Seismic inversion method, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a seismic inversion method, electronic equipment and a storage medium, and relates to the field of seismic interpretation. The method comprises the following steps: performing well-seismic calibration processing according to logging data and time-domain seismic records to obtain time-domain logging data; obtaining a frequency domain seismic record according to the time domain seismic record; horizon information is extracted from the time domain seismic record, and an elastic impedance initial model is generated according to the time domain logging data and the horizon information; constructing a positive operator according to the elastic impedance initial model and the time domain seismic record to synthesize a pre-stack time domain seismic record and a pre-stack frequency domain seismic record; performing multi-target gradient descent solving processing to obtain a Pareto solution set; and carrying out sorting decision processing to obtain a target solution, and outputting the target solution as a pre-stack elastic impedance inversion result. The problem of single-domain control is avoided by constructing an inversion basis and performing multi-target gradient descent solution and sorting in combination with time-frequency domain data, the robustness of an inversion result is ensured, and accurate reservoir feature description under deep high noise is realized.

Description

Technical Field

[0001] This application relates to the field of seismic interpretation, and more particularly to a seismic inversion method, electronic device, and storage medium. Background Technology

[0002] With the continued growth in global demand for oil and gas resources and the advancement of exploration technologies, deep oil and gas reservoirs have gradually become a key target for exploration and development. However, the complex geological structures of deep reservoirs and the low signal-to-noise ratio of seismic data seriously affect the accuracy of reservoir characteristic descriptions.

[0003] Existing techniques are based on pre-stack elastic impedance inversion, which describes reservoir properties by extending the Aki-Richards equations and combining multi-angle seismic information. Hybrid time-frequency domain inversion further integrates time-domain and frequency-domain data, uses the L2 norm to construct an objective function, and optimizes the inversion results using a gradient descent algorithm. Existing techniques rely on multi-domain data collaboration, but the inversion process still mainly depends on traditional optimization algorithms.

[0004] Because the L2 norm objective function is sensitive to the difference in data magnitude between the time and frequency domains, the time-domain data, with its energy concentrated at the arrival time of the reflection layer (large amplitude), causes the inversion results to over-rely on time-domain information. Simultaneously, the gradient descent algorithm fails to effectively balance the contributions of data from multiple domains, weakening the noise resistance advantage of the frequency domain. Ultimately, the inversion results are dominated by a single domain, making it difficult to achieve high-precision, high-resolution feature characterization in high-noise deep reservoirs. This results in insufficient accuracy in reservoir feature description for existing hybrid time-frequency domain inversion methods in deep, high-noise environments, hindering the efficiency and accuracy of deep oil and gas reservoir exploration and development. Summary of the Invention

[0005] This application provides a seismic inversion method, electronic equipment, and storage medium to achieve accurate reservoir feature description in deep, high-noise environments.

[0006] In a first aspect, embodiments of this application provide a seismic inversion method, including:

[0007] Acquire well logging data and time-domain seismic records, and perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data;

[0008] The time-domain seismic records are converted and processed to obtain frequency-domain seismic records.

[0009] Stratigraphic information is extracted from the time-domain seismic records, and an initial elastic impedance model is generated based on the time-domain well logging data and the stratigraphic information.

[0010] Based on the initial model of elastic impedance and time-domain seismic records, a forward modeling operator is constructed.

[0011] Based on the seismic wavelet, the elastic impedance initial model, and the forward modeling operator, pre-stack time-domain seismic records are synthesized.

[0012] Based on the seismic wavelet, the elastic initial impedance initial model, and the forward modeling operator, a pre-stack frequency domain seismic record is synthesized.

[0013] Based on the pre-stack time-domain seismic records, the pre-stack frequency-domain seismic records, the forward modeling operator, the elastic impedance initial model, and the pre-established initial objective function, a multi-objective gradient descent solution is performed to obtain the Pareto solution set.

[0014] The Pareto solution set is used to perform sorting and decision-making processes to obtain the target solution, which is then output as the pre-stack elastic impedance inversion result.

[0015] In one possible implementation, the step of performing well-seismic calibration processing based on the well logging data to obtain time-domain well logging data includes:

[0016] Data extraction and processing are performed on the well logging data to obtain P-wave velocity, S-wave velocity, and density curves.

[0017] Based on the P-wave velocity, S-wave velocity, and density curves, the time-depth relationship conversion curve is used to convert and process the data to obtain time-domain logging data.

[0018] In one possible implementation, the calculation expression for constructing the forward modeler based on the initial elastic impedance model and time-domain seismic records is as follows:

[0019]

[0020] In the formula, Represents time-domain earthquake records. Indicates the forward operand operator. This represents the initial model of elastic impedance. This represents random noise.

[0021] In one possible implementation, the pre-built initial objective function is:

[0022]

[0023] In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward modeling operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

[0024] In one possible implementation, the Pareto solution set includes a plurality of nondominated solutions and the objective function value of each nondominated solution on the corresponding objective function;

[0025] Accordingly, the process of sorting and deciding based on the Pareto solution set to obtain the target solution includes:

[0026] Construct a solution set matrix based on all non-dominated solutions and all objective function values;

[0027] Based on the solution set matrix, determine the probability matrix and information entropy;

[0028] Based on the probability matrix and information entropy, determine the degree of difference and weight reflected by the entropy;

[0029] Based on the solution set matrix, the degree of difference reflected by the entropy, and the weights, determine the first weighted distance and the second weighted distance for each non-dominated solution;

[0030] The ranking score of the non-dominated solution is calculated using the superior-inferior solution distance method based on the first weighted distance and the second weighted distance of each non-dominated solution.

[0031] Based on the ranking scores of all non-dominated solutions, select one non-dominated solution as the target solution.

[0032] In one possible implementation, determining the first weighted distance and the second weighted distance for each non-dominated solution based on the solution set matrix, the difference reflected by the entropy, and the weights includes:

[0033] Based on the multiple objective function values ​​in the solution set matrix, determine the ideal solution and negative ideal solution for each objective function;

[0034] Based on the solution set matrix and the ideal solution of each objective function, determine the first weighted distance of each non-dominated solution, and based on the solution set matrix, the negative ideal solution of each objective function, the degree of difference reflected by entropy, and the weight, determine the second weighted distance of each non-dominated solution.

[0035] In one possible implementation, the formula for determining the first weighted distance of each non-dominated solution based on the solution set matrix and the ideal solution of each objective function, and the formula for determining the second weighted distance of each non-dominated solution based on the solution set matrix, the negative ideal solution of each objective function, the difference reflected by entropy, and the weight, is as follows:

[0036]

[0037] In the formula, Indicates the first weighted distance. Indicates the second weighted distance. =2, This represents the objective function value after range normalization. The weights that represent entropy. Represents the ideal solution. This represents the negative ideal solution.

