Seismic reservoir characterization method and system based on adaptive differentiable short-time fourier transform

Abstract

The application discloses a kind of based on adaptive differentiable short-time Fourier transform Seismic Reservoir Characterization method and system, belong to oil and gas exploration and development technical field, the method first obtains the seismic data to be analyzed, constructs the differentiable short-time Fourier transform model of window length and window function time shift according to time position and frequency index joint adaptive adjustment;Multi-objective composite loss function containing Rényi entropy loss, kurtosis loss and regularization constraint is constructed, and window length and window function time shift are iteratively optimized;The time-frequency spectrum of seismic data is calculated using the optimized model;Based on the time-frequency spectrum, extract seismic attenuation attribute, and according to seismic attenuation attribute, reservoir oil and gas content is characterized.The application can alleviate the problem that time resolution and frequency resolution are mutually restricted in traditional time-frequency analysis, improve the accuracy of seismic attenuation estimation and complex reservoir interpretation.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas exploration and development technology, specifically relating to a seismic reservoir characterization method and system based on adaptive differentiable short-time Fourier transform. Background Technology

[0002] As core strategic resources for ensuring the stable operation of the national economy and safeguarding national energy security, oil and natural gas have irreplaceable strategic significance for stable supply and sustainable development. Currently, the focus of oil and gas exploration has gradually shifted from conventional reservoirs to unconventional oil and gas reservoirs such as shale oil, tight sandstone, and thin interbedded layers. These reservoirs are characterized by strong heterogeneity and complex structures, making it difficult for traditional exploration methods to accurately identify and delineate them, posing significant challenges to oil and gas reservoir exploration and interpretation. Accurately extracting information reflecting reservoir characteristics from seismic data helps to achieve precise identification and location of oil and gas reservoirs, thereby significantly improving the accuracy of oil and gas exploration.

[0003] Time-frequency analysis, as a crucial technique for characterizing the localized properties of signals, can convert one-dimensional seismic signals into two-dimensional time-frequency spectra, effectively representing the time-frequency localization characteristics of non-stationary seismic signals. Seismic attenuation properties, as a key attribute for predicting reservoir hydrocarbon content, directly depend on the resolution of the time-frequency transform for estimation accuracy. Constrained by the Heisenberg uncertainty principle, traditional time-frequency analysis methods such as short-time Fourier transform and S-transform struggle to simultaneously achieve adequate time-frequency resolution, resulting in insufficient resolution of the attenuation characterization profile and difficulty in accurately depicting the distribution range of complex reservoirs. While existing differentiable time-frequency analysis has been applied in signal processing fields such as audio processing and mechanical fault detection, the window lengths of the window functions used can only achieve optimization in the time dimension, making it difficult to adapt to the non-stationary characteristics of seismic signals. Therefore, there is an urgent need to develop a time-frequency analysis method that can adapt to the non-stationary characteristics of seismic signals while simultaneously improving time-frequency resolution, thereby enhancing the accuracy of seismic attenuation estimation and reservoir interpretation.

[0004] Specifically, the traditional short-time Fourier transform is limited by a fixed window length. When analyzing non-stationary seismic signals, the local features corresponding to different time locations and frequency components differ, making it difficult to dynamically adjust the analysis scale based on these local features using a fixed window length. For shallow high-frequency components, a shorter window length is needed to capture thin-layer details; for deep low-frequency components, a longer window length is needed to improve frequency resolution. Therefore, a single fixed window length cannot simultaneously meet the time-frequency analysis requirements of both high-frequency and low-frequency components, resulting in a mutual constraint between time resolution and frequency resolution.

[0005] While existing differentiable time-frequency analysis can improve upon the limitations of traditional fixed-parameter time-frequency analysis through parameter optimization, its window function parameters can typically only be optimized globally or in the time dimension. This makes it difficult to jointly adjust the local variations of seismic signals in both time and frequency dimensions. When seismic signals at different time locations and frequency indices exhibit different local time-frequency characteristics, optimizing the window function parameters along only one dimension is insufficient to ensure the analysis window is perfectly suited to the corresponding local seismic signal characteristics. Therefore, dynamically adjusting the window length and window function time shift based on the local frequency characteristics of the seismic signal is one of the key challenges in improving the adaptability of time-frequency analysis for non-stationary seismic signals.

[0006] On the other hand, time-frequency analysis of complex seismic signals requires not only improving time-frequency resolution but also considering the energy focusing ability of the time spectrum, the ability to identify weak reflection signals, the ability to characterize transient responses, and the ability to suppress noise. For local geological features such as thin interbedded layers and reservoir boundaries, it is necessary to highlight the corresponding transient responses as much as possible; for weak reflection signals that are easily masked by strong reflections or noise, it is necessary to improve their identifiability in the time-frequency analysis results; at the same time, it is also necessary to suppress sidelobe diffusion, background noise, and instability caused by parameter variations. Therefore, it is difficult to simultaneously meet the needs of different characterization objectives in the processing of complex seismic signals by optimizing parameters only for a single time-frequency feature.

[0007] Furthermore, seismic attenuation estimation typically requires analysis based on the energy difference between high-frequency and low-frequency components during seismic wave propagation. Seismic waves attenuate as they propagate through subsurface media due to inherent properties such as rock, with attenuation being particularly pronounced when passing through oil and gas reservoirs. Therefore, estimating the attenuation intensity allows for the study of the spatial distribution of oil and gas-bearing reservoirs. Since both low-frequency and high-frequency components originate from time-frequency analysis of seismic data, the resolution, energy focusing, and noise resistance of the time-frequency spectrum further influence the extraction results of seismic attenuation attributes. When time-frequency analysis results suffer from energy diffusion, insufficient local feature resolution, or noise interference, the vertical time resolution and lateral continuity of the attenuation results are easily insufficient, thus affecting the accurate characterization of reservoir distribution and oil and gas content.

