A cheap real-time optimal control setup for microalgae cultures: Simulation and experimental results

Model-free extremum seeking (ES) has become a popular dynamic optimization strategy as it avoids the time- and resource-consuming task of developing a process model and identifying its parameters. This study explores the possibility of developing an ES control strategy for continuous cultures of micro-algae using minimum investment and, in particular, exploiting a simple “in-house” RGB sensor for biomass concentration. Besides the action on the dilution rate, variation in the incident light is also considered, and the lower sensitivity of the productivity with respect to this input is tackled by a Newton approach where the second-order information is inferred from an adaptive Hammerstein model included in the ES scheme. The setup is validated in simulation and experimental studies, demonstrating the strategy’s good performance. Initialization is a critical factor to ensure fast convergence, and, to ensure robustness, bounds on the Hessian estimates have to be imposed.