[0038] In one possible implementation, determining a non-dominated solution as the target solution from all non-dominated solutions based on their ranking scores includes:

[0039] Sort all non-dominated solutions according to their sorting scores to obtain a sequence of non-dominated solutions.

[0040] The non-dominated solution with the largest sorted score is obtained from the non-dominated solution sequence and taken as the target solution.

[0041] Secondly, embodiments of this application provide a seismic inversion device, comprising:

[0042] The acquisition module is used to acquire well logging data and time-domain seismic records, and perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data.

[0043] The conversion module is used to perform data conversion processing on the time-domain seismic records to obtain frequency-domain seismic records;

[0044] The model building module is used to extract stratigraphic information from the time-domain seismic records and generate an initial elastic impedance model based on the time-domain well logging data and stratigraphic information.

[0045] The construction module is used to construct forward modeling operators based on the initial elastic impedance model and time-domain seismic records;

[0046] The generation module is used to synthesize pre-stack time-domain seismic records based on the seismic wavelet, the initial elastic impedance model, and the forward modeling operator.

[0047] The generation module is also used to synthesize pre-stack frequency domain seismic records based on the seismic wavelet, the elastic initial impedance initial model, and the forward modeling operator.

[0048] The analysis module is used to perform multi-objective gradient descent solution processing based on the pre-stack time domain seismic record, the pre-stack frequency domain seismic record, the forward modeling operator, the elastic impedance initial model, and the pre-built initial objective function to obtain the Pareto solution set;

[0049] The processing module is used to perform sorting decision processing based on the Pareto solution set to obtain the target solution, and output the target solution as the pre-stack elastic impedance inversion result.

[0050] In one possible implementation, the conversion module is specifically used for:

[0051] Data extraction and processing are performed on the well logging data to obtain P-wave velocity, S-wave velocity, and density curves.

[0052] Based on the P-wave velocity, S-wave velocity, and density curves, the time-depth relationship conversion curve is used to convert and process the data to obtain time-domain logging data.

[0053] In one possible implementation, the construction module constructs the computational expression for the forward operator as follows:

[0054]

[0055] In the formula, Represents time-domain earthquake records. Indicates the forward operand operator. This represents the initial model of elastic impedance. This represents random noise.

[0056] In one possible implementation, the initial objective function pre-built in the analysis module is:

[0057]

[0058] In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward calculus operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

[0059] In one possible implementation, the processing module is specifically used for:

[0060] Construct a solution set matrix based on all non-dominated solutions and all objective function values;

[0061] Based on the solution set matrix, determine the probability matrix and information entropy;

[0062] Based on the probability matrix and information entropy, determine the degree of difference and weight reflected by the entropy;

[0063] Based on the solution set matrix, the degree of difference reflected by the entropy, and the weights, determine the first weighted distance and the second weighted distance for each non-dominated solution;

[0064] The ranking score of the non-dominated solution is calculated using the superior-inferior solution distance method based on the first weighted distance and the second weighted distance of each non-dominated solution.

[0065] Based on the ranking scores of all non-dominated solutions, select one non-dominated solution as the target solution.

[0066] In one possible implementation, the processing module is further configured to:

[0067] Based on the multiple objective function values ​​in the solution set matrix, determine the ideal solution and negative ideal solution for each objective function;

[0068] Based on the solution set matrix and the ideal solution of each objective function, determine the first weighted distance of each non-dominated solution, and based on the solution set matrix, the negative ideal solution of each objective function, the degree of difference reflected by entropy, and the weight, determine the second weighted distance of each non-dominated solution.

[0069] In one possible implementation, the formula for calculating the second weighted distance for each non-dominated solution in the processing module is as follows:

[0070]

[0071] In the formula, Indicates the first weighted distance. Indicates the second weighted distance. =2, This represents the objective function value after range normalization. The weights that represent entropy reflect. Represents the ideal solution. This represents the negative ideal solution.

[0072] In one possible implementation, the processing module is further configured to:

[0073] Sort all non-dominated solutions according to their sorting scores to obtain a sequence of non-dominated solutions.

[0074] The non-dominated solution with the largest sorted score is obtained from the non-dominated solution sequence and taken as the target solution.

[0075] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0076] The memory stores computer-executed instructions;

[0077] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0078] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0079] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0080] The seismic inversion method, electronic device, and storage medium provided in this application acquire well logging data and time-domain seismic records, and perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data; perform data conversion processing based on the time-domain seismic records to obtain frequency-domain seismic records; extract stratigraphic information from the time-domain seismic records, and generate an initial elastic impedance model based on the time-domain well logging data and stratigraphic information; construct a forward modeling operator based on the initial elastic impedance model and the time-domain seismic records; synthesize pre-stack time-domain seismic records based on the seismic wavelet, the initial elastic impedance model, and the forward modeling operator; synthesize pre-stack frequency-domain seismic records based on the seismic wavelet, the initial elastic impedance model, and the forward modeling operator; perform multi-objective gradient descent solution processing based on the pre-stack time-domain seismic records, the pre-stack frequency-domain seismic records, the forward modeling operator, the initial elastic impedance model, and the pre-built initial objective function to obtain a Pareto solution set; perform sorting decision processing based on the Pareto solution set to obtain the objective solution, and output the objective solution as the pre-stack elastic impedance inversion result. By acquiring well logging data and performing well-seismic calibration and data conversion, well logging and seismic data are fully integrated; an initial elastic impedance model and forward modeling operator are generated, and the inversion basis is constructed based on physical principles; multi-objective gradient descent solution and ranking decision are performed by combining time-frequency domain data, effectively avoiding single-domain control problems, improving the accuracy, resolution and reliability of the inversion results, enabling more accurate description of deep oil and gas reservoir characteristics, ensuring the robustness of the final inversion results while achieving accurate reservoir characteristic description in a high-noise environment at depth. Attached Figure Description

[0081] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0082] Figure 1 A flowchart illustrating the seismic inversion method provided in this application;

[0083] Figure 2 A schematic diagram of the elastic impedance model provided in this application;

[0084] Figure 3 A schematic diagram of time-domain seismic records provided for this application;

[0085] Figure 4 A schematic diagram of frequency domain seismic records provided for this application;

[0086] Figure 5 A schematic diagram of the initial model of elastic impedance provided in this application;

[0087] Figure 6 A schematic diagram of the elastic impedance inversion results provided in this application;

[0088] Figure 7 This is a schematic diagram of the weighted elastic impedance inversion results using existing technology.