[0008] Therefore, existing seismic reservoir characterization techniques still face the following interrelated technical challenges: First, traditional time-frequency analysis uses a fixed window length, making it difficult to simultaneously consider both time and frequency resolution; second, the parameter optimization dimensions of existing differentiable time-frequency analysis are limited, making it difficult to adapt to the non-stationary changes of seismic signals in both the time and frequency dimensions; third, while improving time-frequency resolution, it is also necessary to consider different needs such as energy focusing, weak reflection signal identification, local transient feature enhancement, and noise suppression; fourth, the accuracy of time-frequency analysis will further affect the extraction of low-frequency and high-frequency components and the estimation of seismic attenuation attributes, ultimately affecting the prediction of hydrocarbon content in complex reservoirs and the accurate interpretation of reservoir distribution range.

[0009] Based on the above problems, it is necessary to further improve the adaptability of seismic signal time-frequency analysis to local non-stationary characteristics, reasonably adjust the analysis scale under different time locations and different frequency components, and improve the resolution and characterization accuracy of the obtained time spectrum, so as to provide a more reliable time-frequency analysis basis for subsequent seismic attenuation attribute extraction and hydrocarbon characterization of complex reservoirs. Summary of the Invention

[0010] The technical problem to be solved by this invention is to address the shortcomings of the prior art by providing a seismic reservoir characterization method and system based on adaptive differentiable short-time Fourier transform. This method solves the technical problems of traditional seismic time-frequency analysis, which, when using fixed or single-dimensional optimized window parameters, is unable to simultaneously consider time resolution and frequency resolution, and cannot adaptively adjust the analysis scale according to the local characteristics of non-stationary seismic signals at different time-frequency locations. Consequently, it leads to insufficient resolution of seismic attenuation attributes and limited accuracy in characterizing the hydrocarbon content of complex reservoirs.

[0011] The present invention adopts the following technical solution: The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform includes the following steps: Acquire the earthquake data to be analyzed; An adaptive differentiable short-time Fourier transform model is constructed. The model adopts a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted according to the time position and frequency index. A multi-objective composite loss function is constructed, which includes Rényi entropy loss, kurtosis loss, and regularization constraints. The window length and window function time shift are iteratively optimized based on the multi-objective composite loss function. The optimized window length and window function time shift are substituted into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; Seismic attenuation attributes are extracted based on the time-frequency spectrum, and reservoir hydrocarbon-bearing properties are characterized based on the seismic attenuation attributes.

[0012] Furthermore, the construction of the adaptive differentiable short-time Fourier transform model includes: For the single-channel seismic signal in the seismic data to be analyzed, corresponding window lengths are set at different time locations and at different frequency indices, and the window function time shift is set; A Gaussian window function that is differentiable with respect to the independent variable and the parameters is used as the window function, wherein the time shift of the window function is used to characterize the sliding length of the window function; For single-channel seismic signals The adaptive differentiable short-time Fourier transform is:

[0013] in, This represents the time shift of the window function. Indicates the first The time position, the Window length at each frequency index This represents a window function that is differentiable with respect to the independent variable and the parameter.

[0014] Furthermore, the construction of the multi-objective composite loss function includes: Set the weight coefficients for Rényi entropy loss, kurtosis loss, and regularization constraint respectively; The multi-objective composite loss function is constructed based on the Rényi entropy loss, kurtosis loss, regularization constraint, and corresponding weight coefficients.

[0015] Furthermore, the determination of the Rényi entropy loss includes: The normalized energy distribution is determined based on the time spectrum output by the adaptive differentiable short-time Fourier transform model. The Rényi entropy loss is determined based on the normalized energy distribution and the entropy order.

[0016] Furthermore, the kurtosis loss is determined based on the kurtosis of the time spectrum at different time indices and frequency indices, and the negative average kurtosis is used as the kurtosis loss; The regularization constraints include regularization terms determined based on the sum of the absolute values ​​of the model parameters and regularization terms determined based on the sum of the squares of the model parameters.

[0017] Furthermore, the iterative optimization of the window length and window function time shift based on the multi-objective composite loss function includes: The gradients of the multi-objective composite loss function with respect to the window length and the time shift of the window function are determined based on the chain rule. Based on the gradient, the AdamW optimizer is used to iteratively update the window length and window function time shift until the multi-objective composite loss function converges.

[0018] Furthermore, the AdamW optimizer is trained 500 times, the learning rate is initialized to 0.001, the window length is initialized to 64-128 sampling points, and the window function time shift is initialized to 1 / 8-1 / 4 of the window length.

[0019] Furthermore, the calculation of the time spectrum of the seismic data to be analyzed includes: Read the time variable and trace number of the seismic data to be analyzed; Substitute the time-shifted window length and window function after iterative optimization convergence into the adaptive differentiable short-time Fourier transform model; Based on the time variable and the number of traces, the seismic data to be analyzed is calculated trace by trace to obtain the time spectrum.

