Seismic exploration well depth analysis method based on complex surface environment

Abstract

The invention relates to the technical field of geological exploration, and particularly discloses a seismic exploration well depth analysis method based on a complex surface environment, which is used for solving the problems of blind well depth setting and low efficiency of a traditional artificial experience method. Comprising the steps of analyzing geological conditions of a complex earth surface and a shallow surface layer, constructing a middle-deep layer geological model, designing a partition well depth test, collecting multi-gradient well depth test data, performing automatic seismic reflection data analysis based on deep learning, performing intelligent well depth optimization inversion, performing field verification on the optimized well depth, constructing a three-dimensional data body and performing geological application analysis. According to the method, the well depth gradient and the landform recognition parameters are dynamically optimized for different landform areas such as a bedrock exposed area, a farmland section, a weathered object area and a gully area through a dynamic model fed back in real time, and automatic adaptive adjustment of the seismic exploration well depth is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic exploration, and more particularly to a seismic exploration well depth analysis method based on a complex surface environment. Background Art

[0002] In the open CNKI paper "3D Seismic Exploration Acquisition Parameter Design - Taking the Changping Mine Field in the Southern Qinshui Basin as an Example", the author took the coalbed methane field in the Qinshui Basin as an example, and conducted an analysis and study on the well depth and charge parameters using an 8-line 12-shot beam 3D observation system in bedrock exposed areas, farmland areas, weathered areas and valley areas. By analyzing the imaging effect, the excitation well depth and charge that obtained relatively good imaging effects in different near-surface environments were obtained from the excitation well depth tests, charge tests, receiving condition tests and optimal observation range tests in different environments. This facilitates the construction of a 3D profile wave group with obvious characteristics, good phase axis continuity, strong energy, rich geological information, and clear shallow and deep seismic interface reflection wave groups for exploration areas with similar geological conditions. Seismic time profiles are used to facilitate the construction of a three-dimensional data volume for this area, laying the foundation for the subsequent establishment of a seismic sequence framework, geological inversion, and prediction. However, manually evaluating the imaging effect from each test point in each experiment to obtain the excitation well depth for different environments is inefficient, causes high fatigue to personnel, and has poor accuracy. This is especially true in areas with complex terrain, large relative drop, drastic lateral lithologic changes, and loose Quaternary structures. The seismic reflection wave energy absorption and attenuation are severe, the seismic wave field is complex, and the wavelet consistency is poor. The exploration target layers are numerous and the occurrence conditions vary greatly. How to collect effective features from the massive amount of seismic data, experimental data, and image data to refine the seismic data, experimental data, and image data and analyze the appropriate seismic exploration well depth has become a difficult problem in the acquisition stage. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a seismic exploration well depth analysis method based on a complex surface environment. Through a dynamic model with real-time feedback, the well depth gradient and landform identification parameters are dynamically optimized for different landform areas such as bedrock exposed areas, farmland areas, weathered areas and valley areas, thereby realizing automatic adaptive adjustment of the seismic exploration well depth.

[0004] In this technical solution, the shallow stratum is the stratum of 5m-200m, and the medium-deep stratum is the stratum of 200m-1000m.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A seismic exploration well depth analysis method based on a complex surface environment comprises the following steps:

[0007] Step 1: Analyze complex surface and shallow geological conditions: Obtain surface orthophotos and digital elevation models through aerial photography to identify landform features. Use high-frequency surface radar to detect the thickness and distribution of loose structures in the shallow layer. Through on-site sampling and laboratory analysis, clarify the distribution of rock and soil particles and the degree of looseness and compaction. Analyze complex surface and shallow geological conditions, and use dynamic models to drive refined image segmentation and real-time parameter adjustment to improve the accuracy of surface and shallow geological analysis.

[0008] Step 2: Construct a mid-to-deep geological model: Historical drilling data is used to determine the target stratum's burial depth, thickness, and surrounding lithology. Cross-well logging data is used to determine the characteristics of the T3 wave reflection interface and its shielding effect on the overlying coal seam. A mid-to-deep seismic geological model is then established. Accurate quantitative modeling is then performed using drilling and logging data to accurately analyze the T3 wave shielding effect.

[0009] Step 3: Design zoning well depth test: Based on the analysis results of steps 1 and 2, use the geographic information system to divide the area. According to different seismic wave field characteristics, design several different levels of well depth gradients. Integrate real-time dynamic model feedback with refined adjustment of stepped well depth gradients to achieve high-precision and dynamic adaptation of well depth zoning.

[0010] Step 4: Collect multi-gradient well depth test data: Conduct seismic excitation tests according to the well depth gradient in each test zone, and use node-type three-dimensional seismic sensors to collect shallow and medium-deep seismic reflection data;

[0011] Step 5: Automated seismic reflection data analysis based on deep learning: Build an automatic recognition system based on a convolutional neural network. Use the labeled training data to train the model. Use the trained automatic recognition system to automatically extract waveform continuity, signal-to-noise ratio, and attenuation rate, perform data scoring, and optimize the model using transfer learning.

[0012] Step 6: Intelligent optimization and inversion of well depth: A machine learning mapping model is established by combining the model evaluation results of step 5 with the corresponding well depth data. The XGBoost intelligent optimization algorithm is used to perform well depth inversion analysis to determine the optimal well depth parameters for each partition.

[0013] As a further embodiment of the present invention, the method further comprises:

[0014] Step 7: Field verification of optimized well depth and construction of 3D data: Conduct actual field testing using the optimized well depth data, reacquire data, and use data processing software to create 3D seismic time profiles. Analyze wavefield characteristics and wavelet consistency. Use 3D seismic data to refine wavefield imaging and clarify shallow and mid-depth characteristics.

