Well seismic velocity error extrapolation method based on VAR-RBF

Abstract

The invention discloses a well seismic velocity error extrapolation method based on VAR-RBF, and aims to improve the precision of a seismic velocity model. The method comprises three main steps: firstly, performing pre-stack depth migration based on an initial seismic velocity model, and creating a construction model through horizon pickup; secondly, stiffness conversion is carried out on logging speed data to achieve Backus smoothing, and a smooth equivalent logging speed curve is obtained; and finally, calculating a well seismic velocity error by using a seismic velocity profile, the smoothed well logging velocity curve and the construction model, and extrapolating the velocity error to a well-free area through a VAR-RBF method so as to update and optimize a velocity model. The method effectively solves the problem of inconsistency of well seismic velocities, enhances the reliability of the velocity model, is suitable for depth inversion and velocity optimization under complex geological conditions, and improves the accuracy of seismic imaging and stratum prediction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of seismic exploration, and particularly relates to a well-seismic velocity error extrapolation method based on VAR-RBF. BACKGROUND

[0002] With the continuous development of exploration technology, seismic exploration work goes from medium-deep layer to ultra-deep layer, and the exploration target geologic body becomes more complex, and the requirement for seismic imaging precision is also continuously improved. The velocity model, as the basis of seismic imaging, its precision directly affects the accuracy and stability of the imaging result. The logging data, as one of the important data of seismic exploration, provides high-precision velocity, density and layering information, which is basically consistent with the real situation underground. Using it to constrain the construction of the seismic velocity model can improve the longitudinal precision of the velocity model and improve the lateral resolution, and then can effectively improve the authenticity of the seismic imaging result. However, this well-seismic joint modeling method still faces many problems in practical application.

[0003] The commonly used method in extrapolating logging information is the Kriging method and the random simulation and random inversion method. However, the Kriging interpolation has certain limitations: the Kriging method has high computational complexity, and when a large-scale model is constructed by using the Kriging method, the matrix operation of the Kriging method will consume huge computing resources, affecting the processing efficiency; the Kriging method has strong dependence on the variogram function, and the selection and fitting degree of the variogram function greatly affects the accuracy of the interpolation result; the Kriging method has poor extrapolation, and its accuracy is low when interpolating outside the known data range, especially in the sparse well point data, and the Kriging method does not consider the spatial correlation of the data itself, and some meaningful abnormal velocity values are often discarded. The Kriging method can obtain high-precision and reliable interpolation results under the premise of sufficient data support and using reasonable variogram function, but its applicability is poor in the case of sparse data or uneven distribution. However, the results of the random simulation and inversion are sensitive to the input parameters, and improper parameter selection may lead to model distortion, and the method usually requires a large amount of computing resources, so the development of the method is limited.

[0004] In addition, there is another interpolation method based on fitting, that is, radial basis function interpolation (Radial basis function, RBF), which does not depend on the data grid and saves the traditional interpolation grid division step. Compared with the Kriging interpolation extrapolation and the random simulation and random inversion method, the radial basis interpolation method has lower requirements for the distribution of data, and it can flexibly process irregularly distributed logging data, and its calculation efficiency is higher and is suitable for quickly constructing a velocity model. SUMMARY

[0005] The application aims to provide a well-seismic velocity error extrapolation method based on VAR-RBF, so as to realize well logging data extrapolation under uneven well distribution and insufficient well logging data, and obtain a smooth and stable fitting curve when the well logging data is insufficient.

[0006] To solve the technical problem, the technical scheme of the application is:

[0007] A well-seismic velocity error extrapolation method based on VAR-RBF, the method comprises:

[0008] S1: performing pre-stack depth migration based on an initial seismic velocity model, picking up horizons to create a structure model;

[0009] S2: obtaining well logging velocity data and the structure model, performing stiffness conversion to realize Backus smoothing, and obtaining a smooth equivalent well logging velocity curve;

[0010] S3: using a seismic velocity profile, the smoothed well logging velocity and the structure model, calculating well-seismic velocity error, using a VAR-RBF method to extrapolate the velocity error to a well-free area, and updating and optimizing the velocity model.

