Method for improving matching of forward data and observation data in full waveform inversion

Abstract

The application belongs to the technical field of exploration data processing, and particularly discloses a method for improving the matching of forward data and observation data in full waveform inversion, which changes the method for establishing a velocity model, matches the characteristics of the forward data and the observation data, eliminates the influence of strong-amplitude direct wave interference in the forward data, avoids the situation that the objective function in the full waveform inversion cannot correctly reflect the change of the velocity, and thus makes the velocity model after the full waveform inversion closer to the real underground medium velocity model. The application is suitable for improving the data matching in the full waveform inversion.

Description

Technical Field

[0001] The present invention belongs to the technical field of exploration data processing, and relates to a method for improving data matching in full waveform inversion, in particular to a method for improving the matching between forward modeling data and observation data in full waveform inversion. Background Art

[0002] In recent years, with the continuous development of exploration technology, the demand for imaging accuracy has become increasingly higher. However, due to the lack of shallow reflection information in seismic data and the low signal-to-noise ratio of subsurface reflection waves, establishing high-precision velocity models is relatively difficult. Currently, full waveform inversion (FWI), one of the cutting-edge technologies in velocity modeling, compares the simulated wavefield with the actual seismic data and iteratively matches them to obtain subsurface medium parameters, thereby obtaining velocity updates and establishing a high-precision velocity model. FWI has achieved initial production capabilities in marine data processing, but its application to terrestrial data has not yet achieved a breakthrough.

[0003] The core goal of full-waveform inversion (FWI) technology is to minimize the discrepancy between forward-modeled and observed data. When applied to real-world land data, forward-modeled data often contain strong-amplitude direct waves, while direct waves in observed data decay rapidly. This results in poor matching of early-arrival wave characteristics between the two. This introduces non-velocity factors into the FWI process, affecting its accuracy and potentially leading to significant velocity errors.

[0004] At present, in full waveform inversion of onshore seismic data, in order to solve the problem of feature matching differences between forward data and observation data, traditional processing methods use the excision method or Q absorption attenuation compensation method to avoid the feature differences between the two sets of data and minimize the impact on the accuracy of the velocity model obtained by full waveform inversion.

[0005] The excision method removes direct waves with inconsistent characteristics between the forward modeled and observed data, retaining the consistent features between the two data sets. This difficulty arises from the fact that direct waves, part of the seismic early arrivals, are easily identified at far offsets but difficult to distinguish at close offsets. Furthermore, they overlap with the return wave information, making it difficult to distinguish them effectively, making it difficult to determine the excision parameters. Improper excision also has different consequences. Excessive excision prevents the full utilization of early arrival wave information, while insufficient excision leads to overlap between direct and return wave information, resulting in erroneous inversion results.

[0006] The Q absorption attenuation compensation method also has its drawbacks. For example, it doesn't completely suppress the direct wave. In the forward modeled data after Q absorption attenuation compensation, the direct wave is only partially suppressed, and its characteristics still differ significantly from the observed data, making it impossible to obtain accurate inversion results. Furthermore, Q absorption attenuation compensation is inefficient, requiring multiple iterations for full waveform inversion, which significantly increases computational costs and hinders its industrial application. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for improving the matching of forward data and observation data in full waveform inversion. By changing the method for establishing the initial velocity model, the characteristics of the forward data and the observation data are made consistent, the influence of the direct wave in the forward data is eliminated, and the objective function constructed in the full waveform inversion is avoided from correctly reflecting the change in velocity. The velocity model after full waveform inversion is thus closer to the actual underground medium velocity model, so as to achieve the purpose of applying full waveform inversion to onshore seismic data.

[0008] To achieve the above object, the present invention provides a method for improving the matching between forward modeling data and observation data in full waveform inversion, the method comprising the following steps:

[0009] Using the surface elevation of the seismic exploration observation data, a small-scale smoothing process is performed, and the common center point gather and common shot point data are respectively corrected to the small smoothing surface of the surface elevation to obtain the corrected observation shot gather data.

[0010] An initial velocity model is established using common midpoint gathers, prestack time migration RMS velocity and first arrival information.

[0011] Gaussian beam migration prestack depth migration is performed using the common center gather and initial velocity model to obtain the common reflection angle deviation gather.