[0015] Step 8. Geological application analysis: Based on the optimized three-dimensional data body, carry out detailed seismic sequence grid construction and geological inversion analysis to clarify the precise distribution and occurrence conditions of coal seams, analyze geological structural changes and coalbed methane occurrence characteristics, and provide a geological structural interpretation report.

[0016] As a further solution of the present invention, in step 1, the surface orthophoto and digital elevation model are obtained by aerial photography, and the specific identification means for identifying landform features is to use a supervised classification method to obtain input feature data from the image and identify the landform types, including:

[0017] The image areas with RGB brightness greater than 200, color standard deviation less than 20, texture contrast greater than 0.7 and texture entropy less than 3 are identified as bedrock exposed areas;

[0018] Image areas with RGB color standard deviation less than 30, regularity coefficient greater than 0.8, and boundary sharpness index greater than 0.75 in the image are identified as farmland areas;

[0019] Image areas with RGB color standard deviation greater than 40, crack density greater than 5 / ㎡, and texture entropy greater than 5 in the image are identified as weathered areas;

[0020] The image areas with DEM slope greater than 15°, landform depth difference greater than 2 meters, and valley width standard deviation less than 0.5m are identified as valley areas.

[0021] As a further solution of the present invention, in step 1:

[0022] The input feature data of the bedrock exposed area are dynamically adjusted for RGB brightness ±10, color standard deviation ±5, texture contrast ±0.1, and texture entropy ±0.5 through the regional historical geological feature learning model;

[0023] The input feature data of farmland plots were adaptively adjusted using the agricultural cycle change model to adjust the color standard deviation ±5, regularity coefficient ±0.1, and boundary sharpness index ±0.05;

[0024] The input characteristic data of the weathering area are dynamically adjusted by the weathering process dynamic monitoring model to adjust the color standard deviation ±5, crack density ±1 / ㎡, and texture entropy ±0.5;

[0025] In the gully area, the landform evolution trend analysis model was used to dynamically adjust the slope by ±2°, the depth difference by ±0.5m, and the standard deviation of the gully width by ±0.1m.

[0026] As a further solution of the present invention, the regional historical geological characteristics learning model is a machine learning model constructed by regional stability index and historical rock weathering rate;

[0027] The agricultural cycle change model is a time series analysis model established based on seasonal crop planting patterns, crop types and growth cycles;

[0028] The weathering process dynamic monitoring model is a numerical model based on historical data analysis based on real-time rainfall, temperature and historical weathering rate;

[0029] The landform evolution trend analysis model is a real-time analysis model constructed based on historical erosion, rainfall data and landform evolution trends.

[0030] As a further solution of the present invention, in step 1, a high-precision orthophoto and digital elevation model of the surface are obtained by using drone aerial photography, and the specific image processing steps for identifying landform features and achieving preliminary regional segmentation are as follows:

[0031] Step 11: Perform radiometric correction and orthorectification on the acquired aerial image data;

[0032] Step 12: Use the Mean-shift algorithm to segment the original image into several homogeneous area blocks based on color similarity and spatial proximity;

[0033] Step 13: extract the quantitative features of the images of each segmented area and preliminarily establish an image feature library of different landform categories;

[0034] Step 14: classify the preliminary segmented areas using a random forest classification algorithm, and preliminarily identify the types of landform areas based on the classification results;

[0035] Step 15: Adaptively perform fine image processing on the preliminarily identified areas: construct a machine learning model based on the historical geological stability index and rock weathering rate, dynamically adjust the RGB brightness, color standard deviation, texture contrast and texture entropy thresholds, optimize the thresholds through the feedback results of regional characteristics, finely adjust the segmentation boundaries of the bedrock exposed areas, construct a time series analysis model of seasonal agricultural cycle changes, dynamically adjust the color standard deviation, regularity coefficient and boundary sharpness index, finely segment and identify farmland boundaries and crop types, construct a dynamic data value analysis model based on the regional historical weathering rate and real-time climate monitoring data, dynamically adjust the crack identification density and texture entropy index, finely adjust the contour boundaries of the weathered area, and construct a landform evolution trend analysis model in combination with digital elevation data. Adjust the slope, landform depth difference and valley width thresholds in real time to finely depict the contour details of the valley area.

[0036] As a further solution of the present invention, in step one, a high-frequency surface radar is used to detect the thickness and spatial distribution characteristics of the loose structure of the shallow surface layer, and the characteristic data obtained include the thickness of the loose layer, the intensity value of the reflection wave at the interface of the loose layer, and the distribution of the radar wave propagation velocity; through on-site sampling and laboratory analysis, the distribution of rock and soil particles and the degree of loose compaction are clarified, and the data obtained include the rock and soil particle size distribution, grading coefficient, penetration resistance value, and permeability coefficient.

[0037] As a further solution of the present invention, in step 2, the process of constructing the medium-deep seismic geological condition model includes:

[0038] Step 21, historical drilling data analysis: Integrate existing historical drilling data to quantitatively determine the target stratum burial depth, coal seam thickness, upper and lower surrounding rock types, and lithologic physical properties, including density, porosity, and acoustic wave velocity;

[0039] Step 22, using the cross-well logging method to determine the T3 wave characteristics: using the cross-well logging method of acoustic logging, density logging, and resistivity logging, determine the acoustic wave velocity, density, and resistivity of the T3 wave reflection interface, analyze the T3 wave interface reflection wave intensity, wave continuity index, and its energy attenuation ratio of the seismic reflection wave of the underlying coal seam, form a shielding effect intensity map, and quantitatively analyze the target formation characteristics and the T3 wave shielding effect.