[0089] Figure 8 A schematic diagram of the seismic inversion device provided in this application;

[0090] Figure 9 A schematic diagram of the structure of the electronic device provided in this application.

[0091] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0092] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0093] First, let me explain the terms used in this application:

[0094] Elastic Impedance (EI): A parameter that comprehensively considers the elastic parameters of rock, such as P-wave velocity, S-wave velocity, and density, and takes into account the influence of seismic wave incident angle. It is used to quantitatively characterize the relationship between the elastic characteristics of rock and seismic response, and can more accurately correlate the physical properties of underground rocks with seismic data, thus aiding in reservoir identification.

[0095] Well-seismic calibration: By matching well logging curves (such as sonic and density) with seismic reflection characteristics, a correspondence between the well logging depth domain and the seismic time domain is established, allowing high-resolution well logging information to be integrated into seismic data, providing a basic spatiotemporal correlation for subsequent inversion.

[0096] Forward modeling operator (G): Based on the principle of seismic wave propagation, it constructs a mathematical mapping from the underground elastic impedance model (m) to the seismic record (d), simulates the propagation of seismic waves underground and generates the principle seismic record, and is the core link connecting the model and the actual observation data.

[0097] Pareto Set: In multi-objective optimization, a set of non-dominated solutions. Any solution in the set is not completely surpassed by any other solution across all objectives (such as time-domain or frequency-domain inversion errors), reflecting the optimal trade-offs between different objectives and providing diverse decision choices.

[0098] Entropy Weight TOPSIS: A decision-making algorithm that integrates the entropy weight method and the distance method between best and worst solutions. The entropy weight method adaptively determines the target weight based on the data dispersion, while TOPSIS selects the optimal solution from the multi-objective solution set by calculating the distance between the solution and the ideal / negative ideal solution and ranking them, thus achieving objective and unbiased decision-making.

[0099] Gradient descent: an iterative optimization algorithm that optimizes the model by calculating the gradient of the objective function and adjusting model parameters (such as elastic impedance) along the negative gradient direction, thereby gradually minimizing the objective function (such as time-frequency domain inversion error). It is a commonly used method for inversion solutions.

[0100] Time-domain / frequency-domain seismic records: The time domain records the amplitude changes of seismic waves with time as the axis, reflecting the time sequence characteristics of seismic wave propagation; the frequency domain, after Fourier transform, shows the distribution of frequency components of seismic waves. The combination of the two can complement each other in terms of noise resistance and resolution.

[0101] Non-dominated solution: A solution in the Pareto solution set that, in multi-objective optimization, no other solution is better than the other in all objectives. It is a "better solution" under the trade-off of multiple objectives, constitutes the Pareto front, and supports the selection of the optimal solution.

[0102] In existing technologies, conventional hybrid time-frequency domain seismic inversion methods utilize the L2 norm to construct a mismatch function and employ gradient descent to solve the objective function. However, because the time-domain data volume is much larger than the frequency-domain data, and the L2 norm objective function further amplifies this difference in magnitude, while time-domain energy is concentrated near the arrival times of individual reflection layers, resulting in large amplitude values, and frequency-domain energy is dispersed across different frequency components with relatively small amplitude values ​​at each frequency point, the constructed objective function shifts the optimization results towards the time domain. The inversion results are controlled by a single domain, and in high-noise environments at depth, it fails to fully combine the noise resistance advantages of time-domain inversion with the high resolution advantages of frequency-domain inversion, resulting in poor reservoir characteristic description accuracy and failing to meet the demands for detailed geophysical data in deep oil and gas reservoir exploration and development.

[0103] The seismic inversion method provided in this application fully integrates well logging and seismic data by acquiring well logging data and performing well-seismic calibration and data conversion; it generates an initial elastic impedance model and forward modeling operator, and constructs the inversion basis based on physical principles; it combines time-frequency domain data to perform multi-objective gradient descent solution and ranking decision, effectively avoiding single-domain control problems, improving the accuracy, resolution and reliability of the inversion results, and enabling more accurate description of deep oil and gas reservoir characteristics. It ensures the robustness of the final inversion results while achieving accurate reservoir characteristic description in a high-noise environment at depth.

[0104] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0105] Figure 1 A flowchart illustrating the seismic inversion method provided in this application is shown below. Figure 1 As shown, this embodiment provides a seismic inversion method, which includes the following steps:

[0106] Step S101: Acquire well logging data and time-domain seismic records, and perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data.

[0107] Specifically, well logging data contains information on the physical properties of subsurface rock formations and can be obtained from the oilfield. Well-seismic calibration, based on the principle of seismic wave propagation, establishes a correspondence between well logging depth and seismic time by comparing well logging curves with seismic reflection characteristics, converting well logging data into time-domain well logging data. By acquiring time-domain well logging data containing precise well logging information and eliminating the time-depth differences between well logging data and seismic data through well-seismic calibration, the two are aligned in the time domain, enhancing the seismic data's ability to characterize subsurface geological structures, providing fundamental data support for subsequent inversion, and improving the accuracy of subsequent inversion.

[0108] Time-domain logging data can be represented using the Robinson convolution model:

[0109]

[0110] in, For time-domain logging data; It is random noise; For seismic wavelets; The reflection coefficient; This is the integral variable, used to represent the time offset of the seismic wavelet;

[0111] seismic wavelet Passing through different underground reflective interfaces (corresponding to reflection coefficients) After reflection, plus random noise Time-domain logging data are generated by overlaying data in the time domain. This model can be used to obtain time-domain logging data through forward modeling.

[0112] Step S102: Perform data conversion processing based on the time-domain seismic records to obtain the frequency-domain seismic records.

[0113] Specifically, Fourier transform is used to convert time-domain seismic records from the time dimension to the frequency dimension, decomposing the time-varying seismic signal into a superposition of different frequency components to obtain a frequency-domain seismic record. Extracting frequency-domain features compensates for the insufficient resolution in the time domain and enhances noise robustness.

[0114] Step S103: Extract stratigraphic information from time-domain seismic records and generate an initial elastic impedance model based on time-domain well logging data and stratigraphic information.

