CN-118837953-B - Method, device and storage medium for acquiring offset velocity model

CN118837953BCN 118837953 BCN118837953 BCN 118837953BCN-118837953-B

Abstract

The method comprises the steps of denoising and gather-extracting seismic data to obtain a first common offset gather, inputting the common offset data in the first common offset gather into an initial offset speed model to obtain offset speed output by the offset speed model, homing the first common offset gather according to the offset speed to obtain a second common offset gather, iteratively updating the offset speed model according to the second common offset gather and a target loss function, and obtaining a corresponding target offset speed model after determining that the iteration update times reach a threshold value, thereby achieving the technical effect of improving the acquisition efficiency of the offset speed model.

Inventors

Yuan sanyi
ZHANG HAIFENG
WANG DANYANG
WANG SHANGXU

Assignees

中国石油大学(北京)

Dates

Publication Date: 20260505
Application Date: 20240822

Claims (10)

1. The method for acquiring the offset speed model is characterized by comprising the following steps of: denoising and gather-extracting the seismic data to obtain a first common offset gather, wherein the first common offset gather comprises a plurality of common offset data, and each common offset data corresponds to one offset; Inputting a plurality of pieces of common offset data in the first common offset gather to an initial offset speed model to obtain an offset speed output by the offset speed model, wherein the offset speed model comprises an offset speed solver and a fusion unit, the offset speed solver is used for obtaining speed characteristics corresponding to each piece of common offset data, and the fusion unit is used for fusing a plurality of speed characteristics to obtain an offset speed; Performing homing processing on the first common offset gather according to the offset speed to obtain a second common offset gather, wherein the homing processing comprises prestack time offset and reaction correction, and the second common offset gather is the common offset gather of the first common offset gather after homing processing; and carrying out iterative updating on the offset speed model according to the second common offset distance gather and a target loss function, and acquiring a corresponding target offset speed model after determining that the iterative updating times reach a threshold value, wherein the target loss function is a loss function based on offset speed.
2. The method of claim 1, wherein the offset velocity model further comprises: The system comprises a first common offset gather, a reflection travel unit, a hyperbolic activation function and a second common offset gather, wherein the first common offset gather is used for reflecting the first common offset gather to the ground surface, the second common offset gather is used for reflecting the second common offset gather to the ground surface, and the second common offset gather is used for reflecting the second common offset gather to the ground surface; The position transformation unit is used for carrying out longitudinal position transformation on each piece of common offset data according to each piece of data during reflection travel so as to obtain transformed target common offset data; the offset speed solver is used for obtaining the speed characteristic according to a longitudinal position transformation mode corresponding to the hyperbolic activation function and the target common offset data, and comprises a U-net convolutional neural network.
3. The method of claim 2, wherein the data corresponding to the first set of common offset gathers is obtained by: wherein t (t 0 , x, h) represents the data reflection travel time, h represents the offset distance, x represents the spatial horizontal position, t 0 represents the double travel time corresponding to the zero offset distance, and v represents the speed of the medium above the interface.
4. A method according to claim 3, wherein the transformed target co-offset data is obtained by: Wherein I output (y, x) represents the target co-offset data, I input represents the co-offset data, y and x represent the longitudinal time and the spatial horizontal position of the imaging point, respectively, T represents a travel time matrix calculated based on a travel time physical law, dt represents a time sampling interval, and round represents a rounding function.
5. The method of claim 1, wherein the objective loss function is determined from an unsupervised loss function and an auxiliary data loss function; The unsupervised loss function is an offset velocity based loss function; The auxiliary data loss function is a loss function based on target common offset data; and determining a corresponding target loss function according to the unsupervised loss function and the auxiliary data loss function.
6. The method of claim 5, wherein the unsupervised loss function is obtained by: where L us represents the unsupervised loss function, v represents the network calculated offset speed, Representing the laplace operator, MSE representing root mean square error, x i representing a local window centered on x, And Representing the average of I and J respectively, And Representing the standard deviation corresponding to I and J, cov IJ representing the covariance between I and J, c 1 and c 2 representing constants, M representing the target co-offset data consisting of noff imaging profile stitches, I representing the 1 st through noff-1 st co-offset gathers, J representing the 2 nd through noff co-offset gathers.
7. The method according to claim 6, characterized in that the auxiliary data loss function is obtained by: Wherein L a represents the auxiliary data loss function, A represents the auxiliary data set and consists of noff partial overlapping section transverse stitching.
8. The method of claim 7, wherein the objective loss function is obtained by: Where L represents the target loss function, L us represents the unsupervised loss function, L a represents the auxiliary data loss function, and λ represents the hyper-parameters set to balance the two loss function weights.
9. The device for acquiring the offset speed model is characterized by comprising a memory and a processor; the memory stores computer-executable instructions; the processor executing computer-executable instructions stored in the memory, causing the processor to perform the method of any one of claims 1-8.
10. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-8.

