Quasi-real-time imaging method of underground medium structure under action of fluid in reservoir area

Abstract

A quasi-real-time imaging method for an underground medium structure under the action of fluid in a reservoir area comprises the following steps: continuously acquiring continuous seismic waveform data through an earthquake monitoring station network arranged in the reservoir area, automatically identifying P-wave and S-wave seismic phases by adopting an artificial intelligence technology, and generating a standardized data set when the P-wave and S-wave seismic phases are picked up; on the basis, ray distribution parameters are calculated through ray distribution quantitative modeling; ray distribution parameters are judged in real time by establishing a multi-dimensional data quality control standard, Q value inversion is automatically triggered when the ray distribution parameters reach the standard, and data are intelligently supplemented until requirements are met when the ray distribution parameters do not reach the standard; calculating Q value distribution by adopting a full-automatic inversion algorithm; and finally, generating and outputting an imaging result of the underground medium structure of the reservoir area based on the Q value distribution, and forming a continuous imaging cycle by periodically and repeatedly executing the process. According to the design, the blank of dynamic imaging of a medium structure is filled, the monitoring timeliness is greatly improved, and meanwhile, the reliability and precision of an imaging result are guaranteed through quantitative quality control and full-process automation.

Description

TECHNICAL FIELD

[0001] The application relates to a quasi-real-time imaging method of an underground medium structure and belongs to the technical field of underground medium monitoring. BACKGROUND

[0002] After the water storage of a reservoir, the reservoir water load and the long-term seepage diffusion along the rock mass fissure will continuously change the physical properties of the underground medium, and the energy attenuation characteristics (represented by the quality factor Q value) are particularly sensitive. The Q value directly reflects the absorption capacity of the underground medium to the seismic wave, and the water filling of the fissure will cause the Q value to decrease significantly, which makes the dynamic change of the Q value a key indicator for identifying fluid penetration and warning leakage risk. The mainstream method for capturing such Q value changes in the prior art mainly relies on traditional seismic body wave tomography. This method cannot actively initiate monitoring and must passively rely on natural earthquakes or significant microseismic events to provide data. During the intermittent period of the event, the system is in a monitoring blind area, and there is a significant "time blanking period". During this period, the gradual and continuous geological change process of reservoir water penetration cannot be effectively captured, resulting in the inability to meet the urgent need for real-time or quasi-real-time monitoring of the state of the underground medium under the dynamic water storage and operation conditions of the reservoir, and the monitoring timeliness is insufficient.

[0003] A water conservancy project seepage intelligent monitoring system is disclosed in Chinese patent application No. CN202210276305.X, filed on March 21, 2022. The system directly and contactively collects the seepage liquid in the soil by drilling a hole in the soil of the reservoir area and burying a physical sampling system composed of a "seepage monitoring assembly" and a "greenhouse control assembly". The system actively absorbs the moisture in the surrounding soil using the moisture-absorbing material in the assembly, and then discharges, collects and measures the flow of the absorbed liquid through the gas pressure-driven extrusion device. Meanwhile, the system is assisted by multiple layers of temperature and humidity sensors to assist in determining the seepage direction. The main advantage is that the system can directly obtain seepage liquid samples through physical means to achieve high-precision measurement of the seepage direction and flow. However, this monitoring method relies on physical contact, mechanical action and liquid collection. Its passive and sampling working mode cannot achieve continuous data collection, and there is an inherent "time blanking period". Therefore, the method still has the following defects:

[0004] The monitoring timeliness is insufficient.

[0005] The information disclosed in this background section is intended only to increase an understanding of the general context of the present application, and it should not be taken as an acknowledgement or any form of suggestion that this information forms prior art that is already widely known in the art. SUMMARY

[0006] The invention aims to overcome the problem of insufficient monitoring timeliness in the prior art, and provide a quasi-real-time imaging method for underground medium structure under fluid action in a reservoir area with strong monitoring timeliness.

