基于多录井参数特征同步的溢流事故监测研究

陈青; 黄志强; 孔祥伟; 何弦桀; 徐洲; 安果涛

doi:10.21656/1000-0887.450125

基于多录井参数特征同步的溢流事故监测研究

doi: 10.21656/1000-0887.450125

陈青^{1, 2,},
黄志强^{1, 2},
孔祥伟^{1, 2, ,},
何弦桀³,
徐洲^{1, 2},
安果涛^{1, 2}

1.
长江大学石油工程学院, 武汉 430100
2.
长江大学油气钻完井技术国家工程研究中心, 武汉 430100
3.
中国石油集团川庆钻探工程有限公司钻采工程技术研究院, 四川德阳 618399

基金项目:

国家自然科学基金 51904261

详细信息

作者简介:
陈青(1997—)，男，博士生(E-mail: chenq.st@yangtzeu.edu.cn)

通讯作者:
孔祥伟(1982—)，男，教授，博士生导师(通讯作者. E-mail: 501074@yangtzeu.edu.cn)

中图分类号: O368;O29
计量
- 文章访问数: 65
- HTML全文浏览量: 27
- PDF下载量: 14
- 被引次数: 0
出版历程
- 收稿日期: 2024-05-07
- 修回日期: 2024-06-13
- 刊出日期: 2025-02-01

Study on Overflow Accident Monitoring Based on Synchronous Features of Multiple Well Logging Parameters

CHEN Qing^{1, 2
,},
HUANG Zhiqiang^{1, 2},
KONG Xiangwei^{1, 2
, ,},
HE Xianjie³,
XU Zhou^{1, 2},
AN Guotao^{1, 2}

1.
Petroleum Engineering College, Yangtze University, Wuhan 430100, P.R.China
2.
National Engineering Research Center for Oil & Gas Drilling and Completion Technology, Yangtze University, Wuhan 430100, P.R.China
3.
Research Institute of Drilling and Production Engineering Technology, CNPC Chuanqing Drilling Engineering Co. , Ltd. , Deyang, Sichuan 618399, P.R.China

摘要

摘要: 依据录井参数进行溢流事故的判断十分依赖坐岗人员的经验，且现场采集的综合录井参数信噪严重，参数变化特征不明显，溢流监测准确率低. 通过低通滤波处理和局部加权线性回归，去除现场综合录井参数曲线的高频信号和低频信噪，经归一化处理，得到了多参数同步的溢流识别方法，并结合GCN图形匹配和BRNN双向传递的特点，建立了GCN-BRNN相融合的模型，提高了溢流事故监测的准确率. 结果表明，通过局部加权线性回归处理后能够使曲线变化特征更加明显，且归一化后的多参数同步监测比单一参数监测的准确率更高；以川西某井的综合录井数据为例进行溢流识别测试，与原先模型相比，结合后的模型溢流识别准确率更高，可达到85%；储层特征会影响录井参数的采集精度，储层分布结构越均匀、性质越稳定，溢流监测的准确率越高. 经JT井现场应用，溢流事故识别准确率≥89%，实际溢流风险与模型识别结果一致. 该方法能有效处理多源信息间的冲突，提高溢流监测的识别精度，对现场结合录井参数的溢流事故监测方法具有指导意义.
- 钻井溢流监测 /
- 参数特征 /
- 滤波去噪 /
- 局部加权线性回归 /
- 神经网络
Abstract: Judging overflow accidents based on well logging parameters relies heavily on the experience of on-duty personnel, and the comprehensive well logging parameters collected in-situ have severe noises and unclear parameter change characteristics, resulting in low accuracy of overflow monitoring. A multi-parameter synchronous overflow identification method was obtained through low-pass filtering and locally weighted linear regression to remove the high-frequency signals and low-frequency noises of the in-situ comprehensive well logging parameter curves, and after normalization processing. Combined with the characteristics of the GCN graph matching and the BRNN bidirectional transmission, the GCN-BRNN fusion model was established to improve the accuracy of overflow accident monitoring. The results show that, the local weighted linear regression can make the curve change characteristics more obvious, and the accuracy of the multi-parameter synchronous monitoring after normalization is higher than that of the single-parameter monitoring. With the comprehensive well logging data of a well in western Sichuan as an example, compared with the original model, the combined model has a higher accuracy reaching 85% in overflow identification. The characteristics of the reservoir affect the accuracy of logging parameter collection; the more uniform the reservoir distribution structure is and the more stable the properties are, the higher the accuracy of overflow monitoring will be. After in-situ application in the JT well, the identification accuracy of overflow accidents is ≥89%, and the actual overflow risk is consistent with the model identification results. This method can effectively handle conflicts between multiple sources of information, improve the accuracy of overflow monitoring, and provide guidance for in-situ overflow accident monitoring methods combined with well logging parameters.
- drilling overflow monitoring /
- parameter characteristic /
- filtering denoising /
- locally weighted linear regression /
- neural network

