Method and device for constructing seismic de-noising data set
By constructing noisy and effective signal regions in actual seismic data, calculating relative and absolute energy, and establishing a seismic denoising sample dataset, the problem of denoising in complex seismic data is solved, achieving high-precision denoising effect and improving the denoising performance of deep learning models.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA NAT PETROLEUM CORP
- Filing Date
- 2024-11-06
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies struggle to construct high-precision denoised datasets from complex seismic data, limiting the industrial application of deep learning-based seismic denoising techniques, especially when dealing with complex seismic noise.
By automatically constructing noisy regions and effective signal regions in actual seismic data, calculating relative and absolute energy, and establishing a seismic denoising sample dataset, a deep learning model is trained using this dataset.
It significantly improves the accuracy of earthquake denoising, enhances the denoising effect of deep learning models, and can effectively suppress complex earthquake noise and restore effective signals.
Smart Images
Figure CN121995491A_ABST
Abstract
Description
Technical Field
[0001] This article relates to the field of oil and gas geophysical exploration technology, and in particular to a method and apparatus for constructing seismic denoising datasets. Background Technology
[0002] Given the numerous challenges faced by domestic oil and gas exploration, including complex surface conditions, intricate geological conditions, and concealed oil and gas reservoirs, recovering effective seismic data from complex raw seismic data has become a crucial foundation for overcoming these difficulties. High-energy seismic background noise severely interferes with the ability of seismic data to represent subsurface geological information, and can even create false impressions, misleading oil and gas exploration. Seismic background noise exists at all times during seismic acquisition, distributed before and after the arrival of the first arrival wave. Its sources are diverse, including various non-source-related noises such as wind rustling and animal activity; and source-related noises such as interference from adjacent shots and secondary sources. Therefore, it is characterized by variable energy intensity, wide frequency band distribution, coexistence of coherent and random characteristics, and complex probabilistic features (non-Gaussian, non-stationary), making suppression very difficult. Furthermore, with the increasing prevalence of "wide-bandwidth, high-resolution" acquisition technologies, the volume of seismic data is growing exponentially, and the efficient processing of massive amounts of seismic data presents new challenges to denoising techniques. Therefore, how to efficiently and accurately suppress complex seismic noise and recover effective signals has become one of the key technologies urgently needing to be solved in the field of oil and gas exploration.
[0003] In the field of seismic noise suppression, the use of deep learning to suppress complex seismic noise is constantly evolving. However, deep learning is a data-driven algorithm, and the completeness and realism of the training dataset largely determine its denoising effect. Seismic data, however, is vibration signal acquired by instruments in complex environments, making it difficult to construct a sample label dataset containing clean signals.
[0004] Therefore, how to construct a seismic denoising dataset using actual data is an urgent problem to be solved. Summary of the Invention
[0005] This application provides a method and apparatus for constructing an earthquake denoising dataset. The method constructs a high-precision dataset of noisy data-noise data using actual data. The denoising network model trained using this dataset significantly improves the denoising accuracy compared with the denoising network model trained using simulated data.
[0006] In a first aspect, this application provides a method for constructing an earthquake denoising dataset, the method comprising:
[0007] Obtain actual seismic data from the seismic work area to be processed;
[0008] Select at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data, and identify the noise region;
[0009] The relative energy and absolute energy are calculated based on the actual earthquake data, and the effective signal sample area is determined based on the relative energy and the absolute energy.
[0010] An earthquake denoising sample dataset is established based on the noise region and the effective signal sample region.
[0011] Optionally,
[0012] The step of selecting at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data and identifying the noise region includes:
[0013] Select at least two seismic traces from the actual seismic data, and obtain point information of at least two points from the seismic traces along the first and last points of the seismic data.
[0014] Calculate the offset between two data points based on the seismic record time and trace number of the two points;
[0015] Calculate the self-excitation and self-recovery time and formation velocity based on the offset between two data points;
[0016] The theoretical first arrival time of all traces in the actual seismic data is calculated based on the self-excitation and self-reception time and the formation velocity.
[0017] Optionally,
[0018] The point information includes: earthquake record time and trace number.
[0019] Optionally,
[0020] The calculation of the offset between two data points based on the seismic recording time and trace number of the two points includes:
[0021] The coordinates of the excitation point and the receiver point of each channel are determined based on the channel number corresponding to each data point.
[0022] Based on the coordinates of the channel excitation point and the channel receiver point of the two data points, the corresponding offset is calculated using the offset formula.