[0040] As a further solution of the present invention, in step three, the method for performing well depth test zoning based on the data and features obtained in steps one and two includes:

[0041] Step 31, importing the acquired data into the GIS system: importing the loose layer thickness, interface reflection wave intensity, radar wave propagation velocity, particle distribution, and penetration resistance data acquired in step 1, and the mid-deep coal seam burial depth, coal seam thickness, surrounding rock type and lithologic physical properties, and T3 wave shielding effect intensity data acquired in step 2 into the GIS system;

[0042] Step 32, designing test well depth zones: In a GIS environment, spatial interpolation is used to perform spatial continuity analysis on the imported data to generate a continuous spatial distribution layer. The overlay analysis function of the GIS system is used to overlay the key parameter layers of the shallow layer and the medium-deep layer, and regional analysis is performed based on parameter similarity.

[0043] Step 33, set the well depth gradient for each partition: combine the thickness of the shallow surface loose structure and the buried depth data of the medium and deep coal seams in the partition to determine each adapted well depth gradient, and output a well depth test partition map clearly marked under different surface and geological conditions on the GIS system platform.

[0044] As a further solution of the present invention, in step 33, the specific method for determining each adapted well depth gradient by combining the data of each morphological zone, the thickness of the shallow surface loose structure within the zone, and the buried depth of the medium-deep coal seam within the zone includes:

[0045] Bedrock exposed area: The initial well depth gradient is 5m-10m, and the gradient range is dynamically adjusted to ±1m based on the feedback of the real-time updated rock weathering rate model;

[0046] Farmland: The initial well depth gradient is set at 10m-15m, and the gradient range is dynamically adjusted to ±1.5m based on real-time agricultural cycle change model feedback;

[0047] Weathering area: The initial well depth gradient is set at 15m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time weathering process dynamic monitoring model;

[0048] Valley area: The initial well depth gradient is set at 10m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time landform evolution trend analysis model.

[0049] The technical effect of the seismic exploration well depth analysis method based on complex surface environment of the present invention is as follows: the present invention comprehensively utilizes multi-source geological data and a dynamic model driving mechanism of real-time feedback, accurately adapts to the differentiated characteristics of different landforms in bedrock exposed areas, farmland areas, weathered areas and valley areas, realizes the refinement and automatic dynamic adjustment of well depth gradient parameters, effectively improves the accuracy, reliability and analysis efficiency of well depth parameter determination in complex surface environments, effectively overcomes the problems of low efficiency, high fatigue and insufficient accuracy of traditional manual experience judgment process, improves the quality of seismic wave field imaging in complex terrain areas, and provides a reliable and high-quality data foundation for subsequent fine seismic sequence framework construction, geological inversion analysis and coalbed methane exploration prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a seismic exploration well depth analysis method based on a complex surface environment according to the present invention;

[0051] Figure 2 This is a flowchart of the specific operations of image processing for identifying landform features and achieving preliminary regional segmentation in the present invention;

[0052] Figure 3 A diagram showing the construction process of the deep-seismic geological condition model of the present invention;

[0053] Figure 4 This is a flow chart of the method for performing well depth test zoning based on the data and features obtained in steps one and two of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0055] like Figure 1 As shown, the present invention proposes a seismic exploration well depth analysis method based on a complex surface environment, comprising the following steps:

[0056] Step 1: Analyze complex surface and shallow geological conditions: Obtain surface orthophotos and digital elevation models through aerial photography to identify landform features. Use high-frequency surface radar to detect the thickness and distribution of loose structures in the shallow layer. Through on-site sampling and laboratory analysis, clarify the distribution of rock and soil particles and the degree of looseness and compaction. Analyze complex surface and shallow geological conditions, and use dynamic models to drive refined image segmentation and real-time parameter adjustment to improve the accuracy of surface and shallow geological analysis.

[0057] Step 2: Construct a mid-to-deep geological model: Historical drilling data is used to determine the target stratum's burial depth, thickness, and surrounding lithology. Cross-well logging data is used to determine the characteristics of the T3 wave reflection interface and its shielding effect on the overlying coal seam. A mid-to-deep seismic geological model is then established. Accurate quantitative modeling is then performed using drilling and logging data to accurately analyze the T3 wave shielding effect.

[0058] Step 3: Design zoning well depth test: Based on the analysis results of steps 1 and 2, use the geographic information system to divide the area. According to different seismic wave field characteristics, design several different levels of well depth gradients. Integrate real-time dynamic model feedback with refined adjustment of stepped well depth gradients to achieve high-precision and dynamic adaptation of well depth zoning.

[0059] Step 4: Collect multi-gradient well depth test data: Conduct seismic excitation tests according to the well depth gradient in each test zone, and use node-type three-dimensional seismic sensors to collect shallow and medium-deep seismic reflection data;

[0060] Step 5: Automated seismic reflection data analysis based on deep learning: Build an automatic recognition system based on a convolutional neural network. Use the labeled training data to train the model. Use the trained automatic recognition system to automatically extract waveform continuity, signal-to-noise ratio, and attenuation rate, perform data scoring, and optimize the model using transfer learning.

[0061] Step 6: Intelligent optimization and inversion of well depth: A machine learning mapping model is established by combining the model evaluation results of step 5 with the corresponding well depth data. The XGBoost intelligent optimization algorithm is used to perform well depth inversion analysis to determine the optimal well depth parameters for each partition.

[0062] It should be noted that the method further includes:

[0063] Step 7: Field verification of optimized well depth and construction of 3D data: Conduct actual field testing using the optimized well depth data, reacquire data, and use data processing software to create 3D seismic time profiles. Analyze wavefield characteristics and wavelet consistency. Use 3D seismic data to refine wavefield imaging and clarify shallow and mid-depth characteristics.

[0064] Step 8. Geological application analysis: Based on the optimized three-dimensional data body, carry out detailed seismic sequence grid construction and geological inversion analysis to clarify the precise distribution and occurrence conditions of coal seams, analyze geological structural changes and coalbed methane occurrence characteristics, and provide a geological structural interpretation report.

