Plant-Wide Optimization of Sterling Ethanol LLC

By Srinivas Budaraju, Maina Macharia and Dave Kramer | June 02, 2008
Sterling Ethanol LLC was built to help meet the United States' growing demand for ethanol and to diversify agricultural investment in the northeastern Colorado region known for its high corn yields and large-scale cattle feedlots. Owned and operated by local investors, the 40 MMgy facility started up in November 2005, just seven months after the start of construction. The feat is a record for the ethanol industry.

The facility, located two hours northeast of Denver, is accomplishing its goals. Since the first truck of ethanol left the plant Nov. 23, 2005, it has continued to supply fuel and improve corn prices for the region. However, plant management felt it could do more, leading to a project that implemented a plant-wide advanced process control (APC) solution to maximize production, monitor yield and minimize energy costs per unit of ethanol. The project resulted in an 18 percent increase in throughput. The following describes the facility's process design and production challenges, and the steps taken to improve them.

Process Description
Sterling Ethanol is a dry-grind ethanol plant employing the unit operations shown in Figure 1. The mash preparation has four key sub-processing units: milling, slurry mixing, cook and liquefaction. Sterling Ethanol's corn is ground and hydrolyzed with hot water in the slurry mixing tanks. Enzymes are also added. The corn is further fractured through the cook section by pumping it through a hydroheater, providing the enzymes more access to the starch. In the cook section the corn slurry mixture is heated to sterilize the fermentation feed and to activate enzymes that make the starch more soluble. The excess heat is used as direct steam injection and is the heat source for the third distillation tower and side stripper in the distillation units. The flow from the cook process goes to holding tanks where glucoamylase enzyme is added to break down the starch into glucose. Yeast digest the glucose and convert it to ethanol.

The liquefaction slurry is pumped through heat recovery exchangers as fermentation feed. Sterling has several large fermentors and each is allowed to ferment for 45 to 50 or more hours depending on production rate decisions.

The fermentation process is controlled based on sugar availability, enzyme dosing strategies and temperature management during fermentation. After the fermentation is complete a fermentor product is "dropped" to a large beer well. This vessel is continuously fed to a beer column. The beer well is the feedstock reservoir for the distillation unit, and it allows distillation operation to run continuously without having to slow down or speed up for the batch fermentor cycles.

The beer column feed contains approximately 10 percent unfermentable solids, 12.5 percent ethanol and 87.5 percent water. The Sterling plant column's overhead vapor feeds a vacuum rectifier distillation column while the bottom stream consisting mostly of water and solids (known as whole stillage) is directed to inventory tanks. We will follow the alcohol process flow and then return to the stillage processing.

In the rectification tower, the alcohol is separated close to the azeotropic point. At the azeotropic point, ethanol and water cannot be separated any further by distillation. This product is referred to as 190-proof alcohol (95 percent alcohol by volume).

From an intermediate inventory tank the 190-proof product feeds three parallel molecular sieve absorbers after a sieve feed vaporizer to dehydrate the alcohol to less than 1 percent water by weight.

The side stripper recovers the trace amounts of ethanol off the bottoms flow. The side stripper at Sterling receives its feed and overhead cooling from the rectifier bottoms flow controller. It is reboiled using vapors from the cook flash tank or steam directly from the boiler systems. Water for the side stripper bottoms is recycled back to the cook water section.

Stillage off the beer column bottom is pumped to a whole stillage tank. The whole stillage is then processed into cattle feed that has a maximum moisture content specification for the distillers wet grains with solubles. This is done through centrifuges and evaporators. The DWGS is then sold to local cattle feedlots.

Liquid off the centrifuges is known as thin stillage. Thin stillage must be concentrated to syrup before it is added to the DWGS. This is done with two-stage evaporators. Some of the thin stillage is recycled as backset to the fermentation feed system to minimize freshwater use. Backsets also have some nutritional value to the fermentation operation.