HTML全文

图 1 GNN结构

Figure 1. The graph neural network architecture

下载: 全尺寸图片幻灯片

图 2 BRNN结构

Figure 2. The bidirectional recurrent neural network architecture

下载: 全尺寸图片幻灯片

图 3 求解流程

Figure 3. The solving flowchart

下载: 全尺寸图片幻灯片

图 4 溢流段录井数据滤波处理

注为了解释图中的颜色，读者可以参考本文的电子网页版本，后同.

Figure 4. Filtered logging data for the overflow section

下载: 全尺寸图片幻灯片

图 5 溢流段录井数据LOESS处理

Figure 5. LOESS processing of overflow section logging data

下载: 全尺寸图片幻灯片

图 6 溢流段录井参数局部特征曲线

Figure 6. Local feature curves of logging parameters in the overflow section

下载: 全尺寸图片幻灯片

表 1 多井历史综合录井数据

Table 1. Multi-well historical composite well logging data

time /s	standpipe pressure /MPa	total hydrocarbons (×10^-6)	mud pit volume /m³	inlet-outlet flow differential /(L·s^-1)
0	10.09	0.62	100.54	0.6
20	10.41	0.64	100.56	0.66
40	10.57	0.75	100.52	0.66
60	10.99	1.08	100.29	0.51
80	11.11	1.36	100.29	0.74
100	11.21	1.55	100.21	0.78
120	11.3	1.59	100.11	0.86
140	11.35	1.58	100.02	0.9
160	11.39	1.46	100.08	0.85
180	11.27	1.35	100.05	0.9
200	11.4	1.35	100.08	0.88
220	11.32	1.42	100.06	0.85
240	11.4	1.39	100.03	0.85
260	11.36	1.33	100.06	0.86
280	11.34	1.27	100.15	0.81
300	11.34	1.23	100.05	0.8
⋮	⋮	⋮	⋮	⋮

下载: 导出CSV

表 2 溢流发生后特征参数变化分析

Table 2. Analysis of feature parameter changes after overflow occurrence

feature parameter	change characteristics after overflow occurrence	reason
standpipe pressure	decrease	overflow intrusion reduces static fluid column pressure
total hydrocarbons	increase	overflow is often accompanied by abnormal increases in total hydrocarbons content and gas composition of various hydrocarbon classes
inlet-outlet flow differential	increase	overflow intrusion into annulus causes gas volume expansion, displacing drilling fluid
total pit volume	increase	formation fluids intrude into annulus, increasing return volume

下载: 导出CSV

表 3 模型最终预测结果对比

Table 3. Comparison of final predicted results from models

model	P	R	F₁	δ_Acc
GCN	0.63	0.62	0.62	0.64
BRNN	0.75	0.77	0.76	0.77
GCN-BRNN	0.8	0.85	0.83	0.85

下载: 导出CSV

表 4 部分井历史数据识别结果

Table 4. Identification results of partial well historical data

well ID	reservoir type	overflow frequency	identified overflow frequency	accuracy /%
PZ1-X	shale	11	10	91
GS1-Y	carbonate rock	16	15	94
YT2-X	sandstone	6	6	100

下载: 导出CSV

表 5 JT井模型应用情况

Table 5. The application status of the JT well model

depth /m	time	abnormal type	occurrences /s	warnings	early warning	accuracy /%
5 275.88~5 282.20	2023-06-28	overflow	2	2	78(yes)	100
5 284.33~5 303.40	2023-06-30	overflow	3	3	42(yes)	100
5 305.75~5 318.19	2023-07-09	overflow	4	4	83(yes)	100
5 319.11~5 324.92	2023-07-11	overflow	8	9	60(yes)	89
5 326.36~5 413.26	2023-07-12	overflow	5	5	80(yes)	100

下载: 导出CSV

参考文献(35)

