胡驰,李新虎,李晓君,李健,郭杰

• 1:西安科技大学地质与环境学院
• 2:陕西省煤炭绿色开发地质保障重点实验室
• 3:国土资源部煤炭资源勘查与综合利用重点实验室
• 4:甘肃煤炭地质勘查院

摘要(Abstract)：

为了解决煤层含气量定量解释问题,将煤层测井数据与煤心解吸数据作为输入和输出参数,构建深度置信网络(DBN),进而预测煤层含气量。研究以甘肃合水地区测井数据为例,筛选出该地区120组煤层样品作为DBN样本分析数据。选择短源距自然伽马、自然伽马、密度、长源距自然伽马和浅侧向5条测井曲线,作为DBN的输入参数,煤层气含量作为DBN的输出参数,研究RBM数量和隐藏神经元数量对计算结果的影响。并通过概率统计法、BPNN、DBN和SVM计算了30组煤层的煤层气含量,比较不同方法的预测效果。结果表明:(1)受限玻尔兹曼机(RBM)对DBN计算结果的精度有一定的影响,RBM数量达到7层时,预测结果准确性更高;(2)选择合适的隐藏层神经元数量,可以保证计算结果的精度和稳定性,神经元数量为20时,预测结果精度更高,稳定性更好;(3)RBM使得DBN的准确性高于BPNN,此外,DBN的计算准确性和稳定性高于概率统计法和SVM。

关键词(KeyWords)： 煤层气;测井解释;深度置信网络;合水地区

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金青年基金(编号:41502159);; 国土资源部煤炭资源勘查与综合利用重点实验室开放课题基金项目(编号:KF2018-4)

