Advanced modelling, monitoring, and process control of bioconversion systems
Schmitt, Elliott Carl
MetadataShow full item record
Production of fuels and chemicals from lignocellulosic biomass is an increasingly important area of research and industrialization throughout the world. In order to be competitive with fossil-based fuels and chemicals, maintaining cost-effectiveness is critical. Advanced process control (APC) and optimization methods could significantly reduce operating costs in the biorefining industry. Two reasons APC has previously proven challenging to implement for bioprocesses include: lack of suitable online sensor technology of key system components, and strongly nonlinear first principal models required to predict bioconversion behavior. To overcome these challenges batch fermentations with the acetogen Moorella thermoacetica were monitored with Raman spectroscopy for the conversion of real lignocellulosic hydrolysates and a kinetic model for the conversion of synthetic sugars was developed. Raman spectroscopy was shown to be effective in monitoring the fermentation of sugarcane bagasse and sugarcane straw hydrolysate, where univariate models predicted acetate concentrations with a root mean square error of prediction (RMSEP) of 1.9 and 1.0 g L-1 for bagasse and straw, respectively. Multivariate partial least squares (PLS) models were employed to predict acetate, xylose, glucose, and total sugar concentrations for both hydrolysate fermentations. The PLS models were more robust than univariate models, and yielded a percent error of approximately 5% for both sugarcane bagasse and sugarcane straw. In addition, a screening technique was discussed for improving Raman spectra of hydrolysate samples prior to collecting fermentation data. Furthermore, a mechanistic model was developed to predict batch fermentation of synthetic glucose, xylose, and a mixture of the two sugars to acetate. The models accurately described the bioconversion process with an RMSEP of approximately 1 g L-1 for each model and provided insights into how kinetic parameters changed during dual substrate fermentation with diauxic growth. Model predictive control (MPC), an advanced process control strategy, is capable of utilizing nonlinear models and sensor feedback to provide optimal input while ensuring critical process constraints are met. Using the microorganism Saccharomyces cerevisiae, a commonly used microorganism for biofuel production, and work performed with M. thermoacetica, a nonlinear MPC was implemented on a continuous membrane cell-recycle bioreactor (MCRB) for the conversion of glucose to ethanol. The dilution rate was used to control the ethanol productivity of the system will maintaining total substrate conversion above the constraint of 98%. PLS multivariate models for glucose (RMSEP 1.5 g L-1) and ethanol (RMSEP 0.4 g L-1) were robust in predicting concentrations and a mechanistic kinetic model built accurately predicted continuous fermentation behavior. A setpoint trajectory, ranging from 2 - 4.5 g L-1 h-1 for productivity was closely tracked by the fermentation system using Raman measurements and an extended Kalman filter to estimate biomass concentrations. Overall, this work was able to demonstrate an effective approach for real-time monitoring and control of a complex fermentation system.
- Forestry