Description

Method, device and storage medium for acquiring offset velocity model Technical Field The present application relates to the field of drilling engineering, and in particular, to a method, an apparatus, and a storage medium for acquiring an offset velocity model. Background With the continuous improvement of the technology level, the demand of people for petroleum geophysical exploration is increasing, in petroleum geophysical exploration, seismic imaging is carried out by correcting underground original seismic data, accurate homing of a petroleum and gas reservoir image is achieved, corresponding reflection interfaces are marked, storage positions of oil, gas and water are assisted to be identified, however, imaging is very sensitive to a speed model, and an inaccurate speed model can cause larger deviation of imaging results of the petroleum and gas reservoir form or drilling tracks. The accurate speed model is the key for acquiring the imaging profile with high fidelity, high signal to noise ratio and high resolution, so that the acquisition of the offset speed model becomes a direction with application prospect; in the prior art, the acquisition of the offset speed model is mainly based on the speed of single-point position pickup of a single gather, and full-wavefield information is forward matched through a full-waveform inversion technology to reconstruct the offset speed model; In the prior art, the full waveform inversion technology not only consumes a large amount of calculation resources, but also highly depends on complex forward modeling calculation, so that the matching difficulty is high, the calculated amount is large, the processing time is long, the requirement of quick acquisition of the offset speed model in practical application is difficult to meet, and the technical problem of low acquisition efficiency of the offset speed model exists. Disclosure of Invention The application provides a method, equipment and a storage medium for acquiring an offset speed model, which are used for achieving the technical effect of improving the acquisition efficiency of the offset speed model. In a first aspect, the present application provides a method for acquiring an offset velocity model, including: denoising and gather-extracting the seismic data to obtain a first common offset gather, wherein the first common offset gather comprises a plurality of common offset data, and each common offset data corresponds to one offset; inputting a plurality of common offset data in a first common offset gather to an initial offset speed model, and obtaining an offset speed output by the offset speed model, wherein the offset speed model comprises an offset speed solver and a fusion unit, the offset speed solver is used for obtaining a speed characteristic corresponding to each common offset data, and the fusion unit is used for fusing the plurality of speed characteristics to obtain an offset speed; performing homing processing on the first common offset distance gather according to the offset speed to obtain a second common offset distance gather, wherein the homing processing comprises prestack time offset and reaction correction, and the second common offset distance gather is a common offset distance gather after the homing processing of the first common offset distance gather; And carrying out iterative updating on the offset speed model according to the second common offset distance gather and the target loss function, and obtaining a corresponding target offset speed model after determining that the iterative updating times reach a threshold value, wherein the target loss function is a loss function based on the offset speed. Optionally, the offset velocity model further comprises: The reflection travel unit is used for acquiring a plurality of corresponding travel time data according to each common offset data in the first common offset road set, wherein the travel time data are used for indicating the time required by the seismic waves under each offset to reflect from the underground interface to the ground surface; determining a plurality of data reflection travel times corresponding to the first common offset gather according to each travel time data, and determining a corresponding hyperbolic activation function according to each data reflection travel time; the position transformation unit is used for carrying out longitudinal position transformation on each piece of common offset data according to each piece of data during reflection travel so as to obtain transformed target common offset data; And the offset speed solver is used for acquiring the speed characteristics according to the longitudinal position transformation mode corresponding to the hyperbolic activation function and the target co-offset data, and comprises a U-net convolutional neural network. Optionally, the data corresponding to the first common offset gather is obtained by reflecting travel by: Where t (t 0, x, h) represents th