关键词: 深度学习, 核磁共振测井, 渗透率, 深度置信网络 , 深度置信-核极限学习机

Abstract: In view of the problem of low accuracy of the existing NMR logging permeability model in tight sandstone reservoirs, the relationship between nuclear magnetic resonance T2 spectrum and permeability is derived based on the transverse relaxation theory of nuclear magnetic resonance and the Kozeny-Carman equation in this paper. And based on the theoretical analysis, the reasons for the low accuracy of the model are determined. The deep confidence - kernel-extreme learning machine model is proposed by improving the deep learning algorithm, and it is used to predict the reservoir permeability of nuclear magnetic resonance logging with the convolution neural network. The permeability data of 200 rock specimens in the tight gas reservoir in a certain area and the corresponding T2 spectra of NMR logging were modeled and the model was applied to the evaluation of permeability in this area. The results show that the accuracy of the deep learning algorithm is higher than that of the existing NMR logging permeability model and the shallow layer machine learning model. And the accuracy of the deep confidence-kernel-extreme learning machine model proposed by the author is higher than that of the convolution neural network model, which shows that it is more suitable for the prediction of reservoir permeability. The above shows that deep learning theory can be effectively used in oil exploration and development. It can improve the interpretation accuracy of reservoir parameters. This is an inspiration for the interpretation of reservoir parameters.

Key words: deep learning, NMR logging, permeability, deep relief network, deep relief kernel extreme learning machine network