Real‐time optimization and control for polyhydroxybutyrate fed‐batch production at pilot plant scale

Abstract

BACKGROUND: Polyhydroxyalkanoate-type polymers (PHAs) are claimed as an alternative for replacing conventional plastics; but the economic feasibility of its production should still be assured in order to accomplish such a massive goal. In a previous work on fed-batch production of PHAs, it was shown that the process yield and the properties of the biopolymer obtained highly depend on the type of feeding strategy applied. Such results evidenced the need for implementing optimization-based control strategies to control the process in real-time, rather than implementing pre-defined feeding policies. RESULTS: In this study, real-time optimization for the fed-batch production of PHAs at pilot plant scale from a low-cost substrate (vinasses-molasses mixture) is presented. Two optimization-based control strategies were developed to calculate the optimal feed flow rate. One strategy considered only the process productivity as an objective, whereas the other incorporated not only productivity but also a profitability-oriented objective. The latter resulted in a productivity of 0.364 g/L h and a maximal polymer concentration of 62.03% cell dry weight, versus 0.043 g/L h and 6.25% attained by applying a traditional feedback control law. CONCLUSION: Implementation of closed-loop based-optimal control policies in PHA production has led to higher yield and polymer concentration, confirming that an improved operation of fed-batch bioprocesses is reached when the feeding policy is updated in real-time, according to an optimization goal. Despite implementation costs, applying this approach might be the best alternative for simultaneously facing uncertainties and achieving optimal operation at an industrial scale, especially when living microorganisms are used as factories.

Advanced Kinetic Modeling of Bio-co-polymer Poly(3-hydroxybutyrate-co-3-hydroxyvalerate) Production Using Fructose and Propionate as Carbon Sources

“…Because of the compact model structure and smart coupling to available online measurements the approach can be applied as a soft sensor for the prediction of biomass, substrates, product yield and the composition of the biopolymer (e.g., 3HV/3HB ratio). First steps in this direction where already successfully applied for a vinasse-molasse PHA process [30,31]. In comparison to the model approach used in [30,31], our more complex model represents an excellent basis to predict 3HV and 3HB amounts in the polymer chains.…”

“…In the mathematical model presented here, we applied a time-dependent, kinetic parameter for the formation of residual biomass from fructose and propionic acid in C. necator, in order to map the dynamics of the present metabolic activity without detailed metabolic information. As the researchers at the University of Antioquia (Colombia) already have shown [30,31], the online measurement for the CO 2 content in exhaust gas serves as an excellent measure of the dynamic growth rate. Here, we also apply this correlation and hence, the model is a suitable candidate for the online estimation of PHBV product yields.…”

Digital Twin in biomanufacturing: challenges and opportunities towards its implementation

“…Those parameters can then be used for monitoring purposes or even be employed in feedback control. In industrial settings, soft sensors must be based on hardware that is robust and straightforward to operate and models that are realistically validated [34,35]. Ideally, among other benefits, soft sensors should lead to a simpler analytical system, decreased operational and maintenance effort, reduced equipment costs, and bring about a greater degree of automation.…”

“…Real-time optimization (RTO), also called on-line optimization, continuously evaluates and changes a process' operating conditions to maximize its productivity (under operational constraints) [62]. RTO is usually applied in highly automated production plants [63] following one of two approaches: the two-layer (cascade) structure or the direct on-line optimizing control method [35,63]. In the first approach, the upper layer executes the optimization to identify optimal set-points for a set of controlled variables.…”

“…The latter approach entails that the manipulated variables are used directly as the decision variables of an optimization problem. The objective function is then calculated over a particular prediction horizon (based upon a nonlinear dynamic process model) [35]. Ochoa et al [64] have compared both approaches by applying them to a bioethanol production process (plant-wide optimization).…”