Dynamic process intensification: Fundamentals and implementation to ternary distillation

Abstract

Distillation remains a key technology for separating liquid mixtures. Its versatility comes with the disadvantage of high energy consumption. We previously used empirical arguments to introduce dynamic process intensification (DPI) as a strategy for improving the energy efficiency of binary distillation. In this article, we focus on ternary distillation; we begin by providing a rigorous basis for DPI, then formulate the problem of identifying the operating states for DPI as a nonlinear optimization problem. Via an extensive case study considering a hydrocarbon mixture, we demonstrate that reboiler energy use can be reduced by more than 2.3% relative to an equivalent column operating at steady state. We prove that this result is due to the fact that DPI aims to satisfy product flow and quality constraints on average in time, affording broader opportunities for optimizing column economics than steady‐state operation, where constraints are met strictly at all times.

Dynamic Process Intensification via Data-Driven Dynamic Optimization: Concept and Application to Ternary Distillation

“…Here, we will focus on a new DPI concept that explicitly considers the process dynamics in identifying a (more economic) transient operating pattern. Below, we follow the same notation used in ref to illustrate the proposed dynamic optimization-based DPI (Do-DPI) in the context of operation at a reference steady state and static DPI. The reference/nominal steady-state operating point is labeled as Π* and characterized in terms of three variables: quality variable(s) χ*, operating condition(s) Ω*, and operating cost κ*, combined as .…”

“…The DPI concept has been applied to instances of conventional distillation columns, demonstrating, for example, 1.6% total energy savings for a propanol−acetic acid mixture, 7 1.4% savings for a methanol−1-propanol mixture, 9 and 1.7% savings for a cyclohexane−toluene−m-xylene mixture 10 with minimal extra capital investment, which consists of buffer tanks for blending the auxiliary products. The aforementioned savings were based on determining a feasible/optimal pair of auxiliary states through investigating the static nonlinearity of the respective columns and did not account for the impact of dynamic transitions when designing the auxiliary states.…”

“…■ DO-DPI OF A TERNARY COLUMN SEPARATING CYCLOHEXANE−TOLUENE−M-XYLENE USING A DATA-DRIVEN LOW-ORDER MODEL In our previous work, 10 we applied static DPI (4) to a ternary distillation column separating cyclohexane (C6)−toluene (TOL) and m-xylene (mX), achieving total (i.e., considering both condenser and reboiler duties) energy savings of nearly 1.7% relative to the nominal steady state, and a 2.3% reduction in energy use when considering only reboiler duty. In this section, we utilize the same distillation column as a testbed for the proposed Do-DPI strategy.…”

“…An upper bound of about one week was set based on the results for static DPI presented in our previous work. 10 The dynamic optimization is subject to six tight endpoint quality constraints, which force the time-averaged product qualities and flow rates to be within 0.5% of the reference state values for all three product streams.…”

“…Second, the values of the biases were updated by solving the steady-state design optimization problem in eq 3 and allowing the decision variables to change within the same limits specified in problem (13), a strategy that is described in more detail in our previous work. 10 This imposes an "asymmetry" in the bias terms relative to the steady-state values; in this case, this asymmetry consists of lowering the column pressure, with a favorable impact on energy consumption. With the bias values updated based on the static optimization calculations, only the parameters of the sine wave component of the inputs were used as decision variables in solving (13).…”