Frequency dispersion inversion method and apparatus, and computing device

Abstract

The invention discloses a frequency dispersion inversion method and apparatus, and a computing device. The method comprises the steps of obtaining target frequency dispersion data obtained based on acoustic logging data; inputting any group of candidate stratum parameter values in the candidate parameter space into an inversion model obtained by pre-training, obtaining candidate theoretical frequency dispersion data corresponding to the group of candidate stratum parameter values output by the inversion model, and calculating a fitting residual error corresponding to the group of candidate stratum parameter values in combination with the candidate theoretical frequency dispersion data and the target frequency dispersion data; identifying a target stratum parameter value corresponding to the target frequency dispersion data from the candidate stratum parameter values contained in the candidate parameter space according to fitting residual errors corresponding to the candidate stratum parameter values in the candidate parameter space; and inputting the target stratum parameter value into the inversion model, and obtaining target theoretical frequency dispersion data corresponding to the target frequency dispersion data output by the inversion model. By adopting the scheme, the frequency dispersion inversion efficiency can be greatly improved.

Description

Technical Field

[0001] The present application relates to the field of exploration technology, and in particular to a dispersion inversion method, apparatus, computing device, computer storage medium and computer program product. Background Art

[0002] Dispersion inversion is an important method for processing sonic logging data. Specifically, it obtains a known dispersion curve through sonic logging data, inversely infers the known dispersion curve to obtain a complete dispersion curve, and obtains corresponding formation parameters and so on.

[0003] The commonly used dispersion inversion method in the prior art is an inversion method based on the wellbore acoustic field equation. This inversion method establishes an inversion model through the wellbore acoustic field equation, and uses optimization algorithms such as simulated annealing and particle swarm optimization to search the model solution space, thereby obtaining the physical properties of the downhole formation.

[0004] However, the inventors discovered during the implementation process that the prior art has the following defects: the calculation complexity of the borehole acoustic field equation in the prior art is very high, and the parameter optimization process is time-consuming and lengthy; and the solution convergence speed of optimization algorithms such as simulated annealing and particle swarm optimization is slow, which leads to low efficiency of dispersion inversion using the prior art. Summary of the invention

[0005] In view of the above problems, the present application is proposed to provide a dispersion inversion method, apparatus, computing device, computer storage medium and computer program product that overcome the above problems or at least partially solve the above problems.

[0006] According to a first aspect of the present application, a dispersion inversion method is provided, comprising:

[0007] Obtain target dispersion data based on sonic logging data;

[0008] Input any group of candidate formation parameter values ​​in the candidate parameter space into the pre-trained inversion model, obtain candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculate the fitting residual corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data;

[0009] identifying, according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space, a target formation parameter value corresponding to the target dispersion data from the candidate formation parameter values ​​included in the candidate parameter space;

[0010] The target formation parameter value is input into the inversion model, and target theoretical dispersion data corresponding to the target dispersion data output by the inversion model is obtained.

[0011] In an optional embodiment, the step of inputting any group of candidate formation parameter values ​​in the candidate parameter space into a pre-trained inversion model, obtaining candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculating fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; and identifying target formation parameter values ​​corresponding to the target dispersion data from the candidate formation parameter values ​​contained in the candidate parameter space according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space, comprising:

[0012] Dividing the candidate parameter space into a plurality of primary grids at a first resolution;

[0013] For any first-level grid, the candidate formation parameter value of the first-level grid is input into the pre-trained inversion model, the candidate theoretical dispersion data corresponding to the first-level grid output by the inversion model is obtained, and the fitting residual corresponding to the first-level grid is calculated by combining the candidate theoretical dispersion data with the target dispersion data;

[0014] Determine a target primary grid with the smallest fitting residual, and divide the target primary grid into a plurality of secondary grids using a second resolution;

[0015] For any secondary grid, the candidate formation parameter value of the secondary grid is input into the pre-trained inversion model, the candidate theoretical dispersion data corresponding to the secondary grid output by the inversion model is obtained, and the fitting residual corresponding to the secondary grid is calculated by combining the candidate theoretical dispersion data with the target dispersion data;

[0016] The candidate formation parameter value of the target secondary grid with the smallest fitting residual is taken as the target formation parameter value.

[0017] In an optional implementation, the inversion model is obtained by training in the following manner:

[0018] generating multiple sets of sample formation parameter values;

[0019] The theoretical dispersion data of samples corresponding to the formation parameter values ​​of each group of samples are calculated using the borehole acoustic field algorithm;

[0020] Generate a sample set according to the multiple groups of sample formation parameter values ​​and corresponding sample theoretical dispersion data;

[0021] The inversion model is trained using the sample set.

[0022] In an optional implementation, the loss function used by the inversion model during training is composed of the L1 norm of the inversion model output data and the sample label and a monotonicity constraint function.

[0023] In an optional implementation, the acquiring target dispersion data based on the sonic logging data includes:

[0024] Based on the acoustic logging data, waveform array data of any logging depth is obtained;

[0025] After processing the waveform array data using a spectrum correlation function, original dispersion data is obtained;

[0026] The target dispersion data is generated by processing the original dispersion data based on a density clustering algorithm.

[0027] In an optional implementation, after acquiring the target theoretical dispersion data corresponding to the target dispersion data output by the inversion model, the method further includes:

[0028] Calculate the fitting residual between the target dispersion data and the target theoretical dispersion data;

[0029] A quality control parameter is calculated based on the fitting residual and the target dispersion data.

[0030] According to a second aspect of the present application, a dispersion inversion device is provided, comprising:

[0031] An acquisition module, used to acquire target dispersion data based on acoustic logging data;

[0032] A parameter determination module, for inputting any group of candidate formation parameter values ​​in the candidate parameter space into a pre-trained inversion model, obtaining candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculating fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; and identifying target formation parameter values ​​corresponding to the target dispersion data from the candidate formation parameter values ​​contained in the candidate parameter space according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space;

[0033] A theoretical dispersion determination module, used for inputting the target formation parameter value into the inversion model, and obtaining target theoretical dispersion data corresponding to the target dispersion data output by the inversion model;

[0034] Inversion model, used to output theoretical dispersion data.

