基于人工智能的抽油机井结蜡预警方法

邴绍强

邴绍强. 基于人工智能的抽油机井结蜡预警方法[J]. 石油钻探技术, 2019, 47(4): 97-103. DOI: 10.11911/syztjs.2019093
引用本文: 邴绍强. 基于人工智能的抽油机井结蜡预警方法[J]. 石油钻探技术, 2019, 47(4): 97-103. DOI: 10.11911/syztjs.2019093
BING Shaoqiang. An Early Warning Method Based on Artificial Intelligence for Wax Deposition in Rod Pumping Wells[J]. Petroleum Drilling Techniques, 2019, 47(4): 97-103. DOI: 10.11911/syztjs.2019093
Citation: BING Shaoqiang. An Early Warning Method Based on Artificial Intelligence for Wax Deposition in Rod Pumping Wells[J]. Petroleum Drilling Techniques, 2019, 47(4): 97-103. DOI: 10.11911/syztjs.2019093

基于人工智能的抽油机井结蜡预警方法

基金项目: 中国石化科技攻关项目“勘探开发智能化关键技术研究”(编号:P14130)部分研究内容
详细信息
    作者简介:

    邴绍强(1974—),男,山东莒县人,1996年毕业于石油大学(华东)采油工程专业,2007年获中国石油大学(华东)油气田开发专业硕士学位,高级工程师,主要从事信息自动化研究工作。E-mail:bingshaoqiang.slyt@sinopec.com

  • 中图分类号: TE358+.2

An Early Warning Method Based on Artificial Intelligence for Wax Deposition in Rod Pumping Wells

  • 摘要:

    针对依靠现场经验确定的清蜡周期不准确而导致蜡卡躺井的问题,开展了基于人工智能的抽油机井结蜡预警方法研究。利用皮尔逊相关系数分析方法,分析了17项油井自动采集参数与结蜡程度的关联性,确定了7项主控参数,创建了结蜡预警规则模型;将7项主控参数的合并指标进行归一化处理得到结蜡综合特征指标(WPSC),并利用结蜡预警规则模型产生的样本数据建立了结蜡井WPSC样本集,选用长短时记忆神经网络(LSTM)对样本集进行训练,得到了WPSC机器学习模型,用其可以定量预测抽油机井的结蜡程度。该方法在胜利油田桩23区块的现场应用结果表明,油井清蜡周期得到延长,且有效避免了蜡卡躺井。研究结果表明,基于人工智能的抽油机井结蜡预警方法实现了油井结蜡程度的定量化预测与预警,对精准确定清蜡时机具有较好的指导作用。

    Abstract:

    The timing of wax clearance is usually determined in the field by observations, which is inaccurate and may cause wells to fail due to wax locking. To solve this problem, a wax deposition early warning method based on artificial intelligence for rod pumping wells has been developed. The Pearson correlation coefficient analysis method is used to conduct correlation analysis between 17 points of data automatically acquired by oil well condition and degree of wax deposition, and seven main control parameters are determined. A wax deposition early warning rule protocol was established on this basis. The wax deposition synthetic characteristics (WPSC) index was obtained by normalizing the merged indexes of seven main control parameters, and the WPSC sample set of wells with paraffin problems was established by using the sample data generated by the wax deposition early warning rule model. Long-term and short-term memory (LSTM) neural networks were selected to train the sample set, and the machine learning model of WPSC is obtained, which could quantitatively predict the degree of wax deposition in the rod pumping wells. The field trial of this method in Block Zhuang 23 of the Shengli Oilfield showed that by using this method, the wax clearance timing as prolonged and the wax locking in failed wells was effectively avoided. Research results showed that the wax deposition early warning method based on artificial intelligence for rod pumping wells achieves quantitative prediction and provides effective early warning on the wax deposition degree in oil wells, and could be used as an effective guide in precise the selection of wax clearance timing.

