A physical constraint-based intelligent inversion method for core resistivity

Abstract

The present application relates to the field of physical exploration technology, in particular to a core resistivity intelligent inversion method based on physical constraints, comprising data preprocessing, multi-physical field coupling constraint system construction, adaptive boundary initial model establishment, initial intelligent inversion and linkage calibration iterative inversion. The core resistivity intelligent inversion method based on physical constraints integrates core resistivity observation data and micro-physical structure related data, incorporates multi-physical field influencing factors, constructs a multi-field coupling constraint system, formulates a targeted constraint strategy in combination with the physical property differences of different regions of the core, dynamically allocates constraint weights, and makes the inversion process closely match the real physical state of the core, thereby effectively improving the physical rationality and reliability of the inversion result and meeting the application requirements under complex geological environments.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration technology, and in particular to a smart inversion method for core resistivity based on physical constraints. Background Technology

[0002] Core resistivity inversion technology is a key technology in fields such as oil and gas reservoir evaluation and underground resource exploration. Its core value lies in accurately inferring key parameters such as pore structure, fluid saturation, and mineral composition inside the core by analyzing the resistivity response characteristics of the core, thus providing a scientific basis for resource reserve estimation and development plan formulation.

[0003] As exploration and development extend to deeper and more complex reservoirs, the complexity of the geological environment in which the core samples are located increases significantly, placing stringent demands on the accuracy, stability, and physical rationality of resistivity inversion.

[0004] The core problem with current mainstream intelligent core resistivity inversion methods lies in the lack of a systematic physical constraint mechanism that deeply integrates with the physical properties of the core. This causes the optimization process of the intelligent algorithm to deviate from actual physical laws, and the reliability and accuracy of the inversion results are insufficient to meet the needs of practical applications. Specifically, existing intelligent inversion technologies mostly focus on the iterative optimization of the algorithm itself, but fail to integrate the core physical properties of the core and the interaction of multiple physical fields as constraints into the entire inversion process. On the one hand, the inversion process relies solely on resistivity observation data, failing to effectively integrate microscopic physical structure data such as pore distribution and particle arrangement obtained from core CT scans, and also failing to consider the coupling influence of multiple physical field factors such as fluid seepage characteristics and temperature and pressure changes on resistivity. This results in a single constraint dimension, which cannot comprehensively depict the true physical state of the core. On the other hand, even if some methods introduce simple constraints, they do not formulate differentiated constraint strategies based on the differences in physical characteristics of different regions of the core. The allocation of constraint weights is rigid, making it difficult to address the core influencing factors in different regions in a targeted manner.

[0005] Therefore, a smart inversion method for core resistivity based on physical constraints is needed. Summary of the Invention

[0006] The purpose of this invention is to solve the problems pointed out in the background art, and to propose a smart inversion method for core resistivity based on physical constraints.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a smart core resistivity inversion method based on physical constraints, comprising the following steps: S1: Data preprocessing, extracting mid-to-low frequency data from core resistivity observation data, limiting the observation frequency band span and the number of frequency points, converting the resistivity amplitude and phase of each frequency point into real and imaginary parts, extracting pore distribution and particle arrangement data obtained from core CT scans, integrating observation data and extracted data, and generating a core basic electrical property dataset; S2: Construction of a multi-physics field coupled constraint system. Based on the core basic electrical data set, fluid seepage characteristic data and temperature and pressure change data are introduced. Core area types are divided according to porosity and mineral composition distribution. Constraint weights corresponding to each physical quantity are dynamically allocated. A multi-field coupled constraint model is established and the multi-field coupled constraint weight matrix and regional constraint priority table are output. S3: Adaptive boundary initial model establishment. Data is collected through a small number of pilot experiments. Combined with the core basic electrical dataset, the initial boundary threshold of the inversion parameters is set, the boundary confidence coefficient is calculated, and abnormal thresholds that deviate significantly from the reasonable range are removed to generate the initial constraint boundary interval. S4: Initial intelligent inversion. Input the multi-field coupling constraint weight matrix, the regional constraint priority table, and the initial constraint boundary interval. The first round of inversion operation is performed using a hybrid intelligent algorithm guided by physical priors. The physical prior guidance module filters the input inversion parameters for rationality. The optimization calculation module processes the filtered parameters using a sparrow search algorithm combined with a micro CNN architecture. The model calculation data is compared with the observation data to obtain the data residuals. The initial inversion parameter set, the filtered effective parameter set, and the corresponding residual values ​​are output. S5: Linked calibration iterative inversion. Based on the initial inversion parameter set, the filtered effective parameter set, and the residual value, the constraint weights of each region in the multi-field coupling constraint weight matrix are adjusted. An outlier identification algorithm is introduced to remove experimental error interference data. The boundary adjustment amplitude coefficient is calculated, and the initial constraint boundary interval is calibrated to obtain the dynamic constraint boundary interval. Based on the adjusted matrix, the regional constraint priority table, and the dynamic constraint boundary interval, intelligent inversion is performed again. The weight adjustment, boundary calibration, and inversion operation process are repeated until the residual value is lower than the set threshold or the preset iteration number limit is reached. The final inversion parameter set is then output.