[0012] Using the initial velocity model and the common reflection angle deviation gather, the optimized velocity model is obtained through multi-azimuth traveltime tomographic inversion iterative processing, and the negative gradient velocity is filled to establish the initial full waveform inversion velocity model.

[0013] The forward data is obtained through wavelet extraction and forward processing, and frequency iteration is performed using full waveform inversion. According to the convergence of the objective function and the accuracy of the full waveform inversion velocity model obtained by iteration, the frequency iteration is repeated until the full waveform inversion updated velocity model is obtained.

[0014] As a limitation of the present invention, the method comprises the following steps performed in sequence:

[0015] S1. Small-scale smoothing of surface elevation and gather correction processing

[0016] The elevation information of the shot point and the receiver point is smoothed on a small scale to obtain a small smooth surface of the surface elevation;

[0017] Correct the common center point gather to the small smooth surface of the surface elevation to obtain the corrected common center point gather A1, and obtain the pre-stack time migration root mean square velocity field, pre-stack time migration results and interpreted geological horizons;

[0018] Correct the common shot point data to the small smooth surface of the surface elevation to obtain the corrected observation shot gather data B1;

[0019] S2. Initial velocity model establishment

[0020] Using the common center point gather A1 and prestack time migration root mean square velocity, the constrained velocity inversion technique was used to obtain the initial depth domain velocity model. The velocity model C1 was established using the well logging information and geological constraints of the work area.

[0021] Obtain the first arrival information and establish a surface velocity model through tomographic inversion. The surface velocity model is embedded into the velocity model C1 to obtain the initial velocity model C2 embedded in the surface layer.

[0022] S3. Speed ​​model optimization

[0023] Using the common center gather A1 and the initial velocity model C2, Gaussian beam migration prestack depth migration is performed to obtain the common reflection angle deviation gather D1.

[0024] Using the initial velocity model C2 and the common reflection angle deviation gather, the optimized velocity model C3 is obtained through multi-azimuth traveltime tomographic inversion iterative processing.

[0025] S4. Initial full waveform inversion velocity model establishment

[0026] The velocity model C3 is filled with negative gradient velocity using the smooth surface as the interface, and a filling velocity body with decreasing velocity values ​​is established as the initial full waveform inversion velocity model C4;

[0027] S5. Wavelet Extraction

[0028] Determine the initial inversion frequency E1, use the corrected observation shot gather data B1 and the initial velocity model C4, select the near offset distance for wavelet extraction, and obtain the inversion wavelet F1;

[0029] The initial inversion frequency E1 starts from the low frequency of the seismic observation data, preferably 3 Hz or 2 Hz;

[0030] S6. Perform forward modeling

[0031] Using the initial full waveform inversion velocity model C4 and the inversion wavelet F1, forward modeling is performed to obtain forward modeling data;

[0032] S7. Perform full waveform inversion

[0033] The full waveform inversion processing is performed using the initial full waveform inversion velocity model C4, the inversion wavelet F1 and the corrected observation shot gather data B1. The next frequency iteration is determined based on whether the objective function converges in the full waveform inversion:

[0034] If the objective function does not converge, continue the iteration of this frequency to obtain the full waveform inversion update velocity model of this frequency;

[0035] If the objective function converges, the updated full-waveform inversion velocity model C5 of the current frequency is obtained, and it is determined whether the accuracy of the full-waveform inversion velocity model C5 needs to be iterated to a higher frequency:

[0036] If iteration to a higher frequency is required, an inversion frequency E2 with a frequency higher than the current frequency is selected to replace the current frequency, and the initial full-waveform inversion velocity model C4 is replaced by the full-waveform inversion velocity model C5. Steps S5-S7 are repeated until a final full-waveform inversion updated velocity model that does not require iteration is obtained.

[0037] The higher frequency is 1 Hz-2 Hz higher than the current frequency.

[0038] When the first iteration is performed, this frequency refers to the initial inversion frequency E1.

[0039] As a limitation of the present invention, the smooth diameter of the small-scale smooth surface treatment is not greater than 500m;

[0040] After the small-scale smooth surface processing, the number of rapid mutation points of near-surface elevation is zero.

