VSP longitudinal and transverse direct wave event pickup method
Through the DTW mechanism and multi-shot VSP gather preprocessing method, the efficiency and accuracy issues of picking the vertical and horizontal direct wave events in the VSP wave field are solved, and efficient and accurate automatic picking is achieved, reducing manual intervention and costs.
Patent Information
- Application Number
- CN202410342677.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-25
- Publication Date
- 2025-09-26
AI Technical Summary
Existing technologies make it difficult to efficiently and stably pick up the vertical and horizontal direct wave events in the VSP wave field, especially in multi-shot VSP gather data where there is mutual interference between wave field events and coupling noise, resulting in low manual picking efficiency.
A VSP longitudinal and transverse direct wave event picking method based on the DTW mechanism is adopted. Through multi-shot VSP gather preprocessing, frequency-wavenumber domain filtering, extreme point distribution calculation, grid point voting and gradient constraint, the picking results are optimized to avoid axis jumping and deviation.
The efficient and accurate picking of vertical and horizontal direct wave events in multi-shot VSP gather data is achieved, which reduces manual intervention, improves picking accuracy and stability, and saves labor costs.
Smart Images
Figure CN120703838A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil and gas exploration, and in particular to a method for picking up VSP longitudinal and transverse direct arrival wave events. Background Art
[0002] In surface seismic exploration, the first arrival is the seismic wave generated by the source, reaching the geophone after a single reflection. This first arrival contains information such as formation velocity. Accurately picking first arrivals provides the foundation for subsequent processing and interpretation of seismic data. Traditional methods for picking first arrivals include the energy ratio method, which exploits the distinct onset of the first arrival. Subsequent research has proposed STA / LTA-based methods for picking first arrivals, determining their locations by calculating short-term and long-term averages. First arrivals exhibit certain regularity in waveform transformation during propagation. Duan Jianhua uses the fractal dimension method for picking, which does not rely on waveform correlation between adjacent traces and mimics the human brain's approach to seismic first arrival recognition. With the widespread application of neural networks, some researchers have proposed neural network-based first arrival picking methods, which have shown promising results. However, since these methods primarily rely on high first arrival energy and distinct onset, the wavefield events in VSP recordings often intersect, making first arrival picking less effective.
[0003] Dynamic time warping (DTW), a pattern matching algorithm proposed by Sakoe and Chiba in 1978, was initially applied to speech recognition. In geophysics, Anderson et al. applied the DTW algorithm to cross-well log correction in 1983. They analyzed several geophysical applications of the DTW algorithm and demonstrated its effectiveness in waveform classification and feature extraction. Dave applied the DTW algorithm to estimate the relative time (or depth) offset between two seismic images. Lü Xuesong compared the advantages and disadvantages of first-arrival picking and DTW, conducted experiments on picking reflected seismic waves, and used high-order DTW to extract scaled waveform features, demonstrating the algorithm's robustness to noise. The DTW algorithm works well for data with no waveform interference. However, in VSP records, where wavefield events interfere with each other, relying solely on the DTW algorithm for picking events cannot achieve the desired results.
[0004] Vertical Seismic Profiling (VSP) is a seismic exploration technique that excites seismic waves on the ground and receives them through geophones placed in wells. Compared to conventional surface seismic observation systems that receive and acquire data on the surface, VSP utilizes geophones deployed underground to directly measure the propagation characteristics of direct P- and S-waves in the depth domain. Accurately identifying and picking up direct P- and S-waves in the VSP wavefield is a prerequisite for obtaining quantitative multi-wave propagation parameters in the depth domain.
[0005] There are two main traditional methods for picking direct P- and S-wave events in VSP gather data. The first is manual picking, which, while highly accurate, is time-consuming and labor-intensive. The other is picking using ground seismic first arrivals. Currently, picking direct S-wave events in VSP wavefields still relies on manual picking.
