An intelligent matching method for pre-stack seismic data of P-wave and converted wave

Abstract

The invention discloses a method for intelligently matching pre-stack seismic data of longitudinal waves and converted waves, and belongs to the field of exploration geophysics. The method comprises the following steps: pre-processing real longitudinal wave seismic data and real converted wave seismic data to obtain standard converted wave seismic data and real converted wave seismic data in the longitudinal wave time domain; segmenting the real converted wave seismic data in the longitudinal wave time domain to obtain segmented converted wave seismic data; constructing a neural network model, inputting the endpoint time vectors of the segmented converted wave seismic data into the neural network model to obtain time-adjusted converted wave seismic data; and updating the parameters of the neural network model based on the standard converted wave seismic data and the time-adjusted converted wave seismic data to obtain final converted wave seismic data.

Description

Technical Field

[0001] The invention belongs to the technical field of exploration geophysics, and in particular relates to an intelligent matching method for pre-stack seismic data of longitudinal waves and converted waves. Background Art

[0002] Reservoir elastic parameters play an important role in the process of oil and gas exploration and development. They are the key basis for evaluating reservoir characteristics, predicting oil and gas distribution, and optimizing development plans. P-wave and converted wave joint seismic AVO (amplitude variation with offset) inversion is one of the important means to obtain reservoir elastic parameters. In the joint inversion process, in order to ensure the accuracy and reliability of the inversion results, it is necessary to match the P-wave and converted wave prestack seismic data to the same time domain. This is because only when the reflection phase axis from the same reflection interface underground has the same two-way travel on the P-wave and converted wave profiles, the elastic parameters obtained by inversion can truly reflect the physical characteristics of the reservoir. The conventional matching method usually compresses the converted wave prestack seismic data into the P-wave time domain and then adjusts it manually. The latter process is not only time-consuming and labor-intensive, but also relies heavily on personal experience, so it is difficult to ensure the accuracy of the matching. This uncertainty in accuracy may introduce additional errors, affect the accuracy of subsequent inversion results, and reduce the overall efficiency of exploration and development work. In addition, with the continuous development of oil and gas exploration technology, the accuracy requirements for reservoir elastic parameter inversion are increasing. Especially in the exploration of complex geological structures such as shale reservoirs, the heterogeneity and anisotropy of the reservoir make the limitations of traditional matching methods more prominent. Therefore, a more efficient and accurate P-wave and converted wave prestack seismic data matching method is urgently needed to meet the needs of modern oil and gas exploration and development. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method for intelligent matching of P-wave and converted wave pre-stack seismic data to solve the above problems existing in the prior art.

[0004] To achieve the above object, the present invention provides a method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves, comprising the following steps:

[0005] Preprocessing the real P-wave seismic data and the real converted wave seismic data to obtain standard converted wave seismic data and real converted wave seismic data in the P-wave time domain;

[0006] Processing the real converted wave seismic data in the longitudinal wave time domain in segments to obtain segmented converted wave seismic data;

[0007] Constructing a neural network model, inputting the endpoint time vectors of the segmented converted wave seismic data into the neural network model to obtain the converted wave seismic data after time adjustment;

[0008] The neural network model parameters are updated based on the standard converted wave seismic data and the converted wave seismic data after time adjustment to obtain final converted wave seismic data.

[0009] Optionally, the process of preprocessing the real P-wave seismic data to obtain the standard converted wave seismic data includes:

[0010] Extracting seismic wavelets based on the real longitudinal wave seismic data;

[0011] Based on the real P-wave seismic data, velocity and density are obtained by using seismic AVO inversion;

[0012] Based on the velocity and density, calculating the converted wave reflection coefficient using the Zoeppritz equation;

[0013] The conversion wave reflection coefficient is convolved and synthesized with the seismic wavelet to obtain standard conversion wave seismic data.

[0014] Optionally, the process of obtaining true converted wave seismic data in the P-wave time domain includes:

[0015] The real converted wave seismic data are resampled by using the fractional extraction method, and the real converted wave seismic data are compressed from the converted wave time domain to the longitudinal wave time domain to obtain the real converted wave seismic data in the longitudinal wave time domain.

[0016] Optionally, the neural network model includes: an input layer, a plurality of hidden layers, a plurality of dropout layers and an output layer;

[0017] The input layer includes neurons, and the hidden layers include , and neurons, and the output layer includes A neuron.

[0018] Optionally, the process of inputting the endpoint time vectors of the segmented converted wave seismic data into the neural network model to obtain the time-adjusted converted wave seismic data includes:

[0019] Inputting the endpoint time vector of the segmented converted wave seismic data into the neural network model to obtain an adjusted endpoint time vector;

[0020] Sequentially split the adjusted endpoint time vector to obtain split endpoint time vectors;

[0021] The split endpoint time vectors replace the endpoint time vectors of the segmented converted wave seismic data one by one to obtain the segmented converted wave seismic data after the replacement time;

[0022] The segmented converted wave seismic data after the replacement time are combined, and the overlapping parts are superimposed to obtain the converted wave seismic data after the adjustment time.

