Fault Diagnosis of Sucker Rod Pumping System Using Improved Continuous Hidden Markov Model
-
摘要: 利用人工智能诊断有杆抽油系统故障时,描述工况的示功图成为机器学习的对象,提取示功图特征和建立诊断模型是主要步骤。已有的依据阀门工作位置的几何特征未能直接反映示功图的面积,因此提出一组改进的训练特征,并采用连续隐马尔可夫模型(CHMM)建立诊断模型。为了使参数的初始化更可靠,使用与混合高斯模型相关联的K-means聚类算法。将本文提出的诊断方法用于真实油井的示功图集进行测试,结果表明,本文方法不仅有效,而且改进的训练特征和建模方法都提高了诊断的准确率。Abstract: When using artificial intelligence to diagnose the fault of sucker rod pumping system, dynamometer cards describing the working condition become the object of machine learning, and the main steps are to extract the feature of dynamometer card and build the diagnosis model. The existing geometric features based on valve working position cannot directly reflect the area of dynamometer card, so a group of improved training features are proposed, and the continuous hidden Markov model (CHMM) is used to establish the diagnosis model. In order to make the initialization of parameters more reliable, the K-means clustering algorithm associated with the Gaussian mixture model is used. The diagnosis method proposed in this paper is used to test the dynamometer card set of real oil wells, the results show that this method is effective, and the improved training features and modeling methods can enhance the accuracy of fault diagnosis
-
表 1 训练数据集
Table 1. Training data set
编号 工况类型 示功图数量 1 正常 20 2 气体影响 32 3 供液不足 28 4 固定阀漏 16 5 游动阀漏 24 6 油管漏 22 7 油杆断 32 总数 174 表 2 训练示功图几何特征的平均值
Table 2. Average value of geometric features for training dynamometers
工况 几何特征 SABCD SL-up SR-up SR-down SL-down DBC DDA 正常 68.47 8.38 6.75 11.14 5.51 8.27 7.83 气体影响 60.36 7.08 8.58 23.45 4.35 8.29 5.14 供液不足 51.70 5.54 5.88 33.70 8.81 8.72 3.02 固定阀漏 62.41 7.48 8.57 9.52 3.85 9.11 8.08 游动阀漏 63.28 9.75 5.85 5.11 5.32 8.84 9.50 油管漏 54.16 8.97 20.07 7.85 9.52 10.3 10.1 油杆断 28.82 29.5 29.86 3.67 3.59 9.50 9.65 表 3 实测示功图诊断结果
Table 3. Diagnostic results of measured dynamometers
诊断工况
实际工况正常 气体
影响供液
不足固定
阀漏游动
阀漏油管漏 油杆断 正常 5 0 0 0 0 0 0 气体影响 1 6 0 0 0 0 0 供液不足 0 0 6 0 0 0 0 固定阀漏 0 0 0 6 0 0 0 游动阀漏 0 0 0 0 6 0 0 油管漏 0 0 0 0 0 6 0 油杆断 0 0 0 0 0 0 6 表 4 KB-CHMM诊断方法与其他诊断方法的对比
Table 4. Comparison of KB-CHMM diagnostic method with other diagnostic methods
方法 训练的几何特征 曲线矩 文献[13] 本文提出 HMM 64.2% 71.4% 75.2% CHMM 71.4% 82.1% 83.6% CSA-CHMM 82.1% 94.1% 96.2% KB-CHMM 83.9% 96.2% 97.6% -
[1] 萧河. 2020年全球能源需求将出现70年最大降幅[J]. 中国石化, 2020(5): 86.XIAO H. Global energy demand will show the largest decline in 70 years in 2020[J]. Sinopec Monthly, 2020(5): 86. (in Chinese) [2] 贾承造. 中国石油工业上游发展面临的挑战与未来科技攻关方向[J]. 石油学报, 2020, 41(12): 1445-1464.JIA C Z. Development challenges and future scientific and technological researches in China's petroleum industry upstream[J]. Acta Petrolei Sinica, 2020, 41(12): 1445-1464. (in Chinese) [3] 李阳, 廉培庆, 薛兆杰, 等. 大数据及人工智能在油气田开发中的应用现状及展望[J]. 中国石油大学学报(自然科学版), 2020, 44(4): 1-11.LI Y, LIAN P Q, XUE Z J, et al. Application status and prospect of big data and artificial intelligence in oil and gas field development[J]. Journal of China University of Petroleum, 2020, 44(4): 1-11. (in Chinese) [4] 张凯, 赵兴刚, 张黎明, 等. 智能油田开发中的大数据及智能优化理论和方法研究现状及展望[J]. 中国石油大学学报(自然科学版), 2020, 44(4): 28-38.ZHANG K, ZHAO X G, ZHANG L M, et al. Current status and prospect for the research and application of big data and intelligent optimization methods in oilfield development[J]. Journal of China University of Petroleum, 2020, 44(4): 28-38. (in Chinese) [5] 杨剑锋, 杜金虎, 杨勇, 等. 