[0006] Due to the interference between transmitted and reflected waves in the VSP wavefield, accurate shear wave event detection using layer picking is difficult. Furthermore, interference factors such as coupling noise in the VSP wavefield complicate the process of picking both vertical and horizontal direct wave events. With the advancement of variable-offset VSP exploration technology, hundreds or even thousands of VSP gathers are generated in a single work area. Therefore, research on efficient and stable picking of vertical and horizontal direct wave events is needed to free geologists from the tedious manual wave picking process and save labor costs. Accurately picking vertical and horizontal direct wave events in the VSP wavefield has become a major challenge in current VSP data processing. Summary of the Invention
[0007] The present invention aims to provide a method for picking VSP longitudinal and transverse direct wave events. The method is based on the DTW mechanism and effectively picks VSP longitudinal and transverse direct wave events on multi-shot VSP gather data. This method effectively avoids axis jumping and tracking problems in the process of picking wave field events, thereby ensuring the accuracy of picking effective waves in the VSP wave field.
[0008] The present invention is achieved through the following technical solutions: A method for picking up the vertical and horizontal direct arrival wave events of a VSP comprises the following steps: S1. Preprocess the multi-shot VSP gathers and rearrange the gather order; S2. Preprocess some abnormal traces in the multi-shot VSP gather data to reduce interference factors; S3, using frequency-wavenumber domain filtering to separate the uplink and downlink wave fields; S4. Calculate the distribution of extreme points of the VSP wave field event to be picked; S5, fuse multiple shot data and calculate grid point coordinates as voters; S6. Use adjacent extreme value points as voters to vote; S7. Vote for all extreme points of the current track, count the voting results of the current track, and use the extreme point with the majority of votes as the seed point for picking the next track data; S8. Perform gradient constraints and optimize the picking results.
[0009] Furthermore, in step S1, the preprocessing method is to sort the multi-shot VSP gather data in ascending order according to the distance between the shot point and the wellhead, and reorganize the multi-shot VSP gather data.
[0010] Furthermore, in step S2, the preprocessing method includes two steps: abnormal channel interpolation and gain compensation; Ⅰ. Anomaly trace interpolation: Use adjacent trace interpolation to pre-process anomaly traces at the same depth in multi-shot data. II. Gain compensation: Amplitude gain compensation is used to pre-process the abnormal channels at the same depth of multi-shot data.
[0011] Furthermore, the interpolation methods for abnormal channels are divided into linear interpolation and bilinear interpolation. The linear interpolation calculates the estimated value by the slope of the known data points; The bilinear interpolation is applicable to seismic records that have linear changes in time and space. The bilinear interpolation is based on four nearest neighbor points and estimates the value of the missing trace by weighted average of the four known data points.
[0012] Furthermore, the gain compensation method uses an attenuation function to adjust the seismic records.
[0013] Furthermore, the attenuation function is an exponential attenuation function, and the amplitude of the seismic record adjusted by the attenuation function is recorded as A, which satisfies the following relationship: , Among them, A(t) represents the amplitude of the seismic record after adjustment by the attenuation function, A0 is the amplitude of the original seismic record, t is time, and b is the attenuation coefficient.
[0014] Furthermore, the steps for adjusting the seismic records using the attenuation function are as follows: 1) Perform Fourier transform on the original seismic record to obtain the amplitude spectrum in the frequency domain; 2) Calculate the equalization objective function of the amplitude spectrum, usually selecting a reference amplitude spectrum or a theoretical amplitude spectrum as the target; 3) Calculate the equalization factor based on the objective function and use it to adjust each frequency component; 4) applying the equalization factor to the frequency domain amplitude spectrum of the original seismic record to obtain the equalized amplitude spectrum; 5) Perform inverse Fourier transform on the equalized amplitude spectrum to obtain the gain-compensated seismic record.
[0015] Furthermore, in step S3, the method for separating the uplink and downlink wave fields is: using a two-dimensional Fourier transform on the time-space domain to obtain a frequency-wavenumber domain, performing a filtering operation, and retaining or removing frequency components; Apply two-dimensional inverse Fourier transform to the filtered frequency domain and convert the image from the frequency domain back to the time-space domain of the original VSP gather data to obtain the VSP single-shot gather data that only retains the downlink wave. The signal is then converted from the frequency domain back to the time domain to remove the interference of the upgoing wave field event, and the vertical and horizontal direct wave events are picked up using this data.