[0008] Preferably, in S1, the range of mid-to-low frequency band data interception is limited to low frequency to within 10³Hz, the observation frequency band span is not less than four orders of magnitude, and the number of frequency points is not less than 20.

[0009] Preferably, in step S1, the data conversion adopts an amplitude and phase linkage calculation method, and the real and imaginary part values ​​are directly derived from the amplitude and phase values.

[0010] Preferably, in S2, the core region type is divided into high-porosity region, low-porosity region, and temperature- and pressure-sensitive region. Different regions correspond to different physical quantity constraints. The high-porosity region focuses on the constraint of the correlation between resistivity and saturation, the low-porosity region focuses on the constraint of mineral composition, and the temperature- and pressure-sensitive region focuses on the constraint of environmental factor correction.

[0011] Preferably, in step S3, the inversion parameters include zero-frequency resistivity, polarizability, time constant, and frequency correlation coefficient. The initial boundary threshold is set with reference to the statistical distribution range of the corresponding parameters in the basic electrical data set of the core sample. The boundary confidence coefficient is calculated by the degree of fit between the pilot experimental data and historical data.

[0012] Preferably, in step S4, the rationality screening of the physical prior guidance module is based on the core resistivity constitutive relationship, eliminating parameters that exceed the allowable range of physical properties. The screened parameters form a set of effective parameters, which serves as input data for the optimization calculation module.

[0013] Preferably, in step S5, the boundary adjustment amplitude coefficient is positively correlated with the residual value. The larger the residual value, the larger the boundary adjustment amplitude, and the smaller the residual value, the smaller the boundary adjustment amplitude, thereby gradually narrowing the parameter search range.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: 1. This invention integrates core resistivity observation data with microscopic physical structure data, incorporates multi-physics field influencing factors, constructs a multi-field coupled constraint system, formulates targeted constraint strategies based on the differences in physical characteristics of different regions of the core, and dynamically allocates constraint weights, so that the inversion process closely follows the real physical state of the core, effectively improving the physical rationality and reliability of the inversion results and meeting the application needs in complex geological environments.

[0015] 2. This invention establishes realistic initial constraint boundaries by screening observation data and transforming its representation form. It also selects inversion parameters based on physical laws, employs a hybrid algorithm to perform inversion operations, and dynamically adjusts constraint weights and boundary ranges through an iterative process to gradually narrow the parameter search interval, reduce experimental error interference, and improve the calculation accuracy of inversion parameters and the overall efficiency of inversion operations. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the workflow of the method of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a technical solution: a smart inversion method for core resistivity based on physical constraints, comprising the following steps: S1: Data preprocessing, using a bandpass filter to precisely filter core resistivity observation data for 1Hz- For mid-to-low frequency data in the Hz range, the observation frequency band must cover at least four consecutive orders of magnitude (e.g., 1Hz-10Hz-). Hz- (Hz), the frequency points are distributed with equal logarithmic intervals, with at least 5 frequency points set for each order of magnitude, and the total number of frequency points is not less than 20. The frequency ratio of adjacent frequency points is kept consistent to ensure data uniformity. Based on the complex resistivity expression Amplitude obtained through observation and phase ,use , The linkage calculation formula is converted into a real and imaginary part representation. Before the conversion, a phase correction coefficient is introduced to correct the system error. After the conversion, the rationality of the data is verified by correlation analysis, and abnormal data pairs with correlation coefficients lower than 0.9 are removed. By analyzing the grayscale values ​​of core CT scan images, pore distribution data such as pore size, connectivity, and spatial distribution coordinates, as well as particle arrangement data such as particle morphology, particle size distribution, and stacking method, were extracted. After standardizing all data in the [0,1] interval, the observation data and the extracted data were integrated to generate a core basic electrical property dataset containing the real / imaginary part of resistivity, pore parameters, and particle parameters.

