地球科学进展

地球科学进展 ›› 2018, Vol. 33 ›› Issue (6): 614 -622. doi: 10.11867/j.issn.1001-8166.2018.06.0614

研究论文 上一篇    下一篇

BP神经网络在富钾卤水中的应用研究
陈科贵 1( ), 李进 1, *( ), 黄长兵 2, 陈愿愿 3, 王刚 4, 刘阳 1   
  1. 1.西南石油大学地球科学与技术学院,四川 成都 610500
    2.中国石油化工集团中原油田, 河南 濮阳 457000
    3.川庆钻探工程有限公司地球物理勘探公司,四川 成都 610213
    4.中国石油新疆油田分公司勘探开发研究院,新疆 克拉玛依 834000
  • 收稿日期:2018-02-03 修回日期:2018-05-09 出版日期:2018-06-20
  • 通讯作者: 李进 E-mail:chenkegui@21cn.com;1094129014@qq.com
  • 基金资助:
    *国家自然科学基金项目“四川盆地油钾兼探的地球物理评价方法研究”(编号:41372103)资助.

Application Research of BP Neural Network in Potassium-Rich Brine

Kegui Chen 1( ), Jin Li 1, *( ), Changbing Huang 2, Yuanyuan Chen 3, Gang Wang 4, Yang Liu 1   

  1. 1. School of Geoscience and Technology, Southwest Petroleum University, Chengdu 610500, China
    2.Sinopec Group Zhongyuan Oilfield, Puyang He’nan 457000, China;
    3.Geophysical Exploration Company, Chuanqing Drilling Engineering Company Limited, Chengdu 610213, China
    4.Research Institute of Exploration and Development, PetroChina Xinjiang Oilfield Company, Karamay Xinjiang 834000, China
  • Received:2018-02-03 Revised:2018-05-09 Online:2018-06-20 Published:2018-07-23
  • Contact: Jin Li E-mail:chenkegui@21cn.com;1094129014@qq.com
  • About author:

    First author:Chen Kegui(1959-), male, Zigong City, Sichuan Province, Professor. Research areas include petroleum geology, technology of well logging reservoir evaluation and survey research potash in Sichuan. E-mail:chenkegui@21cn.com

  • Supported by:
    Project supported by the National Natural Science Foundation of China “The study on geophysical evaluation method of oil and potash in Sichuan Basin (No.41372103).
全文 ( 2 ) 下载PDF ( 774 )

富钾卤水是一种重要的液态钾盐资源,是四川盆地主要的找钾方向之一。川东地区三叠纪地层内与含盐系相邻的碳酸盐岩储层卤水矿化度极高,卤水资源丰富,开采潜力巨大,是我国目前钾盐勘探研究的重点区域。针对常规测井解释方法识别卤水层速度慢、准确率不高等特点,提出建立BP神经网络模型开展富钾卤水层的识别与划分。以BP神经网络理论和测井解释原理为基础,对卤水层识别影响最大的测井曲线值作为输入,建立BP神经网络模型,开展深层卤水层和富钾卤水层的识别和划分,并用准确的录井结果验证模型性能。测试发现,模型识别卤水的准确率为85.7%;改进的富钾卤水模型识别准确率为89.1%。结果表明,BP神经网络技术在四川盆地钾盐的勘探开发过程中具有广阔的应用前景。

关键词: 富钾卤水, 深层卤水, BP神经网络模型, 测井响应

Potassium-rich brine, an important source of liquid potassium salt, is one of the major potassium-seeking directions in the Sichuan Basin. The potassium-rich brine in Triassic carbonate formation of the eastern Sichuan has a high degree of salinity, high exploitation potential and huge resources. It is one of the key areas of potassium salt exploration in China. Aiming at the characteristics of conventional logging interpretation method to identify deep brine with slow speed and low accuracy, it was proposed to establish BP neural network model to carry out the identification and division of potassium-rich brine in Sichuan Basin. Based on the theory of BP neural network and logging interpretation methods, a neural network model with logging curves as input was built and applied to the deep brine and potassium-rich brine. The discrimination results were compared with logging data. The accuracy rate of the model reaches 85.7%,and the accuracy rate of the improved potassium-rich brine model achieves 89.1%. This study demonstrates that BP neural network has a wide application prospect in the exploration and development of potassium salt in Sichuan Basin.

