Marine seismic data reconstruction method, device, electronic equipment and storage medium
Through the marine seismic data reconstruction method, anomaly detection and generation models are used to process seismic data when the source energy is insufficient, the data morphological distortion problem is solved, the data accuracy and accuracy are improved, and the seabed geological structure and mineral deposit distribution are achieved. reflect.
Patent Information
- Application Number
- CN202410224254.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-28
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-02-28
AI Technical Summary
During the collection of marine seismic data, due to insufficient supply of seismic air pressure, the existing technology lacks efficient, accurate and economical solutions, resulting in low data accuracy and poor accuracy, and cannot accurately reflect the geological structure and mineral reservoirs under the seabed and mineral storage. distributed.
A marine seismic data reconstruction method is adopted to determine whether there is abnormal data by preprocessing the original data. If it exists, multi-source fusion information is obtained for reconstruction; if it does not exist, the seismic data generation model is trained in real time, and the generation network and discrimination are used. The network generates reconstructed data, combining anomaly detection models and seismic data generation models to improve the accuracy and accuracy of the data.
It realizes efficient, accurate and economical reconstruction of earthquake data when the source energy is insufficient, improves the accuracy and accuracy of the data, makes full use of multi-source information such as abnormal sources, similar sources and fixed-depth orientation, and enhances the prior test of data. Use of knowledge.
Smart Images

Figure CN118191945B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of marine geophysical data acquisition, and in particular to a method, device, electronic equipment and storage medium for reconstructing marine seismic data. Background Art
[0002] In order to obtain the distribution of submarine energy mineral deposits and submarine geological structures, two-dimensional and three-dimensional towing cable technology is usually used to collect marine seismic data. Two-dimensional and three-dimensional towing cable technology usually involves at least one operating vessel towing at least one source and at least one tow cable at the stern. The source is used to transmit energy to the seabed, and the tow cable is used to receive reflected waves, refracted waves, etc. caused by the geological structure of the seabed or below the seabed. However, due to reasons such as waves, weather, instruments and equipment, during the operation of marine seismic data collection, there is usually a moment when the air pressure supply of the source is insufficient, resulting in morphological distortion of the collected seismic data.
[0003] When there is a problem of morphological distortion of seismic data due to insufficient air pressure supply at the source at a certain moment or at several moments, the traditional approach is to re-go online and re-collect seismic data or use mathematical methods such as statistics and numerical simulation to calculate seismic data. The method of re-going online and re-collecting not only wastes a lot of manpower and material resources, but also wastes precious offshore operation windows. The use of mathematical calculation methods such as statistics and numerical simulation is performed during data post-processing, not the original seismic data. The data has been processed many times, resulting in error accumulation. At the same time, this method only uses simple mathematical methods, and does not use comprehensive use of abnormal sources, depth and orientation of similar sources, seismic data and other multi-source fusion information, resulting in a lack of prior knowledge of the calculated seismic data. Therefore, the seismic data obtained using this method has low precision and poor accuracy, and cannot accurately reflect the seabed geological structure and mineral distribution.
[0004] Generative adversarial neural network technology has been used in the post-processing of seismic data. Random signals are input into the trained adversarial neural network model, and seismic trace data is generated through the adversarial neural network model. It is mainly used in the case where a certain seismic trace data is missing, or for the case of denoising seismic data during data post-processing. However, there is no efficient, accurate and economical solution in the prior art for data anomalies caused by insufficient source energy. Summary of the invention
[0005] The present invention provides a method, device, electronic equipment and storage medium for reconstructing marine seismic data, which can efficiently, accurately and economically reconstruct seismic data when the source energy is insufficient.
[0006] According to one aspect of the present invention, a method for reconstructing marine seismic data is provided, the method comprising:
[0007] According to a preset period, the collected original seismic data set is obtained;
[0008] Preprocessing the original seismic data set to obtain an initial seismic data set;
[0009] Inputting the initial seismic data set into an anomaly detection model to determine whether there is abnormal data in the initial seismic data set;
[0010] If it exists, then obtaining multi-source fusion information associated with the abnormal data; inputting the multi-source fusion information into the seismic data generation model to generate reconstructed data corresponding to the abnormal data;
[0011] If it does not exist, the initial earthquake data set is used as sample data to train the earthquake number generation model in real time.
