Method and apparatus for well logging environment analysis using convolution kernel upscaling transformation and multidimensional feature threshold

By employing convolution kernel upscaling transformation and multidimensional feature threshold analysis, the problem of identifying environmental factors in horizontal and highly deviated wells was solved, enabling visualization and quantitative calculation of logging influencing factors, and improving the accuracy of logging data analysis and the reliability of interpretation.

CN123115772APending Publication Date: 2026-11-17CHINA OILFIELD SERVICES LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611562799.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to accurately identify and quantify the impact of various environmental factors on logging-while-drilling electromagnetic resistivity in horizontal and highly deviated wells, resulting in high uncertainty in logging interpretation. Existing methods also struggle to achieve effective correction and inversion in complex environments.

Method used

By employing convolution kernel dimensionality upscaling transformation and multidimensional feature threshold analysis, a formation model parameter library is constructed, well logging response data is simulated, and convolution kernels are used to perform data dimensionality upscaling and feature extraction on well logging curves. An empirical knowledge base with rich levels of influencing factors is established, and matching analysis is performed in conjunction with measured well logging curves to identify the main controlling environmental influencing factors.

Benefits of technology

It enables visualization and quantitative calculation of well logging influencing factors in complex environments, improves the accuracy of well logging data analysis and the reliability of interpretation, reduces the impact of human factors, and provides an efficient environmental impact analysis solution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115772A_ABST
    Figure CN123115772A_ABST
Patent Text Reader

Abstract

This invention discloses a method and apparatus for analyzing well logging environments using convolutional kernel dimensionality upscaling and multidimensional feature thresholds. The method includes: constructing a formation model parameter library; simulating well logging responses to obtain a multi-source distance well logging response dataset; expanding the electromagnetic resistivity curves of each multi-source distance while drilling to obtain a one-dimensional well logging curve dataset and obtaining the influence weights of model parameters; performing a convolutional transformation on the one-dimensional well logging curve dataset to obtain a two-dimensional image; dividing the two-dimensional image into regions to determine the feature point sampling area, extracting the first image feature parameters, calculating the first richness level, and combining it with the influence weights of model parameters to determine the richness level of influencing factors and construct an empirical knowledge base; acquiring the measured target well logging curve of the target area to obtain the second image feature parameters and the second richness level, querying the richness level of influencing factors matched in the empirical knowledge base, and determining the influence weights of target model parameters and the conditions of target model parameters to analyze the well logging environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of oilfield development technology, specifically to a method and apparatus for well logging environment analysis using convolution kernel upscaling transformation and multidimensional feature thresholds. Background Technology

[0002] Electromagnetic resistivity logging (EMR) technology, with its high timeliness and rich logging curves, has been widely used in reservoir exploration and geological steering. With advancements in rotary steerable technology and improvements in oil drilling techniques, EMR logging has become a crucial and essential logging component for horizontal and highly deviated wells, demonstrating a broad market prospect and development potential. However, EMR logging, based on high-frequency signal excitation and induced voltage calibration of resistivity, is significantly affected by the complex formation structure. For example, the propagation of electromagnetic waves in the wellbore, invasion zone, and formation interface can lead to changes in phase and amplitude, resulting in the measured resistivity not accurately reflecting formation characteristics. Furthermore, in actual logging environments, complex geological conditions and multiple interference factors make it difficult to accurately identify environmental interference sources based on real-time uploaded curves. This poses a serious challenge to the interpretation of abnormal responses, the formulation of correction strategies, and the accuracy of geological steering decisions, increasing the uncertainty of post-drilling interpretation and evaluation.

[0003] Currently, existing solutions to the problem of wellbore environment influences logging response in electromagnetic resistivity logging during drilling include single-factor wellbore environment correction and multi-factor formation parameter comprehensive inversion techniques, which have achieved certain results in vertical wells. However, their application in horizontal and highly deviated wells has significant limitations: single-factor environment correction is based on simplified assumptions of the model and is usually designed for vertical well environments. It struggles to adapt to the complex multi-factor environments of highly deviated and horizontal wells, often misinterpreting the combined effect of multiple influencing factors as the effect of a single factor. This leads to overcorrection or undercorrection during the correction process, affecting the analysis of the logging environment. In real-time well site data processing, due to limited data upload volume and constraints imposed by the ambiguity of solutions, multi-parameter inversion techniques often fail to obtain stable solutions in practical applications. While reducing the number of unknowns is a common strategy for solving ambiguity problems, blindly simplifying the model without accurately identifying the controlling influencing factors may have the opposite effect, weakening the accuracy of correction and inversion, and hindering effective analysis of the logging environment.

[0004] During logging in highly deviated and horizontal wells, the electromagnetic resistivity logging curves are affected by various environmental factors. Existing techniques cannot clearly identify and quantify these influencing factors, leading to significant confusion in logging interpretation and increasing the risk of inversion failure due to inaccurate control of influencing factors. Currently, automatic identification of complex environmental influencing factors in horizontal and highly deviated wells remains an industry challenge, primarily relying on manual experience, which is limited by the widespread adoption of methods and human factors. Summary of the Invention

[0005] In view of the above problems, embodiments of the present invention are proposed to provide a method and apparatus for well logging environment analysis using convolution kernel upscaling transformation and multidimensional feature thresholds to overcome or at least partially solve the above problems.

[0006] According to one aspect of the present invention, a method for well logging environment analysis using convolution kernel dimensionality upscaling transformation and multidimensional feature thresholding is provided, the method comprising: Construct a stratigraphic model parameter library containing different model parameter conditions for the target area; The simulated drilling electromagnetic resistivity logging instrument obtains a multi-source distance logging response dataset containing various multi-source distance drilling electromagnetic resistivity curves based on the logging response of different model parameter conditions in the formation model parameter library. For each multi-source distance logging response dataset containing multi-source distance logging-while-drilling electromagnetic resistivity curve, after sorting by detection depth, an M-fold interpolation expansion is performed to obtain a one-dimensional logging curve dataset. Simultaneously, the influence weights of each model parameter are calculated based on the corresponding model parameter conditions; where M is a positive integer greater than or equal to 3. A convolution transformation is performed on the one-dimensional logging curve dataset using an M×M convolution kernel to upgrade the one-dimensional curve data to a two-dimensional image. The two-dimensional image is then divided vertically to determine the feature point sampling area, and the first image feature parameters are extracted. The first image complexity is calculated by combining the corresponding image pixel values ​​and normalized to obtain the first richness level. A one-to-one mapping relationship is established between the feature combination formed by the first image feature parameters and the first richness level and the corresponding model parameter influence weights to obtain the richness level of influencing factors. An empirical knowledge base containing the richness levels of influencing factors corresponding to each drilling-while-drilling electromagnetic resistivity curve is constructed. Obtain the measured target logging curve of the target area, calculate the corresponding second image feature parameters and second richness level based on the measured target logging curve, query the experience knowledge base and match the corresponding richness level of influencing factors, determine the influence weight of target model parameters and the conditions of target model parameters, and complete the logging environment analysis.