[0016] Furthermore, in step S4, the location of the extreme value point is calculated for each data set of a shot data, thereby determining the distribution of points that the current trace may be passed by the wavefield event.
[0017] Furthermore, in step S5, the multi-shot data is fused, the data is extracted using a grid, and the coordinates of the grid points are calculated as voters. The specific operation is as follows: a. Mesh definition: Before generating a mesh, you need to define the mesh's geometric properties, including its boundaries and distribution. You also need to determine the mesh type, which includes uniform, non-uniform, and adaptive. b. Mesh generation: Divide the 3D data volume into discrete grid points according to the selected grid definition; c. Grid point data extraction: When the grid is generated, extract the data value of each grid point.
[0018] Furthermore, in step S6, the specific operation method is as follows: starting from the shallowest detector data of the zero-offset VSP channel gather, the vertical and horizontal direct wave phase axes of the VSP wave field are picked up. When the phase axis is updated to a deeper position, the neighboring extreme points of the extreme point selected in step S4 are used as voters to vote for the extreme point of the current channel number.
[0019] Furthermore, in step S7, after all extreme value points of the current detector have finished voting, the position of the next deep detector event is determined, the seed point is updated, and step S6 is repeated to continue picking up the wavefield event.
[0020] Furthermore, in step S8, by calculating the difference array of the spatial position information of the picked wave field events, the events with abnormal jumps are identified and the picked wave field events with abnormal jumps are removed.
[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects: In the present invention, starting from the multi-shot VSP gather data, the valid event axes on the VSP gather data are picked. After pre-processing the gather data, the extreme points to be voted are first selected using a grid, and then each extreme point is voted using a dynamic time planning algorithm. The more votes an extreme point obtains, the higher the similarity between the extreme point and the surrounding area. Since the extreme points to be voted are output from the previous iterative update, the higher the number of votes, the greater the probability of being located in the current wave field event axis. Finally, the abnormal wave field event axes are detected and picked, and invalid picking results are removed. For the wave field event axis picking of VSP gather data, the gather data of adjacent shots are used as auxiliary picking in this scheme. It is proved through application on actual multi-shot VSP gather data that: 1. In the present invention, multi-shot VSP gather data can be used to pick the P- and T-wave direct events of single-shot gather data, effectively avoiding the instability in the process of picking the P- and T-wave direct events of the VSP wave field and improving the accuracy of the final picking of P- and T-waves.
[0022] 2. In the present invention, after picking the vertical and horizontal direct wave events of the VSP wave field of one shot, it provides effective reference information on the distribution of wave field events for picking other shots, thus realizing the picking of vertical and horizontal direct wave events of multi-shot and multi-dimensional VSP gather data. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is the wavefield diagram before FK domain filtering.
[0024] Figure 2 This is a diagram of the downlink wave field after FK domain filtering.
[0025] Figure 3 This is an illustration of the upgoing wavefield after FK domain filtering.
[0026] Figure 4 is a diagram of the FK domain before filtering.
[0027] Figure 5 It is a diagram of the FK domain of the downlink wavefield.
[0028] Figure 6 It is a diagram of the upgoing wavefield FK domain.
[0029] Figure 7 It is the grayscale image of the VSP wave field gather.
[0030] Figure 8 It is the waveform diagram of the VSP wave field gather.
[0031] Figure 9 This is a schematic diagram of multi-shot VSP gather data.
[0032] Figure 10 It is a schematic diagram of neighborhood extreme point voting.
[0033] Figure 11 It is a display of wave field event tracking results (multiple shots).
[0034] Figure 12 This is a display of the wave field event tracking results (single shot).
[0035] Figure 13 are two discrete signals to be matched.
[0036] Figure 14 It is the matching result of DTW algorithm.
[0037] Figure 15 This is a diagram of the results of picking the longitudinal and transverse coaxial axes of direct-arrival waves from multiple-gun (110-gun) VSP.
[0038] Figure 16 This is a diagram of the results of picking the longitudinal and transverse coaxial axes of direct-arrival waves from multiple guns (117 guns) VSP.