Key words: Potassium-rich brine, Deep brine, BP neural network model, Well logging response.

中图分类号: 

图1 研究区构造图(据参考文献[ 10 ]修改)
Fig.1 Research area structure location map (modified after reference[10])
图1 研究区构造图(据参考文献[ 10 ]修改)
Fig.1 Research area structure location map (modified after reference[10])
图2 BP神经网络结构
Fig.2 Structure of BP neural network
图2 BP神经网络结构
Fig.2 Structure of BP neural network
图3 Rt对数归一化和线性归一化对比图
Fig.3 Correlation between logarithmic normalized andlinear normalized of Rt
图3 Rt对数归一化和线性归一化对比图
Fig.3 Correlation between logarithmic normalized andlinear normalized of Rt
表1 部分学习样本
Table 1 Partial learning sample
GR/API AC/(μs/m) CNL/% DEN/(g/cm3) Rt/(Ω·m) 岩性
57.230 176.136 16.192 2.834 139.737 碳酸盐岩
55.340 176.720 17.254 2.751 145.236 碳酸盐岩
12.477 169.263 1.331 2.984 2 114.281 硬石膏
13.391 168.738 1.357 2.987 2 215.452 硬石膏
165.743 265.141 19.127 2.328 16.579 泥岩
171.175 258.133 18.089 2.393 16.855 泥岩
表1 部分学习样本
Table 1 Partial learning sample
GR/API AC/(μs/m) CNL/% DEN/(g/cm3) Rt/(Ω·m) 岩性
57.230 176.136 16.192 2.834 139.737 碳酸盐岩
55.340 176.720 17.254 2.751 145.236 碳酸盐岩
12.477 169.263 1.331 2.984 2 114.281 硬石膏
13.391 168.738 1.357 2.987 2 215.452 硬石膏
165.743 265.141 19.127 2.328 16.579 泥岩
171.175 258.133 18.089 2.393 16.855 泥岩
表2 部分样本测试结果
Table 2 Partial sample test results
岩性 BP神经网络实际输出结果 期望输出 测试结果
泥岩 0.9532 0.0386 0.0082 1 0 0 泥岩
泥岩 0.9352 0.0351 0.0297 1 0 0 泥岩
硬石膏 0.0281 0.9436 0.0283 0 1 0 硬石膏
硬石膏 0.0084 0.9282 0.0634 0 1 0 硬石膏
碳酸盐岩 0.0147 0.0199 0.9654 0 0 1 碳酸盐岩
碳酸盐岩 0.0148 0.0429 0.9423 0 0 1 碳酸盐岩
表2 部分样本测试结果
Table 2 Partial sample test results
岩性 BP神经网络实际输出结果 期望输出 测试结果
泥岩 0.9532 0.0386 0.0082 1 0 0 泥岩
泥岩 0.9352 0.0351 0.0297 1 0 0 泥岩
硬石膏 0.0281 0.9436 0.0283 0 1 0 硬石膏
硬石膏 0.0084 0.9282 0.0634 0 1 0 硬石膏
碳酸盐岩 0.0147 0.0199 0.9654 0 0 1 碳酸盐岩
碳酸盐岩 0.0148 0.0429 0.9423 0 0 1 碳酸盐岩
表3 学习样本示例
Table 3 Examples of study samples
GR/API AC/(μs/m) CNL/% DEN/(g/cm3) Rt/(Ω·m) 流体性质
10.256 165.726 15.474 2.641 15.502 卤水层
8.254 183.072 15.594 2.642 16.352 卤水层
7.267 176.904 15.813 2.646 13.183 卤水层
5.325 78.231 2.257 2.849 139.737 干层
6.254 88.621 4.252 2.841 140.643 干层
8.623 67.391 3.254 2.841 141.549 干层
表3 学习样本示例
Table 3 Examples of study samples
GR/API AC/(μs/m) CNL/% DEN/(g/cm3) Rt/(Ω·m) 流体性质
10.256 165.726 15.474 2.641 15.502 卤水层
8.254 183.072 15.594 2.642 16.352 卤水层
7.267 176.904 15.813 2.646 13.183 卤水层
5.325 78.231 2.257 2.849 139.737 干层
6.254 88.621 4.252 2.841 140.643 干层
8.623 67.391 3.254 2.841 141.549 干层
图4 神经网络误差收敛曲线
Fig.4 Neural network error convergence curve
图4 神经网络误差收敛曲线
Fig.4 Neural network error convergence curve
图5 峰x5井卤水层识别结果对比
Fig.5 Correlation of brine discrimination in Feng x5 well
图5 峰x5井卤水层识别结果对比
Fig.5 Correlation of brine discrimination in Feng x5 well
表4 卤水层识别正确率