[0012] According to another aspect of the present invention, there is provided a marine seismic data reconstruction device, the device comprising:
[0013] A data acquisition module is used to acquire the collected original seismic data set according to a preset period;
[0014] A preprocessing module, used for preprocessing the original seismic data set to obtain an initial seismic data set;
[0015] A judgment module, used for inputting the initial seismic data set into an anomaly detection model to judge whether there is abnormal data in the initial seismic data set;
[0016] A reconstruction data generation module is used to obtain multi-source fusion information associated with the abnormal data if it exists; input the multi-source fusion information into the seismic data generation model to generate reconstruction data corresponding to the abnormal data;
[0017] A training module is used to use the initial earthquake data set as sample data to train the earthquake number generation model in real time if it does not exist.
[0018] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:
[0019] at least one processor; and
[0020] a memory communicatively connected to the at least one processor; wherein,
[0021] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a marine seismic data reconstruction method described in any embodiment of the present invention.
[0022] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement a marine seismic data reconstruction method described in any embodiment of the present invention when executed.
[0023] The technical solution of the embodiment of the present invention obtains the collected original seismic data set according to a preset period; pre-processes the original seismic data set to obtain an initial seismic data set; inputs the initial seismic data set into an anomaly detection model to determine whether there is abnormal data in the initial seismic data set; if there is, obtains multi-source fusion information associated with the abnormal data; inputs the multi-source fusion information into a seismic data generation model to generate reconstructed data corresponding to the abnormal data; if there is no abnormal data, the initial seismic data set is used as sample data to train the earthquake number generation model in real time. This technical solution can efficiently, accurately and economically reconstruct seismic data when the source energy is insufficient.
[0024] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0026] Figure 1 is a flow chart of a method for reconstructing marine seismic data provided according to Embodiment 1 of the present invention;
[0027] Figure 2 is a structural schematic diagram of a marine seismic data reconstruction device provided according to Embodiment 2 of the present invention;
[0028] Figure 3 It is a structural schematic diagram of an electronic device for implementing a method for reconstructing marine seismic data according to an embodiment of the present invention. DETAILED DESCRIPTION
[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0031] Embodiment 1
[0032] Figure 1 This is a flow chart of a method for reconstructing marine seismic data according to the first embodiment of the present invention. This embodiment is applicable to the case of reconstructing marine seismic data. The method can be executed by a marine seismic data reconstruction device. The marine seismic data reconstruction device can be implemented in the form of hardware and / or software. The marine seismic data reconstruction device can be configured in a device. For example, the device can be a background server or other device with communication and computing capabilities. Figure 1 As shown, the method includes:
[0033] S110, acquiring the collected original seismic data set according to a preset period.
[0034] In this solution, the collected original seismic data is mainly obtained by relying on the ocean towing cable technology. Among them, the towing cable technology is usually composed of an offshore operating vessel towing at least one cable and at least one group of seismic sources at the stern. Each cable is equipped with multiple data collectors at equal intervals. The data collected by each data collector is called a seismic channel data. During the operation, the seismic source is usually triggered to emit energy to the seabed at a preset distance or at a preset time. The multiple data collectors on each cable collect the reflected energy and convert it into a digital signal. After that, it is collected to the data acquisition front end through multiple digital packets to form a SEGD file. Among them, after each triggering of the seismic source, a SEGD file will be generated for each cable.
[0035] Furthermore, the data acquisition front end opens a corresponding cache to store SEGD files. When the number of SEGD files in the cache reaches a preset value, the SEGD files in the data acquisition front end cache are migrated to permanent storage media and servers for storage. Therefore, according to the preset period, the original seismic data set collected is obtained. When the SEGD file is detected to be written to the permanent storage medium or server, all newly written SEGD files are obtained, and all SEGD files are used as the original seismic data set within the preset period of the original earthquake.
[0036] S120: Preprocess the original seismic data set to obtain an initial seismic data set.
[0037] In this embodiment, the original seismic data set is preprocessed, mainly using a bandpass filter to perform frequency filtering on each SEGD file in the original seismic data set. Since a bandpass filter is usually used to perform frequency filtering on the original seismic data during the post-processing of seismic data, in order to improve the efficiency of seismic data generation, a bandpass filter is used to slim down the original seismic data before the seismic data is generated to obtain filtered seismic data.