[0065] It should be noted that in step 1, the surface orthophotos and digital elevation models are obtained through aerial photography. The specific identification method for identifying landform features is to use supervised classification methods to obtain input feature data from the images and identify the landform types, including:

[0066] The image areas with RGB brightness greater than 200, color standard deviation less than 20, texture contrast greater than 0.7 and texture entropy less than 3 are identified as bedrock exposed areas;

[0067] Image areas with RGB color standard deviation less than 30, regularity coefficient greater than 0.8, and boundary sharpness index greater than 0.75 in the image are identified as farmland areas;

[0068] Image areas with RGB color standard deviation greater than 40, crack density greater than 5 / ㎡, and texture entropy greater than 5 in the image are identified as weathered areas;

[0069] The image areas with DEM slope greater than 15°, landform depth difference greater than 2 meters, and valley width standard deviation less than 0.5m are identified as valley areas.

[0070] A supervised classification method is used to perform preliminary zoning of aerial orthophotos and DEM data. RGB brightness, color standard deviation, texture contrast, texture entropy, slope, landform depth difference and valley width standard deviation are used to accurately identify bedrock exposed areas, farmland areas, weathered areas and valley areas. Combined with dynamic adjustment model feedback, the segmentation boundaries are refined in real time, effectively improving the adaptability of landform zoning and seismic well depth, and realizing the refinement of zoning and intelligent optimization of well depth gradients under complex surface conditions.

[0071] It should be noted that in step 1:

[0072] The input feature data of the bedrock exposed area are dynamically adjusted for RGB brightness ±10, color standard deviation ±5, texture contrast ±0.1, and texture entropy ±0.5 through the regional historical geological feature learning model;

[0073] The input feature data of farmland plots were adaptively adjusted using the agricultural cycle change model to adjust the color standard deviation ±5, regularity coefficient ±0.1, and boundary sharpness index ±0.05;

[0074] The input characteristic data of the weathering area are dynamically adjusted by the weathering process dynamic monitoring model to adjust the color standard deviation ±5, crack density ±1 / ㎡, and texture entropy ±0.5;

[0075] In the gully area, the landform evolution trend analysis model was used to dynamically adjust the slope by ±2°, the depth difference by ±0.5m, and the standard deviation of the gully width by ±0.1m.

[0076] By introducing regional historical geological feature learning models, agricultural cycle change models, weathering process dynamic monitoring models and landform evolution trend analysis models, the thresholds of various image feature parameters of bedrock exposed areas, farmland areas, weathering areas and valley areas are dynamically adjusted, and the landform recognition accuracy and contour details are optimized in real time, effectively enhancing the zoning accuracy and robustness, so that well depth analysis can be accurately matched to geologically complex areas, significantly improving the accuracy of seismic imaging and the efficiency of geological information extraction, and ensuring the accuracy and automation of well depth optimization in complex environment seismic exploration.

[0077] It should be noted that the regional historical geological characteristics learning model is a machine learning model constructed by using the regional stability index and historical rock weathering rate;

[0078] The agricultural cycle change model is a time series analysis model established based on seasonal crop planting patterns, crop types and growth cycles;

[0079] The weathering process dynamic monitoring model is a numerical model based on historical data analysis based on real-time rainfall, temperature and historical weathering rate;

[0080] The landform evolution trend analysis model is a real-time analysis model constructed based on historical erosion, rainfall data and landform evolution trends.

[0081] The regional historical geological characteristics learning model uses the regional stability index and historical rock weathering rate because they reflect the long-term stability and weathering rate of the regional rock mass, are directly related to the weathering degree and stability changes of the rock mass, and can dynamically and accurately distinguish the weathering differences between the bedrock exposed area and other areas; the agricultural cycle change model selects seasonal crop planting patterns, crop types and growth cycles. These parameters can accurately reflect the seasonal changes and regular characteristics of farmland, which is conducive to the adaptive adjustment of the image features of the farmland area and ensures the recognition accuracy; the weathering process dynamic monitoring model selects real-time rainfall, temperature and historical weathering rate. These parameters can dynamically reflect the intensity and speed of the weathering process changes, and timely capture the changing characteristics of the development of rock cracks in the weathering area to adapt to the actual changes in the characteristics of the weathering area; the landform evolution trend analysis model selects historical erosion, rainfall data and landform evolution trends because these data can dynamically reflect the continuous changing trends of the slope, width and erosion depth of the valley area, thereby determining the contour characteristics of the valley area in real time and accurately.

[0082] It should be noted that if Figure 2 As shown in the figure, in step 1, high-precision orthophotos and digital elevation models of the surface are obtained by drone aerial photography, and the specific image processing steps for identifying landform features and achieving preliminary regional segmentation are as follows:

[0083] Step 11: Perform radiometric correction and orthorectification on the acquired aerial image data;

[0084] Step 12: Use the Mean-shift algorithm to segment the original image into several homogeneous area blocks based on color similarity and spatial proximity;

[0085] Step 13: extract the quantitative features of the images of each segmented area and preliminarily establish an image feature library of different landform categories;

[0086] Step 14: classify the preliminary segmented areas using a random forest classification algorithm, and preliminarily identify the types of landform areas based on the classification results;

[0087] Step 15: Adaptively perform fine image processing on the preliminarily identified areas: construct a machine learning model based on the historical geological stability index and rock weathering rate, dynamically adjust the RGB brightness, color standard deviation, texture contrast and texture entropy thresholds, optimize the thresholds through the feedback results of regional characteristics, finely adjust the segmentation boundaries of the bedrock exposed areas, construct a time series analysis model of seasonal agricultural cycle changes, dynamically adjust the color standard deviation, regularity coefficient and boundary sharpness index, finely segment and identify farmland boundaries and crop types, construct a dynamic data value analysis model based on the regional historical weathering rate and real-time climate monitoring data, dynamically adjust the crack identification density and texture entropy index, finely adjust the contour boundaries of the weathered area, and construct a landform evolution trend analysis model in combination with digital elevation data. Adjust the slope, landform depth difference and valley width thresholds in real time to finely depict the contour details of the valley area.