油气行业数字化转型研究与实践[J]. 石油学报, 2021, 42(2): 248-258.YANG J F, DU J H, YANG Y, et al. Research and practice on digital transformation of the oil and gas industry[J]. Acta Petrolei Sinica, 2021, 42(2): 248-258. (in Chinese) [6] 曾涛, 刘晗光, 高坚. 斯伦贝谢公司数字化转型的经验与启示[J]. 国际石油经济, 2021, 29(1): 94-99.ZENG T, LIU H G, GAO J. Experience and inspiration from Schlumberger's digital transition[J]. International Petroleum Economics, 2021, 29(1): 94-99. (in Chinese) [7] 张朝林, 范玉刚. CEEMD与卷积神经网络特征提取的故障诊断方法研究[J]. 机械科学与技术, 2019, 38(2): 178-183.ZHANG Z L, FAN Y G. Fault diagnosis of a bearing using feature extraction method based on CEEMD algorithm and CNN[J]. Mechanical Science and Technology for Aerospace Engineering, 2019, 38(2): 178-183. (in Chinese) [8] HOANG D T, KANG H J. A survey on deep learning based bearing fault diagnosis[J]. Neurocomputing, 2019, 335: 327-335. doi: 10.1016/j.neucom.2018.06.078 [9] LI K, HAN Y, WANG T. A novel prediction method for down-hole working conditions of the beam pumping unit based on 8-directions chain codes and online sequential extreme learning machine[J]. Journal of Petroleum Science and Engineering, 2018, 160: 285-301. doi: 10.1016/j.petrol.2017.10.052 [10] LI K, GAO X W, ZHOU H B, et al. Fault diagnosis for down-hole conditions of sucker rod pumping systems based on the FBH-SC method[J]. Petroleum Science, 2015, 12(1): 135-147. doi: 10.1007/s12182-014-0006-5 [11] 高银中, 雷小强, 杨江天. 基于泵示功图的抽油机井井口产量计算[J]. 石油矿场机械, 2008, 37(5): 88-92.GAO Y Z, LEI X Q, YANG J T. Production evaluation of oil wells using dynamometer card[J]. Oil Field Equipment, 2008, 37(5): 88-92. (in Chinese) [12] LIANG H, LI X M. Accurate extraction of valve opening and closing points based on the physical meaning of surface dynamometer card[J]. Petroleum Exploration and Development, 2011, 38(1): 109-115. doi: 10.1016/S1876-3804(11)60018-9 [13] ZHENG B Y, GAO X W. Sucker rod pumping diagnosis using valve working position and parameter optimal continuous hidden Markov model[J]. Journal of Process Control, 2017, 59: 1-12. doi: 10.1016/j.jprocont.2017.09.007 [14] 周斌, 王延江, 刘伟锋, 等. 基于Hessian正则化支持向量机的多视角协同识别抽油机井工况方法[J]. 石油学报, 2018, 39(12): 1429-1436.ZHOU B, WANG Y J, LIU W F, et al. A working condition recognition method of sucker-rod pumping wells based on Hessian-regularized SVM and multi-view co-training algorithm[J]. Acta Petrolei Sinica, 2018, 39(12): 1429-1436. (in Chinese) [15] XU P, XU S J, YIN H W. Application of self-organizing competitive neural network in fault diagnosis of suck rod pumping system[J]. Journal of Petroleum Science and Engineering, 2007, 58(1-2): 43-48. doi: 10.1016/j.petrol.2006.11.008 [16] YUWONO M, QIN Y, ZHOU J, et al. Automatic bearing fault diagnosis using particle swarm clustering and Hidden Markov Model[J]. Engineering Applications of Artificial Intelligence, 2016, 47: 88-100. doi: 10.1016/j.engappai.2015.03.007 [17] HUANG D R, KE L Y, CHU X Y, et al. Fault diagnosis for the motor drive system of urban transit based on improved Hidden Markov Model[J]. Microelectronics Reliability, 2018, 82: 179-189. doi: 10.1016/j.microrel.2018.01.017 [18] 肖文斌. 基于耦合隐马尔可夫模型的滚动轴承故障诊断与性能退化评估研究[D]. 上海: 上海交通大学, 2011: 30.XIAO W B. Study on coupled hidden Markov model based rolling element bearing fault diagnosis and performance degradation assessment[D]. Shanghai: Shanghai Jiao Tong University, 2011: 30. (in Chinese)