Geosteering method for low-permeability sandstone reservoir gas horizontal well

Abstract

The invention discloses a geosteering method for a low-permeability sandstone reservoir gas horizontal well, and belongs to the technical field of oil and gas development. The invention discloses a geosteering method for a low-permeability sandstone reservoir gas horizontal well. The geosteering method comprises the following steps: designing an initial trajectory based on seismic inversion and adjacent well data; gR, RT, DT and CNL data while drilling are collected in real time; space-time calibration is implemented through a deep matching algorithm, and quality control is performed by using multi-parameter cross validation; calculating a reservoir matching degree index SMI based on a geological rule and a machine learning model, and generating an adjustment suggestion when the SMI is lower than a threshold value; verifying feasibility by combining drilling engineering constraints, and dynamically optimizing the trajectory by taking a high-quality reservoir contact rate greater than or equal to 90% as a target; and after drilling is completed, the reservoir drilling encounter rate is quantitatively analyzed. According to the method, multi-source data accurate fusion and intelligent trajectory adjustment are achieved, the problem of low-permeability sandstone reservoir trajectory control is solved, and the drilling efficiency and the recovery efficiency are improved.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas development technology, specifically relating to a geological steering method for horizontal wells in low-permeability sandstone reservoirs. Background Technology

[0002] Low-permeability sandstone reservoirs are major enrichment areas for unconventional natural gas, and their efficient development is crucial for alleviating the energy crisis. Horizontal well technology, due to its ability to significantly increase the reservoir contact area, has become the core means of developing this type of reservoir. The accuracy of geological steering directly determines the drilling rate of high-quality reservoirs in horizontal wells, which is a key factor affecting single-well production and development economics.

[0003] Current traditional geosteering methods for low-permeability sandstone reservoirs mainly rely on macroscopic seismic prediction and experience-based identification from adjacent wells, which suffer from three major technical bottlenecks: low data fusion efficiency, with inconsistent depth benchmarks between seismic inversion data (macroscale) and logging-while-drilling data (microscale), leading to ambiguous reservoir boundary identification due to spatiotemporal matching errors; insufficient real-time performance, with long delays in cuttings identification and logging signal transmission, making it difficult to respond promptly to lateral changes in the reservoir, easily missing high-quality reservoirs or drilling into mudstone interlayers; and crude trajectory optimization, relying on manual experience to set thresholds for parameters such as resistivity and gamma, without combining dynamic adjustments to formation water salinity, and without quantifying drilling engineering constraints (such as build-up rate limits and drilling fluid safety windows), often resulting in excessive trajectory adjustments or substandard reservoir contact rates.

[0004] Therefore, there is an urgent need for a geological steering method that integrates multi-source data spatiotemporal precision calibration, machine learning intelligent identification, and dynamic trajectory optimization under engineering constraints to solve the technical challenges of low drilling rate and high drilling risk in horizontal wells of low-permeability sandstone reservoirs. Summary of the Invention

[0005] This invention addresses the aforementioned pain points by providing a geological steering method for horizontal wells in low-permeability sandstone reservoirs that integrates high-frequency logging-constrained seismic inversion, real-time spatiotemporal calibration, machine learning reservoir matching degree calculation, and dynamic trajectory optimization under engineering constraints. Through precise fusion of multi-source data and intelligent decision-making, it maximizes the contact rate of high-quality reservoirs and minimizes drilling risks.

[0006] The present invention is as follows: A geological steering method for horizontal wells in low-permeability sandstone reservoirs is characterized by comprising the following steps: S1. Based on the seismic inversion results constrained by high-frequency logging data and reservoir parameter data of adjacent wells, design the initial trajectory of the horizontal well; S2. Real-time acquisition of logging-while-drilling data, including gamma ray (GR), resistivity (RT), sonic transit time (DT), and neutron porosity (CNL) signals; S3. Perform spatiotemporal calibration and quality control on the seismic inversion data, adjacent well reservoir parameter data, and real-time logging data from step S1 and step S2: The spatiotemporal calibration refers to using a depth matching algorithm to unify the depth benchmark between the macroscopic scale of seismic logging and the microscopic scale of logging; The quality control refers to implementing multi-parameter cross-validation, and triggering data retesting when the gamma ray GR value fluctuation exceeds a preset threshold, or when the correlation between resistivity RT and sonic transit time DT is lower than the set standard. S4. Based on the reliable data processed in step S3, and combined with the preset geological rules for identifying low-permeability sandstone reservoirs, a machine learning algorithm is applied to calculate the reservoir matching index (SMI). When the SMI is lower than a preset threshold, a trajectory adjustment suggestion is generated. S5. Verify the feasibility of the adjustment suggestions generated in step S4 in conjunction with drilling engineering constraints, and perform dynamic optimization of the horizontal segment trajectory with the goal of maximizing the contact rate of high-quality reservoirs. S6. Based on the logging data obtained after drilling is completed, the actual reservoir drilling rate is quantitatively analyzed to provide a basis for optimizing subsequent construction parameters.