[0041] As a further limitation of the present invention, in step S4, the filling negative gradient velocity is a filling thickness of not less than 1000 m and a gradient value of not less than 1000 m / s. The "negative" in the negative gradient refers to a value calculated from sea level as 0, with positive values ​​for downward and negative values ​​for upward. Therefore, the filling negative gradient velocity is an upward filling velocity.

[0042] As a further limitation of the present invention, in step S5, the near offset distance is no greater than 500 m.

[0043] As a further limitation of the present invention, frequency iteration in full waveform inversion should be performed from low to high, and each frequency should be subjected to wavelet extraction in step S5.

[0044] Due to the adoption of the above solution, the present invention has the following beneficial effects compared with the prior art:

[0045] The application modifies the velocity model by the gradient velocity model filling mode, eliminates the direct wave feature difference existing in the forward data and the observation data, so that the early wave features of the forward simulation data and the actual observation data are matched and consistent, without false image, and solves the matching problem of the features of the forward simulation data and the actual observation data in the land full waveform inversion, so that the influence of the direct wave in the forward simulation data is avoided in the full waveform inversion, and the precision of the full waveform inversion is affected.

[0046] The application is suitable for the land full waveform inversion in the exploration data processing, and is used for improving the matching of the forward data and the observation data in the full waveform inversion. BRIEF DESCRIPTION OF DRAWINGS

[0047] The application will be further described in combination with the drawings and specific embodiments.

[0048] Figure 1 The flow chart of the full waveform inversion in the embodiments of the application is shown in the figure.

[0049] Figure 2 The initial full waveform inversion velocity model C4 model in the embodiments of the application is shown in the figure.

[0050] Figure 3 The final full waveform inversion updated velocity model in the embodiments of the application is shown in the figure. DETAILED DESCRIPTION

[0051] The application will be further described in combination with the embodiments, but those skilled in the art should understand that the application is not limited to the following embodiments, and any improvement and equivalent change made on the basis of the specific embodiments of the application is within the protection scope of the claims of the application.

[0052] Embodiment Method for improving the matching of the forward data and the observation data in the full waveform inversion

[0053] The embodiment is a method for improving the matching of the forward data and the observation data in the full waveform inversion, and the method is used for the land full waveform inversion in the eastern Bohai Bay Basin.

[0054] The eastern Bohai Bay Basin has superior reservoir forming conditions, is a key zone for risk exploration of deep natural gas, has great exploration potential, develops multiple large-scale geological bodies such as the Paleozoic buried hill, the sandstone and gravel fan body of the Shahejie Formation and volcanic intrusive rock, and is a key target for risk exploration, but the imaging effect of the seismic data obtained by the current data processing method is poor, the external morphology and the internal structure of the buried hill are difficult to be determined, the low-frequency information of the internal seismic data of the buried hill is greatly lost, the imaging effect of the top surface morphology and the internal structure of the buried hill is poor, and it is difficult to meet the needs of the fine description of the internal structure of the buried hill, the fine structural interpretation and the fine geological body description.

[0055] Data processing is performed by improving the matching between forward modeling data and observation data in full waveform inversion. The process is as follows: Figure 1 , the specific method is as follows:

[0056] S1. Small-scale smoothing of surface elevation and gather correction processing

[0057] The elevation information of the shot points and receiver points is smoothed on a small scale with a smoothing diameter of 500 m until the number of rapid mutation points in the near-surface elevation is zero.

[0058] Get the small smooth surface of the surface elevation;

[0059] Correct the common center point gather to the small smooth surface of the surface elevation to obtain the corrected common center point gather A1, and obtain the pre-stack time migration root mean square velocity field, pre-stack time migration results and interpreted geological horizons;

[0060] Correct the common shot point data to the small smooth surface of the surface elevation to obtain the corrected observation shot gather data B1;

[0061] S2. Establishment of initial velocity model for prestack depth migration

[0062] Using the common center point gather A1 and prestack time migration root mean square velocity, the constrained velocity inversion technique was used to obtain the initial depth domain velocity model. The velocity model C1 was established using the well logging information and geological constraints of the work area.

