Multi-channel inversion method and system based on improved genetic algorithm

Abstract

The invention discloses a multi-channel inversion method and system based on an improved genetic algorithm, and relates to the technical field of geophysical inversion, and the method comprises the steps: employing the improved genetic algorithm to achieve the inversion of a seismic record during the inversion of the seismic record, i.e., adjusting the fitness value of the inversion process, carrying out the descending order of the fitness function values of all individuals, and carrying out the calculation of the fitness function values of all individuals. The fitness value of the individual corresponding to the front quarter of the fitness function value is increased by a first preset value, the fitness function value of the individual corresponding to the rear quarter of the fitness function value is decreased by a second preset value, and the fitness value of the individual corresponding to the middle fitness function value is not changed; then carrying out selective inheritance by adopting a new fitness function value; therefore, the excellent individuals cannot be eliminated due to too small quantity, and the bad individuals are easier to eliminate due to reduction of the fitness value, so that the genetic algorithm can better play a preferable effect, and the genetic algorithm is faster to converge.

Description

Technical Field

[0001] The present invention relates to the field of geophysical inversion technology, and in particular to a multi-channel inversion method and system based on an improved genetic algorithm. Background Art

[0002] Seismic exploration has always been the most important and effective method for detecting oil and gas reservoirs in geophysical exploration. The so-called seismic exploration technology is a method of inferring the properties of underground media and the structure of strata by the response of seismic waves. Its main working process is: seismic waves are generated by artificial means such as industrial blasting or piling. Seismic waves propagate downward along the strata. When they encounter the interface of elastic media, they will be reflected. At this time, the precision instruments on the ground will record the vibration of each sampling point on the ground. By analyzing the characteristics of the recorded sampling points, we can help us find the location of oil and gas reservoirs, predict the type of oil and gas reservoirs, and judge the lithological characteristics of the strata. Seismic inversion technology is one of the important means to achieve reservoir description through seismic technology. This technology uses surface recorded seismic data, constructs the target function through previous experience, and uses drilling data and logging data as constraints to invert the underground space structure and the physical characteristics of the underground medium and perform imaging. Seismic inversion has promoted the development of seismic lithology analysis technology and seismic fluid analysis technology, and provided an important basis for the exploration and development of oil and gas reservoirs.

[0003] In the field of petroleum physical exploration, inversion technology can establish an elastic parameter model of underground media from observed geophysical data through mathematical and physical relationships. Compared with post-stack seismic data, pre-stack seismic data has high fidelity and more elastic parameter information, and has received widespread attention and long-term development.

[0004] The classification of seismic inversion methods is numerous and complex. People have proposed many classification methods based on different input data, different problems to be solved and different implementation methods.

[0005] The traditional prestack inversion method is mainly AVO (Amplitude Varies with Offset) inversion, which uses the Zoeppritz equation and various linearized approximate formulas to describe the relationship between the reflection coefficient or seismic wave amplitude with the offset and the incident angle to invert the elastic parameters on both sides of the interface. Forward modeling is the basis of inversion. However, the Zoeppritz equation is based on the assumption of a single interface, which means that multiple waves, transmission losses, thin-layer interference effects and other propagation effects cannot be considered. On the one hand, the forward modeled wave field is different from the actual wave field, and on the other hand, it will cause the loss of inversion information and even instability of the inversion. The reflectivity method is a method that can efficiently and accurately simulate the seismic wave field of layered media. In the forward modeling process, the full wavefield seismic record including various propagation effects is considered, and part of the wavefield can be simulated separately as needed. Compared with the simulation of prestack seismic records using the Zoeppritz equation, the reflectivity method can establish a more reasonable relationship between model parameters and prestack seismic records, so it is suitable for accurate prestack seismic record synthesis.