[0007] To achieve the above object, the technical solution of the invention is:

[0008] A quasi-real-time imaging method for underground medium structure under fluid action in a reservoir area, the method comprising the following steps:

[0009] S1, first receive continuous seismic waveform data collected by a seismic monitoring network arranged in the reservoir area, then standardize and preprocess the continuous seismic waveform data to obtain a standardized waveform data set, then identify P and S wave phases contained in microseismic signals in the standardized waveform data set, then pick up P and S wave arrival time data from the identified P and S wave phases, and then standardize and process the P and S wave arrival time data to obtain a standardized arrival time data set;

[0010] S2, first simulate and generate multiple groups of virtual source-station pairs by using a ray distribution quantification modeling algorithm based on the standardized waveform data set and the standardized arrival time data set, and then calculate ray distribution parameters by using a ray tracing model with the virtual source-station pairs as input;

[0011] S3, judge whether the ray distribution parameters obtained in S2 meet a pre-defined multi-dimensional data quality control standard, if yes, trigger a Q value inversion process; if not, first update the standardized waveform data set by supplementing continuous seismic waveform data, then update the standardized arrival time data set based on the updated standardized waveform data set, then update the ray distribution parameters based on the updated standardized waveform data set and the standardized arrival time data set, until the updated ray distribution parameters meet the multi-dimensional data quality control standard, and then trigger the Q value inversion process;

[0012] S4, when the Q value inversion process in S3 is triggered, the following calculation is performed: based on the standardized waveform data set and the standardized arrival time data set that meet the multi-dimensional data quality control standard in S3, calculate the Q value distribution of the underground medium in the reservoir area by using a full-automatic inversion algorithm;

[0013] S5, generate and output the imaging result of the underground medium structure in the reservoir area based on the Q value distribution of the underground medium in the reservoir area obtained in S4, and the method ends.

[0014] In S2, the ray distribution parameters include: the propagation path of each ray, the underground medium grid cell through which each ray propagates, and the propagation azimuth angle of each ray in the underground medium grid cell through which it propagates.

[0015] The S3 includes the following multi-dimension data quality control standards:

[0016] Ray coverage uniformity standard: the coefficient of variation of the number of ray coverages of all subsurface medium grid cells is not greater than 0.3;

[0017] Ray density standard: the subsurface medium grid cells in the shallow depth range of 0-5 km are covered by at least 3 rays, and the subsurface medium grid cells in the deep depth range of 5-10 km are covered by at least 2 rays;

[0018] Ray azimuth coverage standard: the propagation azimuth coverage of all rays in a single subsurface medium grid cell is not less than 180 degrees.

[0019] In the S3, if the ray distribution parameters obtained in the S2 do not meet the pre-defined multi-dimension data quality control standards, the specific processing procedure is as follows: first, output the identification of the subsurface medium grid cell to be supplemented, then extend the observation time of the station corresponding to the subsurface medium grid cell to supplement the acquisition of continuous seismic waveform data, then input the supplemented continuous seismic waveform data into the procedure of the S1 to update the standardized waveform data set and the standardized travel time data set, then based on the updated standardized waveform data set and the standardized travel time data set, re-execute the S2 to update the ray distribution parameters, and finally execute the S3 to judge whether the updated ray distribution parameters meet the multi-dimension data quality control standards.

[0020] In the S4, the specific steps for calculating the subsurface medium Q value distribution in the reservoir area by using the full-automatic inversion algorithm are as follows:

[0021] S41, first calculate the cross-correlation function between stations based on the standardized waveform data set, then extract the Rayleigh wave or Love wave dispersion curve, and then determine the main frequency f0 of seismic wave propagation;

[0022] S42, first calculate the seismic wave propagation time t using the standardized travel time data, then combine the main frequency f0, and then calculate the seismic wave attenuation parameter t* by the formula:

[0023] S43, convert the calculated t* and the standardized travel time data set into physical constraints, and integrate them into the prior model based on wave theory to construct a target model for inversion;

[0024] S44, input the target model into the Markov Chain Monte Carlo algorithm to automatically iteratively solve the Q value, and obtain the subsurface medium Q value distribution in the reservoir area.

[0025] ​In the S42, when the seismic wave attenuation parameter t* is derived and calculated, the specific derivation and calculation logic is: first, the amplitude envelope attenuation curve of the standardized waveform data is normalized, the exponential decay fitting is used to preliminarily calculate the t* initial value, then the seismic wave propagation time t is combined, and the t* initial value is corrected through the correction formula: The t* initial value is corrected, wherein Δt is the waveform propagation time difference, A0 is the amplitude at the virtual seismic source, and A is the station received amplitude.

