基于LOCAL算法的秸秆水分和热值的近红外光谱快速分析  

RAPID MEASUREMENT FOR MOISTURE AND CALORIFIC VALUE OF STRAW BASED ON NEAR INFRARED SPECTROSCOPY AND LOCAL ALGORITHM

作　　者：皇才进[1,2] 韩鲁佳[1] 刘贤[1] 马玲娟[3] 

机构地区：[1]中国农业大学工学院,北京100083 [2]中国农业机械化科学研究院,北京100083 [3]农业部农业机械试验鉴定总站,北京100122

出　　处：《红外与毫米波学报》2009年第3期184-187,共4页Journal of Infrared and Millimeter Waves

基　　金：国家“十一五”科技支撑计划项目(2006BAD12B04-02B,2006BAD07A14)

摘　　要：水分和热值是与秸秆能源转化利用相关的重要特性指标,采用近红外光谱技术结合LOCAL算法来予测秸杆水份和热值.首先将样品分别按水分含量和热值大小均分为高、中、低三个子集分段建模,结果高、中、低含量样品建立的模型的交互验证标准差(SECV)几乎都小于全部样品模型的SECV,表明了秸秆水分和热值变幅对近红外光谱模型的预测精度有较大的影响.针对化学值变幅对模型精度的影响,引入LOCAL算法实现近红外光谱技术预测,快速分析测定秸秆的水分和热值.研究结果表明,LOCAL算法为每个预测样品选择合理的定标集,与常规的PLS和MPLS模型相比,有效提高了近红外光谱技术预测精度,在秸秆近红外光谱定量分析中有着广阔的应用前景.Moisture and Calorific value, which are two of the most important properties of straw for energy conversion process, were predicted by near infrared spectroscopy （NIRS） technique and LOCAL algorithm. Firstly, the samples were evenly divided into 3 subsets according to the chemical values, named high, mid and low concentration respectively, to build global partial least square regression（PLS） calibrations. Standard errors of cross validation （SECV） of the three calibrations based on subsets were lower than that of calibration based on the whole samples, which suggested that the variation of moisture and calorific value affected the accuracy of NIRS calibrations. Then, LOCAL algorithm was introduced to near infrared spectroscopy analysis for rapid measurement of the moisture content and calorific value of straw samples. By the use of LOCAL algorithm, the prediction accuracy is improved compared to the PLS and MPLS models for both moisture and calorific value of straw. It is therefore concluded that LOCAL algorithm has a broad application prospect in quantitative analysis of straw.

关 键 词：近红外光谱 LOCAL算法 水分 热值 

分 类 号：S216.2[农业科学—农业机械化工程]