Transforming data to information: A parallel hybrid model for real-time state estimation in lignocellulosic ethanol...

Area of Science:

• Biotechnology
• Chemical Engineering
• Process Monitoring

Background:

• Lignocellulosic fermentation for fuels and chemicals faces challenges due to complex and variable media.
• Traditional monitoring methods lack actionable data for advanced control strategies.

Purpose of the Study:

• To develop a hybrid-modeling approach for real-time monitoring of cellulose-to-ethanol fermentation.
• To estimate glucose, xylose, and ethanol concentrations accurately.

Main Methods:

• A hybrid approach combining a data-driven partial least squares (PLS) model and a kinetic model.
• Utilized a continuous-discrete extended Kalman filter for model reconciliation.
• Employed attenuated total reflectance mid-infrared spectroscopy for real-time spectral data collection.

Main Results:

• The hybrid model significantly outperformed the standalone PLS model in accuracy.
• Improved prediction accuracy compared to the internal kinetic model.
• Demonstrated robust and consistent estimations even with measurement disturbances.

Conclusions:

• The developed hybrid model offers a robust solution for real-time monitoring of lignocellulosic fermentations.
• Enables the implementation of advanced feedback control schemes for improved process efficiency.
• Advances the field of bioprocess monitoring and control.