Elastic parameter prediction method based on azimuthal anisotropy

Abstract

The invention provides an elastic parameter prediction method based on azimuthal anisotropy, and the method comprises the steps: S1, obtaining five-dimensional seismic data, and carrying out the optimization processing of the five-dimensional seismic data; s2, according to the characteristics of actual seismic data, the five-dimensional seismic data are divided into directions and incident angles; s3, extracting an intercept attribute P and a gradient attribute G of the five-dimensional seismic data in each direction; s4, performing linear fitting on the logging data and the seismic attributes in multiple directions by adopting a multi-correlation filtering attribute conversion method to obtain a fitting result of the elastic parameters and the seismic attributes; and S5, predicting an elastic parameter according to the fitting result. And the reservoir prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the fields of oil and gas reservoir development and mineral engineering, and particularly to a method for predicting elastic parameters based on azimuthal anisotropy. Background Art

[0002] OVT-domain azimuthal AVO seismic data contains three-dimensional spatial coordinates as well as rich azimuth and incident angle (offset) information, and can better analyze the variation information of kinematic and dynamic parameters such as travel time, velocity, amplitude, frequency, and phase of seismic waves when propagating in underground media with spatial directions. The incident angle information in seismic data is correlated with the lithology, fluid composition, and elastic properties of the target geological body, while the azimuth information is related to the development characteristics, elastic properties, and anisotropic parameters of fractures and cracks in the formation. Using OVT-domain azimuthal AVO seismic gathers and their attributes, seismic data interpretation can be carried out, including structural interpretation, stratigraphic interpretation, lithologic interpretation, fluid interpretation, fracture identification, and in-situ stress research. By using the difference in seismic response information at multiple incident angles, AVO attributes and elastic parameter information for identifying formation lithology and fluid characteristics can be obtained.

[0003] Wide-azimuth seismic exploration refers to a seismic exploration method in which the ratio of the lateral to longitudinal arrangement in the observation system is greater than 0.5. Compared with traditional narrow-azimuth seismic exploration, wide-azimuth seismic exploration has many advantages: wide-azimuth exploration can increase exploration illumination and obtain a more complete seismic wave field; wide-azimuth seismic data can study the variation of amplitude with offset and azimuth (Amplitude variation with offset and azimuth, AVOAZ) and the variation of formation velocity with azimuth (Velocity variation with angle, VVA), thereby enhancing the ability to identify underground anisotropy and fluid distribution.

[0004] Five-dimensional seismic data provides high-dimensional information such as azimuth and incident angle. Compared with traditional seismic data, five-dimensional seismic data has many advantages: five-dimensional seismic data can perform omnidirectional observation, increase acquisition illumination, and obtain a more complete seismic wave field; five-dimensional seismic data can study the variation of amplitude with offset and azimuth and the variation of formation velocity with azimuth, thereby enhancing the ability to identify faults, fractures, formation lithology, and fluids; five-dimensional seismic data has higher steep-dip imaging ability and richer amplitude imaging information. However, how to explore and innovate in theory, method, and technology to extract the extremely rich information in five-dimensional data, how to fully consider the important azimuth and offset information in wide-azimuth seismic data, and how to better utilize the rich azimuthal anisotropy information in five-dimensional seismic data. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a method for predicting elastic parameters based on azimuthal anisotropy that overcomes the above problems or at least partially solves the above problems.

[0006] According to one aspect of the present invention, there is provided a method for predicting elastic parameters based on azimuthal anisotropy, the prediction method comprising:

[0007] Step S1: Obtain five-dimensional seismic data and optimize the five-dimensional seismic data;

[0008] Step S2: Divide the five-dimensional seismic data into azimuths and incident angles according to the characteristics of the actual seismic data;

[0009] Step S3: Extract the intercept attribute P and the gradient attribute G of the five-dimensional seismic data in each azimuth;

[0010] Step S4: Use the multi-correlation filtering attribute conversion method to linearly fit the well logging data with the seismic attributes in multiple azimuths to obtain a fitting result;

[0011] Step S5: According to the fitting result, use the seismic attributes in each azimuth of the five-dimensional seismic data to obtain the elastic parameters of seismic prediction.

[0012] Optionally, the fitting result specifically includes: the linear relationship between the elastic parameters and the seismic attributes.

[0013] Optionally, the elastic parameters specifically include: the longitudinal wave impedance and the shear wave impedance of the seismic wave.