Production Challenges
When Sterling encountered challenges increasing production rates, management considered resolving them internally with engineering studies and operational tests. Both consume time and require capital decisions. APC technology is commonly applied in refining and petrochemical plants and is known for its model predictive control of processes for better coordination of controls, allowing for higher efficiencies, yields and production rates. Simply increasing fermentation rates and dropping fermentors faster does increase production, but has a negative impact on ethanol yields. Model predictive control can simultaneously improve yields and capacity, and a calibrated plant model can be used to trade-off capacity and yield for best economic results. Pavilion Technologies agreed on the need to optimize the entire plant operation with APC technology.

Figure 1

Dry-grind ethanol plant production flow

Being a new plant, there were several operational challenges, including low fermentor yields, overpurified ethanol products and low throughputs. The objective was to ramp up the profitability of the plant.

Plant APC Optimization
The APC solution called for three controllers connected together so that when one changes the feed rate or other objectives, the other controllers follow suit. This coordinated control ensures smooth operation and serves to identify bottlenecks that can be eliminated by process modifications. The three controllers include one focused on the batch fermentation, which is the heart of the plant, and two more managing the continuous plant sections on either side of the batch fermentors. The cook/milling section is controlled to manage and stabilize fermentation feed, along with coordinating the fermentation feed on target with the rest of the facility. The unsteady-state, time-dependent batch fermentation quality controllers optimize end of batch yields within the current plant fermentor cycle time. A single back-end controller manages all continuous plant sections after the fermentor, including distillation, molecular sieves, evaporation, centrifuges and stillage tanks.

Pavilion's non-linear process optimizer was crucial for the success of the plant-wide optimizer because while the plant had an objective of increasing production rates, the fermentation batch yields more ethanol from the starch if allowed to remain longer in the batch. However, longer batch times reduce overall ethanol throughput, so tradeoffs have to be made between higher production rates and gradually decreasing yield. This necessitated the use of a non-linear optimizer.

Table 1

Fermentation improvements through the course of APC implementation

Optimization Results
The APC and plant optimizer system stabilized the plant, increasing overall production, yield and energy efficiencies. The distillation-stillage (back-end) controller was implemented with the control objectives to:
Increase ethanol production

Improve control of the distillation columns to produce 190-proof product within specification limits

Operate the pressure-swing adsorption molecular sieves to produce 200-proof alcohol at the lowest operating cost and thus increase sieve ethanol yields by producing product at specification limit

Manage the molecular sieve feedstock inventory tank, utilize its surge volume and to link sieve production rates to the distillation tower feeds. This allows for steadier plant controls in the distillation section.

Respond to disturbances, especially those caused by fermentation drops to the beer well and beer mash train exchanger switches

Utilize the surge volume of the syrup tanks, and whole and thin stillage.

The specific energy of the plant's distillation section is controlled by the addition of the right proportions of heat/feed ratio. Changes in the feed to the distillation tower provide feedfoward information. The downstream rectification section's heat and evaporator heat sources are adjusted to provide the optimum heat. All relationships of 190 proof, stripper distillation tower temperature profile and the hydraulic limits are accounted for, allowing the APC to make correct adjustments to keep these variables on specification. The APC project helped identify operational problems with the side stripper distillation tower separation. However, the APC system was able to operate even with malfunctioning equipment and still stabilize the operation. On a mini-turn-around, the trays were cleaned and corrected, and the full potential of the distillation column realized.

When feed rates are adjusted to the distillation, the beer column bottoms (whole stillage) rates change. Inventory tanks in whole stillage and liquid residue of the centrifuge systems (thin stillage) are adjusted to ensure the distillation stillage. The APC solution only changes rates within the tank limits with a calculated estimate of upper and lower inventory capacities calculated over a predictive time horizon. In addition, the evaporator's energy requirements and distillation energy requirements are balanced by the APC strategy.

The ethanol loss via the side stripper bottoms decreased so significantly that the methanator, which wasn't working properly before the controller installation, started working efficiently. 200-proof product operated closer to the maximum 0.9 percent Karl Fischer upper limit, allowing for an increase in ethanol yield of the distillation process.

The specification limit for certified storage tanks of ethanol is a Karl Fischer value and the APC was able to run at the limit without violating it because of the predictive capability of the distillation-stillage APC controller.

The difference between the pre-APC mean value and the post-APC mean value shows a 46 percent increase in the Karl Fischer values of the final ethanol. The post-APC mean value for the Karl Fischer is run at the specification limit.

