BP神经网络算法在单井流动单元识别中的应用  

Identify Flow Unit Of Single Well Through Bp Neural Networks

作　　者：王治国[1] 宋考平[1] 张春鹤[2] 刘刚 王爱明[1] 

机构地区：[1]东北石油大学石油工程学院,黑龙江大庆163318 [2]大庆钻探工程公司钻井四公司,吉林松原138000 [3]大庆油田采油八厂,黑龙江大庆163000

出　　处：《数学的实践与认识》2011年第5期97-101,共5页Mathematics in Practice and Theory

基　　金：国家自然科学基金重点项目"低渗透油层提高提高驱油效率的机理研究(50634022)"

摘　　要：流动单元的准确划分不仅是建立精细地质、搞清剩余油分布的必要条件,而且关系到后期挖潜剩余油措施的选择.本文应用BP神经网络算法,以取芯井参数聚类分析结果为学习样本,对非取芯井进行流动单元划分,实例表明与常规的判别分析方法相比,在划分结果上基本一致,但是BP神经网络算法具有更准确的样本学习能力,不存在回判错误的可能,为单井流动单元划分提出了新的思路.Accurate Classification of flow unit is not only the base of building fine geologic model and learning distribution of remainder oil saturation, it also concerns choosing digging measure of remainder oil. This paper uses Learning Algorithm of BP Neural Networks to classify non- core well, using result of clustering analysis of core well data as Learning Patterns. The actual example shows that results of BP Neural Networks and Discriminatory Analysis are similar. But the BP neural networks have more exact ability of sample learning. Error of verified sample does not exist. It provides a new method for flow unit classification of single well.

关 键 词：BP神经网络 流动单元 聚类分析 判别分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]