摘要
在充分探讨BP神经网络基本原理的基础上,提出了用改进的神经网络进行水淹层识别的一种方法。研究中为了解决网络中由于学习率ε的不稳定而导致的网络振荡问题,采用了一种新型的动态学习率方法。实际应用结果表明,该方法的运用进一步提高了水淹层解释精度,通过对南阳油田4口井的试油结果相对比,其符合率达到80%。
With the consideration of fundamental principles in the BP neural network,a method which uses the improved neural network to identify the water- flooded zone is proposed. In order to solve the network's oscillation problem caused by the instability of learning rate- ε,new method that is dynamic learning rate is adopted.The application results shows that this method has improved the water- flooded zone interpretation accuracy.Compared with the four wells’production test results in Nanyang oilfield,it shows that the agreement rate has reached to 80%.
出处
《大庆石油地质与开发》
CAS
CSCD
2004年第3期44-45,共2页
Petroleum Geology & Oilfield Development in Daqing
基金
高等学校优秀青年教师教学科研奖励计划资助。
关键词
神经网络
动态学习率
水淹层
测井解释
BP neural network
dynamic learning rate
log interpretation
water- flooded zone identification
