A new method for establishing high-precision 3D velocity field in horizontal well deployment area

Abstract

The application discloses a new high-precision three-dimensional velocity field establishment method for horizontal well deployment area, which comprises the following steps: firstly, determining two or more oil layer group level seismic standard layers near a target layer; secondly, establishing a reasonable constraint framework model in the vertical direction; thirdly, calibrating a velocity spectrum by using a fitting relationship; fourthly, establishing an initial three-dimensional velocity model; fifthly, performing filtering processing on the velocity curve of each well by using a geological mode constraint, so as to obtain a smoothed velocity curve capable of reflecting reservoir changes; sixthly, training the model by using a multi-layer feedforward neural network intelligent algorithm; seventhly, applying the training result of the model to the initial three-dimensional velocity model; and eighthly, obtaining a high-precision three-dimensional velocity field. The method fully applies various information such as well time-depth relationship, three-dimensional velocity spectrum, seismic interpretation results and the like, reduces the error between the depth domain result and the well by means of pseudo well velocity analysis, velocity curve filtering and artificial intelligence algorithm fusion, and thus the accuracy of velocity establishment for the horizontal well area is improved.

Description

Technical Field

[0001] This invention belongs to the field of reservoir geophysics technology, specifically relating to a new method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area. Background Technology

[0002] Horizontal wells have become one of the effective ways to increase oilfield reserves and production. The layout of horizontal wells requires the establishment of a precise three-dimensional velocity field. First, it meets the requirements of high-precision structural mapping and guides the horizontal wells to accurately enter the target. Second, it converts time-domain seismic data and reservoir prediction three-dimensional lithology data into the depth domain, providing a basis for reservoir modeling in horizontal well areas, horizontal well trajectory optimization design, and drilling tracking.

[0003] Current conventional velocity field modeling methods typically use the top and bottom surfaces of reservoir groups as research units, focusing primarily on single factors such as well time-depth relationships or superimposed velocity spectra, and relying mainly on conventional linear or inverse distance weighted interpolation methods. These methods fail to effectively reflect the spatial variations in reservoir velocity, impacting the accuracy of structural mapping of the target layer's top surface in horizontal wells. In cases of low well density or numerous deviated wells, especially in blocks with significant velocity variations, the accuracy of conventional velocity fields cannot meet the design requirements of horizontal wells. Furthermore, using conventional velocity fields for time-depth conversion results in significant errors between the depth domain and wellbore data, leading to inaccurate velocity establishment in the horizontal well area. Therefore, we need to propose a new method for establishing a high-precision three-dimensional velocity field in the horizontal well deployment area to address these issues and ensure it meets the design requirements of horizontal wells. Summary of the Invention

[0004] The purpose of this invention is to provide a new method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area. This method fully utilizes various information such as the time-depth relationship of the well, the three-dimensional velocity spectrum, and seismic interpretation results. Through pseudo-well velocity analysis, velocity curve filtering, and the fusion of artificial intelligence algorithms, it provides reliable technical support for horizontal well deployment and precise development measures adjustment and potential tapping, thereby solving the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area includes the following steps:

[0007] S1. Based on the time-depth relationship obtained from the fine calibration results of the synthetic seismic record, determine two or more oil layer group-level seismic standard layers near the target layer;

[0008] S2. Under the constraints of two or more oil layer groups, taking into account factors such as seismic waveform characteristics and reservoir characteristics, a reasonable vertical constraint framework model is established through the well-seismic matching unit analysis method.

[0009] S3. Using the multi-well velocity consistency analysis method, cross-analysis and trend fitting are performed on the velocities obtained through the time-depth relationship of well points and the velocities obtained through the velocity spectrum. The velocity spectrum is calibrated using the fitting relationship. Virtual wells are established in sparse well areas through layer control and equal time-depth relationship with the same phase. Virtual wells are established in deviated well areas through layer control and perpendicular time-depth relationship with the deviated well. The velocity spectrum is calibrated using the virtual wells.

[0010] S4. Sample the corrected velocity spectrum onto the constraint frame model established in S2 to obtain the initial three-dimensional velocity model;

[0011] S5. Transform the acoustic curves at the well points to obtain the velocity curves for each well.