[0023] The offset formula is as follows:
[0024]
[0025] Where x is the offset distance between the two points, (sx,sy) are the coordinates of the channel excitation point, and (rx,ry) are the coordinates of the channel receiver point.
[0026] Optionally,
[0027] The calculation of self-excitation and self-recovery time and formation velocity based on the offset between two data points includes:
[0028] Obtain the relationship between offset and time;
[0029] Based on the offset between two data points and the trace number data, the self-excitation and self-recovery time and formation velocity are determined using the formula relating the offset and time.
[0030] The formula relating the offset and time is:
[0031]
[0032] Where x represents the offset of the data point, ti represents the seismic recording time, the trace number data are (x1,ti1) and (x2,ti2), t0 is the self-excitation and self-reception time, v is the formation velocity, x1 is the offset of the first data point, x2 is the offset of the second data point, ti1 is the seismic recording time of the first data point, and ti2 is the seismic recording time of the second data point.
[0033] Optionally,
[0034] After calculating the self-excitation and self-recovery time and formation velocity based on the offset between two data points, the method further includes:
[0035] Determine whether the theoretical arrival line of all seismic traces is above the actual arrival time of the earthquake based on the theoretical arrival time.
[0036] If the theoretical first arrival line is below the actual earthquake first arrival time, then two seismic traces should be selected to recalculate the theoretical first arrival line;
[0037] If the theoretical first arrival line of all seismic traces is above the actual first arrival time of the earthquake, then the data before that theoretical first arrival line is determined to be a noise region.
[0038] Optionally,
[0039] The step of calculating relative and absolute energy based on the actual seismic data, and determining the effective signal sample area based on the relative and absolute energy, includes:
[0040] Calculate the energy before the first arrival and the energy of the entire seismic trace based on the actual earthquake data.
[0041] The relative and absolute energies are calculated based on the energy prior to arrival and the energy of the entire seismic trace.
[0042] Based on the relative energy or the absolute energy, and using a pre-set relative energy threshold or absolute energy threshold, determine whether each seismic trace is a valid signal seismic trace;
[0043] After performing the above judgment steps on all seismic trace data, if the effective signal seismic traces are continuous and the number exceeds the minimum regional threshold, then the current region is determined to be an effective signal sample region.
[0044] Optionally,
[0045] The calculation of the energy before the first arrival and the energy of the entire seismic trace based on the actual seismic data includes:
[0046] The energy before arrival e fb :
[0047] e fb =max{|d i |,i=0,1,…,T′};
[0048] Among them, e fb The energy before the first arrival is T, and T′ is the time of the first arrival line in this theory;
[0049] The energy e of the entire seismic channel all :
[0050] e all =max{|d i |,i=0,1,…,T}
[0051] Among them, e all The energy of the entire seismic trace is given by T, where T is the maximum recording time of that trace, || represents the absolute value, and d is the energy of the entire seismic trace. i This represents the value of the seismic sampling point at time i.
[0052] Optionally,
[0053] The relative energy e r for:
[0054]
[0055] In the above formula, e r e is relative energy fb For the energy before arrival, e all The energy of the entire earthquake channel;
[0056] The absolute energy e a for:
[0057] e a =e fb
[0058] In the above formula, ea As absolute energy, e fb The energy before the arrival of the first arrival.
[0059] Optionally,
[0060] The step of determining whether each seismic trace is a valid signal seismic trace based on the relative energy or the absolute energy, using a pre-set relative energy threshold or absolute energy threshold, includes:
[0061] The judgment is made based on the relative energy or absolute energy, combined with the judgment rules of the pre-set relative energy threshold or absolute energy threshold;
[0062] If the relative energy corresponding to the seismic trace is less than the relative energy threshold, or the absolute energy corresponding to the seismic trace is less than the absolute energy threshold, then the seismic trace is determined to be a valid signal seismic trace.
[0063] The judgment rule is as follows:
[0064]
[0065] Where f is the indicator of a valid signal seismic trace, 1 indicates a valid signal seismic trace, and 0 indicates a non-valid signal seismic trace, v r As the relative energy threshold, v a This is the absolute energy threshold.
[0066] Secondly, embodiments of the present invention also provide an apparatus for constructing an earthquake denoising dataset, characterized in that the apparatus includes: a memory and a processor; the memory is used to store a program for constructing an earthquake denoising dataset, and the processor is used to read and execute the program for constructing an earthquake denoising dataset, and execute the method described in any one of the above embodiments.
[0067] Thirdly, embodiments of the present invention also provide a computer-readable storage medium storing a data processing program, wherein the data processing program is executed by a processor using the method for constructing an earthquake denoising dataset as described in any of the above embodiments.