[0014] Optionally, in step S1: the optimization process of the five-dimensional seismic data specifically includes:

[0015] Predict coherent signals through seismic data denoising, and the method for predicting elastic parameters based on azimuthal anisotropy suppresses random noise;

[0016] Use the inverse Q filtering method to process the actual seismic data, compensate for amplitude attenuation and frequency loss, improve the phase characteristics of the record, thereby improving the continuity of the event axis, increasing the energy of weak reflection waves, and the signal-to-noise ratio and resolution of the seismic data;

[0017] On the stacked record, establish a model trace using forward and backward prediction, and obtain the residual normal moveout according to the maximum similarity coefficient criterion;

[0018] Adopt a Butterworth filter for filtering and interpolation to complete high-precision dynamic stretching correction, thereby eliminating the influence of residual normal moveout.

[0019] Optionally, in step S2: dividing the five-dimensional seismic data into azimuths and incident angles according to the characteristics of the actual seismic data specifically includes:

[0020] According to the characteristics of actual seismic data, the five-dimensional seismic data is divided by azimuth and incident angle to obtain the seismic data after stacking by azimuth and angle.

[0021] Optionally, step S3: Extracting the intercept attribute P and gradient attribute G under each azimuth of the five-dimensional seismic data specifically includes:

[0022] Extract the intercept attribute P and gradient attribute G under each azimuth of the five-dimensional seismic data;

[0023] Combined with geological, logging and seismic data, through fine interpretation, the prestack gather data and the method of AVO attribute crossplot are used for description.

[0024] Optionally, step S1: Optimizing the five-dimensional seismic data specifically includes:

[0025] Seismic data denoising processing;

[0026] Coherent signals are predictable in the x-direction of the domain, while random noise is not. Utilizing the similarity between traces, a prediction filter factor at a certain fixed frequency is designed and calculated; then the prediction filter factor is convolved with each trace to predict the coherent signal and suppress the random noise;

[0027] Use filtering to process the actual seismic data;

[0028] Seismic data resolution improvement processing;

[0029] The actual seismic data is processed using the inverse Q filtering method;

[0030] Seismic data event flattening processing;

[0031] In the processing of actual seismic data, the residual normal moveout method is used to remove the residual normal moveout;

[0032] On the stacked record, a model trace is established using forward and backward prediction, the residual normal moveout is obtained according to the maximum similarity coefficient criterion, and a Butterworth filter is used for filtering and interpolation to complete high-precision dynamic stretching correction.

[0033] Optionally, step S4: Using the multi-correlation filtering attribute conversion method to linearly fit the logging data with the seismic attributes in multiple azimuths specifically includes:

[0034] Elastic parameters, as quantitative characteristics of reservoir physical properties, are related to seismic parameters;

[0035] Assume that the seismic parameters and elastic parameters are two random processes. Predicting the elastic parameters from the seismic parameters is reduced to a linear filtering problem of one random process;

[0036] The random process serves as the input of the filter, and the output is obtained through the filter, which is the estimated value of the reservoir parameters, realizing the direct conversion from seismic parameters to elastic parameters;

[0037] To design the filter, a filtering factor is selected to minimize the error between the estimated value of the reservoir parameters and the true value of the parameters in the sense of least squares. This process is called the correlation filtering for reservoir parameter conversion.

[0038] Optionally, the step S4: using the multi-correlation filtering attribute conversion method to linearly fit the logging data with seismic attributes in multiple azimuths further includes:

[0039] Directly predicting the reservoir parameters based on seismic attributes in the case of lacking well data;

[0040] The prediction effect depends on whether the seismic attributes selected to predict the reservoir parameters can truly reflect the changes in the reservoir parameters.

[0041] Optionally, the generalized form of the convolution operator is:

[0042] L = ω 0 + ω 1 * A 1 + ω 2 * A 2 + ω 3 * A 3 ;

[0043] where, * represents convolution, and ω i is an operator with a specific length.

[0044] An elastic parameter prediction method based on azimuthal anisotropy provided by the present invention, the prediction method includes: step S1: obtaining five-dimensional seismic data and optimizing the five-dimensional seismic data; step S2: dividing the five-dimensional seismic data into azimuths and incident angles according to the characteristics of the actual seismic data; step S3: extracting the intercept attribute P and gradient attribute G of each azimuth of the five-dimensional seismic data; step S4: using the multi-correlation filtering attribute conversion method to linearly fit the logging data with seismic attributes in multiple azimuths to obtain a fitting result; step S5: according to the fitting result, using the seismic attributes of each azimuth of the five-dimensional seismic data to fit and obtain the elastic parameters predicted by the earthquake. Using the seismic attributes of each azimuth of the five-dimensional seismic data by the correlation filtering attribute conversion method to obtain the elastic parameter information predicted by the earthquake, so as to achieve the purpose of improving the reservoir prediction accuracy.

