基于支持向量机的分层注水效果预测模型设计与实现

Design and Implementation of Stratified Waterflooding Effect Prediction Model Based on Support Vector Machine

将影响分层注水效果的井控储量、渗透率变异系数、连通率、原油粘度、月产油量、含水率、连通油井数、分注的分段数作为输入参数,将评价分层注水效果好坏的无因次增油量作为输出参数,建立了基于支持向量机的分层注水效果预测模型。选用油田实施井例建立了支持向量机的学习样本和检验样本,使用支持向量机的回归训练算法对学习样本进行学习训练,然后对检验样本进行预测运算,结果表明支持向量机方法能够达到较高的预测精度。与油藏数值模拟法和BP神经网络法计算结果进行对比,表明了支持向量机方法的预测精度高于其他两种方法,可以用来预测分层注水的效果,指导油田进行分层注水选井工作。

Factors of stratification injection effects, such as well controlled reserves, permeability variation coefficient, connectivity, oil viscosity, monthly production rate, water content ratio, connecting well number, number of sections for stratification injection are used for input parameters, the dimensionless oil increment for evaluating stratification injection effect is used as output parameter. A model for predicting the effect of stratification injection is established based on support vector machine, learning and inspecting samples are established for the support vector machine by using operational well examples in oilfields. Regression training algorithm of the support vector machine is used for a learning and training of the sample, and the inspecting sample is predicated and operated. The result shows that the method of vector machine has higher prediction accuracy. And it is contrasted with that of reservoir simulation method and BP neural network method, the result indicates that the accuracy of vector machine method is higher than that of those 2 methods. The method can be used to predict the effect of stratification water-injection and guide the well selection for the stratification injection.