论文部分内容阅读
本文用实倒说明了一种侧向测井反演的神经网络方法。文中所用的倒向测井曲线是采用有限差分方法模拟得到的,模拟曲线用作反传播神经网络的输入。由神经网络产生一个初始预测的地层模型,该模型再用作Marquardt反演的输入。神经网络对实际测井曲线中的总体数据特征和细微的数据特征都作出反应,并产生根据训练过程中存储在网络中的知识推断得到的响应。采用合成数据和现场数据对神经网络侧向测井反演进行检验。根据采用现场数据试算得到的最终地层模型模拟出的侧向测井曲线与实际测井数据吻合较好。
This paper illustrates a neural network method for lateral log inversion. The downhole logging curves used in this paper are modeled using the finite difference method and the simulated curves are used as inputs to the anti-propagation neural network. An initially predicted stratigraphic model is generated by the neural network, which is then used as input to the Marquardt inversion. The neural network responds to the overall data features and subtle data features in the actual log and generates inferred responses based on knowledge stored in the network during training. The synthetic logging data and field data are used to test the lateral log inversion of neural network. The lateral log curves simulated from the final formation model based on field data are in good agreement with the actual log data.