基于稀疏滤波的潜油电泵故障诊断方法

ESP fault diagnosis method based on sparse filtering

韦龙贵 付军 黄新春 张光一 高小永 张誉 丁昆鹏 李启帆 陈硕

WEI Longgui FU Jun HUANG Xinchun ZHANG Guangyi GAO Xiaoyong ZHANG Yu DING Kunpeng LI Qifan CHEN Shuo

中海油能源发展股份有限公司工程技术分公司 中国石油大学(北京)信息科学与工程学院

CNOOC EnerTech-Drilling & Production Co., Tianjin 300452, China College of Information Science and Engineering, China University of Petroleum (Beijing), Beijing 102249, China

潜油电泵作为一种常见的人工举升装置，由于其强提液能力而广泛应用在海上油田，但海洋环境复杂且潜油电泵故障类型繁多与故障数据匮乏等原因使其在海上油田应用存在着一定的局限性。针对难以有效地在线诊断与定量分析潜油电泵电流信号等问题，提出了一种需要极少超参数调节的稀疏滤波特征提取方法，该方法对多种工况电流信号进行了有效的特征提取与模式识别，得到了一个准确、高效的实时诊断模型，通过对现场数据进行分析，验证了该方法的有效性。实验结果表明，该方法可有效地提取特征并实现海上油田潜油电泵10种故障状态的故障诊断，诊断准确率高达99.1%。随着数据的不断丰富和故障种类的不断完善，可实现更高效、准确的故障诊断。

Electrical submersible pump (ESP), as a kind of artificial lift device, is widely applied in offshore oilfields due to its strong liquid lifting capability. However, marine environment is complex, there are a variety of ESP faults, and the fault data are lack, so ESP cannot be effectively applied in offshore oilfields, and its application is limited to some extent. The effective on-line diagnosis and quantitative analysis of ESP’s current signals can be hardly realized. To solve these problems, this paper put forward a sparse filtering characteristics extraction method needing rare hyperparameter tuning, which conducts effective characteristics extraction and mode identification on current signals under various working conditions, so as to obtain an accurate and efficient real-time diagnosis model. In addition, its effectiveness was verified by analyzing the field data. It is experimentally indicated that this method can effectively extract characteristics and diagnose 10 kinds of ESP fault states in offshore oilfields with diagnosis accuracy high up to 99.1%. More efficient and accurate fault diagnosis can be realized as the more and more data is acquired and fault types are clarified better.

10.13639/j.odpt.2023.01.015