[0011] Further, the step S1 specifically comprises: obtaining an initial seismic velocity model and a seismic data gather, performing pre-stack depth migration using a Kirchhoff algorithm, generating a common imaging point gather, stacking the common imaging point gather to obtain a clear imaging profile; performing manual horizon picking and fault delineation on the imaging profile, building a stratigraphic sequence and a fault structure, and forming a structure model.

[0012] Further, the step S2 comprises: based on well logging velocity data of multiple wells and the structure model, performing stiffness conversion to convert well logging velocity and density into elastic parameters, and further calculating a stiffness coefficient of each layer; applying a Backus smoothing algorithm, setting a sliding window in each stratum, and weighting and averaging the stiffness coefficient and the density according to the thickness; obtaining smoothed equivalent well logging P-wave velocity and S-wave velocity by inverse calculation of the equivalent stiffness tensor and the equivalent density; and outputting a smooth velocity curve of each well, which is aligned by layer.

[0013] Further, the step S3 comprises: using the seismic velocity profile obtained in step S1, the smoothed well logging velocity obtained in step S2 and the structure model (as a structure constraint) to perform error calculation, calculating the well-seismic velocity error on each formation layer block of each well, taking the error as a seed point, normalizing to a unified formation level; then performing VAR-RBF extrapolation, constructing an RBF interpolator with a variable sigma asymmetric Gaussian kernel, adaptively adjusting the left and right influence factors according to the spatial distribution of information points, using the structure model to perform hierarchical constraint, and extrapolating the velocity error to the well-free area within each formation block; finally, updating the velocity model, superimposing the extrapolated error value into the initial seismic velocity model to obtain the optimized seismic velocity body, and outputting the extrapolated well-seismic velocity error body and the updated velocity model, which can be used for the next round of pre-stack migration or inversion processing.

[0014] Further, the step S2 specifically comprises:

[0015] S201: converting each physical property of the formation to the stiffness domain to calculate the equivalent stiffness parameter of the formation based on the relationship between the elastic modulus and the velocity of the formation; assuming that a certain formation is composed of n thin layers, each layer has a thickness d i , a density p i , a P-wave velocity V p,i , and a S-wave velocity V s,i , where the layer number i = 1, 2, 3,..., n, and the Lame parameter l can be obtained according to the relationship between the elastic modulus and the velocity as follows:

[0016]

[0017] The shear modulus m is expressed as:

[0018]

[0019] The stiffness coefficient of each layer is expressed as:

[0020]

[0021] S202: based on the stiffness coefficient and the formation thickness, a sliding window H is set, the window includes the current layer point and H layers of adjacent layers around it, and the total thickness D in the window is expressed as:

[0022]

[0023] The equivalent stiffness tensor can be obtained by weighted average:

[0024]

[0025] The weighted average of the equivalent density is:

[0026]

[0027] According to the equivalent stiffness tensor, further calculate the equivalent longitudinal wave velocity and transverse wave velocity:

[0028]

[0029]

[0030] Further, the step S3 specifically comprises:

[0031] S301: Radial basis function RBF interpolation extrapolation algorithm is a spatial data interpolation method with radial basis function as the core, and its formula is:

[0032] J(x) = Σk i *F i (x i ,x) (23)

[0033] Where F i (x i ,x) is the basis function of the center point at x i , k i is the weight of each function, and J(x) is the radial basis interpolation function to be solved; taking the Gaussian kernel as an example, given n seed points (x i ,y i )i=0,1,2…n, all n points are regarded as center points, then the Gaussian kernel function is expressed as:

[0034]

[0035] Where σ is the factor controlling the influence range of the Gaussian function, then:

[0036]

[0037] That is:

[0038]

[0039] In the actual work area, on the basis of the Gaussian kernel radial basis function, the VAR-RBF method with variable influence range factor σ in asymmetric form is proposed; in this method, for any adjacent seed points A, B and C, the variable influence range factor σ is determined according to the distribution characteristics between the seed points, and for the same seed point, the σ values on the left and right sides are asymmetric, so as to fully consider the non-uniformity between the seed points. Similarly, for various seed points (x i ,y i )i=0,1,2…n, the variable σ i,dir factor is calculated according to the distance between adjacent seed points, and its expression is:

[0040] σi,left = x i - x i-1 , sigma i,right = x i+1 - x i (27)

[0041]

[0042] Variable sigma i,dir The Gaussian kernel function constructed by the variable sigma can better adapt to the distribution characteristics of different intervals.