[0012] S6. Perform geological model-constrained filtering on the velocity curves of each well to eliminate random interference and non-reservoir influence factors, and obtain smooth velocity curves that can reflect reservoir changes.

[0013] S7. Using the initial three-dimensional velocity model obtained in S4 as input, perform machine learning training with the filtered velocity curve in S6. Through attribute analysis, select the single or multiple attributes that have the best correlation with the target velocity curve. Use a multi-layer feedforward neural network intelligent algorithm to train the model. Apply the model training results to the initial three-dimensional velocity model to obtain a high-precision three-dimensional velocity field. If the high-precision three-dimensional velocity field meets the requirements, output the high-precision three-dimensional velocity field; if the high-precision three-dimensional velocity field does not meet the requirements, repeat S3 to S7 to finally obtain the high-precision three-dimensional velocity field.

[0014] Preferably, in step S1, the fine calibration result of the synthetic seismic record is an optimized result obtained by comparing and adjusting it with the actual seismic record. The seismic record is generated by simulating the phenomenon of seismic wave propagation and reflection underground. Fine calibration refers to adjusting the amplitude, arrival time, and waveform parameters of the synthetic seismic record to make these parameters consistent with the actual seismic record. Seismic interpreters perform fine calibration on the synthetic seismic record according to the underground geological conditions to obtain more accurate seismic reflections and characteristics.

[0015] Preferably, when determining the two or more oil-bearing group-level seismic standard layers near the target layer, the synthetic seismic records near the target layer are first obtained, and then the fine calibration results of the synthetic seismic records are obtained by comparing and adjusting them with the actual seismic records. The fine calibration results are then combined with the development seismic horizon interpretation to establish the time-depth relationship of the seismic records. Based on the time-depth relationship and the development seismic horizon interpretation, the two or more oil-bearing group-level seismic standard layers near the target layer are determined.

[0016] Preferably, in step S2, the oil layer group is an accumulation of oil layers with certain continuity and geological characteristics in an oil and gas field, and each oil layer group has reservoir characteristics with different thickness, porosity, permeability and oil saturation.

[0017] Preferably, before establishing the constraint framework model, wellbore data of the target area is collected, including well logging curves, geological stratification, and lithological interpretation data. The seismic data is then preprocessed, including noise removal and frequency division. The processed seismic data is then divided according to the target layer to form well-seismic matching units. Vertically, the upper and lower boundaries of the target layer are determined based on wellbore interpretation and seismic time-depth conversion, and the seismic attributes within the target layer are extracted and analyzed. Each well-seismic matching unit is analyzed, taking into account seismic waveform characteristics and reservoir development characteristics. Based on the degree of matching between seismic records and wellbore data, the relationship between reservoir attributes and seismic response is determined. Based on the analysis results of the well-seismic matching units, a vertical constraint framework model is established.

[0018] Preferably, after the constraint framework model is established, it is evaluated and adjusted by verifying and comparing it with unused well and seismic data. Based on the verification results, the model parameters are modified and optimized to obtain a more accurate constraint framework model.

[0019] Preferably, in step S3, the multi-well velocity consistency analysis is performed by comparing velocity data measured at different well points, analyzing the time-depth relationship between well points, and then calibrating and standardizing the velocity data to obtain a consistent velocity model.

[0020] Preferably, the velocity spectrum is a way of expressing energy information and wave velocity in seismic data. Before performing cross-hatching analysis and trend fitting, velocity data and depth information of multiple well points are obtained based on the time-depth relationship. Then, by associating the depth of each well point with the velocity at the corresponding depth position, a continuous velocity-depth curve is established. By associating the velocity in the velocity spectrum data with the well point velocity at the corresponding depth position, velocity values ​​at different depths in the velocity spectrum are obtained. The velocity obtained from the cross-hatching analysis is trend-fitted using linear regression and polynomial fitting methods to find the fitting relationship between the velocity at the well point and the velocity in the velocity spectrum. Through the fitting relationship, the velocity in the velocity spectrum is mapped to the corresponding velocity value at the well point, thereby obtaining the calibrated velocity spectrum.