[0006] In the prior art, the method of constructing an elastic parameter model is generally used to realize multi-channel pre-stack wave inversion. For example, Chinese invention patent CN115480311A discloses "a multi-channel pre-stack waveform inversion method and device", which includes: step 1, obtaining angle gathers based on elastic wave equations; step 2, obtaining a multi-channel inversion target function based on the angle gathers; step 3, removing the correlation between elastic parameters from the multi-channel inversion target function; step 4, selecting an elastic parameter initial model, and obtaining the disturbance amount in the elastic parameter relative to the initial model based on the multi-channel inversion target function after removing the correlation between elastic parameters and the elastic parameter initial model; step 5, repeatedly iterating step 4 to obtain a multi-channel pre-stack waveform inversion result. Although this patent solves the problems of large inversion error and insufficient inversion resolution to a certain extent by constructing an elastic model and iterating, the effect of reducing inversion is not obvious during multi-channel inversion. Therefore, the prior art urgently needs a multi-channel inversion method and system that can reduce inversion error. Summary of the invention

[0007] The purpose of the present invention is to provide a multi-channel inversion method and system based on an improved genetic algorithm, which significantly improves the inversion effect during multi-channel inversion and further improves the inversion accuracy. To achieve the above purpose, the present invention provides the following technical solutions:

[0008] The present invention provides a multi-channel inversion method based on an improved genetic algorithm, the method comprising the following steps:

[0009] Step S1, synthesizing multi-channel seismic records using a reflectivity method;

[0010] Step S2, obtaining a final velocity model based on the pre-stack waveform inversion of the multi-channel seismic records using an improved genetic algorithm; comprising:

[0011] Step S2.1, determining the initial values ​​of the inversion parameters using the logging data;

[0012] Step S2.2, encoding the inversion parameters to generate an initial model population;

[0013] Step S2.3, given measured data, the measured data are compared with the multi-channel seismic records to form individuals, and the fitness function value of the objective function of each individual is calculated;

[0014] Step S2.4, arranging the fitness function value of each individual in the initial model population in descending order, and adjusting the fitness function value at the same time;

[0015] Step S2.5, based on the adjusted fitness function value, performing genetic operation on each individual in the initial model population to obtain a new population;

[0016] Step S2.6: determine whether the new population meets the accuracy requirement. If so, output the final velocity model. If not, continue iterating until the accuracy requirement is met.

[0017] Furthermore, the step S2.4 comprises:

[0018] The fitness function value of each individual in the initial model population is arranged in descending order, and then for the individuals corresponding to the fitness function values ​​in the first quarter, their fitness values ​​are increased by a first preset value, and for the individuals corresponding to the fitness function values ​​in the second quarter, their fitness function values ​​are reduced by a second preset value, and the fitness values ​​of the individuals corresponding to the middle fitness function values ​​remain unchanged, and finally a new fitness function value is obtained; wherein the first preset value and the second preset value are values ​​set in advance.

[0019] Furthermore, the step S1 comprises:

[0020] Step S1.1, calculating the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain;

[0021] Step S1.2, multiplying the longitudinal wave reflection coefficient by the source wavelet to obtain a reflection spectrum;

[0022] Step S1.3, performing inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain;

[0023] Step S1.4, changing the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain;

[0024] Step S1.5, inversely transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, where the reflection wave record in the xt domain is the multi-channel seismic record synthesized by the reflectivity method.

[0025] Furthermore, the step S1.1 comprises:

[0026] First, a layered formation model is designed to study the propagation of longitudinal waves and transverse waves after a plane wave is incident on the formation; then, the relationship between displacement and stress on the contact surface of each layer in the layered formation model is obtained through equations; finally, the reflection coefficient is calculated based on the relationship between displacement and stress on the contact surface of each layer; the reflection coefficient includes: longitudinal wave reflection coefficient, transverse wave reflection coefficient, longitudinal wave transmission coefficient and transverse wave transmission coefficient.

[0027] Furthermore, in step S1.2, the reflection spectrum is a reflection spectrum in the frequency-slowness domain.

[0028] Furthermore, the step S2.3 includes:

[0029] Step S2.3.1, setting the objective function of each individual in the initial model population, the objective function is used to characterize the error between the measured data and the synthetic seismic record; the expression of the objective function is:

[0030] Objvalue = norm(c-data);

[0031] Wherein, Objvalue represents the objective function, norm represents the norm value of the error adopted by the objective function, c represents the synthetic seismic record obtained in step S1, and data represents the measured data;

[0032] Step S2.3.2, calculate the fitness function value of the objective function of each individual in the initial model population; the expression of the fitness function is:

[0033] Fitt=1 / (Objvalue*Objvalue+0.001);

[0034] Where Fitt represents the fitness function.

