径向基函数神经网络法致密砂岩储层相对渗透率预测与含水率计算

致密砂岩储层具有物性差、孔隙结构复杂、非均质性强等特点，导致利用传统方法难以精确预测或计算其相对渗透率和含水率。为此，文中提出基于径向基函数（RBF）的神经网络预测相对渗透率方法：在介绍RBF神经网络原理的基础上，选择高斯函数和最近邻聚类算法构建网络模型；以含水饱和度、核磁束缚水饱和度、孔隙度、渗透率等四参数为输入，油、水相对渗透率为输出，根据误差分析确定最佳相对渗透率预测网络模型及参数；最后采用分流量方程计算得到储层含水率。将该方法应用于鄂尔多斯盆地陇东地区延长组长8储层，预测的油、水相对渗透率与相渗实验结果一致，计算的含水率与测试结果吻合。

Prediction of relative permeability and calculation of water cut of tight sandstone reservoir based on radial basis function neural network

Tight sandstone reservoir is characterized by poor physical properties,complex pore structures and strong heterogeneity. It is difficult for conventional methods to predict or estimate the relative permeability and water cut. This paper proposes to use the radial basis function (RBF) neural network to predict the relative permeability of tight sandstone reservoir. Based on the RBF neural network,we choose the Gaussian function and the nearest neighbor algorithm to build a network model,and take water saturation,nuclear magnetic irreducible water saturation,porosity and permeability as inputs and relative oil and water permeability as output to define the best relative permeability network model and parameters after error analysis,and finally calculate the water cut using the split flow equation. For the Chang 8 reservoir of the Yanchang Formation in the Longdong area of the Ordos Basin,the relative oil and water permeability predicted by this method is consistent with the experimental results,and the water cut calculated is consistent with the measured value too.