Demystifying Plant Simulation Models

The proliferation of data-driven business principles has brought us new ways of improving performance, developing scenarios, and uncovering latent trends.

In refining alone, numerous models exist to design new construction, troubleshoot, optimize production, and simulate changes in operating conditions. There are steady-state models, dynamic models, two-dimensional and three-dimensional models, hydraulic models, economic models, logistic models, atmospheric dispersion models, and the list goes on.

While this mess of models might seem confusing, they can be bifurcated into two simple categories. Each category comes with a set of data sources, requirements, and use cases.


The first type of model is a property-based model.

These models use the process’ chemical and physical properties to estimate a refinery’s unit operation.

They are especially useful when designing new equipment and processes, troubleshooting current operations, and modeling alternate scenarios. Process adjustment simulations in property-based models allow teams to explore operations scenarios without risking production or safety. This makes them a critical decision-making tool.

Software such as Aspen HYSYS aids in creating process simulation models | Photo Credit: Aspen HYSYS

Process simulators are one example. Process simulators describe processes in flow diagrams where unit operations are positioned and connected by product or educt streams. Typically aided by software, model owners attempt to solve mass and energy balance to find a stable operating point.

Because a refinery’s unit operations are complex, property-based models generally are limited to sections, or “units,” within a refinery, although refinery-wide models do exist. Property-based models may also be more granular than a unit. Certain models, such as dynamic simulation or computational fluid dynamics, call for more detailed input and higher computation requirements.

Example of a computational fluid dynamics model (CFD) simulating incompressible airflow through a wind turbine | Photo Credit: Simscale

The heightened level of detail, processing power, and the need to fit results to operational data typically require property-based models to run offline.


The second type of model is a data-based model.

Data-based models have a wide scale of impact throughout industrial settings. They may guide high-level refinery operations amidst changing market conditions or help operations improve performance in real-time. Refiners use these models in combatting the constant challenge to optimize economic and operational performance.

Data-based models incorporate large volumes of diverse information such as cost and composition of inputs, refinery productivity and constraints, and downstream sales prices to guide operations. These models generally rely on parameters from historical correlation and experimental data, which can have a limiting effect.

APCs have a wide range of analytical capabilities for operations support | Photo credit:

A common data-based model is a linear program (LP) to simulate site-wide economic planning functions, including a current, short-term, and long-term approach to maximizing margins, forecasting financials, evaluating feedstock, and material logistics planning. Data-based models also support operations, such as process control models (i.e., advanced process control (APC) & operator training simulators (OTS)). Process control models offer real-time process operations performance management to improve optimization against process constraints such as throughput, yield, and product quality, among other variables.

Because these models generally require less computational power, they generally apply to larger scopes (e.g., plant-wide) than property-based models. These simulations can be used in both online & offline environments since their input data is already fitted to operational data.


It is readily apparent that planning, design, and operations will continue to be rooted in data-based decision-making. We'd be remiss not to mention that the quality of the data sources (both operational data and property-based data) has the highest impact on their ability to project scenarios and improve performance. As we like to say, "garbage in, garbage out."

We believe there will continue to be opportunities to blend these two types of models. A world of new opportunities exists to infuse data-based models with more granular operations data. This operations data should be sourced and managed from a solid data foundation with effective management and integration of asset data. This model-mending will improve logistics and execution, strengthen economic planning, design smarter projects, and build a safer, more effective, and more efficient operation.