基于最小二乘支持向量机的油页岩含油率近红外光谱分析

doi:10.7503/cjcu20160344

高等学校化学学报 ›› 2016, Vol. 37 ›› Issue (10): 1792.doi: 10.7503/cjcu20160344

基于最小二乘支持向量机的油页岩含油率近红外光谱分析

张福东, 刘杰, 王智宏()

吉林大学仪器科学与电气工程学院, 长春 130021

收稿日期:2016-05-16 出版日期:2016-10-10 发布日期:2016-09-20
作者简介:联系人简介: 王智宏, 女, 博士, 教授, 博士生导师, 主要从事近红外光谱仪器研制及应用技术研究.E-mail:zhwang@jlu.edu.cn.
基金资助:
国家潜在油气资源(油页岩勘探开发利用)产学研用合作创新子课题(批准号: OSR-02-04)和吉林省科技发展计划项目重大科技专项(批准号: 20116014)资助

Analysis of Oil Yield from Oil Shale Minerals Based on Near-infrared Spectroscopy with Least Squares Support Vector Machines^†

ZHANG Fudong, LIU Jie, WANG Zhihong^*()

College of Instrumentation Science & Electrical Engineering, Jilin University, Changchun 130021, China

Received:2016-05-16 Online:2016-10-10 Published:2016-09-20
Contact: WANG Zhihong E-mail:zhwang@jlu.edu.cn.

摘要/Abstract

摘要：

为了提高油页岩含油率近红外光谱分析建模的预测精度和稳定性, 开展了基于最小二乘支持向量机(LS-SVM)建模方法的对比研究. 采用主成分-马氏距离(PCA-MD)和基于蒙特卡洛采样(MCS)2种方法进行了奇异样本的检测, 采用径向基核函数的LS-SVM、偏最小二乘(PLS)和反向传播神经网络(BPANN)3种方法进行建模方法对比. 结果表明, 对于64个油页岩岩芯样本, 与PCA-MD方法相比, 采用MCS方法剔除奇异样本后所建PLS模型的预测精度提高了28%. 对于MCS方法剔除奇异样本后的58个样品, 采用Kennard-Stone法划分了44个样品的校正集和14个样品的预测集, 采用2阶导数和标准化预处理方法, 建立了100个LS-SVM的校正模型, 模型的预测决定系数R²平均值达到0.90以上, 高于PLS和BPANN模型的对应值; 且R²的变化量(0.02)小于BPANN模型的对应值(0.32). 因此, MCS奇异样本检测结合LS-SVM方法可提高油页岩含油率样本建模的精度和稳定性.

关键词: 最小二乘支持向量机, 油页岩, 含油率, 近红外光谱分析, 奇异样本

Abstract:

In order to improve the prediction accuracy and precision of near-infrared(NIR) spectroscopy model for analyzing the oil yield from oil shale, sixty-four oil shale samples from the No.2 well drilling of Fuyu oil shale base were analyzed based on least squares support vector machines(LS-SVM) calibration models. The Principal component-mahalanobis distance(PCA-MD) method and the Monte-Carlo sampling-based detection of outliers(MCS) method were investigated as means of removing the outliers. The modeling methods of radial basis function-based LS-SVM, partial least squares(PLS) and back propagation neural network (BPANN) were compared. The results showed that, compared with PCA-MD, the prediction accuracy of PLS models based on MCS was improved by 28%. The samples after eliminating the outliers were divided into the calibration set with 44 samples and the prediction set with 14 samples using the Kennard Stone method. One hundred LS-SVM calibration models were established based on preprocessing method of second-derivative and autoscaling. The mean determination coefficient(R²) were more than 90% and higher than PLS and BPANN models, and the fluctuation of R² were less than BPANN models. Thus, LS-SVM regression with MCS method can improve the accuracy and precision of oil yield of oil shale modeling.

Key words: Least squares support vector machine, Oil shale, Oil yield, Near-infrared spectroscopy, Outlier

中图分类号:

O657.33

TrendMD:

张福东, 刘杰, 王智宏. 基于最小二乘支持向量机的油页岩含油率近红外光谱分析. 高等学校化学学报, 2016, 37(10): 1792.