[0068] Compared with related technologies, this application provides a method and apparatus for constructing an earthquake denoising dataset. The method includes: acquiring actual earthquake data from the seismic work area to be processed; selecting at least two seismic traces from the actual earthquake data to determine the theoretical first arrival line of the actual earthquake data and identifying the noise region; calculating the relative energy and absolute energy based on the actual earthquake data, and determining the effective signal sample region based on the relative energy and absolute energy; and establishing an earthquake denoising sample dataset based on the noise region and the effective signal sample region. This application automatically constructs a high-precision dataset of noisy data-noise data using actual data. The denoising network model trained using this dataset significantly improves the denoising accuracy compared to the denoising network model trained using simulated data.
[0069] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. Other advantages of this application can be realized and obtained by means of the solutions described in the description and the accompanying drawings. Attached Figure Description
[0070] The accompanying drawings are used to provide an understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.
[0071] Figure 1 This is a flowchart illustrating the method for constructing an earthquake denoising dataset according to an embodiment of this application;
[0072] Figure 2 This is a schematic diagram of an apparatus for constructing an earthquake denoising dataset according to an embodiment of this application;
[0073] Figure 3 This is a schematic diagram illustrating the method of picking the theoretical first solstice line in some exemplary embodiments;
[0074] Figure 4 This is a schematic diagram of the theoretical first-arrival line check in some exemplary embodiments;
[0075] Figure 5 This is a schematic diagram of the standard theoretical first solstice line in some exemplary embodiments;
[0076] Figure 6 This is a schematic diagram of a binary input denoising network in some exemplary embodiments;
[0077] Figure 7 This is a schematic diagram of raw seismic data in some exemplary embodiments;
[0078] Figure 8 This is a schematic diagram of data after background noise suppression using a constructed seismic denoising dataset in some exemplary embodiments;
[0079] Figure 9 This is a schematic diagram illustrating background noise used in some exemplary embodiments to suppress background noise by constructing an earthquake denoising dataset. Detailed Implementation
[0080] This application describes several embodiments, but these descriptions are exemplary and not restrictive, and it will be apparent to those skilled in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with, or may replace, any feature or element of any other embodiment.
[0081] This application includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features, and elements disclosed in this application may also be combined with any conventional features or elements to form a unique inventive scheme as defined by the claims. Any feature or element of any embodiment may also be combined with features or elements from other inventive schemes to form another unique inventive scheme as defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this application may be implemented individually or in any suitable combination. Therefore, the embodiments are not limited except by the limitations imposed by the appended claims and their equivalents. Furthermore, various modifications and changes may be made within the scope of the appended claims.
[0082] Furthermore, in describing representative embodiments, the specification may have presented methods and / or processes as a specific sequence of steps. However, the method or process should not be limited to the specific order of steps described herein, to the extent that it does not depend on such a specific order. As will be understood by those skilled in the art, other sequences of steps are also possible. Therefore, the specific order of steps set forth in the specification should not be construed as a limitation of the claims. Moreover, the claims concerning the method and / or process should not be limited to the steps performed in the written order, and those skilled in the art will readily understand that these orders can be varied and still remain within the spirit and scope of the embodiments of this application.
[0083] To address the problem of seismic noise suppression, the general approach is to find a unique feature of the signal that the noise lacks (and vice versa) and use this feature for noise suppression. These features typically require a mathematically constructed model for description; therefore, conventional noise suppression techniques are also known as model-driven methods. Based on the data space, denoising techniques can be broadly categorized into time-domain and transform-domain methods. Time-space domain noise attenuation methods include stacking, median filtering, diffusion filtering, and structure-guided filtering. Transform-domain methods transform the seismic record from the time-space domain to other domains for signal-to-noise separation, including KL transform, fx-domain predictive filtering, Cazor filtering, wavelet transform, curvelet transform, empirical mode decomposition, singular value decomposition, and local time-frequency decomposition.
[0084] Deep learning, a rapidly developing machine learning algorithm in recent years, has demonstrated its powerful potential for nonlinear representation and deep data mining through revolutionary successes in fields such as image classification and speech recognition. In seismic exploration, deep learning technology has been rapidly applied in areas such as seismic fault identification, first arrival picking, and attribute recognition. Its application in seismic noise suppression is also gradually expanding. As a novel technology, deep learning exhibits strong potential in seismic exploration and possesses broad development prospects.