[0039] Figure 17 This is a diagram showing the results of picking up the longitudinal and transverse coaxial axes of direct-arrival waves from multiple guns (119 guns) VSP.
[0040] Figure 18 This is a diagram showing the results of picking up the longitudinal and transverse coaxial axes of the VSP direct-arrival waves from multiple guns (126 guns).
[0041] Figure 19 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0042] The present invention will be further described in detail below with reference to the examples, but the embodiments of the present invention are not limited thereto.
[0043] Example 1 To facilitate public understanding of the present solution, this embodiment takes a method for picking up the longitudinal and transverse direct arrival wave events of a VSP as an example to further illustrate the present solution.
[0044] refer to Figure 19 , the scheme specifically includes the following steps: Step 1: Extract track gathers according to offset distance to prepare for picking up multi-shot data.
[0045] Considering that the multi-shot VSP gather data are actually acquired from the wellhead by exciting the source in two directions along the lateral line, but there is a nonlinear relationship between the number and the offset distance, which affects the simultaneous picking of the vertical and horizontal direct wave phase axes of the multi-shot VSP gather data, this scheme adopts the method of reordering the multi-shot gather data by offset distance from near to far.
[0046] The details are as follows: a) Read the trace header information of a shot data, obtain the 3D coordinate information of the shot point and the 3D coordinate information of the shallowest depth detector, and thus calculate the offset distance d.
[0047] , , .
[0048] b) Rearrange the order of multiple shot data according to the offset distance from small to large.
[0049] Step 2: Preprocess some abnormal traces in the multi-shot VSP gather data to reduce the picking failure caused by noise interference during the picking process.
[0050] The preprocessing method includes the following two steps: I. Abnormal Channel Interpolation Since the detectors are distributed in the well and are easily affected by various noises, the data received by the detector may have outliers, and the channel where the detector is located on the multi-shot data is also an outlier. By using the detector data of the adjacent channel to interpolate the channel, it is possible to effectively avoid picking up wave field event anomalies due to abnormal data noise and fill in abnormal or missing channels in the seismic record. Abnormal channels are usually due to data missing or quality problems caused by various reasons, such as sensor failure, inhomogeneous underground media, etc. The interpolation method is mainly divided into linear interpolation and bilinear interpolation according to the characteristics of the abnormal channel. The details are as follows: a) The process of linear interpolation is as follows: Assume that the known data points are (x0, y0) and (x1, y1), where x0 and x1 are the independent variables of the known data points (such as time or space coordinates), and y0 and y1 are the corresponding dependent variables (such as earthquake record values).
[0051] Assume that the point to be estimated is x, which is between x0 and x1. According to the linear relationship, the estimated value is calculated by the slope of the known data points. In the time domain, the linear interpolation formula is: , In the spatial domain, the linear interpolation formula is: , b) The process of bilinear interpolation is as follows: Bilinear interpolation is suitable for situations where seismic records vary linearly in both time and space. It interpolates data based on the four nearest neighbor points. Assuming the location of the missing trace lies between the locations of the four known data points, bilinear interpolation estimates the value of the missing trace by taking a weighted average of the four known data points.
[0052] The specific steps are as follows: The data point position of the missing channel is expressed as ( x , y ), the position of the known complete data point is ( x 1, y 1), ( x 2, y 2), ( x 3, y3), ( x 4, y 4).
[0053] Calculate the distances between the four nearest neighbors and the missing path location: , Perform a weighted average of the four nearest neighbor points to obtain the estimated value of the missing channel: , , , , in, F ( x, y ) represents the estimated value of the missing channel.
[0054] The bilinear interpolation method utilizes the local linear characteristics in space, which can better estimate the value of the missing channel and maintain the smoothness of the data.
[0055] II. Gain Compensation Amplitude gain compensation is used to preprocess anomaly traces at the same depth in multi-shot data. This method corrects for amplitude attenuation in seismic recordings. Due to absorption and scattering in the subsurface, seismic waves gradually weaken during propagation, causing the amplitude of the recorded seismic signal to decay over time. Gain compensation aims to correct for this amplitude attenuation, resulting in a more balanced energy distribution over time in the seismic recording.