[0063] Obtain the first arrival information and establish a surface velocity model through tomographic inversion. The surface velocity model is embedded into the velocity model C1 to obtain the initial velocity model C2 embedded in the surface layer, which is the initial velocity model of prestack depth migration.

[0064] S3. Prestack Depth Migration Velocity Model Optimization

[0065] Using the common center gather A1 and the initial velocity model C2, Gaussian beam migration prestack depth migration is performed to obtain the common reflection angle deviation gather D1.

[0066] Using the initial velocity model C2 and the common reflection angle deviation trace gather, the optimized velocity model C3 is obtained through multi-azimuth traveltime tomographic inversion iterative processing, which is the pre-stack depth migration velocity model.

[0067] S4. Establishment of initial velocity model for full waveform inversion

[0068] The velocity model C3 is divided into two parts, the smooth surface is used as the interface, and the negative gradient velocity is filled 1000m above the interface, where the gradient value is 1000m / s. The filling velocity body C4 with decreasing velocity value is established as the initial full waveform inversion velocity model C4, as shown in the following example: Figure 2 As shown;

[0069] The "negative" in negative gradient means starting from sea level as 0, with positive values ​​downward and negative values ​​upward. Therefore, the filling speed of negative gradient is upward filling.

[0070] S5. Wavelet Extraction

[0071] The low frequency 3 Hz of the seismic observation data is determined as the initial inversion frequency E1, the observed shot gather data B1 and the initial velocity model C4 are corrected, and the near offset distance is selected as 500m to perform wavelet extraction and obtain the inversion wavelet F1;

[0072] S6. Perform forward modeling

[0073] Using the initial full waveform inversion velocity model C4 and the inversion wavelet F1, forward modeling is performed to obtain forward modeling data;

[0074] S7. Perform full waveform inversion

[0075] The full waveform inversion processing is performed using the initial full waveform inversion velocity model C4, the inversion wavelet F1 and the corrected observation shot gather data B1. The next frequency iteration is determined based on whether the objective function converges in the full waveform inversion:

[0076] If the objective function does not converge, continue the iteration at the current frequency, i.e., 3 Hz, to obtain the full waveform inversion update rate model at this frequency;

[0077] If the objective function converges, the updated full-waveform inversion velocity model C5 of the current frequency is obtained, and it is determined whether the accuracy of the full-waveform inversion velocity model C5 needs to be iterated to a higher frequency:

[0078] If iteration to a higher frequency is required, an inversion frequency E2 with a frequency 1 Hz higher than the current frequency is selected, that is, a 4 Hz frequency is used to replace the current frequency. The full waveform inversion velocity model C5 is used to replace the initial full waveform inversion velocity model C4. Steps S5-S7 are repeated until the final full waveform inversion updated velocity model that does not require iteration is obtained, as shown in FIG. Figure 3 shown.

[0079] The working area of ​​this embodiment is the eastern Bohai Bay Basin. The initial full waveform inversion velocity model C4 obtained through steps S2-S4 is as follows: Figure 2 As shown, Figure 3 It is obtained after steps S4-S7. By comparison, it can be seen that Figure 3 The final full waveform inversion updated velocity model is shown to have rich shallow details and the resolution of the velocity model is significantly improved.

[0080] In addition, in another embodiment, the smoothing diameter of the small-scale smooth surface processing is selected to be no greater than 500m, in step S5, the near offset is no greater than 500m, in step S7, the higher frequency is 2Hz higher than the current frequency, and the other steps are basically the same as the above embodiment. Each embodiment achieves a resolution higher than Figure 2 The final full waveform inversion updates the velocity model.

[0081] It should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art may still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for improving the matching of forward modeling data and observation data in full waveform inversion, characterized in that: The method comprises the following steps: Using the surface elevation of the seismic exploration observation data, a small-scale smoothing process is performed, and the common center point gather and common shot point data are respectively corrected to the small smoothing surface of the surface elevation to obtain the corrected observation shot gather data. An initial velocity model is established using common midpoint gathers, prestack time migration RMS velocity and first arrival information. Gaussian beam migration prestack depth migration is performed using the common center gather and initial velocity model to obtain the common reflection angle deviation gather. Using the initial velocity model and the common reflection angle deviation gather, the optimized velocity model is obtained through multi-azimuth traveltime tomographic inversion iterative processing, and the negative gradient velocity is filled to establish the initial full waveform inversion velocity model. The forward data is obtained through wavelet extraction and forward processing, and frequency iteration is performed using full waveform inversion. According to the convergence of the objective function and the accuracy of the full waveform inversion velocity model obtained by iteration, the frequency iteration is repeated until the full waveform inversion updated velocity model is obtained.

