Digital Twin for Autonomous Decision-Making in Gas Lift Operations: Improving Reliability and Adaptability

Abstract

A virtual representation of a complete cyber-physical system introduces several opportunities, such as enabling real-time monitoring of physical systems and ongoing learning to deliver accurate and dependable information. This approach is often referred to as the creation of a digital twin (DT). Nevertheless, challenges emerge, particularly with the computational requirements of implementing AI-driven models in real-time data exchange contexts, as is common with DTs. This research presents a DT framework tailored for optimal and autonomous decision-making within a gas-lift process, with a focus on increasing the adaptability of the DT system. The proposed solution integrates Bayesian inference, Monte Carlo (MC) simulations, transfer learning, and online learning alongside techniques for dimensionality reduction and cognitive modelling. These approaches contribute to the development of a reliable and efficient DT. The framework aims to deliver a system that can adapt to dynamic environments, account for prediction uncertainties, and improve decision-making processes in complex, real-world applications.