[0021] Preferably, in step S4, when the initial three-dimensional velocity model is established, the velocity values ​​are sampled onto the spatial grid of the constraint frame model based on the velocity information in the velocity spectrum to obtain an initial velocity model. Using seismic phase information and seismic wave reflection simulation methods, the initial velocity model is matched with seismic data through velocity model inversion technology. Then, the obtained initial velocity model is verified and corrected using seismic data, well logging data, or other geological information to obtain the final initial three-dimensional velocity model.

[0022] Preferably, in step S5, the acoustic wave curve at the well point is the amplitude-time relationship curve of the acoustic wave signal in the well obtained by seismic exploration, logging or acoustic wave propagation experiment. When transforming the acoustic wave curve at the well point, a Fourier transform is applied to the acoustic wave curve at the well point, and the amplitude spectrum part of the transform result is taken. According to the properties of the Fourier transform, the spectrum information is reversed back to the time domain, that is, the spectrum is converted into a velocity curve through inverse Fourier transform. The amplitude spectrum part of the transform result represents the energy distribution of each frequency in the acoustic wave signal.

[0023] The present invention proposes a novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area, which has the following advantages compared with existing technologies:

[0024] This invention first identifies two or more seismic standard layers of oil-bearing formations near the target layer, then establishes a reasonable vertical constraint framework model. It performs cross-sectional analysis and trend fitting on velocities obtained from well point time-depth relationships and velocity spectra, calibrates the velocity spectrum using the fitted relationship, and establishes an initial three-dimensional velocity model. Next, it transforms the sonic curves at the well points to obtain velocity curves for each well. The velocity curves of each well are then filtered using geological model constraints to eliminate random interference and non-reservoir influences, resulting in smooth velocity curves that reflect reservoir changes. A multi-layer feedforward neural network intelligent algorithm is used to train the model, and the training results are applied to the initial three-dimensional velocity model to obtain a high-precision three-dimensional velocity field. By fully utilizing various information such as well time-depth relationships, three-dimensional velocity spectra, and seismic interpretation results, and through pseudo-well velocity analysis, velocity curve filtering, and artificial intelligence algorithm fusion, the error between depth domain results and wellbore data is reduced, thereby improving the accuracy of velocity establishment in horizontal well areas. This provides reliable technical support for horizontal well deployment and precise development measures adjustment and potential tapping. Attached Figure Description

[0025] Figure 1 This is a flowchart of the present invention;

[0026] Figure 2 This is a schematic diagram of the constraint frame model of the present invention;

[0027] Figure 3 This is a schematic diagram of the velocity field profile established in this invention;

[0028] Figure 4 This is a schematic diagram of the time-depth conversion of the reservoir prediction profile after the present invention;

[0029] Figure 5 This is a schematic diagram of the framework model of existing technology;

[0030] Figure 6 A schematic diagram of the velocity field profile established using existing technology;

[0031] Figure 7 This is a schematic diagram of time-depth conversion for reservoir prediction profiles using existing technologies. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] This invention provides, for example Figure 1-4 The method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area, as shown, includes the following steps:

[0034] S1. Based on the time-depth relationship obtained from the fine calibration results of the synthetic seismic record, determine two or more oil layer group-level seismic standard layers near the target layer;

[0035] The fine calibration results of synthetic seismic records are the optimized results obtained by comparing and adjusting them with actual seismic records. Seismic records are generated by simulating the propagation and reflection of seismic waves underground. Fine calibration refers to adjusting the amplitude, arrival time, and waveform parameters of the synthetic seismic records to ensure that these parameters match those of actual seismic records. Seismic interpreters perform fine calibration of synthetic seismic records based on underground geological conditions to obtain more accurate seismic reflections and characteristics.

[0036] When determining two or more oil-bearing group-level seismic standard layers near the target layer, the synthetic seismic records near the target layer are first obtained. Then, by comparing and adjusting with the actual seismic records, the fine calibration results of the synthetic seismic records are obtained. The fine calibration results are then combined with the interpretation of development seismic horizons to establish the time-depth relationship of the seismic records. Based on the time-depth relationship and the interpretation of development seismic horizons, two or more oil-bearing group-level seismic standard layers near the target layer are determined.

[0037] S2. Under the constraints of two or more oil layer groups, taking into account factors such as seismic waveform characteristics and reservoir characteristics, a reasonable vertical constraint framework model is established through the well-seismic matching unit analysis method.