[0023] Optionally, the process of updating the neural network model parameters based on the standard converted wave seismic data and the time-adjusted converted wave seismic data includes:

[0024] Calculating the cross-correlation between the standard converted wave seismic data and the time-adjusted converted wave seismic data;

[0025] The absolute value of the difference between the cross-correlation and 1 is used as the objective function to iteratively update the neural network model parameters until the objective function converges to obtain the final converted wave seismic data.

[0026] The present invention also discloses a computer terminal device, comprising:

[0027] one or more processors;

[0028] A memory, coupled to the processor, for storing one or more programs;

[0029] When the one or more programs are executed by the one or more processors, the one or more processors implement an intelligent matching method for pre-stack seismic data of longitudinal waves and converted waves.

[0030] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, a method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves is implemented.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] The method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves provided by the present invention significantly improves the efficiency and accuracy of seismic data matching through advanced deep learning technology. The method first pre-processes the real longitudinal wave seismic data and the real converted wave seismic data to obtain standard converted wave seismic data and real converted wave seismic data in the longitudinal wave time domain, thereby ensuring the consistency and comparability of the data. Through the segmentation processing technology, the real converted wave seismic data is divided into multiple small segments, so that the neural network model can more finely adjust the endpoint time characteristics of each segment of data, thereby achieving more accurate matching. The constructed neural network model can automatically learn and adjust the endpoint time vector, reduce manual intervention, reduce human errors, and improve the degree of automation of matching. Finally, by iteratively updating the parameters of the neural network model, the final converted wave seismic data that is highly consistent with the standard converted wave seismic data is obtained, which greatly improves the accuracy and efficiency of oil and gas exploration and development, and provides strong technical support for the exploration of complex geological structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0034] Figure 1 It is a flow chart of the intelligent matching method of pre-stack seismic data of P-wave and converted wave according to an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of segmented processing of real converted wave seismic data according to an embodiment of the present invention;

[0036] Figure 3 A schematic diagram of segmented converted wave seismic data before and after adjusting the endpoint time according to an embodiment of the present invention;

[0037] Figure 4 It is a schematic diagram of obtaining converted wave seismic data after adjusting the time by combining the segmented converted wave seismic data after replacing the time according to an embodiment of the present invention;

[0038] Figure 5 The seismic data obtained based on the test logging data in the embodiment of the present invention, wherein Figure (a) is the real P-wave seismic data; Figure (b) is the standard converted wave seismic data; Figure (c) is the real converted wave seismic data; Figure (d) is the real converted wave seismic data compressed to the P-wave time domain; Figure (e) is the converted wave seismic data after time adjustment;

[0039] Figure 6 The seismic data with an incident angle of 15° according to an embodiment of the present invention, wherein Figure (a) is real P-wave seismic data; Figure (b) is standard converted wave seismic data; Figure (c) is real converted wave seismic data; Figure (d) is real converted wave seismic data compressed to the P-wave time domain; Figure (e) is converted wave seismic data after time adjustment;

[0040] Figure 7 The figures are converted wave seismic traces before and after adjustment according to an embodiment of the present invention, wherein Figure (a) shows a standard converted wave seismic trace and a converted wave seismic trace before adjustment by the method; Figure (b) shows a standard converted wave seismic trace and a converted wave seismic trace after adjustment by the method. DETAILED DESCRIPTION

[0041] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0042] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0043] Embodiment 1

[0044] The present invention proposes an intelligent matching method for pre-stack seismic data of P-wave and converted waves, aiming to solve the problem of time-consuming and low-precision matching of pre-stack seismic data of P-wave and converted waves in the prior art based on advanced deep learning methods. Through an automated segmented matching process, the efficiency of matching work is improved, the influence of human factors is reduced, the matching degree of pre-stack seismic data of P-wave and converted waves is improved, and then the accuracy of joint inversion elastic parameters is improved, providing more accurate and efficient technical support for oil and gas exploration and development.

[0045] like Figure 1 As shown, in this embodiment, a method for intelligent matching of P-wave and converted wave pre-stack seismic data is provided, and automatic and accurate matching of P-wave and converted wave pre-stack seismic data in the time domain is realized by constructing and training a deep learning model. Specifically, the following steps are included: pre-processing the real P-wave seismic data and the real converted wave seismic data to obtain standard converted wave seismic data and real converted wave seismic data in the P-wave time domain; segmenting the real converted wave seismic data in the P-wave time domain to obtain segmented converted wave seismic data; constructing a neural network model, inputting the time vector of the segmented converted wave seismic data into the neural network model to obtain the converted wave seismic data after time adjustment; updating the parameters of the neural network model based on the standard converted wave seismic data and the converted wave seismic data after time adjustment to obtain the final converted wave seismic data.