Table 4 Accuracy of brine discrimination
测井 测试点 符合点 正确率/%
峰x5井 89 75 84.3
峰x7井 234 206 88.0
峰x8井 322 285 88.5
峰x12井 251 217 86.5
峰x13井 176 143 81.3
表4 卤水层识别正确率
Table 4 Accuracy of brine discrimination
测井 测试点 符合点 正确率/%
峰x5井 89 75 84.3
峰x7井 234 206 88.0
峰x8井 322 285 88.5
峰x12井 251 217 86.5
峰x13井 176 143 81.3
图6 PX井富钾卤水层识别结果对比
Fig.6 Correlation of potassium-rich brine discrimination in PX well
图6 PX井富钾卤水层识别结果对比
Fig.6 Correlation of potassium-rich brine discrimination in PX well
[1] Cao Qin,Zhou Xun,Zhang Huan,et al. Hydrochemical characteristics and genesis of the subsurface brines in the Wolonghe brine-bearing structure of Sichuan Basin[J]. Geological Bulletin of China,2015,34(5):990-997.
[曹琴,周训,张欢,等. 四川盆地卧龙河储卤构造地下卤水的水化学特征及成因[J]. 地质通报,2015,34(5):990-997.]
URL    
[2] Hu Suhua, Feng Yinghui,Shen Qianling, et al. Preliminary approach on the explanatory method of well logging in the potassium-rich brine reservoir in Chengdu salt basin[J]. China Well and Rock Salt,2008,39(6):11-14.
[胡素华,冯迎辉,申潜玲,等. 成都盐盆富钾卤水储层测井解释方法初探[J]. 中国井矿盐,2008,39(6):11-14.]
doi: 10.3969/j.issn.1001-0335.2008.06.004     URL    
[3] Huang Hua,Zhang Shiwan,Zhang Lianyuan, et al. Logging identification of brine mineral bed in the sandstone of Qianjiang formation,Qianjiang depression[J]. Geology of Chemical Minerals,2013,35(2):67-71.
[黄华,张士万,张连元,等. 潜江凹陷潜江组砂岩卤水矿床测井识别方法研究[J]. 化工矿产地质,2013,35(2):65-71.]
doi: 10.3969/j.issn.1006-5296.2013.02.001     URL    
[4] Sun Dapeng.Distribution and evolution characteristics of China’s iodine-rich Brines: 3.Formation condition of the brines and iodine exploration orientation[J]. Journal of Salt Lake Research,2007,36(1),20-28.
[孙大鹏. 我国高碘卤水分布规律及其形成(续完)——高碘卤水的形成条件和找碘方向[J]. 盐湖研究,2007,36(1):20-28. ]
[5] Mu Yanzong,Yan Zhen,Bu Lingzhong,et al. Progress in study of potash resources of oil(gas) field brine in China[J]. Advances in Earth Science,2016,31(2):147-160.
[穆延宗,乜贞,卜令忠,等.我国油(气)田水钾资源研究进展[J].地球科学进展,2016,31(2):147-160.]
doi: 10.11867/j.issn.1001-8166.2016.02.0147    
[6] Lin Yaoting,Yan Yangji,Wu Yinglin.High-grade Brine in the Sichuan basin: Hydrogeochemistry, origin and its resource significance[J]. Sedimentary Facies and Paleogeography,1996,17(4):12-22.
[林耀庭,颜仰基,吴应林.四川盆地某地富矿卤水水文地球化学特征及其成因资源意义[J].岩相古地理,1996,17(4):12-22.]
[7] Jalalalhosseini S M,Ali H,Mostafazadeh M.Predicting porosity by using seismic multi-attributes and well data and combining these available data by geostatistical methods in a South Iranian Oil Field[J]. Petroleum Science and Technology,2014,32(1): 29-37.
doi: 10.1080/10916466.2011.584102     URL    
[8] Baneshi M.The determination of lithofacies using an optimized neural network and well log data[J]. Petroleum Science and Technology,2014,32(8):897-903.
