State observers for model-based optimization in a microalgae culture

GORRINI, F.A.; FIGUEROA, J.L.; VANDE WOUWER, A.; BIAGIOLA, S.

CABA

Congreso; WCCE11 - 11th WORLD CONGRESS OF CHEMICAL ENGINEERING; 2023

ASOCIACIÓN ARGENTINA DE INGENIEROS QUÍMICOS

The algae growth process is essential for the production of biofuels/bioproducts from microalgae. Providing non-optimal conditions for biomass growth is one of the hurdles to achieving consistently high algal production rates. Though an optimal model-based controller strategy can be developed for production optimization [1], both a reliable model of the process and faithful information from the internal state of the process are necessary. However, the variables accounting for the culture evolution might be difficult to measure due to the lack of specific instruments, high sensor costs, or infeasibility of installation in the process.In the present work, a microalgae culture of Scenedesmus obliquus sp. is addressed for biomass production. Biomass concentration, intracellular quota, nitrogen concentration, and the acclimation photon flux density are the relevant variables that provide information about the internal state of microalgae culture. However, only substrate and biomass concentration measurements are available from hardware-based sensors. Therefore, two model-based observers are designed to be used as software sensors for culture optimization purposes: an Extended Kalman Filter (EKF) and a Receding Horizon Observer (RHO). The mathematical model of the process based on experimental data, along with the parameter’s values and their confidence intervals is based on [2]. This model describes the ability of microalgae to store nutrients as well as the influence of incident light on microalgae growth.As regards the EKF, a valuable and methodic way to select the filter parameters is implemented. The design parameters are tuned based on magnitudes from the system itself, avoiding a trial-and-error sintonization procedure.On the other hand, an RHO estimation algorithm is developed and tuned to take advantage of the knowledge of the nonlinear dynamics of the process. The exact nominal model provided as a set of ordinary differential equations, as well as information about model-parameters’ uncertainty can be used for the estimation. The algae production optimization problem is addressed by a nonlinear model predictive controller (NMPC), coupled with each state observer – EKF and RHO. In both situations, the incident light and the dilution rate are both manipulated for control purposes while biomass and substrate concentrations are the available measurements of the bioprocess.Observers assessment is accomplished for different scenarios such as open and closed-loop performance. For the open loop situation, the observers are validated and compared against bioreactor experimental data. Parameter uncertainty as well as unknown initial state-variables conditions are considered for observer testing and simulation results are shown. As regards the whole observer-based control structure, it is assessed by means of numerical simulation. In this sense, results show significant and promising improvements in productivity can be achieved with an observer-based NMPC structure.