  • 图  1   结蜡预警规则模型运行示意

    Figure  1.   Operation schematic diagram of wax deposition early warning rule model

    图  2   HJH82-X11井结蜡预警规则模型

    Figure  2.   Wax deposition early warning rule model of Well HJH82-X11

    图  3   GN24P102井躺井前30 d的WPSC数值曲线

    Figure  3.   WPSC numerical curve of 30 days before Well GN24P102 fails

    图  4   H148井WPSC实际值与预测值对比曲线(10月20日–11月8日)

    Figure  4.   Correlation curve between actual and predicted values of WPSC in Well H148 (October 20–November 8)

    图  5   H148井WPSC实际值与预测值对比曲线(11月9日–28日)

    Figure  5.   Correlation curve between actual and predicted values of WPSC in Well H148 (November 9–28)

    图  6   H148井WPSC预测值曲线

    Figure  6.   WPSC predictive value curve of Well H148

    表  1   300口典型结蜡井属性数据相关性分析结果

    Table  1   Correlation analysis of attribute data of 300 typical paraffin troubled wells

    属性 相关系数
    D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 Z1 Z2 Z3 Z4 J1 J2 J3
    D1 1.000 –0.875 0.872 0.983 0.89 0.213 –0.819 0.975 0.898 0.879 0.955 –0.963 0.982 0.919 0.241 0.016 0.167
    D2 –0.875 1.000 –0.762 –0.602 0.835 0.192 –0.824 –0.819 0.773 0.878 –0.829 0.821 –0.896 –0.748 0.169 0.071 0.172
    D3 0.872 –0.762 1.000 0.886 0.891 0.663 –0.838 0.837 0.861 0.859 0.819 –0.823 0.869 0.804 0.321 0.021 0.116
    D4 0.983 –0.602 0.886 1.000 0.892 0.812 –0.826 0.915 0.873 0.843 0.912 –0.917 0.962 0.943 0.371 0.018 0.132
    D5 0.890 0.835 0.891 0.892 1.000 0.289 0.296 0.865 0.832 0.867 0.893 0.712 0.871 0.861 0.126 0.031 0.128
    D6 0.213 0.192 0.663 0.812 0.289 1.000 0.278 0.263 0.389 0.363 0.219 0.132 0.267 0.389 0.131 0.021 0.127
    D7 –0.819 –0.824 –0.838 –0.826 0.296 0.278 1.000 0.302 –0.886 –0.872 –0.761 –0.781 –0.769 –0.732 0.191 0.032 0.139
    D8 0.975 –0.819 0.837 0.915 0.865 0.263 0.302 1.000 0.832 0.859 0.972 –0.959 0.971 0.929 0.221 0.036 0.137
    D9 0.898 0.773 0.861 0.873 0.832 0.389 –0.886 0.832 1.000 0.886 0.781 0.486 0.792 0.625 0.132 0.042 0.069
    D10 0.879 0.878 0.859 0.843 0.867 0.363 –0.872 0.859 0.886 1.000 0.812 0.638 0.831 0.781 0.256 0.062 0.136
    Z1 0.955 –0.829 0.819 0.912 0.893 0.219 –0.761 0.972 0.781 0.812 1.000 –0.992 0.996 0.965 0.269 0.026 0.113
    Z2 –0.963 0.821 –0.823 –0.917 0.712 0.132 –0.781 –0.959 0.486 0.638 –0.992 1.000 –0.986 –0.958 0.189 0.032 0.142
    Z3 0.982 –0.896 0.869 0.962 0.871 0.267 –0.769 0.971 0.792 0.831 0.996 –0.986 1.000 0.993 0.306 0.019 0.201
    Z4 0.919 –0.748 0.804 0.943 0.861 0.389 –0.732 0.929 0.625 0.781 0.965 –0.958 0.993 1.000 0.025 0.066 0.139
    J1 0.241 0.169 0.321 0.371 0.126 0.131 0.191 0.221 0.132 0.256 0.269 0.189 0.306 0.025 1.000 0.021 0.136
    J2 0.016 0.071 0.021 0.018 0.031 0.021 0.032 0.036 0.042 0.062 0.026 0.032 0.019 0.066 0.021 1.000 0.132
    J3 0.167 0.172 0.116 0.132 0.128 0.127 0.139 0.137 0.069 0.136 0.113 0.142 0.201 0.139 0.136 0.132 1.000
     注:D1为上行电流;D2为下行电流;D3为A项电流;D4为A项平均电流;D5为耗电量;D6为AB项电压;D7为功率因数;D8为周期内有功功率平均值;D9为无功功率;D10为周期内无功功率平均值;Z1为最大载荷;Z2为最小载荷;Z3为载荷差;Z4为功图面积;J1为井口回压;J2为井口套压;J3为井口温度。
    下载: 导出CSV
  • [1] 钟张起, 侯读杰, 王建飞, 等. 抽油机井结蜡定量化分析新技术[J]. 石油机械, 2014, 42(5): 68–71. doi: 10.3969/j.issn.1001-4578.2014.05.015

