CN-121995458-A - Surface wave vision speed analysis method, device, equipment and storage medium based on deep learning

CN121995458ACN 121995458 ACN121995458 ACN 121995458ACN-121995458-A

Abstract

The application discloses a surface wave vision speed analysis method, device, equipment and storage medium based on deep learning, which comprises the steps of extracting part of seismic data as a data set, predicting the data to be processed by utilizing the U-net model, converting a prediction result into a t-x domain by utilizing inverse RT conversion, converting the t-x domain prediction result into vision speed information, preprocessing the converted data set, extracting part of the data set as a training set according to the preprocessing result, taking the rest as a verification set, constructing a deep neural network, training a U-net model, converting the data to be processed into the t-v domain by utilizing the RT conversion, predicting the data to be processed, converting the t-x domain prediction result into the vision speed information, and performing surface wave compression on the surface wave vision speed result. The application is well suitable for the surface wave analysis work of the seismic data of different terrains, reduces the manual workload and improves the analysis efficiency and the accuracy of analysis results.

Inventors

WANG JIAQI
WANG ZHAOLEI
GENG WEIFENG
XUE GUIREN
SUN YUNSONG
BIAN CE

Assignees

中国石油天然气集团有限公司
中国石油集团东方地球物理勘探有限责任公司
中油油气勘探软件国家工程研究中心有限公司

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (10)

1. The surface wave vision speed analysis method based on deep learning is characterized by comprising the following steps of: Extracting part of the seismic data as a data set, and extracting the non-extracted seismic data as data to be processed; Analyzing the data set to obtain a corresponding visual speed label; transferring the data set and the corresponding visual speed label to a t-v domain through RT transformation; Preprocessing the converted data set; Extracting part of the data set as a training set and the rest as a verification set according to the preprocessing result; Constructing a deep neural network and training a U-net model; Converting the data to be processed into a t-v domain through RT conversion; predicting data to be processed by using the trained U-net model; Converting the prediction result into a t-x domain through inverse RT transformation; Converting the t-x domain prediction result into view velocity information to obtain a face wave view velocity result; and carrying out surface wave suppression on the surface wave apparent velocity result.
2. The deep learning-based surface wave velocity analysis method according to claim 1, wherein before extracting the seismic data, static correction processing is performed on the seismic data, and extraction is performed according to three keywords of a shot number, a detector line number and a channel number based on the seismic data after the static correction processing, so as to obtain a plurality of data sets.
3. The method for analyzing the surface wave velocity based on the deep learning of claim 2, further comprising the step of extracting additional position information of data in the data set, wherein the additional position information comprises shot coordinates, geophone coordinates, elevation and projection offset information; Constructing a deep neural network, training a U-net model, comprising: And inputting the training set and the additional position information into a deep neural network to train a U-net model.
4. The face velocity analysis method based on deep learning according to claim 3, wherein the projection offset information is classified into left and right parts according to positive and negative, the projection offset is marked as T1, T2, T3,..tn, the projection offset is marked as positive as T '1, T' 2, T '3,..t' n; analyzing the data set to obtain a corresponding speed of sight tag, including: And analyzing the extracted data set through interaction software, wherein the analysis determines the data of the data set according to the gun number and the detector line number, and the envelope of the maximum apparent velocity of the surface wave is respectively picked up for the positive part and the negative part of the projection gun offset, and the envelope is an apparent velocity label of the data.
5. The deep learning based face velocity analysis method of claim 1, wherein the preprocessing is a gain and filter operation on the converted data set.
6. The deep learning based face velocity analysis method of claim 1, wherein the loss function in the training U-net model employs a binary cross entropy loss function: where loss is the loss function, and, For (i, j) a point-of-view speed label, And (3) outputting a training result for the U-net model to be trained for the point (i, j).
7. The deep learning based face velocity analysis method of claim 1, further comprising smoothing the prediction result to remove outliers before converting the prediction result to the t-x domain by inverse RT transformation.
8. A deep learning-based facial wave velocity analysis apparatus, the apparatus comprising: the data acquisition module is used for extracting part of the seismic data, wherein the extracted seismic data are data sets, and the non-extracted seismic data are data to be processed; the data analysis module is used for analyzing the data set to obtain a corresponding visual speed label; The data conversion module is used for the mutual conversion of the t-x domain and the t-v domain and comprises the steps of converting a data set and a corresponding visual speed label into the t-v domain through RT conversion, converting data to be processed into the t-v domain through RT conversion, and converting a prediction result into the t-x domain through reverse RT conversion; the data processing module is used for carrying out gain and filtering operation on the converted data set, carrying out smoothing processing on a prediction result and removing abnormal values; The data training module is used for training the U-net model and predicting the data to be processed by utilizing the trained U-net model; and the surface wave suppression module is used for performing surface wave suppression on the surface wave apparent velocity result.
9. An electronic device comprising a processor, a memory, and a computer program stored on the memory and executable on the processor, the processor implementing the deep learning based face speed analysis method of any one of claims 1 to 7 when the program is executed.
10. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the deep learning-based face-wave velocity analysis method according to any one of claims 1 to 7.

