AI-Driven Automation of Rod Lift Systems Using Non-Fixed Stroke Length Pumping Units

Abstract

This paper presents the integration of artificial intelligence (AI)-enabled pump-off controllers (POCs) with non-fixed stroke length rod pumping systems to enable fully automated corrective actions in rod lift operations. Conventional beam pumping systems, while capable of diagnosing conditions such as gas interference, pump tagging, and rod loading anomalies, lack the mechanical flexibility required to implement real-time corrective actions. This results in deferred production, increased intervention frequency, and higher operating costs. Field applications across multiple wells demonstrated measurable improvements, including increased pump fillage from 34% to 52%, elimination of pump tagging events, prevention of rod compression failures, and recovery of stuck pumps without service rig intervention. Economic benefits exceeded $70,000 per event, with production gains greater than $2,700 per day in select cases. The results demonstrate that integrating AI diagnostics with dynamic mechanical control enables autonomous optimization of rod lift systems, significantly improving production efficiency, reducing costs, and enhancing operational safety.