迁移深度神经网络的页岩总孔隙度预测  

Prediction for Total Porosity of Shale Based on Transfer Deep Neural Network

作　　者：汪敏 杨桃 唐洪明[2,3] 闫建平[2,3] 廖纪佳[3] WANG Min;YANG Tao;TANG Hongming;YAN Jianping;LIAO Jijia(School of Electrical Engineering and Information,Southwest Petroleum University,Chengdu,Sichuan 610500,China;National Key Laboratory of Oil&Gas Reservoir Geology and Exploitation,Southwest Petroleum University,Chengdu,Sichuan 610500,China;School of Geoscience and Technology,Southwest Petroleum University,Chengdu,Sichuan 610500,China)

机构地区：[1]西南石油大学电气信息学院,四川成都610500 [2]油气藏地质及开发工程全国重点实验室·西南石油大学,四川成都610500 [3]西南石油大学地球科学与技术学院,四川成都610500

出　　处：《西南石油大学学报（自然科学版）》2023年第6期69-79,共11页Journal of Southwest Petroleum University(Science & Technology Edition)

基　　金：国家自然科学基金(62006200);中国石油西南石油大学创新联合体科技合作项目(2020CX020000)。

摘　　要：页岩孔隙度是表征页岩储层孔隙结构的关键指数之一,对孔隙度进行准确预测是开展储层评价的重要一步。页岩孔隙度的准确确定需要以取芯井数据为依据,如何基于极少量取芯井数据,准确预测整口井的孔隙度,是一个有意义的问题。提出了全新迁移深度神经网络模型,基于少量测井和岩芯数据,实现孔隙度的准确预测。首先,根据皮尔逊相关系数法,优选源井深度神经网络的输入测井参数;然后,提出一种新的方法,计算源井与目标井测井数据分布相似性,定量衡量两井间的地质差异;最后,用少量的与源井测井数据分布相似的目标井测井数据再训练源井预测网络,构建预测目标井孔隙度迁移深度神经网络。在A2、B2两口井的测试结果表明:1)该方法只需要10%的数据量,就到达了绝对均值误差为0.0329和决定系数为0.8416的预测性能。2)提出的计算两井相似性的方法可以有效衡量井间差异。源井测井数据与目标井测井数据分布越相似,迁移学习网络的孔隙度预测精度越高。所提出的模型能有效减小对测井和岩芯数据的依赖,极大降低页岩气勘探开发成本。Porosity is one of the key indicators to characterize the por structure of shale reservoirs.Quantitative prediction research on porosity is an important step in reservoir evaluation.The accurate value of shale porosity must be obtained through core analysis.How to obtain an accurate prediction of the porosity of the entire well based on a very small amount of coring well data is a significant problem.This paper proposes a new transfer deep neural network model,based on a small amount of core and logging data,to achieve accurate prediction of porosity.Firstly,according to Pearson correlation coefficient method,the logging parameters suitable for the source well deep neural network are selected as the input for the model.Secondly,a new method is proposed to calculate the similarity of the well logging data distribution between the source well and the target well,quantitatively measure the geological difference between two wells.Thridly,retrain the source well prediction network with a small amount of target well logging data similar to the source well logging data distribution,and build a transfer deep neural network for predicting the porosity migration of the target well.The test results of A2 and B2 show that:1)This method requires only 10% of the data volume,and reaches the performance of an absolute mean error of 0.0329 and a coefficient of determination of 0.8416;2)The proposed method for calculating the similarity of two wells can effectively measure the difference between wells.The more similar the distribution of the source well logging data and the target well logging data,the higher the accuracy of porosity prediction of the transfer learning network.The proposed model can effectively reduce the dependence on logging and core data,and greatly reduce shale gas exploration and development costs.

关 键 词：页岩气 测井 页岩孔隙度预测 井间差异 深度迁移学习 

分 类 号：TE122[石油与天然气工程—油气勘探]