[0007] Preferably, the adjacent well reservoir parameter data includes permeability, gas saturation, and formation water salinity data; the resistivity threshold used to identify the reservoir in the geological rules is dynamically adjusted based on the formation water salinity obtained or predicted in real time.

[0008] Preferably, the spatiotemporal calibration and quality control process must meet the real-time requirements: the cuttings depth identification delay is <5 meters; the logging signal transmission delay from downhole to the surface processing system is <30 seconds.

[0009] Preferably, in the quality control process, the clay content is corrected using neutron porosity (CNL) logging data; when the corrected density (DEN) and neutron porosity (CNL) data meet the preset gas reservoir response characteristic conditions, the well section is marked as a potential gas reservoir; the potential gas reservoir is a high-quality reservoir.

[0010] Preferably, the input feature parameters of the machine learning model for the reservoir matching index SMI include: gamma GR curve variation features, resistivity RT gradient features, acoustic transit time DT variation features, and neutron-density intersection features.

[0011] Preferably, when the reservoir matching index (SMI) is lower than a preset threshold, the following operations are automatically performed: merging seismic inversion data and real-time logging data to generate a spatial probability distribution cloud map of sand bodies around the wellbore; the cloud map is generated by merging seismic wave attribute volumes and logging inversion results, and the predicted reservoir boundaries and their uncertainty range are marked in a visual manner.

[0012] Preferably, the spatial probability distribution cloud map of the sand body is input into the machine learning model in step S4 to update the calculation results of the reservoir matching index SMI; at the same time, the real-time drilling trajectory and the predicted trajectory after error calibration are superimposed and displayed on the design trajectory map.

[0013] Preferably, the drilling engineering constraints specifically include: the maximum single-meter build-up rate for trajectory adjustment is limited to 5° / meter; the horizontal drilling target is a high-quality reservoir contact rate ≥90%; and the drilling fluid equivalent circulation density must be less than the specified safety factor of the formation fracture pressure gradient.

[0014] Preferably, the trajectory dynamic optimization includes: generating at least 3 candidate trajectory adjustment schemes based on the latest data every 10-20 meters of drilling unit length; evaluating each candidate scheme according to the drilling engineering constraints, and quantifying the engineering risk level through a weighted algorithm, with weighting factors including: trajectory adjustment range (weight 0.4), formation uncertainty coefficient (weight 0.3), and drilling fluid safety window margin (weight 0.3).

[0015] Preferably, the trajectory adjustment operation specifically includes: an upward obstacle avoidance operation: when a mudstone interlayer is identified, an upward adjustment is performed at a build-up rate of ≤3° / m; a downward tracking operation: when a high-permeability sand body region is identified, a downward adjustment is performed at a build-up rate of ≤5° / m; wherein: the criteria for determining the high-permeability sand body region are: sonic transit time > 280 μs / m and bulk density < 2.4 g / cm³. 3 The criteria for determining the mudstone interlayer are: gamma GR value > 100 API and the intersection value of neutron porosity CNL and density DEN is within the preset mudstone response range.

