SMART PETROLEUM SYSTEMS: LEVERAGING MACHINE LEARNING AND DIGITAL TWINS FOR ENHANCED OIL RECOVERY AND GAS EMISSION CONTROL

The evolution to smart and sustainable oilfield operations requires the implementation of advanced technologies with the ability to maximize production while minimizing the environmental imprint. The current study hypothesizes a hybrid method that bridges machine learning (ML) and digital twin (DT) technologies for visualizing an intelligent petroleum system towards the maximization of oil recovery (EOR) and reduction of greenhouse gas emissions. A mixed-method methodology was followed, with an initial qualitative systematic review of the literature to identify main themes of predictive maintenance, flare monitoring, and real-time optimization. These results fed into developing a quantitative simulation model based on synthetic and public data. ML models such as Artificial Neural Networks (ANN), XGBoost, and Long Short-Term Memory (LSTM) networks were trained to predict reservoir performance and issue warnings for abnormal CO₂ and CH₄ emissions. The best models were incorporated into a DT prototype developed in MATLAB Simulink and Python that would emulate and control the most influential production parameters in real time. It was optimized with Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) and provided a 13.7% improvement in oil recovery and a 28.1% decrease in CO₂ emission at optimized conditions. The findings verify that ML-DT integration enhances proactive data-driven decision-making, enhancing operational efficiency and environmental responsibility. The study provides a scalable framework for smart petroleum system deployment and enriches the literature in digital transformation for the energy industry.