[0007] Optionally, The simulation of the logging response of the electromagnetic resistivity logging instrument while drilling (ESL) under different model parameter conditions in the formation model parameter library yields a multi-source distance logging response dataset containing ESL curves for each multi-source distance. This dataset further includes: Based on the parameter conditions of each set of models, a corresponding formation model is constructed. The logging response of the electromagnetic resistivity logging instrument under each formation model is calculated using the finite element numerical simulation method, and the multi-source distance electromagnetic resistivity curves corresponding to different source distances are obtained.

[0008] Optionally, for each multi-source distance logging response dataset, after sorting by detection depth, M-fold interpolation expansion is performed to obtain a one-dimensional logging curve dataset. Simultaneously, the model parameter influence weights of each model parameter are calculated based on the corresponding model parameter conditions, further including: Arrange the electromagnetic resistivity curves of each source distance logging response data set during drilling in order of increasing depth to obtain a curve column consisting of N curves. The curve series is extended by M-fold linear interpolation along the detection depth dimension to obtain a one-dimensional logging curve dataset. The set of function variables is obtained based on the well diameter, eccentricity, invasion depth, invasion zone resistivity, formation resistivity, anisotropy, surrounding rock resistivity, layer thickness, formation dip angle, and boundary distance in the model parameter conditions. The rate of change in logging response caused by the change of a single model parameter is calculated based on the set of function variables. The total change in response is obtained by accumulating the rate of change in logging response of all model parameters. The influence weight of each model parameter on the logging response is obtained based on the ratio of the rate of change in logging response of each model parameter to the total change in response.

[0009] Optionally, The convolution transformation of the one-dimensional well logging curve dataset based on the M×M size convolution kernel further includes: The one-dimensional logging curve dataset is arranged in columns to form an initial two-dimensional matrix. An M×M convolution kernel is used to perform a convolution operation on the initial two-dimensional matrix to obtain two-dimensional wave train data that characterizes the correlation features between curves. Using the curve column direction of the one-dimensional logging curve dataset as the horizontal axis and the depth direction as the vertical axis, and the logging depth sampling interval as the pixel spacing, the convolution calculation value of the two-dimensional wave train data is used as the pixel grayscale fill value to generate a two-dimensional image.

[0010] Optionally, the two-dimensional image is divided into regions to determine the feature point sampling area, the first image feature parameters are extracted, and the first image complexity is calculated by combining the corresponding image pixel values. After normalization, the first richness level is obtained. The feature combination formed by the first image feature parameters and the first richness level is mapped one-to-one with the influence weight of the corresponding model parameters to obtain the richness level of the influencing factors. The empirical knowledge base containing the richness levels of the influencing factors corresponding to each drilling electromagnetic resistivity curve is further included: The two-dimensional image is divided into P regions along the vertical direction, and the preset local regions corresponding to the center and edge positions of each region are used as feature point sampling regions. A two-dimensional image is converted to grayscale, and the grayscale value of each pixel is mapped to a preset value range of [0, 255] to obtain the corresponding pixel value; The slope, mean, standard deviation, volatility, amplitude, width, and steepness of pixel values ​​are extracted within the sampling area of ​​each feature point as the first image feature parameters. Along the vertical direction of the two-dimensional image, calculate the sum of squares of the gray level differences between two horizontally adjacent points in each row of pixels, and use this as the first image complexity for each row. The first image complexity of all rows of the two-dimensional image is normalized and mapped to the preset interval [0, 1] to obtain the first richness level; The feature combination consisting of the first image feature parameters and the first richness level is used to establish a one-to-one mapping relationship with the influence weights of each model parameter under the corresponding model parameter conditions to obtain the richness level of the influencing factors. An experience knowledge base is constructed based on the richness level of influencing factors corresponding to each drilling electromagnetic resistivity curve. The experience knowledge base uses the first richness level as the first index and the feature vector composed of the first image feature parameters as the secondary index. Both the first index and the secondary index are sorted in ascending order of numerical value.

[0011] Optionally, the measured target logging curve of the target area is obtained, and the corresponding second image feature parameters and second richness level are calculated based on the measured target logging curve. The corresponding richness level of influencing factors is obtained by querying the experience knowledge base and matching, and the influence weight of the target model parameters and the conditions of the target model parameters are determined. The logging environment analysis is further completed by including: Obtain the measured logging curves of the target area, and select the measured logging curves at the corresponding depths of the drilling electromagnetic resistivity curves used to construct the experience knowledge base as the measured target logging curves. The measured target logging curve is expanded by M times interpolation, convolved by an M×M size convolution kernel, and the data is upgraded in dimensionality. The second image feature parameters are then extracted and combined with the image pixel values ​​after the data upgrade to calculate and normalize the second richness level. The second richness level and the second image feature parameters are nonlinearly fitted and matched with the first richness level and the first image feature parameters of the experience knowledge base, respectively, to determine the richness level of the influencing factor with the highest matching degree. Based on the richness level of the influencing factor with the highest matching degree, the corresponding target model parameter influence weight and target model parameter conditions are extracted. Based on the target model parameter influence weight, the main environmental influencing factors are identified, and the well logging environment analysis is completed based on the target model parameter conditions.

[0012] According to another aspect of the present invention, a device for analyzing well logging environment using convolution kernel dimensionality upscaling transformation and multidimensional feature thresholds is provided, the device comprising: The parameter library construction module is suitable for building a stratigraphic model parameter library containing different model parameter conditions for the target area; The logging response simulation module is suitable for simulating the logging response of the drilling electromagnetic resistivity logging instrument based on different model parameter conditions in the formation model parameter library, and obtaining a multi-source distance logging response dataset containing the drilling electromagnetic resistivity curves of each multi-source distance. The knowledge base construction module is suitable for sorting the multi-source distance logging response dataset for each multi-source distance logging electromagnetic resistivity curve after drilling by depth and then performing M-fold interpolation expansion to obtain a one-dimensional logging curve dataset. Simultaneously, it calculates the influence weights of each model parameter based on the corresponding model parameter conditions, where M is a positive integer greater than or equal to 3. Based on an M×M convolution kernel, it performs a convolution transformation on the one-dimensional logging curve dataset, upgrading the one-dimensional curve data to a two-dimensional image. The two-dimensional image is then divided vertically to determine the feature point sampling area, extracting the first image feature parameters. The first image complexity is calculated by combining the corresponding image pixel values ​​and normalized to obtain the first richness level. The feature combination formed by the first image feature parameters and the first richness level is mapped one-to-one with the corresponding model parameter influence weights to obtain the richness level of influencing factors. This constructs an empirical knowledge base containing the richness levels of influencing factors corresponding to each drilling electromagnetic resistivity curve. The matching analysis module is suitable for obtaining the measured target logging curve of the target area, calculating the corresponding second image feature parameters and second richness level based on the measured target logging curve, querying the experience knowledge base and matching to obtain the corresponding richness level of influencing factors, determining the influence weight of target model parameters and the conditions of target model parameters, and completing the logging environment analysis.