[0016] Compared with the prior art, the advantages of the present invention are as follows: (1) In this invention, a seismic inversion model is constrained by high-frequency logging data, combined with reservoir parameters from adjacent wells: permeability, gas saturation, and formation water salinity; a multi-scale data fusion system is constructed, and a depth matching algorithm is used to unify the depth benchmarks of macroscopic seismic and micro-observation wells, achieving real-time control of cuttings depth identification delay < 5 meters and logging signal transmission delay < 30 seconds; when the gamma GR value fluctuation exceeds the threshold or the correlation between resistivity and acoustic transit time is insufficient, a multi-parameter cross-validation and retesting mechanism is triggered, and the clay content is simultaneously corrected through neutron porosity, accurately marking the density DEN < 2.4 g / cm³. 3 Furthermore, the potential gas layer has a CNL greater than 15%. This technology overcomes the problems of inconsistent macro-micro data benchmarks and ambiguous reservoir boundary identification in traditional methods, improving the accuracy of reservoir parameter identification by more than 25% and providing real-time and reliable data support for dynamic trajectory adjustment.

[0017] (2) In this invention, a reservoir matching index (SMI) is innovatively introduced. A machine learning model is constructed based on the GR curve variation characteristics, RT gradient, DT deviation, and neutron-density intersection characteristics to quantify reservoir quality into a continuous value of 0-1. When the SMI is lower than a preset threshold, the seismic wave attribute volume and well logging inversion results are automatically fused to generate a spatial probability distribution cloud map of sand bodies around the wellbore, visually marking the uncertainty range of the reservoir boundary, and updating the SMI calculation results in real time. This mechanism replaces the traditional rough mode of manually setting thresholds based on experience, and realizes dynamic evaluation of reservoir quality through data-driven approaches, improving the accuracy of trajectory adjustment suggestions by 30%. In low-permeability sandstone reservoirs, this model can accurately identify sonic transit time > 280 μs / m and bulk density < 2.4 g / cm³. 3 In high-permeability sandstone areas, the guide trajectory tracks high-quality reservoirs with a build-up rate of ≤5° / meter, avoiding drilling into mudstone interlayers with gamma GR > 100 API.

[0018] (3) In this invention, engineering constraints such as the maximum build-up rate limit of trajectory adjustment ≤5° / m, the target of high-quality reservoir contact rate ≥90%, and the drilling fluid safety density window are embedded into the dynamic optimization process. At least 3 candidate schemes are generated every 10 to 20 meters of drilling. The engineering risk is quantitatively assessed by weighting the trajectory adjustment amplitude weight 0.4, the formation uncertainty weight 0.3, and the drilling fluid safety margin weight 0.3. This mechanism solves the problem of unquantified engineering constraints in trajectory adjustment in traditional directional drilling. In implementation, it can ensure that the build-up rate is ≤5° / m when cutting down to track high-permeability sand bodies, and at the same time, the drilling fluid equivalent circulating density is less than the formation fracture pressure gradient safety threshold. Finally, the high-quality reservoir contact rate in the horizontal section exceeds 90%, which is 15% to 20% higher than the traditional method. The smoothness of the wellbore trajectory meets the engineering construction requirements, which greatly reduces the risks of stuck pipe and lost circulation, and takes into account both development efficiency and drilling safety. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a geological steering method for horizontal wells in low-permeability sandstone reservoirs. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be explained and described below. However, the following embodiments are only preferred embodiments of the present invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments in the implementation methods without creative effort are all within the protection scope of the present invention.

[0021] Example 1: I. Basic Information and Scenarios 1. Scenario: Select a low-permeability sandstone gas reservoir block in a basin. The target layer is the He8 sandstone section, with a burial depth of 2800-2900 meters. The average permeability of the reservoir is 0.5 mD, the gas saturation is 65%, and there are mudstone interlayers and high-permeability sandstone lenses.

[0022] 2. Basic Information: 2-1. Seismic Inversion Data: Based on gamma GR and resistivity RT logging data constraints with a 1-meter sampling interval, a model-driven inversion algorithm combined with a sparse pulse constraint algorithm was used to invert the sandstone thickness and wave impedance properties. Specific steps included: An initial model was constructed using the impedance trend of adjacent wells, and a sparse reflection coefficient sequence was extracted by L1 norm deconvolution. The number of reflection coefficients is iteratively adjusted to minimize the residual between the synthetic seismic record and the original data, ultimately generating the wave impedance volume; The wave impedance threshold is set to 12,000 to 15,000 kg / m 3 ·m / s, corresponding to sandstone reservoirs, core analysis shows that the gas saturation in this threshold range is ≥60%.