[0043] S302: When the well seismic velocity error is extrapolated by constraining the construction, the stratum needs to be standardized, and the top interface of the same set of stratum is standardized to the same horizontal plane, assuming that the well W i W j Through the stratum h1 and h2, the seismic velocity of each point on the well site is V i W j The seismic velocity of each point on the well site is V si,d V sj,d The well velocity smoothed by the Backus well velocity is V wi,d V wj,d The well seismic velocity error AV i,d = V si,d -V wi,d And AV j,d = V sj,d -V wj,d The well point velocity error is regarded as a seed point, and the VAR-RBF method is used to extrapolate to the entire layer block, so as to realize the updating of the velocity model.

[0044] Compared with the prior art, the advantages of the present application are:

[0045] Improve the accuracy of the velocity model: by means of the variable kernel radius basis function (VAR-RBF) method, the initial seismic velocity model is optimized by accurate extrapolation of the well seismic velocity error, and the calculation accuracy of the velocity field is significantly improved, which provides a more reliable basis for depth domain seismic imaging and reservoir prediction.

[0046] Solve the inconsistency of well seismic velocity: by introducing the Backus smoothing algorithm, the discrete velocity of the well is converted into a smooth and representative continuous velocity body, which solves the inconsistency problem between the well seismic velocity and enhances the rationality of the model in the sparse area of the logging data.

[0047] Innovative VAR-RBF extrapolation method: a VAR-RBF method based on variable sigma asymmetric Gaussian kernel is proposed, which not only overcomes the limitations of traditional RBF interpolation in uneven well pattern distribution, but also fully considers the non-uniformity and local variation trend of seismic data and well point distribution, and enhances the adaptability and accuracy of error extrapolation.

[0048] Introducing geological structure constraints: in the error extrapolation process, the stratigraphic structure model is used as a constraint condition to ensure that the error extrapolation result conforms to the geological law and reduces the unreasonable error propagation in the extrapolation process.

[0049] Suitable for complex geological conditions: the application shows excellent applicability in complex structure areas (such as fault zones and multi-stratum sequence areas), improves the accuracy of reservoir characterization and oil and gas exploration, and has significant engineering application value.

[0050] In summary, the method combines VAR-RBF, high-precision logging data processing and geological structure constraints, and has scientific method, stable results and wide application range, which is of great significance in the field of pre-stack depth migration and velocity modeling. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The main flowchart of the well-seismic velocity error extrapolation method based on VAR-RBF;

[0052] Figure 2 (a), the initial depth domain velocity model used in the embodiment, Figure 2 (b) is the imaging result obtained by using the initial velocity model for Kirchhoff pre-stack depth migration, and the structure is picked up;

[0053] Figure 3 Backus logging data smoothing, wherein the blue is the actual logging curve and the red is the smoothed logging curve;

[0054] Figure 4 The fitting effect diagram of various kernel functions under the condition of more seed points;

[0055] Figure 5 The fitting effect diagram of various kernel functions under the condition of less seed points;

[0056] Figure 6 Stratigraphic standardization diagram, wherein (a) is a stratigraphic original appearance diagram, and (b) is a stratigraphic appearance diagram after standardization processing;

[0057] Figure 7 The well-seismic velocity error plane obtained by fitting;

[0058] Figure 8 The updated velocity model. DETAILED DESCRIPTION

[0059] The specific embodiments of the present application will be described below with reference to the accompanying drawings:

[0060] It should be noted that the structures, proportions, sizes, etc. shown in the present specification are only used to cooperate with the content disclosed in the present specification, so that people skilled in the art can understand and read, and are not used to limit the conditions that can be implemented by the present application. Any modification of structure, change of proportion relationship or adjustment of size, which does not affect the effects that can be produced by the present application and the purposes that can be achieved, should still fall within the scope of the technical content disclosed by the present application.

[0061] At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" referred to in the present specification are only for the convenience of clear description, and are not used to limit the scope of the present application. The change or adjustment of the relative relationship, without substantially changing the technical content, is also considered as the scope of the present application.