[0115] Specifically, reflection characteristics of stratigraphic interfaces are identified from time-domain seismic records to extract stratigraphic information. Combined with elastic parameters of the strata from well logging data, an initial elastic impedance model reflecting the distribution of elastic properties of subsurface strata is constructed based on the principle of elastic impedance. This generated initial elastic impedance model provides fundamental support for inversion, effectively reducing the blind spots in the inversion process and improving its efficiency and accuracy. The process of generating the initial elastic impedance model can be implemented using existing software. After inputting time-domain well logging data and stratigraphic information, the initial elastic impedance model is automatically generated, such as... Figure 2 The figures show the automatically generated initial models of elastic impedance at 8°(a), 16°(b), and 24°(c).

[0116] Step S104: Construct forward modeling operators based on the initial elastic impedance model and time-domain seismic records.

[0117] Specifically, based on the forward modeling principle of seismic wave propagation, and considering an initial elastic impedance model and time-domain seismic records, a forward modeling operator is constructed to simulate the propagation process of seismic waves in the subsurface medium. By establishing a mapping relationship from the subsurface medium model to the seismic records, it is convenient to subsequently adjust the model parameters for inversion by comparing the forward modeling results with the actual records. The constructed forward modeling operator can simulate seismic wave propagation, providing a theoretical basis for inversion calculations.

[0118] Step S105: Synthesize pre-stack time-domain seismic records based on seismic wavelet, initial elastic impedance model, and forward modeling operator.

[0119] Specifically, by combining forward modeling operators with an initial elastic impedance model, and following the propagation laws of seismic waves, pre-stack time-domain seismic records are simulated and generated in the time domain, resulting in a principle-based time-domain seismic record based on the initial model. For example... Figure 3 As shown in the figure, the time-domain seismic records are displayed at different stratigraphic angles (8°(a), 16°(b), 24°(c)) with a signal-to-noise ratio of 5. In this embodiment, the Riker wavelet can be used as the seismic wavelet when creating the simulation input model. Actual data can be obtained by extracting statistical wavelets from the seismic data.

[0120] Step S106: Synthesize pre-stack frequency domain seismic records based on seismic wavelet, elastic initial impedance initial model and forward modeling operator.

[0121] Specifically, by utilizing forward modeling operators and an initial model of elastic initial impedance in the frequency domain, a principle frequency domain seismic record based on the initial model is obtained. This record is then compared with the actual frequency domain record, allowing for analysis of the initial model from a frequency perspective, thus providing more multi-dimensional data support for inversion. For example... Figure 4As shown in the figure, the frequency domain seismic records at different stratigraphic angles (8°(a), 16°(b), 24°(c)) with a signal-to-noise ratio of 5 are displayed. In this embodiment, the Riker wavelet can be used as the seismic wavelet when creating the simulation input model. Actual data can be obtained by extracting statistical wavelets from the seismic data.

[0122] Step S107: Perform multi-objective gradient descent solution processing based on pre-stack time-domain seismic records, pre-stack frequency-domain seismic records, forward modeling operators, elastic impedance initial model, and pre-built initial objective function to obtain the Pareto solution set.

[0123] Specifically, based on the principles of multi-objective optimization and the gradient descent algorithm, and using pre-stack time-domain seismic records, pre-stack frequency-domain seismic records, forward modeling operators, an initial elastic impedance model, and a pre-established initial objective function as a foundation, the Pareto solution set is obtained by adjusting the model parameters to minimize the objective function value. The Pareto solution set contains multiple optimal solutions under different objectives, providing multiple options for subsequent decision-making and improving the reliability of the inversion results.

[0124] The process derives the gradients of the objective function in the time and frequency domains using the chain rule:

[0125]

[0126]

[0127] Gradient vectors of the two objective functions: and By defining a search direction, both objective functions can be reduced simultaneously. However, the two gradient directions may be inconsistent or even conflicting. Therefore, a common search direction is constructed by combining gradients:

[0128] Normalize the gradient:

[0129]

[0130] Constructing the gradient matrix:

[0131]

[0132] The goal is to find a weight vector. This makes the combination direction Minimize. Equilibrium is achieved by defining the following quadratic programming problem:

[0133]

[0134] The objective function can be to minimize the norm of the combined direction, i.e., to find a unit direction that, in the least-squares sense, most closely approximates the common contribution of the two gradients. Constraints ensure that the weights are positive and sum to 1. This can be solved using quadratic programming algorithms (such as the interior-point method) to obtain the optimal weights. Post-calculation of combined directions:

[0135]

[0136] The model is then updated using gradient descent.

[0137]

[0138] in, This refers to the step size. To accelerate the process of obtaining the optimal solution, a decaying step size is used.

[0139]

[0140] The elastic impedance model obtained by each iteration of the initial model of low-frequency elastic impedance and its corresponding objective function value They will all be analyzed. Corresponding time domain objective function value, The corresponding objective function value in the frequency domain is selected and attempted to be added to the Pareto solution set. The Pareto solution set is maintained through non-dominated sorting and crowding distance, ensuring that the size of the solution set does not exceed a preset value (e.g., 50).

[0141] To determine whether a solution is superior to another, construct a Pareto front using the following dominance criterion:

[0142] Suppose there are two candidate solutions, namely and :

[0143]

[0144] Right now Not inferior to in all objectives And it is better than at least one objective. Therefore, the solution is considered to be superior. Dominate , These should be removed. The Pareto front retains only the set of solutions not dominated by any other solution, and maintains solution diversity through crowding distance sorting.

[0145] Step S108: Perform sorting decision processing based on the Pareto solution set to obtain the target solution, and output the target solution as the pre-stack elastic impedance inversion result.

[0146] Specifically, by comprehensively analyzing and sorting the solutions in the Pareto solution set, a target solution that better matches the actual underground geological conditions is obtained, and accurate pre-stack elastic impedance inversion results are output, providing reliable geological data for the exploration and development of deep oil and gas reservoirs.

[0147] The seismic inversion method provided in this invention fully integrates well logging and seismic data by acquiring well logging data and performing well-seismic calibration and data conversion; it generates an initial elastic impedance model and forward modeling operator, and constructs the inversion basis based on physical principles; it combines time-frequency domain data to perform multi-objective gradient descent solution and ranking decision, effectively avoiding single-domain control problems, improving the accuracy, resolution and reliability of the inversion results, and enabling more accurate description of deep oil and gas reservoir characteristics. It ensures the robustness of the final inversion results while achieving accurate reservoir characteristic description in a high-noise environment at depth.

[0148] This embodiment provides a detailed description of the process in the above embodiment of performing well-seismic calibration processing on well logging data to obtain time-domain well logging data. The specific implementation of this process includes the following steps:

[0149] Step a1: Extract and process the data based on the logging data to obtain the P-wave velocity, S-wave velocity, and density curves.

