基于随机森林和树突网络的煤镜质组反射率估计

袁懿琳; 赵荣焕; 何坤; 黄秀; 王洪栋; 邹亮

doi:10.13272/j.issn.1671-251x.18082

基于随机森林和树突网络的煤镜质组反射率估计

doi: 10.13272/j.issn.1671-251x.18082

袁懿琳^1,,
赵荣焕²,
何坤¹,
黄秀¹,
王洪栋^2, ,,
邹亮²

1.
中国石油天然气股份有限公司勘探开发研究院, 北京　100083
2.
中国矿业大学信息与控制工程学院, 江苏徐州　221116

基金项目: 中国石油天然气股份有限公司科学研究与技术开发项目(2021DJ0107)；科技部科技创新2030—“新一代人工智能”重大项目(2020AAA0107300)；徐州市基础研究计划项目(KC22020)。

详细信息

作者简介:
袁懿琳（1993—），女，北京人，工程师，主要从事有机地球化学研究工作，E-mail：yuanyilin@petrochina.com.cn

通讯作者:
王洪栋（1986—），男，山东临沂人，讲师，博士，现主要从事机器视觉与图像处理研究工作，E-mail：zs10060188@cumt.edu.cn

中图分类号: TD94
计量
- 文章访问数: 170
- HTML全文浏览量: 56
- PDF下载量: 17
- 被引次数: 0
出版历程
- 收稿日期: 2023-02-22
- 修回日期: 2023-08-10
- 网络出版日期: 2023-09-04

Estimation of coal vitrinite reflectance based on random forest and dendritic network

1.
Research Institute of Petroleum Exploration and Development, PetroChina Company Limited, Beijing 100083, China
2.
School of Information and Control Engineering, China University of Mining and Technology, Xuzhou 221116, China

摘要

摘要: 镜质组平均最大反射率是表征煤化程度的重要指标，在确定煤级、鉴别混煤和指导炼焦配煤中起关键作用。传统反射率测定方法费时耗力，且测量结果的主观性较强，致使实验室间鉴定结果的可比性差。针对该问题，提出一种基于随机森林（RF）和树突网络（DDNet）的煤镜质组反射率估计方法，主要包括煤岩显微图像分割、镜质组识别和镜质组平均最大反射率预测3个部分。利用手肘法和K−Means算法对显微图像聚类，以实现不同显微组分区域的分割；采用人工少数类过采样法（SMOTE）对少数类样本过采样，以改善煤岩中镜质组与非镜质组区域样本的不均衡问题；利用基于DDNet的回归算法实现镜质组平均最大反射率的估计，构建回归模型时从镜质组区域中选择多个41×41像素的方形窗口并提取其灰度特征，以提高算法的鲁棒性，其决定系数达到0.990。实验结果表明：采用手肘法自动确定K−Means算法的参数K，具有良好的自适应能力，能够自动区分不同类别数的显微组分；SMOTE方法可有效避免模型因过度学习样本先验信息而导致对多数类识别好、少数类识别差的问题，提高分类准确度，其中基于RF的识别模型准确率达到97.0%；建立了7种回归估计模型，其中DDNet回归模型性能最佳，决定系数达到0.990，预测结果与实际值高度契合，验证了所提方法的可行性。
- 镜质组平均最大反射率 /
- K−Means聚类 /
- 随机森林 /
- 树突网络 /
- SMOTE /
- DDNet回归算法
Abstract: The mean maximum vitrinite reflectance is an important indicator of the degree of coalification, and plays a key role in determining coal grade, identifying mixed coal, and guiding coking coal blending. The traditional reflectance measurement methods are time-consuming and labor-intensive. The subjectivity of measurement results is strong, resulting in poor comparability of identification results between laboratories. To address this issue, a method for estimating coal vitrinite reflectance based on random forests(RF) and dendritic networks(DDNet) is proposed. It mainly includes three parts: coal rock microscopic image segmentation, vitrinite recognition, and mean maximum vitrinite reflectance prediction. The elbow method and K-Means algorithm are used to achieve segmentation of different maceral regions of the clustering microscopic images. The artificial minority oversampling method (SMOTE) is used to oversample minority samples to improve the imbalance between vitrinite and nonvitrinite regional samples in coal and rock. The DDNet-based regression algorithm is used to estimate the mean maximum vitrinite reflectance. When building a regression model, multiple 41×41 pixel square windows are selected from the vitrinite regions to extract their grey scale features. It improves the robustness of the algorithm, with a determination coefficient of 0.990. The experimental results show that using elbow method to automatically determine the parameter K of the K-Means algorithm, which has good adaptive capability. It can automatically distinguish different types of microscopic components. The SMOTE method can effectively avoid the problem of over-learning sample prior information, which leads to good recognition of the majority class and poor recognition of the minority class. It improves classification accuracy. Among them, the recognition model based on RF has an accuracy rate of 97.0%. Seven regression estimation models have been established, among which the DDNet regression model has the best performance, with a determination coefficient of 0.990. The predicted results are highly consistent with the actual values, verifying the feasibility of the proposed method.
- mean maximum vitrinite reflectance /
- K-Means clustering /
- random forest /
- dendritic network /
- SMOTE /
- DDNet regression algorithm