[0035] In an optional embodiment, the parameter determination module is used to: divide the candidate parameter space into multiple first-level grids with a first resolution; for any first-level grid, input the candidate formation parameter value of the first-level grid into the inversion model obtained by pre-training, obtain the candidate theoretical dispersion data corresponding to the first-level grid output by the inversion model, and calculate the fitting residual corresponding to the first-level grid by combining the candidate theoretical dispersion data with the target dispersion data; determine the target first-level grid with the smallest fitting residual, and divide the target first-level grid into multiple second-level grids with a second resolution; for any second-level grid, input the candidate formation parameter value of the second-level grid into the inversion model obtained by pre-training, obtain the candidate theoretical dispersion data corresponding to the second-level grid output by the inversion model, and calculate the fitting residual corresponding to the second-level grid by combining the candidate theoretical dispersion data with the target dispersion data; and use the candidate formation parameter value of the target second-level grid with the smallest fitting residual as the target formation parameter value.

[0036] In an optional embodiment, the device further comprises: a model training module for generating multiple sets of sample formation parameter values;

[0037] The theoretical dispersion data of samples corresponding to the formation parameter values ​​of each group of samples are calculated using the borehole acoustic field algorithm;

[0038] Generate a sample set according to the multiple groups of sample formation parameter values ​​and corresponding sample theoretical dispersion data;

[0039] The inversion model is trained using the sample set.

[0040] In an optional implementation, the loss function used by the inversion model during training is composed of the L1 norm of the inversion model output data and the sample label and a monotonicity constraint function.

[0041] In an optional implementation, the acquisition module is used to: obtain waveform array data of any logging depth based on the acoustic logging data;

[0042] After processing the waveform array data using a spectrum correlation function, original dispersion data is obtained;

[0043] The target dispersion data is generated by processing the original dispersion data based on a density clustering algorithm.

[0044] In an optional embodiment, the device includes: a quality control module, which is used to calculate the fitting residual between the target dispersion data and the target theoretical dispersion data; and calculate the quality control parameter based on the fitting residual and the target dispersion data.

[0045] According to a third aspect of the present application, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0046] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned dispersion inversion method.

[0047] According to a fourth aspect of the present application, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned dispersion inversion method.

[0048] According to a fifth aspect of the present application, a computer program product is provided, comprising at least one executable instruction, wherein the executable instruction enables a processor to perform operations corresponding to the above-mentioned dispersion inversion method.

[0049] The dispersion inversion method, apparatus, computing device, computer storage medium and computer program product provided in the present application utilize an inversion model to perform dispersion inversion on target dispersion data, thereby obtaining target theoretical dispersion data and target formation parameters. The inversion process can perform dispersion inversion without relying on prior parameters, and there is no need to perform complex solution space searches and solve wellbore acoustic field equations, which greatly improves the efficiency of dispersion inversion and is suitable for large-scale application and implementation.

[0050] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0052] Figure 1 A schematic diagram of a process flow of a dispersion inversion method provided in Example 1 of the present application is shown;

[0053] Figure 2 A schematic flow chart of a method for generating target dispersion data provided in Embodiment 1 of the present application is shown;

[0054] Figure 3 A schematic diagram of a waveform array provided in Embodiment 1 of the present application is shown;

[0055] Figure 4 A schematic diagram of original dispersion data provided in Example 1 of the present application is shown;

[0056] Figure 5 A schematic diagram of clustering of original dispersion data provided in Example 1 of the present application is shown;

[0057] Figure 6 A schematic diagram of target dispersion data provided in Example 1 of the present application is shown;

[0058] Figure 7 A schematic diagram of target theoretical dispersion data provided in Example 1 of the present application is shown;

[0059] Figure 8 A schematic flow chart of a method for determining a target formation parameter value provided in the second embodiment of the present application is shown;

[0060] Fig. 9 A schematic flow chart of a method for determining a target formation parameter value provided in Embodiment 3 of the present application is shown;

[0061] Fig.10 A schematic flow chart of an inversion model training method provided in Embodiment 4 of the present application is shown;

[0062] Fig.11 A schematic structural diagram of an inversion model provided in Embodiment 4 of the present application is shown;

[0063] Fig.12 A schematic diagram showing the change of the loss function of an inversion model provided in the fourth embodiment of the present application during the training process is shown;

[0064] Fig.13 A schematic flow chart of a dispersion inversion method provided in Embodiment 5 of the present application is shown;

[0065] Fig.14 A schematic diagram of actual data dispersion inversion effect provided in Example 5 of the present application is shown;

[0066] Fig.15 A schematic structural diagram of a dispersion inversion device provided in Embodiment 6 of the present application is shown;

[0067] Fig.16 A structural schematic diagram of a computing device provided in Embodiment 7 of the present application is shown. DETAILED DESCRIPTION

[0068] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0069] Embodiment 1

[0070] Figure 1 FIG. 1 is a flow chart of a dispersion inversion method provided in Example 1 of the present application. Specifically, Figure 1 As shown, the method comprises the following steps:

[0071] Step S101, obtaining target dispersion data based on acoustic logging data.

[0072] The target dispersion data is the data to be subjected to dispersion inversion, and is also the known dispersion data on which dispersion inversion is based. It can be understood as a part of the dispersion data in the complete dispersion data. The complete dispersion data and the corresponding formation parameters can be obtained by inverting the target dispersion data.