[0035] Furthermore, in step S2.5, the genetic operation is to perform crossover and mutation using a crossover operator and a mutation operator.

[0036] Furthermore, in step S2.6, the accuracy requirement is to meet the signal-to-noise ratio and error convergence requirements of the seismic data.

[0037] The present invention also provides a multi-channel inversion system based on an improved genetic algorithm, the system comprising: a synthesis unit and an inversion unit; wherein:

[0038] A synthesis unit for synthesizing multi-channel seismic records using a reflectivity method;

[0039] An inversion unit is used to obtain a final velocity model based on the pre-stack waveform inversion of the seismic record based on an improved genetic algorithm; comprising:

[0040] An acquisition module, which establishes an initial velocity model using well logging data, and performs forward modeling operations based on the initial velocity model to obtain synthetic seismic records;

[0041] A generation module, based on the initial velocity model, determines parameters of the improved genetic algorithm, encodes the parameters, and generates an initial model population;

[0042] A first calculation module is provided with measured data, and compares the measured data with the synthetic seismic record, and calculates the fitness function value of the objective function of each individual in the initial model population;

[0043] An adjustment module arranges the fitness function value of each individual in the initial model population in descending order and adjusts the fitness function value at the same time;

[0044] An operation module, based on the adjusted fitness function value, performs genetic operation on each individual in the initial model population to obtain a new population;

[0045] The judgment module judges whether the new population meets the accuracy requirement. If so, the final velocity model is output; if not, the iteration is continued until the accuracy requirement is met.

[0046] Furthermore, the synthesis unit includes: a second calculation module, a first acquisition module, a second acquisition module, a third acquisition module and a fourth acquisition module; wherein,

[0047] The second calculation module is used to calculate the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain;

[0048] A first acquisition module is used to multiply the longitudinal wave reflection coefficient with the source wavelet to obtain a reflection wave spectrum;

[0049] A second acquisition module is used to perform inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain;

[0050] A third acquisition module is used to change the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain;

[0051] The fourth acquisition module is used to inversely transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, and the reflection wave record in the xt domain is the multi-channel seismic record synthesized by the reflectivity method.

[0052] Technical effects and advantages of the present invention:

[0053] When inverting earthquake records, the present invention adopts an improved genetic algorithm to realize the inversion of earthquake records, that is, the fitness value is adjusted, the fitness function value of each individual is arranged in descending order, and then for individuals corresponding to the fitness function values ​​of the first quarter, the fitness value is increased by a first preset value, and for individuals corresponding to the fitness function values ​​of the second quarter, the fitness function value is reduced by a second preset value, and the fitness value of individuals corresponding to the middle fitness function value remains unchanged; then the new fitness function value is adopted for selection and inheritance; thereby, excellent individuals will not be eliminated due to too small number, and poor individuals will be more easily eliminated due to reduced fitness values, so that the genetic algorithm can better play the optimization effect and also make the genetic algorithm converge faster.

[0054] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 A flow chart of a multi-channel inversion method based on an improved genetic algorithm provided by the present invention;

[0057] Figure 2 A flow chart of obtaining a final velocity model by prestack waveform inversion based on an improved genetic algorithm provided by the present invention;

[0058] Figure 3 A comparison diagram of the original genetic algorithm inversion and the improved genetic algorithm inversion provided by the present invention;

[0059] Figure 4 A schematic diagram of a multi-channel inversion system based on an improved genetic algorithm provided by the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] In order to solve the deficiencies of the prior art, the present invention provides a multi-channel inversion method based on an improved genetic algorithm. Figure 1 A flow chart of a multi-channel inversion method based on an improved genetic algorithm provided by the present invention is as follows: Figure 1 As shown, the method comprises the following steps:

[0062] Step S1, synthesizing multi-channel seismic records using a reflectivity method;

[0063] Specifically, step S1 includes steps S1.1-S1.5;

[0064] Step S1.1, calculating the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain;

[0065] Specifically, the step S1.1 includes: firstly designing a layered formation model to study the propagation of longitudinal waves and transverse waves after a plane wave is incident on the formation; then obtaining the relationship between displacement and stress on the contact surface of each layer in the layered formation model through a wave equation; finally, obtaining the reflection coefficient according to the relationship between displacement and stress on the contact surface of each layer;