Slurry/water balance APC (continuous) were fundamental to the improvement of the fermentation process. The slurry controls the quality of feedstock to the fermentation process.

It accomplishes the following objectives:
Balances load and energy between mills to achieve efficient milling

Manages the fermentation inventory (fermentation gap control) with the water balance

Controls the liquefaction solids and slurry percentage solids to provide consistent slurry and liquefaction solids to the fermentors

Optimizes backset to ensure a consistent feedstock to fermentation

Manages energy utilization with backset

Controls the contributions of water sources to maintain a consistent water quality in the fermentation feed

Stabilizes slurry pH and temperature conditions to cook and liquefaction

Slurry solids variation was reduced substantially, providing a more stable input to the fermentors. This increased the activity and reduced stress factors on the yeast to improve the fermentation. Table 1 shows the changes in the slurry solids to the fermentors, indicating a significant reduction in the variation of the solids content.

Process Optimization Results
The APC project is showing production improvements. The controller uptime is near 100 percent, which indicates a high operator acceptance. The payback period for the project is a few months.

The batch fermentation APC is most difficult to control as it is not solvable with traditional APC approaches. However, Sterling's plant personnel and Pavilion focused on this section to run tests and trials to achieve sequential improvements through the project implementation staging.

The fermentation controller had the following control objectives:
Maximize end-batch ethanol concentration

Calculate optimal temperature profile

Accelerate and control starch hydrolsis to glucose

Accelerate and control glucose fermentation to ethanol

To calculate optimal glucoamalyze enzyme feed strategy.

Any small improvement in Sterling's fermentors and beer well has a significant impact on the overall profitability of the plant. In addition, each fermentor holds significant corn inventory such that intelligent, designed plant testing is required to avoid losing revenue and weeks of infections and/or stillage handling issues from a single "bad" fermentor.

Many fermentation control challenges initially existed. Fermentation is a biological process and not easy to model because of numerous trace quality factors that impact the yeast activity and the fermentation environment. In addition, there are many unmeasured disturbances variables: corn and water quality, and unknown fermentable starch quantities. The data used to model, which is historically archived by plants, was sparse due to the sampling regiments. U.S. dry-grind ethanol plants use high-performance liquid chromatography (HPLC) data which is only sampled at set schedules. Sampling and HPLC analysis take hours, so this lab-quality data was not sufficient for real-time control of the plant. Online in-situ analyzers to measure accurate key fermentation product qualities, dextrins, glucose, lactose, acetic acid and ethanol were not available. Also the process was known to be nonlinear with complex interactions of all feed quality and environmental conditions of the fermentors.

Pavilion leveraged first-principle modeling using scientific principles of biological fermentation processes in a hybrid combination with an analysis of the HPLC empirical data to ensure that the fermentation models were customized for the specific feedstock. The models were then run to establish the optimum staging for glucoamylase and temperature trajectory modulations. Pavilion developed an innovative unique batch model predictive control algorithm and online HPLC data calculations that allowed for the implementation of real-time quality control for the fuel ethanol fermentors. It's believed that the fuel ethanol industry is unique in this advanced use of such a batch model-based control algorithm.
Table 1 summarizes the fermentation improvements through the course of the project implementation.

The most dramatic improvement in the process was the overall production rate which increased by 17.6 percent production increase in the plant.

Energy improvements that are generally demonstrated from the APC were not auditable because the plant added a steam turbine, which increased energy costs (natural gas) but produced electricity for the facility and therefore reduced electrical power import. This was a significant cost saving to Sterling Ethanol.

Overall, the APC project achieved an increase in fermentation ethanol yield, an 18 percent increase in production rate, improved energy efficiencies (impacted by significant process change), and higher ethanol and DWGS quality control.

The APC was able to hold plant capacity above 18 percent of the pre-APC performance. The APC project has improved the plant controls substantially and achieved the economic objectives set out by Sterling Ethanol. The controller uptime is near 100 percent, which indicates a high operator acceptance.

Srinivas Budaraju is a senior applications engineer with Pavilion Technologies. Maina Macharia is the manager of project engineering for Pavilion Technologies. Dave Kramer is the president/general manager and director of Sterling Ethanol LLC.