Background: In the pursuit of sources of energy, biofuel pellet is emerging as a promising resource because of its easy storage and transport, and lower pollution to the environment. The composition of biomass has important implication for energy conversion processing strategies. Current standard chemical methods for biomass composition are laborious, time-consuming, and unsuitable for high-throughput analysis. Therefore, a reliable and efficient method is needed for determining lignocellulose composition in biomass and so to accelerate biomass utilization. Here, near-infrared hyperspectral imaging (900-1700 nm) together with chemometrics was used to determine the lignocellulose components in different types of biofuel pellets. Partial least-squares regression and principal component multiple linear regression models based on whole wavelengths and optimal wavelengths were employed and compared for predicting lignocellulose composition.
Results: Out of 216 wavelengths, 20, 10 and 17 were selected by the successive projections algorithm for cellulose, hemicellulose and lignin, respectively. Three simple and satisfactory prediction models were constructed, with coefficients of determination of 0.92, 0.84 and 0.71 for cellulose, hemicellulose and lignin, respectively. The relative parameter distributions were quantitatively visualized through prediction maps by transferring the optimal models to all pixels on the hyperspectral image.
Conclusions: Hence, the overall results indicated that hyperspectral imaging combined with chemometrics offers a non-destructive and low-cost method for determining biomass lignocellulose components, which would help in developing a simple multispectral imaging instrument for biofuel pellets online measurement and improving the production management.
Keywords: Biofuel pellet; Biomass; Hyperspectral imaging; Image processing analysis; Lignocellulose components; Wavelength selection.