[0150] Specifically, well logging data contains a variety of physical information about underground rock formations. By using existing data extraction techniques and based on the correspondence between different well logging methods and the physical properties of rock formations, P-wave velocity, S-wave velocity, and density curves can be separated from the raw well logging data. These curves reflect the elastic properties of the rock formations, providing a data basis for subsequent calculations and conversions, improving the accuracy and reliability of subsequent steps, and thus ensuring the accuracy of well-seismic calibration.

[0151] Step a2: Convert the P-wave velocity, S-wave velocity and density curves according to the time-depth relationship conversion curve to obtain time-domain logging data.

[0152] Specifically, based on the time-depth conversion principle of seismic wave propagation, there is a corresponding relationship between the propagation time and depth of seismic waves in the subsurface medium. By extracting the P-wave velocity, S-wave velocity, and density curves, and combining them with existing mathematical models of seismic wave propagation time and depth, well logging data in the depth domain is converted to the time domain, resulting in time-domain well logging data. This eliminates the difference between well logging depth and seismic time, achieving well-seismic matching, and enabling high-resolution well logging information to be integrated into the seismic data system, providing fundamental data for subsequent time-domain-based seismic inversion.

[0153] This embodiment illustrates the calculation expression for the forward modeling operator constructed based on the initial elastic impedance model and time-domain seismic records in the above embodiments. The expression is as follows:

[0154]

[0155] In the formula, Represents time-domain earthquake records. Indicates the forward operand operator. This represents the initial model of elastic impedance. This represents random noise.

[0156] Specifically, the computational expression for the forward operator is obtained by discretizing the Robinson convolution model.

[0157] The forward operand can be represented as:

[0158]

[0159]

[0160] in, For seismic wavelets; : indicates the first The elastic impedance value at each sampling point (or corresponding to a specific underground location). The time-domain forward operator can be obtained from the expression of the forward operator. The frequency-domain forward operator can be obtained by transforming the time-domain forward operator from the time dimension to the frequency dimension.

[0161] The initial model m of elastic impedance is used to describe the distribution of elastic properties of underground rock strata, and the forward modeling operator... Based on the principle of seismic wave propagation, a mapping relationship is constructed through mathematical calculations to convert the elastic properties of underground rock strata into the response after seismic wave propagation, i.e., time-domain seismic records. Random noise This simulates unavoidable interference factors during actual exploration, including environmental noise and instrument errors. Random noise is added when processing this data in the software to simulate noise reduction. Time-domain seismic records simulate the propagation of seismic waves from the hypocenter through subsurface media containing rock strata with varying elastic properties, ultimately being received by seismic instruments to form seismic records.

[0162] In some optional implementations, the pre-built initial objective function in the above embodiments is:

[0163]

[0164] In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward calculus operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

[0165] The time-domain seismic record is transformed from the time dimension to the frequency dimension using Fourier transform, decomposing the time-varying seismic signal into a superposition of different frequency components to obtain the frequency-domain seismic record. Extracting frequency-domain features compensates for the insufficient resolution of the time domain and enhances noise robustness.

[0166] Specifically, frequency domain seismic records can be represented as:

[0167]

[0168] in, Represents angular frequency. For seismic wavelets The spectrum; Time-domain reflection coefficient sequence Spectrum:

[0169]

[0170] Discretize the expression to obtain its discrete form:

[0171]

[0172] Expand the complex matrix in the formula into expressions for its real and imaginary parts:

[0173]

[0174] The existing objective function for hybrid time-domain and frequency-domain seismic inversion is expressed as follows:

[0175]

[0176] Represents time-domain earthquake records. Represents the forward calculus operator in the time domain. Represents frequency domain seismic records. This represents the frequency domain forward operator. Used to control the contribution of time-domain and frequency-domain inversion results to the objective function. Represents the regularization term, existing conventional initial model constraints. .

[0177] However, existing objective functions suffer from a problem due to the inconsistency in magnitude between the time and frequency domains. The mismatch between the time and frequency domain functions constructed using the L2 norm results in a significant difference in magnitude, leading to inversion results often being heavily controlled and influenced by a single domain. This fails to fully combine the noise resistance advantages of time-domain inversion with the high resolution advantages of frequency-domain inversion. Therefore, the pre-built initial objective function constructed in this embodiment is:

[0178]

[0179] In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward calculus operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

[0180] Specifically, combining time-frequency domain collaboration with model constraints can effectively reduce inversion ambiguity, making the elastic impedance model closer to the real subsurface medium (e.g., clearer depiction of deep reservoir boundaries and thin interbedded structures). It can adapt to complex deep geological conditions (high noise, tectonic fracturing, etc.), and compared to existing technologies where inversion results are ambiguous and highly susceptible to noise interference, it can output a more reliable elastic impedance distribution, assisting in the accurate identification of oil and gas reservoirs. Initial model constraints can reduce invalid iterations, and the balancing factor simplifies the difficulty of multi-domain adaptation and parameter tuning, making the inversion process more efficient, improving result stability, and reducing reliance on human experience, thus better meeting the needs of industrial-scale batch processing.

[0181] This embodiment details the process of sorting and making decisions based on the Pareto solution set to obtain the objective solution in the above embodiment. The Pareto solution set includes multiple non-dominated solutions and the objective function value of each non-dominated solution on the corresponding objective function. The specific implementation of this process includes the following steps:

[0182] Step c1: Construct the solution set matrix based on all non-dominated solutions and all objective function values.

[0183] Specifically, a standardized data structure is formed by arranging each non-dominated solution in the Pareto solution set and its corresponding objective function value (such as time domain error and frequency domain error) in matrix form. By transforming the multi-objective optimization results into a quantifiable mathematical matrix, input is provided for subsequent entropy weighting and TOPSIS algorithms, achieving a structured representation of the data and ensuring the consistency and comparability of different objective function values.

[0184] By combining the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) algorithm with entropy weighting on the Pareto set, the importance of each objective function is automatically determined, achieving a decision-making process without human bias. Assume there are N non-dominated solutions in the final Pareto solution set, and the corresponding two objective function values ​​(time domain error and frequency domain error) form a matrix:

[0185]

[0186] in, For the first The solution is at the th solution. The value of the objective function The objective function is a dual objective function optimization in both the time and frequency domains.

[0187] Step c2: Determine the probability matrix and information entropy based on the solution set matrix.

[0188] Specifically, the objective function value is transformed into a probability distribution through a probability matrix, eliminating the influence of dimensional differences on the weights. Furthermore, the data dispersion of each objective function is quantified through information entropy, laying the foundation for objectively determining the weights, avoiding biases caused by subjective weighting, and making the subsequent evaluation of non-dominated solutions more consistent with the actual data distribution, thereby improving the scientific nature of decision-making.