[0038] Oil-bearing formations are oil-bearing layers in oil and gas fields that have certain continuity and geological characteristics. Each oil-bearing formation has different reservoir characteristics such as thickness, porosity, permeability, and oil saturation.

[0039] Before establishing the constraint framework model, wellbore data for the target area is collected, including well logging curves, geological stratification, and lithological interpretation data. Seismic data is then preprocessed, including noise removal and frequency division. The processed seismic data is then divided according to the target layer to form well-seismic matching units. Vertically, the upper and lower boundaries of the target layer are determined based on wellbore interpretation and seismic time-depth conversion. Seismic attributes within the target layer are extracted and analyzed. Each well-seismic matching unit is analyzed, comprehensively considering seismic waveform characteristics and reservoir development characteristics. Based on the degree of matching between seismic records and wellbore data, the relationship between reservoir attributes and seismic response is determined. Based on the analysis results of the well-seismic matching units, a vertical constraint framework model is established.

[0040] After the constraint framework model is established, it is evaluated and adjusted by verifying and comparing it with unused well and seismic data. Based on the verification results, the model parameters are modified and optimized to obtain a more accurate constraint framework model.

[0041] S3. Using the multi-well velocity consistency analysis method, cross-analysis and trend fitting are performed on the velocities obtained through the time-depth relationship of well points and the velocities obtained through the velocity spectrum. The velocity spectrum is calibrated using the fitting relationship. Virtual wells are established in sparse well areas through layer control and equal time-depth relationship with the same phase. Virtual wells are established in deviated well areas through layer control and perpendicular time-depth relationship with the deviated well. The velocity spectrum is calibrated using the virtual wells.

[0042] Multi-well velocity consistency analysis is a method that compares velocity data measured at different well points, analyzes the time-depth relationship between well points, and then calibrates and standardizes the velocity data to obtain a consistent velocity model.

[0043] Velocity spectrum is a way of expressing energy information and wave velocity in seismic data. Before performing cross-hatching analysis and trend fitting, velocity data and depth information of multiple well points are obtained based on time-depth relationship. Then, by associating the depth of each well point with the velocity at the corresponding depth position, a continuous velocity-depth curve is established. By associating the velocity in the velocity spectrum data with the well point velocity at the corresponding depth position, velocity values ​​at different depths in the velocity spectrum are obtained. Linear regression and polynomial fitting methods are used to perform trend fitting on the velocity obtained from cross-hatching analysis to find the fitting relationship between the velocity at the well point and the velocity in the velocity spectrum. Through the fitting relationship, the velocity in the velocity spectrum is mapped to the corresponding velocity value at the well point, thereby obtaining the calibrated velocity spectrum.

[0044] S4. Sample the corrected velocity spectrum onto the constraint frame model established in S2 to obtain the initial three-dimensional velocity model;

[0045] When establishing the initial three-dimensional velocity model, the velocity values ​​are sampled onto the spatial grid of the constrained frame model based on the velocity information in the velocity spectrum to obtain an initial velocity model. Using seismic phase information and seismic wave reflection simulation methods, the initial velocity model is matched with seismic data through velocity model inversion technology. Then, the obtained initial velocity model is verified and corrected using seismic data, well logging data, or other geological information to obtain the final initial three-dimensional velocity model.

[0046] S5. Transform the acoustic curves at the well points to obtain the velocity curves for each well.

[0047] The acoustic curve at a well point is the amplitude-time relationship curve of the acoustic signal in the well obtained through seismic exploration, logging, or acoustic propagation experiments. When transforming the acoustic curve at a well point, a Fourier transform is applied to the acoustic curve at the well point, and the amplitude spectrum part of the transform result is taken. According to the properties of the Fourier transform, the spectrum information is reversed back to the time domain, that is, the spectrum is converted into a velocity curve through inverse Fourier transform. The amplitude spectrum part of the transform result represents the energy distribution of each frequency in the acoustic signal.

