Digital Twin Architecture for Refineries Using Petroleum Refining Library

       A refinery digital twin is more than a process simulation model. Depending on the engineering objectives, refinery digital twins can be organized into different architectural model classes. While simulation reproduces material flows and equipment behavior, real operational decisions also require optimization, forecasting, production planning, and scenario analysis.
       Petroleum Refining Library (PRL), built on AnyLogic, is a software framework for developing digital twins of refineries and other industrial oil and gas production systems. At the core of every digital twin is a process simulation model that continuously maintains the current state of the production system. Analytical models access this shared state, solve specific engineering problems, and automatically update operating parameters, creating a closed-loop digital twin rather than a standalone simulation. This architecture enables engineers to combine process simulation, process optimization, and decision support within a single extensible digital twin platform.

       Process simulation accurately reproduces material flows, equipment behavior, and production dynamics. However, it cannot determine the optimal operating strategy. Refineries continuously solve engineering problems such as feedstock allocation, blend optimization, production planning, production scheduling, and scenario evaluation. These tasks require specialized analytical models that work with the operating state of the digital twin. In Petroleum Refining Library, process simulation and analytics are tightly integrated. The simulation model maintains the complete state of the production system, while analytical models retrieve the required data, solve specific engineering problems, and automatically update operating parameters. This creates a closed-loop digital twin where simulation and analytics continuously work together within a hybrid modeling framework that integrates continuous process simulation, discrete-event simulation, and optimization.

       Petroleum Refining Library organizes a refinery digital twin as a multi-layer architecture, where each layer has a distinct responsibility. This separation simplifies model development, improves maintainability, and allows simulation and analytical models to evolve independently.
The architecture consists of four layers:

• Data Layer – stores process data, equipment parameters, production plans, and simulation results.
• Process Simulation Layer – represents the physical production system through interconnected process simulation modules for continuous process simulation and material flow simulation.
• Analytical Layer – contains optimization, forecasting, planning, and custom engineering models that solve specific operational problems.
• Decision Support Layer – combines simulation and analytical results to support operational and strategic decision-making.

Unlike traditional simulation models, these layers operate as a single integrated digital twin. The process simulation layer continuously maintains the simulation state of the production system, while analytical models update operating parameters based on their calculations. This closed-loop architecture keeps the digital twin synchronized with production logic throughout the simulation.

       The Process Simulation Layer is the foundation of every digital twin built with Petroleum Refining Library. It represents the production system as a network of interconnected process modules linked by material flows. Each module is a digital twin of a physical asset, enabling dynamic process simulation of process units, tank farms, feed sources, blending systems, and loading facilities. Together, they reproduce refinery operations while continuously maintaining the current state of the production system.
Unlike generic simulation libraries, Petroleum Refining Library modules include industry-specific logic for oil and gas production, including material transformation, flow distribution, inventory management, and operational constraints. Since all modules are fully compatible with AnyLogic and Java, engineers can extend existing functionality or develop custom components that reflect the unique requirements of a specific refinery.
       The Process Simulation Layer serves as the single source of truth for the entire digital twin, providing the current production state used by all analytical models.

       The Analytical Layer transforms the process simulation model into an active decision-making tool. Instead of modeling physical processes, it solves specific engineering problems using the model state of the digital twin. Typical applications include feedstock allocation, blend optimization, production scheduling, production planning, production forecasting, and user-defined analytical models. Unlike external optimization tools, analytical models in Petroleum Refining Library are fully integrated with the simulation. They retrieve the required data from the digital twin, perform calculations, and automatically update operating parameters. This creates a closed-loop workflow where simulation and analytics continuously interact throughout the simulation.

       The Analytical Layer can incorporate different classes of analytical models depending on the objectives of the digital twin. Petroleum Refining Library provides an open architecture that allows multiple analytical approaches to operate on the same process simulation model.
       Optimization Models
       Optimization models determine the best operating strategy while satisfying technical and operational constraints using mathematical optimization techniques. Typical applications include feed allocation, product blending, production planning, and production scheduling. The interaction between optimization and simulation is described in Hybrid Simulation and Optimization for Refineries.
       Predictive Models
Predictive models estimate future process behavior and support production forecasting and operational planning. They can forecast production performance, inventory levels, equipment behavior, and other key process indicators.
       Custom Analytical Models
       Beyond optimization and prediction, engineers can implement refinery-specific analytical applications, including business rules, engineering calculations, economic evaluations, and decision-support algorithms. Because Petroleum Refining Library is fully compatible with Java and AnyLogic, these models can be integrated directly into the digital twin without changing the underlying process simulation.