[0085] Deep learning is a data-driven algorithm, and the completeness and realism of the training dataset largely determine its denoising effect. However, earthquake data is vibration signal acquired by instruments in complex environments, making it difficult to construct a clean sample label dataset. The challenge of constructing earthquake data label datasets severely restricts the industrial application of deep learning-based earthquake denoising technology. Currently, solutions to the dataset construction challenge are basically divided into three categories: the first is to use simulated data; the second is to use data before and after conventional processing as the training dataset, with deep learning technology acting as an alternative mapping to improve processing efficiency; and the third is to use unsupervised (or weakly supervised) deep learning techniques to solve (or reduce) the dependence on the dataset. However, these methods are only suitable for random noise, and it is difficult to suppress coherent noise using self-supervised deep learning techniques.
[0086] To address the challenges of suppressing earthquake background noise and constructing realistic earthquake noise training sets, this invention proposes a method for constructing a background noise dataset. Through ingenious process design, a high-precision dataset of noisy data and simulated noise data is automatically constructed using only real-world data. The denoising network model trained using this dataset significantly improves denoising accuracy compared to a denoising network model trained using simulated data.
[0087] This invention provides a method for constructing an earthquake denoising dataset, such as... Figure 1 As shown, the method includes steps S100-S130:
[0088] S100: Obtain actual seismic data in the seismic work area to be processed;
[0089] S110: Select at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data, and identify the noise region;
[0090] S120: Calculate the relative energy and absolute energy based on the actual earthquake data, and determine the effective signal sample area based on the relative energy and absolute energy;
[0091] S130: Establish an earthquake denoising sample dataset based on the noise region and the effective signal sample region.
[0092] In one exemplary embodiment, selecting at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data and identifying noise regions includes:
[0093] Step 1: Select at least two seismic traces from the actual seismic data, and obtain point information for at least two points along the first and last arrival points of the seismic data from the seismic traces;
[0094] The point information includes: earthquake record time and trace number.
[0095] Step 2: Calculate the offset between the two data points based on the seismic record time and trace number of the two points;
[0096] Step 3: Calculate the self-excitation and self-recovery time t0 and the formation velocity v based on the offset between the two data points;
[0097] Step 4: Calculate the theoretical first arrival time of all traces in the actual seismic data based on time t0 and velocity v;
[0098] Step 5: Determine whether the theoretical arrival lines of all seismic traces are above the actual earthquake arrival time based on the theoretical arrival time.
[0099] Step 6: If the theoretical first arrival line is below the actual earthquake first arrival time, then select two seismic traces and repeat steps 1-5 to recalculate the theoretical first arrival line.
[0100] Step 7: If the theoretical first arrival line of all seismic traces is above the actual first arrival time of the earthquake, then the data before the theoretical first arrival line is determined as a noise region.
[0101] In one exemplary embodiment, calculating the offset between two data points based on the seismic recording time and trace number of the two points includes:
[0102] Step 1: Determine the coordinates of the excitation point and the receiver point of each channel based on the channel number corresponding to each data point;
[0103] Step 2: Calculate the corresponding offset distance using the offset distance formula based on the coordinates of the excitation point and the receiver point of the two data points.
[0104] The offset formula is as follows:
[0105]
[0106] Where x is the offset distance between the two points, (sx,sy) are the coordinates of the channel excitation point, and (rx,ry) are the coordinates of the channel receiver point.
[0107] In one exemplary embodiment, calculating the self-excitation and self-recovery time t0 and the formation velocity v based on the offset between two data points includes:
[0108] Step 1: Obtain the relationship between offset and time;
[0109] Step 2: Based on the offset between the two data points and the trace numbers (x1, ti1) and (x2, ti2), determine the self-excitation and self-recovery time t0 and the formation velocity v using the formula relating offset and time.
[0110] The formula relating the offset and time is:
[0111]
[0112] Where x represents the offset of the data point, ti represents the seismic recording time, (x1,ti1) and (x2,ti2), t0 is the self-excitation and self-reception time, the formation velocity v, x1 is the offset of the first data point, x2 is the offset of the second data point, ti1 is the seismic recording time of the first data point, and ti2 is the seismic recording time of the second data point.
[0113] In one exemplary embodiment, calculating relative energy and absolute energy based on the actual seismic data, and determining the effective signal sample area based on the relative energy and absolute energy, includes:
[0114] Step 1: Calculate the energy e before the first arrival based on the actual earthquake data. fb and the energy e of the entire earthquake channel all ;
[0115] In this step, the energy e before the initial arrival... fb :
[0116] e fb =max{|d i|,i=0,1,…,T′};
[0117] Among them, e fb The energy before the first arrival is T, and T′ is the time of the first arrival line in this theory;
[0118] The energy e of the entire seismic channel all :
[0119] e all =max{|d i |,i=0,1,…,T}
[0120] Where T′ is the time of the theoretical arrival line;
[0121] `max{}` represents the maximum value of the sequence, `| |` represents the absolute value, and `d`... i The value of the seismic sampling point at time i is represented, T′ is the time of the theoretical first arrival line of the trace, and T is the maximum recording time of the trace.
