基于谱区优选的近红外光谱快速预测玉米秸秆中木质纤维素含量的研究  

Research on Rapid Determination of Lignocellulosic Contents in Corn Stover Using Near Infrared Spectroscopy Based on Spectral Intervals Selection

作　　者：许永花[1] 王娜[2] 刘金明[2,3] XU Yong-Hua;WANG Na;LIU Jin-Ming(School of Electrical and Information,Northeast Agricultural University,Harbin 150030,China;College of Information and Electrical Engineering,Heilongjiang Bayi Agricultural University,Daqing 163319,China;National Coarse Cereals Engineering Research Center,Daqing 163319,China)

机构地区：[1]东北农业大学电气与信息学院,哈尔滨150030 [2]黑龙江八一农垦大学信息与电气工程学院,大庆163319 [3]国家杂粮工程技术中心,大庆163319

出　　处：《分析化学》2022年第10期1587-1596,共10页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金项目(No.52076034);黑龙江省省院科技合作项目(No.YS20B01);黑龙江八一农垦大学人才启动计划课题项目(No.XDB202006)资助。

摘　　要：在生物燃气生产过程中,玉米秸秆中的木质纤维素(纤维素、半纤维素和木质素)成分含量对厌氧发酵性能具有重要影响。针对传统方法测定本质纤维素的耗时、成本高等问题,本研究分析了近红外光谱(NIRS)结合化学计量学进行玉米秸秆中木质纤维素含量快速检测的可行性。为提高NIRS模型的检测精度和效率,将遗传模拟退火算法(GSA)、区间偏最小二乘法(iPLS)和支持向量机(SVM)相结合,构建遗传模拟退火区间支持向量机(GSA-iSVM)进行NIRS特征谱区和SVM参数的同步优化,并与反向区间偏最小二乘法(BiPLS)、遗传模拟退火区间偏最小二乘法(GSA-iPLS)的优选特征谱区的建模性能进行对比,确定基于GSA-iSVM建立的纤维素和木质素校正模型性能最佳,基于GSA-iPLS建立的半纤维素校正模型性能最佳。纤维素、半纤维素和木质素最佳校正模型验证集的预测决定系数(R^(2))分别为0.910、0.990和0.939,预测均方根误差(RMSEP)分别为0.881%、0.707%和0.249%,剩余预测偏差(RPD)分别为3.283、10.235和4.270。结果表明,NIRS与GSA特征谱区智能搜索相结合,可作为可靠的替代策略,用于测量厌氧发酵工艺中预处理玉米秸秆中的木质纤维素成分含量。The contents of lignocellulosic components(including cellulose,hemicellulose and lignin)have an important influence on the methane yield of anaerobic digestion(AD)with corn stover(CS)as feedstocks in biogas industry.Aiming at the time-consuming and high-cost issues of traditional chemical analytical techniques,the feasibility of near infrared spectroscopy(NIRS)combined with chemometrics methods to measure the contents of lignocellulose in corn stover was analyzed in this work.To improve the detection accuracy and efficiency of NIRS regressive model,the genetic simulated annealing interval support vector machine(GSA-iSVM)was constructed using genetic simulated annealing algorithm(GSA)combined with interval partial least squares(iPLS)and support vector machine(SVM),which was used for synchronous optimization of the NIRS characteristic intervals and SVM parameters.By comparison with the modeling performance of the characteristic spectral intervals selected by backward interval partial least squares and genetic simulated annealing interval partial least squares(GSA-iPLS),it was found that the calibration model for cellulose and lignin established by GSA-iSVM had the best predicted accuracy,and that of hemicellulose established by GSA-iPLS performed best.For the validation set,the determination coefficients of prediction,root mean squared error of prediction and residual predictive deviation of the best calibration models were 0.910,0.881%and 3.283 for cellulose;0.990,0.707%and 10.235 for hemicellulose;and 0.939,0.249%and 4.270 for lignin,respectively.The results indicated that NIRS coupled with characteristic intervals intelligent selection of GSA could be used as a reliable alternative strategy to measure contents of lignocellulosic components in the pretreated CS in AD process.

关 键 词：玉米秸秆 木质纤维素 近红外光谱 特征谱区 偏最小二乘 支持向量机 

分 类 号：TK6[动力工程及工程热物理—生物能] O657.33[理学—分析化学]