【摘要】 以认识油藏水淹层的水淹情况、指导石油勘探开发为目标,对比常用的多元回归分析方法,引入反向传播神经网络技术。针对油藏水淹层的测井资料,选取感应电导率、声波时差和电阻率作为特征变量,利用聚类分析法,根据水淹层测井数据的亲疏关系进行分类,分类结果作为神经网络结构输出,对测井数据进行训练学习,提高水淹层识别准确率。研究结果表明基于聚类分析的神经网络技术,可以很好地对油层水淹情况进行分析。更多还原

【Abstract】 A technology of back propagation neural network is presented by comparing with the multiple regression analysis.This technology can not only identify the flooding of water-flooding layer but also guide the hydrocarbon exploration and development.In addition,the induction conductivity,acoustic time and resistivity are selected to be characteristic variables;and then well-logging data of water-flooded layer is classified by clustering analysis;the classified result can be as output of neural network structure and the well-logging data can be trained in order to increase identification accuracy.Result shows the neural network technology based on clustering can analyze water flooding of oil layers very well..更多还原