i-CDU is the first application of process digital twin technology for atmospheric and vacuum columns. The product is focused on the production and operational needs of the leading refining unit, the crude distillation column, and it utilises artificial intelligence, neural networks, and other technologies to accurately model and optimise the unit's dynamic production process in real time. This allows for a deep understanding of changes within the unit, optimising its closed-loop operation, ensuring smooth performance, improving product yield and reducing energy consumption.
Based on a unified model, the product offers a decision-making foundation for operations, planning and management. It enables integrated management and control of atmospheric and vacuum production processes.
| Decision making |
Visualised decision making analysis
|
||
| Planning Management |
Monthly/weekly planning optimisation
|
||
| Production Management |
|
||
| Operation |
|
Digital Twin
The process digital twin primarily utilises artificial intelligence (AI) technology to model the unit. Both the fractionation towers and heat exchange network are represented by artificial neural network (ANN) models. The main model comprises 500 to 800 artificial neuron networks, containing 3 to 4 million parameters and 80,000 to 120,000 neurons. Additionally, the core model of the process digital twin integrates ten other models, including the material balance model, the energy balance model and the pressure drop empirical model for the entire vacuum column.
Augmented Database for Model Training
The atmospheric and vacuum AI models consider changes in crude oil properties, product qualities, equipment performances, operating parameters and other indicators beyond the unit's actual processing range. It enhances operating data based on the enterprise's real-world conditions. Historical data from thousands of site scenarios are amplified to the order of millions, providing a comprehensive and robust dataset.
Real-Time Simulation
During steady-state operation, the i-CDU product calibrates the current unit operations in real time every 30 minutes, outputting over 1,200 operation parameters. It systematically corrects on-site instrument readings to predict 600-800 unmeasurable data points. This includes critical information such as the atmospheric tower over-vaporisation rate, vacuum tower wash oil spray density, gas and liquid phase loads of key trays/packings, pump-around cooling ratios, and the overlap of adjacent product properties.
When processing the same oil from the same tank, without other external interferences, the process digital twin models reveal significant fluctuations in crude oil properties over time. This underscores the challenge of optimising operations in real time, where the feed is constantly changing.
Intelligent Decision-Making Centre - System-Wide Optimisation
The product performs data reconciliation and real time optimisation every 30 mins, considering on-site equipment constraints and integrating the unit's long-term operational requirements. A tailor-made optimisation strategy is applied to dozens of adjustable on-site variables, aiming to enhance the unit's economic benefits. Using a mathematical programming algorithm, the optimisation process determines the direction and magnitude of changes in operating conditions, providing control targets for iES intelligent execution. The model accuracy is comparable to that of the rigorous simulation, with a solution speed of under 5 minutes and an optimisation execution frequency of every 0.5 to 1 hour.
