抽油杆泵，也称为游梁泵，是从地下油井开采石油的主要人工举升方法。井下游梁泵组件的维护问题可以通过绘制位移和载荷的测功图来诊断。每天手动收集和分析这些图需要耗费大量的时间和精力。本文展示了机器学习 (ML) 在预测泵状态、为每个测功图提供详细的概率评估、自动化实时分析以及促进泵损坏的早期检测方面的有效性。
数据自动化和收集。通过将现场抽油机生成的 dynacard 数据传输到开放平台通信 (OPC) 服务器，然后将其存储在数据库中，可以实现数据收集和传输的自动化。
Sucker rod pumping, also known as beam pumping, is the major artificial lifting method used for extracting oil from subsurface wells. Maintenance issues in downhole beam-pump components can be diagnosed using dynamometer cards that plot displacement and load. Manually collecting and analyzing these cards daily requires significant effort and time. This paper demonstrates the effectiveness of machine learning (ML) in predicting pump status, offering detailed probabilistic assessments for each dynacard, automating real-time analysis, and facilitating early detection of pump damage.
Data Automation and Collection. The automation of data collection and transfer is facilitated by transferring dynacard data generated by beam pumps in the field to an open-platform-communications (OPC) server, where it is subsequently stored in a database.
