Analytical Evaluation of Gas Lift Performance Optimisation Under Reservoir Decline Conditions Using Nodal Analysis and Machine Learning Regression in OML 83, Edo State

Gas lift system performance optimisation becomes increasingly challenging as reservoir pressure declines and reservoir inflow performance relationships evolve during field maturation. Combining conventional nodal analysis with machine learning regression offers a pathway to adaptive optimisation that accounts for the dynamic relationship between reservoir behaviour and surface injection parameters. This study analytically evaluates gas lift performance optimisation approaches using nodal analysis and machine learning regression applied to production data from OML 83, Edo State, South South Nigeria. Production histories, wellhead pressure records, gas injection rates, and periodic inflow performance tests from forty-six gas lift wells spanning 2017 to 2023 were compiled. Conventional nodal analysis using Prosper software was applied to establish current operating points and identify wells with suboptimal gas injection rates. A random forest regression model was trained using 80 percent of the well-period dataset and validated on the remaining 20 percent to predict optimal gas injection rate as a function of current reservoir pressure, water cut, wellbore geometry, and wellhead back-pressure. The random forest model achieved a root mean square error of 0.42 million standard cubic feet per day in injection rate prediction versus an RMSE of 0.91 for the nodal analysis approach when applied without updated inflow performance data. Integration of the machine learning predictions with a real-time injection controller interface is demonstrated for three pilot wells. Keywords: gas lift optimisation, nodal analysis, machine learning, OML 83, Edo State.