[0026] In the S43, the construction logic of the prior model is: first, the travel time residual of the t* correction value and the standardized arrival time data set is taken as the core input, then the theoretical attenuation characteristics of the seismic wave propagation are calculated through the wave equation forward model, and the prior probability distribution of the Q value inversion is established.

[0027] In the S44, the iteration process of the Markov chain Monte Carlo algorithm takes the calculation error minimization of t* as the objective function, and the convergence determination standard is that the Q value calculation result variance of 500 continuous iterations is less than or equal to 0.01.

[0028] In the S1, the identification of the seismic phase and the picking of the arrival time data are realized by using a pre-trained artificial intelligence algorithm model, and the geological condition parameters of the reservoir area are introduced into the model as the model regularization constraint.

[0029] The method further comprises S6: periodically repeating S1 to S5 to form a periodic imaging cycle, in each cycle, after obtaining the current cycle reservoir area underground medium structure imaging result through S5, the current cycle reservoir area underground medium structure imaging result is compared with the past cycle reservoir area underground medium structure imaging result, and the change rate of the Q value at the same spatial position in the current cycle and the past cycle is calculated, when the absolute value of the change rate is greater than or equal to 20%, an abnormal early warning information is output.

[0030] Compared with the prior art, the method has the following beneficial effects:

[0031] 1. In the method for quasi-real-time imaging of underground medium structure under fluid action in a reservoir area, continuous seismic waveform data is continuously collected by a seismic monitoring network arranged in the reservoir area, the original continuous seismic waveform data is first standardized and preprocessed to obtain a standardized waveform data set, and the P wave and S wave phases and their arrival times in the microseismic signal are automatically identified and picked up using artificial intelligence technology to generate a standardized arrival time data set; then, based on the standardized waveform data set and the standardized arrival time data set, a ray distribution quantitative modeling algorithm is used to simulate and generate multiple virtual source-station pairs, and ray tracing model calculation is used to obtain ray distribution parameters; the ray distribution parameters are then judged in real time by establishing multi-dimensional data quality control standards, and when the parameters fully meet the standards, the Q value inversion process is automatically triggered, and when the parameters do not fully meet the standards, the standardized waveform data set, the standardized arrival time data set and the ray distribution parameters are intelligently updated by supplementing continuous seismic waveform data until the requirements are met; then, based on the standardized data set confirmed by quality control, a full-automatic inversion algorithm is used to calculate the Q value distribution of the underground medium in the reservoir area; finally, the underground medium structure imaging result in the reservoir area is generated based on the Q value distribution of the underground medium in the reservoir area and is output, and a continuous imaging cycle is formed by periodically repeating the above process. The advantages of the present application are:

[0032] First, by combining the automatic identification of phases by artificial intelligence, automatic quality control triggered inversion and full-process imaging cycle, the fundamental change from intermittent monitoring depending on seismic events to active continuous monitoring is realized, the "time blanking period" of traditional methods is effectively eliminated, and the monitoring timeliness is improved in quality;

[0033] Second, by establishing quantitative multi-dimensional data quality control standards and intelligent data supplement mechanism, the quality of inversion data is guaranteed, and the continuous seismic waveform data is fully utilized, which significantly improves the data utilization rate and the reliability of the inversion result.

[0034] Therefore, the present application not only realizes quasi-real-time dynamic imaging of the underground medium structure in the reservoir area through full-process automation and imaging cycle mechanism, greatly improves the monitoring timeliness, but also guarantees the accuracy and reliability of the imaging result through intelligent quality control and data optimization, providing an effective technical means for reservoir safety operation.

[0035] 2、The method for quasi-real-time imaging of underground medium structure under fluid action in a reservoir area, by establishing a ray distribution parameter system comprising a ray propagation path, a covering grid unit and a propagation azimuth, and constructing a three-dimensional quantitative quality control standard composed of ray coverage uniformity, ray density and ray azimuth coverage, differentiating the ray density threshold values of 3 or more and 2 or more for the shallow 0-5km and deep 5-10km respectively, while requiring the ray coverage variation coefficient of each underground medium grid unit to be less than or equal to 0.3 and the azimuth coverage to be greater than or equal to 180°, the ray distribution quality is strictly controlled through a series of quantitative standards, on the basis of ensuring the uniformity of ray spatial distribution and the integrity of directional coverage, the ray coverage density in the shallow area is effectively improved, thereby significantly improving the resolution and accuracy of the shallow stratum Q value inversion, and the imaging result can accurately match the special requirements of the shallow permeability monitoring in the reservoir area. Therefore, by constructing a scientific and reasonable ray distribution quality control system, the monitoring resolution of the shallow stratum is particularly optimized while ensuring the reliability of the global inversion, realizing the technical leap from "generalized imaging" to "accurate monitoring".

