基于改进U-Net卷积神经网络的储层预测

传统的U-Net卷积神经网络大多存在深层网络梯度消失的问题。本文在U-Net卷积神经网络中加入残差模块，提出了一种改进U-Net卷积神经网络。残差模块保证了U-Net卷积神经网络在误差反向传播过程中梯度的存在，在一定程度上可以缓解梯度消失的问题。最后将改进U-Net卷积神经网络应用于实际储层预测中，实际数据测试结果表明基于改进U-Net卷积神经网络在岩性识别以及“甜点”预测上均能取得较好的效果。

Reservoir Prediction Based on Improved U-Net Convolutional Neural Network

Most of the traditional U-Net convolutional neural networks have the problem that the gradient of the deep network disappears. In this paper, a residual module is added to the U-Net convolutional neural network, and an improved U-Net convolutional neural network is proposed. The residual module guarantees the existence of the gradient of the U-Net convolutional neural network in the process of error back-propagation, which can alleviate the problem of gradient disappearance to a certain extent. Finally, the improved U-Net convolutional neural network is applied to the actual reservoir prediction. The actual data measurement shows that the improved U-Net convolutional neural network can achieve better results in lithology identification and "Sweet Point" prediction.