基于RNN-FCNN的多尺度油水层识别方法

Multiscale oil-water layer recognition method based on RNN-FCNN

对油藏开采领域而言,快速、准确地识别油水层有利于节省大量的人力物力,提高不可再生资源的开采率。现有的层位识别方法未能考虑测井数据的序列关系,并且对所有层位进行统一识别,导致物性相近的层位识别易混淆,识别效果存在局限性。针对测井数据的特点,本文提出基于循环神经网络(RNN)和全连接神经网络(FCNN)的多尺度油水层识别方法。该方法首先基于RNN建立粗粒度的识别模型,再通过串联FCNN的方式实现更细粒度的层位识别,不仅考虑了测井数据在空间上的关联性,同时以多尺度方法识别易混淆的层位。解决了测井数据特征提取困难、层位识别率低的问题。本文在真实测井数据上进行了实验验证,实验结果表明本文方法油水层识别效果良好,有较强的实用性。

For the field of reservoir exploitation,the rapid and accurate identification of oil-water layer is conductive to saving a lot of manpower and material resources and improving the exploitation rate of non-renewable resources.The existing methods fail to consider the sequence relationship of logging data and identify all layers uniformly,which leads to the confusion of the identification of layers with similar physical properties,so the recognition effect has limitation.In view of the characteristics of logging data,this work proposes a multiscale oil-water layer identification method based on recurrent neural network(RNN)and fully connected neural network(FCNN).The method firstly establishes a coarse-grained recognition model based on RNN,and then realizes finer-grained layer recognition by means of serial FCNN.It not only considers the spatial correlation of logging data,but also identifies confusing layers by multiscale method.This method solves the problem that the feature extraction of logging data is difficult and the recognition rate of the layer is low.In this work,the experimental verification is carried out on real logging data.The experimental results show that the method has good identification effect of oil-water layer and strong practicability.