Reduced dynamic modeling approach for rectification columns based on compartmentalization and artificial neural networks

Abstract

The availability of reduced‐dimensional, accurate dynamic models is crucial for the optimal operation of chemical processes in fast‐changing environments. Herein, we present a reduced modeling approach for rectification columns. The model combines compartmentalization to reduce the number of differential equations with artificial neural networks to express the nonlinear input–output relations within compartments. We apply the model to the optimal control of an air separation unit. We reduce the size of the differential equation system by 90% while limiting the additional error in product purities to below 1 ppm compared to a full‐order stage‐by‐stage model. We demonstrate that the proposed model enables savings in computational times for optimal control problems by ~95% compared to a full order and ~99% to a standard compartment model. The presented model enables a trade‐off between accuracy and computational efficiency, which is superior to what has recently been reported for similar applications using collocation‐based reduction approaches.

Renewable production of ammonia and nitric acid

“…In real life, the plant operator needs to exploit economic incentives from the intermittent supply and at the same time meet the demand, while coordinate up‐ and down‐stream processes and manage the inventory in between. To achieve this, we expect three essential technology advances: first, a cross‐sector, real‐time capable information structure that bridges the production, the management and the external players (e.g., grid), as proposed by Backx et al; second, tailored models that capture transient behaviors on appropriate time scales of supply and demand, as demonstrated by Schäfer et al; and third, optimization‐based decision tools under uncertainties that are computationally tractable and robust, as reviewed in Dias and Ierapetritou . More open questions and challenges on this topic are reviewed by Daoutidis et al and Mitsos et al At the end, it is the plant manager who needs to welcome intentional dynamics and enable the changes.…”

A framework of hybrid model development with identification of plant‐model mismatch

“…Algorithms in supervised training play an important role, as they dictate the black‐box training process and therefore heavily impact the performance. Among all algorithms, artificial neural network (ANN) and support vector regression (SVR) have been employed in many chemical engineering applications due to its relatively easy implementation, high flexibility and adaptability 4,6,17,47‐51 . When the hybrid modeling concept was first developed in chemical engineering field, ANN was the first algorithm to be used 45,52,53 …”

The Potential of Hybrid Mechanistic/Data‐Driven Approaches for Reduced Dynamic Modeling: Application to Distillation Columns

“…To overcome this limitation, they propose the use of explicit functions interpolating between precalculated solutions, which exhibits an unfavorable scaling with both the level of detail of the table and the number of components in the column. However, we recently extended this concept by integrating machine learning techniques [22].…”

“…We herein confine to discussing its properties. A detailed derivation as well as a numerical study on its computational performance can be found in [22]. .…”