Online sensor for monitoring a microalgal bioreactor system using support vector regression

In this work, Raman spectroscopy and a machine learning technique known as support vector regression (SVR) are used for building an online sensor to monitor the heterotrophic algal culture conditions in a computer-interfaced bench-scale microalgal bioreactor system, for the production of bio-oil. Monitoring of process conditions in algal cultures is required to enable the use of different control strategies to maximize oil productivity. In order to correlate the Raman spectra with culture conditions, three independent experimental datasets are used. The effect of several preprocessing techniques, including Savitzky-Golay filtering, baseline correction, and standard normal variate transformation, on the goodness of fit is evaluated. A multivariate sensor for real time online monitoring of the concentrations of biomass, glucose and percentage oil content is successfully built and validated. The advantages of using the proposed real-time on-line sensor are illustrated in an experimental microalgal bioreactor system.