[0020] Furthermore, the step of extracting seismic attenuation attributes based on the time-frequency spectrum and characterizing reservoir hydrocarbon content based on the seismic attenuation attributes includes: Based on the spectral analysis results of the earthquake data to be analyzed, low-frequency and high-frequency frequencies that meet the 3dB bandwidth criterion are selected; Based on the time spectrum, extract the low frequency and the low frequency respectively. and high frequency Corresponding low-frequency components and high-frequency components ; According to the low-frequency components and high-frequency components The seismic attenuation properties at the target layer are calculated as follows:

[0021] in, Represents the time spectrum. As a Taoist name, For time variables, Scaling factor The target layer is used as a reference layer; a scaling factor determined based on the reference layer is used to eliminate the inherent energy differences of the target layer. The distribution range of oil and gas reservoirs is determined based on the strong attenuation anomaly zones in the seismic attenuation attributes.

[0022] Secondly, embodiments of the present invention provide a seismic reservoir characterization system based on adaptive differentiable short-time Fourier transform, comprising: The data module is used to acquire the earthquake data to be analyzed. The model module is used to construct an adaptive differentiable short-time Fourier transform model. The model adopts a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted according to the time position and frequency index. The parameter module is used to construct a multi-objective composite loss function that includes Rényi entropy loss, kurtosis loss and regularization constraints, and to iteratively optimize the window length and window function time shift based on the multi-objective composite loss function; The calculation module is used to substitute the optimized window length and window function time shift into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; The characterization module is used to extract seismic attenuation attributes based on the time spectrum and characterize the reservoir's hydrocarbon content based on the seismic attenuation attributes.

[0023] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described adaptive differentiable short-time Fourier transform seismic reservoir characterization method.

[0024] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described adaptive differentiable short-time Fourier transform seismic reservoir characterization method.

[0025] Compared with the prior art, the present invention has at least the following beneficial effects: This seismic reservoir characterization method, based on adaptive differentiable short-time Fourier transform (SFT), constructs an SFT model and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted based on time location and frequency index. Iterative optimization of these parameters is achieved using a multi-objective composite loss function incorporating Rényi entropy loss, kurtosis loss, and regularization constraints. This approach overcomes the limitations of traditional SFT methods that use fixed window lengths and existing differentiable time-frequency analysis methods that only optimize parameters globally or in the time dimension, enabling the analysis window to adapt to different local time-frequency characteristics of the seismic signal. The optimized model is used to obtain the time-spectrum and extract seismic attenuation attributes, alleviating the constraint between time resolution and frequency resolution, improving the resolution of the time-spectrum and the accuracy of seismic attenuation estimation, and providing more accurate time-frequency information for reservoir hydrocarbon characterization.

[0026] Furthermore, for single-channel seismic signals, corresponding window lengths and window function time shifts are set at different time locations and frequency indices. A Gaussian window function that is differentiable with respect to the independent variables and parameters is adopted, allowing the window length and window function time shift to participate in gradient backpropagation and parameter optimization, and enabling the analysis scale to be adjusted according to local characteristics of the seismic signal. For shallow high-frequency components, a shorter window length can be used to capture thin-layer details; for deep low-frequency components, a longer window length can be used to improve frequency resolution, adapting to the analysis needs of non-stationary seismic signals at different time and frequency locations.

[0027] Furthermore, by setting weight coefficients for Rényi entropy loss, kurtosis loss, and regularization constraints respectively, and constructing a multi-objective composite loss function from each loss term and its weight coefficients, the degree of influence of different optimization objectives is adjusted according to the emphasis of time-spectrum representation. This allows different objectives such as energy focusing, weak reflection identification, transient feature enhancement, and parameter constraints to participate in the window function parameter optimization in a coordinated manner, avoiding the problem that it is difficult to adapt to the needs of complex seismic signal processing by using only a single evaluation index.

[0028] Furthermore, the Rényi entropy loss is determined based on the normalized energy distribution and entropy order of the time spectrum. Its purpose is to use the concentration of the time spectrum energy distribution as the basis for parameter optimization, guide the signal energy to gather near the time frequency ridge, and suppress sidelobe diffusion and background noise. At the same time, by adjusting the entropy order, the sensitivity to different energy regions is enhanced, the role of weak reflection signals in the optimization process is strengthened, and the energy focusing, sparsity and noise resistance of the time spectrum are improved.

[0029] Furthermore, negative mean kurtosis is used as the kurtosis loss, and regularization constraints based on the sum of the absolute values ​​of model parameters and the sum of squares of model parameters are set. Among them, the kurtosis loss is used to enhance the transient response of the seismic signal, making local geological features such as reservoir boundaries and thin interbedded layers stand out in the time-frequency spectrum; the regularization based on the sum of the absolute values ​​of parameters is used to promote the sparsity of the results and suppress noise interference, while the regularization based on the sum of squares of parameters is used to limit the parameter amplitude, improve the uniformity of parameter distribution, and constrain the smoothness of time-frequency results.

[0030] Furthermore, the gradient of the multi-objective composite loss function with respect to the window length and window function time shift is determined using the chain rule. The AdamW optimizer is then employed to iteratively update these parameters based on the gradient, establishing a differentiable optimization path between the multi-objective composite loss and the time-frequency transformation parameters. This allows the window length and window function time shift to be automatically adjusted based on the seismic signal and loss function feedback. The AdamW optimizer employs an adaptive learning rate and weight decay strategy, which helps improve the convergence speed and stability of the parameter iteration process and reduces the limitations imposed by manually determining the window function parameters.

[0031] Furthermore, the training epochs, initial learning rate, window length initialization range, and window function time shift initialization range of the AdamW optimizer are limited, providing clear initial parameters and training conditions for the iterative optimization of window length and window function time shift, so that the optimization process can iteratively update according to the predetermined parameter range until the loss function converges.

