Microalgae are well known photosynthetic microbes used as cell factories for the production of relevant biotechnological compounds (e.g. pigments, unsaturated fatty acids) and environmental applications (e.g. wastewater treatment, lipids for biofuel). Despite these outstanding characteristics, industrial scale production of microalgae is still at an early stage of development. Moreover, the economically viable support of the sustainable large-scale production of microalgae at the present state of technology is a challenge.
One important bottleneck in this support is the lack of suitable on-line sensors for the reliable monitoring of biological parameters, mostly concentration of intracellular components, in microalgae bioprocesses. The use of spectroscopic sensors and non-optical methods (e.g. NMR, Mass spectrometry) provides interesting alternatives to overcome this situation. Sensors based on optical methods (e.g. spectroscopy of absorption, reflectance and scattering, fluorescence, microscopy, and multispectral imaging) have proven their applicability and reliability to target relevant biological variables in microalgae bioprocess for example biomass, cell concentration, nutrient consumption (e.g. NO3, glucose), intracellular products (pigments and lipids), physiological state (e.g. photosynthetic efficiency), cell morphology, and presence of contaminants.
Together with spectroscopic sensors, the application of software sensors, i.e., estimators based on model and data driven process models, using physical signal inputs of multiple sensors could complement the on-line monitoring as it allows the estimation of the desired biological process variables in real or almost real time which are often hidden in the spectroscopic data and only indirectly accessible. But which techniques are used in the development of a software sensor? For that purpose, different techniques are employed including artificial neural network (ANN), deep neural network, adaptive interval observer, Luenberger observer, Kalman filter, Principal Component Analysis (PCA), and PLS regression model. Also, these techniques could be employed in combination, in parallel or in sequence. For example, multiple linear regression in combination with mechanistic model or support vector regression together with random forest regression.
Considering the forgoing, the development of a non-invasive optical on-line biosensors represents an interesting opportunity within the context of microalgae bioprocesses. The expected results in this work include the development of a novel on-line sensor system combining hardware sensors and advanced real-time data processing, using software sensors, for a fast automated estimation of different relevant biological process parameters.