HTML全文

图 1 MMVR估计流程

Figure 1. Estimation process of mean maximum vitrinite reflectance

下载: 全尺寸图片幻灯片

图 2 DDNet网络结构

Figure 2. The structure of dendrite net

下载: 全尺寸图片幻灯片

图 3 K−Means聚类结果

Figure 3. K−Means clustering results

下载: 全尺寸图片幻灯片

图 4 煤岩显微图像的MMVR分析结果

Figure 4. MMVR analysis results of the coal photomicrograph

下载: 全尺寸图片幻灯片

图 5 DDNet模型的预测值与真实值对比

Figure 5. Comparison of predicted value of DDNet and the ground truth

下载: 全尺寸图片幻灯片

图 6 煤岩MMVR估计软件界面

Figure 6. MMVR estimation software interface for coal and rock

下载: 全尺寸图片幻灯片

表 1 过采样、下采样处理前后结果对比

Table 1. Comparison of experimental results before and after oversampling and down-sampling

数据处理	分类算法	准确率	查准率	召回率	F₁分数
处理前	CART	0.95±0.02	0.88±0.06	0.87±0.07	0.87±0.05
	KNN	0.95±0.01	0.90±0.05	0.86±0.05	0.88±0.04
	SVM	0.95±0.01	0.88±0.06	0.88±0.06	0.88±0.04
	RF	0.96±0.01	0.92±0.05	0.85±0.07	0.88±0.04
RUS	CART	0.93±0.02	0.77±0.08	0.92±0.05	0.83±0.05
	KNN	0.95±0.02	0.82±0.06	0.93±0.04	0.87±0.04
	SVM	0.95±0.02	0.81±0.08	0.95±0.04	0.87±0.05
	RF	0.95±0.02	0.82±0.06	0.96±0.04	0.89±0.04
SMOTE结合RUS	CART	0.94±0.02	0.83±0.06	0.87±0.06	0.85±0.04
	KNN	0.96±0.01	0.86±0.06	0.92±0.05	0.89±0.03
	SVM	0.96±0.01	0.88±0.05	0.90±0.04	0.89±0.03
	RF	0.97±0.01	0.91±0.05	0.93±0.04	0.92±0.03
SMOTE	CART	0.95±0.01	0.85±0.05	0.88±0.05	0.87±0.04
	KNN	0.96±0.01	0.87±0.05	0.91±0.05	0.89±0.03
	SVM	0.96±0.01	0.88±0.05	0.90±0.05	0.89±0.03
	RF	0.97±0.01	0.92±0.04	0.93±0.04	0.92±0.03

下载: 导出CSV

表 2 回归算法测试结果对比

Table 2. Comparison of test results of regression algorithms

算法	均方误差	平均绝对误差	决定系数
SVR	0.014	0.080	0.885
AdaBoost	0.010	0.078	0.919
KNN	0.010	0.071	0.925
Gradient Boosting	0.009	0.071	0.926
RF	0.009	0.070	0.926
FNN	0.033	0.064	0.735
DDNet	0.001	0.027	0.990

下载: 导出CSV

参考文献(24)