[0032] Furthermore, after the parameter optimization converges, the optimized window length and window function time shift are substituted into the adaptive differentiable short-time Fourier transform model, and time-frequency calculations are performed for each trace according to the time variable and trace number of the seismic data to be analyzed. The optimized local adaptive parameters are actually applied to the entire seismic data, so that each seismic trace obtains the corresponding time spectrum according to the optimized analysis scale, and time-frequency analysis results are obtained for subsequent low-frequency and high-frequency component extraction and attenuation analysis.

[0033] Furthermore, based on the spectral analysis results of seismic data, low-frequency and high-frequency components satisfying the 3dB bandwidth criterion are selected. The corresponding low-frequency and high-frequency components are then extracted from the time-frequency spectrum, and these components are used to calculate the seismic attenuation properties at the target layer. The propagation attenuation is characterized by the energy difference between the high-frequency and low-frequency components of the seismic wave. A scaling factor is determined using a reference layer to eliminate the influence of the inherent energy difference of the target layer on the attenuation estimation. Furthermore, strong attenuation anomaly zones are used to characterize the distribution of hydrocarbon-bearing reservoirs, transforming the high-resolution time-frequency analysis results into a basis for interpreting the hydrocarbon-bearing properties of the reservoirs.

[0034] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0035] In summary, this invention employs a time-frequency joint adaptive differentiable short-time Fourier transform and multi-objective composite loss optimization to dynamically adjust the window length and window function time shift, thereby obtaining a high-resolution time spectrum and extracting seismic attenuation properties. This alleviates the problem of limited time-frequency resolution and enables accurate characterization of the hydrocarbon content of complex reservoirs.

[0036] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of a noisy synthetic seismic record in an embodiment of the present invention; Figure 2 This is a temporal spectrum diagram of the noisy synthetic seismic record obtained based on the method in an embodiment of the present invention; Figure 3 This is a schematic diagram of the 60Hz single-frequency component extracted from the time spectrum of a noisy synthetic seismic record according to the method described in this embodiment of the invention; Figure 4 This is a schematic diagram of the actual three-dimensional seismic data volume of the research area in an embodiment of the present invention; Figure 5 This is a schematic diagram of the low-frequency components obtained 45ms below layer H1 by the method described in this embodiment of the invention; Figure 6 This is a schematic diagram of the high-frequency components obtained 45ms below layer H1 by the method described in this embodiment of the invention; Figure 7 This is a schematic diagram of the attenuation results obtained 45ms below layer H1 by the method described in this embodiment of the invention; Figure 8 The method described in the embodiments of the present invention is based on Figure 4 A schematic diagram of attenuation results obtained from the Zhonglianjing profile; Figure 9 A schematic diagram illustrating the technical principle of the seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform provided in this embodiment of the invention. Figure 10 A flowchart of a seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform provided in an embodiment of the present invention; Figure 11 A schematic diagram of a computer device provided in an embodiment of the present invention; Figure 12 This is a block diagram of an electronic device according to an embodiment of the present invention.

[0038] Among them, 60. Computer equipment; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access memory unit; 6202. Cache memory unit; 6203. Read-only memory unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0041] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0042] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0043] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0044] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0045] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0046] This invention provides a seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform. It constructs an adaptive differentiable short-time Fourier transform model where the window length and window function time shift can be adaptively adjusted jointly with time and frequency. A multi-objective composite loss function is designed, incorporating Rényi entropy loss, kurtosis loss, and regularization constraints. Gradient descent, backpropagation, and the AdamW optimizer are used to iteratively optimize the window function parameters. This allows for the use of short windows to capture thin-layer details for shallow high-frequency components and long windows to improve frequency resolution for deep low-frequency components, thereby reducing the limitations of fixed window lengths and single-dimensional parameter optimization. The analysis scale can be adaptively adjusted according to the local time-frequency characteristics of the seismic signal. Specifically, Rényi entropy loss improves the focusing of time-frequency energy, enhances the identification of weak reflection signals, and suppresses background noise; kurtosis loss enhances transient response and highlights local geological features such as reservoir boundaries and thin interbedded layers; and regularization constraints improve the sparsity, smoothness, and stability of the time-frequency results. The various loss terms work synergistically to adapt to the processing requirements of complex seismic signals. Based on this, the optimized window length and time shift of the window function are substituted into the adaptive differentiable short-time Fourier transform to obtain a high-resolution time spectrum channel by channel. Seismic attenuation attributes are then extracted based on the low-frequency and high-frequency components of the time spectrum, and strong attenuation anomalies are used to characterize the distribution of hydrocarbon-bearing reservoirs. This effectively alleviates the problem of the mutual constraint between time resolution and frequency resolution in seismic signal time-frequency analysis, reduces the subjectivity and limitations of manually setting window function parameters, improves the resolution, energy focusing, and noise resistance of the time spectrum, and gives the extracted seismic attenuation results high vertical time resolution and strong lateral continuity. This, in turn, improves the accuracy of seismic attenuation estimation and the accuracy of identifying hydrocarbon-bearing properties in complex reservoirs and characterizing reservoir distribution ranges.

[0047] Please see Figure 9 First, the seismic data to be analyzed is acquired, and an adaptive differentiable short-time Fourier transform model is constructed. The model employs a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted based on the time location and frequency index, allowing the window function parameters to be adjusted according to the local time-frequency characteristics of the seismic signal.

[0048] After the model is built, a multi-objective composite loss function is constructed, which includes Rényi entropy loss, kurtosis loss, and regularization constraints. The window length and window function time shift are then iteratively optimized based on this multi-objective composite loss function. Specifically, the Rényi entropy loss constrains the energy distribution of the time spectrum, the kurtosis loss characterizes the transient features of the seismic signal, and the regularization constraints constrain the model parameters. All loss terms participate in the optimization process of the window length and window function time shift.

