A method and system for earthquake velocity prediction based on big data

Through big data methods, a seismic historical and geological data is processed, a seismic velocity prediction model is constructed, which solves the problem of insufficient accuracy and real-time prediction of earthquake velocity in the existing technology, and achieves efficient intelligent prediction.

CN118604886BActive Publication Date: 2024-11-15SEISMOLOGICAL BUREAU OF GANSU PROVINCE CHINA EARTHQUAKE ADMINISTRATION
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410675136.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-29
Publication Date
2024-11-15
Estimated Expiration
2044-05-29

AI Technical Summary

Technical Problem

The existing seismic velocity prediction methods have problems such as difficult data acquisition and integration, uneven data quality, and insufficient prediction accuracy and real-time performance.

Method used

Using a big data-based method, by acquiring and preprocessing the historical and geological data of the earthquake, conducting full waveform inversion, parameter selection, velocity change exploration and model optimization, constructing an earthquake velocity prediction model, using random forest algorithms, generating adversarial networks and Algorithms such as recurrent neural networks make predictions.

Benefits of technology

It improves the accuracy and real-time prediction of earthquake velocity, realizes intelligent prediction of earthquake velocity, adapts to data needs of different standards, and is universal.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118604886B_ABST
    Figure CN118604886B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for predicting earthquake velocity based on big data, including obtaining historical data and geological data of earthquakes, preprocessing the historical data and the geological data, performing full waveform inversion on the historical data according to an inversion model to obtain propagation parameters, selecting the propagation parameters according to the influence degree to obtain rapid change parameters, using geophysical exploration to perform velocity change exploration on the geological data to obtain variable data, correcting the earthquake velocity according to the rapid change parameters and the variable data to obtain variable velocity, constructing an earthquake velocity prediction model according to the variable velocity, optimizing the earthquake velocity prediction model according to the velocity error, inputting the data to be predicted into the optimized earthquake velocity prediction model, and outputting the predicted velocity. This method can not only improve the accuracy of earthquake velocity prediction based on big data, but also has good interpretability, and can be directly applied to an earthquake velocity prediction system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of prediction, and in particular to a method and system for predicting earthquake velocity based on big data. Background Art

[0002] Earthquakes, as a highly destructive natural disaster, have always brought immeasurable loss of life and property to human society. Although the current level of science and technology has made significant progress, we still cannot directly prevent earthquakes. However, accurate earthquake prediction, especially the prediction of earthquake velocity, can provide us with precious escape time, thereby effectively reducing the losses caused by earthquake disasters.

[0003] Traditional earthquake prediction methods mostly rely on the experience of seismologists and limited earthquake data, and their prediction accuracy and real-time performance are greatly limited. In recent years, with the rapid development of big data technology, its application in various fields has achieved remarkable results. Especially in the field of earthquake prediction, the application of big data technology provides new possibilities for earthquake prediction.

[0004] There are still many challenges in the current earthquake velocity prediction method. On the one hand, it is difficult to obtain and integrate earthquake data, and the data quality is uneven; on the other hand, the existing earthquake prediction model still needs to improve the prediction accuracy and real-time performance of earthquake velocity. Therefore, this study aims to construct an earthquake velocity prediction method based on big data to improve the accuracy and real-time performance of earthquake prediction and provide strong support for earthquake disaster prevention and mitigation. Summary of the invention

[0005] The purpose of the present invention is to provide a method for earthquake velocity prediction based on big data.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The present invention comprises the following steps:

[0008] Acquire historical earthquake data and geological data, and pre-process the historical data and the geological data;

[0009] Performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapidly changing parameters;

[0010] Performing velocity change exploration on the geological data by geophysical exploration to obtain variable factor data, and correcting the seismic velocity according to the velocity change parameter and the variable factor data to obtain variable velocity;

[0011] Constructing a seismic velocity prediction model according to the variable velocity, and optimizing the seismic velocity prediction model according to the velocity error;

[0012] The data to be predicted is input into the optimized seismic velocity prediction model, and the predicted velocity is output.

