a predictive response that can be used directly in conjunction with a solver for single-well and multi-well optimization (handling well level and group level constraints by need). The application operates under iterative optimization cycles that progress towards system optimality. Even though well conditions are constantly changing over time, and consequently system optimality, these changes are reflected in the high-frequency data gathered by the application running on the gateway on site. Due to the iterative nature of the process, the solver can recognize these changes and react accordingly, adjusting based on the new system conditions in a closed-loop manner. This paper presents the methodology and the results of a case study of eight wells, including both, single and multi-well optimization. All these wells are unconventional horizontal wells from the Permian basin in Texas, US. Regardless of the complexities associated with unconventional wells, noted by severe slugging and fast changing well conditions, in all the cases the results were outstanding. For the single well optimization, the candidate well was able to outperform the remaining wells in the pad by 5% in production improvement. For the multi-well optimization results vary from 5% to 25% production improvements. The full execution and optimization process was done in a fully autonomous manner, removing completely office and field personnel, as well as the need for well modeling from the optimization process. This solution demonstrates a fully autonomous and Data-Driven Gas-Lift Optimization workflow, from data gathering and processing, edge computation, multi-well optimization based on field constraints, to the direct well implementation via closed-loop control.