[Biomass Compositional Analysis Using Sparse Partial Least Squares Regression and Near Infrared Spectrum Technique].

Area of Science:

• Renewable Energy
• Biomass Characterization
• Chemometrics

Background:

• Forest biofuels are a crucial renewable energy source.
• Accurate fuel characteristic analysis is vital for biofuel utilization.
• Traditional proximate analysis is time-consuming.

Purpose of the Study:

• To develop a novel method for proximate analysis of forest biofuel (sawdust).
• To evaluate the effectiveness of Sparse Partial Least Squares Regression (SPLS) compared to other chemometric methods.
• To establish a predictive model for fuel characteristics using Near Infrared (NIR) spectroscopy.

Main Methods:

• Collected proximate analysis data (moisture, ash, volatile, fixed carbon) for 80 sawdust samples.
• Acquired Near Infrared (NIR) spectra using a Nicolet NIR spectrometer.
• Applied wavelet transform for data filtering and divided samples into training and validation sets.
• Constructed prediction models using SPLS, Principle Component Regression (PCR), Partial Least Squares Regression (PLS), and Least Absolute Shrinkage and Selection Operator (LASSO).

Main Results:

• SPLS demonstrated superior performance in predicting fuel characteristics compared to PCR, PLS, and LASSO.
• SPLS effectively selected relevant grouped wavelengths, enhancing prediction accuracy.
• Selected wavelengths were found to correlate with moisture absorption peaks.

Conclusions:

• SPLS is a powerful tool for reducing dimensionality in complex NIR spectral data.
• SPLS facilitates interpretation of the relationship between spectral data and composition concentration.
• This approach holds significant potential for advancing NIR applications in biofuel analysis.