[0045] The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the content of the specification. And in order to make the above and other objects, features and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention are specifically given below. Brief Description of the Drawings

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0047] Figure 1 It is a flowchart of a method for predicting elastic parameters based on azimuthal anisotropy provided by an embodiment of the present invention;

[0048] Figure 2 It is an effect diagram of processing actual seismic data using filtering in a specific embodiment of the present invention, including the comparison of the effects before and after processing and the noise components removed by this method. As the random noise is suppressed, the signal-to-noise ratio is improved;

[0049] Figure 3 It is a comparison effect diagram before and after processing actual seismic data using the inverse Q filtering method in a specific embodiment of the present invention. The energy of weak reflection waves, the signal-to-noise ratio and resolution of seismic data are improved;

[0050] Figure 4 It is a comparison effect diagram before and after performing residual normal moveout processing on actual seismic data in a specific embodiment of the present invention. The event is flattened and the correction effect of the residual static correction amount is significant;

[0051] Figure 5 It is the seismic data after azimuthal and angular stacking in a specific embodiment of the present invention. The selected angle range is: 6 - 38°, and the selected azimuth range is: 11 - 56°, 123 - 168°. The azimuthal and angular division ranges selected for the reservoir section are: azimuth division: 11 - 34°; 34 - 56°; 123 - 146°; 146 - 168°, and angle division: 6 - 16°; 17 - 27°; 28 - 38°;

[0052] Figure 6 It is the intercept attribute and gradient attribute diagrams obtained from seismic data with different azimuths and incident angles of five-dimensional seismic data in a specific embodiment of the present invention, including the intercept attribute and gradient attribute diagrams at azimuth angles of 23°, 45°, 135°, and 157°;

[0053] Figure 7Effect diagram of simulating the target logging curve by a linear equation at each time sampling point under the assumption of three relevant filtering attributes;

[0054] Figure 8 Comparison between the target logging (left) and seismic attributes (right). It can be seen that the frequency components of the target logging curve are much higher than those of the seismic attributes, and there are differences in frequency components between them, which need to be eliminated by a convolution operator;

[0055] Figure 9 Correlation between seismic attributes and the target logging curve using a 5-point convolution operator. The 5-point convolution operator in the figure is closely related to the seismic wavelet;

[0056] Figure 10 In a specific embodiment of the present invention, 5 groups of seismic attribute data are selected from the intercept and gradient attributes obtained in four directions. The longitudinal wave impedance in the logging data is fitted with these 5 groups of seismic attribute data using the relevant filtering attribute conversion method, and the fitting result diagram of the longitudinal wave impedance and these five groups of seismic attributes;

[0057] Figure 11 Correlation coefficient diagram between the logging longitudinal wave impedance and the result obtained by attribute conversion in a specific embodiment of the present invention. The correlation coefficient is 0.85;

[0058] Figure 12 In a specific embodiment of the present invention, 5 groups of seismic attribute data are selected from the intercept and gradient attributes obtained in four directions. The shear wave impedance in the logging data is fitted with these 5 groups of seismic attribute data using the relevant filtering attribute conversion method, and the fitting result diagram of the longitudinal wave impedance and these five groups of seismic attributes;

[0059] Figure 13 Correlation coefficient diagram between the logging shear wave impedance and the result obtained by attribute conversion in a specific embodiment of the present invention. The correlation coefficient is 0.83;

[0060] Figure 14 Prediction result diagram of the longitudinal wave impedance using the fitting result of the logging data and the five-dimensional seismic data in a specific embodiment of the present invention;

[0061] Figure 15 Prediction result diagram of the shear wave impedance using the fitting result of the logging data and the five-dimensional seismic data in a specific embodiment of the present invention. Detailed implementation manner

[0062] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0063] The terms "comprising" and "having" and any variations thereof in the description, claims, and drawings of the present invention are intended to cover non-exclusive inclusion. For example, including a series of steps or units.