[0038] S130: Input the initial seismic data set into an anomaly detection model to determine whether there is abnormal data in the initial seismic data set.
[0039] In this scheme, during the marine seismic data collection operation, the operating environment is relatively complex. Strong winds, waves, long-term wear and tear of instruments and equipment, etc., can cause air compressor failures, resulting in insufficient source energy, which distorts the shape of seismic data. In order to detect whether there is seismic data with morphological distortion caused by insufficient source energy in the initial seismic data set, an anomaly detection model is constructed, which is at least one of a clustering model and a neural network model.
[0040] When the anomaly detection model is a clustering model, the source number of the initial seismic data set is obtained, and for the initial seismic data corresponding to each source number, a preset number of seismic trace data are randomly extracted from the initial seismic data as the data to be clustered for the source number. The data to be clustered for each source number are clustered according to the data peak value or the data mean. If the data to be clustered corresponding to a certain source number is not in the same category as the data to be clustered corresponding to other source numbers, it is determined that the initial seismic data corresponding to the source number is morphologically distorted and is abnormal data.
[0041] When the anomaly detection model is a neural network model, each initial seismic data in the initial seismic data set is input into the neural network model, and the neural network model outputs whether the initial seismic data is morphologically distorted, and outputs the source number of the morphologically distorted earthquake. The neural network model is a multi-layer convolutional neural network model, whose input is the seismic profile generated by the initial seismic data, and the characteristic data of the seismic profile is extracted through multiple convolutional layers, and the characteristic data is input into the softmax layer, and the probability of the initial seismic data being morphologically distorted is output. When the probability exceeds the preset threshold, the initial seismic data is judged to be morphologically distorted and is abnormal data.
[0042] When the anomaly detection model is a clustering model and a neural network model, a voting method is used to determine whether the initial seismic data is morphologically distorted, thereby obtaining abnormal data.
[0043] S140, if it exists, obtaining multi-source fusion information associated with the abnormal data; inputting the multi-source fusion information into the seismic data generation model to generate reconstructed data corresponding to the abnormal data.
[0044] Specifically, the multi-source fusion information associated with the abnormal data includes the first seismic wavelet data, the first depth-determined data corresponding to the abnormal earthquake source, the first initial seismic data, the second depth-determined data corresponding to the nearby earthquake source, and the second initial seismic data; wherein the second initial seismic data does not contain any abnormality.
[0045] Optionally, obtain multi-source fusion information associated with abnormal data, including:
[0046] Acquire second initial seismic data, and extract first seismic wavelet data and second depth determination data through the second initial seismic data;
[0047] First initial seismic data is acquired, and first depth determination data is extracted through the first initial seismic data.
[0048] In this scheme, it is assumed that the anomaly detection model detects that the initial seismic data corresponding to the Sth source number and the S+1th source number have anomalies, and the initial seismic data corresponding to the S+2th, S-1th, and S-2th source numbers do not have anomalies, then the S-1th source is taken as a similar source, and the initial seismic data corresponding to the S-1th source number is obtained as the second initial seismic data, and the first seismic wavelet data and the second fixed depth data are extracted based on the second initial seismic data, and the initial seismic data corresponding to the Sth source number is obtained as the first initial seismic data, and the first fixed depth data is extracted based on the first initial seismic data.
[0049] The depth data can be extracted based on the initial seismic data by using the field matching method or the OCR recognition method. When the field matching method is used, the depth data transmitted by the depth-determining orientator in each initial seismic data is obtained by matching the SEGD file with the depth field. When the OCR recognition method is used, the OCR recognition model is pre-trained, and a seismic profile is generated based on the initial seismic data, and the seismic profile is input into the OCR recognition model, thereby outputting the depth data transmitted by the depth-determining orientator in the initial seismic data.
[0050] Each cable is equipped with a depth directional device. The depth directional data transmitted by each depth directional device includes depth data. Therefore, each initial seismic data includes multiple depth data, which can be expressed as: DD = {(depth k )},depth k represents the depth data transmitted by the kth depth-fixing directional device. At this time, the depth data of the i-th seismic trace can be expressed as: represents rounding up, and m represents the number of seismic channels on each cable, which is usually 12 or 24.