[0036] 3、The method for quasi-real-time imaging of underground medium structure under fluid action in a reservoir area, by adopting a Bayesian constrained full-automatic Q value inversion algorithm, extracting the seismic wave dispersion characteristics and main frequency parameters based on a standardized waveform data set, deriving the seismic wave attenuation parameter t* in combination with the standardized travel time data set, and further accurately optimizing the t* parameter through the amplitude envelope attenuation curve fitting and propagation parameter correction mechanism; at the same time, the corrected t* value and the travel time residual are taken as the core physical constraints and are integrated into the prior model based on wave theory, ensuring that the inversion model is consistent with the real seismic wave propagation physical law; finally, the Markov chain Monte Carlo algorithm with the minimum t* calculation error as the target is used for automatic iterative solution, and a strict numerical convergence criterion is set; this series of collaborative optimization makes the inversion process based on full automation, significantly reduces the sensitivity of Q value inversion to data noise by introducing wave theory prior and physical constraints, and still obtains stable and reliable high-precision Q value distribution results under complex geological background. Therefore, the present application not only realizes the automation of the Q value inversion process, but also fundamentally improves the physical reasonableness, numerical stability and anti-interference ability of the Q value inversion result through multiple optimization and physical constraints at the algorithm level, providing accurate and reliable core data basis for quasi-real-time imaging of underground medium structure in the reservoir area. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is the flowchart of the present application. DETAILED DESCRIPTION

[0038] The present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0039] Referring to Figure 1 A quasi-real-time imaging method for underground medium structure under fluid action in a reservoir area, the method comprising the following steps:

[0040] S1, first receive continuous seismic waveform data collected by a seismic monitoring network arranged in the reservoir area, then standardize and preprocess the continuous seismic waveform data to obtain a standardized waveform data set, then identify P and S wave phases contained in microseismic signals in the standardized waveform data set, then pick up P and S wave arrival time data from the identified P and S wave phases, and then standardize and process the P and S wave arrival time data to obtain a standardized arrival time data set;

[0041] S2, first simulate and generate a plurality of virtual source-station pairs based on the standardized waveform data set and the standardized arrival time data set using a ray distribution quantification modeling algorithm, and then calculate ray distribution parameters by ray tracing model taking the virtual source-station pairs as input;

[0042] S3, determine whether the ray distribution parameters obtained in S2 meet a pre-defined multi-dimensional data quality control standard, if yes, trigger a Q value inversion process; if not, first update the standardized waveform data set by supplementing continuous seismic waveform data, then update the standardized arrival time data set based on the updated standardized waveform data set, then update the ray distribution parameters based on the updated standardized waveform data set and the standardized arrival time data set, until the updated ray distribution parameters meet the multi-dimensional data quality control standard, and then trigger the Q value inversion process;

[0043] S4, after S3 triggers the Q value inversion process, the following calculation is performed: based on the standardized waveform data set and the standardized arrival time data set that meet the multi-dimensional data quality control standard determined in S3, calculate the Q value distribution of the underground medium in the reservoir area using a fully automatic inversion algorithm;

[0044] S5, generate and output the underground medium structure imaging result in the reservoir area based on the Q value distribution of the underground medium in the reservoir area obtained in S4, and the method ends.

[0045] In S2, the ray distribution parameters include: the propagation path of each ray, the underground medium grid cell through which each ray propagates, and the propagation azimuth angle of each ray in the underground medium grid cell through which it propagates.

[0046] In S3, the multi-dimensional data quality control standard includes:

[0047] Ray coverage uniformity standard: the coefficient of variation of the number of ray coverages of all underground medium grid cells is not greater than 0.3;

[0048] Ray density standard: at least 3 rays cover the underground medium grid unit in the range of 0-5 km depth, and at least 2 rays cover the underground medium grid unit in the range of 5-10 km depth;

[0049] Ray azimuth coverage standard: the propagation azimuth of all rays in a single underground medium grid unit covers no less than 180 degrees.

