当前位置：首页 > 过刊浏览->2026年第52卷第1期
地面驱动螺杆泵油井产液计量模型及应用
Liquid production measurement model of surface driven screw pump wells and its application
浏览(474) 下载(3)
- DOI:
- 10.3969/j.issn.1001-2206.2026.01.009
- 作者:
- 王天宇, 何思琦, 鲍玲, 刘书军, 管桐, 李恬颍
WANG Tianyu, HE Siqi, BAO Ling, LIU Shujun, GUAN Tong, LI Tianying
- 作者单位:
- 1. 中国石油华北油田分公司油气工艺研究院, 河北任丘 062550;2. 中国石油华北油田分公司勘探开发研究院, 河北任丘 062550;3. 中国石油华北油田分公司, 河北任丘 062550
1. Oil & Gas Technology Research Institute of PetroChina Huabei Oilfield Company, Renqiu 062550, China;2. Exploration and Development Research Institute of PetroChina HuaBei Oilfield Company, Renqiu 062550, China;3. PetroChina HuaBei Oilfield Company, RenQiu 062550, China
- 关键词:
- 螺杆泵井; 计量模型; 电参量; 大数据; 远程预测
screw pump well;measurement model;electrical parameter;big data;remote prediction
- 摘要:
随着数字油田建设的深入推进，传统螺杆泵油井计量方式（如罐车计量、标定车计量）在实时性、连续性和经济性方面的局限日益凸显。提出一种“物理模型+数据驱动”的双轨计量体系，创新性地将电参量模型与XGBoost机器学习模型进行加权融合。通过建立单位转速-单位沉没压力标准化体系，解决传统电参法在变工况下的适应性难题；采用特征工程优化和动态参数调优策略，使XGBoost模型在样本量有限的条件下仍保持94.7%的预测准确率。现场试验表明：双模型协同计量可使系统误差从单一模型的8%~20%降至5%以内，尤其对于低产液井（产液量<5 t/d）仍能控制在7%以内的误差水平。计量模型已在辛集、深州油田的150口螺杆泵井实现工业化应用，累计减少计量车巡检里程2.7×104 km，年节约运营成本超400万元。研究成果可为常规油气开发中的低成本数字化计量提供技术参考。
With the in-depth advancement of digital oilfield construction, the limitations of traditional screw pump well measurement methods (such as tank truck measurement and calibration vehicle measurement) have become increasingly prominent in terms of realtime performance, continuity and economy. A“physical model + data-driven”dual-track measurement system was proposed to innovatively integrate the electrical parameter model with the XGBoost machine learning model through weighted fusion. By establishing a standardized system of unit rotational speed-submerged pressure, the adaptability problem of the traditional electric parameter method under variable working conditions was solved. The strategies of feature engineering optimization and dynamic parameter tuning were adopted to allow the XGBoost model to maintain a prediction accuracy rate of 94.7% under the condition of limited sample size. Field tests show that the collaborative measurement of the dual models can reduce the systematic error from 8%~12% to 5% or less of the single model, especially when the wells with low liquid production can control its error level within 7%. The model has been industrially applied in 150 screw pump wells in Xinji and Shenzhou Oilfields, cumulatively reducing the inspection mileage of metering vehicles by 27 000 kilometers and saving over 4 million yuan in operating costs per year. A reusable technical paradigm is provided for low-cost digital measurement in conventional oil and gas development.