Real-time estimation of glucose concentration in algae cultivation system using Raman spectroscopy

• Published in: Bioresource technology Vol. 142; pp. 131 - 137
• Main Authors: Oh, Se-Kyu, Yoo, Sung Jin, Jeong, Dong Hwi, Lee, Jong Min
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.08.2013; Elsevier
• Subjects: Biological and medical sciences; Cultivation; Data processing; Design engineering; Fundamental and applied biological sciences. Psychology; Glucose; Glucose - analysis; Least squares method; Least-Squares Analysis; Microalgae; Microalgae - chemistry; On-line monitoring; Partial Least Squares; Raman spectroscopy; Real time; Regression; Rolling-Circle Filter; Spectrum Analysis, Raman - methods; Microalgae; On-line monitoring; Partial Least Squares; Rolling-Circle Filter; Surveillance; Filter; Raman spectrometry; Alga; Glucose; Microorganism; Real time; Partial least squares; Culture
• ISSN: 0960-8524; 1873-2976
• DOI: 10.1016/j.biortech.2013.05.008

•We propose a framework for estimation of glucose concentration.•Prediction performance is improved by applying pre- and post-processing steps.•Prediction performance was evaluated using samples of two different experiments.•The background effect has the greatest impact on prediction performance. This work proposes a soft-sensor design for real-time estimation of glucose concentration under mixotrophic conditions using Raman spectroscopy. The suggested approach applies a Rolling-Circle Filter (RCF), Partial Least Squares (PLS), and a successive Savitzky–Golay (SG) smoothing filter. RCF is used to remove the background effects of Raman spectrum in the pre-processing step. PLS is used to reduce the dimensionality of spectrum data and relate them to the concentration. The SG filter is further employed as a post-processing step in a successive manner to adjust predicted glucose concentrations. Two sets of experiments using artificial assays and samples from a microalgae cultivation system were performed for verification. The proposed approach showed improved prediction performances compared to other data processing and regression techniques.