Seismic data reconstruction method based on generative adversarial network

Abstract

The invention belongs to the technical field of oilfield seismic big data reconstruction, in particular to a seismic data reconstruction method based on a generative adversarial network. The seismic data reconstruction method comprises the following steps: seismic slice data reduced to the uniform size are used as a training set; the training set is trained by using a deep convolution generative adversarial network, and the Wasserstein distance is used as the training evaluation index of aseismic data generation model; and the seismic data generation model is applied to reconstruct the seismicdata, and the back propagation algorithm and the optimization algorithm based on the standard gradient are applied to optimize the gradient of the objective function so that the difference between the reconstructed data and the missing data is minimized. The method has the beneficial effects that the problem that the traditional seismic data reconstruction algorithm needs to meet the limitation of Nyquist sampling theorem is solved; the problem that it is difficult to select the sparse base for seismic data reconstruction by using the compression sensing algorithm is solved; and the problem that the compression sensing algorithm and the traditional reconstruction algorithm have poor reconstruction effect under the condition of extremely low sampling rate is solved.

Description

technical field [0001] The invention belongs to the technical field of oilfield seismic big data reconstruction, and in particular relates to a seismic data reconstruction method based on a generative countermeasure network. Background technique [0002] Data reconstruction is an important part of data processing. In the signal field, the signal data collected due to factors such as environment, equipment, and human factors may not be complete. If incomplete data is used for data interpretation and analysis, there will be large deviations in the analysis results, so the data needs to be reconstructed before data interpretation and analysis. In addition, for seismic exploration, a large amount of data acquisition work, a large amount of data will generate huge costs in various links such as acquisition, storage and transportation. Therefore, on the one hand, it is hoped to reduce the collected data as much as possible, and on the other hand, it is hoped that the reconstruct...

AI Technical Summary

Problems solved by technology

This type of method usually treats irregularly sampled data as regular data, and interpolates through Gaussian windows, which is easy to introduce errors

The second type of method is based on the wave equation, that is, iteratively solves an inverse problem through DMO or AMO forward and inversion operators. This type of method uses the physical properties of wave propagation to reconstruct the seismic wave field, but requires prior information about the underground structure. And the amount of calculation is very large

The recently developed Shearlet transform is more sensitive to directionality. Compared with the Curvelet transform, it can represent seismic signals more sparsely, making seismic data reconstruction based on compressed sensing better, but there are still problems that cannot be adaptively selected according to the data to be processed. The problem

Images