ethanol, The distillation and dehydration operations consist of three columns and a parallel grouping of molecular sieves

The distillation and dehydration operations consist of three columns and a parallel grouping of molecular sieves (Figure 1).

http://www.chemicalprocessing.com/Media/0710/miw_fig1.jpg
The distillation system is fed with the fermentation process product, a mash that contains 10% to 14% alcohol, water and all non-fermentable solids from the corn and yeast cells (A). The first column, called the beer column, splits the mash into 190-proof (95 vol. %) alcohol and a residue mash called stillage (B). Whole stillage is transferred from the base of the beer column to the co-product processing area. The alcohol goes to a second column or rectifier. The bottoms from the rectifier pass to a side-stripper column  it sends alcohol back to the rectifier, while its bottom stream is recycled to the liquefaction process to reduce heat consumption.


Meanwhile, overhead from the rectifier goes to the dehydration step to yield product that exceeds 99 vol. % alcohol. Without that final drying process, the ethanol produced by fermentation wouldn稚 be a viable fuel.


BSE痴 dehydration process consists of three vessels, each containing Type 3A zeolite beads, which require regeneration. The units cycle through a sequence of in service, ready and regeneration.


Pressurized, superheated 190-proof alcohol from the distillation system is fed into one of the vessels. Water is exothermically adsorbed on the beads while ethanol passes through the bed. When the beads near full loading, the feed alcohol is diverted to a second vessel ready to take over. The anhydrous ethanol vapor goes to the process for heat recovery, is cooled back to a liquid and sent to storage tanks. Meanwhile, the molecular sieves in the first vessel are regenerated, removing the water by adding energy with recycled product.


The molecular sieves water removal ability is impacted by feed temperature, pressure, residence time and composition. The APC application was needed to improve the performance of the system by controlling these variables to achieve maximum water removal at lower energy consumption.


Achieving effective control


Dealing with azeotropic ethanol/water separation and the pressure-swing operations of the molecular sieves required a complex model. Pavilion痴 application relies on a nonlinear Model Predictive Control (MPC) methodology  a control algorithm, based on a dynamic model of the process, predicts and optimizes the process痴 future response. The MPC algorithm finds the process痴 optimum operating point by computing a sequence of manipulated variables moves in each control interval. MPC can handle the constraint challenges and disturbances of a highly nonlinear, multivariable system such as the distillation/sieves arrangement in the ethanol process.


The BSE application relies on a combination of first principles and empirical models identified from input/output measurements. These models where created with Pavilion痴 state-of-the-art empirical modeling technology combined in a hybrid fashion with well-known fundamental models of ethanol/water separation and sieve adsorption.
Empirical modeling alone is often used in MPC because solving a set of complex differential equations in first principles models for the calculation of the optimal sequences at each control interval can be unfeasible.

However, first principles models can add great value to the application by providing understanding of process behavior outside of previous operating ranges (e.g., at 5% higher capacity). Pavilion痴
application for the ethanol industry combines comprehensive empirical models and available first-principles models with a constraint-handling multivariable control algorithm.


The first principles models were developed based on heat- and mass-transfer systems and validated using operating data from plant tests also used to develop the empirical models.


The MPC application obtains the current process information and, based on its dynamic model, calculates a sequence of future process moves over a specified time period. An optimization algorithm determines the correct of the manipulated variables so that several objectives can be achieved simultaneously. The control system dynamically compares the predicted trajectory with the process response and corrects for any differences detected. Because of the dynamic trajectory compensation and the taking into account of process constraints, the control system is stable and operationally reliable. Using small steps, the stable process is led gradually to its optimal operating point while larger steps can bring prompt corrections if disturbances affect the system.