[0048] S6. For each well, the velocity curve is filtered using a geological model constraint to eliminate random interference and non-reservoir influences, resulting in a smooth velocity curve that reflects reservoir changes. Based on the characteristics of the velocity curve and user requirements, a high-pass filter, low-pass filter, or scale-space filter is selected for filtering. A high-pass filter removes low-frequency components while retaining high-frequency variations, helping to eliminate low-frequency random noise or non-reservoir influences and highlighting the characteristics of rapid changes in the reservoir. A low-pass filter removes high-frequency components while retaining low-frequency variations, smoothing the velocity curve and eliminating high-frequency random interference, making reservoir changes more clearly visible. A scale-space filter performs multi-scale filtering, decomposing the velocity curve into components of different scales. By gradually reducing the scale, different frequency components of reservoir changes and noise can be separated.

[0049] S7. Using the initial 3D velocity model obtained in S4 as input, perform machine learning training with the filtered velocity curve in S6. Through attribute analysis, select the single or multiple attributes with the best correlation to the target velocity curve. Use a multilayer feedforward neural network intelligent algorithm to train the model. Apply the model training results to the initial 3D velocity model to obtain a high-precision 3D velocity field. If the high-precision 3D velocity field meets the requirements, output the high-precision 3D velocity field. The high-precision 3D velocity field output includes seismic data volume time-depth conversion and layer time-depth conversion. If the high-precision 3D velocity field does not meet the requirements, repeat S3 to S7 to finally obtain the high-precision 3D velocity field.

[0050] The following uses the Deng III section of the Shuang 68 horizontal well deployment area in Daqing Oilfield as an example to illustrate the implementation process of the method of the present invention.

[0051] 7.1 Research Background

[0052] With the completion of development wells in the Shuang68 block, a more accurate understanding of the reservoir, oil-water distribution, and structure has been achieved. Given the characteristics of the localized development of the main oil-bearing layer, it is necessary to deploy horizontal wells in structurally favorable locations to increase production capacity and net present value, thereby meeting the economic evaluation requirements of the development plan.

[0053] Currently, the top and bottom surfaces of oil-bearing formations are usually used as research units. Velocity fields are established using single data such as velocity spectrum or well-seismic calibration time-depth relationship, and methods such as linear or inverse distance weighting. However, this method cannot meet the design accuracy requirements of horizontal wells, especially in blocks with low well density or many deviated wells, and particularly in blocks with large velocity variations. Therefore, taking the Deng III oil-bearing formation in the Shuang 68 horizontal well deployment area of ​​Daqing Oilfield as an example, this study conducted a research on a high-precision three-dimensional velocity field establishment method for horizontal well deployment areas.

[0054] 7.2 Implementation Content

[0055] Traditional methods for establishing velocity fields in horizontal well areas are based on the control of the top DII and bottom DIII marker layers. Figure 5 As shown in the figure, the velocity obtained from the calibration of the time-depth relationship of the known well synthetic seismic records is used as input data, and the three-dimensional velocity field is obtained by interpolation through the inverse distance weighting algorithm.

[0056] This invention first establishes a vertically refined and reasonable framework model under the control of the top DII and bottom DIII marker layers, comprehensively considering factors such as seismic waveform characteristics and reservoir characteristics, through well-seismic matching element analysis method. Figure 2 Secondly, the velocity obtained through the well point time-depth relationship and the velocity obtained through the velocity spectrum are subjected to intersection analysis and trend fitting. The velocity spectrum is corrected to the well point velocity range using this fitting relationship. In deviated well areas or areas with low well network density, virtual wells are established using layer control + deviated well vertical time-depth relationship and layer control + same phase equal time and equal depth relationship to correct the velocity spectrum and sample it into the above framework model as the initial input velocity model. Then, the velocity curve obtained by transforming the well point sonic curve is subjected to geological model constraint filtering to obtain a smooth velocity curve as the well point velocity input. Finally, intelligent algorithms such as multilayer feedforward neural networks are used to establish the nonlinear relationship between the initial input velocity model at the well point and the well point velocity curve, and this relationship is applied to the entire initial velocity model to obtain a high-precision three-dimensional velocity field.

