Improved prediction of biomass composition for switchgrass using reproducing kernel methods with wavelet compressed FT-NIR spectra

Fourier transform near-infrared (FT-NIR) technique is an effective approach to predict chemical properties and can be applied to online monitoring in bio-energy industry. High dimensionality and collinearity of the FT-NIR spectral data makes it difficult in some applications to construct the reliable prediction model. In this study, two nonlinear kernel methods with wavelet-compressed data, Kernel Partial Least Squares (KPLS) regression and Kernel Ridge Regression (KRR), are presented to resolve those data into a few predictors and then, more sophisticated models are created to capture the nonlinear relationships between the spectral data and concentrations determined by wet chemistry. A wavelet transform is adopted as a preprocessing procedure to reduce the data size for supporting real-time implementation of assessing biomass properties with FT-NIR spectroscopy. A real-life data of switchgrass is presented to illustrate the performance of the developed models and the results advocated that the use of nonlinear kernel procedure with wavelet compression improved the prediction performance of the model.