The Future of Plunger Lift Control Using Artificial Intelligence

Abstract Plunger lift is a form of artificial lift popular due to its low installation and operating costs. Historically plunger lift was reserved for low producing stripper gas wells due to limitations on liquid rate and gas volume requirements. More recently, the use of continuous run plungers and SCADA systems have extended application of this method of artificial lift to wells producing higher volume of liquids. Today, gas-assisted plunger lift (GAPL) and plunger-assisted gas lift (PAGL) technologies allow wells with insufficient gas volume produced by the reservoir or low gas-liquid ratios (GLR) to also lift liquids to the surface with plungers. Plunger lift is currently used in every major shale play in the U.S. While plunger systems are attractive for their low cost and relative simplicity, several challenges prevent engineers and optimizers from operating wells equipped with these systems at peak production with minimum lifting cost ($/Mcf). One of the primary challenges encountered by plunger lift operators is selecting the appropriate algorithm to control the various aspects of a plunger cycle. Once selected, frequent setpoint adjustments are often necessary to accommodate varying well conditions and production loss as the well moves along its own natural decline curve. Dozens of plunger lift control algorithms have been developed to account for different well conditions and optimization protocols. However, challenges exist that prevent optimization at scale which include: varying degrees of operator knowledge, time availability, number of wells, changing well conditions, data quality, data accessibility, varying plunger lift controllers, lack of API standards, limited understanding of downhole conditions, etc. To address these challenges, a plunger lift optimization software was developed. One aspect of the software is enabling setpoint optimization at scale. This paper will present the methodology to do so, detailing three separate areas working in unison to offer significant value to plunger lift well operators. First, a novel physics engine was developed to deliver superior downhole insights. The physics engine incorporates improved analytical models for horizontal wellbores from literature and implements improved mass balance and thermodynamic equations of state, which allow for improved calculations of critical flow rate, critical lift pressure, and plunger fall velocity. Second, dynamic well optimization was employed to drive optimization decisions and provide anomaly detection to users. The well optimization model dynamically runs calculations over the data, alerting users to key anomalous conditions and provides insights into well instability and sub-optimized states. Third, artificial intelligence was deployed to drive further optimization and allow setpoints to optimize continually over time. Layered on top of improved physics and well insights, artificial intelligence and numerical optimization continually search for the optimal setpoint values and plunger selection to maximize production rates across entire fields.