[0066] Furthermore, the reflection coefficient includes: a longitudinal wave reflection coefficient, a transverse wave reflection coefficient, a longitudinal wave transmission coefficient and a transverse wave transmission coefficient;

[0067] Step S1.2, multiplying the longitudinal wave reflection coefficient by the source wavelet to obtain a reflection spectrum;

[0068] Preferably, the reflection spectrum is a reflection spectrum in the frequency-slowness domain

[0069] Step S1.3, performing inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain;

[0070] Preferably, the step S1.2 obtains the reflection wave spectrum in the frequency-slowness domain, which needs to be converted into the reflection wave spectrum in the time-slowness domain for calculating the subsequent reflection wave record. Therefore, the present invention adopts the inverse Fourier transform method to obtain the reflection wave record in the time-slowness domain;

[0071] Step S1.4, changing the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain;

[0072] Step S1.5, inversely transforming the seismic record in the τ-p domain to obtain a reflection wave record in the xt domain;

[0073] The reflection wave record in the xt domain is a seismic record synthesized by the reflectivity method;

[0074] It is worth emphasizing that the seismic record is a multi-channel seismic record.

[0075] Step S2: obtaining a final velocity model through pre-stack waveform inversion based on an improved genetic algorithm.

[0076] Specifically, Figure 2 The flowchart of obtaining the final velocity model by prestack waveform inversion based on the improved genetic algorithm provided by the present invention is as follows: Figure 2 As shown, step S2 includes steps S2.1-S2.6;

[0077] Step S2.1, establishing an initial velocity model using well logging data, and performing forward modeling based on the initial velocity model to obtain synthetic seismic records;

[0078] Preferably, the initial velocity model includes: an initial longitudinal wave velocity model and an initial shear wave velocity model;

[0079] Step S2.2, determining the parameters of the improved genetic algorithm, encoding the parameters, and generating an initial model population;

[0080] Preferably, all parameters of the improved genetic algorithm are encoded according to actual conditions, then connected to form a chromosome, and an initial population is randomly generated.

[0081] Preferably, real numbers are used for encoding, and the real number encoding specifically includes: the inverted parameters are represented by real numbers;

[0082] Step S2.3, given measured data, the measured data is compared with the synthetic seismic record, and the fitness function value of the objective function of each individual is calculated;

[0083] Preferably, the measured data is a measured seismic record, which is used for comparison with the synthetic seismic record;

[0084] Specifically, the step S2.3 includes steps S2.3.1-S2.3.2;

[0085] Step S2.3.1, setting an individual objective function, wherein the objective function is used to characterize the error between the measured data and the synthetic seismic record;

[0086] Furthermore, the objective function is expressed as:

[0087] Objvalue = norm(c-data);

[0088] Wherein, Objvalue represents the objective function, norm represents the norm value of the error adopted by the objective function, c represents the synthetic seismic record obtained in step S1, and data represents the measured data;

[0089] Step S2.3.2, calculating the fitness function value of the objective function of each individual;

[0090] Preferably, the fitness function used to calculate the fitness function value is expressed as:

[0091] Fitt=1 / (Objvalue*Objvalue+0.001);

[0092] In the formula, Fitt represents the fitness function, and Objvalue represents the objective function;

[0093] Step S2.4, arranging the fitness function value of each individual in descending order, and adjusting the fitness function value at the same time;

[0094] The present invention adopts a strategy of adjusting the fitness value to ensure that the algorithm works better;

[0095] Preferably, the strategy for adjusting the fitness value is specifically as follows: the fitness function value of each individual is arranged in descending order, and then for the individuals corresponding to the fitness function values ​​of the first quarter, the fitness value thereof is increased by a first preset value, and for the individuals corresponding to the fitness function values ​​of the second quarter, the fitness function value thereof is reduced by a second preset value, and the fitness value of the individuals corresponding to the fitness function values ​​in the middle remains unchanged; and then the new fitness function value is used for selection and inheritance;

[0096] Preferably, the first preset value and the second preset value are both values ​​set in advance;

[0097] Step S2.5, performing genetic operations to obtain a new population;

[0098] Specifically, the genetic operation is to use a crossover operator and a mutation operator to perform crossover and mutation;

[0099] The selected individuals in the group need to undergo a crossover operation to generate new individuals. Crossover essentially involves randomly selecting two individuals, exchanging some of their genes according to the crossover probability and a certain exchange method, and thus generating two individuals with different external characteristics from the original individuals. In the early stages of iteration, a high crossover probability can quickly increase the richness of the group and speed up the group's evolution; in the later stages of iteration, only a smaller crossover probability can be used to avoid destroying excellent individuals.