[0073] Specifically, the target dispersion data is obtained from the acoustic logging data. The acoustic logging data is the waveform data obtained by the acoustic logging instrument to detect the preset formation in the logging, and the target dispersion data is obtained by corresponding processing of the waveform data. Among them, the embodiment of the application does not limit the specific processing method for obtaining the target dispersion data based on the acoustic logging data, and those skilled in the art can select the corresponding processing method according to the actual situation. For example, Figure 2 Steps S1011-S1013 are shown to obtain target dispersion data.

[0074] S1011, based on the acoustic logging data, waveform array data of any logging depth is obtained.

[0075] By using the acoustic logging instrument to measure the preset layer section in the logging, the waveform array data corresponding to different logging depths can be obtained, that is, each logging depth will correspond to the corresponding waveform array data. For example, the waveform array data at a certain logging depth is as follows: Figure 3 As shown. Figure 3 It can be seen from the figure that the waveform array data at a logging depth is composed of waveform data obtained by detection by different receivers.

[0076] S1012, after processing the waveform array data using the spectrum correlation function, original dispersion data is obtained.

[0077] This step is used to convert the waveform array data into dispersion data. The dispersion data directly converted from the waveform array data is referred to as raw dispersion data in this application. The dispersion data described in this application is usually frequency-slowness domain data, or frequency-velocity domain data, which is not limited in this application.

[0078] Specifically, the spectrum correlation function can be used to realize the conversion of waveform array data to original dispersion data.

[0079]

[0080] Where ρ(ω,s) represents the correlation between frequency and slowness; ω represents the circular frequency; s represents the slowness; N represents the number of receivers of the logging instrument; d represents the receiver spacing; X n (ω) represents the spectrum of the waveform collected by the nth receiver; Represents X n The complex conjugate of (ω). Thus, the maximum point of the coherence function corresponding to each frequency constitutes the original dispersion signal. The original dispersion data of a certain logging depth obtained in this step can be Figure 4 shown.

[0081] S1013, processing the original dispersion data based on a density clustering algorithm to generate target dispersion data.

[0082] like Figure 4 As shown, there will be some noise in the original dispersion data. In order to improve the subsequent dispersion inversion accuracy, this step can further reduce the noise of the original dispersion data to remove the noise in the target dispersion data. The target dispersion data specifically corresponds to the logging depth, that is, the target dispersion data corresponding to any logging depth can be obtained through this implementation process.

[0083] This embodiment does not limit the specific noise reduction processing method, but in order to improve the noise reduction effect, this embodiment can preferably use a density clustering algorithm (such as the DBSCAN algorithm, etc.) to process the original dispersion data. During the processing, the data set is divided according to the distribution density of the data points, and high-density points are divided into different clusters, and low-density points are divided into noise. Figure 5 As shown in , the data points can be divided into corresponding data sets or data clusters through clustering algorithms. Further, valid data clusters can be screened out from the data clusters, that is, valid signals can be screened out. For example, the frequency and speed limit ranges can be set to eliminate the noise and the clusters containing data points less than the threshold, and the remaining clusters constitute the target dispersion signal. Figure 5 The cluster in the trapezoidal area is the selected valid data cluster. After removing other data points outside the valid data cluster, we get Figure 6The target dispersion data schematic diagram shown in FIG. 1 shows that the present embodiment can effectively filter out the noise in the original dispersion data, thus providing a basis for improving the accuracy of subsequent dispersion inversion.

[0084] Step S102, input any group of candidate formation parameter values ​​in the candidate parameter space into the pre-trained inversion model, obtain the candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculate the fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data and the target dispersion data.

[0085] The present application constructs and trains an inversion model based on a machine learning algorithm in advance. The inversion model can predict theoretical dispersion data that matches the input formation parameters, and the theoretical dispersion data is complete dispersion data. Thus, the present application uses the inversion model to perform dispersion inversion on the target dispersion data, and determines the target formation parameter value and target theoretical dispersion data corresponding to the target dispersion data. Among them, the target formation parameter value is the specific value of the formation parameter determined by inverting the target dispersion data, and the target theoretical dispersion data is the complete dispersion data obtained by performing dispersion inversion on the target dispersion data.

[0086] Specifically, a candidate parameter space is constructed, which is a set of all possible values ​​of the target formation parameters, and can also be understood as a value range of the target formation parameters. Wherein, the target formation parameters can be one or more, and the candidate parameter space is a value space composed of the value ranges of each target formation parameter, and the number of dimensions of the candidate parameter space matches the number of formation parameters. For example, if the target formation parameters include well diameter, well fluid sound velocity, and formation shear wave sound velocity, the value ranges of the well diameter, well fluid sound velocity, and formation shear wave sound velocity are determined respectively, and then a three-dimensional candidate parameter space is constructed.

[0087] Optionally, since there are many actual formation parameters, in order to simplify the implementation process of this method and improve the inversion efficiency, the sensitivity analysis of the formation parameters can be performed on the theoretical dispersion curve first, and then they can be sorted in order from high to low sensitivity, and the formation parameters with the top N sensitivities can be selected, and a candidate parameter space can be constructed for the top N formation parameters.

[0088] A point in the candidate parameter space corresponds to a combination of candidate formation parameter values. s _j,v f _k) is the well diameter is R_i, the sound velocity of the fluid in the well is v s _j, and the formation shear wave velocity is v f A set of candidate formation parameter values ​​for _k.

[0089] The candidate formation parameter values ​​are input into the inversion model, and the inversion model outputs candidate theoretical frequency data corresponding to the candidate formation parameter values. The candidate theoretical dispersion data is the complete dispersion data corresponding to the set of candidate formation parameter values.

[0090] The fitting residual of the output candidate theoretical dispersion data and the target dispersion data is further calculated, and the fitting residual characterizes the difference between the candidate theoretical dispersion curve and the target dispersion data. Each set of candidate formation parameter values ​​input to the inversion model can have a corresponding fitting residual. Among them, the embodiment of the present application does not limit the calculation method of the fitting residual, for example, the Euclidean norm of the candidate theoretical dispersion data and the target dispersion data can be used as the corresponding fitting residual, and so on.