[0189] Calculate the probability matrix With information entropy :

[0190]

[0191] in, It is a probability matrix. For information entropy, This represents the objective function value after range normalization.

[0192] Step c3: Based on the probability matrix and information entropy, determine the degree of difference and weight reflected by the entropy.

[0193] Specifically, the difference degree of each objective is calculated based on information entropy. Difference degree = 1 - information entropy. The larger the difference degree, the stronger the ability of the objective function index to distinguish the solution. Then, the weight of each objective is obtained through normalization. By converting information entropy into difference degree that can be directly used for weighting, the subsequent analysis of non-dominated solutions can reasonably reflect the actual importance of each objective, avoid decision-making bias caused by improper subjective weight setting, and improve the accuracy of ranking.

[0194] Entropy reflects the degree of difference (i.e., the importance of the target) and its weight:

[0195]

[0196] in, For the degree of difference, As weight.

[0197] Step c4: Based on the solution set matrix, the degree of difference reflected by the entropy, and the weights, determine the first weighted distance and the second weighted distance for each non-dominated solution.

[0198] Specifically, the first weighted distance and the second weighted distance correspond to the weighted distances from the non-dominated solution to the ideal point and the negative ideal point, respectively. The first weighted distance (distance to the ideal point) is calculated by multiplying the difference between each objective function value of each non-dominated solution and the optimal value of that objective function (the ideal point, such as minimum error in the time domain or minimum error in the frequency domain) by the corresponding objective weight and then summing them up. The second weighted distance (distance to the negative ideal point) is a weighted sum of the differences with the worst value of the objective function (the negative ideal point). By analyzing the objective performance of each objective function weight and non-dominated solution through weighted distance analysis, the multi-objective optimization results are transformed into comparable distance indicators, simplifying the complex comparison of the merits of multiple objectives into a comparison of distance magnitudes and streamlining the basis for decision-making.

[0199] Step c5: Calculate the ranking score of non-dominated solutions using the superior-inferior solution distance method based on the first weighted distance and the second weighted distance of each non-dominated solution.

[0200] Specifically, the TOPSIS algorithm combines the first weighted distance (to the ideal solution) and the second weighted distance (to the negative ideal solution) of each solution to calculate a comprehensive score. By combining the distance relationships between positive and negative ideal solutions, the solutions are globally ranked. The higher the score, the closer the solution is to the ideal state, providing an objective basis for prioritization.

[0201] The TOPSIS score is defined as follows:

[0202]

[0203] in The higher the score, the closer the solution is to the optimal solution. Finally, the solution with the highest score is selected as the optimal elastic impedance inversion result.

[0204] Step c6: Based on the ranking scores of all non-dominated solutions, determine one non-dominated solution as the target solution from all non-dominated solutions.

[0205] Specifically, in the Pareto solution set of multi-objective optimization, the higher the ranking score of the non-dominated solution, the closer its overall performance is to the ideal state (the optimal balance of each objective function). Based on the score, the final target solution is determined. The target solution achieves a relatively optimal balance in multiple objectives such as time domain and frequency domain, providing accurate model results for applications such as seismic inversion.

[0206]

[0207] The optimal solution is The entropy-weighted TOPSIS method not only considers the distribution information and differences between objective functions, but also effectively identifies objective solutions in the multi-objective optimization solution set through normalization, weighted distance measure, and similarity calculation.

[0208] This invention achieves data-driven multi-objective solution sets by constructing standardized solution set matrix data, objectively weighting using entropy weighting, and quantifying and ranking using TOPSIS, thus avoiding the limitations of existing technologies that rely on human experience. Furthermore, by measuring the importance of objectives through difference and comprehensively analyzing the performance of solutions using weighted distance, it selects target solutions that balance the contradiction between time-domain noise resistance and frequency-domain high resolution, effectively reducing subjective bias and improving the scientific rigor and practicality of deep and complex reservoir inversion results.

[0209] This embodiment details the process of determining the first weighted distance and the second weighted distance of each non-dominated solution based on the solution set matrix, the difference reflected by the entropy, and the weights in the above embodiment. The specific implementation of this process includes the following steps:

[0210] Step d1: Based on the multiple objective function values ​​in the solution set matrix, determine the ideal solution and negative ideal solution for each objective function.

[0211] Specifically, the ideal solution is the optimal value of each objective function (e.g., minimizing the minimum value or maximizing the maximum value), while the negative ideal solution is the worst value of each objective function (e.g., minimizing the maximum value or maximizing the minimum value). By traversing the objective function values ​​in the solution set matrix, the theoretically optimal and worst benchmark points for each objective function are determined as references for distance calculation.

[0212] Step d2: Based on the solution set matrix and the ideal solution of each objective function, determine the first weighted distance of each non-dominated solution, and based on the solution set matrix, the negative ideal solution of each objective function, the degree of difference reflected by entropy, and the weight, determine the second weighted distance of each non-dominated solution.

[0213] Specifically, the first weighted distance measures how close the non-dominated solution is to the optimal state; the smaller the distance, the better the solution. The second weighted distance measures how far the non-dominated solution is from the worst state; the larger the distance, the better the solution. These two dimensions quantify the quality of the solutions, while the weights are adjusted using entropy difference to avoid bias caused by different degrees of dispersion of the objective function.

[0214] This invention clarifies the optimization direction by defining ideal and negative ideal states through extreme values, and uses the entropy weight method for objective weighting combined with Euclidean distance to quantify performance differences. This transforms non-dominated solutions in multi-objective optimization into quantifiable and comparable numerical indicators, realizing data-driven sorting of multi-objective solution sets. The sorting results more comprehensively reflect the situation of non-dominated solutions in the multi-objective space, providing a more scientific and objective method for selecting target solutions for pre-stack elastic impedance inversion.

[0215] In this embodiment, based on the solution set matrix and the ideal solution of each objective function, the first weighted distance of each non-dominated solution is determined. Furthermore, based on the solution set matrix, the negative ideal solution of each objective function, the difference reflected by entropy, and the weight, the formula for calculating the second weighted distance of each non-dominated solution is as follows:

[0216]

[0217] In the formula, Indicates the first weighted distance. Indicates the second weighted distance. =2, This represents the objective function value after range normalization. The weights that represent entropy. Represents the ideal solution. This represents the negative ideal solution.