[0051] The OCR recognition model includes a regional detection network and a fixed-depth directional data recognition network. The regional detection network is used to identify the seismic profile to obtain a fixed-depth directional data regional image. The fixed-depth data recognition network is used to identify the regional image and output the recognition result, which is the depth data. Among them, the regional detection network and the fixed-depth data recognition network are neural network models.
[0052] Furthermore, conventional techniques may be used to extract seismic wavelet data from initial seismic data, which is not specifically limited in this embodiment.
[0053] Optionally, the seismic data generation model includes a feature extraction network, a mapping network, a generation network, and a discriminant network. The feature extraction network is a convolutional neural network, which is used to extract feature data on the seismic profile; the mapping network is used to map the feature data into feature vectors, and the generation network is used to receive serialized seismic wavelet data, fixed depth data, and feature vectors; and to generate data for each seismic trace, wherein the generation network is composed of a bidirectional multi-layer long short-term memory network based on an attention mechanism, including a forward n-layer LSTM network and a reverse n-layer LSTM network, and the forward i-th layer LSTM network and the reverse i-th layer LSTM are used to generate the i-th seismic trace data y i ; The discriminant network is a deep LSTM network.
[0054] Optionally, the multi-source fusion information is input into a seismic data generation model to generate reconstructed data corresponding to the abnormal data, including:
[0055] Extracting first feature data from the first initial seismic data by the feature extraction network;
[0056] Mapping the first feature data into a first feature vector through the mapping network;
[0057] Inputting the serialized first seismic wavelet data, the first eigenvector, the first depth determination data and the second depth determination data into the generation network to obtain generation data;
[0058] Inputting the generated data and the second initial seismic data into the first discriminant network, and judging whether the generated data meets the preset conditions by the first discriminant network;
[0059] If it meets the requirements, the generated data is used as the reconstructed data and the reconstructed data is output;
[0060] If it does not meet the requirements, the parameters of the seismic data generation model are adjusted to obtain the generated data again.
[0061] Specifically, the first feature data in the first initial seismic profile is extracted through the feature extraction network. The first initial seismic profile is generated according to the first initial seismic data, and the first initial seismic profile is input into the convolutional neural network for feature extraction to obtain the feature data of each seismic trace. The feature data of the i-th seismic trace can be expressed as z i =(z i1 ,z i2 ,…,z il ). The characteristic data can be the time corresponding to multiple maximum values on the seismic profile, or the time when the black pixel exceeds the preset threshold. il Indicates the time corresponding to the lth maximum value on the ith seismic trace.
[0062] The first feature data is mapped into a first feature vector through a mapping network. The mapping network receives the first feature data output by the feature extraction network and maps the first feature data into a feature vector V i =(V i1 ,V i2 ,…,V it ,…,V iT ), specifically:
[0063]
[0064] The values of α and β can be set in advance or through the human-machine interface during the generation process.
[0065] The first seismic wavelet data, the first eigenvector, the first depth determination data and the second depth determination data after serialization are input into the generation network to obtain the generated data. The seismic wavelet data is serialized into a seismic wavelet sequence x, and the seismic wavelet sequence x, the first eigenvector V i, the first fixed depth data DD 1 ={(depth k 1 )} and the second fixed depth data DD 2 ={(depth k 2 )} is input to the first generation network, then the output of the forward i-th layer LSTM network can be expressed as:
[0066]
[0067] in, represents the internal weight of the forward i-th layer LSTM network, After the convolution operation, Indicates that first With V i Perform an inner product operation, then perform a convolution operation with x, and then take the value of the convolution result at time t; Represents an aggregation operation, Represents the weight between the forward i-th layer LSTM network and the forward i-1-th layer LSTM network, called the external weight; represents the bias between the forward i-th layer LSTM and the forward i-1-th layer LSTM network, represents the vector of the forward i-th layer LSTM network at time t-1, Represents the vector of the forward i-1th layer LSTM network at the tth time.
[0068] The output of the reverse i-th layer LSTM network can be expressed as:
[0069]
[0070] in, represents the internal weight of the reverse i-th layer LSTM network, Indicates that first With V i Perform an inner product operation, then perform a convolution operation with x, and then take the value of the convolution result at time t; Represents the weight between the reverse i+1th layer LSTM network and the reverse i-th layer LSTM network, called the external weight; represents the bias between the reverse i+1th layer LSTM network and the reverse i-th layer LSTM network, represents the vector of the reverse i-th layer LSTM network at time t-1, Represents the vector of the reverse i-1th layer LSTM network at the tth time.