[0056] The specific gain compensation method can use the attenuation function to adjust the seismic record. The common attenuation function is the exponential decay function (EDF), which is as follows: , A(t) represents the amplitude of the seismic record after adjustment by the attenuation function, A0 is the amplitude of the original seismic record, t is the time, and b is the attenuation coefficient.
[0057] Amplitude Spectrum Equalization (ASE). ASE performs frequency domain analysis on seismic records, calculates the amplitude spectrum of the seismic records, and makes adjustments based on the shape of the amplitude spectrum. The specific steps are as follows: 1) Perform Fourier transform on the original seismic record to obtain the amplitude spectrum in the frequency domain; 2) Calculate the equalization objective function of the amplitude spectrum, usually selecting a reference amplitude spectrum or a theoretical amplitude spectrum as the target; 3) Calculate the equalization factor based on the objective function to adjust each frequency component; 4) applying the equalization factor to the frequency domain amplitude spectrum of the original seismic record to obtain the equalized amplitude spectrum; 5) Perform inverse Fourier transform on the equalized amplitude spectrum to obtain the gain-compensated seismic record.
[0058] Step 3: Use frequency-wavenumber domain filtering to separate the uplink and downlink wave fields.
[0059] The time-space domain is transformed using a two-dimensional Fourier transform to obtain the frequency-wavenumber domain (Frequency-Wavenumber Domain). The frequency-wavenumber domain is also called the FK domain. Before performing FK domain filtering, in order to facilitate the simultaneous picking of longitudinal and transverse direct-arrival waves, we directly add the Z component data and the X component data to obtain the magic path. Unlike the first arrival picking and layer picking of ground earthquakes, there are uplink and downlink wavefields in the VSP wavefield. These wavefield events with different trends cross-interfere with each other. During the process of picking the wavefield event, the phenomenon of axis jumping is prone to occur. In the present invention, the FK domain filtering method is first adopted to separate the uplink and downlink wavefields, and then the wavefield where the target wavefield event is located is picked, which reduces the problem of picking difficulties caused by the cross-interference of the wavefield event.
[0060] Based on the concept of Fourier Transform, the signal is converted from the time domain to the frequency domain, filtered, and then the signal is converted back from the frequency domain to the time domain.
[0061] First, we represent the VSP gather data of a shot as DV(t,x). For VSP gather data, the information of the VSP wave field in the time and space domains is stored, where t represents the time domain, indicating the data reception time; x represents the depth domain, indicating the depth of the detector.
[0062] Apply a two-dimensional Fourier transform to the original data to obtain the frequency wavenumber domain representation of the image F ( u , v ),in, u ,v denote the frequency and wave number in the FK domain, respectively.
[0063] , Where j is an imaginary unit, the inner layer integrates over the time domain, and the outer layer integrates over the space domain.
[0064] In the frequency domain, we can perform filtering operations. Filtering can be achieved by multiplication, which is represented by the frequency wave number domain. F ( u,v ) and filter function H ( u,v ) to obtain the filtered frequency domain representation G ( u,v ).
[0065] , in, H ( u,v ) is a filter function that describes the frequency components that you want to keep or remove. In the present invention, H ( u, v ) is defined as follows: , Through the action of the filtering function, only the downlink event wave field remains in the FK domain data. The vertical and horizontal direct wave fields to be picked up in the present invention are all in the filtered data.
[0066] Finally, the filtered frequency domain representation is G ( u,v ) Apply two-dimensional inverse Fourier transform to convert the image from the frequency domain back to the time-space domain of the original VSP gather data, and obtain the VSP single-shot gather data that only retains the downlink wave DV' ( t,x ), since the interference of the upgoing wave field phase axis is removed, the data is used to pick the vertical and horizontal direct wave phase axes.
[0067] Figure 1-3 Shows the time and space domain comparison before and after using FK domain filtering. Figure 1 is the time-space domain before filtering, Figure 2 is the downlink wave field retained after FK domain filtering, Figure 3 It is the upgoing wave field retained after FK domain filtering.