[0013] According to another aspect of the present invention, a computing device is provided, 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 through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method.

[0014] According to another aspect of the present invention, a computer storage medium is provided, wherein at least one executable instruction is stored therein, the executable instruction causing a processor to perform operations corresponding to the above-described convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method.

[0015] According to another aspect of the present invention, a computer program product is provided, including at least one executable instruction that causes a processor to perform operations corresponding to the above-described convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method.

[0016] The convolution kernel-based dimensionality-upgrading transformation and multi-dimensional feature threshold logging environment analysis method and apparatus provided by embodiments of the present invention associates and transforms logging curve datasets using convolution kernels to form two-dimensional images. This increases the vertical and horizontal correlation of data while preserving the differences between curves at different depths, mapping the influence of environmental factors on one-dimensional logging curves onto the two-dimensional image, thus visualizing the comprehensive influencing factors of logging curves and enhancing the richness of logging curve difference information. By calculating image feature parameters in the two-dimensional image, the richness level of logging influencing factors is quantitatively calculated, enabling rapid identification of comprehensive influencing factors in complex environments. Without increasing the number of logging items, through data association and transformation, two-dimensional comparisons of information between logging curves at different depths are performed, amplifying the role of influencing factors on different logging curves, while distinguishing the differences in the influence of different factors in the image direction. This achieves the differentiation and weight calculation of multiple influencing factors, providing an efficient solution for comprehensive environmental impact analysis of drilling resistivity logging data and accurate post-drilling interpretation and evaluation.

[0017] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more obvious and understandable, specific implementation methods of the embodiments of the present invention are described below. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1A flowchart of a well logging environment analysis method based on convolution kernel dimension upscaling transformation and multidimensional feature thresholding according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of a two-dimensional cross-section of the stratigraphic model is shown; Figure 3 This diagram illustrates the variation of simulated logging curves with depth. Figure 4 This diagram illustrates the horizontally spaced arrangement of data columns after the well logging curves have been expanded. Figure 5 This diagram illustrates the 2D image result after dimensionality upscaling using the response convolution kernel. Figure 6 A schematic diagram illustrating the calculation of image feature parameters for a two-dimensional image is shown. Figure 7 An example diagram illustrating the application effect based on this invention is shown; Figure 8 A schematic diagram of a well logging environment analysis device based on convolution kernel dimensionality upscaling transformation and multidimensional feature thresholding according to an embodiment of the present invention is shown. Figure 9 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0020] Figure 1 A flowchart of a well logging environment analysis method based on convolution kernel dimensionality upscaling transformation and multidimensional feature thresholding according to an embodiment of the present invention is shown, as follows: Figure 1 As shown, the method includes the following steps: Step S101: Construct a stratigraphic model parameter library containing different model parameter conditions for the target area.

[0021] For the target area (taking sandstone and mudstone reservoirs as an example), a formation model parameter library covering different model parameter conditions is constructed based on formation properties and well condition parameters. The model parameter conditions include multiple types of formation and wellbore parameters, and the range of values ​​and key value points for each parameter are shown in Table 1. Table 1

[0022] As shown in Table 1, the first column lists the model parameters, the second column lists the range of values, the third column lists the key value points, and the fourth column lists the units of the model parameters. Model parameters include, in addition to those listed in Table 1 such as well diameter, mud resistivity, eccentricity, invasion depth, invasion zone resistivity, anisotropy coefficient, target layer resistivity, surrounding rock resistivity, and well inclination, as well as comprehensive factors such as layer thickness, formation dip angle, and boundary distance. The above are just examples; specific settings should be determined based on the implementation situation and are not limited here.

[0023] Different formation models can be constructed based on varying model parameter conditions, facilitating well logging curve analysis based on accurate model parameters. A two-dimensional profile of the formation model is shown below. Figure 2 As shown, it includes well diameter 101, eccentricity distance 102, intrusion zone 103, layer thickness 104, formation dip angle 105, upper surrounding rock 106, lower surrounding rock 107, target layer 108, measurement point 109, and boundary distance 110.

[0024] Step S102: Simulate the logging response of the drilling electromagnetic resistivity logging instrument based on different model parameter conditions in the formation model parameter library to obtain a multi-source distance logging response dataset containing the drilling electromagnetic resistivity curves of each multi-source distance.

[0025] To ensure the accuracy of the correspondence between the logging curves obtained from the logging response and the formation parameters, different formation models can be constructed by traversing different model parameter conditions. A logging-while-drilling (LWD) electromagnetic resistivity instrument can then be used to log the data on each of these models. Numerical simulations, such as those using the finite element method, can be employed to simulate the logging response of the instrument, ultimately yielding multi-source-distance LWD electromagnetic resistivity curves at different logging depths. Traversing different model parameter conditions yields different LWD electromagnetic resistivity curves, which are then synthesized into a multi-source-distance logging response dataset.

[0026] For each model parameter condition, the following process is performed to construct its corresponding formation model, and the logging response of the drilling electromagnetic resistivity logging instrument under each formation model is simulated using preset numerical simulations to obtain multi-source-distance drilling electromagnetic resistivity curves with different source distances: When simulating the logging response of a well logging instrument, a cylindrical coordinate system is used as the solution coordinate system, and the solution domain for the three-dimensional forward modeling while drilling is a cylinder. To improve the calculation accuracy, a perfectly matched cylindrical layer is added. The intrinsic parameters of the anisotropic medium in the perfectly matched layer have the following forms:

[0027] in, , These are the relative magnetic permeability and relative permittivity of the formation, respectively. , These are the vacuum permeability and vacuum permittivity, respectively. Let be the tensor constitutive parameter of the perfectly matched layer in diagonal form, which contains information about the conformal perfectly matched layer. Outside the matched layer, its value is I (identity matrix). , These are tensor permeability and tensor permittivity, respectively.

[0028] Based on Maxwell's equations, in regions containing anisotropic media such as perfectly matched layers, the following Maxwell's equations are satisfied:

[0029]

[0030] in, , Let be the electric field strength and magnetic field strength of the electromagnetic wave scattering field, respectively, and j be the imaginary unit. Angular frequency, The relative magnetic permeability of the formation. The relative permittivity, For curl operator, The constitutive parameters of the diagonally matched layer, in the form of a single-axis tensor in the local coordinate system, are as follows:

[0031] in, Let be the triple orthogonal local coordinates, where a, b, and c are the absorption coefficient tensor components of the perfectly matched layer in the three local coordinate directions. Their values ​​gradually change with the position within the matched layer. A second-order polynomial absorption profile is typically used to achieve broadband low-reflection characteristics. Outside the matched layer, the values ​​are 1. When using this model, the uniaxial tensor in the local coordinate system needs to be transformed to a Cartesian coordinate system, i.e., a similarity transformation is performed using the Jacobian matrix.

[0032] in, Given the Jacobian matrix of the coordinate transformation, and under the condition of triple orthogonality of the local coordinate systems, it satisfies... x, y, z are Cartesian coordinates.

