Surrogate Modeling for Control of Microbial Biopolymer Production Process

In this contribution, the Dynamic Mode Decomposition with control (DMDc) is used to derive a surrogate model of a continuous PHA biopolymer production process based on a recently published complex process model. Here, snapshot simulation data of the original model is processed to obtain a linear surrogate model formulation using delay coordinates. The quality of the surrogate is statistically validated within simulation studies. Additionally, the influence of the of the order of delay coordinates is investigated. It is shown, that the highly nonlinear dynamics of the PHA-manufacturing process can be approximated accurately by the DMD-based model even for large variations of initial conditions and control variables. This offers the opportunity to apply well-studied and established tools from robust and optimal control in future investigations.