[0122] Step 2: Based on the energy e before the initial arrival... fb and the energy e of the entire earthquake channel all Calculate the relative energy e r and absolute energy e a ;
[0123] In this step, the relative energy e r for:
[0124]
[0125] In the above formula, e r e is relative energy fb For the energy before arrival, e all The energy of the entire earthquake channel;
[0126] The absolute energy e a for:
[0127] e a =e fb
[0128] In the above formula, e a As absolute energy, e fb The energy before the arrival of the first arrival.
[0129] Step 3: Based on the relative energy e r Or absolute energy e a Using a pre-set relative energy threshold v r Or absolute energy threshold v a Determine whether each seismic trace is a valid signal seismic trace;
[0130] The specific judgment process in this step is as follows:
[0131] According to the relative energy e r Or absolute energy e a Using a pre-set relative energy threshold v r Or absolute energy threshold v a Make a judgment based on the judgment rules;
[0132] If the relative energy corresponding to the seismic trace is less than the relative energy threshold, or the absolute energy is less than the absolute energy threshold, then the seismic trace is determined to be a valid signal seismic trace.
[0133] The judgment rule is as follows:
[0134]
[0135] Where f is a marker for valid signal seismic traces, 1 indicates a valid signal seismic trace, and 0 indicates a non-valid signal seismic trace; v r As the relative energy threshold, v a This is the absolute energy threshold. Generally, the relative energy threshold v... r The default value is 0.1, and the absolute energy threshold v a The default value is 0.001.
[0136] Step 4: After performing the above judgment steps for all seismic trace data, if the effective signal seismic traces are continuous and the number exceeds the minimum regional threshold, then the current region is determined to be an effective signal sample region.
[0137] The background noise dataset construction method implemented in this embodiment has the following technical effects:
[0138] A high-precision dataset of noisy data and noisy data was automatically constructed using only real-world data.
[0139] The denoising network model trained using this dataset significantly improves denoising accuracy compared to the denoising network model trained using simulated data.
[0140] Secondly, embodiments of the present invention also provide an apparatus for constructing an earthquake denoising dataset, such as... Figure 2 As shown, the device includes a memory 200 and a processor 210; the memory is used to store a program for constructing a seismic denoising dataset, and the processor is used to read and execute the program for constructing a seismic denoising dataset, and to execute the method described in any of the above embodiments.
[0141] Thirdly, embodiments of the present invention also provide a computer-readable storage medium storing a data processing program, wherein the data processing program is executed by a processor using the method for constructing an earthquake denoising dataset as described in any of the above embodiments.
[0142] Example 1
[0143] This example demonstrates a method for constructing a seismic denoising dataset. The specific implementation process is as follows:
[0144] Step 1: Construct the noise region:
[0145] Step 1.1 Select multiple seismic traces from the actual seismic data, and obtain two points (tr1, ti1) and (tr2, ti2) along the first arrival of the earthquake, where tr is the earthquake recording time and ti is the trace number. For example... Figure 3 As shown, Figure 3 The arrow in the upper middle figure represents the first data point, (tr1, ti1). Figure 3 The arrow in the upper middle figure represents the second data point, (tr2, ti2).
[0146] Based on the characteristics of seismic data records, the coordinates of the excitation point and receiver point of a trace can be retrieved in the trace header by the trace number. Here, the coordinates of the excitation point are denoted as (sx1, sy1) and (sx2, sy2), and the coordinates of the receiver point are denoted as (rx1, ry1) and (rx2, ry2).
[0147] Step 1.2 Calculate the offset distance
[0148] Calculate the offsets x1 and x2 for each data point using the following formulas:
[0149]
[0150] Where (sx,sy) and (rx,ry) represent the coordinates of the excitation point and the receiver point, respectively.
[0151] Step 1.3 Calculate time t0 and velocity v;
[0152] Calculate time t0 and velocity v for (x1, ti1) and (x2, ti2) respectively, using the following formulas:
[0153]
[0154] Where x and ti represent the offset and seismic recording time corresponding to a certain point, respectively; x1 is the offset of the first data point, x2 is the offset of the second data point, ti1 is the seismic recording time of the first data point, and ti2 is the seismic recording time of the second data point.