[0218] Specifically, the differences between multiple time-frequency domain objectives are transformed into intuitive distance indicators for quantitative comparison. Dynamically adjusting entropy weights is more suitable for deep, noisy, and multifaceted reservoirs. In the inversion of deep sandstone and conglomerate reservoirs, the frequency domain weights, which are sensitive to details, can be strengthened, or the time domain noise resistance can be emphasized, making distance calculations more consistent with actual geological constraints, providing a basis for subsequent ranking, and avoiding the influence of subjective bias caused by manual weighting.

[0219] This embodiment details the process of determining a non-dominated solution as the target solution from all non-dominated solutions based on the ranking scores of all non-dominated solutions in the above embodiment. The specific implementation of this process includes the following steps:

[0220] Step e1: Sort all non-dominated solutions according to their sorting scores to obtain a sequence of non-dominated solutions.

[0221] Specifically, based on the ranking scores of all non-dominated solutions (the comprehensive score calculated by the distance method between superior and inferior solutions), the solution set is sorted in descending order of score to form a sequence of non-dominated solutions. If solutions with the same score exist, they are further distinguished by the difference degree (i.e., weight) of the entropy weight method or by local comparison of the objective function values. Transforming the multi-objective solution set into an ordered sequence clarifies the priority of solutions, providing a direct basis for subsequent decision-making.

[0222] Step e2: Match the non-dominated solution with the largest sorted score from the non-dominated solution sequence and use it as the target solution.

[0223] Specifically, the solution with the highest ranking score is selected as the target solution from the sorted sequence of non-dominated solutions. If multiple solutions have the same score and are tied for first place, the solution with the smaller difference or better noise resistance is selected. This method quickly selects the solution with the best overall performance from the Pareto front, avoiding bias caused by human intervention.

[0224] like Figure 5 , Figure 6 and Figure 7 As shown, to compare with existing technologies, the same initial model and the same hyperparameters are used for illustrative purposes:

[0225] Figure 6 The inversion results for pre-stack elastic impedance at 8°(a), 16°(b), and 24°(c) under a signal-to-noise ratio of 5 are shown. The inversion method in this embodiment has both noise resistance and high resolution, and can completely depict the underground geological structure even under high noise conditions.

[0226] Figure 7 For existing hybrid time-frequency domain inversion methods, the weight distribution is selected. , , This is used to control the extent to which frequency domain seismic records participate in the inversion. Due to the influence of noise, the accuracy of conventional frequency domain inversion results is limited by the quality of the seismic records, and it is impossible to reasonably balance the relationship between the noise resistance and resolution of the inversion results.

[0227] To quantitatively compare the accuracy of the present invention with that of existing single-time-domain elastic impedance inversion results, single-frequency-domain elastic impedance inversion results, and traditional hybrid time-frequency-domain elastic impedance inversion results, the 8° elastic impedance inversion results of the above inversion methods with CDP=150 were selected and compared and analyzed using cross-correlation coefficients.

[0228] vector with vector The correlation coefficient is defined as:

[0229]

[0230] In the formula, For vectors with vector covariance, and They are vectors with vector The variance.

[0231] The cross-correlation coefficient results show that the correlation coefficient of this invention is as high as 0.965, which is significantly better than other control methods. This indicates that its inversion results are in better agreement with the real elastic impedance model. It can ensure the robustness of the inversion results while achieving accurate reservoir characteristic description in a deep, high-noise environment.

[0232] This invention sorts non-dominated solutions based on their ranking scores and selects the solution with the highest score as the target solution. Using quantitative indicators as the sole criterion, it avoids subjective biases caused by human experience, ensuring the objectivity and scientific rigor of the target solution selection. The target solution selected from the Pareto solution set exhibits noise resistance and high resolution, providing scientifically reliable elastic impedance inversion results for geological scenarios such as complex reservoir characterization.

[0233] Figure 8 This is a schematic diagram of the seismic inversion device provided in this application. Figure 8 As shown, the seismic inversion device 80 includes:

[0234] The acquisition module 801 is used to acquire well logging data and time-domain seismic records, and to perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data.

[0235] The conversion module 802 is used to perform data conversion processing on the time-domain seismic records to obtain the frequency-domain seismic records.

[0236] The model building module 803 is used to extract stratigraphic information from time-domain seismic records and generate an initial elastic impedance model based on time-domain well logging data and stratigraphic information.

[0237] Module 804 is used to construct forward modeling operators based on the initial elastic impedance model and time-domain seismic records.

[0238] The generation module 805 is used to generate pre-stack time-domain seismic records based on seismic wavelets, elastic impedance initial models, and forward modeling operators.

[0239] The generation module 805 is also used to generate pre-stack frequency domain seismic records based on seismic wavelets, elastic initial impedance initial models, and forward modeling operators.

[0240] Analysis module 806 is used to perform multi-objective gradient descent solution processing based on pre-stack time domain seismic records, pre-stack frequency domain seismic records, forward modeling operators, elastic impedance initial model and pre-built initial objective function to obtain Pareto solution set;

[0241] The processing module 807 is used to perform sorting decision processing based on the Pareto solution set, obtain the target solution, and output the target solution as the pre-stack elastic impedance inversion result.

[0242] In one possible implementation, the conversion module 802 is specifically used for:

[0243] Data extraction and processing are performed based on well logging data to obtain P-wave velocity, S-wave velocity, and density curves.

[0244] The time-domain logging data is obtained by converting the P-wave velocity, S-wave velocity, and density curves according to the time-depth relationship conversion curve.

[0245] In one possible implementation, the construction module 804 constructs the computational expression for the forward operator as follows:

[0246]

[0247] In the formula, Represents time-domain earthquake records. Indicates the forward operand operator. This represents the initial model of elastic impedance. This indicates preset random noise.

[0248] In one possible implementation, the initial objective function pre-built in the analysis module is:

[0249]

[0250] In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward calculus operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

[0251] In one possible implementation, the processing module 807 is specifically used for:

[0252] Construct a solution set matrix based on all non-dominated solutions and all objective function values;

[0253] Based on the solution set matrix, determine the probability matrix and information entropy;

[0254] Based on the probability matrix and information entropy, determine the degree of difference and weight reflected by the entropy;

[0255] Based on the solution set matrix, the degree of difference reflected by entropy, and the weights, determine the first weighted distance and the second weighted distance for each non-dominated solution;

[0256] The ranking score of non-dominated solutions is calculated using the superior-inferior solution distance method based on the first weighted distance and the second weighted distance of each non-dominated solution.

[0257] Based on the ranking scores of all non-dominated solutions, select one non-dominated solution as the target solution.