[0013] Furthermore, the method of performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters includes:

[0014] The finite element method is used to perform gravity gradient forward modeling and calculate the gravitational potential of the observation point:

[0015]

[0016] Where the gravitational constant is H, the horizontal coordinate is q and the vertical coordinate is u, the gravity value generated by the underground medium at the observation point (q,u) is Δg(q,u), the horizontal coordinate of the observation point after segmentation is λ, the vertical coordinate of the observation point after segmentation is τ, and the underground medium density of the observation point coordinate after segmentation is φ(λ,τ);

[0017] Calculate the vertical component of the gravity anomaly and the vertical component of the gravity gradient:

[0018]

[0019]

[0020] The vertical component of the gravity anomaly is g u , the vertical component of the gravity gradient tensor is K uu ;

[0021] Calculate the kernel function matrix:

[0022]

[0023] The kernel function matrix is ​​S, and the simulated seismic signal is calculated:

[0024]

[0025] The simulated earthquake signal at time d The stability constant is a, and the weighting parameter is The initial seismic signal is F o , the natural water length is e, and the objective function is constructed, the expression is:

[0026]

[0027] F(d)=f(d)+ωE{f(d)}

[0028] The objective function is The seismic signal at time d is F(d), the real part of the seismic signal F(d) at time d is f(d), the imaginary part of the seismic signal F(d) at time d is E{f(d)}, the imaginary unit is ω, and the real part of the simulated seismic signal at time d is The imaginary part of the simulated seismic signal F(d) at time d is

[0029] Compute the inversion gradient:

[0030]

[0031] The inversion gradient is ε, and the observed value of the gravity gradient is The simulated value of the gravity gradient is

[0032] The inversion gradient is used to optimize the inversion model according to the objective function value until the objective function value is minimized and then the optimization is stopped;

[0033] The historical data are input into the optimized inversion model and the propagation parameters are output.

[0034] Furthermore, the method of selecting the propagation parameter according to the influence to obtain the rapidly changing parameter includes:

[0035] Normalize the propagation parameters and construct a decision matrix with the normalized propagation parameters;

[0036] Calculate the importance of propagation parameters:

[0037]

[0038] The adjustment factor of the sth propagation parameter is p s , the correlation between the sth propagation parameter and the seismic velocity is The number of propagation parameters is The correlation between the cth propagation parameter and the seismic velocity is

[0039] Calculate the decision value based on the normalized propagation parameter and importance:

[0040]

[0041] The normalized sth propagation parameter is G s , the decision value of the sth propagation parameter is c s ;

[0042] Construct a standard decision matrix based on the decision value and calculate the grey correlation degree of positive and negative ideal solutions:

[0043]

[0044]

[0045] The grey correlation degree of the positive ideal solution in the i-th row and y-th column is The grey relational degree of the negative ideal solution in row i and column y is The importance of the propagation parameters in column y is The decision value of row i and column y is c iy , the propagation parameter of the i-th row is z i , the propagation parameter z i The probability of occurrence is h(z i ), the resolution coefficient is ζ;

[0046] Calculate the impact:

[0047]

[0048] The first weight is The second weight is The negative decision value of column y is The positive decision value of column y is The positive influence of the propagation parameter in the i-th row is

[0049] The influence degrees are sorted in descending order, and the propagation parameters with influence degrees greater than 0.507 are output as rapidly changing parameters.

[0050] Furthermore, the method of obtaining variable velocity by correcting the seismic velocity according to the velocity variation parameter and the variable factor data comprises:

[0051]

[0052] The speed at time d+1 is The earthquake velocity at time d is V d , the jth speed-changing parameter is The weight of the jth fast-changing parameter is The number of rapidly changing parameters is σ and the decay rate is ξ.

