Comparison of different machine learning algorithms to estimate liquid level for bioreactor management

Estimating the liquid level in an anaerobic digester can be disturbed by its closedness, bubbles and scum formation, inhomogeneity of digestate. In our previous study, a soft-sensor approach using seven pressure meters has been proposed as alternative for real-time estimation. Here, machine learning techniques were used to improve estimation accuracy optimize number sensors required this approach. Four algorithms, multiple linear regression (MLR), artificial neural network (ANN), random forest (RF), support vector (SVM) with radial basis function kernel compared purpose. All models outperformed cubic model developed among which ANN RF performed best. Variable importance analysis suggested that readings from top (in headspace) most significant, while other showed varying significance levels depending on type. The sensor experienced both headspace phases variation incurred higher error than sensors. results ML provide effective tool estimate optimizing reducing rate.