[1]	孔祥伟, 刘祚才, 靳彦欣. 川渝裂缝性地层自动压井环空多相压力波速特性研究[J]. 应用数学和力学, 2022, 43 (12): 1370-1379. doi: 10.21656/1000-0887.430006 KONG Xiangwei, LIU Zuocai, JIN Yanxin. Study on multiphase pressure wave velocity characteristics of automatic kill annulus in Chuanyu fractured formation[J]. Applied Mathematics and Mechanics, 2022, 43 (12): 1370-1379. (in Chinese) doi: 10.21656/1000-0887.430006
[2]	SCHAFER D M, LOEPPKE G E, GLOWKA D A, et al. An evaluation of flowmeters for the detection of kicks and lost circulation during drilling[C]//IADC/SPE Drilling Conference. New Orleans, Louisiana: SPE, 1992: SPE-23935-MS.
[3]	刘书杰, 杨向前, 郭华, 等. 井控溢流快速判断方法研究[J]. 煤炭技术, 2017, 36 (5): 296-298. LIU Shujie, YANG Xiangqian, GUO Hua, et al. Research for judgment method of well control overflow[J]. Coal Technology, 2017, 36 (5): 296-298. (in Chinese)
[4]	付加胜, 刘伟, 韩霄松, 等. 基于CNN-LSTM融合网络的溢流早期预测深度学习方法[J]. 石油机械, 2021, 49 (6): 16-22. FU Jiasheng, LIU Wei, HAN Xiaosong, et al. CNN-LSTM fusion network based deep learning method for early prediction of overflow[J]. China Petroleum Machinery, 2021, 49 (6): 16-22. (in Chinese)
[5]	常杨, 郭修成, 李永钊, 等. 基于随钻工程参数测量数据的钻井风险识别试验[J]. 钻采工艺, 2022, 45 (5): 150-153. CHANG Yang, GUO Xiucheng, LI Yongzhao, et al. Drilling risk analysis test based on engineering parameter measurement while drilling[J]. Drilling & Production Technology, 2022, 45 (5): 150-153. (in Chinese)
[6]	姚浩, 夏桂然, 刘泽佳, 等. 基于机器学习的黏钢构件黏接层缺陷识别方法研究[J]. 应用数学和力学, 2024, 45 (4): 429-442. doi: 10.21656/1000-0887.440365 YAO Hao, XIA Guiran, LIU Zejia, et al. A defect identification method for bonding layers of adhesive steel members based on machine learning[J]. Applied Mathematics and Mechanics, 2024, 45 (4): 429-442. (in Chinese) doi: 10.21656/1000-0887.440365
[7]	姚明辉, 王兴志, 吴启亮, 等. 基于RBF神经网络的压气机叶片面压力场预测研究[J]. 应用数学和力学, 2023, 44 (10): 1187-1199. doi: 10.21656/1000-0887.440054 YAO Minghui, WANG Xingzhi, WU Qiliang, et al. RBF neural network based prediction on blade surface pressure fields in compressors[J]. Applied Mathematics and Mechanics, 2023, 44 (10): 1187-1199. (in Chinese) doi: 10.21656/1000-0887.440054
[8]	李杨, 闫冬梅, 刘磊. 基于输出层具有噪声的DQN的无人车路径规划[J]. 应用数学和力学, 2023, 44 (4): 450-460. doi: 10.21656/1000-0887.430070 LI Yang, YAN Dongmei, LIU Lei. UGV path programming based on the DQN with noise in the output layer[J]. Applied Mathematics and Mechanics, 2023, 44 (4): 450-460. (in Chinese) doi: 10.21656/1000-0887.430070
[9]	王沐晨, 李立州, 张珺, 等. 基于卷积神经网络气动力降阶模型的翼型优化方法[J]. 应用数学和力学, 2022, 43 (1): 77-83. doi: 10.21656/1000-0887.420137 WANG Muchen, LI Lizhou, ZHANG Jun, et al. An airfoil optimization method based on the convolutional neural network aerodynamic reduced order model[J]. Applied Mathematics and Mechanics, 2022, 43 (1): 77-83. (in Chinese) doi: 10.21656/1000-0887.420137
[10]	周济民, 张海晨, 王沫然. 基于物理经验模型约束的机器学习方法在页岩油产量预测中的应用[J]. 应用数学和力学, 2021, 42 (9): 881-890. doi: 10.21656/1000-0887.420015 ZHOU Jimin, ZHANG Haichen, WANG Moran. Machine learning with physical empirical model constraints for prediction of shale oil production[J]. Applied Mathematics and Mechanics, 2021, 42 (9): 881-890. (in Chinese) doi: 10.21656/1000-0887.420015