ZHANG Fudong, LIU Jie, WANG Zhihong. Analysis of Oil Yield from Oil Shale Minerals Based on Near-infrared Spectroscopy with Least Squares Support Vector Machines^†. Chem. J. Chinese Universities, 2016, 37(10): 1792.

图/表 12

Fig.1 Original NIR spectra of 58 oil shale samples

Fig.2 Frequency distribution of the optimum latent variable number of components based on RDCV method

Fig.3 Frequency distribution of the RMSEP of oil yield for oil shale

Fig.4 MD values between each sample spectrum and the center of spectra of all samples

Fig.5 Result of variance of residuals versus mean of residuals on oil shale data

Fig.6 Frequency distribution of the RMSEP for oil shale data using PCA-MD

Fig.7 Frequency distribution of the RMSEP for oil shale data using MCS

Table 1 Statistical results of the oil yield in the oil shale samples

Sample set	Range	Mean	Standard deviation	Variation coefficient
All	0.45—10.87	3.33	1.92	0.58
Calibration set	0.45—10.87	3.31	1.93	0.58
Prediction set	1.48—9.09	3.36	1.95	0.58

Table 2 Prediction results for the oil yield in the oil shale samples

Model	Optimal model		Model averaging		Maximum variation
Model	RMSEP_opt	$R opt 2$	RMSEP_ave	$R ave 2$	ΔRMSEP	ΔR²
PLS	0.36	0.85
BPANN	0.24	0.93	0.47	0.80	0.33	0.32
LS-SVM	0.27	0.92	0.28	0.91	0.03	0.02

Table 2 Prediction results for the oil yield in the oil shale samples

Model	Optimal model		Model averaging		Maximum variation
Model	RMSEP_opt	$R opt 2$	RMSEP_ave	$R ave 2$	ΔRMSEP	ΔR²
PLS	0.36	0.85
BPANN	0.24	0.93	0.47	0.80	0.33	0.32
LS-SVM	0.27	0.92	0.28	0.91	0.03	0.02

