基于支持向量机的二氧化碳非混相驱效果预测

目前国内缺乏一种快速、准确预测CO2非混相驱油效果的方法,为了解决这一问题,选取剩余地层压力与混相压力之比、孔隙度、渗透率、油藏中深、地层平均有效厚度、地层温度、原油相对密度、含油饱和度、原油黏度、渗透率变异系数、注采比、注入速度和水气交替注入比等13个地质及工程参数作为输入参数,平均单井日增油量作为输出参数构建了预测CO2非混相驱效果的支持向量机预测模型.以国内6个CO2非混相驱项目和1个CO2混相驱项目为学习样本,2个CO2非混相驱项目和1个CO2混相驱项目为检测样本检测了支持向量机预测模型的准确度,结果表明,3个检测样本的预测值与实际值的平均相对误差为5.57%,满足工程要求.利用该模型预测了腰英台油田CO2非混相驱井组的增产效果,与实际增产效果相比,相对误差仅为1.30%.这表明,采用支持向量机方法对CO2非混相驱油效果进行预测可行且有效. 

Prediction of the Effect CO2 Immiscible Flooding Based on Support Vector Machine

In order to predict the effect of CO2 immiscible flooding rapidly and accurately, a prediction model based on support vector machine was established. It takes 13 geological and engineering parameters (i.e. the ratio of residual formation pressure and CO2 miscibility pressure, porosity, permeability, reservoir mid-depth, net pay, formation temperature, relative density of crude oil, oil saturation, oil viscosity, coefficient of permeability variation, injection-production ratio, injection rate, and the ratio of water/gas alternating injection) as input parameters, and the average daily oil increment per well as output parameter. with six CO2 immiscible flooding projects and 1 CO2 miscible flooding project as training samples, and two CO2 immiscible flooding projects and one CO2 miscible flooding project as testing samples in China, the accuracy of the model was verified. The results showed that average relative error between predicted value and actual value of above 3 samples was 5.57%, which met the engineering requirement. The model was applied to predict the effect of CO2 immiscible flooding in Yaoyingtai Oilfield, indicating a relative error of only 1.30% in relation with the actual value. It suggested that the method based on support vector machine is feasible and effective to predict the effect of CO2 immiscible flooding.