Predicting anaerobic digestion stability in load-flexible operation using gas phase indicators and classification algorithms

This study investigates early warning indicators for process instabilities in anaerobic digestion caused by shock-loadings in biogas plants, focussing on gas-phase parameters to avoid substrate analyses. With the increasing use of renewable energy sources, improved energy management is essential. Biogas plants can stabilise power grids when operated flexibly. Six laboratory-scale anaerobic filters with varying organic loading rates were used to simulate load-flexible operation. Gas parameters (CH₄, CO₂, H₂, and volume) were monitored at 40-minute intervals. The analysis showed that gas quality variability can serve as an early warning, with increased variability preceding disturbances. Machine learning classifiers, i.e. Support Vector Machine, Random Forest, and Multi-Layer Perceptron, were used to distinguish between stable and unstable states achieving 80% accuracy. Compared to conventional methods (e.g., volatile fatty acid to alkalinity ratio), these methods offer a cost-effective, rapid approach for monitoring load-flexible biogas plants, providing insights without frequent laboratory analyses and high temporal resolution.