       The Decision Support Layer combines information from the process simulation and analytical models to support operational and strategic decision-making. It provides a unified environment for evaluating scenarios, comparing operating strategies, and assessing the impact of engineering decisions. By separating simulation from analytics while keeping them tightly integrated, Petroleum Refining Library enables digital twins that are accurate, extensible, and adaptable to the unique requirements of each refinery.

       The Data Layer enables all layers of the architecture to operate on a consistent and synchronized representation of the production system. The Data Layer provides the information required by every component of the digital twin. It stores equipment parameters, material properties, production plans, operating constraints, simulation results, and historical process data while integrating information from ERP, MES, SCADA, process historians, and laboratory information systems. Rather than serving individual applications, the Data Layer supplies a common information foundation for the entire digital twin. This ensures that the process simulation, analytical models, and decision support tools operate with consistent and synchronized data.

       Every refinery has unique production processes, operating rules, and optimization objectives. While Petroleum Refining Library provides industry-specific simulation modules and analytical capabilities, it is designed to be extended rather than limited to predefined functionality. Built on AnyLogic and fully compatible with Java, Petroleum Refining Library allows engineers to develop custom process modules, simulation models, optimization algorithms, predictive models, business rules, and system integrations without modifying the core library. This enables organizations to implement refinery-specific logic while reusing the same digital twin architecture.
As operational requirements evolve, new analytical models can be added without redesigning the process simulation model, making the digital twin scalable, maintainable, and adaptable throughout its lifecycle.

       Petroleum Refining Library is more than a process simulation library. It provides a complete framework for building digital twins that integrate simulation, analytics, and decision support into a single engineering environment. The same framework supports Functional Object Models, Interaction Models, Enterprise Models, and Enterprise Network Models described in the refinery digital twin classification. By combining a shared process simulation model with extensible analytical applications, Petroleum Refining Library enables engineers to optimize refinery operations, evaluate future scenarios, and continuously improve production performance. Its open architecture, based on AnyLogic and Java, allows digital twins to evolve alongside changing operational requirements without redesigning the core simulation model.

1. What is a refinery digital twin?
A refinery digital twin is a virtual representation of a production system that combines process simulation, operational data, and analytical models. It enables engineers to analyze current operations, evaluate future scenarios, optimize production, and support decision-making.

2. What is Petroleum Refining Library?
Petroleum Refining Library (PRL) is a software framework for building digital twins of refineries and other oil and gas production systems. It provides industry-specific process simulation modules, analytical integration, and an extensible architecture based on AnyLogic and Java.

3. How is a digital twin different from a process simulation model?
A process simulation model reproduces the physical behavior of a production system. A digital twin extends simulation by integrating optimization, forecasting, production planning, and decision support into a single continuously updated environment.

4. What types of analytical models can be integrated with Petroleum Refining Library?
Petroleum Refining Library supports optimization models, predictive models, production planning algorithms, blending optimization, forecasting, and custom engineering applications. Additional analytical models can be developed using Java and integrated into the same digital twin.

5. Can Petroleum Refining Library be customized?
Yes. Petroleum Refining Library is fully compatible with Java and AnyLogic, allowing engineers to develop custom process modules, analytical models, optimization algorithms, business rules, and integrations with external enterprise systems.

6. Which refinery processes can be modeled?
Petroleum Refining Library can model entire refinery processes, including process units, material transportation, tank farms, product blending, loading facilities, feedstock supply, production planning, and other continuous production systems.

7. Does Petroleum Refining Library support optimization?
Yes. Petroleum Refining Library integrates optimization directly with the process simulation model. Optimization algorithms use the current state of the digital twin to determine optimal operating parameters, which are then applied automatically to the simulation.

8. Which software is Petroleum Refining Library built on?
Petroleum Refining Library is built on the AnyLogic simulation platform and uses Java for customization and extension. This allows organizations to leverage AnyLogic's simulation capabilities while adding refinery-specific process and analytical logic.

9. Can Petroleum Refining Library integrate with external systems?
Yes. Petroleum Refining Library can exchange data with enterprise systems such as ERP, MES, historians, laboratory information systems (LIMS), databases, and other industrial software through standard Java integration mechanisms.

10. Who is Petroleum Refining Library designed for?
Petroleum Refining Library is intended for process engineers, simulation specialists, optimization engineers, digital transformation teams, and software developers building digital twins for refineries and other oil and gas production facilities.

11. Is Petroleum Refining Library suitable for industrial digital twins?
Yes. Although primarily designed for refineries, Petroleum Refining Library can also be used to develop industrial digital twins for other continuous-flow oil and gas production systems.