[0062] Example 1

[0063] As shown in the present application, a well-seismic velocity error extrapolation method based on VAR-RBF includes the following steps: Figure 1

[0064] S1: Pre-stack depth migration is performed based on an initial seismic velocity model, and horizon is picked to create a structure model.

[0065] S2: The logging velocity is preprocessed based on the Backus smoothing algorithm to solve the inconsistency between the well-seismic velocity.

[0066] S3: Based on the conventional RBF function extrapolation interpolation method, a variable kernel radial basis function (VAR-RBF, Variable Kernel Radial Basis Function) interpolation extrapolation method is proposed. And the VAR-RBF method is used to extrapolate the well-seismic velocity error under the constraint of the structure model, so as to optimize the velocity model.

[0067] Preferably, S1 includes the following sub-steps:

[0068] S11: Kirchhoff pre-stack depth migration is performed using an initial depth domain velocity model, and the migrated gathers are stacked to obtain an imaging profile.

[0069] S12: The migrated imaging profile is manually picked up and faulted to create a structure model.

[0070] Preferably, S2 includes the following sub-steps:

[0071] ​S21: convert each physical property of the formation to the stiffness domain to calculate the equivalent stiffness parameters of the formation based on the relationship between the elastic modulus and the velocity of the formation. Assuming that a certain formation is composed of n thin layers, each layer has a thickness d i , a density p i , a P-wave velocity V p,i , and a S-wave velocity V s,i , where the layer number i = 1, 2, 3,..., n, and the Lame parameter λ can be obtained according to the relationship between the elastic modulus and the velocity as follows:

[0072]

[0073] The shear modulus μ can be expressed as:

[0074]

[0075] The stiffness coefficient of each layer can be expressed as:

[0076]

[0077] S22: based on the stiffness coefficient and the thickness of the formation, a sliding window H is set, which includes the current layer point and its surrounding H adjacent layers, and the total thickness D in the window can be expressed as:

[0078]

[0079] The equivalent stiffness tensor can be obtained by weighted average:

[0080]

[0081] The weighted average of the equivalent density is:

[0082]

[0083] According to the equivalent stiffness tensor, the equivalent P-wave velocity and S-wave velocity can be further calculated:

[0084]

[0085] Preferably, S3 includes the following sub-steps:

[0086] S31: the radial basis function (RBF) interpolation extrapolation algorithm is a spatial data interpolation method based on radial basis function. Its basic principle is to use a set of known points to estimate the value of unknown points by establishing a continuous interpolation function based on radial basis function. It is a symmetric local function that only depends on the Euclidean distance between points, and its formula is:

[0087]

[0088] where Fi (x i ,x) is the base function with the center point at x i k i is the weight of each function, and J(x) is the radial basis interpolation function to be solved. Common radial kernel functions include linear kernel, Gaussian kernel, quadratic polynomial kernel, inverse quadratic polynomial kernel, and thin plate spline kernel, etc. Different kernel function forms have different smoothness and applicable scenarios. Taking the Gaussian kernel as an example, given n seed points (x i ,y i )i=0,1,2…n, all n points are regarded as center points, then the Gaussian kernel function can be expressed as:

[0089]

[0090] where σ is a factor that controls the range of influence of the Gaussian function, then:

[0091]

[0092] that is:

[0093]

[0094] However, in actual work areas, due to uneven well distribution and less logging data, the RBF method with fixed influence range factor σ has certain defects. First, the fitting accuracy is low, and the transformation trend cannot be effectively obtained; second, the fitted curve is not smooth enough, and overfitting or underfitting phenomenon is easy to occur. To overcome the above defects, the VAR-RBF method with variable influence range factor σ in asymmetric form is proposed based on the Gaussian kernel radial basis function. In this method, for any adjacent seed points A, B and C, the variable influence range factor σ is determined according to the distribution characteristics between the seed points, and for the same seed point, the σ values on the left and right sides are asymmetric, so as to fully consider the non-uniformity between the seed points. Similarly, for each seed point (x i ,y i )i=0,1,2…n, the variable σ i,dir factor is calculated according to the distance between adjacent seed points, and its expression is:

[0095] σ i,left =x i -x i-1 , σ i,right =x i+1 -x i (41)

[0096]

[0097] The Gaussian kernel function constructed by the variable σ i,dir can better adapt to the distribution characteristics of different intervals.

