神经网络方法在致密油藏非线性渗流模型中的应用  

Application of neural network method in nonlinear percolation model of tight reservoir

作　　者：丁荷颖 宋付权 朱维耀[2] 汪勇 孙业恒 He-ying Ding;Fu-quan Song;Wei-yao Zhu;Yong Wang;Ye-heng Sun(School of Petrochemical and Energy Engineering,Zhejiang Ocean University,Zhoushan 316022,China;School of Civil and Environmental Engineering,Beijing University of Science and Technology,Beijing 100083,China;Exploration and development scientific research institute of Shengli oil field branch of Sinopec,Dongying 257015,China)

机构地区：[1]浙江海洋大学石油化工与环境学院,舟山316022 [2]北京科技大学土木与环境工程学院,北京100083 [3]中国石油化工股份有限公司胜利油田分公司勘探开发研究院,东营257015

出　　处：《水动力学研究与进展（A辑）》2022年第6期837-845,共9页Chinese Journal of Hydrodynamics

基　　金：国家自然科学基金面上项目(12272350);国家重大专项(2017ZX05072005)

摘　　要：微纳米尺度下,流体的流动规律呈现非线性渗流规律,故需基于微纳米空间中受限流体的非牛顿流体特征,建立液体的非线性渗流模型,模型中的参数需要通过流动实验得到。为了更方便地获取模型参数,预测流体在致密油藏孔隙中的渗流特征,该文基于神经网络预测方法,用现有的实验拟合数据进行机器学习,从而预测出模型参数。并以胜利油田某致密油藏为研究对象,选取77块致密岩芯并测量其物性参数,将渗透率、孔隙半径和稠度系数作为输入参数,建立了预测模型参数的神经网络模型。研究结果表明:使用反向传播神经网络方法,可以准确预测致密油藏中油和水流动时的非线性渗流模型参数,未来可用于预测油藏储层中其他物性特征,具有较高的准确性和广阔的前景。The flow law of fluid at micro-nano scale presents nonlinear seepage law.Based on the characteristics of non-Newtonian fluid of confined fluid in micro-nano space,a nonlinear seepage model of liquid is established.The parameters in the model need to be obtained through flow experiments.In order to obtain model parameters more conveniently and predict the seepage characteristics of fluid in tight reservoir pores,based on neural network prediction method,the existing experimental fitting data are used for machine learning to predict model parameters.Taking a tight reservoir in Sheng Li Oilfield as the research object,77 tight cores are selected and their physical parameters are measured.Taking permeability,pore radius and consistency coefficient as input parameters,a neural network model for predicting model parameters is established.The research shows that the Back Propagation Neural Network(BPNN)method can accurately predict the nonlinear seepage model parameters of oil and water flow in tight reservoirs.It can be used to predict other physical properties in reservoirs in the future,with high accuracy and broad prospects.

关 键 词：致密油藏 神经网络方法 模型参数预测 非线性渗流模型 

分 类 号：TE319[石油与天然气工程—油气田开发工程] TP183[自动化与计算机技术—控制理论与控制工程]