[0064] The technical solutions of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

[0065] As Figure 1 shown, Figure 1 is a flowchart of a specific embodiment of the elastic parameter prediction method based on azimuthal anisotropy of the present invention.

[0066] In step 1, by optimizing the 3D seismic data, the signal-to-noise ratio and resolution of the data are improved:

[0067] ① Denoising processing of seismic data

[0068] Coherent signals are predictable in the x-direction of the frequency domain, while random noise is not. The similarity between traces can be utilized to design and obtain a prediction filtering factor at a certain fixed frequency; then the prediction filtering factor is convolved with each trace to predict the coherent signal and suppress random noise, thereby improving the signal-to-noise ratio. Using filtering to process the actual seismic data, the results are as Figure 2 .

[0069] ② Resolution improvement processing of seismic data

[0070] Inverse Q filtering is a technique for compensating the absorption attenuation effect of the earth. It can not only compensate for amplitude attenuation and frequency loss, but also improve the phase characteristics of the record, thereby improving the continuity of the event axis, increasing the energy of weak reflection waves, and the signal-to-noise ratio and resolution of seismic data. Using the inverse Q filtering method to process the actual seismic data, the processing results are as Figure 3 .

[0071] ③ Flattening processing of seismic data event axis

[0072] In actual seismic data processing, since there is always a certain amount of residual normal moveout in the data after normal moveout correction, which seriously affects the inversion effect, it is necessary to adopt the method of residual normal moveout correction to remove the residual normal moveout. On the stacked record, a model trace is established by forward and backward prediction, and the residual normal moveout is obtained according to the maximum similarity coefficient criterion. Then, a Butterworth filter is used for filtering interpolation to complete high-precision dynamic stretching correction, thereby eliminating the influence of residual normal moveout. Residual normal moveout correction is performed on the actual seismic profile to obtain Figure 4 。

[0073] The process proceeds to step 2.

[0074] In step 2, according to the characteristics of the actual seismic data, the five-dimensional seismic data is divided by azimuth and incident angle. Since the offset distribution is uneven and the range is narrow, the stacked trace gather is preferably selected. The finally selected angle range is: 6 - 38°, and the selected azimuth range is: 11 - 56°, 123 - 168°. The divided azimuth and angle ranges selected according to the reservoir section are: azimuth division: 11 - 34°; 34 - 56°; 123 - 146°; 146 - 168°, and angle division: 6 - 16°; 17 - 27°; 28 - 38°. Figure 5 is the seismic data after stacked by azimuth and angle. The process proceeds to step 3.

[0075] In step 3, the intercept attribute (P) and gradient attribute (G) attributes under each azimuth of the five-dimensional seismic data are extracted. The mutual relationship between AVO information, reservoir lithology, and hydrocarbon-bearing property is the AVO attribute analysis and interpretation technology. Through research and comparison, geological meanings are assigned to different AVO attributes. AVO analysis must be combined with the geological characteristics and geophysical features of the study area, combined with geological, logging, and seismic data, and through fine interpretation, methods such as prestack trace gather data and AVO attribute crossplot are used for description.

[0076] The basic attributes of AVO are mainly the intercept section (P) and the gradient section (G).

[0077] ① Intercept attribute (P)

[0078] For the Aki-Richards approximate formula, let:[[]]

[0079]

[0080] P can also be in another form:

[0081]

[0082] P is the intercept in the Aki-Richards approximate formula. The P section can be regarded as the case where the slope G is equal to zero, that is, the zero-offset reflection section, which can reflect the change of wave impedance above and below the interface.

[0083] ②Gradient attribute (G)

[0084] When the incident angle is less than 30°, the Aki-Richards approximation simplifies to:

[0085]

[0086] Let:

[0087]

[0088] G is the slope in the Aki-Richards approximation. The G profile is called the gradient profile and is a function of the longitudinal wave velocity change rate, the shear wave velocity change rate, and the density change rate, and also represents the change rate of the amplitude with the incident angle.

[0089] Calculate the intercept attribute and the gradient attribute for the seismic data with different azimuths and different incident angles of the five-dimensional seismic data as Figure 6 shown. The process proceeds to step 4.

[0090] In step 4, the correlation filtering attribute conversion method is used to fit the well logging data with the seismic attributes in multiple azimuths.