[0088] This method achieves preliminary identification of landform types through the correction of high-precision aerial images and DEM data, preliminary regional segmentation using the Mean-shift algorithm, and the random forest classification algorithm. It also uses the historical geological stability index and rock weathering rate model, the agricultural cycle change time series analysis model, the weathering process dynamic monitoring model, and the real-time analysis model of landform evolution trends to dynamically and adaptively adjust the image feature thresholds of bedrock exposed areas, farmland areas, weathered areas, and valley areas. This method solves the technical problems of difficult manual identification, large subjective errors, and insufficient regional division accuracy caused by fuzzy landform boundaries and large changes in geological conditions in complex terrain areas. It significantly improves the accuracy and efficiency of landform regional boundary identification and fine characterization, ensures the subsequent seismic well depth parameter gradient design is accurate and has strong real-time adaptability, and ultimately realizes the automation, efficiency, and accuracy of seismic exploration well depth parameter determination in complex surface environments, laying a solid data foundation for high-quality seismic exploration imaging and subsequent fine geological interpretation.

[0089] It should be noted that in step one, a high-frequency surface radar is used to detect the thickness and spatial distribution characteristics of the loose structure of the shallow surface layer. The characteristic data obtained include the thickness of the loose layer, the intensity value of the reflection wave at the interface of the loose layer, and the distribution of the radar wave propagation velocity. Through on-site sampling and laboratory analysis, the distribution of rock and soil particles and the degree of loose compaction are clarified. The data obtained include the particle size distribution of rock and soil particles, grading coefficient, penetration resistance value, and permeability coefficient.

[0090] This method uses high-frequency surface radar to obtain the thickness of the loose layer, the intensity of the interface reflection wave, and the radar wave propagation velocity. Combined with the particle size distribution, gradation coefficient, penetration resistance value, and permeability coefficient obtained through field sampling and laboratory analysis, it accurately quantifies the seismic physical characteristics of the loose structure of the shallow surface layer. It solves the problems of severe seismic wave attenuation and poor wavelet consistency caused by the complex rock and soil structure of the Quaternary shallow surface layer. It provides reliable data support for the fine delineation of shallow seismic well depth zoning and the optimization of source parameters, significantly improving the accuracy of seismic exploration in complex areas.

[0091] It should be noted that if Figure 3 As shown in FIG, in step 2, the process of constructing the medium-deep seismic geological condition model includes:

[0092] Step 21, historical drilling data analysis: Integrate existing historical drilling data to quantitatively determine the target stratum burial depth, coal seam thickness, upper and lower surrounding rock types, and lithologic physical properties, including density, porosity, and acoustic wave velocity;

[0093] Step 22, using the cross-well logging method to determine the T3 wave characteristics: using the cross-well logging method of acoustic logging, density logging, and resistivity logging, determine the acoustic wave velocity, density, and resistivity of the T3 wave reflection interface, analyze the T3 wave interface reflection wave intensity, wave continuity index, and its energy attenuation ratio of the seismic reflection wave of the underlying coal seam, form a shielding effect intensity map, and quantitatively analyze the target formation characteristics and the T3 wave shielding effect.

[0094] The reflection characteristics of the T3 wave interface are analyzed because the T3 wave interface is usually an interface with significant differences in lithologic properties between the coal seam top and the siltstone roof. The high-intensity reflection wave generated by this interface has a significant shielding and energy attenuation effect on the seismic reflection wave of the underlying coal seam, which will lead to a decrease in the seismic imaging quality of the medium-deep target coal seam in seismic exploration. Therefore, quantitative analysis of the reflection wave intensity, continuity and shielding effect of the T3 wave through acoustic logging, density logging and resistivity logging can effectively overcome the interference and data quality attenuation problems of the seismic wave field under complex geological conditions, significantly improve the identification accuracy and imaging quality of the target coal seam, and provide clear data support for the accurate determination of the well depth gradient.

[0095] It should be noted that if Figure 4 As shown, in step three, the method for performing well depth test zoning based on the data and features obtained in steps one and two includes:

[0096] Step 31, importing the acquired data into the GIS system: importing the loose layer thickness, interface reflection wave intensity, radar wave propagation velocity, particle distribution, and penetration resistance data acquired in step 1, and the mid-deep coal seam burial depth, coal seam thickness, surrounding rock type and lithologic physical properties, and T3 wave shielding effect intensity data acquired in step 2 into the GIS system;

[0097] Step 32, designing test well depth zones: In a GIS environment, spatial interpolation is used to perform spatial continuity analysis on the imported data to generate a continuous spatial distribution layer. The overlay analysis function of the GIS system is used to overlay the key parameter layers of the shallow layer and the medium-deep layer, and regional analysis is performed based on parameter similarity.

[0098] Step 33, set the well depth gradient for each partition: combine the thickness of the shallow surface loose structure and the buried depth data of the medium and deep coal seams in the partition to determine each adapted well depth gradient, and output a well depth test partition map clearly marked under different surface and geological conditions on the GIS system platform.

[0099] This method integrates and imports data on the thickness of shallow loose structure, interface reflection wave intensity, radar wave propagation velocity, rock and soil particle distribution, compaction degree, coal seam burial depth, coal seam thickness, surrounding rock physical properties and T3 wave shielding effect intensity into the GIS system, and uses spatial interpolation and overlay analysis to perform preliminary and fine regional zoning. The well depth gradient is dynamically determined according to the combined differences in shallow and medium-deep geological conditions, achieving refined distribution and adaptive dynamic feedback adjustment of well depth parameters in different geomorphological areas. This overcomes the problem of traditional methods that are difficult to accurately quantify the difficulty in adapting well depth parameters due to the differences in complex surface conditions and geological conditions, and significantly improves the intelligence of well depth zoning and the quality of seismic imaging.