[0053] Furthermore, the method for constructing a seismic velocity prediction model according to the variable velocity includes:

[0054] The objective function is constructed according to the variable speed, and the expression is:

[0055]

[0056] The objective function is The speed at time d is The predicted speed at time d is The loss function is The adjustment constant is a, and the control weight is

[0057] Constructing a seismic velocity prediction model based on the objective function. The seismic velocity prediction model includes a random forest algorithm, a generative adversarial network, a recurrent neural network algorithm, and a particle swarm optimization algorithm;

[0058] The random forest algorithm divides the earthquake history data into training data and test data in a ratio of 5:2;

[0059] Generative adversarial networks obtain abnormal data by comparing training data with normal geological data;

[0060] The recurrent neural network algorithm learns the velocity law of abnormal data and predicts earthquake velocity based on the velocity law;

[0061] The particle swarm optimization algorithm adjusts the hyperparameters of the recurrent neural network algorithm based on the error between the predicted seismic velocity of the test data and the actual seismic data.

[0062] Furthermore, the method for optimizing the seismic velocity prediction model according to the velocity error includes:

[0063] Perform chaotic mapping on the search sub-population, the expression is:

[0064]

[0065] The chaotic random number from 0 to 1 is b, and the w+1th dimension is ε w+1 , the wth dimension is ε w ;

[0066] Compute the position of the searcher:

[0067]

[0068] The random number from 1 to 2 is v1, and the search subposition of the a-th search sub in the w-th dimension in the t-th iteration of the exploration phase is The searcher position of the a-th searcher in the w-th dimension at the t-th iteration is The searcher type of the a-th searcher in the t-th iteration and the w-th dimension is D a,w (t), the first random number from 0 to 1 is r1;

[0069] Update search sub-location:

[0070]

[0071]

[0072]

[0073] In the exploration phase, the updated position of the wth dimension of the ath searcher in the tth iteration is The flight parameter is γ, the first normal distribution random number is κ, the second normal distribution random number is ι, and the standard deviation of the first normal distribution random number κ is θ κ , the standard deviation of the second normal distribution random number ι is θ ι , the gamma function is The modulation constant is x, the attenuation factor is μ, the inertia weight is l, and the maximum number of iterations is t max , the flight coefficient is χ;

[0074] Calculate the location of the explorer during the development phase:

[0075]

[0076] The position of the wth dimension of the ath searcher in the tth iteration of the development phase is The number of iterations is t, and the lower bound of the w-th dimension is The upper bound of the w-th dimension is

[0077] Calculate the updated development phase search sub-position:

[0078]

[0079] In the development phase, the updated position of the wth dimension of the ath searcher in the tth iteration is The natural constant is e, and the spiral coefficient is k;

[0080] Calculate the velocity error of the searcher:

[0081]

[0082] The seismic velocity error of the a-th searcher is ρ a , the second random number from 0 to 1 is r2, and the prediction speed of the a-th searcher is The predicted seismic velocity of the a-th searcher is The actual seismic velocity of the ath searcher is

[0083] Continue iterating until the speed error is minimized.

[0084] In the second aspect, a seismic velocity prediction system based on big data includes:

[0085] Data acquisition module: used to obtain historical earthquake data and geological data, and pre-process the historical data and geological data;

[0086] Data acquisition module: used for performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapid change parameters;

[0087] Velocity correction module: used for performing velocity change exploration on the geological data by geophysical exploration to obtain variable factor data, and correcting the seismic velocity according to the velocity change parameter and the variable factor data to obtain variable velocity;

[0088] Model building module: used to build a seismic velocity prediction model according to the variable velocity, and optimize the seismic velocity prediction model according to the velocity error;

[0089] Result output module: used to input the data to be predicted into the optimized seismic velocity prediction model and output the predicted velocity.

