基于偏最小二乘回归法的煤中硫含量近红外检测

发布时间：2019-06-12 05:23

【摘要】：煤炭中的硫会对用煤企业的生产设备造成破坏,影响生产效率。此外,煤炭燃烧生成的硫氧化物会污染环境,危害人体健康,破坏生态系统。为了有效解决这些问题,需要测定煤的硫含量。目前常用的测定方法都需要从煤场取样然后拿到实验室进行分析测定,耗时较长,分析结果存在一定的滞后性。近红外光谱分析技术是一种快速、在线分析技术,可以在极短的时间内测定煤的硫含量,对及时优化使用条件,实现煤炭的充分燃烧具有重要意义。本论文介绍了近红外光谱分析技术的原理、特点、算法和评价指标,研究了利用近红外光谱分析技术检测煤中硫含量的快速无损检测方法。在严格控制实验环境的条件下,采集了221个煤样的近红外漫反射光谱,并测定了每个煤样的硫含量。这221个煤样包含特低硫煤、低中硫煤和中硫煤三种煤,从中选取176个作为校正集,用于建立硫分的回归模型,剩下的45个煤样作为验证集,用于检测模型的预测能力和回归效果。本研究首先选用偏最小二乘回归法建立硫分的回归模型,取得了较好的预测效果。然后运用不同异常点剔除方法优化该模型,模型结果显示马氏距离优化效果最佳。在此基础上考察不同光谱预处理方法对模型的优化程度,结果显示标准化方法最适用于该模型。最后在两次优化之后,比较不同波段筛选方法对模型的影响,结果显示相关系数法对模型有一定的优化效果。利用偏最小二乘回归法结合这三种最佳方法对煤样的硫分进行回归建模,得到了回归效果好、预测能力强的回归模型。
[Abstract]:Sulfur in coal will destroy the production equipment of coal enterprises and affect the production efficiency. In addition, sulfur oxides from coal combustion will pollute the environment, endanger human health and destroy ecosystems. In order to solve these problems effectively, it is necessary to determine the sulfur content of coal. At present, the commonly used determination methods need to be sampled from the coal site and taken to the laboratory for analysis and determination, which takes a long time and there is a certain lag in the analysis results. Near infrared spectroscopy (NIR) is a rapid and on-line analysis technology, which can be used to determine the sulfur content of coal in a very short time. It is of great significance to optimize the operating conditions in time and realize the full combustion of coal. In this paper, the principle, characteristics, algorithm and evaluation index of near infrared spectroscopy are introduced, and the rapid nondestructive testing method of sulfur content in coal by near infrared spectroscopy is studied. Under the condition of strict control of the experimental environment, 221 coal samples were collected and the sulfur content of each coal sample was determined. The 221 coal samples include ultra-low sulfur coal, low medium sulfur coal and medium sulfur coal. 176 of them are selected as correction sets to establish the regression model of sulfur content, and the remaining 45 coal samples are used as verification sets to detect the prediction ability and regression effect of the model. In this study, the partial least square regression method was used to establish the regression model of sulfur content, and good prediction results were obtained. Then the model is optimized by different outliers elimination methods, and the results show that the Mahalanobis distance optimization effect is the best. On this basis, the optimization degree of different spectral preprocessing methods to the model is investigated, and the results show that the standardized method is the most suitable for the model. Finally, after two optimizations, the effects of different band selection methods on the model are compared, and the results show that the correlation coefficient method has a certain optimization effect on the model. The partial least square regression method combined with these three best methods is used to model the sulfur content of coal samples, and a regression model with good regression effect and strong prediction ability is obtained.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O657.33;TQ533

【参考文献】