Description

Surface wave vision speed analysis method, device, equipment and storage medium based on deep learning Technical Field The application relates to the technical field of seismic data processing in oil and gas exploration, in particular to a surface wave apparent velocity analysis method, device and equipment based on deep learning and a storage medium. Background In the process of seismic data acquisition, a lot of noise interference can be encountered, noise can be roughly divided into two types of coherent noise and random noise according to a noise propagation mechanism, the coherent noise is characterized by interference expressed in a certain regularity in space and time directions, a surface wave is typical coherent noise, the noise is effectively and accurately removed, and the noise characteristic can be accurately described as an important premise. The method for suppressing the surface waves is mainly to suppress the surface waves by utilizing the difference between the surface waves and effective waves in view speed, frequency and propagation time and adopting the technologies of filtering, various transformation and the like. In practical applications, these methods all need parameters such as maximum apparent velocity and frequency of the surface wave. The accuracy of these parameters is decisive for the denoising effect. Therefore, how to accurately acquire the characteristics of the surface wave becomes an important step for realizing noise suppression. In these noise characteristics, the apparent velocity varies with a range of factors such as subsurface conditions, propagation conditions, etc., and varies greatly at different locations in the same work area. Currently, in the conventional processing, the maximum value of the apparent velocity of the surface wave is usually manually analyzed and determined for a plurality of representative data, and usually, a three-dimensional work area is only processed by using three to four different sets of parameters in a blocking manner, or only one apparent velocity parameter is used for processing all data, so that the processing method is difficult to consider all data, and when the data change is severe, the effect of noise suppression is affected, and meanwhile, effective signals are possibly damaged. However, if the processor is required to perform fine analysis on the data of each gun, the time cost and the labor cost are too great. In order to achieve the purpose of accurately analyzing the surface wave velocity maximum value of each single shot of seismic data by reasonable time and labor cost, an artificial intelligence method is introduced, and a surface wave velocity analysis method based on a deep neural network is achieved. The method aims to provide more accurate surface wave velocity information for the traditional denoising algorithm, and further improve the data processing efficiency and accuracy. Disclosure of Invention In order to solve the problems, the application provides a surface wave view velocity analysis method, a device, equipment and a storage medium based on deep learning, which aim to provide more accurate surface wave view velocity information for a traditional denoising algorithm and further improve data processing efficiency and accuracy. The first aspect of the embodiment of the invention provides a surface wave vision speed analysis method based on deep learning, which comprises the following steps: Extracting part of the seismic data as a data set, and extracting the non-extracted seismic data as data to be processed; Analyzing the data set to obtain a corresponding visual speed label; transferring the data set and the corresponding visual speed label to a t-v domain through RT transformation; Preprocessing the converted data set; Extracting part of the data set as a training set and the rest as a verification set according to the preprocessing result; Constructing a deep neural network and training a U-net model; Converting the data to be processed into a t-v domain through RT conversion; predicting data to be processed by using the trained U-net model; Converting the prediction result into a t-x domain through inverse RT transformation; Converting the t-x domain prediction result into view velocity information to obtain a face wave view velocity result; and carrying out surface wave suppression on the surface wave apparent velocity result. In an alternative embodiment, before extracting the seismic data, static correction processing is performed on the seismic data, and extraction is performed according to three keywords of the gun number, the detector line number and the channel number based on the seismic data after the static correction processing, so as to obtain a plurality of data sets. In an alternative embodiment, the method further comprises the step of extracting additional position information of the data in the data set, wherein the additional position information comprises shot point coordinates, wav