作者(Author): 胡驰,李新虎,李晓君,李健,郭杰

参考文献(References)：

• [1]陈小军，李鹏飞，李萍，等．多元逐步回归分析法在煤层含气量预测中的应用[J]．煤炭工程，2019,51(2):106-111.Chen Xiaojun,Li Pengfei,Li Ping,et al. Application of Multiple Stepwise Regression Analysisin Prediction of Coal Seam Gas Content[J]. Coal Engineering. 2019,51(2):106-111.
• [2]张鹏，王昆，于涛，等．基于多元线性回归分析法的煤层气含量预测[J]．煤炭技术，2016,35(11):112-115.Zhang Peng,Wang Kun,Yu Tao,et al. Content Predicting of CBM Based on Multiple Linear Regression Analysis Method[J]. Coal Technology.2016,35(11):112-115
• [3]曹军涛，赵军龙，王轶平，等．煤层气含量影响因素及预测方法[J]．西安石油大学学报(自然科学版)，2013,28(4):28-34+94+7.Cao Juntao,Zhao Junlong,Wang Yiping,et al. Review of influencing factors and prediction methods of gas content in coal seams and prospect of prediction methods[J]. Journal of Xi'an Shiyou University(Natural Science Edition),2013,28(4):28-34+94+7.
• [4]郭晓龙，李璇，代春萌，等．煤层气地球物理预测方法[J]．天然气地球科学，2017,28(2):287-295.Guo Xiaolong,Li Xuan,Dai Chunmeng,et al. Research on CBM geophysical prediction[J]. Natural Gas Geoscience,2017,28(2):287-295.
• [5]李凡，凌光磊．基于BP神经网络模型的煤层含气量预测[C]．//北京力学会第二十五届学术年会会议论文集．北京力学会:北京力学会，2019:1166-1168Li Fan,Ling Guanglei. Prediction of Coal Seam Gas Content Based on BP Neural Network Model[C]．//Proceedings of the 25th Annual Conference of the Beijing Society of Mechanics. Beijing Society of Mechanics:Beijing Society of Mechanics,2019:1166-1168.
• [6]刘之的，赵靖舟，杨秀春，等．基于灰色关联分析和BP神经网络的煤层含气量预测研究[J]．西安石油大学学报(自然科学版)，2014,29(3):58-62+9.Liu Zhidi,Zhao Jingzhou,Yang Xiuchun,et al. Prediction of methane gas content in coalbed based on the grey relation analysis and BP neural network[J]. Journal of Xi'an Shiyou University(Natural Science Edition),2014,29(3):58-62+9.
• [7]范志辉．不同煤阶煤大分子建模及吸附机理研究[D]．北京:中国地质大学(北京)，2020.Fan Zhihui. Study on macromolecular modeling and adsorption mechanism of coal with different coal rank[D]. Beijing:China University of Geosciences(Beijing),2020.
• [8]叶绍泽，曹俊兴，吴施楷，等．基于深度置信网络的总有机碳含量预测方法[J]．地球物理学进展，2018,36(6):2490-2497.Ye Shaoze,Cao Junxing,Wu Shikai,et al. Prediction method of total organic carbon content based on deepbeliefnets[J]. Progress in Geophysics,2018,33(6):2490-2497.
• [9] Hinton G. A practical guide to training restricted boltzmann machines,vol 7700[M].Springer,Berlin,2012:599-619.
• [10]Krizhevsky A,Sutskever I.Image Net classification with deep convolutional neural networks[J]. Int Conf Neural Inf Process Syst,2012,25(2):1097-1105.
• [11]Holloway K,Hall P,Silva J,et al. Surface reservoir characteristics from subsurface seismic images with deep learning methodologies[R].SAS Institute Inc,2016:1-5.
• [12]Jordan M,Mitchell T. Machine learning:trends,perspectives,and prospects[J]. Science,2015,349(6245):255-260.
• [13]Korjani M,Popa A,et al. A New Approach to Reservoir Characterization Using Deep Learning Neural Networks[C]．//SPE Western Regional Meeting,Anchorage,Alaska,2016:23-26.
• [14]Krizhevsky A,Sutskever I,Hinton G. Image Net classification with deep convolutional neural networks[J]. Advances in neural information processing systems,2012,25(2):1097-1105.
• [15]吴施楷．基于深度学习的天然气储层检测方法研究[D]．成都:成都理工大学，2017.Wu Shikai. Research on gas reservoir identification method based on deep learning[D]. Chengdu:Chengdu University of Technology,2017.
• [16]Ajith A. Artifical neural networks[M]. New York:John Wiley&Sons,2005.
• [17]Bengio Y. Learning deep architectures for AI[J].Foundations and trends in Machine Learning,2009,2(1):121-127.
• [18]Hinton G,Salakhutdinov R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
• [19]Parhat Z,Xiangmin,Zhang fengwei. Study on logging interpretation of coal-bed methane content based on deep learning[M]. Acta Geophysica,2019(67):589-596.
• [20]胡修凤．基于BP神经网络的沁中南煤层气井解吸压力预测及应用[D]．焦作:河南理工大学，2017Hu Xiufeng. Prediction of Coalbed Methane Wells Desorption Pressure Based on BP Neural Networks and Its Applications in the Central一Southern Qinshui Basin[D]. Jiaozuo:Henan Polytechnic University,2017.
• [21]侯颉，邹长春，杨玉卿，等．测井解释中煤层含气量评价方法对比研究[J]．煤炭科学技术，2015,43(12):157-161+156.Hou Ji,Zou Changchun,Yang Yuqin,et al. Comparison study on evaluation methods of coalbed methane gas content with logging interpretation[J]. Coal Science and Technology,2015,43(12):157-161+156.
• [22]陶卿，姚穗，范劲松，等．一种新的机器学习算法:Support Vector Machines[J]．模式识别与人工智能，2000,13(3):285-290.Tao Qin,Yao Shui,Fan jinsong,et al. A new kind of machine learning algorithm:support vectory machines[J]. Pattern Recognition and Artificial Intelligence,2000,13(3):285-290.
• [23]孟召平，田永东，雷旸．煤层含气量预测的BP神经网络模型与应用[J]．中国矿业大学学报，2008(4):456-461.Meng Zhaoping,Tian Yongdong,Lei Yang. Prediction models of coal bed gas content based on BP neural networks and its applications[J]. Journal of China University of Mining&Technology,2008(4):456-461.