基于特征选择、遗传算法和支持向量机的水淹层识别方法

支持向量机是识别水淹层的有效方法,但其预测性能受多种因素的影响。研究提出一种水淹层识别新方法,采用Relief-F算法进行自动化特征选择,通过遗传算法优化模型参数以及使用加权支持向量机改善样本类数据分布不平衡对分类准确率的影响。将该方法应用于克拉玛依油田六中区克下组砾岩油藏水淹级别划分中,结果表明效果良好,增强了支持向量机的预测能力,进一步提高了水淹层解释的精度。

Festure Sslsction ,Genetic Algorithm and Support Vector Machine

The support vector machine (SVM) was proved effective in identifying the water-flooded oil reservoirs but its prediction performance was sensitive to various factors. A new identification method was presented, where feature subsets were automatically selected by using Relief-F; model parameters were optimized by using genetic algorithms; and the classification accuracy of the small size sample was improved through weighted SVM. The method was applied in the evaluation of water-flooded interval for Lower Karamay reservoir of Liuzhong Area in the Karamay Oilfield. The results show that the SVM model with high generalization performance is obtained and hence the accuracy of the identification of water-flooded interval is effectively improved.