基于BP神经网络技术的水淹层评价  

Waterflooded Reservoir Evaluation Based on BP Neural Network Technology

作　　者：郭海敏[1,2] 赵亚宁[1,2] 时新磊[1,2] 彭红浪[1,2] 

机构地区：[1]油气资源与勘探技术教育部重点实验室(长江大学) [2]长江大学地球物理与石油资源学院,湖北荆州434023

出　　处：《石油天然气学报》2010年第5期79-83,共5页Journal of Oil and Gas Technology

基　　金：中国石油"十一五"后三年测井科技项目(2008A-2703)

摘　　要：油层水淹后,油层的电阻率、自然电位、声学性质以及核物理性质等均会发生一系列变化,而且这些变化同油层的物理性质、注入水性质以及注入量等有关。不同的注水时期,这些变化也是不同的,因而使地质情况更加复杂多变。此时如果仅仅依靠常规测井曲线的变化建立模型来评价水淹层,势必造成很大误差。根据常规测井资料,借助BP神经网络,建立了BP网络模型,用建立的模型对某断块的15口具有试油资料的井进行了水淹级别预测,正确率高达80%以上。结果表明,基于BP神经网络的水淹层识别技术具有良好的应用效果。After waterflooding in oil reservoirs,a series of changes,such as resistivity,spontaneous potential,acoustic properties and the nature of nuclear physics were induced,and these changes were related with reservoir physical properties,character of water injection and as well as injection rate and so on.The changes were different in different injection times,it would result in more complex geologic conditions.Therefore if only conventional logging curve was used to establish a model to evaluate the changes in waterflooded reservoirs,serious errors would be caused inevitably.According to conventional logging data,a BP(Back-Propagation)neural network was used to establish a BP network model,waterflooding grade is predicted for 15wells with the oil test data in a block by using the model,its accuracy is 80%.The results show that BP neural network technology has good prospects of application in oilfield.

关 键 词：常规测井曲线 试油资料 测井数据 BP神经网络 水淹层评价 

分 类 号：P631.84[天文地球—地质矿产勘探]