[0023] S2: Construction of a multi-physics field coupled constraint system, based on the core basic electrical data set, and incorporating fluid seepage characteristic data such as permeability, fluid viscosity, and flow velocity, as well as temperature and pressure change data such as ambient temperature, confining pressure, and pore pressure; Based on porosity (high porosity ≥ 30%, low porosity < 10%), mineral composition distribution, and temperature and pressure sensitivity (temperature change rate ≥ 5℃ / 100m or pressure change rate ≥ 1MPa / 100m), core regions were classified into high-porosity regions, low-porosity regions, and temperature and pressure sensitive regions. The Analytic Hierarchy Process (AHP) was used to dynamically allocate the constraint weights of each physical quantity, constructing a multi-field coupled constraint model covering resistivity, porosity, mineral composition, fluid properties, and temperature and pressure. The influence of each physical quantity was quantified to form a constraint weight matrix. Simultaneously, a regional constraint priority table was established (high-porosity region: resistivity-saturation correlation constraint > fluid seepage constraint > mineral composition constraint; low-porosity region: mineral composition constraint > resistivity constraint > temperature and pressure constraint; temperature and pressure sensitive region: temperature and pressure correction constraint > resistivity constraint > porosity constraint). The multi-field coupled constraint weight matrix and the regional constraint priority table were then output.

[0024] S3: Establish an adaptive boundary initial model, conduct no less than 3 sets of core resistivity pilot experiments under different temperature and pressure and different fluid saturation conditions, and collect key parameter data such as zero-frequency resistivity, polarizability, time constant, and frequency correlation coefficient; Based on the mean ±3σ statistical distribution range of the corresponding parameters in the basic electrical data set of the core, and with reference to similar core experimental data, the initial boundary threshold of the inversion parameters is set. The least squares goodness of fit between pilot experimental data and historical data was used. Calculate the boundary confidence coefficient ( (If the condition is deemed valid), the 3σ criterion is used to remove outlier thresholds that exceed the mean ± 3σ range, generating the initial constraint boundary intervals for each inversion parameter.

[0025] S4: Initial intelligent inversion, the first round of inversion operation is performed by a hybrid intelligent algorithm that combines the physical prior guidance module and the optimization calculation module, based on the input multi-field coupling constraint weight matrix, regional constraint priority table and initial constraint boundary interval; The physical prior guidance module, based on the constitutive relation of resistivity in dual-porosity media cores, clarifies the physical value range of the inversion parameters (zero-frequency resistivity). Polarizability 0.01-0.3, time constant The effective parameter set after filtering by frequency correlation coefficient; The optimization calculation module adopts a combination architecture of sparrow search algorithm (global parameter search) and miniature CNN (local parameter optimization) with 3 convolutional layers + 2 fully connected layers. It calculates the residual between the model calculation data and the observed data through the mean square error formula, and outputs the initial inversion parameter set, the filtered effective parameter set and the corresponding residual values.

[0026] S5: Linked calibration iterative inversion, based on the initial inversion parameter set, the selected effective parameter set and residual values, the constraint weight in the region with large residuals is increased by 10%-20% through the constraint weight correction algorithm; The Laida criterion (3σ criterion) is introduced to eliminate experimental error interference in the data; based on quantization relationships... ( The benchmark adjustment factor is 0.1-0.3. The residual value, Calculate the boundary adjustment range coefficient (with a baseline residual value of 0.05). The larger the residual value, the larger the adjustment range, and vice versa. The dynamic constraint boundary range is obtained by calibrating the initial constraint boundary range. Based on the adjusted matrix, the regional constraint priority table, and the dynamic constraint boundary interval, intelligent inversion is performed again, repeating the weight adjustment, boundary calibration, and inversion calculation process until the residual value is obtained. Alternatively, after 50 iterations, the final inversion parameter set can be output.