[0090] The beneficial effects of the present invention are:

[0091] The present invention is a method and system for predicting earthquake velocity based on big data. Compared with the prior art, the present invention has the following technical effects:

[0092] The present invention can improve the accuracy of seismic velocity prediction through preprocessing, full waveform inversion, parameter selection, velocity change exploration, seismic velocity correction, model construction and model optimization steps, thereby improving the accuracy of seismic velocity prediction of big data. The seismic velocity prediction of big data can be optimized, which can greatly save resources and improve work efficiency. It can realize intelligent prediction of seismic velocity and perform velocity correction on seismic velocity prediction of big data in real time, which is of great significance to seismic velocity prediction of big data, can adapt to seismic velocity prediction of big data with different standards and seismic velocity prediction needs of different big data, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a flow chart of the steps of a seismic velocity prediction method based on big data of the present invention. DETAILED DESCRIPTION

[0094] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0095] The present invention provides a method and system for predicting earthquake velocity based on big data, comprising the following steps:

[0096] like Figure 1 As shown, in this embodiment, the following steps are included:

[0097] Acquire historical data and geological data of earthquakes, and pre-process the historical data and geological data; the geological data include topography, strata, faults, rock density, rock composition, elastic modulus, Poisson's ratio, crust thickness, crust movement, plate movement, and plate force; the historical data include time, location, magnitude, earthquake type, earthquake duration, seismic wave propagation velocity, earthquake intensity, propagation direction, acceleration, displacement, ground velocity change, frequency, waveform characteristics, phase, temperature change, electromagnetic field change, gravity field change, and focal depth;

[0098] In the actual assessment, the geological data are the topography and landforms of the junction of mountains and plains, the strata are mainly sedimentary rocks with a small amount of volcanic rocks, the fault has a northeast fault zone, and the average rock density is about 2.65g / cm 3 , rock composition metamorphic rock, elastic modulus 55GPa, Poisson's ratio 0.26, crust thickness 35km, crust movement mainly horizontal movement accompanied by vertical rise and fall, plate movement at the edge of the plate, plates squeeze each other, and the plate is subjected to northeast squeezing pressure; historical data are time 3.15pm, location XXX, magnitude 5.2, earthquake type shallow source earthquake, earthquake duration 18s, seismic wave propagation speed 5.8km / s, earthquake intensity level four, propagation direction northeast, peak acceleration 0.12g, maximum displacement 0.25 meters, ground velocity change 1.2m / s, frequency 2-4Hz, waveform characteristics P wave clear S wave more obvious, phase initial phase is positive, temperature change no significant change, electromagnetic field change slight fluctuation, gravity field change no significant change, focal depth 31km;

[0099] Performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapidly changing parameters;

[0100] In the actual evaluation, the propagation parameters are elastic parameters, viscoelastic parameters, and density. The elastic parameters are elastic modulus 55 GPa, Poisson's ratio 0.26, viscoelastic parameters 3.91×1018 Pa·s, and density 2.65 g / cm 3 , elastic parameter 1.57 degrees; the influence degrees of elastic parameter, viscoelastic parameter and density are 0.513, 0.816 and 0.672 respectively, and the rapidly changing parameters are elastic parameter, viscoelastic parameter and density;

[0101] Performing velocity change exploration on the geological data by geophysical exploration to obtain variable factor data, and correcting the seismic velocity according to the velocity change parameter and the variable factor data to obtain variable velocity;

[0102] In the actual evaluation, the variable data are: variable speed is 5.67km / s;

[0103] Constructing a seismic velocity prediction model according to the variable velocity, and optimizing the seismic velocity prediction model according to the velocity error;

[0104] Inputting the data to be predicted into the optimized seismic velocity prediction model and outputting the predicted velocity;