[0091] Step S103 , identifying a target formation parameter value corresponding to the target dispersion data from the candidate formation parameter values ​​included in the candidate parameter space according to the fitting residual corresponding to the candidate formation parameter value in the candidate parameter space.

[0092] Since the fitting residuals corresponding to the candidate formation parameter values ​​can reflect the difference between the corresponding candidate theoretical dispersion data and the target dispersion data, the target formation parameter corresponding to the target dispersion data can be determined from each candidate formation parameter value according to the fitting residuals. For example, the candidate formation parameter value with the smallest fitting residual can be used as the target formation parameter value.

[0093] Step S104: input the target formation parameter value into the inversion model, and obtain the target theoretical dispersion data corresponding to the target dispersion data output by the inversion model.

[0094] The determined target formation parameters are provided to the inversion model, and the theoretical dispersion data output by the inversion model is the target theoretical frequency data. Figure 7 As shown, by inverting the target dispersion data, we can get Figure 7 The dispersion curve in is the target theoretical dispersion data.

[0095] It can be seen that the dispersion inversion method provided in the embodiment of the present application uses an inversion model to perform dispersion inversion on the target dispersion data, thereby obtaining target theoretical dispersion data and target formation parameters. The inversion process does not rely on prior parameters to perform dispersion inversion, and there is no need to perform complex solution space search and solve the wellbore acoustic field equation, which greatly improves the efficiency of dispersion inversion and is suitable for large-scale application and implementation.

[0096] Embodiment 2

[0097] Figure 8 A schematic flow chart of a method for determining a target formation parameter value provided in the second embodiment of the present application is shown. The method for determining a target formation parameter value is applied in a dispersion inversion process.

[0098] Specifically, since there are a large number of candidate formation parameter values ​​in the candidate parameter space, if the inversion model is used to process each group of candidate formation parameter values ​​in the candidate parameter space, the system processing capacity will be greatly increased and the processing efficiency will be reduced. In order to solve this technical problem, the following steps are proposed to optimize the target formation parameter value determination process in Example 1.

[0099] like Figure 8 As shown, the method comprises the following steps:

[0100] Step S801, dividing the candidate parameter space into a plurality of first-level grids at a first resolution.

[0101] The resolution is used to determine the granularity of the candidate parameter space. The higher the resolution, the finer the granularity, the smaller the grid size obtained after the division, and the more grids obtained. In this embodiment, the candidate parameter space is first divided into two grids at a first resolution, and the obtained grids are called first-level grids. For example, 1000 grids are obtained by dividing the candidate parameter space into 10*10*10 grids.

[0102] Step S802, for any first-level grid, input the candidate formation parameter value of the first-level grid into the pre-trained inversion model, obtain the candidate theoretical dispersion data corresponding to the first-level grid output by the inversion model, and calculate the fitting residual corresponding to the first-level grid by combining the candidate theoretical dispersion data and the target dispersion data.

[0103] For each primary grid, the candidate formation parameter value of the grid is determined. For example, the candidate formation parameter combination corresponding to the grid center can be used as the candidate formation parameter value of the grid. The candidate formation parameter value of the grid is input into the inversion model, and the theoretical dispersion data output by the inversion model is used as the candidate theoretical dispersion data corresponding to the primary grid.

[0104] And the fitting residual corresponding to each primary grid is calculated. Specifically, the fitting residual between the candidate theoretical dispersion data and the target dispersion data corresponding to the primary grid is used as the fitting residual corresponding to the primary grid.

[0105] Step S803, determining a target primary grid with the smallest fitting residual, and dividing the target primary grid into a plurality of secondary grids using a second resolution.

[0106] The first-level grid with the smallest fitting residual in each first-level grid is used as the target first-level grid. The target first-level grid is further divided using the second resolution, and the second resolution is greater than the first resolution. For example, the second resolution can be N times the first resolution. After the target first-level grid is divided, multiple second-level grids are obtained, and the size of the second-level grid is smaller than the grid.

[0107] Step S804, for any secondary grid, input the candidate formation parameter value of the secondary grid into the pre-trained inversion model, obtain the candidate theoretical dispersion data corresponding to the secondary grid output by the inversion model, and calculate the fitting residual corresponding to the secondary grid by combining the candidate theoretical dispersion data and the target dispersion data.

[0108] For each secondary grid, the candidate formation parameter value of the grid is determined. For example, the candidate formation parameter combination corresponding to the grid center can be used as the candidate formation parameter value of the grid. The candidate formation parameter value of the grid is input into the inversion model, and the theoretical dispersion data output by the inversion model is used as the candidate theoretical dispersion data corresponding to the secondary grid.

[0109] And the fitting residual corresponding to each secondary grid is calculated. Specifically, the fitting residual between the candidate theoretical dispersion data and the target dispersion data corresponding to the secondary grid is used as the fitting residual corresponding to the secondary grid.

[0110] Step S805: taking the candidate formation parameter value of the target secondary grid with the smallest fitting residual as the target formation parameter value.

[0111] The secondary grid with the smallest fitting residual in each secondary grid is taken as the target secondary grid, and the candidate formation parameter value of the target secondary grid is taken as the target formation parameter value.

[0112] It can be seen that the target formation parameter value determination method provided in the embodiment of the present application first performs a coarse-grained division on the candidate parameter space to determine the target primary grid with the smallest fitting residual; then the target primary grid is fine-grainedly divided to determine the target secondary grid with the smallest fitting residual, and then the candidate formation parameter value of the target secondary grid with the smallest fitting residual is used as the target formation parameter value, thereby simplifying the target formation parameter value determination process, reducing the system processing volume, and improving the target formation parameter value determination efficiency.

