基于循环神经网络的测井曲线重构技术研究

Research on Logging Curve Reconstruction Technology Based on Recurrent Neural Networks

测井是石油天然气勘探开发中的重要组成部分。实际测井过程中,由于仪器损坏等原因,易产生密度、声波等测井曲线严重失真问题,不利于测井资料的解释和处理工作。传统重构缺失数据的方法不仅难以实现而且成本较大,针对传统重构数据方法的问题,采用循环神经网络对真实的测井曲线进行重构。首先,分析出与缺失曲线相关性较高的一条或多条测井曲线,找出测井曲线之间的内在关系。然后,找出与缺失测井曲线相关性较高的曲线进行训练,最后对缺失测井曲线进行重构。结果表明:循环神经网络能有效的对缺失测井曲线实现高精度重构,在实际测井过程中具有一定应用价值。

Well logging is an important part of oil and gas exploration and development.In the actual logging process,due to instrument damage and other reasons,it is easy to produce serious distortion of logging curves such as density and acoustic waves,which is not conducive to the interpretation and processing of logging data.The traditional method of reconstructing the missing data is not only difficult to implement but also costly.To address the problems of the traditional method of reconstructing data,recurrent neural network is used to reconstruct the real logging curve.First,one or more log curves with high correlation with the missing curve are analyzed to find out the intrinsic relationship between the log curves.Then,the curves with higher correlation with the missing log curves are identified for training,and finally the missing log curves are reconstructed.The results show that the recurrent neural network can effectively reconstruct the missing log curves with high accuracy and has some application value in the actual logging process.