原油产能预测中色谱指纹的特征选取方法研究

通过动态监测油井分油层的产能可以掌握油井的生产状况，监测产能在油田生产中具有重要的意义.目前基于色谱指纹的产能预测越来越受到油田相关部门的重视，但是目前色谱指纹的选取还是依靠专业技术人员进行选取，这导致特征选取存在一定的主观性.到目前为止还没有研究者对色谱指纹特征的选择问题进行系统的研究.为了对色谱指纹特征选取问题进行分析，本文分别采用主成分分析(PCA)、线性相关以及随机森林中的重要性分析理论进行特征选取，最后提出了一种新的联合特征选取方法.对来自南海某油田的配比数据进行了测试，实验结果表明，本文提出的方法取得了很好的试验效果，在原油产能预测领域有较大的应用价值.

Feature selection of chromatographic fingerprints for oil production prediction

It is of importance to monitor the production status of oil wells. Nowadays, more and more oilfields use chromatographic fingerprint to monitor Single zone productivity contribution. However how to select chromatographic fingerprint is still a problem, the current selection of chromatographic fingerprint relies on professional experts, which leads to a certain degree of subjectivity. So far, to our knowledge no research was done on the choice of chromatographic fingerprints. In order to analyze chromatographic fingerprint, principal component analysis (PCA), linear correlation method and variable importance in random forest are used in this paper. Then, a joint feature selection method, which combines two methods, is proposed. Experimental results with oil from an oil field of South China Sea show that the proposed method achieves very good results.