[0049] After optimizing the window length and window function time shift, the optimized window length and window function time shift are substituted into an adaptive differentiable short-time Fourier transform model to perform time-frequency transformation on the seismic data to be analyzed, obtaining the corresponding time spectrum. This allows seismic signals at different time locations and frequency indices to undergo time-frequency analysis using window function parameters adapted to their local characteristics.

[0050] Furthermore, seismic attenuation attributes are extracted based on the obtained time-frequency spectrum. During the propagation of seismic waves in the subsurface medium, the energy changes of high-frequency and low-frequency components can reflect the attenuation of seismic waves. Therefore, the seismic attenuation attributes at the target layer are determined based on the low-frequency and high-frequency components in the time-frequency spectrum, and the reservoir's hydrocarbon-bearing properties are characterized based on the strong attenuation anomaly areas in the seismic attenuation attributes.

[0051] therefore, Figure 9 The technical principle shown forms a continuous processing flow: seismic data to be analyzed – adaptive differentiable short-time Fourier transform model – time-shift iterative optimization of window length and window function – time-frequency spectrum calculation – seismic attenuation attribute extraction – reservoir hydrocarbon characterization. The time-frequency joint adaptive parameter optimization results are used for seismic attenuation attribute extraction to characterize the reservoir hydrocarbon content.

[0052] Please see Figure 10 The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform described in this invention includes the following steps: S1. Obtain the seismic data to be analyzed and construct an adaptive differentiable short-time Fourier transform model. The window function parameters of the model are jointly and adaptively adjusted with time or frequency. As one embodiment, constructing the adaptive differentiable short-time Fourier transform model includes: For the single-channel seismic signal to be analyzed The adaptive differentiable short-time Fourier transform is defined as follows:

[0053] in, This represents the time shift of the window function (reflecting the sliding length of the window function). Indicates the first The time position, the Window length at each frequency index This represents a differentiable window function. The window function... It must satisfy the condition for the independent variable. and parameters Differentiability is ensured to guarantee successful gradient backpropagation. For a variance of... The Gaussian window, whose differentiable window function is defined as:

[0054] in, The standard deviation of the Gaussian window function is typically taken as a value. It can be set according to the actual situation. This represents the distance from the sampling point to the center of the window function.

[0055] The adaptive differentiable short-time Fourier transform differs from the traditional short-time Fourier transform in that the window length and window function time shift can be dynamically adjusted with time or frequency. This reduces the limitation of a fixed window length and the limitation of existing differentiable short-time Fourier transforms, which can only optimize parameters in the global or time dimension. For non-stationary seismic signals, shallow high-frequency components require short windows to capture thin-layer details, while deep low-frequency components require long windows to improve frequency resolution. The adaptive differentiable short-time Fourier transform perfectly meets this requirement.

[0056] S2. Design a composite loss function that includes Rényi entropy loss, kurtosis loss and regularization constraints, and use gradient descent and backpropagation algorithms to iteratively optimize the window length and window function time shift. As one example, designing a composite loss function and optimizing its parameters includes: S201. Constructing a multi-objective composite loss function:

[0057] in, , , , The weight coefficients for each loss term can be set and adjusted according to the specific optimization problem and the different emphases of the time-spectrum representation.

[0058] In seismic signal processing, entropy loss is used to measure the concentration of energy distribution in the time spectrum. Its core function is to improve the energy focusing of the time spectrum, so that the signal energy is concentrated as much as possible near the time-frequency ridge, and to suppress sidelobe diffusion and background noise.

[0059] This invention uses Rényi entropy as entropy loss, defined as:

[0060] in, The normalized energy distribution is defined as follows: , is the time-frequency spectral coefficient. Represents the entropy order, when At this time, the Rényi entropy approaches the Shannon entropy. The core advantage of the Rényi entropy is that it can be adjusted by changing the parameters. Controlling the sensitivity of entropy to different energy regions.

[0061] when In this way, the weight of low-energy regions (i.e., weak reflection signals) in the loss function can be effectively increased, thereby guiding the optimization process to pay more attention to the weak geological information that is masked by strong reflection or noise. In addition, higher-order Rényi entropy has a natural suppression effect on background noise, further enhancing the noise resistance and sparsity of the time spectrum.

[0062] Kurtosis is a statistical measure describing the shape of a signal's probability distribution, used to measure how sharp or flat the distribution is relative to a normal distribution. Introducing kurtosis into the loss function of time-frequency analysis aims to enhance transient response and highlight local geological features such as reservoir boundaries and thin interbedded layers. For random variables... Kurtosis is defined as the ratio of the fourth central moment to the square of the second central moment:

[0063] in, Represents random variables The mean, This indicates the expected operation.

[0064] In the adaptive differentiable short-time Fourier transform, the kurtosis loss is defined as the negative mean kurtosis:

[0065] in, and These are frequency and time indices, respectively (the negative sign is used to maximize kurtosis while minimizing loss).

[0066] Regularization induces sparse solutions for the parameters by penalizing the sum of their absolute values.

[0067] When used in time-frequency analysis, it can promote sparsity of the time-frequency spectrum, suppress noise interference, and improve the focusing of time-frequency energy. It is defined as:

[0068] Regularization limits the magnitude of parameter values ​​by penalizing the sum of squares, making the parameter distribution more uniform and preventing individual parameters from becoming too large, which could lead to model instability. In time-frequency analysis, Regularization is used to constrain the smoothness of time-frequency results and suppress noise amplification. It is defined as follows:

[0069] in, This represents the model parameters to be constrained. This represents the regularization coefficient.