[0046] 1. Preprocessing of P-wave and converted wave pre-stack seismic data

[0047] Acquire real P-wave seismic data and real converted wave seismic data through field acquisition;

[0048] Seismic wavelets are extracted based on real P-wave seismic data, and P-wave velocity, S-wave velocity and density are obtained by seismic AVO inversion based on real P-wave seismic data; the conversion wave reflection coefficient is calculated based on the P-wave velocity, S-wave velocity and density using the precise Zoeppritz equation, and the conversion wave reflection coefficient is convolved with the extracted seismic wavelet to synthesize standard conversion wave seismic data .

[0049] The real converted wave seismic data are resampled by using the fractional extraction method, and the real converted wave seismic data are compressed from the converted wave time domain to the longitudinal wave time domain to obtain the real converted wave seismic data in the longitudinal wave time domain;

[0050] 2. Segment processing of converted wave prestack seismic data in the P-wave time domain

[0051] like Figure 2 As shown, the real converted wave seismic data compressed into the P-wave time domain Split into The length of each segment is [ , ],[ , ],[ , ], ..., [ , ],[ , ],in, , when The length of the segment is less than When . The length of is usually a multiple of 5 according to the length of the actual converted wave data. .also, , The length of is usually a multiple of 5 according to the length of the actual converted wave data. Arrange the start and end time of each segment of converted wave seismic data in order to obtain a length of The time vector .

[0052] 3. Constructing a neural network model for matching pre-stack seismic data of P-wave and converted waves

[0053] The constructed neural network model includes 1 input layer, 3 hidden layers, 3 Dropout layers and 1 output layer. The input layer includes neurons, corresponding to the dimensions of the input data; the three hidden layers include , and There is a Dropout layer after each hidden layer to alleviate overfitting, and the Dropout rate is set to 0.3. The output layer includes neurons, with the same dimension as the input data, are used to generate the final output vector. The activation function in this network model is the ReLU function, and the learning rate is set to 0.001.

[0054] 4. Intelligent matching of pre-stack seismic data of longitudinal waves and converted waves

[0055] First, the length of the real converted wave seismic data obtained after segmentation processing is The time vector Input into the neural network model constructed above to obtain an input length of The adjusted endpoint time vector Then Split into [ , ],[ , ],[ , ], ..., [ , ],[ , ], so that it is consistent with [ , ],[ , ],[ , ], ..., [ , ],[ , ] One-to-one correspondence and replacement of the endpoint time in x, such as Figure 3 shown.

[0056] Then, the segmented converted wave seismic data of the replaced time are combined, and the overlapping parts are superimposed to obtain the converted wave seismic data after the adjusted time. ,like Figure 4 shown.

[0057] Then calculate Standard converted wave seismic data The absolute value of the difference between it and 1 is used as the objective function, and the neural network parameters are iteratively updated until the objective function converges to obtain the final converted wave seismic data, thus realizing the intelligent matching of P-wave and converted wave pre-stack seismic data.

[0058] The technical solution provided in this embodiment is applied to actual logging data for testing. The data includes P-wave velocity, S-wave velocity and density, with a total of 301 sampling points and a sampling interval of 1 ms. First, the P-wave and converted wave reflection coefficients are calculated using the precise Zoeppritz equation, and the latter are mapped to the converted wave time domain. Then, the P-wave reflection coefficient is convolved with the Ricker wavelet with a main frequency of 60 Hz to synthesize the real P-wave seismic prestack data, as shown in Figure 1. Figure 5 As shown in (a) in the figure, the converted wave reflection coefficient is convolved with the Ricker wavelet with a main frequency of 30 Hz to synthesize the real converted wave pre-stack seismic data, as shown in Figure 5 As shown in (c) in .