[0155] Step 1.4 Determine the theoretical first arrival line.
[0156] Obtain the coordinates of the excitation and receiving points of all channels, and then obtain the offset of the channel. Calculate the time of t0 and the formation velocity v according to the above steps and the time point of all channels according to formula (2). The line connecting the time points of all channels is called the theoretical first arrival line.
[0157] Step 1.5 Check whether the theoretical first arrival lines of all seismic traces are above the actual first arrival times of the earthquakes. If so, Figure 4 The situation indicated by the middle arrow (where the theoretical first arrival line is below the actual earthquake first arrival time) requires repeating steps 1.1 to 1.5, as follows: Figure 5 As shown in the schematic diagram of the standard theoretical first arrival line, it is necessary to satisfy the condition that the theoretical first arrival line of all seismic traces is above the actual first arrival time of the earthquake.
[0158] Note: The purpose of obtaining the theoretical first-solar-day line is to capture the background noise before the first solar term. Therefore, the principle for selecting the theoretical first-solar-day line is that it should not be after the first solar term, but it is allowed to be slightly higher. However, if the theoretical first-solar-day line is too high, it will result in fewer noise samples.
[0159] Step 1.6 The region above the theoretical initial arrival line is defined as the noise region.
[0160] Step 2: Constructing the effective signal region:
[0161] In this application, seismic signals other than background noise are considered valid signals. Background noise is interference information related to the acquisition environment but unrelated to underground reflection structures; therefore, its morphology, energy, and other characteristics remain relatively stable during seismic data recording. A significant characteristic distinguishing signals from background noise from seismic records is that background noise has strong energy and its energy remains relatively consistent before and after the first arrival, while signal energy is almost zero before the first arrival. Based on this characteristic, the steps for identifying valid signals in seismic records are as follows:
[0162] Step 2.1 Based on the given seismic trace, calculate the energy e before the first arrival using the following formula. fb and the energy e of the entire earthquake channel all :
[0163] The energy before arrival e fb :
[0164] e fb =max{|d i |,i=0,1,…,T′};
[0165] Among them, e fb The energy before the first arrival is T, and T′ is the time of the first arrival line in this theory;
[0166] The energy e of the entire seismic channel all :
[0167] e all =max{|d i |,i=0,1,…,T}
[0168] `max{}` represents the maximum value of the sequence, `| |` represents the absolute value, and `d`... i The value of the seismic sampling point at time i is represented; T′ is the time of the theoretical first arrival line of the trace; T is the maximum recording time of the trace.
[0169] Step 2.2 Calculate the relative energy e according to the following formula. r and absolute energy e a :
[0170] In this step, the relative energy e r for:
[0171]
[0172] In the above formula, e r e is relative energy fb For the energy before arrival, e all The energy of the entire earthquake channel;
[0173] The absolute energy e a for:
[0174] e a =e fb (5)
[0175] In the above formula, e a As absolute energy, e fb The energy before the arrival of the first arrival.
[0176] Step 2.3 Set the relative energy threshold v r and absolute energy threshold v a And determine whether the seismic trace is a valid signal trace according to the following formula:
[0177]
[0178] Where f is the effective seismic trace indicator, 1 indicates an effective seismic trace, and 0 indicates an ineffective trace; the relative energy threshold v r The default value is 0.1, and the absolute energy threshold v a The default value is 0.001.
[0179] Step 2.4 sets the minimum region threshold m, with a default value of 30. After steps 2.1-2.3, if the seismic traces identified as valid signals are continuous and their number exceeds the minimum region threshold, then they are considered valid signal sample regions.
[0180] Step 3: Construct a noisy dataset
[0181] There are two types of datasets: single-input datasets and binary-input datasets. When using datasets, you should choose one of these two types. The following demonstrates the process of creating a binary-input dataset:
[0182] Step 1: Based on the given seismic data, follow the steps of obtaining the theoretical first arrival line in step 1 and obtaining the effective signal sample area in step 2.
[0183] Step 2: Given the size of the data patch in the dataset (pt, px), randomly set j, select trace j to trace j+px-1 in the seismic data, and randomly select two noise sample patches n1 and n2 above the theoretical first arrival line; randomly select a signal patch s from the effective sample area, where the size of n1, n2 and s are all (pt, px).
[0184] Step 3: Calculate the noisy sample piece d according to the following formula:
[0185] d=s+n1 (7)
[0186] For a single-input dataset, a single sample contains {d; n1}; for a binary-input dataset, a single sample contains {d, n2; n1}.
[0187] Step 4: Given the number of samples n, the default value is 30000. Repeat steps 1-3 until the number of samples meets n.