doi: 10.1080/10916466.2011.604056     URL    
[9] Lin Yaoting,Chen Shaolan.Discussion on the evaporite generating modes,salt forming mechanism and potassium-hunting prospect of lower-middle Triassic in Sichuan Basin[J]. Journal of Salt Lake Research,2008,37(3):1-10.
[林耀庭,陈绍兰.论四川盆地下、中三叠统蒸发岩的生成模式、成盐机理及找钾展望[J].盐湖研究,2008,37(3):1-10.]
URL    
[10] Li Li.Logging Evaluation of Potassium-bearing Formations in Sichuan Basin and Analysis of Potash-forming Geological Conditions in Favorable Areas[D]. Chengdu:Southwest Petroleum University,2015.
[李利. 四川盆地含钾地层测井评价及有利区成钾地质条件分析[D].成都:西南石油大学,2015.]
[11] Chen Kegui,Li Li,Li Chunmei,et al. Evolution of the Potash-Rich Areas in evaporation basin during the Epigenetic Stage with continental block being active[J]. Advances in Earth Science,2014,29(4):515-522.
[陈科贵,李利,李春梅,等.活动陆块背景下蒸发盆地后生阶段富钾区的演变[J].地球科学进展,2014,29(4):515-522.]
doi: 10.11867/j.issn.1001-8166.2014.04.0515     URL    
[12] Cao Qin.Hydrochemical Characteristics and Origin of the Deep-Formation Brines and Hot Springs Occurring in the Triassic Strata in the High-Fold Belts of the Eastern Sichuan Basin[D]. Beijing: China University of Geosciences (Beijing),2014.
[曹琴. 川东高褶带三叠系深层卤水和温泉的水化学特征及成因[D].北京:中国地质大学(北京),2014.]
[13] Zhang Yan,Li Meng,Gong Daxing,et al. Research on identification and prediction method of the Triassic potash deposits in Sichuan Basin[J]. Acta Geologica Sinica,2015,89(11):1 970-1 979.
[张研,李萌,龚大兴,等.四川盆地三叠系隐伏钾盐层识别与预测方法研究[J].地质学报,2015,89(11):1 970-1 979.]
URL    
[14] Chen Kegui,Chen Xu,Zhang Jiahao.Combined methods of permeability logging evaluate in glutenite reservoirs—A case study of Badaowan Formation in Karamay Oilfield[J]. Advances in Earth Science,2015,30(7):773-779.
[陈科贵,陈旭,张家浩.复合渗透率测井评价方法在砂砾岩稠油油藏的应用——以克拉玛依油田某区八道湾组为例[J].地球科学进展,2015,30(7):773-779.]
doi: 10.11867/j.issn.1001-8166.2015.07.0773     URL    
[15] Gao Pengyi.Research on BP Neural Network Classifier Optimization Technology[D]. Wuhan:Huazhong University of Science and Technology,2012.
[高鹏毅. BP神经网络分类器优化技术研究[D].武汉:华中科技大学,2012.]
[16] Chen Kegui,Wu Liulei,Chen Yuanyuan,et al. Classification and recognition of polyhalite in Chuanzhong based on support vector machine[J]. Advances in Earth Science,2016,31(10):1 041-1 046.
[陈科贵,吴刘磊,陈愿愿,等.基于支持向量机的川中杂卤石分类识别研究[J].地球科学进展,2016,31(10):1 041-1 046.]
doi: 10.11867/j.issn.1001-8166.2016.10.1041     URL    
[17] Chen Kegui,Liu Li,Chen Yuanyuan,et al. Research on classification and discrimination of polyhalite with drilling and logging data by BP neural network[J]. Journal of China University of Petroleum (Edition of Natural Science),2016,40(4):66-72.
[陈科贵,刘利,陈愿愿,等. BP神经网络在钻孔测井资料分类识别杂卤石中的研究[J]. 中国石油大学学报:自然科学版,2016,40(4):66-72.]
doi: 10.3969/j.issn.1673-5005.2016.04.008     URL    