[0050] In the S3, if the ray distribution parameters obtained in the S2 do not meet the pre-defined multi-dimensional data quality control standard, the specific processing procedure is: first, output the identification of the underground medium grid unit to be supplemented, then extend the observation time of the station corresponding to the underground medium grid unit to supplement the continuous seismic waveform data, then input the supplemented continuous seismic waveform data into the procedure of the S1 to update the standardized waveform data set and the standardized arrival time data set, then based on the updated standardized waveform data set and the standardized arrival time data set, re-execute the S2 to update the ray distribution parameters, and finally execute the S3 to judge whether the updated ray distribution parameters meet the multi-dimensional data quality control standard.

[0051] In the S4, the specific steps for calculating the reservoir area underground medium Q value distribution by using the full-automatic inversion algorithm are:

[0052] S41, first calculate the cross-correlation function between stations based on the standardized waveform data set, then extract the Rayleigh wave or Love wave dispersion curve, and then determine the main frequency f0 of seismic wave propagation;

[0053] S42, first calculate the seismic wave propagation time t by using the standardized arrival time data, then combine the main frequency f0, and then calculate the seismic wave attenuation parameter t* by the formula:

[0054] S43, convert the calculated t* and the standardized arrival time data set into physical constraints, and integrate them into the prior model based on wave theory to construct the target model for inversion;

[0055] S44, input the target model into the Markov chain Monte Carlo algorithm to automatically iteratively solve Q value, and obtain the reservoir area underground medium Q value distribution.

[0056] In the S42, when deriving and calculating the seismic wave attenuation parameter t*, the specific derivation and calculation logic is: first, preliminarily calculate the t* initial value by using the exponential decay fitting of the amplitude envelope decay curve of the standardized waveform data, then combine the seismic wave propagation time t, and then correct the t* initial value by the correction formula: ; wherein, Δt is the waveform propagation time difference, A0 is the amplitude at the virtual source, and A is the station receiving amplitude. ​

[0057] In the S43, the construction logic of the prior model is that the t* correction value and the travel time residual standardized to the time data set are first taken as core inputs, then the theoretical attenuation characteristics of seismic wave propagation are calculated through the wave equation forward model, and the prior probability distribution of Q value inversion is established.

[0058] In the S44, the iteration process of the Markov chain Monte Carlo algorithm takes the minimum calculation error of t* as the objective function, and the convergence criterion is that the variance of the Q value calculation results of the continuous 500 iterations is less than or equal to 0.01.

[0059] In the S1, the identification of seismic phases and the picking of arrival time data are realized through a pre-trained artificial intelligence algorithm model, and the artificial intelligence algorithm model introduces the geological condition parameters of the reservoir area as a model regularization constraint.

[0060] The method further comprises S6: periodically repeating S1 to S5 to form a periodic imaging cycle, and in each cycle, after obtaining the current cycle reservoir area underground medium structure imaging result through S5, comparing it with the reservoir area underground medium structure imaging result of the previous cycle, and calculating the change rate of Q value at the same spatial position in the current cycle and the previous cycle, when the absolute value of the change rate is ≥20%, outputting an abnormal early warning information.

[0061] Embodiment 1:

[0062] In this embodiment, the underground medium structure of a certain large reservoir area is taken as an example for quasi-real-time imaging to monitor the reservoir water penetration dynamics. The reservoir area is provided with a dense monitoring network composed of 50 seismic stations, which continuously collects continuous seismic waveform data and transmits it to the processing center in real time.

[0063] S1, data acquisition and standardization processing:

[0064] The continuous seismic waveform data transmitted by the seismic monitoring network in real time is first standardized and preprocessed, including unified conversion to SAC format, normalization of signal amplitude to [-1, 1] interval, and alignment of the time stamps of all waveform data based on high-precision GPS clock to generate a standardized waveform data set; then, the standardized waveform data set is input into a pre-trained artificial intelligence algorithm model to automatically identify P and S wave phases in microseismic signals and pick their arrival times. The artificial intelligence algorithm model takes ≤10 minutes to process a single batch (such as continuous 1 hour) of data, and the accuracy rate of seismic phase identification and arrival time picking is stable at more than 92%, which is much higher than manual picking efficiency; finally, the picked arrival time data is arranged and formatted to generate a standardized arrival time data set containing station, event, phase type and accurate arrival time.