[0105] In the actual assessment, the geological data are the topography and landforms of the junction of mountains and plains, the strata are mainly sedimentary rocks with a small amount of volcanic rocks, the fault has a northeast fault zone, and the average rock density is about 2.65g / cm 3 , rock composition metamorphic rock, elastic modulus 55GPa, Poisson's ratio 0.26, crust thickness 35km, crust movement mainly horizontal movement accompanied by vertical rise and fall, plate movement at the edge of the plate, plates squeeze each other, and the plate is subjected to northeast squeezing pressure; historical data are time 15.45pm, location XXX, magnitude 6.1, earthquake type medium source earthquake, earthquake duration 25s, seismic wave propagation speed 6.2km / s, earthquake intensity 5, propagation direction northwest, peak acceleration 0.18g, maximum displacement 0.4m, ground velocity change 1.8m / s, frequency 3-6Hz, waveform characteristics P wave and S wave are more obvious, the initial phase of the phase is negative, the temperature change has no significant change, the electromagnetic field change fluctuates slightly, the gravity field change has no significant change, and the focal depth is 54km;

[0106] The predicted speed at the next moment is 5.97 km / s.

[0107] In this embodiment, the method of performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters includes:

[0108] The finite element method is used to perform gravity gradient forward modeling and calculate the gravitational potential of the observation point:

[0109]

[0110] Where the gravitational constant is H, the horizontal coordinate is q and the vertical coordinate is u, the gravity value generated by the underground medium at the observation point (q,u) is Δg(q,u), the horizontal coordinate of the observation point after segmentation is λ, the vertical coordinate of the observation point after segmentation is τ, and the underground medium density of the observation point coordinate after segmentation is φ(λ,τ);

[0111] Calculate the vertical component of the gravity anomaly and the vertical component of the gravity gradient:

[0112]

[0113]

[0114] The vertical component of the gravity anomaly is g u , the vertical component of the gravity gradient tensor is K uu;

[0115] Calculate the kernel function matrix:

[0116]

[0117] The kernel function matrix is ​​S, and the simulated seismic signal is calculated:

[0118]

[0119] The simulated earthquake signal at time d The stability constant is a, and the weighting parameter is The initial seismic signal is F o , the natural water length is e, and the objective function is constructed, the expression is:

[0120]

[0121] F(d)=f(d)+ωE{f(d)}

[0122] The objective function is The seismic signal at time d is F(d), the real part of the seismic signal F(d) at time d is f(d), the imaginary part of the seismic signal F(d) at time d is E{f(d)}, the imaginary unit is ω, and the real part of the simulated seismic signal at time d is The imaginary part of the simulated seismic signal F(d) at time d is

[0123] Compute the inversion gradient:

[0124]

[0125] The inversion gradient is ε, and the observed value of the gravity gradient is The simulated value of the gravity gradient is

[0126] The inversion gradient is used to optimize the inversion model according to the objective function value until the objective function value is minimized and then the optimization is stopped;

[0127] The historical data are input into the optimized inversion model and the propagation parameters are output.

[0128] In this embodiment, the method of selecting the propagation parameter according to the influence to obtain the rapidly changing parameter includes:

[0129] Normalize the propagation parameters and construct a decision matrix with the normalized propagation parameters;

[0130] Calculate the importance of propagation parameters:

[0131]

[0132] The adjustment factor of the sth propagation parameter is p s , the correlation between the sth propagation parameter and the seismic velocity is V s , the number of propagation parameters is The correlation between the cth propagation parameter and the seismic velocity is V c ;

[0133] Calculate the decision value based on the normalized propagation parameter and importance:

[0134]

[0135] The normalized sth propagation parameter is G s , the decision value of the sth propagation parameter is c s ;

[0136] Construct a standard decision matrix based on the decision value and calculate the grey correlation degree of positive and negative ideal solutions:

[0137]

[0138]

[0139] The grey correlation degree of the positive ideal solution in the i-th row and y-th column is The grey relational degree of the negative ideal solution in row i and column y is The importance of the propagation parameters in column y is The decision value of row i and column y is c iy , the propagation parameter of the i-th row is z i , the propagation parameter z i The probability of occurrence is h(z i ), the resolution coefficient is ζ;

[0140] Calculate the impact:

[0141]

[0142] The first weight is The second weight is The negative decision value of column y is The positive decision value of column y is The positive influence of the propagation parameter in the i-th row is

[0143] The influence degrees are sorted in descending order, and the propagation parameters with influence degrees greater than 0.507 are output as rapidly changing parameters.