[1]	田英奇,张卫华,沈寓韬,等. 镜质组反射率指导优化配煤炼焦方案的研究[J]. 煤炭科学技术,2016,44(4):162-168. TIAN Yingqi,ZHANG Weihua,SHEN Yutao,et al. Research on optimization of coal blending coking guided by vitrinite reflectance[J]. Coal Science and Technology,2016,44(4):162-168.
[2]	俞楠,邹冲,刘诗薇,等. 利用镜质组反射率鉴定兰炭与煤粉互混样的方法解析[J]. 冶金能源,2022,41(5):13-18. YU Nan,ZOU Chong,LIU Shiwei,et al. Analysis of the method for identifying the mixed samples of pulverized coal and semi-coke by vitrinite reflectance[J]. Energy for Metallurgical Industry,2022,41(5):13-18.
[3]	宋孝忠,张群. 煤岩显微组分组图像自动识别系统与关键技术[J]. 煤炭学报,2019,44(10):3085-3097. SONG Xiaozhong,ZHANG Qun. Automatic image recognition system and key technologies of maceral group[J]. Journal of China Coal Society,2019,44(10):3085-3097.
[4]	SANTOS R B M,AUGUSTO K S,IGLESIAS J C Á,et al. A deep learning system for collotelinite segmentation and coal reflectance determination[J]. International Journal of Coal Geology,2022,263:104111-104122. doi: 10.1016/j.coal.2022.104111
[5]	VAN NIEKERK D,MITCHELL G D,MATHEWS J P. Petrographic and reflectance analysis of solvent-swelled and solvent-extracted South African vitrinite-rich and inertinite-rich coals[J]. International Journal of Coal Geology,2009,81(1):45-52. doi: 10.1016/j.coal.2009.10.021
[6]	MLYNARCZUK M,GÓRSZCZYK A,ŚLIPEK B. The application of pattern recognition in the automatic classification of microscopic rock images[J]. Computers & Geosciences,2013,60:126-133.
[7]	WANG Hongdong, LEI Meng, CHEN Yilin, et al. Intelligent identification of maceral components of coal based on image segmentation and classification[J]. Applied Sciences, 2019, 9（16）. DOI: 10.3390/app9163245 .
[8]	LEI Meng, RAO Zhongyu, WANG Hongdong, et al. Maceral groups analysis of coal based on semantic segmentation of photomicrographs via the improved U-net[J]. Fuel, 2021, 294. DOI: 10.1016/j.fuel.2021.120475.
[9]	王培珍,余晨,薛子邯,等. 基于迁移学习的煤岩壳质组显微组分识别模型[J]. 煤炭科学技术,2022,50(1):220-227. WANG Peizhen,YU Chen,XUE Zihan,et al. Transfer learning based identification model for macerals of exinite in coal[J]. Coal Science and Technology,2022,50(1):220-227.
[10]	ENGLAND B M,MIKKA R A,BAGNALL E J,et al. Petrographic characterization of coal using automatic image analysis[J]. Journal of Microscopy,1979,116(3):329-336. doi: 10.1111/j.1365-2818.1979.tb00218.x
[11]	王洪栋. 基于机器学习的煤岩显微图像分析研究[D]. 徐州: 中国矿业大学, 2019. WANG Hongdong. Research on photomicrograph analysis of coal based on machine learning[D]. Xuzhou: China University of Mining and Technology, 2019.
[12]	WANG Hongdong,LEI Meng,LI Ming,et al. Intelligent estimation of vitrinite reflectance of coal from photomicrographs based on machine learning[J]. Energies,2019,12(20):1-16.
[13]	ONUMANYI A J, MOLOKOMME D N, ISAAC S J, et al. AutoElbow: an automatic elbow detection method for estimating the number of clusters in a dataset[J]. Applied Sciences, 2022, 12（15）. DOI: 10.3390/app12157515.
[14]	SAMMOUDA R, EL-ZAART A. An optimized approach for prostate image segmentation using K-means clustering algorithm with elbow method[J]. Computational Intelligence and Neuroscience, 2021, 2021. DOI: 10.1155/2021/4553832.
[15]	BOKHARE A, BHAGAT A, BHALODIA R. Multi-layer perceptron for heart failure detection using SMOTE technique[J]. SN Computer Science, 2023, 4（2）. DOI: 10.1007/s42979-022-01596-x.
[16]	ISHAQ A,SADIQ S,UMER M,et al. Improving the prediction of heart failure patients' survival using SMOTE and effective data mining techniques[J]. IEEE Access,2021,9:39707-39716. doi: 10.1109/ACCESS.2021.3064084
[17]	LIU Gang. It may be time to improve the neuron of artificial neural network[EB/OL]. [2023-01-20]. https://www.techrxiv.org/articles/preprint/It_may_be_time_to_perfect_the_neuron_of_artificial_neural_network/12477266.
[18]	LIU Gang,WANG Jing. Dendrite net:a white-box module for classification,regression,and system identification[J]. IEEE Transactions on Cybernetics,2022,52(12):13774-13787. doi: 10.1109/TCYB.2021.3124328
[19]	MARIUSZ M,MARTA S. The application of artificial intelligence for the identification of the maceral groups and mineral components of coal[J]. Computers & Geosciences,2017,103:133-141.
[20]	高涛,冯松宝. 煤的显微组分特征研究综述[J]. 能源技术与管理,2021,46(4):15-16,20. GAO Tao,FENG Songbao. Literature review on property of macerals of coals[J]. Energy Technology and Management,2021,46(4):15-16,20.
[21]	LI Na,HAO Huizhen,GU Qing,et al. A transfer learning method for automatic identification of sandstone microscopic images[J]. Computers & Geosciences,2017,103:111-121.
[22]	WU Zhuang,JIANG Shanshan,ZHOU Xiaolei,et al. Application of image retrieval based on convolutional neural networks and Hu invariant moment algorithm in computer telecommunications[J]. Computer Communications,2020,150:729-738. doi: 10.1016/j.comcom.2019.11.053
[23]	MEENAKSHI G,GAURAV D. A novel content-based image retrieval approach for classification using GLCM features and texture fused LBP variants[J]. Neural Computing & Applications,2021,33(4):1311-1328.
[24]	WANG Qi, HUANG Wei, ZHANG Xueting, et al. GLCM: global-local captioning model for remote sensing image captioning[J]. IEEE Transactions on Cybernetics （Early Access）, 2022: 1-13. DOI: 10.1109/TCYB.2022.3222606.