[0098] S32: In the extrapolation of well-seismic velocity error constrained by the structure, the stratum needs to be normalized to the same level on the top interface of the same set of stratum. It is assumed that the well W i ,W j Through the stratum h1 and h2, the interlayer W i ,W j The seismic velocity of each point on the well site is V si,d and V sj,d The smoothed logging velocity of Backus is V wi,d and V wj,d Then the well-seismic velocity error ΔV i,d = V si,d -V wi,d and ΔV j,d = V sj,d -V wj,d The well point velocity error is regarded as a seed point, which is extrapolated to the whole layer block by using the VAR-RBF method, so as to realize the updating of the velocity model.

[0099] Embodiment 2:

[0100] This embodiment is applied to the embodiment 1, taking the test model as an example, the well-seismic velocity error extrapolation method based on VAR-RBF is divided into three steps:

[0101] Firstly, the steps in S11 and S12 are implemented by using the seismic shot data and the initial depth domain velocity model, and the structure model is obtained as shown in Figure 2 (b).

[0102] Secondly, the steps in S21 and S22 are implemented by using the current logging data, and the Backus smoothed logging data is obtained as shown in Figure 3 .

[0103] Thirdly, the structure model obtained by S1 is used as a constraint to extrapolate the well-seismic velocity error between the smoothed logging velocity obtained by S2 and the initial depth domain velocity model, and the extrapolation method adopts the VAR-RBF method described in S31, Figure 4 is a schematic diagram of fitting effects of various kernel functions in the case of more seed points, Figure 5 is a schematic diagram of fitting effects of various kernel functions in the case of less seed points, wherein the accuracy of the VAR-RBF extrapolation method is shown in Figure 5 . The steps in S32 are implemented to obtain a new velocity model, wherein the normalization process is shown in Figure 6 , the well-seismic velocity error plane obtained after the extrapolation of the velocity error is shown in Figure 7 , and the velocity model is shown in Figure 8 .

[0104] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In one embodiment, the present application can be implemented in software and can be stored on a computer readable medium, which can include random access memory (RAM), read only memory (ROM), magnetic disk or optical disk, or the like. The software implementation of the present application files can further be transmitted or received over a modem or network connection.

[0105] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks.

[0106] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams block or blocks.

[0108] While the present application has been described with reference to the preferred embodiments thereof, it is to be understood that the application is not limited to the embodiments discussed which can be modified. Changes can be made without departing from the spirit of the application. The present application is thus deemed to cover all modifications of this application that fall within the scope of the appended claims.

[0109] Many other changes and modifications can be made to the application without departing from the spirit thereof. The scope of such application thus not to be taken as limited by the particular embodiments discussed and it is understood that this application is limited only by the scope of the appended claims.

Claims

1. A VAR-RBF based seismic velocity error extrapolation method, characterized in that: The method comprises: S1: Prestack depth migration is performed based on the initial seismic velocity model, and horizons are picked to create a structural model; S2: Obtain logging velocity data and construct a model, perform stiffness conversion to achieve Backus smoothing, and obtain a smooth equivalent logging velocity curve; S3: Calculate the well-seismic velocity error using the seismic velocity profile, smoothed well logging velocity, and structural model. Use the VAR-RBF method to extrapolate the velocity error to the well-free area and update and optimize the velocity model.

2. A VAR-RBF based seismic velocity error extrapolation method according to claim 1, characterized in that: The step S1 specifically includes: obtaining an initial seismic velocity model and seismic data gathers, performing pre-stack depth migration using the Kirchhoff algorithm to generate common imaging point gathers, and superimposing the common imaging point gathers to obtain a clear imaging section; performing manual layer picking and fault characterization on the imaging section to construct a stratigraphic sequence and fault structure to form a structural model.

3. The VAR-RBF-based seismic velocity error extrapolation method according to claim 1, characterized in that: The step S2 includes: performing stiffness conversion based on the logging velocity data and structural model of multiple wells, converting the logging velocity and density into elastic parameters, and further calculating the stiffness coefficient of each layer; applying the Backus smoothing algorithm, setting a sliding window in each layer, and performing weighted averaging of the stiffness coefficient and density according to the thickness; back-calculating the smoothed equivalent logging compressional wave velocity and shear wave velocity from the equivalent stiffness tensor and the equivalent density; and outputting a smoothed velocity curve aligned with each layer on each well.