[0071] After passing through the bidirectional multi-layer long short-term memory network, the output of the i-th seismic trace at the tth time is:
[0072]
[0073] The data of the i-th seismic trace is expressed as:
[0074] y i =(y i0 ,y i1 ,…,y iT );
[0075] The generated seismic trace data are corrected using the first fixed depth data and the second fixed depth data, which can be expressed as:
[0076]
[0077] in, To preset the cable laying depth value, represents the depth data of the i-th seismic trace calculated based on the first fixed depth data, represents the depth data of the i-th seismic trace calculated based on the second fixed depth data, represents rounding up, and m represents the number of seismic channels on each cable.
[0078] The generated data and the second initial seismic data are input into the discriminant network, and the discriminant network is used to determine whether the generated data meets the preset conditions; if so, the generated data is used as the reconstructed data and the reconstructed data is output; if not, the seismic data generation model parameters are adjusted to obtain the generated data again.
[0079] Specifically, the generated data is used as the reconstructed data and the reconstructed data is output, including inputting the generated data and the second initial seismic data into the discriminant network, and judging through the discriminant network whether the seismic trace data generated by the generating network meets the preset conditions. If so, the generated seismic trace data is output as the reconstructed data; if not, the iterative generation continues.
[0080] Among them, the discriminant network is used to discriminate whether the seismic trace data generated by the generating network meets the preset conditions, specifically including: extracting the characteristic vector of the seismic trace data having a common emission point with the second initial seismic data in the generating data through the discriminant network, and discriminating the similarity between the seismic trace data having a common emission point through the characteristic vector, when the similarity reaches a preset threshold, stopping the generating network from generating seismic data, and outputting the generated seismic data as the reconstructed data corresponding to the source energy deficiency S.
[0081] Assuming that the lth seismic trace data in the generated data and the l+4th seismic trace data in the second initial seismic data have a common emission point, the characteristic vector V' of the lth seismic trace data and the characteristic vector V", of the l+4th seismic trace data are extracted through the discriminant network, and the similarity between V' and V", when the similarity reaches a preset threshold, it is judged that the lth seismic trace data is similar to the l+4th seismic trace data. When the similarity between all seismic trace data with a common reflection point reaches a preset threshold, it is determined that the generated seismic data meets the conditions, and the data generated by the generating network is output as reconstructed data.
[0082] Adjusting the parameters of the seismic data generation model to regenerate the generated data includes: when the similarity between all the seismic trace data with common reflection points does not reach a preset threshold, determining that the generated seismic data does not meet the conditions, adjusting the parameters of the seismic data generation model at this time, and regenerating the data through the generation network.
[0083] S150: If it does not exist, use the initial earthquake data set as sample data to train the earthquake number generation model in real time.
[0084] In this scheme, the initial seismic data set is divided into multiple training pairs according to the source sequence number, each training pair includes the first sample initial seismic data corresponding to the first source and the second sample initial seismic data corresponding to the second source, and each training pair is used to train the seismic data generation model. Assuming that the initial seismic data set contains initial seismic data corresponding to 40 source sequences (S, S+1, S+2, ..., S+39), the initial seismic data set is divided into 39 training pairs such as (S, S+1), (S+1, S+2), ..., (S+38, S+39) according to the source sequence number, and any one of each training pair is used as the first source, and the remaining one is used as the second source to train the seismic data generation model.
[0085] Optionally, using the initial earthquake data set as sample data to train an earthquake number generation model in real time includes:
[0086] Dividing the initial seismic data set into a plurality of training pairs according to the sequence number of the seismic sources; wherein the training pairs include a first sample of initial seismic data corresponding to the first seismic source and a second sample of initial seismic data corresponding to the second seismic source;
[0087] Extracting sample seismic wavelet data and first sample depth-determining data based on the first sample initial seismic data, and extracting second sample depth-determining data based on the second sample initial seismic data;
[0088] Performing distortion processing on the first sample initial seismic data to obtain distorted seismic data;
[0089] Inputting the distorted seismic data, sample seismic wavelet data, first sample fixed-depth data, second sample initial seismic data, and second sample fixed-depth data into a seismic data generation model to obtain generated seismic data;
[0090] Extracting seismic data with the same channel number as the first sample initial seismic data from the generated seismic data, and sequentially calculating the similarity between the seismic data with the same channel number;
[0091] If all similarities meet the preset threshold, the seismic data generation model is trained.
