Near-channel residual multiple suppression method and system based on deep learning

By performing AVO background trend correction and spectral equalization on seismic data, a deep learning network was constructed to solve the problem of poor suppression of residual multiples in the near path, thereby improving the consistency of seismic data in phase axis and the accuracy of reservoir prediction.

CN122260468APending Publication Date: 2026-06-23CHINA PETROLEUM & CHEMICAL CORP +1
1 Cites 0 Cited by

Patent Information

Application Number
CN202411888799.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-20
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing technologies are insufficient to effectively suppress residual multiples in the near path, affecting the AVO response characteristics of seismic data and the pre-stack reservoir prediction performance.

Method used

By performing AVO background trend correction, spectral equalization, time-varying wavelet inversion, and time-varying wavelet convolution on actual seismic data, a deep learning network was constructed. Seismic data within the water-bearing background range was used as label data for training, thus constructing a deep neural network for suppressing multiple waves.

Benefits of technology

It significantly improved the suppression effect of near-channel residual multiples and enhanced the consistency of seismic phase axes, providing reliable gather data for subsequent AVO analysis and pre-stack reservoir prediction.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a near-channel residual multiple suppression method and system based on deep learning, and belongs to the field of petroleum and natural gas exploration and development. The near-channel residual multiple suppression method based on deep learning comprises steps 1)-19). According to the method, AVO background trend correction, frequency spectrum equalization processing, time-varying wavelet inversion and time-varying wavelet convolution processing are carried out on far-trace data in actual seismic data, a large amount of label data which are consistent with near-trace data of an actual seismic trace set in feature and do not contain multiple signals are obtained, and deep learning training is carried out only in a water-containing background range. According to the method, the influence of oil and gas on the AVO characteristics of the seismic trace gather is avoided, the deep neural network for suppressing the multiples can be constructed more accurately, the suppressing effect of the residual multiples of the near trace can be improved, and reliable trace gather data can be provided for subsequent AVO analysis and pre-stack reservoir prediction.
Need to check novelty before this filing date? Find Prior Art