[0027] Furthermore, in S1 described above, the mid-to-low frequency data extraction is achieved through a bandpass filter, with the filtering frequency strictly limited to 1Hz-10Hz. Within Hz, the observation frequency band needs to cover at least four consecutive orders of magnitude (e.g., 1Hz—10Hz-10). Hz-10 (Hz) to ensure frequency coverage integrity; frequency points are distributed with equal logarithmic intervals, with at least 5 frequency points configured for each order of magnitude, and a total number of frequency points not less than 20. The frequency ratio of adjacent frequency points is uniformly set to . This ensures that the data is evenly distributed along the frequency axis, providing stable data support for inversion calculations.

[0028] Its data conversion employs a calculation method that links amplitude and phase, based on the mathematical essence of complex resistivity, and utilizes formulas... , ( For the real part, The virtual part, To observe the amplitude, The real and imaginary parts are directly derived for the observed phase; systematic error correction is performed on the phase data before calculation. , To observe the phase, (As a preset system error value), the real and imaginary data are verified through correlation analysis after calculation to ensure the rationality of the data. Abnormal data pairs with a correlation coefficient lower than 0.9 are removed to ensure the accuracy of the conversion results.

[0029] Furthermore, in S2 above, the core region types are divided into high-porosity regions, low-porosity regions, and temperature- and pressure-sensitive regions, with the following criteria for determination: In the high-porosity region (porosity 230%), the focus is on the constraint of the relationship between resistivity and saturation, which is addressed by modifying the Archie formula. ( For the skeletal resistivity, For saturation, Establish a correlation with the saturation index and update the constraint relationship in real time; In the low-porosity region (porosity <10%), the focus is on mineral composition constraints. Based on mineral content data from X-ray diffraction analysis, a linear fitting model is constructed between the content of quartz, feldspar, and clay minerals and resistivity. , For the corresponding mineral resistivity, , , (as a mass percentage), the resistivity is calculated based on the mineral component percentage constrained by the resistivity. Temperature and pressure sensitive areas (temperature change rate > 5°C / 100m or pressure change rate) ( for Temperature coefficient of ° for pressure coefficient, The change in temperature (for pressure change), through Environmentally corrected constraints are applied to resistivity.

[0030] Furthermore, in S3 above, the inversion parameters include zero-frequency resistivity (the resistivity limit when the frequency approaches 0), polarizability (the ratio of polarization intensity to applied electric field intensity), time constant (polarization relaxation time), and frequency correlation coefficient (the attenuation coefficient of resistivity as frequency changes). The initial boundary thresholds are based on the mean ± 3σ statistical distribution of the corresponding parameters in the core basic electrical data set, and corrected by combining similar core experimental data; the boundary confidence coefficients are determined by the least squares goodness of fit between the pilot experimental data and historical data. calculate, The judgment is valid. It is necessary to expand the range of statistical samples and adjust the boundaries; the outlier threshold is eliminated using the 3σ criterion, and finally the initial constraint boundary intervals of each inversion parameter are generated.

[0031] Furthermore, in S4 above, the rationality screening of the physical prior guidance module is based on the constitutive relationship of the resistivity of the core in a dual-porosity medium, clarifying the physical value range of the inversion parameters: Zero-frequency resistivity Polarizability 0.01-0.3, time constant The frequency correlation coefficient is 0.1-1.0. During the screening process, the parameters are first checked to see if they are within the range of values, and then substituted into the constitutive model to calculate the theoretical resistivity. If the initial deviation between the theoretical and observed resistivity exceeds 50%, the parameters are removed. The parameters after double screening form the set of effective parameters after screening, which serves as the input data for the optimization calculation module.

[0032] Furthermore, in S5 above, the boundary adjustment amplitude coefficient is positively correlated with the residual value, as shown by the formula... Quantification ( To adjust the coefficient, The baseline coefficient is 0.1-0.3. This is the current residual value. (The baseline residual value is 0.05). hour The dynamic constraint boundary interval expands outward. times; hour, The boundary interval shrinks inward. This method gradually narrows the parameter search range, improving inversion efficiency and accuracy.