[0092] Specifically, based on the first sample initial seismic data, sample seismic wavelet data and first sample fixed depth data are extracted; based on the second sample initial seismic data, second sample fixed depth data are extracted; the first sample initial seismic data is distorted to obtain distorted seismic data; the distorted seismic data, sample seismic wavelet data, first sample fixed depth data, second sample initial seismic data, and second sample fixed depth data are input into the initial seismic data generation model to obtain generated seismic data; seismic data with the same channel number as the first sample initial seismic data are extracted from the generated seismic data, and the similarities between seismic data with the same channel number are calculated in sequence. If all similarities reach a preset threshold, the seismic data generation model is trained. If a certain similarity does not reach the preset threshold, the model parameters are adjusted until all similarities reach the preset threshold.
[0093] Optionally, the first sample initial seismic data is distorted to obtain distorted seismic data, including:
[0094] Based on the forward modeling method, the second seismic wavelet data when the source energy is sufficient and the third seismic wavelet data when the source energy is a preset ratio are simulated;
[0095] Calculating the difference ratio between the second seismic wavelet data and the third seismic wavelet data;
[0096] The distorted seismic data is obtained by calculation using the difference ratio and each seismic trace data in the initial seismic data.
[0097] Specifically, based on the forward simulation method, the second seismic wavelet data when the source energy is sufficient and the third seismic wavelet data when the source energy is a preset ratio are simulated; the difference ratio between the second seismic wavelet data and the third seismic wavelet data is calculated; and the distorted seismic data is obtained by calculating the difference ratio and each seismic trace data in the initial seismic data.
[0098] The preset ratio may be 70%, 60%, etc., or any ratio set manually. The distortion processing may be performed by multiplying the difference ratio by the seismic trace data when the energy is sufficient, or by multiplying the difference ratio, the weight factor, and the seismic trace data when the energy is sufficient, which is not limited here.
[0099] The difference ratio between the second seismic wavelet data and the third seismic wavelet data can be calculated in the following way:
[0100]
[0101] Wherein, Peak2 represents the peak value of the second seismic wavelet data, and Peak3 represents the peak value of the third seismic wavelet data.
[0102] During the training process, it is assumed that the initial earthquake data corresponding to the current earthquake source is in For the i-th seismic trace data in the seismic data, the training objective function is set as:
[0103]
[0104] The technical solution of the embodiment of the present invention obtains the collected original seismic data set according to the preset cycle; pre-processes the original seismic data set to obtain the initial seismic data set; inputs the initial seismic data set into the anomaly detection model to determine whether there is abnormal data in the initial seismic data set; if there is, obtains the multi-source fusion information associated with the abnormal data; inputs the multi-source fusion information into the seismic data generation model to generate the reconstruction data corresponding to the abnormal data; if there is no, the initial seismic data set is used as sample data to train the earthquake number generation model in real time. By executing this technical solution, based on artificial intelligence, big data and other methods, a seismic data generation model is constructed, and seismic data corresponding to abnormal sources are generated based on the seismic data generation model, thereby improving the efficiency of abnormal seismic data reconstruction; at the same time, using abnormal sources, seismic data of similar sources, depth and orientation and other multi-source fusion information as the input of the seismic data generation model, fully considering the prior knowledge in the seismic operation process, so that the generated seismic data has high accuracy and precision; further, using seismic wavelet data to replace traditional random data as the input of the seismic data generation model further improves the precision and accuracy of the generated seismic data.
[0105] Embodiment 2
[0106] Figure 2 Schematic diagram of the structure of a marine seismic data reconstruction device provided according to the second embodiment of the present invention. Figure 2 As shown, the device comprises:
[0107] The data acquisition module 210 is used to acquire the collected original seismic data set according to a preset period;
[0108] A preprocessing module 220 is used to preprocess the original seismic data set to obtain an initial seismic data set;
[0109] A determination module 230 is used to input the initial seismic data set into an anomaly detection model to determine whether there is abnormal data in the initial seismic data set;
[0110] The reconstructed data generation module 240 is used to obtain multi-source fusion information associated with the abnormal data if it exists; input the multi-source fusion information into the seismic data generation model to generate reconstructed data corresponding to the abnormal data;
[0111] The training module 250 is used to use the initial earthquake data set as sample data to train the earthquake number generation model in real time if it does not exist.