[0113] Embodiment 3

[0114] Fig. 9 A schematic flow chart of a method for determining a target formation parameter value provided in the third embodiment of the present application is shown. The method for determining a target formation parameter value is applied in a dispersion inversion process.

[0115] Specifically, the target formation parameter value determination method in Example 2 can greatly improve the processing efficiency, but in some scenarios (such as a large candidate parameter space, unreasonable segmentation resolution selection, etc.), the method in Example 2 cannot accurately determine the target formation parameter value. Therefore, the embodiment of the present application further optimizes the solution in Example 2 to ensure the determination accuracy of the target formation parameter value.

[0116] like Fig. 9As shown, the method comprises the following steps:

[0117] Step S901, dividing the candidate parameter space into a plurality of i-level grids at an i-th resolution.

[0118] The initial value of i is 1, the i-th resolution is smaller than the i+1-th resolution, and the i-level grid size is larger than the i+1-level grid size.

[0119] Step S902, for any i-level grid, input the candidate formation parameter value of the i-level grid into the pre-trained inversion model, obtain the candidate theoretical dispersion data corresponding to the i-level grid output by the inversion model, and calculate the fitting residual corresponding to the i-level grid by combining the candidate theoretical dispersion data and the target dispersion data.

[0120] For each i-level grid, determine the candidate formation parameter value of the grid. For example, the candidate formation parameter combination corresponding to the grid center can be used as the candidate formation parameter value of the grid. The candidate formation parameter value of the grid is input into the inversion model, and the theoretical dispersion data output by the inversion model is used as the candidate theoretical dispersion data corresponding to the i-level grid.

[0121] And the fitting residual corresponding to each i-level grid is calculated. Specifically, the fitting residual between the candidate theoretical dispersion data and the target dispersion data corresponding to the i-level grid is used as the fitting residual corresponding to the i-level grid.

[0122] Step S903, determining the target i-level grid with the smallest fitting residual.

[0123] The i-level grid with the smallest fitting residual among all i-level grids is taken as the target i-level grid.

[0124] Step S904, determining whether the fitting residual of the target i-level grid is less than a preset threshold; if so, executing step S905; if not, executing step S906.

[0125] The preset threshold can be flexibly set according to actual accuracy requirements, and this application does not limit it here.

[0126] Step S905: taking the candidate formation parameter value of the target i-level grid as the target formation parameter value.

[0127] If the fitting residual of the target i-level grid is less than the preset threshold, it means that the difference between the candidate theoretical frequency data of the target i-level grid and the target dispersion data is small and meets the accuracy requirement, then the candidate formation parameter value of the target i-level grid is used as the target formation parameter value.

[0128] Step S906: Divide the target i-level grid into multiple i+1-level grids using the i+1th resolution, and i is incremented by 1.

[0129] If the fitting residual of the target i-level grid is greater than or equal to the preset threshold, it indicates that the difference between the candidate theoretical frequency data and the target dispersion data of the target i-level grid is large and cannot meet the accuracy requirement. The target i-level grid is then divided into multiple i+1-level grids using the i+1th resolution, and i is incremented by 1 to execute step S902.

[0130] It can be seen that the target formation parameter value determination method provided in the embodiment of the present application introduces a preset threshold value. If the fitting residual of the target i-level grid is less than the preset threshold value at the current segmentation granularity, the candidate formation parameter value of the target i-level grid is used as the target formation parameter value; if the fitting residual of the target i-level grid is greater than or equal to the preset threshold value at the current segmentation granularity, a higher resolution grid segmentation is performed, thereby not only improving the efficiency of determining the target formation parameter value, but also ensuring that the determination accuracy of the target formation parameter value meets the accuracy requirements.

[0131] Embodiment 4

[0132] Fig.10 The flowchart of an inversion model training method provided in the fourth embodiment of the present application is shown. The inversion model training method is executed before the dispersion data is formally inverted.

[0133] Specifically, Fig.10 As shown, the method comprises the following steps:

[0134] Step S1001, constructing an inversion model.

[0135] The inversion model in the embodiment of the present application is specifically a deep neural network model. Fig.11 As shown, the inversion model includes the first fully connected layer (such as Fig.11 The first FC module from top to bottom), ReLU function, the second fully connected layer (such as Fig.11 The second FC module from top to bottom in ), and 8 residual modules connected in sequence (such as Fig.11 The structure of each residual model can be the same, and each residual module can be composed of two groups of batch normalization layers (such as Fig.11 The residual module is composed of a BN module in ), a ReLU activation function layer and a fully connected layer, and the sum of the input and output can be used as the total output of the residual module. Among them, the width of the first fully connected layer is 100*M, and the output of all subsequent fully connected layers is M, where M is the total number of sampling points of the theoretical dispersion curve, for example, M can be 76, etc.

[0136] It should be understood here that Fig.11The model structure shown is only an optional inversion model structure, which is concise and has high processing accuracy. In the actual implementation process, the specific structure of the inversion model can be flexibly set, and this application does not limit this.

[0137] Step S1002, generating multiple groups of sample formation parameter values.

[0138] In order to simplify the model training process and improve the model training accuracy, the embodiment of the present application pre-determines N formation parameters with high sensitivity to the theoretical dispersion data from a large number of formation parameters. Specifically, a sensitivity analysis is performed on the theoretical dispersion, and the formation parameters are sorted in order of sensitivity from high to low, and the first N formation parameters are selected. For example, the highly sensitive formation parameters are selected: well diameter R, formation shear wave velocity v s , fluid velocity in the well v f . Among them, this application does not limit the sensitivity analysis method.

