This work presents a methodology that seeks to be a new standard in modeling identification in anaerobic digestion reactors. Because it is not possible to measure all variables with reliable and cost-efficient real-time methods, a specific structure composed of an asymptotic observer for the concentration of state variables; acidogenic and methanogenic bacterias, unlock the use of new types of raw sludges for industrial control and monitoring purposes. New yield parameters were included in the reduced anaerobic digestion model (ADM2) used as the core, precisely two terms in total alkalinity, to bring about the modeling of additional organic materials at inlet containing proteins or amino acids. The fermentation of these substances introduces ammonium, providing variations in the amount of alkalinity available inside the reaction. The new model is used to solve an optimization problem that calculates the parameters that best fit the dynamics of state variables with the same information taken on the experimental data. The adjustment process started with the genetic algorithm; however, to improve the performance, a novel method is proposed called step-ahead. Together, including the design of an asymptotic observer, numerical simulations demonstrate the strengths of the structure, which constitutes a significant step in paving the way further to implement feasible, cost-effective control and monitoring systems in the industry.

This work presents a nonlinear model predictive control scheme that challenges overcoming the obstacles holding back over decades to develop affordable autonomous control and monitoring systems applied in the large-scale industry. Among the numerous proposals in the literature, most do not consider the significant fluctuation of kinetic parameters in the reduced mathematical model ADM2, widely used for control and monitoring purposes. The prevalent cause, on a basis, is the lack of information caused by some dynamics and parameters that cannot be measured in real-time by reliable sensors. In addition, to make matters worse, those systems inherently act with nonlinear nature and have a high sensitiveness to uncontrollable inputs and perturbations. Therefore, to prevent these drawbacks, this work proposes a new methodology that reconstructs the lack of information from the non-measurable dynamics, concentration of bacterias, and the kinetic parameters related to reaction rates. Simulations results demonstrate the effectiveness of the methodology compared with traditional industrial control schemes.

This work presents a nonlinear model predictive control scheme that challenges overcoming the obstacles holding back over decades to develop affordable autonomous control and monitoring systems applied in the large-scale industry. Among the numerous proposals in the literature, most do not consider the significant fluctuation of kinetic parameters in the reduced mathematical model ADM2, widely used for control and monitoring purposes. The prevalent cause, on a basis, is the lack of information caused by some dynamics and parameters that cannot be measured in real-time by reliable sensors. In addition, to make matters worse, those systems inherently act with nonlinear nature and have a high sensitiveness to uncontrollable inputs and perturbations. Therefore, to prevent these drawbacks, this work proposes a new methodology that reconstructs the lack of information from the non-measurable dynamics, concentration of bacterias, and the kinetic parameters related to reaction rates. Simulations results demonstrate the effectiveness of the methodology compared with traditional industrial control schemes.