[0065] S2, ray distribution parameter calculation:

[0066] Based on the standardized waveform dataset and the standardized arrival time dataset output by S1, the ray distribution quantification modeling algorithm is adopted to simulate a plurality of virtual source-station pairs in the monitoring area, input these virtual pairs into the ray tracing model based on the fast marching method, calculate three core parameters of each ray: ① the propagation path from the virtual source to the corresponding station; ② the underground medium grid cell through which the path passes (the target area 0-10 km deep is divided into a plurality of grid cells according to 1 km*1 km*0.5 km, wherein 0-5 km is the shallow part and 5-10 km is the deep part); ③ the propagation azimuth angle of the ray when it reaches each underground medium grid cell, and finally a comprehensive ray distribution parameter is obtained.

[0067] S3, quality control and inversion triggering judgment:

[0068] A set of quantitative data quality control standards including three dimensions is established:

[0069] ① Ray coverage uniformity: the coefficient of variation of the number of ray coverages of all grid cells is required to be less than or equal to 0.3;

[0070] ② Ray density: the shallow underground medium grid cell (0-5 km) is required to be covered by at least 3 rays, and the deep underground medium grid cell (5-10 km) is required to be covered by at least 2 rays;

[0071] ③ Ray azimuth coverage: the azimuth coverage range of all rays in a single grid cell is required to be greater than or equal to 180°.

[0072] An inversion start automatic decision algorithm is adopted to real-time statistics of the current ray distribution parameter and comparison with the above standards one by one, if all the standards are met, the Q value inversion process is automatically triggered; if not met (for example, a key shallow underground medium grid cell is only covered by 2 rays), a data supplement cycle is automatically executed: the identification of the underground medium grid cell is output, instructing the surrounding stations to extend the observation time to supplement the continuous seismic waveform data; the supplemented continuous seismic waveform data is returned to the S1 process to update the standardized waveform dataset and the standardized arrival time dataset, and then the ray distribution parameter is updated by the S2 process; then it returns to this step to re-judge until all the standards are met, and then the subsequent inversion is triggered.

[0073] S4, automatic Q value inversion calculation:

[0074] When S3 triggers the inversion, a Bayesian constrained fully automatic Q value inversion algorithm is started:

[0075] Firstly, based on the standardized waveform data, the cross-correlation function between all pairs of stations is calculated, the Rayleigh wave dispersion curve is extracted, and the seismic wave propagation main frequency f0 is determined; secondly, the seismic wave propagation time t is calculated by using the standardized arrival time data, and combined with f0, the formula: The seismic wave attenuation parameter t* is derived by calculation; to improve the accuracy, the t* initial value is calculated by exponential decay fitting of the waveform amplitude envelope, and then the formula: The correction is performed; then, the travel time residual of the corrected t* value and the normalized time data is taken as a physical constraint, and is integrated into the prior model constructed based on wave theory to ensure that the model is consistent with the physical law of seismic wave propagation; finally, based on the prior model, the Markov chain Monte Carlo algorithm is used for automatic iterative solving, and the iteration process takes the t* calculation error minimization as the objective function, and when the convergence standard of "the variance of Q value calculation results of 500 consecutive iterations ≤0.01" is met, the calculation is automatically stopped, and the Q value distribution of the underground medium in the reservoir area is output.

[0076] S5, imaging result generation and output:

[0077] Based on the Q value distribution of the underground medium in the reservoir area obtained in step S4, the graphical underground medium structure imaging result of the reservoir area is generated and output. By periodically repeating the whole process of S1 to S5 to form a continuous imaging cycle, the latest underground structure imaging result is output each time. This continuous imaging mechanism based on the whole process cycle can systematically provide sequence data of the dynamic changes of the underground medium structure in the reservoir area, completely fill the "time blank period" of the traditional method, and finally realize the quasi-real-time dynamic imaging of the underground medium structure in the reservoir area under the action of fluid.

[0078] Embodiment 2:

[0079] The basic content is the same as that in embodiment 1: the difference is that:

[0080] In the data processing stage of S1, after the standardization preprocessing of the continuous seismic waveform data is completed, the generated standardized waveform data set is input into a pre-trained artificial intelligence algorithm model optimized for geological conditions.