[0033] According to Maxwell's equations, the three-dimensional electromagnetic wave scattering problem with a perfectly matched layer can be described by the following vector wave equation:

[0034] Using the generalized variational principle, the functional expression corresponding to the above vector wave equation is as follows:

[0035] Where K0 is the free space wavenumber, It is a volumetric infinitesimal element.

[0036] Solving the above functional expression using vector finite element method yields the vector edge electric field strength values ​​for electromagnetic wave scattering, including the conformal perfectly matched layer. After processing the calculated vector edge electric field strength values, the induced electromotive force (EMF) generated by the receiving coil at a certain source distance can be obtained. After obtaining the induced EMF, the phase difference and amplitude resistivity are calculated. When there is an angle between the instrument's transmitting and receiving coils, in the formation medium, the real and imaginary parts of the induced EMF measured by the receiving coil change periodically with the instrument's rotation, which can be expressed in the following form:

[0037]

[0038] ;

[0039] ;

[0040] ; ; in, The complex voltage signal of the receiving coil in direction jj relative to the transmitting coil in direction ii, where the exponents ii and jj take values ​​in the x, y, and z directions of the coordinate system. , , , , , , , , All are complex coefficients. The tool face angle is the azimuth angle of the receiving coil's magnetic moment direction relative to the borehole's high side or relative to the magnetic north pole direction. The angle between the direction of the magnetic moment of the transmitting coil and the axis of the instrument is denoted as . This is the angle between the direction of the magnetic moment of the receiving coil and the axis of the instrument.

[0041] When azimuth resistivity phase difference is defined When it is a transmitting coil, the receiving coil exist The measured phase value, and the receiving coil exist The measured phase difference is:

[0042] in, , They are respectively Azimuth The imaginary and real parts of the receiving coil voltage value , They are respectively Azimuth The imaginary and real parts of the receiving coil voltage value.

[0043] Directional amplitude is defined as When it is a transmitting coil, the receiving line exist The measured amplitude value, compared with the receiving coil exist The ratio of the measured amplitude values ​​is:

[0044]

[0045] When rotating to measure, the electromagnetic resistivity logging instrument divides the 360-degree circumference into 32 zones, corresponding to different azimuths, and obtains the phase difference and amplitude ratio of the 32 azimuths. Then, by looking up the azimuth phase difference and amplitude ratio through a linked list, the apparent resistivity of the azimuth can be obtained.

[0046] Based on the detection capabilities of the logging-while-drilling (LWD) electromagnetic resistivity instrument, N LWD electromagnetic resistivity curves at different detection depths can be selected (N=4 in this embodiment), such as high-frequency short source-distance phase difference (shallow detection mode), high-frequency medium source-distance phase difference resistivity (medium detection mode), high-frequency long source-distance phase difference resistivity (deep detection mode), and high-frequency long source-distance amplitude resistivity (ultra-deep detection mode). The measured values ​​of the four curves are arranged according to the depth of the instrument's measurement point, forming a simulation curve. Figure 3 As shown, the horizontal axis represents apparent resistivity in ohm-meters, which is the measured value of the instrument or the simulation result of the model. The vertical axis represents depth in meters. Figure 3 The curves represent the short source-to-source phase, medium source-to-source phase, long source-to-source phase, and long source-to-source amplitude, respectively.

[0047] After traversing all model parameter conditions, the electromagnetic resistivity curves corresponding to each set of parameters can be obtained, which together constitute the multi-source distance logging response dataset.

[0048] Step S103: For each multi-source distance logging response dataset, sort the multi-source distance logging-while-drilling electromagnetic resistivity curves by detection depth and perform M-fold interpolation expansion to obtain a one-dimensional logging curve dataset. Simultaneously, calculate the model parameter influence weights of each model parameter based on the corresponding model parameter conditions. Perform convolution transformation on the one-dimensional logging curve dataset using an M×M convolution kernel to upgrade the one-dimensional curve data to a two-dimensional image. Divide the two-dimensional image into regions, determine the feature point sampling areas, extract the first image feature parameters, and calculate the first image complexity by combining the corresponding image pixel values. After normalization, obtain the first richness level. Establish a one-to-one mapping relationship between the feature combination composed of the first image feature parameters and the first richness level and the corresponding model parameter influence weights to obtain the richness level of influencing factors. Construct an empirical knowledge base containing the richness levels of influencing factors corresponding to each logging-while-drilling electromagnetic resistivity curve.

[0049] For each multi-source distance logging response dataset, the multi-source distance logging-while-drilling electromagnetic resistivity curve is expanded by M times to obtain a one-dimensional logging curve dataset, and the influence weights of model parameters are calculated. A two-dimensional image is obtained by increasing the dimensionality using convolution kernels, image features are extracted and richness levels are calculated, and finally, an empirical knowledge base is constructed. The specific process is as follows: (1) Curve expansion and influence weight calculation The multi-source-distance electromagnetic resistivity curves are arranged in ascending order of detection depth and descending order of resolution, resulting in a curve column of N curves. The curve column is then extended by M-fold linear interpolation along the detection depth dimension (M=5 in this embodiment) to obtain a one-dimensional logging curve dataset. The extended curves are still sorted in ascending order of detection depth, and the columns are arranged with a fixed spacing, such as... Figure 4 As shown, the horizontal axis represents the apparent resistivity after convolution, which is dimensionless, and the vertical axis represents the depth, in meters.

[0050] For each set of model parameters, the wellbore diameter DH, eccentricity distance ECC, invasion depth Ri, invasion zone resistivity Rxo, formation resistivity Rh, anisotropy λ, surrounding rock resistivity Rs (including upper and lower surrounding rocks), layer thickness thic, formation dip angle dip, and boundary distance DTB are stored in a one-dimensional array to form a set of function variables. ,as follows:

[0051] Define function To determine the logging response values ​​under the corresponding variable set, this embodiment uses the average apparent resistivity of the target well section as the response benchmark. Let... Let be the change in the i-th model parameter. Let the change in the model parameter i cause the change in the logging response, and let the two satisfy:

[0052] Based on the above, the change in instrument response caused by a change in a single model parameter can be simulated and recorded. By normalizing the change in instrument response caused by the change in model parameters, the normalized response change rate is obtained. ,as follows:

[0053] The total change in response is obtained by accumulating the rate of change of each model parameter. ,as follows:

[0054] The influence weights of the model parameters on the logging environment are obtained based on the response change rate and total response change of each model parameter. ,as follows:

[0055] The influence weights of model parameters are associated with the model parameters, and the influence weights of each model parameter can be obtained.

[0056] (2) Convolution kernel upscaling to generate two-dimensional images An M×M convolution kernel is used to perform convolution operations on the one-dimensional well logging curve dataset to enhance the vertical and horizontal correlation of the data, thereby achieving data dimensionality enhancement. This embodiment uses a 5×5 convolution kernel K, as follows:

[0057] First, the expanded one-dimensional curve columns are aligned by depth and arranged into an initial two-dimensional matrix. Then, the initial two-dimensional matrix is ​​convolved using the aforementioned convolution kernel to obtain two-dimensional wave train data that characterizes the correlation features between curves at different detection depths.

