Rapid Discrimination of the Categories of the Biomass Pellets Using Laser-Induced Breakdown Spectroscopy

As a renewable energy alternative to fossil fuels, biomass pellets have attracted much attention due to its promising advantages of convenient use and easy combustion. Rapid discrimination of different biomass pellets and selecting the one with better combustion performance are of great significance to improve the energy utilization. In this study, laser-induced breakdown spectroscopy (LIBS) coupled with chemometrics methods were used to discriminate biomass pellets. Lignocellulose components were firstly determined and further analyzed according to LIBS spectra. Principal component analysis (PCA), partial least squares discrimination analysis (PLS-DA), support vector machines (SVM), radial basis function neural network (RBFNN) and extreme learning machine (ELM) were applied to quantitatively distinguish biomass pellets. The RBFNN model showed a reliable discrimination power, with 100% and 96.88% average recognition accuracy in calibration and prediction sets respectively. Visualization analysis based on the RBFNN model was applied to intuitively discriminate the biomass pellets with different colors. In addition, rice husk with relatively poor combustion performance could be accurately distinguished from wood biomass pellets by all discrimination models. The results indicated that LIBS combined with chemometrics methods could be a novel and reliable approach to discriminate biomass pellets and select the category with better combustion performance.