Can we assess the model complexity for a bioprocess: theory and example of the anaerobic digestion process

Abstract

In this paper we propose a methodology to determine the structure of the pseudo-stoichiometric coefficient matrix K in a mass balance based model, i.e. the maximal number of biomasses that must be taken into account to reproduce an available data set. It consists in estimating the number of reactions that must be taken into account to represent the main mass transfer within the bioreactor. This provides the dimension of K. The method is applied to data from an anaerobic digestion process and shows that even a model including a single biomass is sufficient. Then we apply the same method to the "synthetic data" issued from the complex ADM1 model, showing that the main model features can be obtained with 2 biomasses.

A new adaptive controller for bio-reactors with unknown kinetics and biomass concentration: Guarantees for the boundedness and convergence properties

“…The values of s, u, and f y = λr (2) are available in real time for measurement, where: f y is the measured production of a gaseous component; λ is positive, according to [23] p. 1380 and bounded away from zero; λ ≥ µ λ , where µ λ is an unknown positive constant.…”

“…Mass-balance models involving only one or two microorganism species can provide predictions with acceptable accuracy [23,5,3,2]. In addition, these models are suitable for adaptive and robust control of bioreactors [23,22,36], as they have a simple structure.…”

Modelling Anaerobic Digestion Processes

“…It will ultimately be the objective of the model and the amount and quality of information and data available that will define the optimum level of complexity of anaerobic digestion models. Optimal (minimum) model structure can be determined mathematically (Bernard et al 2006), but model complexity generally exceeds the minimum requirements required to describe the basic kinetics in order to maintain mechanistic relevance (see section 3). It is commonly the modeller's expertise in view of the model objectives and quantity and quality of the information and data available that should determine the level of complexity most suited for a specific modelling application.…”

Lyapunov functions and robustness analysis under Matrosov conditions with an application to biological systems

“…We calibrated the model using real experimental data [2] in order to get realistic parameter values. Finally, we simulated white noises for the δ i , introducing a 1% noise in the controller u and 10% noises to perturb s in and γ; see (37).…”