论文部分内容阅读
长庆气田马五1 储层非均质严重,储层种类多且各类储层的分布规律及特征信息模糊不清,缺乏先验知识与经验模式的特点.本项研究引进了模糊神经网络方法,利用测井数据建立了长庆气田马五1 储层测井判识的模糊神经网络.计算结果表明,其辨识精度( 回判) 达到了97 .45 ％ ,为利用测井进行储层划分提供了一种方便迅速的工具.经过气田近70 口气井的储层分类验证,取得了较为令人满意的结果,表明此分析技术在特定背景条件下较之单一的模糊模式识别或 B P 神经网络的精度更高,方法更先进,它既可以避免测井储层识别中因环境影响岩性、电性特征模糊而引起的误差,还可以避免因噪音数据影响网络学习限入局部极小的可能,且网络在接受了特征数据后学习速度也得到较大提高.“,”The O 1m 5 1 reservoir in Changqing gas field is of the properties such as the deep heterogeneity,multiple categories,blurred and unclear distribution laws and characteristic information,lacking prior knowledge and experiential mode,etc.Through introducing fuzzy nerve network method,by use of log data a log identified fuzzy nerve network for O 1m 5 1 reservoir in Changqing gas field was set up in this research project.It is shown by calculating that its identification accuracy (back discrimination) is up to 97.45%,a convenient and rapid method being provided for dividing the reservoirs by log data.By means of detecting the reservoir classification in near 70 wells of the gas field,some satisfying results were achieved.It is shown that under specified conditions,the accuracy of this method is higher than that of the individual fuzzy mode identification or BP nerve network;this method is more advanced and it may avoid not only the error of log identified reservoir caused by blurred lithological and logging properties owing to the influence of environment,but also the possibility of the network learning's being immersed in local minimum owing to the influence of noise data and the learning velocity of network has been greatly increased after receiving the characteristic data.