Design and Implementation of a Digital Twin for Live Petroleum Production Optimization
The specific use case chosen in the paper for showcasing the methodology is related to set-point changes on the artificial lift 4 equipment of the well for optimizing the production system . Examples of such artificial lift equipment include : Electrical Submersible Pumps ( ESP ), Rod Pumps , Gas Lift , and Plunger Lift .
DIGITAL TWIN SCHEMATIC
The design and implementation of a digital twin for dynamic set-point optimization on a petroleum production system using live IIOT data consists of several steps . These are :
1 . Field Data processing : Collection , profiling , clean-up , transformation and cloud-database maintenance 2 . Simulation : Automated cloud-database triggered field data relevant simulation 3 . Inverse modeling : a . Connecting real-world IIOT data with simulations to learn system unknowns b . Evaluation : Estimate how closely the digital twin mimics the real-world asset from history c . Calibration : Implement initial steps using insights from digital twin to account for uncertainty 4 . AI Model Recommendation : Deploy automated recommendations for set-point adjustments with updates based on dynamic trending of the asset
This paper focuses on the automation of the first two items of this process : Data Processing and Simulation . These steps are described in the context of feeding an AI engine that further consists of inverse modeling and model generated recommendation system . The details of the Inverse modeling and AI model recommendation components are beyond the scope of this paper .
Figure 1 represents a schematic of the overall process . It is important to note that this is a closedloop ongoing process and not a feedforward sequence of steps that ends in a recommendation . This distinction is important for two main reasons :
a . After a model-generated recommendation has been implemented , an effective digital twin that is a live virtual representation of a physical system needs to identify changes in operating state , record and evaluate the response and trigger an ongoing cycle involving data processing , simulation , inverse modeling to adjust the system in case if the previously provided recommendation needs to be followed up with a new recommendation . b . The digital twin can evaluate the impact of all historic set point changes and fine-tune the recommendation system .