[0059] The proposed method is used to match the real P-wave and converted wave pre-stack seismic data obtained above. First, the velocity and density information is inverted based on the P-wave pre-stack seismic data. Then, the converted wave reflection coefficient is calculated using the accurate Zoeppritz equation and convolved with the Ricker wavelet with a main frequency of 60 Hz to synthesize the standard converted wave pre-stack seismic data, as shown in Figure 2. Figure 5 Then, the real converted wave prestack seismic data is resampled using the fractional extraction method and compressed into the P-wave time domain, as shown in (b) in the figure. Figure 5 As shown in (d) in the figure, the converted wave prestack seismic data compressed to the P-wave time domain is different from the standard converted wave prestack seismic data, as shown in Figure 5 The phase axes shown in (b) and (d) are different in time. Finally, the proposed method is used to adjust the converted wave prestack seismic data compressed to the P-wave time domain to obtain the matched converted wave seismic data, as shown in Figure 1. Figure 5 For a clear comparison, seismic data with an incident angle of 15° are extracted one by one and displayed in Figure 6 In the figure, the content represented by each figure is Figure 5 As can be seen from the figure, the converted wave seismic data adjusted by the proposed method matches well with the real P-wave seismic data. Figure 7 is a seismic trace formed by stacking pre-stack seismic data, where: Figure 7 (a) is the standard converted wave seismic trace and the converted wave seismic trace before adjustment by the method proposed in this patent. Figure 7 (b) in the figure is the standard converted wave seismic trace and the converted wave seismic trace adjusted by the method proposed in this patent. As can be seen from the figure, the converted wave seismic data adjusted by the method proposed in this patent has good consistency with the standard converted wave seismic data, and the correlation reaches 0.964, indicating its feasibility and advancement.

[0060] This embodiment proposes an innovative method for intelligent matching of P-wave and converted wave prestack seismic data. This method uses the powerful feature extraction and pattern recognition capabilities of deep neural networks to achieve efficient and accurate matching of P-wave and converted wave prestack seismic data in the time domain. Traditional P-wave and converted wave prestack seismic data matching mainly relies on manual means, which is to observe and compare the time characteristics of different waveforms and manually adjust to achieve matching. In contrast, this method automatically adjusts the time of converted wave prestack seismic data through a deep neural network model, realizes the automation of the matching process, reduces the impact of manual intervention, and thus improves the matching accuracy. In addition, this method also uses a segmented method to process converted wave prestack seismic data, shortens the matching time, improves work efficiency, and reduces matching costs.

[0061] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves, characterized in that: The following steps are involved: Preprocessing the real P-wave seismic data to obtain standard converted wave seismic data; Preprocessing the real converted wave seismic data to obtain real converted wave seismic data in the P-wave time domain; Processing the real converted wave seismic data in the longitudinal wave time domain in segments to obtain segmented converted wave seismic data; Constructing a neural network model, inputting the endpoint time vectors of the segmented converted wave seismic data into the neural network model to obtain the converted wave seismic data after time adjustment; The process of inputting the endpoint time vector of the segmented converted wave seismic data into the neural network model to obtain the converted wave seismic data after time adjustment includes: Inputting the endpoint time vector of the segmented converted wave seismic data into the neural network model to obtain an adjusted endpoint time vector; Sequentially split the adjusted endpoint time vector to obtain split endpoint time vectors; The split endpoint time vectors replace the endpoint time vectors of the segmented converted wave seismic data one by one to obtain the segmented converted wave seismic data after the replacement time; The segmented converted wave seismic data after the replacement time are combined, and the overlapping parts are superimposed to obtain the converted wave seismic data after the adjustment time; The neural network model includes: an input layer, several hidden layers, several Dropout layers and an output layer; The input layer includes neurons, and the hidden layers include , and neurons, and the output layer includes neurons, Represents the number of data in the time vector of the input network; Based on the standard converted wave seismic data and the converted wave seismic data after time adjustment, the neural network model parameters are updated to obtain final converted wave seismic data; The process of updating the neural network model parameters based on the standard converted wave seismic data and the converted wave seismic data after time adjustment includes: Calculating the cross-correlation between the standard converted wave seismic data and the time-adjusted converted wave seismic data; The absolute value of the difference between the cross-correlation and 1 is used as the objective function to iteratively update the neural network model parameters until the objective function converges to obtain the final converted wave seismic data.

2. The method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves according to claim 1, characterized in that: The process of preprocessing the real P-wave seismic data to obtain the standard converted wave seismic data includes: Extracting seismic wavelets based on the real P-wave seismic data; Based on the real P-wave seismic data, velocity and density are obtained by using seismic AVO inversion; Based on the velocity and density, calculating the converted wave reflection coefficient using the Zoeppritz equation; The converted wave reflection coefficient is convolved with the seismic wavelet to obtain standard converted wave seismic data.

3. The method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves according to claim 1, characterized in that: The process of acquiring true converted-wave seismic data in the P-wave time domain includes: The real converted wave seismic data are resampled by using the fractional extraction method, and the real converted wave seismic data are compressed from the converted wave time domain to the longitudinal wave time domain to obtain the real converted wave seismic data in the longitudinal wave time domain.

4. A computer terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent matching method for pre-stack seismic data of longitudinal waves and converted waves as described in any one of claims 1-3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for intelligent matching of pre-stack seismic data of longitudinal waves and converted waves according to any one of claims 1 to 3 is implemented.

Images