[0057] 7.3 Discussion of Results

[0058] Compared with traditional methods for establishing three-dimensional velocity fields in horizontal well areas, the three-dimensional velocity field established using this invention can accurately reflect the spatial variations in velocity within the horizontal well area (e.g., ...). Figure 3 As shown), this solves the problem of local velocity distortion between wells ( Figure 6 After applying this method to convert the reservoir prediction results in horizontal well areas to time and depth, the depth domain results are consistent with the geological information on the wellbore. This effectively solves the problem of inaccurate velocity field establishment caused by the large number of deviated wells and large well spacing in horizontal well areas. It provides timely guidance for the optimization and adjustment of horizontal well trajectories and seismic tracking in the Shuang68 block. The depth error at the entry point is less than 1‰, providing strong technical support for the 88.7% sandstone drilling encounter rate in horizontal wells.

[0059] In summary, as Figure 1As shown, wellbore depth-time conversion is performed based on geological strata, followed by well-seismic matching to establish an accurate time-depth relationship. Fine interpretation of standard layers (two or more oil-bearing formations) is conducted based on the detailed calibration results of synthetic seismic records. A vertically reasonable constraint framework model is established using well-seismic matching unit analysis. Cross-sectional analysis and trend fitting are performed on the velocities obtained from the wellpoint time-depth relationship and the velocity spectrum. The velocity spectrum is calibrated using the fitted relationship, and an initial three-dimensional velocity model is established. The sonic curves at the wellpoints are then transformed to obtain the velocity curves for each well. Geological model-constrained filtering is applied to the velocity curves of each well to eliminate random interference and non-reservoir influences, resulting in smooth velocity curves that reflect reservoir changes. A multilayer feedforward neural network is then employed for intelligent... The algorithm trains the model, optimizes parameters through deep learning, and applies the training results to the initial 3D velocity model to obtain a high-precision 3D velocity field. If the high-precision 3D velocity field meets the requirements, it is output, including seismic data volume time-depth conversion and layer time-depth conversion. If the high-precision 3D velocity field does not meet the requirements, steps S3 to S7 are repeated until the final high-precision 3D velocity field is obtained. By fully utilizing various information such as well time-depth relationship, 3D velocity spectrum, and seismic interpretation results, and through pseudo-well velocity analysis, velocity curve filtering, and artificial intelligence algorithm fusion, the error between depth domain results and wellbore results is reduced, thereby improving the accuracy of velocity establishment in horizontal well areas and providing reliable technical support for horizontal well deployment, precise development measures adjustment and potential tapping.

[0060] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area, characterized in that: Includes the following steps: S1. Based on the time-depth relationship obtained from the fine calibration results of the synthetic seismic record, determine the two oil-bearing group-level seismic standard layers near the target layer; S2. Under the constraint of two oil layer groups, taking into account factors such as seismic waveform characteristics and reservoir characteristics, a vertical constraint frame model is established through the well-seismic matching unit analysis method. S3. Using the multi-well velocity consistency analysis method, cross-analysis and trend fitting are performed on the velocities obtained through the time-depth relationship of well points and the velocities obtained through the velocity spectrum. The velocity spectrum is calibrated using the fitting relationship. Virtual wells are established in sparse well areas through layer control and equal time-depth relationship with the same phase. Virtual wells are established in deviated well areas through layer control and perpendicular time-depth relationship with the deviated well. The velocity spectrum is calibrated using the virtual wells. Multi-well velocity consistency analysis is a method that compares velocity data measured at different well points, analyzes the time-depth relationship between well points, and then calibrates and standardizes the velocity data to obtain a consistent velocity model. S4. Sample the corrected velocity spectrum onto the constraint frame model established in S2 to obtain the initial three-dimensional velocity model; When establishing the initial three-dimensional velocity model, the velocity values ​​are sampled onto the spatial grid of the constraint frame model based on the velocity information in the velocity spectrum to obtain an initial velocity model. Using seismic phase information and seismic wave reflection simulation methods, the initial velocity model is matched with seismic data through velocity model inversion technology. Then, the obtained initial velocity model is verified and corrected using seismic data, well logging data, or other geological information to obtain the final initial three-dimensional velocity model. S5. Transform the acoustic curves at the well points to obtain the velocity curves for each well. S6. Perform geological model-constrained filtering on the velocity curves of each well to eliminate random interference and non-reservoir influence factors, and obtain smooth velocity curves that can reflect reservoir changes. S7. Using the initial 3D velocity model obtained in S4 as input, perform machine learning training with the filtered velocity curve in S6. Select one or more attributes with the best correlation to the target velocity curve through attribute analysis, and train the model using a multi-layer feedforward neural network intelligent algorithm. Apply the model training results to the initial 3D velocity model to obtain a high-precision 3D velocity field. If the high-precision 3D velocity field meets the requirements, output the high-precision 3D velocity field; if the high-precision 3D velocity field does not meet the requirements, repeat S3 to S7 to finally obtain the high-precision 3D velocity field.

