基于人工神经网络的天然气井产量计算方法研究

Research into calculation of natural gas well production based on an artificial neural network

随着石油行业不断向海洋发展,水下油气生产工艺也随之诞生,传统的技术手段面临诸多新的问题。虚拟计量系统已经逐步在国内外的海上油气田生产系统中开始应用。该技术利用油气田的常规基础工艺参数以及从生产控制系统获取的实时仪表数据,通过多种模型实时计算出单井油气水各相的流量。本文主要研究人工神经网络在虚拟计量方面的应用。由于目前常用的井筒模型不能适应产量的瞬时变化,不能及时准确地预测产量,本文引入具有高度非线性预测能力的误差反向传播的人工神经网络方法,以人工调试后的井筒模型结果作为数据样本库,模拟各种影响因素与天然气井产量之间的映射关系,通过学习和训练建立了基于BP神经网络模型的天然气井产量计算模型。预测结果表明:该方法的计算结果与现场物理流量计测量值的相对误差平均值为3.33%,超过80%的数据点相对误差处于±5%内,预测精度较高。综合分析表明,人工神经网络模型能够满足实际生产需要,且该模型结构简单,不拘泥于具体的形式,计算量少。

With the development of the oil industry to the deep water,underwater oil and gas production process have emerged and the traditional technology is facing many new problems.An alternative method for production estimation is represented by a Virtual Metering System(VMS) based on the analysis of the standard process parameters,available in almost all production system.The software is based on a methodology in which several models are included.This article mainly studies the application of an artificial neural network in gas well measurement.Because the existing wellbore models cannot adjust to changes of production in a timely manner nor predict accurately,this article introduced an error back propagation artificial neural network with highly nonlinear predictive ability.Artificially debugged wellbore model results served as a data sample library to simulate the mapping relationship between all kinds of influence factors and the gas well production.A gas well flow calculation model based on a back propagation neural network is set up by learning and training.Predicted results show that compared with a physical flow meter,the average relative error of the calculation results is 3.33%.More than 80% of the data points have a relative error within plus or minus 5%,which indicates a high prediction accuracy.Comprehensive analysis shows that the artificial neural network model can meet the demands of practical production with the advantages of a simple model structure,flexible form and less calculation.Application of the artificial neural network model provides a new tool and method for virtual measurement technology.