[0070] The aforementioned multi-objective composite loss function flexibly adjusts the optimization intensity through weight coefficients, achieving joint optimization of the model. The Rényi entropy loss balances energy focusing and weak reflection signal identification, improving the sparsity and noise resistance of the time-frequency spectrum; the kurtosis loss highlights the transient characteristics of the signal, accurately revealing local details such as thin interlayers and reservoir boundaries; regularization constraints ensure the sparsity and smoothness of the time-frequency results. These various losses complement each other, adapting to the needs of complex seismic signal processing.

[0071] S202. Iterative optimization of the window function parameters is performed using gradient descent and backpropagation algorithms: The partial derivative expression can calculate the gradient of almost any smooth differentiable scalar loss function with respect to the parameters of the adaptive differentiable short-time Fourier transform.

[0072] set up It is a scalar loss function that is almost everywhere smooth and differentiable, and depends on the output of the adaptive differentiable short-time Fourier transform. ,and It is a parameter set The function is represented as ,in, This invention primarily employs gradient descent based on the loss function. Optimize the parameters of the adaptive differentiable short-time Fourier transform. If we... Considered S ( Ω The function of ), i.e. O M ×N Given a matrix, then... It is in R O M ×N The function that it is for S ( Ω The derivative of ) is used S ( S ( Ω )) indicates that it is from O M ×N Linear operations to R. The derivative of the loss function with respect to the time-spectral coefficients is defined as:

[0073] In window length The gradient with respect to the window length, which varies with frequency and time, is:

[0074] Similarly, time shift relative to the window function The gradient is:

[0075] After obtaining the gradient of the loss function with respect to each parameter, the AdamW optimizer is used to update and optimize the parameters, with 500 training epochs. The AdamW optimizer combines adaptive learning rate and weight decay strategies, which can effectively improve the convergence speed and stability of the model. Specifically, for the window length... Window function time shift Its update rules are as follows:

[0076]

[0077] in, This represents the learning rate, initialized to 0.001. This indicates the window length, initialized to 64~128 sampling points. The time shift of the window function is initialized to 1 / 8 to 1 / 4 of the window length. Parameter optimization of the adaptive differentiable short-time Fourier transform is achieved by iteratively updating the parameters until the loss function converges.

[0078] S3. Calculate the time spectrum of the seismic data to be analyzed using the optimized adaptive differentiable short-time Fourier transform; After parameter optimization, the final window length and window function time shift are substituted into the adaptive differentiable short-time Fourier transform formula to calculate the time spectrum of the seismic data channel by channel, and high-resolution time-frequency analysis results are obtained.

[0079] S4. Extract seismic attenuation attributes based on the time spectrum, and characterize the reservoir's hydrocarbon content based on the seismic attenuation attributes.

[0080] As one embodiment, extracting seismic attenuation properties and characterizing the reservoir includes: S401. Select characteristic frequencies and calculate seismic attenuation. Seismic attenuation estimation, a key attribute for hydrocarbon reservoir detection, can be effectively quantified through time-frequency analysis. Its core principle is that seismic waves attenuate as they propagate through the subsurface medium due to inherent properties of rocks, with the attenuation phenomenon being particularly pronounced when passing through oil and gas reservoirs. Therefore, the spatial distribution of oil and gas-bearing reservoirs can be studied by estimating the attenuation intensity.

[0081] The attenuation intensity is qualitatively calculated based on the energy difference between high-frequency and low-frequency components during seismic wave propagation. Firstly, based on spectral analysis of the seismic data, the low-frequency component is selected. and high frequency The frequency must conform to the effective frequency band defined by the 3dB bandwidth criterion and satisfy the symmetry of the spectral amplitude, i.e. To ensure energy balance.

[0082] Based on the calculated time spectrum and low frequency and high frequency Extract the corresponding low-frequency components and high-frequency components .

[0083] Calculate the target layer Earthquake attenuation at location:

[0084] in, Represents the time spectrum. As a Taoist name, It is a time variable.

[0085] scaling factor The term used to eliminate inherent energy differences in the target layer is defined as:

[0086] in, For reference level.

[0087] S402. Characterizing reservoir hydrocarbon-bearing properties based on attenuation attributes The strong attenuation anomaly zone in the attenuation estimation results corresponds to the distribution range of oil and gas reservoirs. By comparing the attenuation response of oil and gas wells with that of dry wells and combining it with well logging interpretation conclusions, the reservoir identification accuracy is verified. The high-resolution attenuation results can clearly depict the vertical distribution characteristics of the reservoirs and improve the ability to identify oil and gas reservoirs.

[0088] In another embodiment of the present invention, a seismic reservoir characterization system based on adaptive differentiable short-time Fourier transform is provided. This system can be used to implement the above-mentioned seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform. Specifically, the seismic reservoir characterization system based on adaptive differentiable short-time Fourier transform includes a data module, a model module, a parameter module, a calculation module, and a characterization module.

[0089] The data module is used to acquire the earthquake data to be analyzed. The model module is used to construct an adaptive differentiable short-time Fourier transform model. The model adopts a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted according to the time position and frequency index. The parameter module is used to construct a multi-objective composite loss function that includes Rényi entropy loss, kurtosis loss and regularization constraints, and to iteratively optimize the window length and window function time shift based on the multi-objective composite loss function; The calculation module is used to substitute the optimized window length and window function time shift into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; The characterization module is used to extract seismic attenuation attributes based on the time spectrum and characterize the reservoir's hydrocarbon content based on the seismic attenuation attributes.