[0081] Model optimization: in the training stage of the artificial intelligence algorithm model, in addition to learning the general seismic waveform features, specific geological condition parameters of the reservoir area are specially introduced as model regularization constraints, which include: rock layer wave velocity difference distribution model (constructed based on previous seismic exploration and drilling data); spatial distribution map of main faults and fracture zones; physical property parameter statistical data of different rock types.

[0082] Optimization effect: In practical application, this optimization enables the artificial intelligence algorithm model to fully consider the complex geological background of the reservoir area when performing seismic phase identification and arrival time picking, which is specifically manifested as: in the known high-speed rock layer distribution area, the model will adjust the judgment criteria for the P-wave first arrival time accordingly; in the vicinity of the fault fracture zone, the model will have higher tolerance and identification ability for waveform complexity; at the lithology change interface, the model will combine geological information to identify the possible wave field changes.

[0083] Embodiment 3:

[0084] The basic content is the same as that in Embodiment 1: the difference is that:

[0085] While generating the current cycle reservoir area underground medium structure imaging result, perform anomaly identification:

[0086] Data comparison: retrieve the current cycle underground medium Q value distribution data, and accurately match and compare with the last cycle underground medium Q value distribution data.

[0087] Change rate calculation: taking each independent underground medium grid element as a basic calculation unit, the following operation is performed: Q value change rate = (current Q value - previous Q value) / previous Q value x 100%.

[0088] Abnormality determination and early warning output: real-time judgment of the calculation result of each underground medium grid element, when the Q value change rate of any grid element is greater than or equal to 20%, it is determined that a significant medium anomaly occurs in the region, and an abnormal early warning information is generated and output immediately, the content structure includes:

[0089] Abnormal area location: three-dimensional geographic coordinate range.

[0090] Abnormal characteristics: Q value average drop of 28%.

[0091] Abnormal range: affecting an area of about 0.1 square kilometers.

[0092] Warning level: yellow warning (determined according to the change amplitude and the importance of the region).

[0093] Processing suggestion: it is suggested to focus on the combination of reservoir water level data and to prepare to start the encryption observation mode. The early warning information is immediately pushed to the reservoir safety operation personnel through the monitoring platform interface pop-up window, sound prompt, short message and other ways.

[0094] The above only describes the preferred embodiments of the present application, and the protection scope of the present application is not limited to the above embodiments, but any equivalent modifications or changes made by those skilled in the art according to the disclosed content of the present application shall be included in the protection scope recited in the claims.

Claims

1. A method for quasi-real-time imaging of subsurface medium structures under fluid action in a reservoir region, characterized by: The method comprises the following steps: S1, first receiving continuous seismic waveform data collected by a seismic monitoring network arranged in a reservoir area, then performing standardization preprocessing on the continuous seismic waveform data to obtain a standardized waveform data set, then identifying P and S wave phases contained in microseismic signals in the standardized waveform data set, then picking up P and S wave arrival time data from the identified P and S wave phases, and then performing standardization processing on the P and S wave arrival time data to obtain a standardized arrival time data set; S2, first simulating a plurality of virtual source-station pairs by using a ray distribution quantization modeling algorithm based on the standardized waveform data set and the standardized arrival time data set, and then calculating ray distribution parameters by using a ray tracing model with the virtual source-station pairs as input; S3, judging whether the ray distribution parameters obtained in S2 meet a pre-defined multi-dimensional data quality control standard, if yes, triggering a Q value inversion process, and if not, first updating the standardized waveform data set by supplementing continuous seismic waveform data, then updating the standardized arrival time data set based on the updated standardized waveform data set, then updating the ray distribution parameters based on the updated standardized waveform data set and the standardized arrival time data set, and repeating the above steps until the updated ray distribution parameters meet the multi-dimensional data quality control standard, and then triggering the Q value inversion process; S4, after the Q value inversion process is triggered in S3, performing the following calculation: calculating a Q value distribution of underground media in the reservoir area by using a full-automatic inversion algorithm based on the standardized waveform data set and the standardized arrival time data set that meet the multi-dimensional data quality control standard in S3; S5, generating and outputting underground media structure imaging results of the reservoir area based on the Q value distribution of the underground media in the reservoir area obtained in S4, and the method ends.

2. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 1, characterized in that: In S2, the ray distribution parameters comprise a propagation path of each ray, an underground media grid cell through which each ray propagates, and a propagation azimuth angle of each ray in the underground media grid cell through which the ray propagates.

3. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 2, characterized in that: In S3, the multi-dimensional data quality control standard comprises: a ray coverage uniformity standard: a variation coefficient of ray coverage times of all underground media grid cells is not greater than 0.3; a ray density standard: underground media grid cells in a shallow depth range of 0-5 km are covered by at least 3 rays, and underground media grid cells in a deep depth range of 5-10 km are covered by at least 2 rays; a ray azimuth angle coverage standard: a propagation azimuth angle coverage range of all rays in a single underground media grid cell is not less than 180 degrees.

4. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 3, characterized in that: In the S3, if the ray distribution parameter obtained in the S2 does not satisfy the predefined multi-dimensional data quality control standard, the specific processing procedure is: first, output the identification of the underground medium grid unit to be supplemented, then extend the observation time of the station corresponding to the underground medium grid unit to supplement the continuous seismic waveform data, then input the supplemented continuous seismic waveform data into the procedure of the S1 to update the standardized waveform data set and the standardized arrival time data set, then based on the updated standardized waveform data set and the standardized arrival time data set, re-execute the S2 to update the ray distribution parameter, and finally execute the S3 to judge whether the updated ray distribution parameter satisfies the multi-dimensional data quality control standard.

5. The method of claim 1, wherein: In the S4, the specific steps of calculating the Q value distribution of the underground medium in the reservoir area by using the full-automatic inversion algorithm are: S41, first calculate the cross-correlation function between stations based on the standardized waveform data set, then extract the Rayleigh wave or Love wave dispersion curve, and then determine the main frequency f0 of seismic wave propagation; S42, first use the normalized arrival time data to calculate the seismic wave propagation time t, combined with the main frequency f0, and then through the formula: Deduce the calculation of the seismic wave attenuation parameter t*; S43, convert the calculated t* and the standardized arrival time data set into physical constraints, and integrate the prior model based on wave theory to construct a target model for inversion; S44, input the target model into the Markov Chain Monte Carlo algorithm to automatically iterate and solve the Q value, and obtain the Q value distribution of the underground medium in the reservoir area.

6. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 5, characterized in that: In the S42, the seismic wave attenuation parameter t* is derived and calculated, and the specific derivation and calculation logic is as follows: firstly, the amplitude envelope attenuation curve of the standardized waveform data is fitted by using exponential attenuation to preliminarily calculate the t* initial value, and then the seismic wave propagation time t is combined to correct the t* initial value through a correction formula: The t* initial value is corrected; wherein, Δt is the waveform propagation time difference, A0 is the amplitude at the virtual seismic source, and A is the station receiving amplitude.

7. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 6, characterized in that: In the S43, the construction logic of the prior model is: first, input the travel time residual of the t* correction value and the standardized arrival time data set as the core, then calculate the theoretical attenuation characteristics of seismic wave propagation through the wave equation forward model to establish the prior probability distribution of Q value inversion.

8. The method for quasi-real-time imaging of subsurface medium structure in a reservoir area under fluid action according to claim 7, characterized in that: In the S44, the iteration process of the Markov Chain Monte Carlo algorithm takes the minimization of the calculation error of t* as the objective function, and the convergence criterion is that the variance of the Q value calculation result of 500 consecutive iterations is less than or equal to 0.

01.

9. The method of claim 1, wherein: In the S1, the identification of seismic phase and the picking of arrival time data are realized by a pre-trained artificial intelligence algorithm model, and the artificial intelligence algorithm model introduces the reservoir geological condition parameters as model regularization constraints.

10. The method of claim 1, wherein: The method further comprises S6: periodically repeating the execution of S1 to S5 to form a periodic imaging cycle, and in each cycle, after obtaining the current cycle reservoir underground medium structure imaging result through S5, comparing it with the past cycle reservoir underground medium structure imaging result, and then calculating the change rate of Q value at the same spatial position in the current cycle and the past cycle, when the absolute value of the change rate is greater than or equal to 20%, outputting an abnormal early warning information.

Images