[18] Li Huaqing.Application of Neural Network in Forecasting Foreign Exchange Trading Signals[D]. Beijing:Beijing University of Posts and Telecommunications,2007.
[黎华清. 神经网络在外汇交易信号预测中的应用[D].北京:北京邮电大学,2007.]
[19] Du Yejun,Zhou Su,Siqinqiqige,et al.Using the synchronous observed data of two observation stations to calculate the delay of ionosphere[J]. Science of Surveying and Mapping,2010,35(Suppl.1):120-121,51.
[都业军,周肃,斯琴其其格,等.人工神经网络在遥感影像分类中的应用与对比研究[J].测绘科学,2010,35(增刊1):120-121,51.]
[20] Hu Jialiang,Gao Yuchao,Yu Jifeng, et al. Lithology identification of unconventional reservoirs based on PCA-BP neural network[J]. Journal of Shandong University of Science and Technology (Natural Science),2016,35(5):9-16.
[胡嘉良,高玉超,余继峰,等.基于PCA-BP神经网络的非常规储层岩性识别研究[J].山东科技大学学报:自然科学版,2016,35(5):9-16.]
URL    
[21] Lin Yaoting,Xu Zulin.Significance of salts preservation condition research on finding potassium of the Triassic in Sichuan Basin[J]. Journal of Salt Lake Research,2009,17(1):6-12.
[林耀庭,许祖霖.论盐类保存条件研究对四川盆地三叠系找钾工作的重要性[J].盐湖研究,2009,17(1):6-12.]
[22] Yang Bin,Yan Lichun,Sun Zhongchun, et al.Neural Network and Its Application in Oil Well Logging[M]. Beijing:Petroleum Industry Press,2005.
[杨斌,匡立春,孙中春,等. 神经网络及其在石油测井中的应用[M].北京:石油工业出版社,2005.]
[23] Li Yu.Evaluating fracturesby logging data and its modeling technology studies[J]. Well Logging Technology,2009,33(6):575-578.
[李毓. 储层裂缝的测井识别及其地质建模研究[J].测井技术,2009,33(6):575-578.]
doi: 10.3969/j.issn.1004-1338.2009.06.017     URL    
[24] Zhang Boxin, Yu Hongguo, Qi Jie.Recognizing volcanic rock lithology in 69 area of Junggar Basin[J]. Journal of Oil and Gas Technology,2010,32(4):97-101,426.
[张伯新,于红果,祁杰. 准噶尔盆地六九区火山岩岩性识别方法[J]. 石油天然气学报,2010,32(4):97-101,426.]
doi: 10.3969/j.issn.1000-9752.2010.04.020     URL    
[25] Zheng Mianping,Qi Wen,Zhang Yongsheng.The current situation of potassium mineral resources in China and the preliminary analysis of potassium application direction[J]. Geological Bulletin of China,2006,25(11):1 239-1 246.
[郑绵平,齐文,张永生. 中国钾盐地质资源现状与找钾方向初步分析[J]. 地质通报,2006,25(11):1 239-1 246. ]
[26] Zhang Chengjiang,Xu Zhengqi,Ni Shijun,et al.Genesis of potassium-bearing brine in Pingluoba structure region,western Sichuan depression[J]. Advances in Earth Science,2012,27(10):1 054-1 060.
[张成江,徐争启,倪师军,等.川西坳陷平落坝构造富钾卤水成因探讨[J].地球科学进展,2012,27(10):1 054-1 060.]
doi: 10.11867/j.issn.1001-8166.2012.10.1054     URL    
[1] 陈科贵, 吴刘磊, 陈愿愿, 王刚. 基于支持向量机的川中杂卤石分类识别研究[J]. 地球科学进展, 2016, 31(10): 1041-1046.
[2] 周游，倪师军，施泽明. 四川盆地东北部某储卤构造深层卤水资源量容积法评价的改进模型研究[J]. 地球科学进展, 2013, 28(6): 703-708.
[3] 张成江，徐争启，倪师军，尹 观. 川西坳陷平落坝构造富钾卤水成因探讨[J]. 地球科学进展, 2012, 27(10): 1054-1060.
阅读次数
全文


摘要

网站版权所有 © 《地球科学进展》编辑部
地址：甘肃省兰州市天水中路8号
QQ：114723517
电话：0931-8762293　E-mail：adearth@lzb.ac.cn
陇ICP备2021001824号-5　 甘公网安备62010202000687