[0090] This invention provides a terminal device comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or function. The processor described in this embodiment can be used for the operation of an adaptive differentiable short-time Fourier transform seismic reservoir characterization method, including: Acquire the seismic data to be analyzed; construct an adaptive differentiable short-time Fourier transform model, wherein the model employs a window function differentiable with respect to independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted based on time location and frequency index; construct a multi-objective composite loss function including Rényi entropy loss, kurtosis loss, and regularization constraints, and iteratively optimize the window length and window function time shift based on the multi-objective composite loss function; substitute the optimized window length and window function time shift into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; extract seismic attenuation attributes based on the time spectrum, and characterize the reservoir hydrocarbon content based on the seismic attenuation attributes.

[0091] Please see Figure 11The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the adaptive differentiable short-time Fourier transform seismic reservoir characterization method of this embodiment. To avoid repetition, details are omitted here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the adaptive differentiable short-time Fourier transform seismic reservoir characterization system of this embodiment. To avoid repetition, details are omitted here.

[0092] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 11 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0093] The processor 61 may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0094] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0095] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0096] Please see Figure 12 The terminal device is an electronic device 600, which is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0097] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 10 The steps are shown in the figure.

[0098] Storage unit 620 may include readable media in the form of volatile storage units, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include read-only memory (ROM) 6203.

[0099] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0100] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0101] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem). This communication can be performed via input / output interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0102] This invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the terminal device and extended storage media supported by the terminal device; it can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). More specific examples of the computer-readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0103] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium can also be any readable medium other than a readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0104] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0105] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the seismic reservoir characterization method with adaptive differentiable short-time Fourier transform in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor in the following steps: Acquire the seismic data to be analyzed; construct an adaptive differentiable short-time Fourier transform model, wherein the model employs a window function differentiable with respect to independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted based on time location and frequency index; construct a multi-objective composite loss function including Rényi entropy loss, kurtosis loss, and regularization constraints, and iteratively optimize the window length and window function time shift based on the multi-objective composite loss function; substitute the optimized window length and window function time shift into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; extract seismic attenuation attributes based on the time spectrum, and characterize the reservoir hydrocarbon content based on the seismic attenuation attributes.

[0106] The databases involved in the various embodiments provided by this invention may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the various embodiments provided by this invention may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0107] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0108] Please see Figure 1 The diagram shows a noisy synthetic seismic record. It uses a Ricker wavelet with a dominant frequency of 30Hz and three sets of reflection coefficient sequences with amplitudes of 0.3 and -0.3; 0.3 and -0.3; and 0.3 and 0.3, respectively. The time intervals corresponding to each set of reflection coefficients are 25ms, 30ms, and 40ms, respectively.

[0109] Please see Figure 2 and Figure 3 The figures show the time spectrum of the noisy synthetic seismic record obtained based on the adaptive differentiable short-time Fourier transform and the corresponding extracted 60Hz single-frequency component, respectively. The red dashed line indicates the position of the reflection in the time domain. It can be observed that the adaptive differentiable short-time Fourier transform has high resolution and strong energy focusing ability, and can accurately identify the wavelets corresponding to the three sets of reflection coefficients, further verifying the feasibility and effectiveness of the method.

[0110] Please see Figure 4 , is a three-dimensional seismic data volume, where pink dots represent the location of well W1 without reservoir, black dots represent the locations of wells W2 and W3 with reservoir, red lines represent the well profiles passing through wells W1, W2 and W3, used for subsequent verification of the effectiveness of the seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform, and light green curves represent layers H1 and H2, respectively.

[0111] Please see Figure 5 and Figure 6The image shows a slice of the low-frequency and high-frequency components along the H1 layer 45ms below the H1 layer, based on the adaptive differentiable short-time Fourier transform. Pink dots represent the location of well W1 (without reservoir), while black dots represent the locations of wells W2 and W3 (containing reservoir). It can be observed that the adaptive differentiable short-time Fourier transform can dynamically adjust the window length according to the local characteristics of the seismic signal, effectively alleviating the contradiction between time and frequency resolution. The low-frequency components can clearly depict the reservoir planar distribution, while the high-frequency components effectively suppress noise interference while preserving thin-layer information.

[0112] Please see Figure 7 The results are qualitative attenuation results along the H1 layer 45ms below, based on the adaptive differentiable short-time Fourier transform. It can be observed that the attenuation results obtained by the method proposed in this invention correspond to reservoir well locations and strong attenuation anomaly zones, while non-reservoir wells show no significant attenuation response. This results in high agreement with actual drilling results, achieving accurate reservoir characterization and interpretation.

[0113] Please see Figure 8 ,for Figure 4 The attenuation results obtained from the Zhonglian well profile are based on the adaptive differentiable short-time Fourier transform. The black curve represents the gamma curve, the pink vertical line represents the location of well W1 (corresponding to high gamma values, as shown in the pink box in the figure) without reservoirs, the black vertical lines represent the locations of wells W2 and W3 (corresponding to low gamma values, as shown in the black box in the figure) containing reservoirs, and the green curves represent stratigraphic positions H1 and H2, respectively. It can be observed that the attenuation results obtained by the method of this invention have high vertical time resolution and strong lateral continuity, improving the accuracy of seismic attenuation estimation and thus enhancing the ability to identify oil and gas reservoirs. The obtained attenuation results are in good agreement with the well logging interpretation results.