[0091] As quantitative characteristics of reservoir physical properties, elastic parameters have a certain relationship with seismic parameters and they are correlated. Assuming that the seismic parameters and elastic parameters are two random processes, predicting elastic parameters from seismic parameters can be reduced to a linear filtering problem of a random process. The random process is the input of the filter, and the output obtained through the filter is the estimated value of the reservoir parameters, realizing the direct conversion from seismic parameters to elastic parameters. To design such a filter, select the filtering factor to minimize the error between the estimated value of the reservoir parameters and the true value of the parameters in the least squares sense. This process is called the correlation filtering of reservoir parameter conversion. It directly predicts reservoir parameters based on seismic attributes in the absence of well data. The prediction effect depends on whether the seismic attributes selected to predict reservoir parameters can truly reflect the changes in reservoir parameters, that is, the correlation between the seismic attributes and reservoir parameters.

[0092] Suppose there are 3 attributes, as Figure 7 shown. At each time sampling point, the target well logging curve can be simulated by a linear equation.

[0093] L(t) = ω 0 + ω 1 A 1 (t) + ω 2 A 2 (t) + ω 3 A 3 (t)

[0094] The weights in the formula can be obtained by minimizing the mean square prediction error:

[0095]

[0096]

[0097] Similar to the single-attribute case, the mean square prediction error calculated using the derived weights is as shown in the above formula, which can be used as a measure of the goodness of fit of this transformation. At this time, the coordinates are the predicted well log values, and the coordinates are the actual well log values.

[0098] However, it can be seen from Figure 8 that the frequency components of the target well log curve are much higher than those of the seismic attribute, and there are differences in their frequency components, which need to be eliminated using a convolution operator. Therefore, point-by-point correlation may not be optimal. An alternative method is to assume that each sample point of the target well log curve is associated with a group of adjacent sample points of the seismic attribute, as shown in Figure 9 shown.

[0099] The classical convolution model in geophysics also requires the use of a convolution operator. For example, if the well log curve is an acoustic impedance curve, then the 5-point convolution operator in Figure 9 is closely related to the seismic wavelet. Generally speaking, for any well log characteristic, each sample point of the well log curve is associated with a continuous segment of seismic samples.

[0100] The generalized form containing the convolution operator is:

[0101] L = ω 0 + ω 1 *A 1 + ω 2 *A 2 + ω 3 *A 3

[0102] In the formula, * represents convolution, and ω i is an operator of a specific length. It should be noted that the number of coefficients has now increased to: (the number of attributes × operator length) + 1. Similarly, the coefficients of the operator can be obtained by minimizing the mean square prediction error:

[0103]

[0104] Select 5 groups of seismic attribute data from the intercept and gradient attributes obtained in four directions. Use the correlation filtering attribute conversion method to fit the longitudinal wave impedance in the well log data with these 5 groups of seismic attribute data to obtain the fitting results of the longitudinal wave impedance with these five groups of seismic attributes. The fitting results are as shown in Figure 10 shown. The correlation coefficient between the longitudinal wave impedance obtained from the well log data and the longitudinal wave impedance obtained by fitting with the seismic attributes is as shown inFigure 11 As shown, it is 0.85.

[0105] Select 5 groups of seismic attribute data from the intercept and gradient attributes in the four acquired directions. Use the relevant filtering attribute conversion method to fit the shear wave impedance in the logging data with these 5 groups of seismic attribute data to obtain the fitting results of the shear wave impedance and these five groups of seismic attributes. The fitting results are as Figure 12 shown. The correlation coefficient between the P-wave impedance obtained from the logging data and the P-wave impedance obtained by fitting with the seismic attributes is as Figure 13 shown, which is 0.83. The process proceeds to step 5.

[0106] Use the fitting results of the logging data and the five-dimensional seismic data, as Figure 14 shown, for the predicted result diagram of the P-wave impedance;

[0107] Use the fitting results of the logging data and the five-dimensional seismic data, as Figure 15 shown, for the predicted result diagram of the shear wave impedance.

[0108] In step 5, predict the elastic parameters using the fitting results of the relevant filtering attribute conversion method. The process ends.

[0109] Beneficial effects: The elastic parameter prediction method based on azimuthal anisotropy in the present invention, on the basis of data optimization and gather analysis to obtain the stacked gather, uses the seismic attributes in each azimuth of the five-dimensional seismic data by the relevant filtering attribute conversion method to obtain seismic elastic parameter information, achieving the purpose of improving the reservoir prediction accuracy.