    ZHONG Zhangqi, HOU Doujie, WANG Jianfei, et al. Quantitative analysis of paraffinning in rod pumped well[J]. China Petroleum Machinery, 2014, 42(5): 68–71. doi: 10.3969/j.issn.1001-4578.2014.05.015

    [2] 邢敦通, 张金元. 原油结蜡规律及清防蜡工艺的研究[J]. 石化技术, 2017, 24(2): 15–22. doi: 10.3969/j.issn.1006-0235.2017.02.013

    XING Duntong, ZHANG Jinyuan. Rules of wax deposit in crude oil and wax removal[J]. Petrochemical Industry Technology, 2017, 24(2): 15–22. doi: 10.3969/j.issn.1006-0235.2017.02.013

    [3] 宋君. 油井结蜡影响因素及防蜡技术研究[J]. 当代化工, 2015(1): 78–79, 82. doi: 10.3969/j.issn.1671-0460.2015.01.027

    SONG Jun. Research on factors influencing oil well paraffin deposit and paraffin control technology[J]. Contemporary Chemical Indu-stry, 2015(1): 78–79, 82. doi: 10.3969/j.issn.1671-0460.2015.01.027

    [4] 郝世彦, 樊平天, 张新涛. 油井远程监控与计量技术在延长油田的应用[J]. 石油钻探技术, 2010, 38(2): 84–87. doi: 10.3969/j.issn.1001-0890.2010.02.020

    HAO Shiyan, FAN Pingtian, ZHANG Xintao. Application of remote well monitoring and measurement technique in Yanchang Oilfield[J]. Petroleum Drilling Techniques, 2010, 38(2): 84–87. doi: 10.3969/j.issn.1001-0890.2010.02.020

    [5] 段友祥, 李钰, 孙歧峰, 等. 改进的Alexnet模型及在油井示功图分类中的应用[J]. 计算机应用与软件, 2018, 35(7): 226–230, 272. doi: 10.3969/j.issn.1000-386x.2018.07.041

    DUAN Youxiang, LI Yu, SUN Qifeng, et al. Improved alexnet model and using in dynamometer card classification[J]. Computer Applications and Software, 2018, 35(7): 226–230, 272. doi: 10.3969/j.issn.1000-386x.2018.07.041

    [6] 焦春晶. 油井生产状态智能识别及辅助决策系统的研究[J]. 石油知识, 2019(2): 36–37.

    JIAO Chunjing. Research on intelligent recognition and assistant decision-making system of oil well production state[J]. Petroleum Knowledge, 2019(2): 36–37.

    [7] 张春琳, 唐道临, 孙成. 基于模式识别技术的油井工况诊断[J]. 现代计算机, 2018(27): 65–68, 100. doi: 10.3969/j.issn.1007-1423.2018.27.013

    ZHANG Chunlin, TANG Daolin, SUN Cheng. Diagnosis of oil well condition based on pattern recognition technology[J]. Modern Computer, 2018(27): 65–68, 100. doi: 10.3969/j.issn.1007-1423.2018.27.013

    [8] 赵可新, 马晓辉, 马彦芬, 等. 人工神经网络在油井故障诊断中的优化[J]. 煤炭技术, 2011, 30(5): 196–198.

    ZHAO Kexin, MA Xiaohui, MA Yanfen, et al. Optimization of ANN in fault diagnosis of oil wells[J]. Coal Technology, 2011, 30(5): 196–198.

    [9]

    MANSHAD A K, ROSTAMI H. Prediction of wax precipitation in crude oil systems using gaussian processes[J]. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects, 2015, 37(1): 84–91. doi: 10.1080/15567036.2011.578109

    [10]

    MANSHAD A K, ASHOORI S, MANSHAD M K, et al. The prediction of wax precipitation by neural network and genetic algorithm and comparison with a multisolid model in crude oil systems[J]. Petroleum Science and Technology, 2012, 30(13): 1369–1378. doi: 10.1080/10916466.2010.499403

    [11] 李华, 刘帅, 李茂, 等. 数据挖掘理论及应用研究[J]. 断块油气田, 2010, 17(1): 88–91.