[0144] In this embodiment, the method for obtaining variable velocity by correcting the seismic velocity according to the velocity variation parameter and the variable factor data includes:

[0145]

[0146] The speed at time d+1 is The earthquake velocity at time d is V d , the jth speed-changing parameter is The weight of the jth fast-changing parameter is The number of rapidly changing parameters is σ and the decay rate is ξ.

[0147] In this embodiment, the method for constructing a seismic velocity prediction model according to the variable velocity includes:

[0148] The objective function is constructed according to the variable speed, and the expression is:

[0149]

[0150] The objective function is The speed at time d is The predicted speed at time d is The loss function is The adjustment constant is a, and the control weight is

[0151] Constructing a seismic velocity prediction model based on the objective function. The seismic velocity prediction model includes a random forest algorithm, a generative adversarial network, a recurrent neural network algorithm, and a particle swarm optimization algorithm;

[0152] The random forest algorithm divides the earthquake history data into training data and test data in a ratio of 5:2;

[0153] Generative adversarial networks obtain abnormal data by comparing training data with normal geological data;

[0154] The recurrent neural network algorithm learns the velocity law of abnormal data and predicts earthquake velocity based on the velocity law;

[0155] The particle swarm optimization algorithm adjusts the hyperparameters of the recurrent neural network algorithm based on the error between the predicted seismic velocity of the test data and the actual seismic data.

[0156] In this embodiment, the method for optimizing the seismic velocity prediction model according to the velocity error includes:

[0157] Perform chaotic mapping on the search sub-population, the expression is:

[0158]

[0159] The chaotic random number from 0 to 1 is b, and the w+1th dimension is ε w+1 , the wth dimension is ε w ;

[0160] Compute the position of the searcher:

[0161]

[0162] The random number from 1 to 2 is v1, and the search subposition of the a-th search sub in the w-th dimension in the t-th iteration of the exploration phase is The searcher position of the a-th searcher in the w-th dimension at the t-th iteration is The searcher type of the a-th searcher in the t-th iteration and the w-th dimension is D a,w (t), the first random number from 0 to 1 is r1;

[0163] Update search sub-location:

[0164]

[0165]

[0166]

[0167] In the exploration phase, the updated position of the wth dimension of the ath searcher in the tth iteration is The flight parameter is γ, the first normal distribution random number is κ, the second normal distribution random number is ι, and the standard deviation of the first normal distribution random number κ is θ κ , the standard deviation of the second normal distribution random number ι is θ ι , the gamma function is The modulation constant is x, the attenuation factor is μ, the inertia weight is l, and the maximum number of iterations is t max , the flight coefficient is χ;

[0168] Calculate the location of the explorer during the development phase:

[0169]

[0170] The position of the wth dimension of the ath searcher in the tth iteration of the development phase is The number of iterations is t, and the lower bound of the w-th dimension is The upper bound of the w-th dimension is

[0171] Calculate the updated development phase search sub-position:

[0172]

[0173] In the development phase, the updated position of the wth dimension of the ath searcher in the tth iteration is The natural constant is e, and the spiral coefficient is k;

[0174] Calculate the velocity error of the searcher:

[0175]

[0176] The seismic velocity error of the a-th searcher is ρ a , the second random number from 0 to 1 is r2, and the prediction speed of the a-th searcher is The predicted seismic velocity of the a-th searcher is The actual seismic velocity of the ath searcher is

[0177] Continue iterating until the speed error is minimized.

[0178] In the second aspect, a seismic velocity prediction system based on big data includes:

[0179] Data acquisition module: used to obtain historical earthquake data and geological data, and pre-process the historical data and geological data;

[0180] Data acquisition module: used for performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapid change parameters;

[0181] Velocity correction module: used for performing velocity change exploration on the geological data by geophysical exploration to obtain variable factor data, and correcting the seismic velocity according to the velocity change parameter and the variable factor data to obtain variable velocity;

[0182] Model building module: used to build a seismic velocity prediction model according to the variable velocity, and optimize the seismic velocity prediction model according to the velocity error;

[0183] Result output module: used to input the data to be predicted into the optimized seismic velocity prediction model and output the predicted velocity.