[0058] Using the curve column direction of the one-dimensional well logging curve dataset as the horizontal axis and the depth direction as the vertical axis, and the well logging depth sampling interval as the pixel spacing, the convolutional values ​​of the two-dimensional wave train data are used as the grayscale fill values ​​for the pixels to generate a two-dimensional image, such as... Figure 5 As shown. Figure 5 The horizontal axis represents the curve data column, which is dimensionless, and the vertical axis represents the depth direction, with the unit being meters. Different colors represent the size of the convolution calculation value of the pixel.

[0059] (3) Image feature extraction and richness level calculation The two-dimensional image is divided into N equal regions along the vertical direction (N=4 in this embodiment). The local regions corresponding to the center and edge positions of each region are used as feature point sampling regions. The two-dimensional image is then converted to grayscale, and the grayscale values ​​of each pixel are linearly mapped to the interval [0, 255] to obtain the corresponding pixel values.

[0060] The pixel values ​​within the feature point sampling area are scanned row by row and column by column to extract the variation features of gray values ​​with the horizontal and vertical coordinates, thus obtaining the first image feature parameters, including slope, standard deviation, mean, volatility, amplitude, width, steepness, etc. The calculation formulas for each parameter are shown in Table 2. Table 2

[0061] In Table 2, x and y correspond to the horizontal and vertical coordinates of the two-dimensional image, respectively. peak For grayscale peak value, y valley For grayscale valley values, x peak The peak value of the x-axis is x. valley denoted as the valley value on the horizontal axis, and n represents the total number of pixels within the sampling area.

[0062] Based on the definitions in Table 2, the first image feature parameters corresponding to the sampling regions of each feature point can be extracted. The distribution of the magnitudes of the first image feature parameters of each pixel in the two-dimensional image is as follows: Figure 6 As shown, the horizontal axis represents the data column, which is dimensionless, and the vertical axis represents the depth, with the unit being meters. Different colors represent the magnitude of the feature parameter.

[0063] For the grayscale-converted 2D image, the sum of squares of the grayscale differences between horizontally adjacent pixels is calculated row by row along the vertical direction, i.e., the detection depth direction, to obtain the first image complexity corresponding to each row. i The first image complexity corresponding to each row is D. yi (f) The calculation formula is:

[0064] Where f(x, yi) is the y-th... i The grayscale value of the pixel in the xth row and xth column; For the yth i The grayscale values ​​of the pixels in the x+1th row are summed over x to obtain all horizontally adjacent pixel pairs in that row.

[0065] After calculating the first image complexity of each row in the 2D image, the first image complexity of all rows is normalized and mapped to the preset interval [0, 1] to obtain the first richness level. The normalized formula becomes as follows:

[0066] Where e is the index of the pixel in the depth direction, e=1, 2, 3, ..., u, u is the total number of pixels in the depth direction of the two-dimensional image; max and min represent the maximum and minimum values ​​of the complexity of all rows in the two-dimensional image, respectively.

[0067] (4) Construction of experience knowledge base The feature combination consisting of the first image feature parameters and the first richness level is mapped one-to-one with the influence weights of each model parameter under the corresponding model parameter conditions to obtain the richness level of the influencing factors corresponding to this set of parameters.

[0068] For all drilling electromagnetic resistivity curves, the above calculations (1) curve expansion and influence weight calculation and (3) image feature extraction and richness level calculation are performed. A one-to-one mapping relationship is established between the feature combination composed of the first image feature parameters and the first richness level and the influence weight of each model parameter of each group of model parameters, so as to obtain the richness level of the influencing factors corresponding to each group of model parameters. Based on this, an experience knowledge base is constructed. The experience knowledge base adopts a hierarchical index structure: the first richness level is used as the first index, and the multidimensional feature vector composed of the first image feature parameters is used as the secondary index. Both the first index and the secondary index are sorted in ascending order of value to improve the retrieval and matching efficiency.

[0069] Step S104: Obtain the measured target logging curve of the target area, calculate the corresponding second image feature parameters and second richness level based on the measured target logging curve, query the experience knowledge base and match the corresponding richness level of influencing factors, determine the influence weight of target model parameters and the conditions of target model parameters, and complete the logging environment analysis.

[0070] The measured target logging curves for the target area are obtained. Using the same processing method as the drilling electromagnetic resistivity curves obtained from formation model logging under different model parameter conditions, the data undergoes M-fold interpolation expansion, M×M convolution kernel convolution transformation, and data dimensionality upscaling. The second image feature parameters are then extracted, and combined with the upscaled image pixel values ​​for calculation and normalization to obtain the second richness level. Corresponding environmental influencing factors are obtained by matching with an empirical knowledge base, completing the logging environment analysis. The specific process is as follows: Obtain the measured logging curves of the target area, and select the measured logging curves at the corresponding depths as the measured target logging curves based on the detection depths of the various electromagnetic resistivity curves used in the construction of the experience knowledge base.

[0071] Using the same calculation process as step S103, the corresponding second image feature parameters and second richness level are calculated based on the measured target logging curve.

[0072] The second richness level and the second image feature parameters are nonlinearly fitted and matched with the first richness level and the first image feature parameters in the experience knowledge base. In this embodiment, the simulated annealing algorithm is used to solve for the optimal matching result. The specific process is as follows: The least squares method is used to establish the energy function as follows:

[0073] Where k is the characteristic parameter index, taking a value of 1-m, and m is the total number of logging characteristic parameters. For model parameter vectors, Model parameters in the experience knowledge base The corresponding combination value of the k-th feature parameter and the richness level, This is the combination value of the k-th characteristic parameter and the abundance level corresponding to the measured target logging curve.

[0074] The matching process is essentially finding the minimum value of the energy function P. The iteration termination condition is set as follows: the change in the energy function is less than a preset threshold or the maximum number of iterations is reached. The iteration steps are as follows: 1. Given the current solution (Initial value), the energy function value is calculated based on the energy function. Simultaneously, a system temperature change function T=T0 is established. a, where a is the temperature decay coefficient (e.g., a=0.9, a range of 0.85-0.99), T0 is the iterative temperature value (initial value is set according to the implementation situation), and T is the calculated current system temperature.

[0075] 2. Determine the obtained energy function value Whether the iteration termination condition is met. The iteration termination condition includes calculating the difference between the energy function value of the current solution and the energy function value of the previous solution to obtain the change in energy function. If the change in energy function is less than a preset threshold (the preset threshold is set according to the implementation situation), then the iteration termination condition is met and the iteration terminates. Alternatively, the current iteration count is counted. If the maximum iteration count is reached, then the iteration termination condition is met and the iteration terminates. If neither of these conditions is met, then step 3 is executed. For If the energy function value of the previous solution does not exist, then there is no need to determine the change in the energy function.

[0076] 3. If the termination condition is not met, for the current solution Applying random perturbations yields new solutions Calculate the energy function value corresponding to the new solution. ,judge Is it less than If yes, proceed to step 4; otherwise, proceed to step 5.

