油田多相流虚拟计量技术研究与应用  

Research and application of multi-phase flow virtual measurement technology in oilfield

作　　者：吕宝新 LYU Baoxin(No.5 Oil Production Plant of Daqing Oilfield Co.,Ltd.)

机构地区：[1]大庆油田有限责任公司第五采油厂

出　　处：《石油石化节能与计量》2025年第4期1-5,共5页Energy Conservation and Measurement in Petroleum & Petrochemical Industry

摘　　要：目前多相流量计普遍存在结构复杂、工艺复杂、价格昂贵、维护困难等问题,严重制约了油田生产效率和开发效果的提升。针对上述问题,采用机理模型和机器学习模型相结合的虚拟计量技术替代传统物理计量装置。首先,基于OLGA软件建立多相流控制方程;其次,通过数据分布类型扩充样本数量,利用BP神经网络完成了输入变量和输出变量的训练工作,建立了机理模型和机器学习模型之间的交互逻辑;最后,将训练好的BP神经网络用于现场多相流量的实时计量。结果表明,OLGA软件在沿程摩阻损失和阀门局部压力损失的模拟结果精度较高,相对误差不超过5%;采用160组仿真数据时,模型拟合效果最佳,油相流量、水相流量、气相流量预测结果的相关系数分别达到了0.9235、0.9543、0.9496;预测值与现场测量值的吻合性良好,可以适应不同工况变化下的流量计量。研究结果可为虚拟计量技术在同类油气田上的应用提供实际参考。At present,There are some problems including structure complexity,process complexity,ex-pensive price and difficult in maintenance in multi-phase flowmeters,which seriously restricts the improve-ment of oilfield production efficiency and development effect.To solve the above problems,the virtual mea-surement technology combining mechanism model and machine learning model is used to replace the traditional physical measurement device.Firstly,the multi-phase flow control equation is established based on OLGA software.Secondly,by expanding the number of samples through data distribution types,the BP neural net-work is used to complete the training of input variables and output variables,and the interaction logic between the mechanism model and the machine learning model is established.Finally,the trained BP neural network is applied to the real-time measurement of multi-phase flow in the field.The results show that the relative error of OLGA software is less than 5%in the simulation results of friction loss and local pressure loss.When 160 sets of simulation data are used,the model fitting effect is the best.The correlation coefficients of oil phase flow,water phase flow and gas phase flow prediction results reached 0.9235,0.9543 and 0.9496 respectively.The predicted value is in good agreement with the measured value and can be adapted to the flow measurement un-der different working conditions.The research results can provide practical reference for the application of virtu-al measurement technology in similar oil and gas fields.

关 键 词：多相流 机器学习 机理模型 虚拟计量 仿真数据 

分 类 号：TE863[石油与天然气工程—油气储运工程]