[0184] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting earthquake velocity based on big data, characterized in that: The following steps are involved: Acquire historical earthquake data and geological data, and pre-process the historical data and the geological data; Performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapidly changing parameters; Performing velocity change exploration on the geological data by geophysical exploration to obtain velocity change parameters, and correcting the seismic velocity according to the velocity change parameters and the velocity change parameters to obtain variable velocity; Constructing a seismic velocity prediction model according to the variable velocity, and optimizing the seismic velocity prediction model according to the velocity error; The data to be predicted is input into the optimized seismic velocity prediction model, and the predicted velocity is output.

2. The earthquake velocity prediction method based on big data according to claim 1, characterized in that: The method of performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters comprises: The finite element method is used to perform gravity gradient forward modeling and calculate the gravitational potential of the observation point: The gravitational constant is H, the horizontal coordinate is q and the vertical coordinate is u. The gravity value generated is , the horizontal coordinate of the observation point after segmentation is , the vertical coordinate of the observation point after segmentation is , the underground medium density of the observation point coordinates after segmentation is ; Calculate the vertical component of the gravity anomaly and the vertical component of the gravity gradient: The vertical component of the gravity anomaly is , the vertical component of the gravity gradient tensor is ; Calculate the kernel function matrix: The kernel function matrix is ​​S, and the simulated seismic signal is calculated: in Simulated earthquake signal at time , the stability constant is , the weight adjustment parameter is , the initial seismic signal is , the natural constant is e, and the objective function is constructed, the expression is: The objective function is , The earthquake signal at time , Earthquake signal at any time The real part of , Earthquake signal at any time The imaginary part of , the imaginary unit is , The real part of the simulated seismic signal at time , Simulate earthquake signals at all times The imaginary part of ; Compute the inversion gradient: The inversion gradient is , the observed value of the gravity gradient is , the simulated value of the gravity gradient is ; The inversion gradient is used to optimize the inversion model according to the objective function value until the objective function value is minimized and then the optimization is stopped; The historical data are input into the optimized inversion model and the propagation parameters are output.

3. The earthquake velocity prediction method based on big data according to claim 1, characterized in that: The method of selecting the propagation parameter according to the influence to obtain the rapidly changing parameter includes: Normalize the propagation parameters and construct a decision matrix with the normalized propagation parameters; Calculate the importance of propagation parameters: The adjustment factor of the sth propagation parameter is , the correlation between the sth propagation parameter and the seismic velocity is , the number of propagation parameters is , the correlation between the cth propagation parameter and the seismic velocity is ; Calculate the decision value based on the normalized propagation parameter and importance: The normalized sth propagation parameter is , the decision value of the sth propagation parameter is ; Construct a standard decision matrix based on the decision value and calculate the grey correlation degree of positive and negative ideal solutions: The grey correlation degree of the positive ideal solution in the i-th row and y-th column is , the grey relational degree of the negative ideal solution in row i and column y is , the importance of the propagation parameter in column y is , the decision value of row i and column y is , the propagation parameter of the i-th row is , propagation parameters The probability of occurrence is , the resolution coefficient is ; Calculate the impact: The first weight is , the second weight is , the negative decision value of column y is , the positive decision value of column y is , the positive influence of the propagation parameter in the i-th row is ; The influence degrees are sorted in descending order, and the propagation parameters with influence degrees greater than 0.507 are output as rapidly changing parameters.

4. The earthquake velocity prediction method based on big data according to claim 1, characterized in that: The method for obtaining variable velocity by correcting the seismic velocity according to the speed variation parameter and the velocity variation parameter comprises: in The speed at time +1 is , The earthquake velocity at time , the jth speed-changing parameter is , the weight of the jth fast-changing parameter is , the number of rapidly changing parameters is , the attenuation rate is .

