基于支持向量机的火烧油层效果预测

火烧油层是一种效果显著的稠油热力开采方式,油藏是否适合火烧油层工艺开发,需要采用适当的方法对火烧油层的效果进行预测,传统做法是通过数值模拟的方法来实现,但人为因素的影响较大,且耗时较长。将统计理论中的支持向量机方法应用于火烧油层效果预测,以国内外42个火烧油层现场试验为例,选取储集层、流体最具典型性和代表性的14个参数作为输入向量,选取空气油比作为输出向量,构建火烧油层效果预测支持向量机模型,并对模型进行了检验,讨论了核函数选取和算法参数选取对预测结果的影响。实际检验结果表明,支持向量机方法预测空气油比最大平均相对误差仅为6.491%,证实了该方法预测火烧油层驱油效果的可行性。图2表2参13

Effect prediction of in-situ combustion based on Support Vector Machine

In-situ combustion is an effective thermal recovery method.In order to decide whether the in-situ combustion is applicable or not,it is necessary to predict the effect of in-situ combustion with proper methods.It was realized with the method of numerical simulation which is time wasting and influenced by artificial factors.The statistic theory about Support Vector Machine(SVM) is applied to predicting the effect of in situ combustion.From 42 domestic and overseas field pilot studies,14 parameters typical of reservoir bed,fluid and operating are used as input vector and the air-oil ratio is used as output vector.The SVM model is then constructed and tested.Influences of kernel function selection and algorithm parameters choice on the prediction result are also discussed.The result shows that the maximum mean relative error of the method is 6.491%,which confirms the feasibility of this method in predicting the in-situ combustion effect.