    LI Hua, LIU Shuai, LI Mao, et al. Theory and application of data mining[J]. Fault-Block Oil & Gas Field, 2010, 17(1): 88–91.

    [12] 王光宏, 蒋平. 数据挖掘综述[J]. 同济大学学报(自然科学版), 2004, 32(2): 246–252. doi: 10.3321/j.issn:0253-374X.2004.02.023

    WANG Guanghong, JIANG Ping. Survey of data mining[J]. Journal of Tongji University(Natural Science), 2004, 32(2): 246–252. doi: 10.3321/j.issn:0253-374X.2004.02.023

    [13] 薛茹. 数据挖掘技术在油田中的应用[J]. 微型电脑应用, 2018, 34(5): 26–28. doi: 10.3969/j.issn.1007-757X.2018.05.008

    XUE Ru. Application of data mining technology in oilfield[J]. Mi-crocomputer Applications, 2018, 34(5): 26–28. doi: 10.3969/j.issn.1007-757X.2018.05.008

    [14] 郭学增. 钻井工程人工智能十年谈[J]. 石油钻探技术, 1999, 27(5): 25–26. doi: 10.3969/j.issn.1001-0890.1999.05.008

    GUO Xuezeng. Review of ten years’ development of artificial intel-ligence in drilling engineering[J]. Petroleum Drilling Techniques, 1999, 27(5): 25–26. doi: 10.3969/j.issn.1001-0890.1999.05.008

    [15] 王小龙. 浅析人工智能在油气行业中的应用[J]. 现代信息科技, 2017, 1(2): 117–119. doi: 10.3969/j.issn.2096-4706.2017.02.049

    WANG Xiaolong. Application of artificial intelligence in oil and gas industry[J]. Modern Information Technology, 2017, 1(2): 117–119. doi: 10.3969/j.issn.2096-4706.2017.02.049

    [16] 孙志远, 鲁成祥, 史忠植, 等. 深度学习研究与进展[J]. 计算机科学, 2016, 43(2): 1–8. doi: 10.11896/j.issn.1002-137X.2016.02.001

    SUN Zhiyuan, LU Chengxiang, SHI Zhongzhi, et al. Research and advances on deep learning[J]. Computer Science, 2016, 43(2): 1–8. doi: 10.11896/j.issn.1002-137X.2016.02.001

    [17] 余凯, 贾磊, 陈雨强, 等. 深度学习的昨天、今天和明天[J]. 计算机研究与发展, 2013, 50(9): 1799–1804. doi: 10.7544/issn1000-1239.2013.20131180

    YU Kai, JIA Lei, CHEN Yuqiang, et al. Deep learning: yesterday, today, and tomorrow[J]. Journal of Computer Research and Deve-lopment, 2013, 50(9): 1799–1804. doi: 10.7544/issn1000-1239.2013.20131180

    [18] 俞祝良. 人工智能技术发展概述[J]. 南京信息工程大学学报(自然科学版), 2017, 9(3): 297–304.

    YU Zhuliang. Review of progress on artificial intelligence[J]. Journal of Nanjing University of Information Science & Tech-nology(Natural Science Edition), 2017, 9(3): 297–304.

    [19] 李泽龙, 杨春节, 刘文辉, 等. 基于LSTM-RNN模型的铁水硅含量预测[J]. 化工学报, 2018, 69(3): 992–997.

    LI Zelong, YANG Chunjie, LIU Wenhui, et al. Research on hot metal Si-content prediction based on LSTM-RNN[J]. CIESC Journal, 2018, 69(3): 992–997.

    [20] 倪铮, 文韬. 一种基于CNN和RNN深度神经网络的天气预测模型:以北京地区雷暴的6小时临近预报为例[J]. 数值计算与计算机应用, 2018, 39(4): 299–309.

    NI Zheng, WEN Tao. A weather prediction model based on CNN and RNN deep neural network: an example is the 6 hour forecast of thunderstorm in Beijing[J]. Journal on Numerical Methods and Computer Applications, 2018, 39(4): 299–309.

图(6)  /  表(1)
计量
  • 文章访问数:  1129
  • HTML全文浏览量:  799
  • PDF下载量:  64
  • 被引次数: 0
出版历程
  • 收稿日期:  2019-01-20
  • 修回日期:  2019-07-04
  • 网络出版日期:  2019-08-07
  • 刊出日期:  2019-06-30

目录

    /

    返回文章
    返回