Less Gas, More Oil: Unlocking Production and Gas Efficiency Through Generative AI-Enabled Gas-Lift Optimization

Abstract

Abstract Artificial Intelligence (AI) is playing an increasingly pivotal role in transforming production optimization within the oil and gas industry by enabling data-driven and near real time decision-making. In gas-lifted wells, where injection gas is often a constrained and valuable resource, AI provides a robust framework to optimize gas allocation, enhance well performance, and maximize overall asset value. Through the application of advanced analytics and intelligent automation, operators can systematically reduce inefficiencies associated with over-injection, improve well stability, and unlock additional production while preserving gas for export and higher-value use. This paper presents a field-deployed generative AI based application developed to optimize gas lift performance and enhance overall production system efficiency. The approach integrates physics-constrained proxy modeling with engineering constraints to evaluate well performance under varying gas-lift operating conditions. A field-deployed Generative AI-enabled decision-support application was developed to interpret validated optimization outputs and provide engineers with structured recommendations for gas-lift allocation. Optimal gas-injection setpoints are determined by the production proxy and field-wide optimization framework, while the GenAI layer provides engineering rationale, operating context, and implementation guidance. The workflow includes data preprocessing, proxy-model development, continuous validation, optimization, and deployment within daily operations. The system dynamically ranks wells based on responsiveness to gas lift, enabling efficient gas redistribution across the field based on historical trends. Field implementation demonstrated substantial improvements in gas lift efficiency and overall production performance. The AI-driven system effectively reduced total gas lift injection rates while maintaining and in several cases increasing oil production, indicating the elimination of over-injection in multiple wells. Optimized gas allocation enhanced well stability, minimized production cycling, and improved overall operational reliability. In addition, the reduction in gas lift requirements enabled the reallocation of surplus gas volumes to export streams, resulting in increased sales gas and improved asset value. The generative AI-based application facilitated faster, more consistent, and data-driven decision-making compared to conventional manual optimization practices. Notably, gas-lift injection reductions reached approximately 90% in the most significant individual-well case, accompanied by measurable increases in wellhead temperature. This behavior indicates the mitigation of gas circulation effects and more efficient lifting performance, ultimately enhancing oil production with minimal gas lift input and OPEX reduction. Overall, the results confirm that integrating AI-driven analytics with sound engineering judgment can deliver sustainable production gains, improve operational efficiency, and optimize resource utilization across gas lift operations. This paper presents a novel integration of production proxy modeling, field-wide optimization, and Generative AI-assisted engineering decision support, moving beyond standalone predictive or optimization workflows. The field-proven workflow demonstrates scalable deployment, near-real-time usability, and measurable value creation through simultaneous production enhancement and gas conservation, contributing to smarter and more sustainable asset management.