[0139] A sample parameter space is further constructed, which is the distribution range of the formation parameters used in the training samples, and the sample parameter space can be consistent with the candidate parameter space. For example, the sample parameter space can be shown in Table 1. As can be seen from Table 1, the sample parameter space contains 3 dimensions, each dimension has a corresponding upper limit and lower limit to constrain the parameter value range, and a corresponding step size is set to control the interval between adjacent parameter values ​​in each dimension. Therefore, multiple groups of sample formation parameter values ​​can be constructed based on the sample parameter space, and each group of sample formation parameter values ​​corresponds to a set of specific values ​​of wellbore diameter, fluid sound velocity in the well, and formation shear wave sound velocity.

[0140] Table 1

[0141]

[0142]

[0143] Step S1003, using a borehole acoustic field algorithm to calculate theoretical dispersion data of samples corresponding to formation parameters of each group of samples.

[0144] For any set of sample formation parameters, the set of sample formation parameters is used as the input of the borehole acoustic field equation, and the sample theoretical dispersion data corresponding to the set of sample formation parameters is obtained through the borehole acoustic field algorithm. Among them, each theoretical dispersion curve can be controlled to consist of the velocities corresponding to 76 frequency points equidistantly distributed from 0 to 15kHz, and the part below the cutoff frequency is assigned according to the formation sound velocity. It should be understood here that the solution of the borehole acoustic field equation here is in the training process of the inversion model, and this stage is the preparatory stage before the formal dispersion inversion, so the solution process of the borehole acoustic field equation here does not affect the efficiency of the formal dispersion inversion.

[0145] Step S1004: Generate a sample set according to multiple groups of sample formation parameters and corresponding sample theoretical dispersion data.

[0146] A sample set is generated for each set of sample formation parameters and the corresponding sample theoretical dispersion data.

[0147] Step S1005: training the inversion model using the sample set.

[0148] Specifically, the sample set can be divided into a training set and a test set according to a preset ratio (such as 95:5), and the samples in the training set are used to train the inversion model, and the samples in the test set are used to test the effect of the inversion model. Among them, the sample formation parameters are used as model inputs, and the corresponding sample theoretical dispersion data can be used as labels.

[0149] In the specific training process, the Adam algorithm can be used to iterate the model parameters, the initial learning rate is set to 0.01, the cosine scheduling is used to decay the learning rate, the batch size can be 64, the number of training rounds can be 200 rounds, etc. The relevant parameters in the training process can be flexibly set, and this application does not limit this.

[0150] In order to improve the prediction accuracy of the inversion model, the loss function used in the inversion model of the embodiment of the present application during the training process is composed of the L1 norm of the inversion model output data and the sample label and the monotonicity constraint function. Among them, the L1 norm of the inversion model output data and the sample label can reflect the model prediction error. In addition, the inventors found that the theoretical dispersion curve is usually monotonically decreasing, so adding a monotonicity constraint function to the loss function can ensure that the theoretical dispersion curve output by the inversion model is a single-point decreasing, thereby improving the dispersion inversion accuracy. Among them, the loss function can be specifically shown in Formula 2:

[0151]

[0152] Among them, v nn Indicates the inversion model output; v synth Represents the sample theoretical dispersion data, that is, the sample label; Loss(v nn , v synth ) represents the loss function; ||v nn -v synth ||1 represents the L1 norm of the inversion model output data and sample labels; represents the monotonicity constraint function.

[0153] By testing the inversion model with samples in the test set, it can be determined that the theoretical dispersion curves output by the inversion model in soft and hard formations are consistent with the label V synth Highly consistent. In addition, Fig.12 The loss changes of the inversion model during training are given, from Fig.12It can be seen that the training set loss and the test set loss keep decreasing synchronously, and there is no underfitting phenomenon. Moreover, in the actual implementation process, this application uses 4471 samples in the test set as model input and uses Intel Xeon W-2245 CPU to calculate the output. When the input batch size is 1, the whole process takes 2822.76ms, and the average single calculation takes 0.631ms. When the input batch size is 4471, the calculation process takes 84.96ms, and the average single calculation takes 0.019ms. Therefore, the inversion model of this application has a higher processing efficiency.

[0154] It can be seen that the inversion model training method provided in the embodiment of the present application takes the formation parameters of each group of samples and the theoretical dispersion data of the samples corresponding to the formation parameters of each group of samples calculated using the wellbore acoustic field algorithm as a sample set, and uses the sample set to train the inversion model, thereby improving the quality of sample data and enhancing the processing accuracy of the inversion model.

[0155] Embodiment 5

[0156] Fig.13 FIG. 5 shows a flow chart of a dispersion inversion method provided in Embodiment 5 of the present application. Fig.13 As shown, the method comprises the following steps:

[0157] Step S1301, obtaining target dispersion data based on sonic logging data.

[0158] Step S1302: determining target formation parameter values ​​and target theoretical dispersion data based on the inversion model.

[0159] The specific implementation process of steps S1301 and S1302 may refer to the description in other embodiments, and will not be described in detail in this embodiment.

[0160] Step S1303, calculating the fitting residual between the target dispersion data and the target theoretical dispersion data, and calculating the quality control parameter based on the fitting residual and the target dispersion data.

[0161] In order to accurately grasp the quality of dispersion inversion, the present embodiment also introduces a quality control parameter for determining the quality of dispersion inversion. The target dispersion data and the target theoretical frequency data correspond to the logging depth, and thus the fitting residual of the target dispersion data and the corresponding target theoretical dispersion data can be expressed as the fitting residual of the corresponding logging depth.

[0162] Specifically, the quality control parameter is determined based on the fitting residual of the target dispersion data and the target theoretical dispersion data and the target dispersion data. For example, the quality control parameter of any logging depth can be determined by the following formula 3:

[0163]

[0164] Among them, QC(i) represents the quality control parameter of the logging depth indexed as i, which can also be called the quality control parameter; DispPts represents the set of dispersion points contained in the target dispersion data of each depth point in the entire well section; max(DispPts) is the maximum value in DispPts, Res is the set of fitting residuals for each depth point in the entire well section; max(Res) is the maximum value in Res.