[0100] It should be noted that in step 33, the specific method for determining each adapted well depth gradient by combining the data of each morphological zone, the thickness of the shallow surface loose structure within the zone, and the buried depth of the medium-deep coal seam within the zone includes:

[0101] Bedrock exposed area: The initial well depth gradient is 5m-10m, and the gradient range is dynamically adjusted to ±1m based on the feedback of the real-time updated rock weathering rate model;

[0102] Farmland: The initial well depth gradient is set at 10m-15m, and the gradient range is dynamically adjusted to ±1.5m based on real-time agricultural cycle change model feedback;

[0103] Weathering area: The initial well depth gradient is set at 15m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time weathering process dynamic monitoring model;

[0104] Valley area: The initial well depth gradient is set at 10m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time landform evolution trend analysis model.

[0105] Because the thickness of the superficial loose structure, rock and soil properties, and reflection characteristics of the mid- to deep-layer geological interfaces vary significantly between exposed bedrock areas, farmland areas, weathered areas, and gully areas, a uniform well depth gradient cannot accurately adapt to the changing geological conditions in each region. Therefore, differentiated well depth gradients are set based on the geological, geomorphological, and dynamic characteristics of different regions. The well depth range is dynamically adjusted in real time based on model feedback to accurately match the well depth parameters to the actual geological conditions, ensure sufficient excitation of seismic reflection wave energy, and improve seismic imaging quality and geological information accuracy. This method sets initial well depth gradients for different geomorphological zones and uses a real-time feedback mechanism and dynamic fine-tuning method to dynamically adjust the well depth parameters based on a regional historical geological characteristic learning model, a time-series analysis model for agricultural cycle changes, a dynamic monitoring model for weathering processes, and a real-time analysis model for geomorphological evolution trends. This method addresses the problems of traditional manual well depth setting methods, which cannot accurately respond to changes in surface conditions in real time and have poor gradient adaptability. It achieves real-time and accurate optimization of well depth parameters, effectively improving seismic data consistency, imaging quality, and the reliability of geological interpretation, and significantly enhancing the efficiency and intelligence of seismic exploration.

[0106] In order to more realistically and meticulously demonstrate the advantages of this method in seismic exploration well depth analysis under complex surface environments, the following four specific examples are given for detailed comparative testing. Before the comparative testing of the examples, the following objective test conditions were agreed upon in advance to ensure the accuracy and reliability of the data:

[0107] (1) The test area was selected in the adjacent area of ​​Changping Mine Field in the southern Qinshui Basin to ensure that the geological environment conditions, shallow surface structure thickness, and coal seam burial depth characteristics within the region were similar;

[0108] (2) All tests used a Geometrics Geode 3D seismic acquisition device with a consistent sampling frequency (1 ms) and record length (2000 ms);

[0109] (3) On-site construction personnel have unified training and consistent professional experience to avoid differences in personnel operations;

[0110] (3) All data were collected under clear weather conditions with no rainfall or significant temperature difference to avoid external interference factors affecting the experimental results.

[0111] Example 1: Bedrock exposed area (area 2.0 km 2 )

[0112] Manual group: Using traditional experience, the test was carried out at a fixed well depth of 7.5m. The average data imaging clarity of the 10 test points was 84.5%, and the waveform continuity was 79.5%. The reflected wave energy attenuation was obvious, and the boundary between the T3 reflected wave and the target coal seam reflected wave was blurred, affecting the imaging quality.

[0113] Test group: This method uses a historical geological feature learning model to analyze the regional stability index (0.85) and historical rock weathering rate (average annual weathering rate of 0.3 cm / a) in real time, dynamically adjusts the RGB brightness to 205 (±8), the color standard deviation to 16 (±3), the texture contrast to 0.72 (±0.08), and the texture entropy to 2.6 (±0.4). Through model feedback, the well depth is optimized to 7±0.5m. The average imaging clarity of the 10 test points is improved to 98.2%, and the waveform continuity is improved to 95.3%, which are 13.7% and 16.5% higher than the manual group, respectively. This solves the problem of single well depth and blurred imaging caused by manual experience, and significantly improves data quality.

[0114] Example 2: Farmland (area 3.5km 2 )

[0115] Manual group: The well depth was randomly set within the range of 10-15m (actually 12m). The average seismic imaging clarity of the 10 test points was 78.3%, and the waveform continuity was 74.8%. Due to the significant influence of seasonal changes, the seismic wave energy loss was obvious.

[0116] Test group: The agricultural cycle change LSTM time series analysis model was applied, and the color standard deviation was fine-tuned to 27 (±4), the regularity coefficient to 0.82 (±0.05), and the boundary sharpness index to 0.77 (±0.04) based on the crop type (corn) and growth cycle (80 days after sowing). The model's real-time feedback optimized the well depth to 11.5±0.75m. The imaging clarity of the 10 test points was improved to 93.5%, and the waveform continuity was improved to 91.2%, which were 15.2% and 16.4% higher than the manual group, respectively. This effectively solved the problem of unstable seismic wave propagation caused by seasonal disturbances in farmland areas.

[0117] Example 3: Weathering area (area 2.8km 2 )

[0118] Manual group: The traditional method fixed the well depth at 17m based on experience. The average waveform continuity of the 10 test points was 69.8% and the signal-to-noise ratio was 63.5%. Due to the large differences in the degree of development of weathering cracks and the rapid attenuation of seismic wave energy, it was difficult to effectively extract the reflected wave information of the target coal seam.

