Surrogate model for improved digitalisation in the anaerobic digestion industry

Purpose and goal:

The main project idea was to evaluate the interconnections between anaerobic digestion (AD) process performance, reactor fluid properties, and stirring power demand. The goals were to first collect the fluctuations of these parameters under different conditions, followed by developing predictive models to describe them. Both goals were achieved, but more work is required to improve the predictive capacity of the models for all possible operational scenarios.

The possibility of including additional model input parameters to improve the predictive power should be evaluated. Expected results and effects:

The main results of this study are that biogas reactor material fluid properties fluctuate with process performance, and that these fluctuations might be possible to record by stirrer motors at laboratory-scale, but additional data is needed. Furthermore, the use of vector autoregression analysis indicated a possibility to correlate the signals from the motor with other process stability indicators, particularly if additional predictive parameters are included. Future work should focus on identifying ways to monitor fluid behaviour at industrial scale.

Approach and implementation:

Process data, such as stirrer motor load, specific gas production, total solids (TS), volatile fatty acids (VFA), and pH, was collected from two continuous stirred-tank biogas reactors (R1 & R2). The process in R2 was intentionally disturbed by spiking its substrate with urea and inducing an ammonia overload, followed by a re-stabilisation to collect information on parameter dynamics during non-steady state operation.

Simple regression analysis was used to find correlations between process parameters, followed by applying a more advanced vector autoregression analysis.