[0165] Fig.14 The figure shows the dispersion inversion effect obtained by processing the dipole acoustic logging data of a 900m long well section using the dispersion inversion method of the present application. Fig.14 In the figure, Waveform represents waveform data, ArrTime represents arrival time, Vs represents the formation shear wave velocity, Dispersion represents the projection of flexural wave dispersion on the slowness axis, and QC represents the quality control parameter. It can be seen that the dispersion inversion of this application is of high quality and has a good inversion effect.

[0166] In an optional implementation, the logging depth that meets the quality requirements (such as greater than the preset quality threshold) and the logging depth that does not meet the quality requirements (such as less than or equal to the preset quality threshold) can be distinguished according to the size relationship between the quality control parameter of the logging depth and the preset quality threshold. For the logging depth that does not meet the quality requirements, corresponding correction processing can be performed, for example, the target dispersion data of the logging depth can be regenerated by adjusting the clustering algorithm parameters, and the target theoretical dispersion data of the target dispersion data can be re-determined.

[0167] It can be seen that the dispersion inversion method provided in the embodiment of the present application can generate quality control parameters corresponding to the logging depth. The quality control parameters can accurately determine the inversion quality of the corresponding logging depth, which is convenient for subsequent processing based on the inversion quality.

[0168] Embodiment 6

[0169] Fig.15 FIG. 1 shows a schematic diagram of the structure of a dispersion inversion device provided in Example 6 of the present application. Fig.15 As shown, the device 1500 includes: an acquisition module 1510 , a parameter determination module 1520 , a theoretical dispersion determination module 1530 , and an inversion model 1540 .

[0170] An acquisition module 1510 is used to acquire target dispersion data based on acoustic logging data;

[0171] The parameter determination module 1520 is used to input any group of candidate formation parameter values ​​in the candidate parameter space into the pre-trained inversion model, obtain the candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculate the fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; and identify the target formation parameter value corresponding to the target dispersion data from the candidate formation parameter values ​​contained in the candidate parameter space according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space;

[0172] Theoretical dispersion determination module 1530 is used to input the target formation parameter value into the inversion model, and obtain target theoretical dispersion data corresponding to the target dispersion data output by the inversion model;

[0173] The inversion model 1540 is used to output theoretical dispersion data.

[0174] In an optional embodiment, the parameter determination module 1520 is used to: divide the candidate parameter space into multiple first-level grids with a first resolution; for any first-level grid, input the candidate formation parameter value of the first-level grid into the inversion model obtained by pre-training, obtain the candidate theoretical dispersion data corresponding to the first-level grid output by the inversion model, and calculate the fitting residual corresponding to the first-level grid by combining the candidate theoretical dispersion data with the target dispersion data; determine the target first-level grid with the smallest fitting residual, and divide the target first-level grid into multiple second-level grids with a second resolution; for any second-level grid, input the candidate formation parameter value of the second-level grid into the inversion model obtained by pre-training, obtain the candidate theoretical dispersion data corresponding to the second-level grid output by the inversion model, and calculate the fitting residual corresponding to the second-level grid by combining the candidate theoretical dispersion data with the target dispersion data; and use the candidate formation parameter value of the target second-level grid with the smallest fitting residual as the target formation parameter value.

[0175] In an optional embodiment, the device further comprises: a model training module (not shown in the figure), which is used to generate multiple groups of sample formation parameter values;

[0176] The theoretical dispersion data of samples corresponding to the formation parameter values ​​of each group of samples are calculated using the borehole acoustic field algorithm;

[0177] Generate a sample set according to the multiple groups of sample formation parameter values ​​and corresponding sample theoretical dispersion data;

[0178] The inversion model is trained using the sample set.

[0179] In an optional implementation, the loss function used by the inversion model during training is composed of the L1 norm of the inversion model output data and the sample label and a monotonicity constraint function.

[0180] In an optional implementation, the acquisition module 1510 is used to: obtain waveform array data of any logging depth based on the acoustic logging data;

[0181] After processing the waveform array data using a spectrum correlation function, original dispersion data is obtained;

[0182] The target dispersion data is generated by processing the original dispersion data based on a density clustering algorithm.

[0183] In an optional embodiment, the device includes: a quality control module (not shown in the figure), which is used to calculate the fitting residual between target dispersion data and target theoretical dispersion data; and calculate quality control parameters based on the fitting residual and the target dispersion data.

[0184] It can be seen that the dispersion inversion device provided in the embodiment of the present application uses an inversion model to perform dispersion inversion on the target dispersion data, thereby obtaining target theoretical dispersion data and target formation parameters. The inversion process does not rely on prior parameters to perform dispersion inversion, and there is no need to perform complex solution space search and solve the wellbore acoustic field equation, which greatly improves the efficiency of dispersion inversion and is suitable for large-scale application and implementation.

[0185] Embodiment 7

[0186] Fig.16 The schematic diagram of the structure of a computing device provided in Embodiment 7 of the present application is shown. The specific embodiments of the present application do not limit the specific implementation of the computing device.

[0187] like Fig.16 As shown, the computing device may include: a processor (processor) 1602 , a communication interface (Communications Interface) 1604 , a memory (memory) 1606 , and a communication bus 1608 .

[0188] The processor 1602, the communication interface 1604, and the memory 1606 communicate with each other via a communication bus 1608. The communication interface 1604 is used to communicate with other devices such as a client or other server network elements. The processor 1602 is used to execute a program 1610, which can specifically execute the relevant steps in the above-mentioned embodiment of the dispersion inversion method for computing a device.

[0189] Specifically, the program 1610 may include program codes, which include computer operation instructions.

[0190] The processor 1602 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0191] The memory 1606 is used to store the program 1610. The memory 1606 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory. The program 1610 may be specifically used to enable the processor 1602 to perform the operations in the above method embodiment.

