(76f) An Affordable Non-Linear APC Technology for Distillation Columns Using First Principle Models

Standard linear APC solutions have been widely applied on distillation columns for decades. However, for many columns the use of non-linear APC technology can bring a lot of extra benefits. This presentation will clarify how this can be achieved using a new technology called INCA MPC4DISTILLATION.

Traditionally, in the context of an APC project, dedicated (automated) plant tests are performed based on which empirical linear models are derived.  This can take anywhere from days to weeks, depending on the complexity and time-to-steady-state of the column under consideration.  Despite this significant investment, the resulting models – due to the fact that these models are linear – will only be sufficiently accurate in a limited operating range and – due to the fact that these models are empirical – will provide limited additional insights into the internal operation of the distillation column.  In distillation columns with very slow dynamics and/or complex and non-linear dynamics this simplification can turn out not to be very favorable.

This paper will discuss a new technology that can deliver a significantly better payback time for a wide range of distillation columns.  This technology uses a combination of historical plant data, column design properties and physical principles to construct an accurate non-linear process model, which is integrated into the INCA non-linear APC controller.  This approach has several direct benefits:

1. More accurate models. The resulting models will be more accurate and valid over a much wider operating range.  For columns that are operated over an extensive range of throughput and quality specs of the main product streams this will result in improved APC performance and higher and more consistent benefits.
2. No need for plant tests. Since the modeling technology mainly relies on physical relationships and historical data, no dedicated plant tests are needed.  The benefits in terms of reduced costs of this approach are obvious.
3. Control of more complex columns.  For high-purity columns with tight composition specifications and e.g. azeotropic distillation the use of first principle models will enable the user to more easily obtain robust control performance, which would otherwise be hard to achieve.
4. Improved process insights.  Due to the physical nature of the models, the user gets soft-sensing capabilities almost for free.  The ability to visualize column temperature, pressure, composition,… profiles in real-time will provide additional insights to the APC engineers, process engineers and operators and will lead to improved technology acceptance and process operation.

This presentation will give an overview of how this new technology is set up and will illustrate the results obtained on a high-purity distillation column.