[0119] Test group: This method is based on a numerical model for dynamic monitoring of weathering processes, combined with real-time rainfall (recent cumulative rainfall of 85 mm) and temperature data (annual average temperature of 15°C), adjusted the image color standard deviation to 45 (±5), the crack density to 6 / ㎡ (±0.8 / ㎡), and the texture entropy to 5.3 (±0.4). Based on feedback, the well depth gradient was dynamically optimized to 17±1m. The waveform continuity of the 10 test points was improved to 88.5%, and the signal-to-noise ratio was improved to 89.6%, which were 18.7% and 26.1% higher than the manual group, respectively. This effectively solved the problems of severe energy attenuation and low signal-to-noise ratio in the weathering crack zone.

[0120] Example 4: Valley area (area 1.5km 2 )

[0121] Manual group: The well depth was randomly set within the range of 10-20m (actually 15m). The wave field consistency of the data at 10 test points was 77.8% and the energy resolution was 74.5%. Due to the drastic changes in the valley terrain and the complex interface undulations, the reflection wave group characteristics were not obvious and the energy loss was serious.

[0122] Test group: Using a real-time analysis model of landform evolution trends, based on historical erosion (average annual erosion of 0.25m / a) and recent rainfall data (accumulated rainfall of 60mm in the last 7 days), the DEM slope was fine-tuned to 16° (±1.5°), the depth difference was 2.2m (±0.3m), and the standard deviation of the valley width was 0.4m (±0.05m). The dynamic feedback optimized well depth gradient was 15±0.5m. The wave field consistency of the data at the 10 test points was improved to 94.3%, and the energy resolution was improved to 92.5%, which were 16.5% and 18% higher than those of the manual group, respectively. This effectively overcomes the problems of discontinuous seismic data and low energy resolution caused by landform changes in the valley area.

[0123] In summary, the present invention comprehensively utilizes multi-source geological data and a dynamic model-driven mechanism with real-time feedback, accurately adapts to the differentiated characteristics of different landforms in bedrock exposed areas, farmland areas, weathered areas and valley areas, and realizes the refinement and automatic dynamic adjustment of well depth gradient parameters, effectively improving the accuracy, reliability and analysis efficiency of well depth parameter determination under complex surface environments, effectively overcoming the problems of low efficiency, high fatigue and insufficient accuracy of the traditional manual experience judgment process, improving the quality of seismic wave field imaging in complex terrain areas, and providing a reliable and high-quality data foundation for subsequent fine seismic sequence framework construction, geological inversion analysis and coalbed methane exploration prediction.

[0124] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0125] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A seismic exploration well depth analysis method based on a complex surface environment, characterized in that: The steps include: Step 1: Analyze complex surface and shallow geological conditions: Obtain surface orthophotos and digital elevation models through aerial photography to identify landform features. Use high-frequency surface radar to detect the thickness and distribution of loose structures in the shallow layer. Through on-site sampling and laboratory analysis, clarify the distribution of rock and soil particles and the degree of looseness and compaction, and analyze the complex surface and shallow geological conditions. Step 2: Construct a medium-deep geological model: Based on historical drilling data, determine the target stratum's burial depth, thickness, and surrounding lithology data. Using cross-well logging data, determine the T3 wave reflection interface characteristics and its shielding effect on the overlying coal seam, and establish a medium-deep seismic geological model. Step 3: Designing zoned well depth tests: Based on the analysis results of steps 1 and 2, use a geographic information system to divide the area and design several different levels of well depth gradients based on different seismic wave field characteristics; Step 4: Collect multi-gradient well depth test data: Conduct seismic excitation tests according to the well depth gradient in each test zone, and use node-type three-dimensional seismic sensors to collect shallow and medium-deep seismic reflection data; Step 5: Automated seismic reflection data analysis based on deep learning: Build an automatic recognition system based on a convolutional neural network. Use the labeled training data to train the model. Use the trained automatic recognition system to automatically extract waveform continuity, signal-to-noise ratio, and attenuation rate, perform data scoring, and optimize the model using transfer learning. Step 6: Intelligent optimization and inversion of well depth: A machine learning mapping model is established by combining the model evaluation results of step 5 with the corresponding well depth data. The XGBoost intelligent optimization algorithm is used to perform well depth inversion analysis to determine the optimal well depth parameters for each partition.

2. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 1, characterized in that: The method further comprises: Step 7: Field verification of optimized well depth and construction of 3D data: Conduct actual field testing using the optimized well depth data, reacquire data, and use data processing software to create 3D seismic time profiles. Analyze wavefield characteristics and wavelet consistency. Use 3D seismic data to refine wavefield imaging and clarify shallow and mid-depth characteristics. Step 8. Geological application analysis: Based on the optimized three-dimensional data body, carry out detailed seismic sequence grid construction and geological inversion analysis to clarify the precise distribution and occurrence conditions of coal seams, analyze geological structural changes and coalbed methane occurrence characteristics, and provide a geological structural interpretation report.

3. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 2, characterized in that: In step 1, aerial photography is used to obtain surface orthophotos and digital elevation models. The specific identification method for landform features is to use supervised classification methods to obtain input feature data from the images and identify the landform types, including: The image areas with RGB brightness greater than 200, color standard deviation less than 20, texture contrast greater than 0.7 and texture entropy less than 3 are identified as bedrock exposed areas; Image areas with RGB color standard deviation less than 30, regularity coefficient greater than 0.8, and boundary sharpness index greater than 0.75 in the image are identified as farmland areas; Image areas with RGB color standard deviation greater than 40, crack density greater than 5 / ㎡, and texture entropy greater than 5 in the image are identified as weathered areas; The image areas with DEM slope greater than 15°, landform depth difference greater than 2 meters, and valley width standard deviation less than 0.5m are identified as valley areas.

4. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 3, characterized in that: In step one: The input feature data of the bedrock exposed area are dynamically adjusted for RGB brightness ±10, color standard deviation ±5, texture contrast ±0.1, and texture entropy ±0.5 through the regional historical geological feature learning model; The input feature data of farmland plots were adaptively adjusted using the agricultural cycle change model to adjust the color standard deviation ±5, regularity coefficient ±0.1, and boundary sharpness index ±0.05; The input characteristic data of the weathering area are dynamically adjusted by the weathering process dynamic monitoring model to adjust the color standard deviation ±5, crack density ±1 / ㎡, and texture entropy ±0.5; In the gully area, the landform evolution trend analysis model was used to dynamically adjust the slope by ±2°, the depth difference by ±0.5m, and the standard deviation of the gully width by ±0.1m.

5. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 4, characterized in that: The regional historical geological characteristics learning model is a machine learning model constructed by using the regional stability index and the historical rock weathering rate; The agricultural cycle change model is a time series analysis model established based on seasonal crop planting patterns, crop types and growth cycles; The weathering process dynamic monitoring model is a numerical model based on historical data analysis based on real-time rainfall, temperature and historical weathering rate; The landform evolution trend analysis model is a real-time analysis model constructed based on historical erosion, rainfall data and landform evolution trends.

6. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 5, characterized in that: In step 1, drones are used to obtain high-precision orthophotos and digital elevation models of the surface, identify landform features, and perform preliminary regional segmentation. The specific steps for image processing are as follows: Step 11: Perform radiometric correction and orthorectification on the acquired aerial image data; Step 12: Use the Mean-shift algorithm to segment the original image into several homogeneous area blocks based on color similarity and spatial proximity; Step 13: extract the quantitative features of the images of each segmented area and preliminarily establish an image feature library of different landform categories; Step 14: classify the preliminary segmented areas using a random forest classification algorithm, and preliminarily identify the types of landform areas based on the classification results; Step 15: Adaptively perform fine image processing on the preliminarily identified areas: construct a machine learning model based on the historical geological stability index and rock weathering rate, dynamically adjust the RGB brightness, color standard deviation, texture contrast and texture entropy thresholds, optimize the thresholds through the feedback results of regional characteristics, finely adjust the segmentation boundaries of the bedrock exposed areas, construct a time series analysis model of seasonal agricultural cycle changes, dynamically adjust the color standard deviation, regularity coefficient and boundary sharpness index, finely segment and identify farmland boundaries and crop types, construct a dynamic data value analysis model based on the regional historical weathering rate and real-time climate monitoring data, dynamically adjust the crack identification density and texture entropy index, finely adjust the contour boundaries of the weathered area, and construct a landform evolution trend analysis model in combination with digital elevation data. Adjust the slope, landform depth difference and valley width thresholds in real time to finely depict the contour details of the valley area.

7. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 3, characterized in that: In step one, a high-frequency surface radar is used to detect the thickness and spatial distribution characteristics of the loose structure of the shallow surface layer. The characteristic data obtained include the thickness of the loose layer, the intensity value of the reflection wave at the interface of the loose layer, and the distribution of the radar wave propagation velocity. Through on-site sampling and laboratory analysis, the distribution of rock and soil particles and the degree of loose compaction are clarified. The data obtained include the particle size distribution of rock and soil particles, grading coefficient, penetration resistance value, and permeability coefficient.

8. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 1, characterized in that: In step 2, the construction process of the medium-deep seismic geological condition model includes: Step 21, historical drilling data analysis: Integrate existing historical drilling data to quantitatively determine the target stratum burial depth, coal seam thickness, upper and lower surrounding rock types, and lithologic physical properties, including density, porosity, and acoustic wave velocity; Step 22, using the cross-well logging method to determine the T3 wave characteristics: using the cross-well logging method of acoustic logging, density logging, and resistivity logging, determine the acoustic wave velocity, density, and resistivity of the T3 wave reflection interface, analyze the T3 wave interface reflection wave intensity, wave continuity index, and its energy attenuation ratio of the seismic reflection wave of the underlying coal seam, form a shielding effect intensity map, and quantitatively analyze the target formation characteristics and the T3 wave shielding effect.

9. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 8, characterized in that: In step three, the method for performing well depth test zoning based on the data and features obtained in steps one and two includes: Step 31, importing the acquired data into the GIS system: importing the loose layer thickness, interface reflection wave intensity, radar wave propagation velocity, particle distribution, and penetration resistance data acquired in step 1, and the mid-deep coal seam burial depth, coal seam thickness, surrounding rock type and lithologic physical properties, and T3 wave shielding effect intensity data acquired in step 2 into the GIS system; Step 32, designing test well depth zones: In a GIS environment, spatial interpolation is used to perform spatial continuity analysis on the imported data to generate a continuous spatial distribution layer. The overlay analysis function of the GIS system is used to overlay the key parameter layers of the shallow layer and the medium-deep layer, and regional analysis is performed based on parameter similarity. Step 33, set the well depth gradient for each partition: combine the thickness of the shallow surface loose structure and the buried depth data of the medium and deep coal seams in the partition to determine each adapted well depth gradient, and output a well depth test partition map clearly marked under different surface and geological conditions on the GIS system platform.

10. The method for analyzing well depth in seismic exploration based on a complex surface environment according to claim 9, characterized in that: In step 33, the specific method for determining each adapted well depth gradient by combining the data of each morphological zone, the thickness of the shallow surface loose structure within the zone, and the buried depth of the medium-deep coal seam within the zone includes: Bedrock exposed area: The initial well depth gradient is 5m-10m, and the gradient range is dynamically adjusted to ±1m based on the feedback of the real-time updated rock weathering rate model; Farmland: The initial well depth gradient is set at 10m-15m, and the gradient range is dynamically adjusted to ±1.5m based on real-time agricultural cycle change model feedback; Weathering area: The initial well depth gradient is set at 15m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time weathering process dynamic monitoring model; Valley area: The initial well depth gradient is set at 10m-20m, and the gradient range is dynamically adjusted to ±2m based on the feedback from the real-time landform evolution trend analysis model.

Images