[0188] Step 4: Verify the denoising effect of the constructed denoised dataset.
[0189] Step 41: Design and Training of Background Noise Suppression Network
[0190] This section is only used to verify the effectiveness of the above dataset in suppressing noise; the network architecture is arbitrary. Classic networks such as DnCNN and UNet can be used for single-input networks, and their network architectures are publicly available. No specific requirements are set here; this is solely for verifying the denoising effect of the constructed denoised dataset.
[0191] Step 42: Binary Input Background Noise Suppression Network
[0192] A binary input noise suppression network requires two inputs, therefore its network architecture is as follows: Figure 6The diagram shows a binary input denoising network. The network employs an Encoder-Decoder structure, consisting of four Feature Encoding Units (FE), four Feature Decoding Units (FD), and two Feature Mapping Units (FM). To avoid feature loss, skip connections are added between feature maps of the same scale. Each Feature Mapping Unit consists of four 3x3 convolutions and three ReLU activation functions. Each Feature Encoding Unit consists of one set of Feature Mapping Units, one ReLU activation function, and one 3x3 convolution with a stride of 2. Each Feature Decoding Unit consists of one set of Feature Mapping Units, one ReLU activation function, and one 3x3 deconvolution with a stride of 2. Input 1 of the network is the noisy seismic data d, and input 2 is the first-arrival background noise n2.
[0193] Step 43: Train the network parameters and test the model using the test set.
[0194] Step 431: Randomly initialize the parameters of the network to obtain the initial model of the network;
[0195] Step 432: Divide the above dataset into a training set and a test set according to a certain ratio;
[0196] Step 433: Input the training set of the earthquake data into the initial model of the network, obtain the data of the output layer, and determine whether the current iteration number meets the maximum number of training iterations (default value is 50). If not, adjust the network parameters of the current network through the backpropagation algorithm until the maximum number of training iterations is reached. The backpropagation algorithm of the network can be implemented by the Adam optimization algorithm.
[0197] Step 44: Use a network model to suppress noise in the actual data.
[0198] Seismic data is input into a network model to obtain seismic data after background noise suppression.
[0199] like Figure 7 As shown, the actual seismic data collected suffers from severe background noise interference, resulting in a low signal-to-noise ratio. The first arrival of the seismic signal is submerged in the background noise, hindering further processing and interpretation of the seismic data. Using the seismic denoising dataset constructed in this embodiment and combined with a binary input noise suppression network, the trained denoising model can effectively identify and suppress background noise, such as... Figure 8 The diagram shown illustrates the seismic data after background noise suppression processing using the seismic denoising dataset constructed in this embodiment. Figure 8 It is clear from the image that the earthquake signal has been restored, and from... Figure 9The schematic diagram of the background noise suppression in the denoised dataset shows that the background noise is significantly suppressed, with almost no effective signal visible, indicating that the present invention has high amplitude preservation. After suppressing background noise using the present invention, the signal-to-noise ratio of the seismic data is higher, providing high-quality basic data for subsequent processing and interpretation, proving the effectiveness of the seismic denoising dataset constructed in this embodiment for noise suppression.
[0200] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.
Claims
1. A method for constructing an earthquake denoising dataset, characterized in that, The method includes: Obtain actual seismic data from the seismic work area to be processed; Select at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data, and identify the noise region; The relative energy and absolute energy are calculated based on the actual earthquake data, and the effective signal sample area is determined based on the relative energy and the absolute energy. An earthquake denoising sample dataset is established based on the noise region and the effective signal sample region.
2. The method for constructing an earthquake denoising dataset according to claim 1, characterized in that, The step of selecting at least two seismic traces from the actual seismic data to determine the theoretical first arrival line of the actual seismic data and identifying the noise region includes: Select at least two seismic traces from the actual seismic data, and obtain point information of at least two points from the seismic traces along the first and last points of the seismic data. Calculate the offset between two data points based on the seismic record time and trace number of the two points; Calculate the self-excitation and self-recovery time and formation velocity based on the offset between two data points; The theoretical first arrival time of all traces in the actual seismic data is calculated based on the self-excitation and self-reception time and the formation velocity.
3. The method for constructing an earthquake denoising dataset according to claim 2, characterized in that, The point information includes: earthquake record time and trace number.