[0100] After crossover, individuals in a population need to undergo mutation operations to produce new individuals with mutation characteristics. Mutation is essentially the random selection of a gene on an individual, and the mutating of some genes of the selected individual based on the mutation probability and a certain mutation method, thereby producing individuals with different external characteristics from the original individuals.

[0101] Step S2.6, judging whether the new population meets the accuracy requirement, if so, outputting the final velocity model, if not, continuing to iterate until the accuracy requirement is met;

[0102] The accuracy requirement is to meet the signal-to-noise ratio and error convergence requirements of seismic data.

[0103] Example

[0104] The embodiment of the present invention provides a multi-channel inversion method based on an improved genetic algorithm, the method comprising the following steps:

[0105] Step S1, synthesizing multi-channel seismic records using a reflectivity method;

[0106] If you want to achieve higher accuracy and efficiency in inversion, forward modeling is the basis. Forward modeling is to obtain data through models, that is, to solve seismic shot records when the medium parameters of the specific structure of the stratum are known. In fact, in the process of seismic inversion, the forward modeling of seismic waves accounts for a high proportion of the calculation amount and calculation time, and the accuracy and efficiency of inversion mainly depend on the accuracy and efficiency of forward modeling. In fact, there are many ways of forward modeling. As a commonly used method of forward modeling, the reflectivity method has received continuous attention from researchers. The reflectivity method uses various information of the full wave field, such as multiple reflection waves and conversion waves in each stratum, which can effectively simulate the stratum.

[0107] In this embodiment, taking the acquisition of reflected waves after a plane wave is incident on a layered stratum as an example, how to synthesize multi-channel seismic records using the reflectivity method is described;

[0108] Specifically, step S1 includes steps S1.1-S1.5;

[0109] Step S1.1, calculate and obtain the longitudinal wave reflection coefficient, transverse wave reflection coefficient, longitudinal wave transmission coefficient and transverse wave transmission coefficient in the frequency domain; specifically, first design a layered formation model to study the propagation of longitudinal waves and transverse waves after a plane wave is incident on the formation; then obtain the displacement and stress relationship on the contact surface of each layer in the layered formation model through an equation; obtain the reflection coefficient according to the displacement and stress relationship on the contact surface of each layer, and the reflection coefficient includes: longitudinal wave reflection coefficient, transverse wave reflection coefficient, longitudinal wave transmission coefficient and transverse wave transmission coefficient;

[0110] Step S1.2, correlating the longitudinal wave reflection coefficient with the source wavelet to obtain a reflection spectrum, wherein the reflection spectrum is a reflection spectrum in the frequency-slowness domain;

[0111] Step S1.3, performing inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain; specifically, step S1.2 obtains the reflection wave spectrum in the frequency-slowness domain, and it is also necessary to use the inverse Fourier transform method to convert the reflection wave spectrum in the frequency-slowness domain into a reflection wave spectrum in the time-slowness domain for calculating subsequent reflection wave records;

[0112] Step S1.4, changing the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain;

[0113] Step S1.5, inversely transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, the reflection wave record in the xt domain is the seismic record synthesized by the reflectivity method, and the seismic record is a multi-channel seismic record.