[0077] 4. If Then accept the new solution. , will new interpretation Updated to the current solution, for Energy function value Continue to determine whether the iteration termination condition is met, that is, continue to execute the operations corresponding to the determination results of whether the iteration termination condition is met in steps 2 and 3.

[0078] 5. If The acceptance of the new solution is determined by the Boltzmann distribution probability, with the acceptance probability being:

[0079] in, To update the energy value of the previous solution, i.e., the corresponding , The energy value of the updated solution, i.e., the corresponding , Let Pr be the Boltzmann constant, and T be the current system temperature calculated based on the system temperature change function. After calculating Pr, the iterative temperature value T0 is updated to T. The system temperature is then reduced based on the temperature decay coefficient and the temperature change function, thus calculating the new current system temperature for use in subsequent iterations.

[0080] Judge Pr, if If yes, then confirm acceptance and... As the updated current solution, for Execute steps 2 and 3 based on the results of the iteration termination condition checks. If... Calculated by probability weighting The obtained numerical value is used as the current solution, and the energy function value is calculated. Then execute the operations corresponding to the judgment results of steps 2 and 3 based on whether the iteration termination condition is met.

[0081] Through the above iterations, the iteration is terminated based on the termination condition, such as the change in the energy function being less than a preset threshold or reaching the maximum number of iterations. After obtaining the richness level of the influencing factors with the highest matching degree, the corresponding target model parameter influence weights and target model parameter conditions are extracted. The target model parameter conditions are used as the environmental parameter inversion result. Parameters with non-zero influence weights and high proportions are identified as the main controlling influencing factors of the well logging environment. By accumulating the influence weights of each target model parameter, the influence degree of each factor is visually displayed by the area proportion of the filled image, achieving quantitative analysis of the influencing factors of the well logging environment. The output format of the analysis results is as follows: Figure 7 As shown, from left to right, the channels are: Exploration Depth (in meters), Lithology (containing curves from left to right representing: Intrusion Depth (in meters); Natural Gamma (in API, American Petroleum Institute units); Anisotropy Coefficient (dimensionless), Drilling Resistivity (in ohms-meters), Drilling Rate (in meters per hour), Resistivity Grid Map (in ohms-meters), Image Features (dimensionless), Confidence Level (presented as a percentage, dimensionless), and Percentage of Influencing Factors (presented as a percentage, dimensionless).

[0082] According to this embodiment, the problem of identifying and quantitatively evaluating the influencing factors in drilling electromagnetic resistivity high-angle wells and horizontal wells is solved. Furthermore, parameter inversion and quantitative calculation of multiple categories of comprehensive influencing factors can be performed on this basis.

[0083] The convolution kernel-based dimensionality-upgrading and multi-dimensional feature threshold logging environment analysis method provided by this invention uses convolution kernels to associate and transform logging curve datasets to form a two-dimensional image. This method increases the vertical and horizontal correlation of data while preserving the differences between curves at different depths, mapping the influence of environmental factors on one-dimensional logging curves onto the two-dimensional image. This visualizes the comprehensive influencing factors of logging curves and enhances the richness of information about differences in logging curves. By calculating the image feature parameters in the two-dimensional image, the richness level of logging influencing factors is quantitatively calculated, enabling rapid identification of comprehensive influencing factors in complex environments. Without adding logging items, through data association and transformation, a two-dimensional comparison of information between logging curves at different depths is performed, amplifying the role of influencing factors on different logging curves. Simultaneously, it distinguishes the differences in the influence of different factors in the image direction, achieving the differentiation and weight calculation of multiple influencing factors. This provides an efficient solution for the comprehensive environmental impact analysis of logging resistivity data during drilling and for accurate post-drilling interpretation and evaluation.

[0084] Figure 8 A schematic diagram of the structure of the well logging environment analysis device with convolution kernel dimensionality upscaling transformation and multidimensional feature threshold provided in an embodiment of the present invention is shown. Figure 8 As shown, the device includes: The parameter library construction module 810 is suitable for constructing a stratigraphic model parameter library containing different model parameter conditions for the target area; The logging response simulation module 820 is suitable for simulating the logging response of the drilling electromagnetic resistivity logging instrument based on different model parameter conditions in the formation model parameter library, and obtaining a multi-source distance logging response dataset containing the drilling electromagnetic resistivity curves of each multi-source distance. The knowledge base construction module 830 is suitable for sorting the multi-source distance logging response dataset for each multi-source distance logging electromagnetic resistivity curve after drilling by depth and then performing M-fold interpolation expansion to obtain a one-dimensional logging curve dataset. Simultaneously, it calculates the influence weights of each model parameter based on the corresponding model parameter conditions, where M is a positive integer greater than or equal to 3. Based on an M×M convolution kernel, it performs a convolution transformation on the one-dimensional logging curve dataset, upgrading the one-dimensional curve data to a two-dimensional image. The two-dimensional image is then divided vertically to determine the feature point sampling area, extracting the first image feature parameters. The first image complexity is calculated by combining the corresponding image pixel values ​​and normalized to obtain the first richness level. The feature combination formed by the first image feature parameters and the first richness level is mapped one-to-one with the corresponding model parameter influence weights to obtain the richness level of influencing factors. This constructs an empirical knowledge base containing the richness levels of influencing factors corresponding to each logging electromagnetic resistivity curve. The matching analysis module 840 is suitable for acquiring the measured target logging curve of the target area, calculating the corresponding second image feature parameters and second richness level based on the measured target logging curve, querying the experience knowledge base and matching to obtain the corresponding richness level of influencing factors, determining the influence weight of target model parameters and the conditions of target model parameters, and completing the logging environment analysis.

[0085] Optionally, the well logging response simulation module 820 is further adapted to: Based on the parameter conditions of each set of models, a corresponding formation model is constructed. The logging response of the electromagnetic resistivity logging instrument under each formation model is calculated using the finite element numerical simulation method, and the multi-source distance electromagnetic resistivity curves corresponding to different source distances are obtained.

[0086] Alternatively, the knowledge base building module 830 is further adapted to: Arrange the electromagnetic resistivity curves of each source distance logging response data set during drilling in order of increasing depth to obtain a curve column consisting of N curves. The curve series is extended by M times linear interpolation along the detection depth dimension to obtain a one-dimensional logging curve dataset. The set of function variables is obtained based on the well diameter, eccentricity, invasion depth, invasion zone resistivity, formation resistivity, anisotropy, surrounding rock resistivity, layer thickness, formation dip angle, and boundary distance in the model parameter conditions. The rate of change in logging response caused by the change of a single model parameter is calculated based on the set of function variables. The total change in response is obtained by accumulating the rate of change in logging response of all model parameters. The influence weight of each model parameter on the logging response is obtained based on the ratio of the rate of change in logging response of each model parameter to the total change in response.