[0192] Embodiment 8

[0193] An embodiment of the present application provides a non-volatile computer storage medium, which stores at least one executable instruction or computer program, and the executable instruction or computer program can enable a processor to perform operations corresponding to the dispersion inversion method in any of the above method embodiments.

[0194] Embodiment 9

[0195] An embodiment of the present application provides a computer program product, which includes at least one executable instruction or computer program, and the executable instruction or computer program can enable a processor to perform operations corresponding to the dispersion inversion method in any of the above method embodiments.

[0196] In summary, according to the computing device, computer storage medium and computer program product provided in this embodiment, an inversion model is used to perform dispersion inversion on the target dispersion data, so as to obtain target theoretical dispersion data and target formation parameters. The inversion process can perform dispersion inversion without relying on prior parameters, and there is no need to perform complex solution space search and solve the wellbore acoustic field equation, which greatly improves the efficiency of dispersion inversion and is suitable for large-scale application and implementation.

[0197] The algorithm or display provided here are not inherently related to any specific computer, virtual system or other equipment. Various general systems can also be used together with the teaching based on this. According to the above description, it is obvious to construct the structure required for this type of system. In addition, the present application embodiment is not directed to any specific programming language yet. It should be understood that various programming languages ​​can be utilized to realize the content of the present application described here, and the above description of specific languages ​​is to disclose the best mode of implementation of the present application.

[0198] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0199] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the embodiments of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the claims below, the inventive aspects are less than all the features of the single embodiment disclosed above. Therefore, the claims following the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0200] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0201] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0202] The various component embodiments of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components according to the embodiments of the present application. The application can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing a part or all of the methods described herein. Such a program implementing the present application can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0203] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be constructed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A dispersion inversion method, characterized in that: include: Obtain target dispersion data based on sonic logging data; Input any group of candidate formation parameter values ​​in the candidate parameter space into the pre-trained inversion model, obtain candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculate the fitting residual corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; identifying, according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space, a target formation parameter value corresponding to the target dispersion data from the candidate formation parameter values ​​included in the candidate parameter space; The target formation parameter value is input into the inversion model, and target theoretical dispersion data corresponding to the target dispersion data output by the inversion model is obtained.

2. The method according to claim 1, characterized in that The step of inputting any group of candidate formation parameter values ​​in the candidate parameter space into the pre-trained inversion model, obtaining candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculating the fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; and identifying the target formation parameter value corresponding to the target dispersion data from the candidate formation parameter values ​​contained in the candidate parameter space according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space includes: Dividing the candidate parameter space into a plurality of primary grids at a first resolution; For any first-level grid, the candidate formation parameter value of the first-level grid is input into the pre-trained inversion model, the candidate theoretical dispersion data corresponding to the first-level grid output by the inversion model is obtained, and the fitting residual corresponding to the first-level grid is calculated by combining the candidate theoretical dispersion data with the target dispersion data; Determine a target primary grid with the smallest fitting residual, and divide the target primary grid into a plurality of secondary grids using a second resolution; For any secondary grid, the candidate formation parameter value of the secondary grid is input into the pre-trained inversion model, the candidate theoretical dispersion data corresponding to the secondary grid output by the inversion model is obtained, and the fitting residual corresponding to the secondary grid is calculated by combining the candidate theoretical dispersion data with the target dispersion data; The candidate formation parameter value of the target secondary grid with the smallest fitting residual is taken as the target formation parameter value.

3. The method according to claim 1, characterized in that The inversion model is trained in the following way: generating multiple sets of sample formation parameter values; The theoretical dispersion data of samples corresponding to the formation parameter values ​​of each group of samples are calculated using the borehole acoustic field algorithm; Generate a sample set according to the multiple groups of sample formation parameter values ​​and corresponding sample theoretical dispersion data; The inversion model is trained using the sample set.

4. The method according to claim 3, characterized in that The loss function used by the inversion model during the training process is composed of the L1 norm of the inversion model output data and the sample label and a monotonicity constraint function.

5. The method according to any one of claims 1 to 4, characterized in that The obtaining of target dispersion data based on acoustic logging data comprises: Based on the acoustic logging data, waveform array data of any logging depth is obtained; After processing the waveform array data using a spectrum correlation function, original dispersion data is obtained; The target dispersion data is generated by processing the original dispersion data based on a density clustering algorithm.

6. The method according to any one of claims 1 to 4, characterized in that After obtaining the target theoretical dispersion data corresponding to the target dispersion data output by the inversion model, the method further includes: Calculate the fitting residual between the target dispersion data and the target theoretical dispersion data; A quality control parameter is calculated based on the fitting residual and the target dispersion data.

7. A dispersion inversion device, characterized in that: include: An acquisition module, used to acquire target dispersion data based on acoustic logging data; A parameter determination module, for inputting any group of candidate formation parameter values ​​in the candidate parameter space into a pre-trained inversion model, obtaining candidate theoretical dispersion data corresponding to the group of candidate formation parameter values ​​output by the inversion model, and calculating fitting residuals corresponding to the group of candidate formation parameter values ​​by combining the candidate theoretical dispersion data with the target dispersion data; and identifying target formation parameter values ​​corresponding to the target dispersion data from the candidate formation parameter values ​​contained in the candidate parameter space according to the fitting residuals corresponding to the candidate formation parameter values ​​in the candidate parameter space; A theoretical dispersion determination module, used for inputting the target formation parameter value into the inversion model, and obtaining target theoretical dispersion data corresponding to the target dispersion data output by the inversion model; Inversion model, used to output theoretical dispersion data.

8. A computing device, characterized in that include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the dispersion inversion method according to any one of claims 1 to 6.

9. A computer storage medium, characterized in that: The storage medium stores at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the dispersion inversion method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the dispersion inversion method according to any one of claims 1 to 6.

Images