[0114] Step S2: obtaining the final velocity model based on prestack waveform inversion using an improved genetic algorithm;

[0115] Specifically, step S2 includes steps S2.1-S2.6;

[0116] Step S2.1, determining the initial values ​​of the inversion parameters using the logging data;

[0117] The inversion parameters include longitudinal wave velocity and shear wave velocity;

[0118] Step S2.2: Encode the inversion parameters to generate an initial model population; specifically, encode the inversion parameters, then connect them to form a chromosome, and randomly generate an initial model population. In order to better achieve local regulation of individuals in the population, this embodiment uses real numbers for encoding, and the real number encoding specifically includes: the inversion parameters are represented by real numbers;

[0119] Step S2.3: Given measured data, the measured data is compared with the multi-channel seismic record to form individuals, and the fitness function value of the objective function of each individual is calculated, wherein the measured data is the measured seismic record, which is used for comparison with the multi-channel seismic record in step 1;

[0120] Specifically, the step S2.3 includes:

[0121] Step S2.3.1, setting an individual objective function, the objective function is used to characterize the error between the measured data and the synthetic seismic record; the expression of the objective function is:

[0122] Objvalue = norm(c-data);

[0123] Where, Objvalue represents the objective function, norm represents the norm value of the error adopted by the objective function, c represents the synthetic seismic record obtained in step 1, and data represents the measured data;

[0124] Step S2.3.2, calculate the fitness function value of the objective function of each individual; the expression of the fitness function is:

[0125] Fitt=1 / (Objvalue*Objvalue+0.001);

[0126] In the formula, Fitt represents the fitness function; Objvalue represents the objective function

[0127] Step S2.4, arrange the fitness function value of each individual in descending order, and adjust the fitness function value at the same time; specifically, in the original genetic algorithm, individuals with higher fitness values ​​are generally selected to be assigned to the next generation with a higher probability, providing a good foundation for subsequent differences and mutations, so that the group can continue to develop in a better direction, reflecting the principle of survival of the fittest in the process of biological evolution; individuals with higher fitness values ​​in the original genetic algorithm will be replicated with a greater probability, but this method will have shortcomings: in the early stage of inheritance, excellent individuals may be directly eliminated due to random selection or too few excellent individuals; resulting in the original genetic algorithm being unable to effectively play the optimization effect; this embodiment adopts a method of adjusting the fitness function to improve the quality of the individuals; The strategy for adjusting the fitness value is to ensure that the algorithm works better; specifically, the strategy for adjusting the fitness value is as follows: arrange the fitness function value of each individual in descending order, and then for the individuals corresponding to the fitness function values ​​of the first quarter, increase their fitness values ​​by a first preset value, and for the individuals corresponding to the fitness function values ​​of the second quarter, reduce their fitness function values ​​by a second preset value, and the fitness values ​​of the individuals corresponding to the middle fitness function values ​​remain unchanged; finally, a new fitness function value can be obtained for the next step of genetic selection; wherein, the first preset value and the second preset value are both values ​​set in advance; the numerical values ​​of the first preset value and the second preset value can be set by technicians in this field during the calculation process according to the optimization effect and convergence speed of the genetic algorithm.

[0128] In fact, since the present embodiment adjusts the fitness function value, excellent individuals will not be eliminated due to their small number, while poor individuals will be more easily eliminated due to their reduced fitness value, thereby enabling the genetic algorithm to better perform the optimization effect and also enable the genetic algorithm to converge faster.

[0129] Step S2.5, perform genetic operations to obtain a new population; specifically, the genetic operations are crossover and mutation using crossover operators and mutation operators; selected individuals in the population must undergo a crossover operation to produce new individuals. Crossover is essentially the random selection of two individuals, exchanging part of their genes according to the crossover probability and a certain exchange method to produce two individuals with different external characteristics from the original individuals. In the early stage of iteration, the crossover probability is large, which can quickly increase the richness of the population and accelerate the speed of population evolution; in the later stage of iteration, only by using a smaller crossover probability can the destruction of excellent individuals be avoided; individuals in the population that have been crossed must undergo a mutation operation to produce new individuals with mutation characteristics. Mutation is essentially the random selection of a gene on an individual, and the mutation of part of the genes of the selected individual according to the mutation probability and a certain mutation method to produce individuals with different external characteristics from the original individual.

[0130] Step S2.6, judging whether the new population meets the accuracy requirement, if so, outputting the final velocity model, if not, continuing iteration until the accuracy requirement is met; the accuracy requirement is to meet the signal-to-noise ratio and error convergence requirements of the seismic data;

[0131] For example, Figure 3 The comparison diagram of the original genetic algorithm inversion and the improved genetic algorithm inversion provided by the present invention is as follows: Figure 3 As shown in FIG. 1 , the inversion results obtained by using the original genetic algorithm and the improved genetic algorithm of this embodiment for a certain block are as follows: Figure 3 It can be seen from the figure that the improved genetic algorithm proposed in this embodiment is used for inversion, and the inversion result is obviously better than the original genetic algorithm.