[0068] Figure 4-6 The comparison of FK domain before and after using FK domain filtering is shown. Figure 4 is the FK domain before filtering, Figure 5It is the downlink wave field FK domain retained after FK domain filtering. Figure 6 It is the upgoing wavefield FK domain retained after FK domain filtering.
[0069] Step 4: Calculate the extreme point distribution of the VSP gather data to be picked.
[0070] In the VSP vertical and horizontal direct waves, the wave field phase axis is in the grayscale image (refer to Figure 7 ) are stripes that are easily recognized by the human eye, corresponding to the waveform (reference Figure 8 ) is the extreme point of the waveform, so by picking the extreme point on the wave field phase axis, you can pick out the stripes on the grayscale image. Figure 7-8 As shown, the red dots indicate the wavefield event picking positions.
[0071] Step 5: Fuse multiple shot data and find the extreme points near the grid point coordinates as voters.
[0072] Due to the large amount of multi-shot VSP data, it is impossible to use all extreme points as voters. Therefore, this solution uses a grid to extract the data. The specific operation method is as follows: 1) Grid Definition: Before generating a grid, you need to define its geometric properties, such as the grid boundary and the distribution of grid points. This can be determined based on specific needs, such as uniform grid, non-uniform grid, or adaptive grid. In this invention, a uniform grid is used.
[0073] 2) Mesh Generation: Divide the 3D data volume into discrete grid points according to the selected grid definition. For example, use an axis-aligned orthogonal grid where each grid point has equal spacing along the three coordinate axes. This can be achieved by splitting the entire data volume into voxels of equal size or using more advanced mesh generation algorithms.
[0074] 3) Grid Point Data Extraction: Once the grid is generated, the data value of each grid point can be extracted from the 3D data volume through indexing. This is achieved by mapping the coordinates of the grid point to the corresponding location in the data volume. For example, for a grid point with coordinates (x, y, z), its data value can be obtained by searching the data volume for the location corresponding to those coordinates.
[0075] Step 6: Use adjacent extreme points as voters to vote.
[0076] Since the vertical and horizontal direct wave events are generated from the earthquake source and do not pass through the wave impedance interface to reach the detector, the signal-to-noise ratio of the vertical and horizontal direct wave events in shallow detectors is higher than that in deep detectors. Based on the distribution characteristics of this type of wave field events, the channel number of the shallowest detector is selected as the starting channel i for wave field event picking.
[0077] Manually select the starting track SP i The extreme points on the vertical and horizontal direct wave phase axes are used as seed points SP s,i,m ,in, m is the serial number of the extreme point, s Is the gun number. Figure 9 , Figure 9 The selected seed point is shown. The position information of the next depth of the wave field phase axis picked from the seed point is recorded as SP s,i+1,n ,in s It's the gun number. i+ 1 is the number of the detector where the current extreme point is located, n is the serial number of the extreme point.
[0078] Select multiple grid points closest to the starting seed point in the multi-shot VSP gather data as the voting extreme value points of the starting seed point, denoted as TSP k,j,h ,in, k Indicates the shot number where the VSP gather data is located. j Indicates the detector number where the extreme point is located, h Indicates the sequence number of the extreme point. These selected multiple grid points meet the following conditions: , , , in T 1. T 2. T 3, which represent the extreme point shot difference threshold, extreme point depth difference threshold, and extreme point time difference threshold, respectively.
[0079] refer to Figure 10 , Figure 10 Shows the relationship between voting extreme points and seed points.
[0080] Through these thresholds, it is ensured that the window data and seed points determined by the two sub-wave periods of the selected voting extreme point are SP s,i,mThese grid points are selected from the extreme points of the multi-shot VSP gather data. These selected extreme points are extreme points near the same depth of adjacent gathers, or extreme points near the same depth of adjacent shots. Therefore, the window data determined by the two cycles of the seismic wavelet around these extreme points and the window data determined by the current starting seed point have a certain degree of similarity. The waveform similarity of adjacent gathers is used in combination with the waveform similarity of adjacent shot gathers to vote and find the extreme point in the next trace that matches the current seed point as the picking result of the current trace.