[0114] In summary, the seismic reservoir characterization method and system based on adaptive differentiable short-time Fourier transform provided by this invention first constructs a differentiable short-time Fourier transform model in which the window length and window function time shift are jointly and adaptively adjusted with time and frequency. Then, a composite loss function incorporating Rényi entropy loss, kurtosis loss, and regularization constraints is designed, and the window length and window function time shift are iteratively optimized using gradient descent and backpropagation algorithms. Next, the optimized transform is used to calculate the high-resolution time spectrum of the seismic data. Finally, seismic attenuation attributes are extracted based on the time spectrum to achieve accurate characterization of reservoir hydrocarbon content. This invention overcomes the limitation of a fixed window length in the short-time Fourier transform, achieving adaptive optimization of time-frequency resolution. It effectively solves the technical problem of existing technologies failing to simultaneously achieve both time-frequency resolution, significantly improving the accuracy of seismic attenuation estimation and the interpretation of complex reservoirs.

[0115] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform, characterized in that, Includes the following steps: Acquire the earthquake data to be analyzed; An adaptive differentiable short-time Fourier transform model is constructed. The model adopts a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted according to the time position and frequency index. A multi-objective composite loss function is constructed, which includes Rényi entropy loss, kurtosis loss, and regularization constraints. The window length and window function time shift are iteratively optimized based on the multi-objective composite loss function. The optimized window length and window function time shift are substituted into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; Seismic attenuation attributes are extracted based on the time-frequency spectrum, and reservoir hydrocarbon-bearing properties are characterized based on the seismic attenuation attributes.

2. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 1, characterized in that, The construction of the adaptive differentiable short-time Fourier transform model includes: For the single-channel seismic signal in the seismic data to be analyzed, corresponding window lengths are set at different time locations and at different frequency indices, and the window function time shift is set; A Gaussian window function that is differentiable with respect to the independent variable and the parameters is used as the window function, wherein the time shift of the window function is used to characterize the sliding length of the window function; For single-channel seismic signals The adaptive differentiable short-time Fourier transform is: in, This represents the time shift of the window function. Indicates the first The time position, the Window length at each frequency index This represents a window function that is differentiable with respect to the independent variable and the parameter.

3. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 1, characterized in that, The construction of the multi-objective composite loss function includes: Set the weight coefficients for Rényi entropy loss, kurtosis loss, and regularization constraint respectively; The multi-objective composite loss function is constructed based on the Rényi entropy loss, kurtosis loss, regularization constraint, and corresponding weight coefficients.

4. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 3, characterized in that, The determination of the Rényi entropy loss includes: The normalized energy distribution is determined based on the time spectrum output by the adaptive differentiable short-time Fourier transform model. The Rényi entropy loss is determined based on the normalized energy distribution and the entropy order.

5. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 3, characterized in that, The kurtosis loss is determined based on the kurtosis of the time spectrum at different time indices and frequency indices, and the negative average kurtosis is used as the kurtosis loss. The regularization constraints include regularization terms determined based on the sum of the absolute values ​​of the model parameters and regularization terms determined based on the sum of the squares of the model parameters.

6. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 1, characterized in that, The iterative optimization of the window length and window function time shift based on the multi-objective composite loss function includes: The gradients of the multi-objective composite loss function with respect to the window length and the time shift of the window function are determined based on the chain rule. Based on the gradient, the AdamW optimizer is used to iteratively update the window length and window function time shift until the multi-objective composite loss function converges.

7. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 6, characterized in that, The AdamW optimizer is trained 500 times, the learning rate is initialized to 0.001, the window length is initialized to 64-128 sampling points, and the window function time shift is initialized to 1 / 8-1 / 4 of the window length.

8. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 1, characterized in that, The calculation of the time spectrum of the seismic data to be analyzed includes: Read the time variable and trace number of the seismic data to be analyzed; Substitute the time-shifted window length and window function after iterative optimization convergence into the adaptive differentiable short-time Fourier transform model; Based on the time variable and the number of traces, the seismic data to be analyzed is calculated trace by trace to obtain the time spectrum.

9. The seismic reservoir characterization method based on adaptive differentiable short-time Fourier transform according to claim 1, characterized in that, The step of extracting seismic attenuation attributes based on the time-frequency spectrum and characterizing reservoir hydrocarbon content based on the seismic attenuation attributes includes: Based on the spectral analysis results of the earthquake data to be analyzed, low-frequency and high-frequency frequencies that meet the 3dB bandwidth criterion are selected; Based on the time spectrum, extract the low frequency and the low frequency respectively. and high frequency Corresponding low-frequency components and high-frequency components ; According to the low-frequency components and high-frequency components The seismic attenuation properties at the target layer are calculated as follows: in, Represents the time spectrum. As a Taoist name, For time variables, Scaling factor The target layer is used as a reference layer; a scaling factor determined based on the reference layer is used to eliminate the inherent energy differences of the target layer. The distribution range of oil and gas reservoirs is determined based on the strong attenuation anomaly zones in the seismic attenuation attributes.

10. A seismic reservoir characterization system based on adaptive differentiable short-time Fourier transform, characterized in that, include: The data module is used to acquire the earthquake data to be analyzed. The model module is used to construct an adaptive differentiable short-time Fourier transform model. The model adopts a window function that is differentiable with respect to the independent variables and parameters, and sets the window length and window function time shift as optimizable parameters that are jointly and adaptively adjusted according to the time position and frequency index. The parameter module is used to construct a multi-objective composite loss function that includes Rényi entropy loss, kurtosis loss and regularization constraints, and to iteratively optimize the window length and window function time shift based on the multi-objective composite loss function; The calculation module is used to substitute the optimized window length and window function time shift into the adaptive differentiable short-time Fourier transform model to calculate the time spectrum of the seismic data to be analyzed; The characterization module is used to extract seismic attenuation attributes based on the time spectrum and characterize the reservoir's hydrocarbon content based on the seismic attenuation attributes.

Images