Fig.8 RMSEP distribution of 100 runs results for BPANN

Fig.9 RMSEP distribution of 100 runs results for LS-SVM

Fig.10 Predicted value vs. reference value of oil yield for optimal LS-SVM model

参考文献 25

[1]	Hou J. L., Ma Y., Li S. Y., Teng J. S., Chem. Ind. Eng. Prog., 2015, 34(5), 1183—1190
	(侯吉礼, 马跃, 李术元, 藤锦生.化工进展, 2015,34(5), 1183—1190)
[2]	Zhao Z. Y., Lin J., Yu Y., Chem. Res. Chinese Universities, 2015, 31(3), 352—356
[3]	Kristin N. A., Dinesh R. K., Kalpana S. K., Spectrochimica Acta Part A Molecular & Biomolecular Spectroscopy, 2012, 89(89), 105—113
[4]	Mike J. A., Firas A., Suresh B., Stephen G., Fuel, 2005, 84(Suppl.14/15), 1986—1991
[5]	Luypaert J., Massart D. L., Heyden Y. V., Talanta, 2007, 72(3), 865—83
[6]	Chu X. L., J. Near Infrared Spec., 2015, 23(5),i —vii
[7]	Melissa J. R., Michael J. A., Andrew R. H., J. Near Infrared Spec., 2002, 10(3), 223—231
[8]	Wang Z. H., Liu J., Wang J. R., Sun Y. Y., Yu Y., Lin J., J. Jilin U. Techno. Ed., 2013, 43(4), 1017—1022
	(王智宏, 刘杰, 王婧茹, 孙玉洋, 于永, 林君.吉林大学学报: 工学版, 2013,43(4), 1017—1022)
[9]	Zhao Z. Y., Lin J., Zhang H. Z., Spectrosc. Spect. Anal., 2014, 34(6), 1707—1710
	(赵振英, 林君, 张怀柱.光谱学与光谱分析, 2014,34(6), 1707—1710)
[10]	Li S. Y., Ji Y. J., Liu W. Y., Wang Z. H., Spectrosc. Spect. Anal., 2013, 33(4), 968—971
	(李素义, 嵇艳鞠, 刘伟宇, 王智宏.光谱学与光谱分析, 2013,33(4), 968—971)
[11]	Liu J., Zhang F. D., Teng F., Li J., Wang Z. H., Spectrosc. Spect. Anal., 2014, 34(10), 2779—2784
	(刘杰, 张福东, 滕飞, 李军, 王智宏.光谱学与光谱分析, 2014,34(10), 2779—2784)
[12]	Kim K., Lee J. M., Lee I. B., Chemometrics & Intelligent Laboratory Systems, 2005, 79(1/2), 22—30
[13]	Olivier D., Cyril R., Alexandra D., Ludovic D., Chemometrics & Intelligent Laboratory Systems, 2009, 96(1), 27—33
[14]	Suykens J.A. K., Gestel T. V., Brabanter J. D., Least Squares Support Vector Machines, World Scientific, Singapore, 2002, 1—24
[15]	Sujay R. N., Paresh C. D., Applied Soft Computing, 2014, 19(6), 372—386
[16]	Li H. D., Liang Y. Z., Xu Q. S., Chemometrics & Intelligent Laboratory Systems, 2009, 95(2), 188—198
[17]	Xie Q. Y., Huang W. J., Liang D., Peng D. L., Huang L. S., Song X. Y., Zhang D. Y., Yang G. J., Spectrosc. Spect. Anal., 2014, 34(2), 489—493
	(谢巧云, 黄文江, 梁栋, 彭代亮, 黄林生, 宋晓宇, 张东彦, 杨贵军.光谱学与光谱分析, 2014,34(2), 489—493)
[18]	Zhang H. Y., Peng Y. K., Wang W., Zhao S. W., Liu Q. Q., Spectrosc. Spect. Anal., 2012, 32(10), 2794—2798
	(张海云, 彭彦昆, 王伟, 赵松玮, 刘巧巧.光谱学与光谱分析, 2012,32(10), 2794—2798)
[19]	Cao D. S., Liang Y. Z., Xu Q. S., Li H. D., Journal of Computational Chemistry, 2010, 31(3), 592—602
[20]	Liu Z. C., Cai W. S., Shao X. G., Science in China Series B: Chemistry, 2008, 51(8), 751—759
[21]	Wang Z. H., Lin J., Wu Z. Y., Zhu H., Zhan X. X., Chinese Journal of Scientific Instrument, 2005,26(11),1135—1138, 1154
	(王智宏, 林君, 武子玉, 朱虹, 占细雄.仪器仪表学报, 2005, 26(11),1135—1138, 1154)
[22]	Shan H. Y., Fei Y. Q., Huan Y. F., Chem. Res. Chinese Universities, 2014, 30(4), 582—586
[23]	Zhang H., He Y. H., Tang D., Li Y. W., Chem. J. Chinese Universities, 2014, 35(6), 1199—1203
	(张汇, 何玉韩, 唐铎, 李彦威.高等学校化学学报, 2014,35(6), 1199—1203)
[24]	Kennard R. W., Stone L. A., Technometrics, 1969, 11, 137—148
25	[25] Pravdova V., Walczak B., Massart D. L., Analytica Chimica Acta, 2002, 456(1), 77—92

基于最小二乘支持向量机的油页岩含油率近红外光谱分析

Analysis of Oil Yield from Oil Shale Minerals Based on Near-infrared Spectroscopy with Least Squares Support Vector Machines^†

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 1

编辑推荐

Metrics

本文评价

基于最小二乘支持向量机的油页岩含油率近红外光谱分析

Analysis of Oil Yield from Oil Shale Minerals Based on Near-infrared Spectroscopy with Least Squares Support Vector Machines†

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 1

编辑推荐

Metrics

本文评价

Analysis of Oil Yield from Oil Shale Minerals Based on Near-infrared Spectroscopy with Least Squares Support Vector Machines^†