(177c) “Closing the loop�: A practical apparoch to achiving Data-Driven feedback control in cellulosic fermentations based on mid-infrared spectroscopy

“Closing the loop”: A practical apparoch to achiving Data-Driven feedback control in cellulosic fermentations based on mid-infrared spectroscopy.

Pau Cabañeros Lopez¹, Isuru Udugama¹, Sune Tjalfe Thomsen², Helena Junicke¹, Krist V. Gernaey¹

¹ Process Systems Engineering Research Center (PROSYS), Department of Chemical and Biochemical Engineering, Technical University of Denmark, Søltofts Plads, Building 229, DK-2800 Kgs. Lyngby, Denmark.

² Department of Geosciences and Natural Resource Management, University of Copenhagen, Frederiksberg, C, Denmark.

The operation of fermentation-based fuel and chemical production processes utlizing lignocellulosic substrates can be challenging due to the inherent variation in the feedstock which contains various inhibitors as well as mixed carbon sources. To minimize the toxic effects of the inhibitors and hence maximize the realized yield lignocellulosic-based fermentations are operated in fed-batch mode with a slow feed of substrate into the fermenter. However, the inability to monitor feedstock variations and take necessary control actions in “real time” results in a loss of production efficiency and equipment utilization. This study addresses these practical needs by developing a cost-efficient process control solution that can monitor and control lingocellulosic fed-batch fermentation in real time. The solution developed uses an ATR-MIR spectrophotometer to collect on-line spectral data which is used to predict the glucose concentration inside the fermenter using a calibrated partial least squares (PLS) model. The glucose concentration is then set to an advanced supervisory control stratergy based on a PID control block that in real time adjusts the flow rate of a peristaltic pump to keep the concentration of glucose inside the fermenter at a set-point.

This approach was validated experimentally in four cellulose-to-ethanol fermentations using a xylose-consuming strain of Saccharomyces cerevisiae and hydrolyzed Danish wheat straw as a substrate in 2.5L Sartorius BIOSTAT A reactors. The developed advanced supervisory control strategy was set up and then further improved over the fermentations to be able to ensure a fast and accurate disturbance rejection even under the complex non-linear kinetics of the system and in presence of measurement noise. The validation experiments showed that by keeping the concentration of glucose at a low set-point, it is possible to ensure the simultaneous detoxification of the inhibitors and the co-consumption of the different carbon sources, resulting in a greatly improved space-time yield and similar final product concentrations compared to standard operation. Specifically, the batch process required 30 hours to ferment 1.3 L of hydrolysate while the controlled fed-batch only required 19 hours to ferment 2.0 L of hydrolysate with the same initial inoculum size, and the same ethanol yield. Moreover, it is also important to note that the strategy proposed is robust to substrate variability and allows to operate the fermenter without the need of predefining feeding strategies.

This work also demonstrates how the on-line monitoring of critical process variables coupled with robust and practical control schemes can contribute to optimize and automate complex fermentation processes with high substrate variability and mixed carbon sources to ensure consistent production.