[0081] When measuring waveform similarity, the dynamic time warping algorithm is used. For two discrete data x of arbitrary length, 1_i , y1_j , and its DTW distance is defined as dist ( x i ,y j ), the state transfer equation is as follows: , , , in, DTW ( x,y )express x and y The DTW distance between two sequences, dist ( x i ,y j )express x i and y j The distance between two points. The schematic diagram of the DTW algorithm is as follows Figure 13-14 As shown, Figure 13 The two discrete signals are intercepted from the VSP gather of the common shot point. The horizontal axis is time and the vertical axis is signal amplitude. Figure 14 The matching results of the DTW algorithm for two signals are shown.
[0082] The DTW algorithm dynamically compresses and stretches two time series to optimally match the amplitudes of the two series, so that the seed points can be matched to SP s,i,m If the DTW distance of the extreme points on the determined waveform data is less than the set threshold, it is considered that the matching degree of the two waveforms is high. SP s,i,m It gets the vote from the extreme point. On the contrary, if the waveform similarity is low, the extreme point SP s,i,mUnable to obtain votes, which indicates extreme points SP s,i,m It may be a noise point or an extreme point of other wave field phase axis.
[0083] refer to Figure 11-12 , Figure 11 The results of picking a wave field event on multi-shot VSP gather data are shown. Figure 12 The results of picking a wave field event in the VSP gather data of a shot are shown.
[0084] Step 7: Count the voting results of the current channel.
[0085] The extreme point with the majority vote is used as the seed point for picking up the next data. Through the voting mechanism, the extreme point that is most similar to the current track number i in track number i+1 can be determined. SP s,i+1,n , set the point as the seed point of the current channel number, repeat step 6, and continue to pick the position of the next wave field phase axis.
[0086] Step 8: Perform gradient constraints and optimize the picking results.
[0087] Check the picked events for each wavefield and remove those with large jumps. For the direct wave events of the VSP wavefield, their physical meaning is the time-distance relationship between the wavelet from the source to the detector. To address this characteristic, the picked events must ensure that there are no sudden changes in velocity. This corresponds to the picked temporal and spatial information, which means that the time differences between the sampling points of adjacent traces are within a certain range.
[0088] refer to Figure 15-18 , Figure 15-18 Shown are the wavefields and picking results for shots 110, 117, 119, and 126, respectively. The different numbers here represent wavefields at different offsets in the VSP data. Changes in the acquisition and observation system cause changes in the apparent velocities of the recorded longitudinal and transverse direct-arrival events. By picking longitudinal and transverse direct-arrival events at different offsets, the effectiveness of the algorithm designed in this paper has been verified to a certain extent.
[0089] Since the present invention uses a voting mechanism to select the most appropriate extreme point, one-dimensional difference can be used to find out the jumps in the picking results, thereby discarding these jump abnormal points and using the most accurate picking result as the VSP wave field phase axis picking result.
[0090] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.
[0091] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.
Claims
1. A method for picking up the longitudinal and transverse direct arrival wave events of a VSP, characterized in that: The steps include: S1. Preprocess the multi-shot VSP gathers and rearrange the gather order; S2. Preprocess some abnormal traces in the multi-shot VSP gather data to reduce interference factors; S3, using frequency-wavenumber domain filtering to separate the uplink and downlink wave fields; S4. Calculate the distribution of extreme points of the VSP wave field event to be picked; S5, fuse multiple shot data and calculate grid point coordinates as voters; S6. Use adjacent extreme value points as voters to vote; S7. Vote for all extreme points of the current track, count the voting results of the current track, and use the extreme point with the majority of votes as the seed point for picking the next track data; S8. Perform gradient constraints and optimize the picking results.
2. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 1, characterized in that: In step S1, the preprocessing method is to sort the multi-shot VSP gather data from small to large according to the distance between the shot point and the wellhead, and reorganize the multi-shot VSP gather data.
3. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 1, characterized in that: In step S2, the preprocessing method includes two steps: abnormal channel interpolation and gain compensation; Ⅰ. Anomaly channel interpolation: Use adjacent channel interpolation to pre-process the anomaly channels of the same depth in multi-shot data; II. Gain compensation: Amplitude gain compensation is used to pre-process the abnormal channels at the same depth of multi-shot data.
4. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 3, characterized in that: The interpolation methods for abnormal channels are divided into linear interpolation and bilinear interpolation. The linear interpolation calculates the estimated value by the slope of the known data points. The bilinear interpolation is applicable to seismic records that have linear changes in time and space. The bilinear interpolation is based on four nearest neighbor points and estimates the value of the missing trace by weighted average of the four known data points.
5. The method for picking up the longitudinal and transverse direct arrival wave events of a VSP according to claim 3, characterized in that: The gain compensation method uses an attenuation function to adjust the seismic record.
6. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 5, characterized in that: The attenuation function is an exponential attenuation function. The amplitude of the seismic record adjusted by the attenuation function is recorded as A, which satisfies the following relationship: ， Among them, A(t) represents the amplitude of the seismic record after adjustment by the attenuation function, A0 is the amplitude of the original seismic record, t is time, and b is the attenuation coefficient.
7. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 6, characterized in that: The steps to adjust the earthquake record using the attenuation function are as follows: 1) Perform Fourier transform on the original seismic record to obtain the amplitude spectrum in the frequency domain; 2) Calculate the equalization objective function of the amplitude spectrum, usually selecting a reference amplitude spectrum or a theoretical amplitude spectrum as the target; 3) Calculate the equalization factor based on the objective function and use it to adjust each frequency component; 4) applying the equalization factor to the frequency domain amplitude spectrum of the original seismic record to obtain the equalized amplitude spectrum; 5) Perform inverse Fourier transform on the equalized amplitude spectrum to obtain the gain-compensated seismic record.
8. The method for picking up the longitudinal and transverse direct arrival wave events of a VSP according to claim 1, characterized in that: In step S3, the method for separating the uplink and downlink wave fields is as follows: using a two-dimensional Fourier transform on the time-space domain to obtain a frequency-wavenumber domain, performing a filtering operation, and retaining or removing frequency components; Apply two-dimensional inverse Fourier transform to the filtered frequency domain and convert the image from the frequency domain back to the time-space domain of the original VSP gather data to obtain the VSP single-shot gather data that only retains the downlink wave. The signal is then converted from the frequency domain back to the time domain to remove the interference of the upgoing wave field events, and the data is used to pick up the vertical and horizontal direct wave events.
9. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 1, characterized in that: In step S4, the location of the extreme point is calculated for each data set of a shot data, thereby determining the distribution of points that the current trace may be where the wavefield event passes through.
10. The method for picking up the longitudinal and transverse direct arrival wave events of a VSP according to claim 1, characterized in that: In step S5, multiple shot data are fused, the data is extracted using a grid, and the coordinates of the grid points are calculated as voters. The specific operation is as follows: a. Mesh definition: Before generating a mesh, you need to define the mesh's geometric properties, including its boundaries and distribution. You also need to determine the mesh type, which includes uniform, non-uniform, and adaptive. b. Mesh generation: Divide the 3D data volume into discrete grid points according to the selected grid definition; c. Grid point data extraction: When the grid is generated, extract the data value of each grid point.
11. A VSP longitudinal and transverse direct arrival wave event picking method according to claim 1, characterized in that: In step S6, the specific operation method is as follows: starting from the shallowest detector data of the zero-offset VSP gather, the vertical and horizontal direct wave events of the VSP wave field are picked. When the deeper positions of the events are updated, the neighboring extreme points of the extreme point selected in step S4 are used as voters to vote for the extreme point of the current channel number.
12. A method for picking up the longitudinal and transverse direct arrival wave events of a VSP according to claim 1, characterized in that: In step S7, after all extreme value points of the current detector have finished voting, the position of the next deep detector event is determined, the seed point is updated, and step S6 is repeated to continue picking wavefield events.
13. The method for picking up the longitudinal and transverse direct arrival wave events of a VSP according to claim 1, characterized in that: In step S8, the events with abnormal jumps are identified by calculating the difference array of the spatial position information of the picked wave field events, and the picked wave field events with abnormal jumps are removed.