[0087] Alternatively, the knowledge base building module 830 is further adapted to: The one-dimensional logging curve dataset is arranged in columns to form an initial two-dimensional matrix. An M×M convolution kernel is used to perform a convolution operation on the initial two-dimensional matrix to obtain two-dimensional wave train data that characterizes the correlation features between curves. Using the curve column direction of the one-dimensional logging curve dataset as the horizontal axis and the depth direction as the vertical axis, and the logging depth sampling interval as the pixel spacing, the convolution calculation value of the two-dimensional wave train data is used as the pixel grayscale fill value to generate a two-dimensional image.

[0088] Alternatively, the knowledge base building module 830 is further adapted to: The two-dimensional image is divided into P regions along the vertical direction, and the preset local regions corresponding to the center and edge positions of each region are used as feature point sampling regions. A two-dimensional image is converted to grayscale, and the grayscale value of each pixel is mapped to a preset value range of [0, 255] to obtain the corresponding pixel value; The slope, mean, standard deviation, volatility, amplitude, width, and steepness of pixel values ​​are extracted within the sampling area of ​​each feature point as the first image feature parameters. Along the vertical direction of the two-dimensional image, calculate the sum of squares of the gray level differences between two horizontally adjacent points in each row of pixels, and use this as the first image complexity for each row. The first image complexity of all rows of the two-dimensional image is normalized and mapped to the preset interval [0, 1] to obtain the first richness level; The feature combination consisting of the first image feature parameters and the first richness level is used to establish a one-to-one mapping relationship with the influence weights of each model parameter under the corresponding model parameter conditions to obtain the richness level of the influencing factors. An experience knowledge base is constructed based on the richness level of influencing factors corresponding to each drilling electromagnetic resistivity curve. The experience knowledge base uses the first richness level as the first index and the feature vector composed of the first image feature parameters as the secondary index. Both the first index and the secondary index are sorted in ascending order of numerical value.

[0089] Optionally, the matching analysis module 840 is further adapted to: Obtain the measured logging curves of the target area, and select the measured logging curves at the corresponding depths of the drilling electromagnetic resistivity curves used to construct the experience knowledge base as the measured target logging curves. The measured target logging curve is expanded by M times interpolation, convolved by an M×M size convolution kernel, and the data is upgraded in dimensionality. The second image feature parameters are then extracted and combined with the image pixel values ​​after the data upgrade to calculate and normalize the second richness level. The second richness level and the second image feature parameters are nonlinearly fitted and matched with the first richness level and the first image feature parameters of the experience knowledge base, respectively, to determine the richness level of the influencing factor with the highest matching degree. Based on the richness level of the influencing factor with the highest matching degree, the corresponding target model parameter influence weight and target model parameter conditions are extracted. Based on the target model parameter influence weight, the main environmental influencing factors are identified, and the well logging environment analysis is completed based on the target model parameter conditions.

[0090] The descriptions of the above modules refer to the corresponding descriptions in the method embodiments, and will not be repeated here.

[0091] This invention also provides a non-volatile computer storage medium storing at least one executable instruction that can perform the operations corresponding to the convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method in any of the above method embodiments.

[0092] This application provides a computer program product, which includes at least one executable instruction or computer program that enables a processor to perform the operations corresponding to the convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method in any of the above method embodiments.

[0093] Figure 9 The diagram illustrates the structure of a computing device according to an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computing device.

[0094] like Figure 9 As shown, the computing device may include: a processor 902, a communication interface 904, a memory 906, and a communication bus 908.

[0095] in: The processor 902, communication interface 904, and memory 906 communicate with each other via communication bus 908.

[0096] The communication interface 904 is used to communicate with other network elements such as clients or other servers.

[0097] The processor 902 is used to execute program 910, which can specifically execute the relevant steps in the above embodiment of the convolution kernel dimensionality transformation and multidimensional feature threshold logging environment analysis method.

[0098] Specifically, program 910 may include program code that includes computer operation instructions.

[0099] Processor 902 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0100] Memory 906 is used to store program 910. Memory 906 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0101] Specifically, program 910 can be used to cause processor 902 to execute the convolution kernel dimensionality increase transformation and multidimensional feature threshold logging environment analysis method in any of the above method embodiments. The specific implementation of each step in program 910 can be found in the corresponding steps and units described in the above embodiments of convolution kernel dimensionality increase transformation and multidimensional feature threshold logging environment analysis, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described equipment and modules can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0102] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the embodiments of the present invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing preferred embodiments of the present invention.

[0103] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0104] Similarly, it should be understood that, in order to streamline the embodiments of the invention and aid in understanding one or more of the various inventive aspects, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed embodiments of the invention require more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0105] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0106] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0107] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention can also be implemented as device or apparatus programs (e.g., computer programs and computer program products) for performing part or all of the methods described herein. Such programs implementing the embodiments of the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0108] It should be noted that the above embodiments are illustrative of the present invention and not restrictive of the invention, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the present invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for analyzing well logging environment using convolution kernel dimensionality upscaling transformation and multidimensional feature thresholds, characterized in that, The method includes: Construct a stratigraphic model parameter library containing different model parameter conditions for the target area; The simulated drilling electromagnetic resistivity logging instrument obtains a multi-source distance logging response dataset containing various multi-source distance drilling electromagnetic resistivity curves based on the logging response of different model parameter conditions in the formation model parameter library. For each multi-source distance logging response dataset, the multi-source distance electromagnetic resistivity curves during drilling are sorted by detection depth and then subjected to M-fold interpolation expansion to obtain a one-dimensional logging curve dataset. At the same time, the model parameter influence weights of each model parameter are calculated according to the corresponding model parameter conditions. The one-dimensional logging curve dataset is transformed into a two-dimensional image by performing convolution transformation based on an M×M size convolution kernel. The two-dimensional image is then divided into regions to determine the feature point sampling area. First image feature parameters are extracted, and the first image complexity is calculated by combining the corresponding image pixel values. After normalization, the first richness level is obtained. The feature combination formed by the first image feature parameters and the first richness level is mapped one-to-one with the influence weight of the corresponding model parameters to obtain the richness level of the influencing factors. An empirical knowledge base containing the richness level of the influencing factors corresponding to each drilling electromagnetic resistivity curve is then constructed. The measured target logging curve of the target area is obtained. Based on the measured target logging curve, the corresponding second image feature parameters and second richness level are calculated. The experience knowledge base is queried and matched to obtain the corresponding richness level of influencing factors. The influence weight of the target model parameters and the target model parameter conditions are determined to complete the logging environment analysis.

2. The method according to claim 1, characterized in that, The simulated drilling electromagnetic resistivity logging instrument, based on logging responses under different model parameter conditions in the formation model parameter library, generates a multi-source distance logging response dataset containing various multi-source distance drilling electromagnetic resistivity curves. This dataset further includes: Based on the parameter conditions of each set of models, a corresponding formation model is constructed. The logging response of the electromagnetic resistivity logging instrument under each formation model is calculated using the finite element numerical simulation method, and the multi-source distance electromagnetic resistivity curves corresponding to different source distances are obtained.

