基于最小二乘支持向量机的低孔低渗薄互层油藏水淹层识别方法

大庆西部外围低孔低渗泥质砂岩泥质含量重、储层厚度薄,水淹层测井响应特征不明显,常规人工经验性解释已不能满足实际开发需要.根据油水性质在测井资料表现上的不同,选取有效表征油水性质的独立测井参数,基于最小二乘支持向量机理论,建立分类器进行油水性质识别分析.以工区现有测试资料的层位作为训练样本进行训练,建立不同水淹级别储层的分类器,并验证分类器精度及有效性,从而对油田生产的待识别油水层进行识别分析.通过对工区的样本学习和预测,并与实际试油资料进行对比,符合率达到92.2%,结果表明,最小二乘支持向量机在低孔低渗薄互层水淹层中可获得良好应用.

A Water-flooded Layer Identification Method of Thin Water-flooded Layer with High Mud Based on Least Square Support Vector Machine in Porosity-low Permeability Reservoirs

Logging response characteristics were complex and not obvious in thin water -flooded layer with high mud and low porosity -low permeability reservoirs in Daqing oilfield .Conventional artificial explana-tion can not satisfy the requirements of development .The theory of classification of support vector machine was used to select relative independent logging parameters for recognizing the fluid properties of low porosity -low permeability reservoirs .The tested fluid properties of layer were used as samples for training .The classifiers of different fluid properties of layers and corresponding support vector machine and its classification were estab -lished .By building up the classifier and its function , recognized layers were analyzed .This model were used to predict layer oiliness and compare with tested fluid properties of layers in some Oilfield .The accuracy of identifi-cation is 92 .2%.The result represented that LSSVM can perfectly resolve identification problem of complex wa-ter-flooded layers .