4. The VAR-RBF-based seismic velocity error extrapolation method according to claim 1, characterized in that: The step S3 includes: using the seismic velocity profile obtained in step S1, the smoothed logging velocity obtained in step S2, and the structural model (as a structural constraint) to perform error calculation, calculating the well seismic velocity error in each stratigraphic block of each well, using the error as a seed point, and normalizing it to a uniform stratigraphic level; then performing VAR-RBF extrapolation, constructing an RBF interpolator with a variable σ asymmetric Gaussian kernel, adaptively adjusting the left and right influencing factors according to the spatial distribution of information points, using the structural model for layered constraints, and extrapolating the velocity error to the well-free area within each stratigraphic block; finally, updating the velocity model, superimposing the extrapolated error value onto the initial seismic velocity model to obtain an optimized seismic velocity volume, outputting the extrapolated well seismic velocity error volume and the updated velocity model, which can be used for the next round of prestack migration or inversion processing.

5. The VAR-RBF-based seismic velocity error extrapolation method according to claim 3 is characterized in that: The step S2 specifically includes: S201: Based on the relationship between the elastic modulus and velocity of the formation, the physical properties of the formation are converted into the stiffness domain to calculate the equivalent stiffness parameters of the formation; assuming that a certain formation is composed of n thin layers, each layer has a thickness of d i , the density is ρ i , the longitudinal wave velocity is V p,i , the shear wave velocity is V s,i , where layer number i = 1, 2, 3, ..., n, according to the relationship between elastic modulus and velocity, the Lamé parameter λ can be obtained as follows: The shear modulus μ is expressed as: The stiffness coefficient of each layer is expressed as: S202: Based on the stiffness coefficient and the formation thickness, a sliding window H is set. The window includes the current layer point and several adjacent layers around it, totaling H layers. The total thickness D in the window is expressed as: Then the equivalent stiffness tensor can be obtained by weighted averaging: The weighted average of the equivalent density is: According to the equivalent stiffness tensor, the equivalent longitudinal wave velocity and shear wave velocity are further calculated:

6. A VAR-RBF based seismic velocity error extrapolation method according to claim 4, characterized in that: The step S3 specifically includes: S301: The radial basis function (RBF) interpolation and extrapolation algorithm is a spatial data interpolation method based on the radial basis function. Its formula is: J(x)=∑k i *F i (x i ,x) (9) Among them F i (x i ,x) is the center point at x i The basis function at k i is the weight of each function, J(x) is the radial basis interpolation function; taking the Gaussian kernel as an example, it is known that n seed points (x i ,y i )i=0,1,2...n, taking n points as the center point, the Gaussian kernel function is expressed as: Where σ is the factor that controls the influence range of the Gaussian function, then: Right now: In the actual work area, this method proposes a VAR-RBF method with an asymmetric form of variable influence range factor σ based on the Gaussian kernel radial basis function. In this method, for any adjacent seed points A, B and C, the variable influence range factor σ is determined according to the distribution characteristics between the seed points. For the same seed point, the σ values ​​on the left and right sides are asymmetric to fully consider the non-uniformity between the seed points. Similarly, for various sub-points (x i ,y i )i=0,1,2…n, variable σ is calculated according to the distance between adjacent seed points i,dir The factor is expressed as: s i,left =x i -x i-1 ,s i,right =x i+1 -x i (13) Using variable σ i,dir The constructed Gaussian kernel function can better adapt to the distribution characteristics of different intervals; S302: When extrapolating the seismic velocity error with the structure as the constraint, the formation needs to be normalized and the top interface of the same formation is normalized to the same horizontal plane. Assume that well W i ,W j Through the strata h1 and h2, between the layers W i ,W j The seismic velocity at each point on the well is V si,d With V sj,d , the smoothed Backus logging speed is V wi,d With V wj,d , then the well seismic velocity error ΔV can be obtained i,d =V si,d -V wi,d and ΔV j,d =V sj,d -V wj,d , the velocity error of the well point is regarded as the seed point and extrapolated to the entire layer block using the VAR-RBF method to achieve velocity model update.

Images