[0023] 2-2. Reservoir parameters of adjacent wells: Adjacent well A has a permeability of 0.8 mD, a gas saturation of 70%, and a formation water salinity of 15,000 mg / L; Adjacent well B has a permeability of 0.3 mD, a gas saturation of 55%, and a formation water salinity of 12,000 mg / L.

[0024] II. Specific Implementation Steps 1. Step S1: Initial trajectory design of horizontal well 1-1. Operational details: Petrel software was used, with the target area being the sandstone thickness greater than 5 meters obtained from seismic inversion. The initial trajectory was designed in combination with the reservoir distribution characteristics of adjacent wells A and B. The trajectory target point was located in the predicted high-permeability zone, with a well inclination angle of 85° to 90° and a horizontal section length of 1500 meters.

[0025] 1-2. Data support: Permeability data from adjacent wells is used to constrain the trajectory to avoid low-permeability zones with permeability less than 0.3 mD; the sandstone distribution probability volume obtained from seismic inversion serves as the basis for trajectory path optimization.

[0026] 2. Step S2: Real-time logging-while-drilling data acquisition 2-1. Equipment and Parameters: The Schlumberger MAXIS 500 logging system was used to collect the following data in real time: Gamma GR: Sampling rate 0.1 meters, measurement range 0 to 200 API; Resistivity RT: Induction logging, resolution 1 meter, measurement range 0.1 to 2000 Ω·m; Acoustic transit time (DT): sampling rate 0.5 meters, measurement range 150 to 400 μs / m; Neutron porosity (CNL): sampling rate 0.5 m, measurement range 0 to 40%.

[0027] 2-2. Transmission delay control: The delay time for logging signals to be transmitted from downhole to the surface processing system is controlled within 25 seconds.

[0028] 3. Step S3: Spatiotemporal calibration and quality control 3-1. Spatiotemporal calibration: The depth moving average matching algorithm is used to unify the depth benchmark between seismic data (vertical resolution 10 meters) and well logging data; the cuttings depth identification delay is controlled within 4 meters based on the cuttings logging depth.

[0029] 3-2. Quality Control: Multi-parameter Cross-validation: When the GR value fluctuates beyond 80 API (preset threshold), the correlation analysis between RT and DT is triggered; if the correlation coefficient between RT and DT is less than 0.7 (set standard), the well section is retested; Clay content correction and potential gas layer labeling: Clay content was corrected using CNL data, calculated using the following formula: ,in The maximum GR value of mudstone from adjacent wells is taken as 150 API. After correction, if the density DEN is less than 2.4 g / cm³... 3 If CNL is greater than 15%, then the well section is marked as a potential gas layer and a high-quality reservoir.

[0030] 4. Step S4: Calculation of Reservoir Matching Index (SMI) and Recommendations for Trajectory Adjustment 4-1. Input features for machine learning models: GR curve variation characteristics: Calculate the GR variance for each 10-meter well section to reflect the degree of drastic lithological changes; RT gradient characteristics: Calculate the rate of change of RT (ΔRT / Δdepth) per 10-meter well section; DT variation characteristics: compare the deviation between the measured DT and the average DT of adjacent wells; Neutron-density intersection characteristics: Based on the intersection point of CNL and DEN, determine whether it falls into the gas layer indicator region: DEN < 2.4 g / cm³ 3 And CNL > 18%.

[0031] 4-2. MI Calculation and Adjustment Trigger: The SMI value ranges from 0 to 1, with a preset threshold of 0.6. When the calculated SMI value for a certain well section is 0.52, the trajectory adjustment mechanism is triggered. By fusing seismic inversion sand body probability volumes with real-time logging inversion results, a spatial probability distribution cloud map of sand bodies within a 50-meter radius around the wellbore is generated. Reservoir boundary uncertainties are marked with different colors, with red areas indicating uncertainties exceeding 30%.

[0032] 5. Step S5: Trajectory Dynamic Optimization and Engineering Constraint Verification 5-1. Candidate Scheme Generation: Three trajectory adjustment schemes are generated every 15 meters of drilling: Option 1: Upward slope of 1.5° / meter, avoiding mudstone interlayers with GR greater than 100 API; Option 2: Cut down 3° / meter, track DT greater than 280 μs / m and DEN less than 2.3 g / cm 3 High-permeability sand body area; Option 3: Maintain the original trajectory and adjust the drilling fluid density to 1.2 g / cm³. 3 .