3. The method according to claim 1, characterized in that, The process involves sorting each multi-source distance logging response dataset's electromagnetic resistivity curves during drilling by detection depth and then performing M-fold interpolation to obtain a one-dimensional logging curve dataset. Furthermore, the model parameter influence weights for each model parameter are calculated based on the corresponding model parameter conditions, including: Arrange the electromagnetic resistivity curves of each of the multi-source distance logging response datasets in order of increasing detection depth to obtain a curve column consisting of N curves. The curve series is extended by M times linear interpolation along the detection depth dimension to obtain a one-dimensional logging curve dataset; Based on the well diameter, eccentricity, invasion depth, invasion zone resistivity, formation resistivity, anisotropy, surrounding rock resistivity, layer thickness, formation dip angle, and boundary distance in the model parameter conditions, a set of function variables is obtained. Based on the set of function variables, the rate of change in logging response caused by the change of a single model parameter is calculated. The total change in response is obtained by accumulating the rate of change in logging response of all model parameters. Based on the ratio of the rate of change in logging response of each model parameter to the total change in response, the model parameter influence weight of each model parameter on the logging response is obtained.

4. The method according to claim 1, characterized in that, The step of performing a convolution transformation on the one-dimensional well logging curve dataset using an M×M-sized convolution kernel to upgrade the one-dimensional curve data into a two-dimensional image further includes: The one-dimensional logging curve dataset is arranged in columns to form an initial two-dimensional matrix. The initial two-dimensional matrix is ​​then convolved using an M×M convolution kernel to obtain two-dimensional wave train data that characterizes the correlation features between curves. Using the curve column direction of the one-dimensional logging curve dataset as the horizontal axis and the depth direction as the vertical axis, and the logging depth sampling interval as the pixel spacing, the convolution calculation value of the two-dimensional wave train data is used as the pixel grayscale fill value to generate a two-dimensional image.

5. The method according to claim 1, characterized in that, The process of dividing the two-dimensional image into regions, determining the feature point sampling region, extracting the first image feature parameters, calculating the first image complexity by combining the corresponding image pixel values, and obtaining the first richness level after normalization; establishing a one-to-one mapping relationship between the feature combination composed of the first image feature parameters and the first richness level and the corresponding model parameter influence weights to obtain the richness level of influencing factors; and constructing an empirical knowledge base containing the richness levels of influencing factors corresponding to each drilling electromagnetic resistivity curve further includes: The two-dimensional image is divided into P regions along the longitudinal direction, and the preset local regions corresponding to the center and edge positions of each region are used as feature point sampling regions. The two-dimensional image is converted to grayscale, and the grayscale value of each pixel is mapped to a preset value range of [0, 255] to obtain the corresponding pixel value; The slope, mean, standard deviation, volatility, amplitude, width, and steepness of pixel values ​​are extracted within the sampling area of ​​each feature point as the first image feature parameters. Along the vertical direction of the two-dimensional image, calculate the sum of squares of the gray level differences between two horizontally adjacent points in each row of pixels, and use this as the first image complexity for each row. The first image complexity of all rows of the two-dimensional image is normalized and mapped to the preset interval [0, 1] to obtain the first richness level; The feature combination formed by the first image feature parameters and the first richness level is used to establish a one-to-one mapping relationship with the influence weights of each model parameter under the corresponding model parameter conditions to obtain the richness level of the influencing factors. An experience knowledge base is constructed based on the richness level of influencing factors corresponding to each electromagnetic resistivity curve during drilling. The experience knowledge base uses the first richness level as the first index and the feature vector composed of the first image feature parameters as the secondary index. Both the first index and the secondary index are sorted in ascending order of numerical value.

6. The method according to any one of claims 1-5, characterized in that, The process of obtaining the measured target logging curve of the target area, calculating the corresponding second image feature parameters and second richness level based on the measured target logging curve, querying the experience knowledge base and matching the corresponding richness level of influencing factors, determining the influence weight of target model parameters and target model parameter conditions, and completing the logging environment analysis further includes: Obtain the measured logging curves of the target area, and select the measured logging curves at the corresponding depths of the drilling electromagnetic resistivity curves used to construct the experience knowledge base as the measured target logging curves. The measured target logging curve is expanded by M times interpolation, convolved by an M×M size convolution kernel, and the data is upgraded. The second image feature parameters are then extracted and combined with the image pixel values ​​after the data upgrade to calculate and normalize the second richness level. The second richness level and the second image feature parameters are respectively nonlinearly fitted and matched with the first richness level and the first image feature parameters of the experience knowledge base to determine the richness level of the influencing factor with the highest matching degree. Based on the richness level of the influencing factor with the highest matching degree, the corresponding target model parameter influence weight and target model parameter conditions are extracted. Based on the target model parameter influence weight, the main environmental influencing factors are identified. Based on the target model parameter conditions, the well logging environment analysis is completed.

7. A device for analyzing well logging environment using convolution kernel upscaling transformation and multidimensional feature thresholds, characterized in that, The device includes: The parameter library construction module is suitable for building a stratigraphic model parameter library containing different model parameter conditions for the target area; The logging response simulation module is suitable for simulating the logging response of the drilling electromagnetic resistivity logging instrument according to different model parameter conditions in the formation model parameter library, and obtaining a multi-source distance logging response dataset containing the drilling electromagnetic resistivity curves of each multi-source distance. The knowledge base construction module is suitable for sorting each multi-source distance logging response dataset's multi-source distance logging-while-drilling electromagnetic resistivity curve by detection depth and then performing M-fold interpolation expansion to obtain a one-dimensional logging curve dataset. Simultaneously, it calculates the model parameter influence weights of each model parameter based on the corresponding model parameter conditions; where M is a positive integer greater than or equal to 3. Based on an M×M convolution kernel, it performs a convolution transformation on the one-dimensional logging curve dataset, upgrading the one-dimensional curve data to a two-dimensional image. The two-dimensional image is then vertically divided to determine the feature point sampling area, extracting the first image feature parameters. The first image complexity is calculated by combining the corresponding image pixel values ​​and normalized to obtain the first richness level. The feature combination formed by the first image feature parameters and the first richness level is mapped one-to-one with the corresponding model parameter influence weights to obtain the richness level of influencing factors. This constructs an empirical knowledge base containing the richness levels of influencing factors corresponding to each logging-while-drilling electromagnetic resistivity curve. The matching analysis module is suitable for obtaining the measured target logging curve of the target area, calculating the corresponding second image feature parameters and second richness level based on the measured target logging curve, querying the experience knowledge base and matching to obtain the corresponding richness level of influencing factors, determining the influence weight of target model parameters and the conditions of target model parameters, and completing the logging environment analysis.

8. A computing device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method as described in any one of claims 1-6.

9. A computer storage medium, characterized in that, The storage medium stores at least one executable instruction that causes the processor to perform the operation corresponding to the convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes at least one executable instruction that causes the processor to perform the operation corresponding to the convolution kernel upscaling transformation and multidimensional feature threshold logging environment analysis method as described in any one of claims 1-6.