[0132] The present invention also discloses a multi-channel inversion system based on an improved genetic algorithm. Figure 4 FIG. 1 is a schematic diagram of a multi-channel inversion system based on an improved genetic algorithm according to the present invention. Figure 4As shown, the system includes: a synthesis unit and an inversion unit; wherein the synthesis unit is used to synthesize multi-channel seismic records by using a reflectivity method; the inversion unit is used to obtain a final velocity model based on the pre-stack waveform inversion of the multi-channel seismic records based on an improved genetic algorithm; it includes: an acquisition module, which uses logging data to establish an initial velocity model, and performs forward modeling operations based on the initial velocity model to obtain synthetic seismic records; a generation module, which determines parameters of the improved genetic algorithm based on the initial velocity model, encodes the parameters, and generates an initial model population; a first calculation module, which gives measured data, compares the measured data with the synthetic seismic records, and calculates the fitness function value of the objective function of each individual in the initial model population; an adjustment module, which arranges the fitness function value of each individual in the initial model population in descending order and adjusts the fitness function value; an operation module, which performs genetic operations on each individual in the initial model population based on the adjusted fitness function value to obtain a new population; a judgment module, which judges whether the new population meets the accuracy requirement, and if so, outputs the final velocity model, and if not, continues to iterate until the accuracy requirement is met.

[0133] Furthermore, the synthesis unit includes: a second calculation module, a first acquisition module, a second acquisition module, a third acquisition module and a fourth acquisition module; wherein the second calculation module is used to calculate the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain; the first acquisition module is used to multiply the longitudinal wave reflection coefficient with the source wavelet to obtain the reflection wave spectrum; the second acquisition module is used to perform inverse Fourier transform on the reflection wave spectrum to obtain the reflection wave record in the time-slowness domain; the third acquisition module is used to change the incident angle of the reflection wave into a ray parameter to obtain the seismic record in the τ-p domain; the fourth acquisition module is used to inverse transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, and the reflection wave record in the xt domain is the multi-channel seismic record synthesized by the reflectivity method.

[0134] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-channel inversion method based on an improved genetic algorithm, characterized in that: The method comprises the following steps: Step S1, synthesizing multi-channel seismic records using a reflectivity method; Step S2, obtaining a final velocity model based on the prestack waveform inversion of the multi-channel seismic records using an improved genetic algorithm; comprising: Step S2.1, determining the initial values ​​of the inversion parameters using the logging data; Step S2.2, encoding the inversion parameters to generate an initial model population; Step S2.3, given measured data, the measured data are compared with the multi-channel seismic records to form individuals, and the fitness function value of the objective function of each individual is calculated; Step S2.4, arranging the fitness function value of each individual in the initial model population in descending order, and adjusting the fitness function value at the same time; Step S2.5, based on the adjusted fitness function value, performing genetic operation on each individual in the initial model population to obtain a new population; Step S2.6: Determine whether the new population meets the accuracy requirement. If so, output the final velocity model. If not, continue iterating until the accuracy requirement is met.

2. The multi-channel inversion method based on improved genetic algorithm according to claim 1, characterized in that: The step S2.4 comprises: The fitness function value of each individual in the initial model population is arranged in descending order, and then for the individuals corresponding to the fitness function values ​​in the first quarter, their fitness values ​​are increased by a first preset value, and for the individuals corresponding to the fitness function values ​​in the second quarter, their fitness function values ​​are reduced by a second preset value, and the fitness values ​​of the individuals corresponding to the middle fitness function values ​​remain unchanged, and finally a new fitness function value is obtained; wherein the first preset value and the second preset value are values ​​set in advance.

3. A multi-channel inversion method based on an improved genetic algorithm according to claim 1 or 2, characterized in that: The step S1 comprises: Step S1.1, calculating the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain; Step S1.2, multiplying the longitudinal wave reflection coefficient by the source wavelet to obtain a reflection spectrum; Step S1.3, performing inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain; Step S1.4, changing the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain; Step S1.5, inversely transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, where the reflection wave record in the xt domain is the multi-channel seismic record synthesized by the reflectivity method.