[0033] 5-2. Engineering constraint verification: Slope rate constraint: Scheme 2 has a slope rate of 3° / m, which meets the constraint of "≤5° / m"; High-quality reservoir contact rate target: Scheme 2 predicts a contact rate of 95%, which is higher than the "≥90%" standard in claim 8; Drilling fluid safety window: Equivalent circulation density of Scheme 2 is 1.15 g / cm³. 3 Less than the formation fracture pressure gradient, 0.015 MPa / m, the threshold value after multiplying by a safety factor of 1.3 is 1.25 g / cm³. 3 .

[0034] 5-3. Weighted Evaluation: Trajectory adjustment range, weight 0.4: Scheme 2 has a moderate adjustment range and scores 80 points; Formation uncertainty coefficient, weight 0.3: Scheme 2 corresponds to a sand body probability of 0.8, with low uncertainty, and a score of 90 points; Drilling fluid safety window margin, weight 0.3: Scheme 2 margin 0.1 g / cm³ 3 The score was 85 points.

[0035] Ultimately, option 2 was chosen to perform the downcut tracking.

[0036] 6. Step S6: Reservoir encounter rate analysis after drilling completion 6-1. Quantitative indicators: The total length of the horizontal section of the well is 1500 meters. A high-quality reservoir was actually encountered, with an SMI ≥ 0.6 and a potential gas layer. The length of the reservoir is 1380 meters. The calculated reservoir encounter rate is 92%, which meets the target of "≥ 90%" in claim 8. 6-2. Basis for subsequent optimization: Comparison with adjacent wells: Adjacent well B, using the traditional geological steering method, achieved a drilling success rate of 75%; 1125 meters of high-quality reservoir were encountered in the 1500-meter horizontal section. In this embodiment, by dynamically adjusting the SMI threshold from 0.6 to 0.55, the drilling success rate was increased to 92%, and the contact rate with high-quality reservoirs reached 95%, verifying the effectiveness of the method. Parameter optimization recommendation: When promoting in similar low-permeability blocks, it is recommended to set the SMI threshold to 0.55 to balance reservoir contact rate and engineering risk.

[0037] Comparative Example 1: Geological Steering Method for Horizontal Wells in Traditional Low-Permeability Sandstone Reservoirs I. Basic Information and Scenarios 1. Scenario: A low-permeability sandstone gas reservoir in an adjacent block of the same basin was selected. The target stratum is the He 8 section sandstone, with a burial depth of 2850-2950 meters. The average permeability of the reservoir is 0.4 mD, the gas saturation is 60%, and there are mudstone interlayers and high-permeability sandstone lenses. The geological conditions are comparable to those in the previous example.

[0038] 2. Basic Information: Seismic inversion data: The conventional sparse pulse inversion algorithm was used, which is not constrained by high-frequency well logging data. The vertical resolution is approximately 15 meters, and the wave impedance threshold is fixed at 13,000 kg / m. 3 ·m / s, without dynamic adjustment based on core analysis; Adjacent well data: Only the permeability and gas saturation data of adjacent well A were referenced; formation water salinity information was not collected.

[0039] II. Specific Implementation Steps S1: Initial trajectory design: Based on the sandstone thickness distribution obtained from seismic inversion, only areas with a thickness greater than 8 meters and the reservoir distribution characteristics of adjacent well A are identified. The initial trajectory is designed using Petrel software. The target point is located in the high wave impedance zone predicted by seismic analysis. The well inclination angle is designed to be 85° to 90°, and the horizontal section length is 1500 meters. Trajectory avoidance is not combined with the low permeability zone data of adjacent well B.

[0040] S2: Logging-while-drilling data acquisition: Using an older logging system, such as the Eclips 5700, the following real-time data was collected: Gamma GR: Sampling rate 0.5 meters, measurement range 0 to 200 API; Resistivity RT: Induction logging, resolution 2 meters, measurement range 0.1 to 1000 Ω·m; Acoustic transit time DT: Sampling rate 1 meter, measurement range 150 to 400 μs / m; Neutron porosity (CNL): sampling rate 1 meter, measurement range 0 to 40%; The delay time for logging signals to be transmitted from downhole to the surface processing system is approximately 150 seconds.

