Rapid Determination Of Agricultural And Forestry Biomass Energy Quality Based On Spectroscopic Analysis

How can we maximize the efficiency Agriculture and forestry waste biomass resource (high density compression molding), which is closely related to the biomass fuel properties (moisture, ash, calorific value, etc).The quality of biochar energy is closely related to these properties (ash, volatile, etc) and varieties. Traditional measurement is chemical detection method in laboratory. But it has some deficiencies, such as time-consuming and high cost, which do not meet the needs of rapid growth for modern biomass energy. Therefore, a fast and accurate testing method for the quality of biomass energy is much-needed. This study aims at spectral modeling for the quality of biomass energy based on visible-near-infrared spectroscopy, main contents and conclusions are listed as follows:(1) The principal components of three kinds of biomass fuel including pine, cedar and cotton stalk were analyzed, and they can clearly distinguished. To the property of biomass samples (removal of water), compared PLSR model of different pretreatment spectrum, and it indicated that the results of spectral data modeling by first derivative are good (RPD>3). Under the condition of wet basis, PLSR model of BOC spectrum. The ash content, volatile matter, fixed carbon and calorific value have a high cross validation precision (R2=0.88-0.98). An artificial neural network (ANN) model combined with PLSR latent variables improved the moisture prediction accuracy (R2=0.94-0.96).(2) The spectrum modeling of agriculture and forestry waste biochar properties (ash content, volatile matter, fixed carbon, calorific value) affected by different pyrolysis were studied. Studies have shown that we can distinguish the cotton stalks biochar samples affected by different pyrolysis. PC-BPANN model was obtained with R2=0.93and RMSEP=0.54; the performance of LV-BPANN model was obviously better than PC-BPANN model about biochar component and calorific value. UVE, SPA and UVE-SPA were used to extract the characteristic wavelengths and the PLSR model was established and analysised, the model performance of UVE-SPA method is best, and its wavelength number was3-6. PLSR model (ash content, volatile matter, fixed carbon and calorific value) were developed with R2of0.83,0.93,0.92and0.77, RPD of3.55,3.30,3.39and2.08. (3) In this study, this is to establish model for identify14kinds of biochar types accurate and fast. PC-SVM model performance was better than PC-LDA, the support vector machine models for the first three PCs (PC-SVM) performed best with lest number of wrongly classified samples in validation set. When there were the first three PCs, PC-BPANN model has the best classification results (R2=0.89). But to the spectral data modeling based on wavelet transform, the LDA model is better than that of the SVM model.