摘要
渤海湾盆地含油层系多、流体性质复杂且纵向多变,精准快速地识别流体难度较大,而胶囊网络深度学习法可提高储层流体录井参数识别精度。胶囊网络通过卷积结构提取不同储层流体的录井参数特征,利用胶囊网络特有向量表达特征的不变性和共变性,有效挖掘出录井评价参数与储层流体性质之间的内在关系,建立高精度的储层流体快速识别模型。以黄河口凹陷常规砂岩储层为例,开展基于胶囊网络的流体录井识别方法研究。首先,对储层流体信息敏感的6种录井评价参数为样本基础,划分黄河口凹陷381个样本数据的训练集和二级构造带X井区132个样本的测试集;再建立基于录井评价参数的胶囊网络储层流体识别模型。与常规线性Fisher判别分析法相比,基于胶囊网络的储层流体识别方法模型正确率达到87.66%,识别精度提高约10%。实践表明,基于胶囊网络的储层流体识别方法能够有效提取录井参数的储层流体敏感性特征,为储层流体识别录井评价方法提供新思路,具备较好的应用前景与推广价值。
There are many oil-bearing series,complex fluid properties and vertical variability in Bohai Bay Basin,so it is difficult to accurately and quickly identify fluids.The depth learning method based on capsule network can improve the identification effect of reservoir fluid logging parameters.The capsule network extracts the logging parameters of different reservoir fluids through the convolution structure,and expresses the invariance and covariation of the features with the capsule vector.The proposed model can effectively dig out the internal relationship between mud logging parameters and reservoir fluid in the spatial sequence structure characteristics,thereby construct a high-precision fluid identification model.By taking the sand reservoir of Huanghekou sag as an example,the application research of capsule network in fluid identification is carried out.Firstly,the training set of 381 sample data in Huanghekou sag and the test set of 132 samples in well block X of secondary structural belt are divided based on six mud logging parameters that are sensitive to fluid information.Then a capsule network reservoir fluid identification model based on logging evaluation parameters is established.Compared with the conventional linear Fisher discriminant analysis method,the model accuracy of reservoir fluid identification method based on capsule network is 87.66%,and the identification accuracy is improved by about 10%.The practice results show that the reservoir fluid identification method based on capsule network can effectively extract the reservoir fluid sensitivity characteristics of logging parameters,provide a new method for reservoir fluid identification,and has a good application prospect and promotion value.
作者
姬建飞
毛敏
杨毅
袁胜斌
郭明宇
李战奎
JI Jianfei;MAO Min;YANG Yi;YUAN Shengbin;GUO Mingyu;LI Zhankui(China-France Bohai Geoservices Co.,Ltd.,Tianjin 300457,China;Tianjin Company,CNOOC(China)Co.,Ltd.,Tianjin 300459,China;Engineering Technology Company of CNOOC Energy Development Co.,Ltd.,Tianjin 300459,China)
出处
《石油地质与工程》
CAS
2022年第2期66-71,共6页
Petroleum Geology and Engineering
基金
中海石油(中国)有限公司“七年行动计划”重大科技专项“渤海油田上产4000万吨新领域勘探关键技术”(CNOOC-KJ-135-ZDXM36-TJ-08-TJ)。
关键词
流体识别
砂岩储层
录井参数
深度学习
胶囊网络
fluid identification
sand reservoir
mud logging parameter
deep learning
capsule network