Prediction of pH and microalgae growth in mixothrophic conditions by nonlinear black-box models for control purposes

Microalgae cultivation processes that recycle waste streams from biorefineries require the availability of flexible control strategies to avoid the rapid decline of the biomass. Unfortunately, the coupling with other processes constrains the possible control actions. We developed some data-driven nonlinear models to predict pH and growth of Chlorella vulgaris nurtured in mixothrophic conditions and fed with both flue gas and wastewater.NARX and Hammerstein-Wiener black-box models are identified for pH, showing a good capability to predict its dynamics, also when subject to pulse feeding. These models can be used to set-up control strategies for this important parameter, using only wastewater flow rate as the manipulated variable and leading to a great advantage in terms of costs, if compared to manipulation of flue gas flow rate or gas composition. A dynamic grey-box model to predict microalgae concentration using only some subsets of the online pH measures is also proposed and validated. Its possible application for optimizing microalgae harvesting is briefly discussed. In order to collect rich data to be used for the identification of the dynamic models, we designed and carried out proper open loop experimental campaigns.