[0033] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart inversion method for core resistivity based on physical constraints, characterized in that, Includes the following steps: S1: Data preprocessing, extracting mid-to-low frequency data from core resistivity observation data, limiting the observation frequency band span and the number of frequency points, converting the resistivity amplitude and phase of each frequency point into real and imaginary parts, extracting pore distribution and particle arrangement data obtained from core CT scans, integrating observation data and extracted data, and generating a core basic electrical property dataset; S2: Construction of a multi-physics field coupled constraint system. Based on the core basic electrical data set, fluid seepage characteristic data and temperature and pressure change data are introduced. Core area types are divided according to porosity and mineral composition distribution. Constraint weights corresponding to each physical quantity are dynamically allocated. A multi-field coupled constraint model is established and the multi-field coupled constraint weight matrix and regional constraint priority table are output. S3: Adaptive boundary initial model establishment. Data is collected through a small number of pilot experiments. Combined with the core basic electrical dataset, the initial boundary threshold of the inversion parameters is set, the boundary confidence coefficient is calculated, and abnormal thresholds that deviate significantly from the reasonable range are removed to generate the initial constraint boundary interval. S4: Initial intelligent inversion. Input the multi-field coupling constraint weight matrix, the regional constraint priority table, and the initial constraint boundary interval. The first round of inversion operation is performed using a hybrid intelligent algorithm guided by physical priors. The physical prior guidance module filters the input inversion parameters for rationality. The optimization calculation module processes the filtered parameters using a sparrow search algorithm combined with a micro CNN architecture. The model calculation data is compared with the observation data to obtain the data residuals. The initial inversion parameter set, the filtered effective parameter set, and the corresponding residual values ​​are output. S5: Linked calibration iterative inversion. Based on the initial inversion parameter set, the filtered effective parameter set, and the residual value, the constraint weights of each region in the multi-field coupling constraint weight matrix are adjusted. An outlier identification algorithm is introduced to remove experimental error interference data. The boundary adjustment amplitude coefficient is calculated, and the initial constraint boundary interval is calibrated to obtain the dynamic constraint boundary interval. Based on the adjusted matrix, the regional constraint priority table, and the dynamic constraint boundary interval, intelligent inversion is performed again. The weight adjustment, boundary calibration, and inversion operation process are repeated until the residual value is lower than the set threshold or the preset iteration number limit is reached. The final inversion parameter set is then output.

2. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S1, the data interception range for the mid-to-low frequency band is limited to low frequency to within 10³Hz, the observation frequency band span is not less than four orders of magnitude, and the number of frequency points is not less than 20.

3. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S1, the data conversion adopts an amplitude and phase linkage calculation method, and the real part and imaginary part values ​​are directly derived from the amplitude and phase values.

4. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S2, the core region is divided into high-porosity region, low-porosity region, and temperature- and pressure-sensitive region. Different regions correspond to different physical quantity constraints. High-porosity region focuses on the constraint of resistivity and saturation correlation, low-porosity region focuses on mineral composition constraint, and temperature- and pressure-sensitive region focuses on environmental factor correction constraint.

5. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S3, the inversion parameters include zero-frequency resistivity, polarizability, time constant, and frequency correlation coefficient. The initial boundary threshold is set with reference to the statistical distribution range of the corresponding parameters in the basic electrical data set of the core. The boundary confidence coefficient is calculated by the degree of fit between the pilot experimental data and historical data.

6. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S4, the rationality screening of the physical prior guidance module is based on the core resistivity constitutive relationship, eliminating parameters that exceed the allowable range of physical properties. The screened parameters form a set of effective parameters, which serves as the input data for the optimization calculation module.

7. The intelligent core resistivity inversion method based on physical constraints according to claim 1, characterized in that: In S5, the boundary adjustment amplitude coefficient is positively correlated with the residual value. The larger the residual value, the larger the boundary adjustment amplitude, and the smaller the residual value, the smaller the boundary adjustment amplitude, thus gradually narrowing the parameter search range.

Images