[0258] In one possible implementation, the processing module 807 is further configured to:

[0259] Based on the multiple objective function values ​​in the solution set matrix, determine the ideal solution and negative ideal solution for each objective function;

[0260] Based on the solution set matrix and the ideal solution of each objective function, determine the first weighted distance of each non-dominated solution, and based on the solution set matrix, the negative ideal solution of each objective function, the degree of difference reflected by entropy, and the weight, determine the second weighted distance of each non-dominated solution.

[0261] In one possible implementation, the formula for calculating the second weighted distance for each non-dominated solution in processing module 807 is as follows:

[0262]

[0263] In the formula, Indicates the first weighted distance. Indicates the second weighted distance. =2, This represents the objective function value after range normalization. The weights that represent entropy reflect. Represents the ideal solution. This represents the negative ideal solution.

[0264] In one possible implementation, the processing module 807 is further configured to:

[0265] Sort all non-dominated solutions according to their sorting scores to obtain a sequence of non-dominated solutions.

[0266] The non-dominated solution with the highest sorted score is obtained from the non-dominated solution sequence and used as the target solution.

[0267] The seismic inversion device provided in this embodiment can be used to perform the above-described seismic inversion method. Its implementation principle and technical effect are similar, and will not be described again in this embodiment.

[0268] Figure 9 A schematic diagram of the hardware structure of the electronic device provided in this application, such as... Figure 9 As shown, the electronic device 90 includes at least one processor 901 and a memory 902. Optionally, the seismic inversion device 90 also includes a communication component 903. The processor 901, memory 902, and communication component 903 are connected via a bus 904.

[0269] In the specific implementation process, at least one processor 901 executes computer execution instructions stored in memory 902, causing at least one processor 901 to perform the method described above.

[0270] The specific implementation process of processor 901 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0271] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0272] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0273] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0274] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0275] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0276] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0277] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0278] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0279] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0280] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0281] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0282] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0283] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A seismic inversion method, characterized in that, include: Acquire well logging data and time-domain seismic records, and perform well-seismic calibration processing based on the well logging data to obtain time-domain well logging data; The time-domain seismic records are converted and processed to obtain frequency-domain seismic records. Stratigraphic information is extracted from the time-domain seismic records, and an initial elastic impedance model is generated based on the time-domain well logging data and the stratigraphic information. Based on the initial model of elastic impedance and time-domain seismic records, a forward modeling operator is constructed. Based on the seismic wavelet, the elastic impedance initial model, and the forward modeling operator, pre-stack time-domain seismic records are synthesized. Based on the seismic wavelet, the elastic initial impedance initial model, and the forward modeling operator, a pre-stack frequency domain seismic record is synthesized. Based on the pre-stack time-domain seismic records, the pre-stack frequency-domain seismic records, the forward modeling operator, the elastic impedance initial model, and the pre-established initial objective function, a multi-objective gradient descent solution is performed to obtain the Pareto solution set. The Pareto solution set is used to perform sorting and decision-making processes to obtain the target solution, which is then output as the pre-stack elastic impedance inversion result.

2. The method according to claim 1, characterized in that, The step of performing well-seismic calibration processing based on the well logging data to obtain time-domain well logging data includes: Data extraction and processing are performed on the well logging data to obtain P-wave velocity, S-wave velocity, and density curves. Based on the P-wave velocity, S-wave velocity, and density curves, the time-depth relationship conversion curve is used to convert and process the data to obtain time-domain logging data.

3. The method according to claim 1, characterized in that, The calculation expression for the forward modeling operator, constructed based on the initial elastic impedance model and time-domain seismic records, is as follows: In the formula, Represents time-domain earthquake records. Indicates the forward operand operator. This represents the initial model of elastic impedance. This represents random noise.

4. The method according to claim 3, characterized in that, The pre-built initial objective function is: In the formula, Represents the objective function in the time domain. Represents the objective function in the frequency domain. This represents the initial model of elastic impedance. , , Represents time-domain earthquake records. This represents frequency domain seismic records, which are obtained by performing a Fourier transform on time domain seismic records. Represents the forward calculus operator in the time domain. This represents the frequency domain forward operator. This indicates that the two data points in parentheses are correlated. This represents the balance factor between the time domain and the frequency domain. This represents the initial model constraints.

5. The method according to claim 1, characterized in that, The Pareto solution set includes multiple non-dominated solutions and the objective function value of each non-dominated solution on the corresponding objective function; Accordingly, the process of sorting and deciding based on the Pareto solution set to obtain the target solution includes: Construct a solution set matrix based on all non-dominated solutions and all objective function values; Based on the solution set matrix, determine the probability matrix and information entropy; Based on the probability matrix and information entropy, determine the degree of difference and weight reflected by the entropy; Based on the solution set matrix, the degree of difference reflected by the entropy, and the weights, determine the first weighted distance and the second weighted distance for each non-dominated solution; The ranking score of the non-dominated solution is calculated using the superior-inferior solution distance method based on the first weighted distance and the second weighted distance of each non-dominated solution. Based on the ranking scores of all non-dominated solutions, select one non-dominated solution as the target solution.

6. The method according to claim 5, characterized in that, The step of determining the first weighted distance and the second weighted distance for each non-dominated solution based on the solution set matrix, the difference reflected by the entropy, and the weights includes: Based on the multiple objective function values ​​in the solution set matrix, determine the ideal solution and negative ideal solution for each objective function; Based on the solution set matrix and the ideal solution of each objective function, determine the first weighted distance of each non-dominated solution, and based on the solution set matrix, the negative ideal solution of each objective function, the degree of difference reflected by entropy, and the weight, determine the second weighted distance of each non-dominated solution.

7. The method according to claim 5, characterized in that, The formula for determining the first weighted distance of each non-dominated solution based on the solution set matrix and the ideal solution of each objective function, and the formula for determining the second weighted distance of each non-dominated solution based on the solution set matrix, the negative ideal solution of each objective function, the difference reflected by entropy, and the weight, is as follows: In the formula, Indicates the first weighted distance. Indicates the second weighted distance. =2, This represents the objective function value after range normalization. The weights that represent entropy. Represents the ideal solution. This represents the negative ideal solution.

8. The method according to claim 4, characterized in that, Based on the ranking scores of all non-dominated solutions, one non-dominated solution is selected as the target solution from all non-dominated solutions, including: Sort all non-dominated solutions according to their sorting scores to obtain a sequence of non-dominated solutions. The non-dominated solution with the largest sorted score is obtained from the non-dominated solution sequence and taken as the target solution.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8.

Images