BP神经网络模型在朝长地区扶杨油层压裂产能预测中的应用  

Application of BP neural network model in fracturing productivity prediction of Fuyang oil layer in Chaochang area

作　　者：庄华[1] 潘保芝[1] 张丽华[1] 

机构地区：[1]吉林大学地球探测科学与技术学院,长春130026

出　　处：《世界地质》2012年第4期785-790,共6页World Geology

基　　金：国家科技重大专项(2009ZX05009-001);中央高校基本科研业务费专项资金项目(201103039)联合支持

摘　　要：松辽盆地朝长地区扶杨油层是典型的低孔低渗砂岩储层,需要进行压裂改造,产能影响因素较多,用常规线性方法进行的产能预测结果往往精度不够。在研究该区测井曲线响应特征的基础上,根据地区经验和灰色关联分析法,优选自然伽马、声波、电阻率、中子和密度5个测井特征参数,及含砂比、破裂压力2个压裂施工参数,与已有的试油结论作为模型的训练样本,建立预测储层压裂改造后单位厚度产液量和单位厚度产油量的BP神经网络模型。在实际应用过程中,不仅能较准确地划分油水层,同时给出了产油量的计算参考值,实现了对低孔低渗砂岩储层压裂产能的有效快速预测。Fuyang oil layer in Chaochang area located in Songliao Basin is a typical sandstone reservoir with low porosity and permeability,which often needs to be fractured.Many factors can affect the productivity.The accuracies of productivity prediction getting from conventional linear methods are not enough.Based on the study of the response characteristics of logging curves in this area,combining with the regional experience and grey relational analysis method,five response characteristic parameters are optimized：GR,AC,LLD,CN and DEN.Sand ratio and fracture pressure are also optimized.Combining the chosen data with the test result as the training samples of the model,the established model can be used to predict liquid production in thickness and oil production in thickness of a fractured reservoir.In the progress of practical application,the model not only can divide the oil layers and water layers precisely,but also can give a reference value of the oil production,which could achieve the effective and rapid fracturing productivity prediction of sandstone reservoir with low porosity and permeability

关 键 词：低孔低渗 砂岩储层 压裂产能 BP神经网络 扶杨油层 

分 类 号：TE357.1[石油与天然气工程—油气田开发工程]