4. The method for constructing an earthquake denoising dataset according to claim 3, characterized in that, The calculation of the offset between two data points based on the seismic recording time and trace number of the two points includes: The coordinates of the excitation point and the receiver point of each channel are determined based on the channel number corresponding to each data point. Based on the coordinates of the channel excitation point and the channel receiver point of the two data points, the corresponding offset is calculated using the offset formula. The offset formula is as follows: Where x is the offset distance between the two points, (sx,sy) are the coordinates of the channel excitation point, and (rx,ry) are the coordinates of the channel receiver point.
5. The method for constructing an earthquake denoising dataset according to claim 2, characterized in that, The calculation of self-excitation and self-recovery time and formation velocity based on the offset between two data points includes: Obtain the relationship between offset and time; Based on the offset between two data points and the trace number data, the self-excitation and self-recovery time and formation velocity are determined using the formula relating the offset and time. The formula relating the offset and time is: Where x represents the offset of the data point, ti represents the seismic recording time, the trace number data are (x1,ti1) and (x2,ti2), t0 is the self-excitation and self-reception time, v is the formation velocity, x1 is the offset of the first data point, x2 is the offset of the second data point, ti1 is the seismic recording time of the first data point, and ti2 is the seismic recording time of the second data point.
6. The method for constructing an earthquake denoising dataset according to claim 5, characterized in that, After calculating the self-excitation and self-recovery time and formation velocity based on the offset between two data points, the method further includes: Determine whether the theoretical arrival line of all seismic traces is above the actual arrival time of the earthquake based on the theoretical arrival time. If the theoretical first arrival line is below the actual earthquake first arrival time, then two seismic traces should be selected to recalculate the theoretical first arrival line; If the theoretical first arrival line of all seismic traces is above the actual first arrival time of the earthquake, then the data before that theoretical first arrival line is determined to be a noise region.
7. The method for constructing an earthquake denoising dataset according to claim 1, characterized in that, The step of calculating relative and absolute energy based on the actual seismic data, and determining the effective signal sample area based on the relative and absolute energy, includes: Calculate the energy before the first arrival and the energy of the entire seismic trace based on the actual earthquake data. The relative and absolute energies are calculated based on the energy prior to arrival and the energy of the entire seismic trace. Based on the relative energy or the absolute energy, and using a pre-set relative energy threshold or absolute energy threshold, determine whether each seismic trace is a valid signal seismic trace; After performing the above judgment steps on all seismic trace data, if the effective signal seismic traces are continuous and the number exceeds the minimum regional threshold, then the current region is determined to be an effective signal sample region.
8. The method for constructing an earthquake denoising dataset according to claim 7, characterized in that, The calculation of the energy before the first arrival and the energy of the entire seismic trace based on the actual seismic data includes: The energy e before arrival fb : e fb =max{|d i |,i=0,1,…,T′}; Among them, e fb The energy before the first arrival is T, and T′ is the time of the first arrival line in this theory; The energy e of the entire seismic channel all : e all =max{|d i |,i=0,1,…,T} Among them, e all The energy of the entire seismic trace is given by T, where T is the maximum recording time of that trace, || represents the absolute value, and d is the energy of the entire seismic trace. i This represents the value of the seismic sampling point at time i.
9. The method for constructing an earthquake denoising dataset according to claim 7, characterized in that, The relative energy e r for: In the above formula, e r e is relative energy fb For the energy before arrival, e all The energy of the entire earthquake channel; The absolute energy e a for: And a =and fb In the above formula, e a As absolute energy, e fb The energy before the arrival of the first arrival.
10. The method for constructing an earthquake denoising dataset according to claim 9, characterized in that, The step of determining whether each seismic trace is a valid signal seismic trace based on the relative energy or the absolute energy, using a pre-set relative energy threshold or absolute energy threshold, includes: The judgment is made based on the relative energy or absolute energy, combined with the judgment rules of the pre-set relative energy threshold or absolute energy threshold; If the relative energy corresponding to the seismic trace is less than the relative energy threshold, or the absolute energy corresponding to the seismic trace is less than the absolute energy threshold, then the seismic trace is determined to be a valid signal seismic trace. The judgment rule is as follows: Where f is the indicator of a valid signal seismic trace, 1 indicates a valid signal seismic trace, and 0 indicates a non-valid signal seismic trace, v r As the relative energy threshold, v a This is the absolute energy threshold.
11. An apparatus for constructing a seismic denoising dataset, characterized in that, The apparatus includes a memory and a processor; the memory is used to store a program for constructing a seismic denoising dataset, and the processor is used to read and execute the program for constructing a seismic denoising dataset, performing the method according to any one of claims 1-10.
12. A computer-readable storage medium storing a data processing program, the data processing program being executed by a processor as the method for constructing an earthquake denoising dataset according to any one of claims 1-10.