Design and Implementation of a Digital Twin for Live Petroleum Production Optimization
to implementing the change . This is also a step towards expert augmented machine learning 6 , where the feedback provided by the expert on the model generated recommendations is utilized to adapt the model to provide improved recommendations in subsequent rounds .
Fig . 9 : Flagging changes and recording the response to the stimulus
SIMULATION
Having covered the collection and processing of physical data , in this section we proceed to describe the methodology for automated generation of virtual data . Simulation represents the virtual data generating component of the digital twin . The oil and gas optimization literature is rich in description of simulation based artificial lift and gas lift set-point optimization approaches . The paper by Rashif et al . 7 summarizes a survey of different gas lift optimization techniques using simulation as a basis . Borden et al . 8 presented a surveillance and workflows-based approach for gas lift optimization . Surendra et al . 9 presented further progress in the field by automating combining the analytics and physics-based modeling / simulation approaches through a case study