4. The multi-channel inversion method based on improved genetic algorithm according to claim 3, characterized in that: The step S1.1 comprises: First, a layered formation model is designed to study the propagation of longitudinal waves and transverse waves after a plane wave is incident on the formation; then, the relationship between displacement and stress on the contact surface of each layer in the layered formation model is obtained through equations; finally, the reflection coefficient is calculated based on the relationship between displacement and stress on the contact surface of each layer; the reflection coefficient includes: longitudinal wave reflection coefficient, transverse wave reflection coefficient, longitudinal wave transmission coefficient and transverse wave transmission coefficient.

5. The multi-channel inversion method based on improved genetic algorithm according to claim 3, characterized in that: In step S1.2, the reflection spectrum is a reflection spectrum in the frequency-slowness domain.

6. The multi-channel inversion method based on improved genetic algorithm according to claim 1, characterized in that: The step S2.3 comprises: Step S2.3.1, setting the objective function of each individual in the initial model population, the objective function is used to characterize the error between the measured data and the synthetic seismic record; the expression of the objective function is: Objvalue = norm(c-data); Wherein, Objvalue represents the objective function, norm represents the norm value of the error adopted by the objective function, c represents the synthetic seismic record obtained in step S1, and data represents the measured data; Step S2.3.2, calculate the fitness function value of the objective function of each individual in the initial model population; the expression of the fitness function is: Fitt=1 / (Objvalue*Objvalue+0.001); Where Fitt represents the fitness function.

7. The multi-channel inversion method based on improved genetic algorithm according to claim 1, characterized in that: In step S2.5, the genetic operation is to use a crossover operator and a mutation operator to perform crossover and mutation.

8. A multi-channel inversion method based on an improved genetic algorithm according to claim 1, 6 or 7, characterized in that: In step S2.6, the accuracy requirement is to meet the signal-to-noise ratio and error convergence requirements of the seismic data.

9. A multi-channel inversion system based on an improved genetic algorithm, characterized in that: The system comprises: a synthesis unit and an inversion unit; wherein, A synthesis unit for synthesizing multi-channel seismic records using a reflectivity method; An inversion unit is used to obtain a final velocity model based on the pre-stack waveform inversion of the seismic record based on an improved genetic algorithm; comprising: An acquisition module, which establishes an initial velocity model using well logging data, and performs forward modeling operations based on the initial velocity model to obtain synthetic seismic records; A generation module, based on the initial velocity model, determines parameters of the improved genetic algorithm, encodes the parameters, and generates an initial model population; A first calculation module is provided with measured data, and compares the measured data with the synthetic seismic record, and calculates the fitness function value of the objective function of each individual in the initial model population; An adjustment module arranges the fitness function value of each individual in the initial model population in descending order and adjusts the fitness function value at the same time; An operation module, based on the adjusted fitness function value, performs genetic operation on each individual in the initial model population to obtain a new population; The judgment module judges whether the new population meets the accuracy requirement. If so, the final velocity model is output; if not, the iteration is continued until the accuracy requirement is met.

10. The multi-channel inversion system based on improved genetic algorithm according to claim 9, characterized in that: The synthesis unit includes: a second calculation module, a first acquisition module, a second acquisition module, a third acquisition module and a fourth acquisition module; wherein, The second calculation module is used to calculate the longitudinal wave reflection coefficient, the transverse wave reflection coefficient, the longitudinal wave transmission coefficient and the transverse wave transmission coefficient in the frequency domain; A first acquisition module is used to multiply the longitudinal wave reflection coefficient with the source wavelet to obtain a reflection wave spectrum; A second acquisition module is used to perform inverse Fourier transform on the reflection wave spectrum to obtain a reflection wave record in the time-slowness domain; A third acquisition module is used to change the incident angle of the reflected wave into a ray parameter to obtain a seismic record in the τ-p domain; The fourth acquisition module is used to inversely transform the seismic record in the τ-p domain to obtain the reflection wave record in the xt domain, and the reflection wave record in the xt domain is the multi-channel seismic record synthesized by the reflectivity method.

Images