[0112] Optionally, the multi-source fusion information includes first seismic wavelet data, first depth-determined data corresponding to an abnormal earthquake source, first initial seismic data, and second depth-determined data and second initial seismic data corresponding to a nearby earthquake source; wherein the second initial seismic data does not contain any abnormality.
[0113] Optionally, the reconstruction data generating module 240 is specifically used for:
[0114] Acquire second initial seismic data, and extract first seismic wavelet data and second depth determination data through the second initial seismic data;
[0115] First initial seismic data is acquired, and first depth determination data is extracted through the first initial seismic data.
[0116] Optionally, the seismic data generation model includes a feature extraction network, a mapping network, a generation network and a discriminant network; wherein the generation network is composed of a bidirectional multi-layer long short-term memory network based on an attention mechanism, and the forward i-th layer LSTM network and the reverse i-th layer LSTM network generate the i-th channel of seismic data; the discriminant network is a deep LSTM network.
[0117] Optionally, the reconstruction data generating module 240 is further used for:
[0118] Extracting first feature data from the first initial seismic data by the feature extraction network;
[0119] Mapping the first feature data into a first feature vector through the mapping network;
[0120] Inputting the serialized first seismic wavelet data, the first eigenvector, the first depth determination data and the second depth determination data into the generation network to obtain generation data;
[0121] Inputting the generated data and the second initial seismic data into the first discriminant network, and judging whether the generated data meets the preset conditions by the first discriminant network;
[0122] If it meets the requirements, the generated data is used as the reconstructed data and the reconstructed data is output;
[0123] If it does not meet the requirements, the parameters of the seismic data generation model are adjusted to obtain the generated data again.
[0124] Optionally, the training module 250 is specifically used for:
[0125] Dividing the initial seismic data set into a plurality of training pairs according to the sequence number of the seismic sources; wherein the training pairs include a first sample of initial seismic data corresponding to the first seismic source and a second sample of initial seismic data corresponding to the second seismic source;
[0126] Extracting sample seismic wavelet data and first sample depth-determining data based on the first sample initial seismic data, and extracting second sample depth-determining data based on the second sample initial seismic data;
[0127] Performing distortion processing on the first sample initial seismic data to obtain distorted seismic data;
[0128] Inputting the distorted seismic data, sample seismic wavelet data, first sample fixed-depth data, second sample initial seismic data, and second sample fixed-depth data into a seismic data generation model to obtain generated seismic data;
[0129] Extracting seismic data with the same channel number as the first sample initial seismic data from the generated seismic data, and sequentially calculating the similarity between the seismic data with the same channel number;
[0130] If all similarities meet the preset threshold, the seismic data generation model is trained.
[0131] Optionally, the training module 250 is further used to:
[0132] Based on the forward modeling method, the second seismic wavelet data when the source energy is sufficient and the third seismic wavelet data when the source energy is a preset ratio are simulated;
[0133] Calculating the difference ratio between the second seismic wavelet data and the third seismic wavelet data;
[0134] The distorted seismic data is obtained by calculation using the difference ratio and each seismic trace data in the initial seismic data.
[0135] An ocean seismic data reconstruction device provided by an embodiment of the present invention can execute an ocean seismic data reconstruction method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.
[0136] Embodiment 3
[0137] Figure 3 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.
[0138] like Figure 3 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0139] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0140] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a method for reconstructing marine seismic data.
[0141] In some embodiments, a method for reconstructing marine seismic data may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for reconstructing marine seismic data described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform a method for reconstructing marine seismic data in any other appropriate manner (e.g., by means of firmware).
[0142] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0143] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0144] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0145] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).
[0146] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.
[0147] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.
[0148] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.
[0149] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for reconstructing marine seismic data, characterized in that: include: According to a preset period, the collected original seismic data set is obtained; Preprocessing the original seismic data set to obtain an initial seismic data set; Inputting the initial seismic data set into an anomaly detection model to determine whether there is abnormal data in the initial seismic data set; If it exists, then obtain multi-source fusion information associated with the abnormal data; input the multi-source fusion information into the seismic data generation model to generate reconstructed data corresponding to the abnormal data, wherein the multi-source fusion information includes the first seismic wavelet data, the first fixed depth data corresponding to the abnormal source, the first initial seismic data, the second fixed depth data corresponding to the adjacent source, and the second initial seismic data; wherein the second initial seismic data does not contain any abnormality; If it does not exist, the initial seismic data set is used as sample data to train the seismic data generation model in real time.