2. The method for improving the matching of forward modeling data and observation data in full waveform inversion according to claim 1, characterized in that: The method comprises the following steps performed in sequence: S1. Small-scale smoothing of surface elevation and gather correction processing The elevation information of the shot point and the receiver point is smoothed on a small scale to obtain a small smooth surface of the surface elevation; Correct the common center point gather to the small smooth surface of the surface elevation to obtain the corrected common center point gather A1, and obtain the pre-stack time migration root mean square velocity field, pre-stack time migration results and interpreted geological horizons; Correct the common shot point data to the small smooth surface of the surface elevation to obtain the corrected observation shot gather data B1; S2. Initial velocity model establishment Using the common center point gather A1 and prestack time migration root mean square velocity, the constrained velocity inversion technique was used to obtain the initial depth domain velocity model. The velocity model C1 was established using the well logging information and geological constraints of the work area. Obtain the first arrival information and establish a surface velocity model through tomographic inversion. The surface velocity model is embedded into the velocity model C1 to obtain the initial velocity model C2 embedded in the surface layer. S3. Speed ​​model optimization Using the common center gather A1 and the initial velocity model C2, Gaussian beam migration prestack depth migration is performed to obtain the common reflection angle deviation gather D1. Using the initial velocity model C2 and the common reflection angle deviation gather, the optimized velocity model C3 is obtained through multi-azimuth traveltime tomographic inversion iterative processing. S4. Initial full waveform inversion velocity model establishment The velocity model C3 is filled with negative gradient velocity using the smooth surface as the interface, and a filling velocity body with decreasing velocity values ​​is established as the initial full waveform inversion velocity model C4; S5. Wavelet Extraction Determine the initial inversion frequency E1, use the corrected observation shot gather data B1 and the initial velocity model C4, select the near offset distance for wavelet extraction, and obtain the inversion wavelet F1; S6. Perform forward modeling Using the initial full waveform inversion velocity model C4 and the inversion wavelet F1, forward modeling is performed to obtain forward modeling data; S7. Perform full waveform inversion The full waveform inversion processing is performed using the initial full waveform inversion velocity model C4, the inversion wavelet F1 and the corrected observation shot gather data B1. The next frequency iteration is determined based on whether the objective function converges in the full waveform inversion: If the objective function does not converge, continue the iteration of this frequency to obtain the full waveform inversion update velocity model of this frequency; If the objective function converges, the updated full-waveform inversion velocity model C5 of the current frequency is obtained, and it is determined whether the accuracy of the full-waveform inversion velocity model C5 needs to be iterated to a higher frequency: If iteration to a higher frequency is required, an inversion frequency E2 with a frequency higher than the current frequency is selected to replace the current frequency, and the initial full-waveform inversion velocity model C4 is replaced by the full-waveform inversion velocity model C5. Steps S5-S7 are repeated until a final full-waveform inversion updated velocity model that does not require iteration is obtained.

3. The method for improving the matching of forward modeling data and observation data in full waveform inversion according to claim 2, characterized in that: The smooth diameter of the small-scale smooth surface treatment is not greater than 500m; After the small-scale smooth surface processing, the number of rapid mutation points of near-surface elevation is zero.

4. The method for improving the matching of forward modeling data and observation data in full waveform inversion according to claim 2, characterized in that: In step S4, the filling negative gradient speed is such that the filling thickness is not less than 1000m and the gradient value is not less than 1000m / s.

5. The method for improving the matching of forward modeling data and observation data in full waveform inversion according to claim 2, characterized in that: In step S5, the near offset distance is no greater than 500m.

6. The method for improving the matching between forward modeling data and observation data in full waveform inversion according to claim 2, characterized in that: In full waveform inversion, frequency iteration should be performed from low to high, and each frequency should be subjected to wavelet extraction in step S5.

Images