2. The novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 1, characterized in that: In step S1, the fine calibration result of the synthetic seismic record is an optimized result obtained by comparing and adjusting it with the actual seismic record. The seismic record is generated by simulating the propagation and reflection of seismic waves underground. Fine calibration refers to adjusting the amplitude, arrival time, and waveform parameters of the synthetic seismic record to make these parameters consistent with the actual seismic record. Seismic interpreters perform fine calibration on the synthetic seismic record based on the underground geological conditions to obtain more accurate seismic reflections and characteristics.

3. The new method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 2, characterized in that: When determining the two or more oil-bearing group-level seismic standard layers near the target layer, the synthetic seismic records near the target layer are first obtained. Then, by comparing and adjusting with the actual seismic records, the fine calibration results of the synthetic seismic records are obtained. The fine calibration results are then combined with the development seismic horizon interpretation to establish the time-depth relationship of the seismic records. Based on the time-depth relationship and the development seismic horizon interpretation, the two or more oil-bearing group-level seismic standard layers near the target layer are determined.

4. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 3, characterized in that: In step S2, the oil layer group is an accumulation of oil layers with certain continuity and geological characteristics in an oil and gas field. Each oil layer group has reservoir characteristics with different thickness, porosity, permeability and oil saturation.

5. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 4, characterized in that: Before establishing the constraint framework model, wellbore data of the target area is collected, including well logging curves, geological stratification, and lithological interpretation data. The seismic data is then preprocessed, including noise removal and frequency division. The processed seismic data is then divided according to the target layer to form well-seismic matching units. Vertically, the upper and lower boundaries of the target layer are determined based on wellbore interpretation and seismic time-depth conversion, and the seismic attributes within the target layer are extracted and analyzed. Each well-seismic matching unit is analyzed, taking into account seismic waveform characteristics and reservoir development characteristics. Based on the degree of matching between seismic records and wellbore data, the relationship between reservoir attributes and seismic response is determined. Based on the analysis results of the well-seismic matching units, a vertical constraint framework model is established.

6. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 5, characterized in that: After the constraint framework model is established, it is evaluated and adjusted by verifying and comparing it with unused well and seismic data. Based on the verification results, the model parameters are modified and optimized to obtain a more accurate constraint framework model.

7. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 6, characterized in that: The velocity spectrum is a way of expressing energy information and wave velocity in seismic data. Before performing cross-hatching analysis and trend fitting, velocity data and depth information of multiple well points are obtained based on the time-depth relationship. Then, by associating the depth of each well point with the velocity at the corresponding depth position, a continuous velocity-depth curve is established. By associating the velocity in the velocity spectrum data with the well point velocity at the corresponding depth position, velocity values ​​at different depths in the velocity spectrum are obtained. Linear regression and polynomial fitting methods are used to perform trend fitting on the velocity obtained from the cross-hatching analysis to find the fitting relationship between the velocity at the well point and the velocity in the velocity spectrum. Through the fitting relationship, the velocity in the velocity spectrum is mapped to the corresponding velocity value at the well point, thereby obtaining the calibrated velocity spectrum.

8. A novel method for establishing a high-precision three-dimensional velocity field in a horizontal well deployment area according to claim 7, characterized in that: In step S5, the acoustic curve at the well point is the amplitude-time relationship curve of the acoustic signal in the well obtained through seismic exploration, logging, or acoustic propagation experiments. When transforming the acoustic curve at the well point, a Fourier transform is applied to the acoustic curve at the well point, and the amplitude spectrum part of the transform result is taken. According to the properties of the Fourier transform, the spectrum information is reversed back to the time domain, that is, the spectrum is converted into a velocity curve through inverse Fourier transform. The amplitude spectrum part of the transform result represents the energy distribution of each frequency in the acoustic signal.

Images