重复压裂选井选层的小波神经网络研究及应用  

STUDY AND APPLICATION OF THE WAVELET NEURAL NETWORK FOR WELL AND LAYER SELECTION IN REFRACTURING

作　　者：屈怀林[1] 郭大立[2] 黄波[1] 熊维莉[1] 

机构地区：[1]西南石油大学研究生院 [2]西南石油大学理学院

出　　处：《石油工业计算机应用》2006年第4期13-15,66,共3页Computer Applications Of Petroleum

摘　　要：重复压裂选井选层需要综合考虑地质特征、油气藏特性、物性参数、测试和生产数据等多种影响因素,难以用传统的数学方法来描述这些参数间的高度非线性映射关系。而以往对重复压裂选井选层的相关研究大都是以BP-神经网络为基础建立模型,但因其算法收敛速度慢、抗干扰能力差等缺点,在应用上受到了一定限制。本文首次将小波神经网络引入重复压裂选井选层,并应用灰色理论进行了参数优选,建立了重复压裂选井选层的小波神经网络模型(该网络模型具有一致逼近和L2逼近能力,以及较强的抗干扰能力),然后在此模型基础上给出了相应的求解算法。最后在现场应用收到了良好的效果,证明了该方法的可靠性与可行性,对重复压裂选井选层有重要的指导意义。Well and layer selection in refracturing requires integrative consideration of multiple influencing factors such as geologic characteristics, reservoir characteristics, characteristical parameters, testing and production data, etc. It is difficult to describe the high nonlinear mapping relationship between these parameters with conventional mathematical technique method.The correlative research of well and layer selection in refracturing in the past is mostly to establish a model based on BP neural network. But it is restricted in the application because its speed in algorithm convergence is slow and the antijamming ability is poor.The wavelet neural network for well and layer selection in refracturing is first introduced in this paper. It uses the grey theory to condut parameter optimization and establish a wavelet neural network model for well and layer selection in refracturing (The network model possess uniform approximation and L2 approximation ability and strong antijamming ability ). Then it gives out a resolution algorithm on the basis of this model.It has gained good effect in on-site application which proved the reliability and viability of it. It is of significance for well and layer selection in refracturing.

关 键 词：重复压裂 选井选层 神经网络 小波网络 模型 

分 类 号：TE357.1[石油与天然气工程—油气田开发工程]