Fast Measurement Of Biomass Compositions Using Near Infrared Spectroscopy Technology

For the sake of fully utilizing forestry and agricultural residues in China,this paper proposes a quantitative analysis method about combining near infrared spectrum technique and chemometrics method for proximate and elemental analysis of biomass. By building Calibration models of quantitative analysis, the thermal chemical characteristic parameters of sawdust that include moisture, ash, volatile,fixed carbon, nitrogen, carbon, elemental sulfur, protium can be predicted fast and precisely. The applicable feasibility analysis of this method is also made and the improvement research is presented. The research is divided into four parts as follows:1. One hundred presentive samples of biomass collected from China are made proximate analysis and elemental analysis according to Chinese standard after all samples are processed and conserved by unify method. All biomass samples collected are grounded and divided into three groups according to their partical diameters through different sieves with mesh number of 20, 40 and 60.2. The NIRS of samples are Collected and pretreated by Wavelet Transform(WT)combined with Standard Normal Variate(SNV), and the calibration models are built through Principal Component Regression(PCR) and Partial Least Squares Regression(PLSR) method. The results are evaluated by comparing the Root Mean Square Error of Prediction( RMSEP), Correlation Coefficient(2R), Relative Prediction Difference(RPD) of the models among different sample sets.3. The quantitative model are constructed by the Sparse method like SPLS, LASSO and Consensus Modeling method like CPLS to analyze the fuel characteristics of sawdust combining NIR Technique. Sparse method is capable of constructing a sparse model with stronger ability in interpretation while retaining good modeling accuracy. The accuracy of prediction results and the robustness of the models can be improved through Consensus Modeling method combining multiple individual PLS models.4. In order to improve the reliability, stability and dynamic adaptability of the modelin the long term, Slope/Bias and PDS methods are used for model optimization and maintenance in Chapter Five.Among all of the multiple linear regression methods used in the experimental results established in this paper, the Spares method combining with Consensus Modeling is the most outstanding one, of which the evaluation indices including, and RPD corresponding to proximate analysis are 0.125, 98.10, 26.40 respectively, for elemental analysis are 0.155, 97.76, 16.19 respectively.It concludes that the model of quantitative analysis based on NIR could accurately and rapidly predict most of the thermal chemical characteristic parameters of biomass and could be put into practical application.