[11]	KAMYAB M, SHADIZADEH S R, JAZAYERI-RAD H, et al. Early kick detection using real time data analysis with dynamic neural network: a case study in Iranian oil fields[C]//Nigeria Annual International Conference and Exhibition. Tinapa-Calabar, Nigeria: SPE, 2010: SPE-136995-MS.
[12]	HARGREAVES D, JARDINE S, JEFFRYES B. Early kick detection for deepwater drilling: new probabilistic methods applied in the field[C]//SPE Annual Technical Conference and Exhibition. New Orleans, Louisiana: SPE, 2001: SPE-71369-MS.
[13]	NHAT D M, VENKATESAN R, KHAN F. Data-driven Bayesian network model for early kick detection in industrial drilling process[J]. Process Safety and Environmental Protection, 2020, 138 : 130-138. doi: 10.1016/j.psep.2020.03.017
[14]	ALOUHALI R, ALJUBRAN M, GHARBI S, et al. Drilling through data: automated kick detection using data mining[C]//SPE International Heavy Oil Conference and Exhibition. Kuwait City, Kuwait: SPE, 2018: SPE-193687-MS.
[15]	袁俊亮, 范白涛, 幸雪松, 等. 基于朴素贝叶斯算法的钻井溢流实时预警研究[J]. 石油钻采工艺, 2021, 43 (4): 455-460. YUAN Junliang, FAN Baitao, XING Xuesong, et al. Real-time early warning of drilling overflow based on naive Bayes algorithm[J]. Oil Drilling & Production Technology, 2021, 43 (4): 455-460. (in Chinese)
[16]	岳炜杰. "三高" 油气井溢流先兆在线监测与预警系统设计与开发[D]. 东营: 中国石油大学(华东), 2014. YUE Weijie. The design and development of online monitoring and warning system for kick foreboding on "three high" wells[D]. Dongying: China University of Petroleum (Huadong), 2014. (in Chinese)
[17]	徐振华. 基于云计算的控压钻井溢流监测诊断系统研究与设计[D]. 成都: 西南石油大学, 2019. XU Zhenhua. Research and design of control pressure drilling overflow monitoring and diagnosis system based on cloud computing[D]. Chengdu: Southwest Petroleum University, 2019. (in Chinese)
[18]	张倩. 基于机器学习的钻井事故识别系统研究与实现[D]. 西安: 西安石油大学, 2021. ZHANG Qian. Research and implementation of drilling accident recongnition system based on machine learning[D]. Xi'an: Xi'an Shiyou University, 2021. (in Chinese)
[19]	肖宏亮. 基于机器学习的石油钻井流监测研究与应用[D]. 北京: 北京邮电大学, 2020. XIAO Hongliang. Research and application of oil well overflow monitoring based on machine learning[D]. Beijing: Beijing University of Posts and Telecommunications, 2020. (in Chinese)
[20]	李玉飞, 张博, 孙伟峰. 基于SVM和D-S证据理论的早期溢流智能识别方法[J]. 钻采工艺, 2020, 43 (5): 27-30. LI Yufei, ZHANG Bo, SUN Weifeng. Research on intelligent early kick identification method based on SVM and D-S evidence theory[J]. Drilling & Production Technology, 2020, 43 (5): 27-30. (in Chinese)
[21]	张旭, 王芷桁, 李鑫, 等. 大数据驱动下的溢流监测预警发展现状[J]. 中国石油和化工标准与质量, 2022, 42 (1): 124-125. ZHANG Xu, WANG Zhiheng, LI Xin, et al. Development status of overflow monitoring and early warning under the driving of big data[J]. China Petroleum and Chemical Standard and Quality, 2022, 42 (1): 124-125. (in Chinese)
[22]	肖阳, 王家豪, 李志刚, 等. 基于大数据的页岩油区块产量差异分析方法研究[J]. 钻采工艺, 2022, 45 (3): 73-78. XIAO Yang, WANG Jiahao, LI Zhigang, et al. Study on production difference analysis method of shale oil play based on big data[J]. Drilling & Production Technology, 2022, 45 (3): 73-78. (in Chinese)
[23]	沈旭东, 刘慧卿, 张郁哲, 等. 基于机器学习的二次采油高耗水层识别方法研究[J]. 钻采工艺, 2022, 45 (4): 74-80. SHEN Xudong, LIU Huiqing, ZHANG Yuzhe, et al. Identification method for high water consumption layer in second oil recovery based on machine learning[J]. Drilling & Production Technology, 2022, 45 (4): 74-80. (in Chinese)