[0110] The above specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are only the specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting elastic parameters based on azimuthal anisotropy, characterized in that: The prediction method comprises: Step S1: acquiring five-dimensional seismic data, and optimizing the five-dimensional seismic data; Step S2: dividing the five-dimensional seismic data into azimuth and incident angle according to the characteristics of the actual seismic data; Step S3: extracting the intercept attribute P and the gradient attribute G at each azimuth of the five-dimensional seismic data; Step S4: linearly fitting the well logging data with seismic attributes in multiple directions using a multi-correlation filtering attribute conversion method to obtain fitting results of elastic parameters and seismic attributes; Step S5: According to the fitting result, the elastic parameters for earthquake prediction are obtained by using the seismic attributes of each direction of the five-dimensional seismic data.

2. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The fitting result specifically includes: a linear relationship between elastic parameters and seismic attributes.

3. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The elastic parameters specifically include: longitudinal wave impedance and transverse wave impedance of earthquake.

4. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S1: optimizing the five-dimensional seismic data specifically includes: The coherent signal is predicted by de-noising the seismic data, and the azimuthal anisotropic elastic parameter prediction method suppresses the random noise; Use the inverse Q filtering method to process actual seismic data, compensate for amplitude attenuation and frequency loss, improve the phase characteristics of the record, thereby improving the continuity of the event axis, increasing the energy of weak reflection waves and the signal-to-noise ratio and resolution of seismic data; The model trace is established by using forward and backward prediction on the stacked records, and the residual dynamic correction is calculated according to the maximum similarity coefficient criterion; The Butterworth filter is used for filtering interpolation to complete high-precision dynamic stretch correction, thereby eliminating the influence of residual dynamic correction.

5. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S2: dividing the azimuth and incident angle of the five-dimensional seismic data according to the characteristics of the actual seismic data specifically includes: According to the characteristics of actual seismic data, the five-dimensional seismic data is divided into azimuth and incident angle to obtain seismic data after azimuth and angle superposition.

6. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S3: extracting the intercept attribute P and the gradient attribute G of the five-dimensional seismic data at each azimuth specifically includes: Extract the intercept attribute P and gradient attribute G attributes at each azimuth of the five-dimensional seismic data; Combining geological, logging and seismic data, the description is carried out through detailed interpretation using pre-stack gather data and AVO attribute intersection diagram methods.

7. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S1: optimizing the five-dimensional seismic data specifically includes: Seismic data denoising; The coherent signal is predictable in the x direction of the domain, while the random noise is unpredictable. By using the similarity between channels, the prediction filter factor at a certain fixed frequency is designed and calculated. Then the prediction filter factor is convolved with each channel to predict the coherent signal and suppress the random noise. Use filtering to process actual seismic data; Seismic data resolution enhancement processing; Using an inverse Q filtering method to process the actual seismic data; Seismic data event leveling processing; In actual seismic data processing, the residual dynamic correction method is used to remove the residual dynamic correction amount; The model trace is established by using forward and backward prediction on the stacked records, the residual dynamic correction is obtained according to the maximum similarity coefficient criterion, and the Butterworth filter is used for filtering interpolation to complete the high-precision dynamic stretch correction.

8. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S4: using a multi-correlation filter attribute conversion method to linearly fit the well logging data with seismic attributes in multiple directions specifically includes: Elastic parameters, as quantitative characteristics of reservoir physical properties, are related to seismic parameters; Assuming that the seismic parameters and elastic parameters are two random processes, the prediction of elastic parameters from seismic parameters can be reduced to a linear filtering problem of a random process. The random process is the input of the filter, and the output obtained through the filter is the estimated value of the reservoir parameters, which realizes the direct conversion from seismic parameters to elastic parameters; To design the filter, the filter factor is selected so that the error between the reservoir parameter estimate and the true value of the parameter is minimized in the least square sense. This process is called correlation filtering of reservoir parameter conversion.

9. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The step S4: linearly fitting the well logging data with seismic attributes in multiple directions using a multi-correlation filtering attribute conversion method also includes: In the absence of well data, reservoir parameters are predicted directly from seismic attributes; The prediction effect depends on whether the seismic attributes selected to predict reservoir parameters can truly reflect the changes in reservoir parameters.

10. The elastic parameter prediction method based on azimuthal anisotropy according to claim 1, characterized in that: The generalized form of the convolution operator is: L=ω0+ω1*A1+ω2*A2+ω3*A3; Among them, * represents convolution, ω i is an operator of a specific length.

Images