5. The earthquake velocity prediction method based on big data according to claim 1, characterized in that: The method for constructing a seismic velocity prediction model according to the variable velocity comprises: The objective function is constructed according to the variable speed, and the expression is: The objective function is , The speed of change at time is , The predicted speed at time , the loss function is , the adjustment constant is , the control weight is ; Constructing a seismic velocity prediction model based on the objective function. The seismic velocity prediction model includes a random forest algorithm, a generative adversarial network, a recurrent neural network algorithm, and a particle swarm optimization algorithm; The random forest algorithm divides the earthquake history data into training data and test data in a ratio of 5:2; Generative adversarial networks obtain abnormal data by comparing training data with normal geological data; The recurrent neural network algorithm learns the velocity law of abnormal data and predicts earthquake velocity based on the velocity law; The particle swarm optimization algorithm adjusts the hyperparameters of the recurrent neural network algorithm based on the error between the predicted seismic velocity of the test data and the actual seismic data.

6. The earthquake velocity prediction method based on big data according to claim 1, characterized in that: The method for optimizing the seismic velocity prediction model according to the velocity error comprises: Perform chaotic mapping on the search sub-population, the expression is: The chaotic random number between 0 and 1 is b, The dimensions are , No. The dimensions are ; Compute the position of the searcher: The random numbers from 1 to 2 are , the position of the search sub-sub in the w-th dimension of the a-th search sub-sub in the t-th iteration of the exploration phase is , the search subposition of the a-th search sub in the w-th dimension in the t-th iteration is , the search subtype of the a-th search sub in the t-th iteration and the w-th dimension in the t-th iteration is , the first random number from 0 to 1 is ; Update search sub-location: In the exploration phase, the updated position of the wth dimension of the ath searcher in the tth iteration is , the flight parameters are , the first normally distributed random number is , the second normal distribution random number is , the first normally distributed random number The standard deviation of , the second normal distribution random number The standard deviation of , the gamma function is , the modulation constant is x, and the attenuation factor is , the inertia weight is , the maximum number of iterations is , the flight coefficient is ; Calculate the location of the explorer during the development phase: The position of the wth dimension of the ath searcher in the tth iteration of the development phase is , the number of iterations is t, and the lower bound of the w-th dimension is , the upper bound of the w-th dimension is ; Calculate the updated development phase search sub-position: In the development phase, the updated position of the wth dimension of the ath searcher in the tth iteration is , the natural constant is e, and the spiral coefficient is k; Calculate the velocity error of the searcher: The seismic velocity error of the a-th searcher is , the second random number from 0 to 1 is , the predicted seismic velocity of the a-th searcher is , the actual seismic velocity of the a-th searcher is ; Continue iterating until the speed error is minimized.

7. A seismic velocity prediction system based on big data, characterized in that: include: Data acquisition module: used to obtain historical earthquake data and geological data, and pre-process the historical data and geological data; Data acquisition module: used for performing full waveform inversion on the historical data according to the inversion model to obtain propagation parameters, and selecting the propagation parameters according to the influence to obtain the rapid change parameters; Velocity correction module: used for performing velocity change exploration on the geological data by geophysical exploration to obtain velocity change parameters, and correcting the seismic velocity according to the velocity change parameters and the velocity change parameters to obtain variable velocity; Model building module: used to build a seismic velocity prediction model according to the variable velocity, and optimize the seismic velocity prediction model according to the velocity error; Result output module: used to input the data to be predicted into the optimized seismic velocity prediction model and output the predicted velocity.

Citation Information

Patent Citations

  • Method and device for determining absolute gravitational acceleration

    CN111679335A

  • Pre-stack seismic velocity inversion method based on attention gate neural network

    CN114942473A

  • Boundary layer tomography method and device

    US20160266267A1

  • Geophysical Inversion with Convolutional Neural Networks

    US20190064389A1