基于示功图的抽油井动液面软测量机理建模

Soft-sensor Mechanism Modeling for Dynamic Fluid Level of Beam Pumping Systems Based on Dynamometer Card

实践中，抽油井动液面都是使用回声仪测试的，存在测试繁琐、效率低、成本高和无法实时在线检测等缺陷。针对现有基于实测示功图的动液面机理软测量方法存在测量精度不高，而数据驱动软测量模型完全依赖历史数据，但油田现场经常无法提供，并且随时间推移易出现模型老化现象等问题，提出在示功图上动态地确定动液面观测特征点（或区域）和提取作为软测量模型辅助变量的抽油井工况信息的方法，建立了新的动液面软测量机理模型，实现对动液面的实时在线检测。油田现场应用结果表明，该软测量模型具有较高的预测精度和较好的泛化能力。可以满足油田工程应用要求。

In practice, the dynamic fluid level of beam pumping systems is traditionally measured onsite with the echometer. However, this method has defects such as inconvenience, inefficiency, high-cost, instantaneity, and so on. In addition, the present soft-sensor mechanism method for the dynamic fluid level based on the measured dynamometer card has disadvantages on precision. Data-driven soft-sensor modeling completely relies on historical data, but it cannot often be acquired from the oilfield, and there is the phenomenon of the data-driven model aging as time goes by. In view of the above-mentioned facts, on the basis of dynamically determining the feature point （or region） for the dynamic fluid level observation and extracting the working condition information of oil wells as assistant variables by means of the measured dynamometer card, a new soft-sensor mechanism method for the dynamic fluid level is proposed in order to determine the real-time dynamic fluid level. The oilfield application results show that the established soft-sensor mechanism model can achieve higher forecasting accuracy and better generalization ability, meeting the oilfield engineering application requirements.