[24]	SABAH M, TALEBKEIKHAH M, AGIN F, et al. Application of decision tree, artificial neural networks, and adaptive neuro-fuzzy inference system on predicting lost circulation: a case study from Marun oil field[J]. Journal of Petroleum Science and Engineering, 2019, 177 : 236-249.
[25]	KHODNENKO I, IVANOV S, PERETS D, et al. Detection of lost circulation in drilling wells employing sensor data using machine learning technique[J]. Procedia Computer Science, 2019, 156 : 300-307.
[26]	盛茂, 李根生, 田守嶒, 等. 人工智能在油气压裂增产中的研究现状与展望[J]. 钻采工艺, 2022, 45 (4): 1-8. SHENG Mao, LI Gensheng, TIAN Shouceng, et al. Research status and prospect of artificial intelligence in reservoir fracturing stimulation[J]. Drilling & Production Technology, 2022, 45 (4): 1-8. (in Chinese)
[27]	任明仑, 宋月丽, 褚伟. 灰铸铁抗拉强度预测的局部加权线性回归建模[J]. 电子测量与仪器学报, 2019, 33 (3): 65-71. REN Minglun, SONG Yueli, CHU Wei. Locally weighted linear regression modeling for tensile strength prediction of gray cast iron[J]. Journal of Electronic Measurement and Instrumentation, 2019, 33 (3): 65-71. (in Chinese)
[28]	杨鹏史, 丁卉, 陈同, 等. 基于局部加权线性回归的城市公交车排放能耗预测[J]. 中山大学学报(自然科学版), 2019, 58 (6): 111-118. YANG Pengshi, DING Hui, CHEN Tong, et al. Estimation of emissions or electricity consumptions of urban buses based on locally weighted linear regression[J]. Acta Scientiarum Naturalium Universitatis Sunyatseni, 2019, 58 (6): 111-118. (in Chinese)
[29]	吴博, 梁循, 张树森, 等. 图神经网络前沿进展与应用[J]. 计算机学报, 2022, 45 (1): 35-68. WU Bo, LIANG Xun, ZHANG Shusen, et al. Advances and applications in graph neural network[J]. Chinese Journal of Computers, 2022, 45 (1): 35-68. (in Chinese)
[30]	杨希洪, 郑群, 章佳欣, 等. 基于特征插值的深度图对比聚类算法[J]. 计算机科学, 2024, 51 (11): 157-165. YANG Xihong, ZHENG Qun, ZHANG Jiaxin, et al. Feature interpolation based deep graph contrastive clustering algorithm[J]. Computer Science, 2024, 51 (11): 157-165. (in Chinese)
[31]	王岩, 吴晓富. 深度神经网络训练中适用于小批次的归一化算法[J]. 计算机科学, 2019, 46 (11A): 273-276. WANG Yan, WU Xiaofu. Novel normalization algorithm for training of deep neural networks with small batch sizes[J]. Computer Science, 2019, 46 (11A): 273-276. (in Chinese)
[32]	王赞, 闫明, 刘爽, 等. 深度神经网络测试研究综述[J]. 软件学报, 2020, 31 (5): 1255-1275. WANG Zan, YAN Ming, LIU Shuang, et al. Survey on testing of deep neural networks[J]. Journal of Software, 2020, 31 (5): 1255-1275. (in Chinese)
[33]	王钰豪, 郝家胜, 张帆, 等. 钻井溢流风险的自适应LSTM预警方法[J]. 控制理论与应用, 2022, 39 (3): 441-448. WANG Yuhao, HAO Jiasheng, ZHANG Fan, et al. Adaptive LSTM early warning method for kick detection in drilling[J]. Control Theory & Applications, 2022, 39 (3): 441-448. (in Chinese)
[34]	郭立力, 赵春江. 十折交叉检验的支持向量机参数优化算法[J]. 计算机工程与应用, 2009, 45 (8): 55-57. GUO Lili, ZHAO Chunjiang. Optimizing parameters of support vector machine's model based on genetic algorithm[J]. Computer Engineering and Applications, 2009, 45 (8): 55-57. (in Chinese)
[35]	LIU Y H, MAHMASSANI H S. Global maximum likelihood estimation procedure for multinomial probit (MNP) model parameters[J]. Transportation Research (Part B): Methodological, 2000, 34 (5): 419-449.