2. The method according to claim 1, characterized in that: Obtain multi-source fusion information associated with abnormal data, including: Acquire second initial seismic data, and extract first seismic wavelet data and second depth determination data through the second initial seismic data; First initial seismic data is acquired, and first depth determination data is extracted through the first initial seismic data.
3. The method according to claim 1, characterized in that The seismic data generation model includes a feature extraction network, a mapping network, a generation network and a discrimination network; wherein the generation network is composed of a bidirectional multi-layer long short-term memory network based on an attention mechanism, and the forward i-th layer LSTM network and the reverse i-th layer LSTM network generate the i-th channel of seismic data; the discrimination network is a deep LSTM network.
4. The method according to claim 3, characterized in that The multi-source fusion information is input into the seismic data generation model to generate reconstructed data corresponding to the abnormal data, including: Extracting first feature data from the first initial seismic data by the feature extraction network; Mapping the first feature data into a first feature vector through the mapping network; Inputting the serialized first seismic wavelet data, the first eigenvector, the first depth determination data and the second depth determination data into the generation network to obtain generation data; Inputting the generated data and the second initial seismic data into the first discriminant network, and judging whether the generated data meets the preset conditions by the first discriminant network; If it meets the requirements, the generated data is used as the reconstructed data and the reconstructed data is output; If it does not meet the requirements, the parameters of the seismic data generation model are adjusted to obtain the generated data again.
5. The method according to claim 1, characterized in that Using the initial earthquake data set as sample data to train an earthquake number generation model in real time includes: Dividing the initial seismic data set into a plurality of training pairs according to the sequence number of the seismic sources; wherein the training pairs include a first sample of initial seismic data corresponding to the first seismic source and a second sample of initial seismic data corresponding to the second seismic source; Extracting sample seismic wavelet data and first sample depth-determining data based on the first sample initial seismic data, and extracting second sample depth-determining data based on the second sample initial seismic data; Performing distortion processing on the first sample initial seismic data to obtain distorted seismic data; Inputting the distorted seismic data, sample seismic wavelet data, first sample fixed-depth data, second sample initial seismic data, and second sample fixed-depth data into a seismic data generation model to obtain generated seismic data; Extracting seismic data with the same channel number as the first sample initial seismic data from the generated seismic data, and sequentially calculating the similarity between the seismic data with the same channel number; If all similarities meet the preset threshold, the seismic data generation model is trained.
6. The method according to claim 5, characterized in that The first sample initial seismic data is distorted to obtain distorted seismic data, including: Based on the forward modeling method, the second seismic wavelet data when the source energy is sufficient and the third seismic wavelet data when the source energy is a preset ratio are simulated; Calculating the difference ratio between the second seismic wavelet data and the third seismic wavelet data; The distorted seismic data is obtained by calculation using the difference ratio and each seismic trace data in the initial seismic data.
7. A marine seismic data reconstruction device, characterized in that: include: A data acquisition module is used to acquire the collected original seismic data set according to a preset period; A preprocessing module, used for preprocessing the original seismic data set to obtain an initial seismic data set; A judgment module, used for inputting the initial seismic data set into an anomaly detection model to judge whether there is abnormal data in the initial seismic data set; A reconstruction data generation module is used to obtain multi-source fusion information associated with the abnormal data if it exists; input the multi-source fusion information into the seismic data generation model to generate reconstruction data corresponding to the abnormal data, wherein the multi-source fusion information includes the first seismic wavelet data, the first fixed depth data corresponding to the abnormal source, the first initial seismic data, the second fixed depth data corresponding to the adjacent source, and the second initial seismic data; wherein the second initial seismic data does not contain any abnormality; A training module is used to use the initial seismic data set as sample data to train the seismic data generation model in real time if it does not exist.
8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory in communication with the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform a marine seismic data reconstruction method according to any one of claims 1-6.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement a marine seismic data reconstruction method according to any one of claims 1 to 6 when executed.