[0041] S3: Data Processing and Reservoir Identification Lacking spatiotemporal calibration: The depth matching algorithm was not used, the depth benchmarks of seismic data and well logging data were not consistent, the cuttings logging depth identification was delayed by about 12 meters, resulting in a reservoir boundary prediction error of more than 8 meters; Quality control was lax: only manual inspection of GR curve anomalies was performed, and multi-parameter cross-validation was not implemented; when the GR value fluctuated by more than 100 API, correlation analysis between RT and DT was not triggered, and the original data was used directly. Reservoir identification rules: The fixed resistivity threshold is 50 Ω·m. The influence of formation water salinity is not considered. Gamma GR < 80 API is identified as sandstone reservoir. The clay content is not corrected by neutron porosity CNL, so potential gas layers cannot be accurately identified.

[0042] S4: Track Adjustment and Engineering Implementation Adjustment mechanism: Relies on geological engineers to manually interpret well logging curves. When the GR value is found to be >100 API, it is judged to be a mudstone interlayer, and an upward adjustment suggestion is manually proposed. There is no quantitative reservoir matching degree assessment. Lack of engineering constraints: The build-up rate was not limited; during one adjustment, the build-up rate reached 7° / meter, causing wellbore trajectory irregularities. The drilling fluid safety window was not assessed, resulting in an equivalent circulating density of 1.35 g / cm³. 3 It is close to the formation fracture pressure threshold, 1.3 g / cm³. 3 ; Candidate scheme generation: Only one adjustment scheme is generated every 50-100 meters of drilling, and the response speed lags behind the changes in the reservoir.

[0043] Key Indicator Comparison Table

[0044] In summary, traditional methods suffer from low reservoir identification accuracy, delayed trajectory adjustment, and high engineering risks due to the lack of multi-source data spatiotemporal calibration, machine learning-based intelligent identification, and dynamic optimization under engineering constraints. This invention, through technological innovation, improves the drilling encounter rate by 17 percentage points while reducing the engineering accident rate by 72%, significantly validating the necessity and advancement of the "precise fusion of multi-source data + intelligent decision-making" technical approach.

[0045] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A geological steering method for horizontal wells in low-permeability sandstone reservoirs, characterized in that, Includes the following steps: S1. Based on the seismic inversion results constrained by high-frequency logging data and reservoir parameter data of adjacent wells, design the initial trajectory of the horizontal well; S2. Real-time acquisition of logging-while-drilling data, including gamma ray (GR), resistivity (RT), sonic transit time (DT), and neutron porosity (CNL) signals; S3. Perform spatiotemporal calibration and quality control on the seismic inversion data, adjacent well reservoir parameter data, and real-time logging data from step S1 and step S2: The spatiotemporal calibration refers to using a depth matching algorithm to unify the depth benchmark between the macroscopic scale of seismic logging and the microscopic scale of logging; The quality control refers to implementing multi-parameter cross-validation, and triggering data retesting when the gamma ray GR value fluctuation exceeds a preset threshold, or when the correlation between resistivity RT and sonic transit time DT is lower than the set standard. S4. Based on the reliable data processed in step S3, and combined with the preset geological rules for identifying low-permeability sandstone reservoirs, a machine learning algorithm is applied to calculate the reservoir matching index (SMI). When the SMI is lower than a preset threshold, a trajectory adjustment suggestion is generated. S5. Verify the feasibility of the adjustment suggestions generated in step S4 in conjunction with drilling engineering constraints, and perform dynamic optimization of the horizontal segment trajectory with the goal of maximizing the contact rate of high-quality reservoirs. S6. Based on the logging data obtained after drilling is completed, the actual reservoir drilling rate is quantitatively analyzed to provide a basis for optimizing subsequent construction parameters.

2. The geological steering method for low-permeability sandstone reservoir gas horizontal wells according to claim 1, characterized in that, The adjacent well reservoir parameter data includes permeability, gas saturation, and formation water salinity data; the resistivity threshold used to identify the reservoir in the geological rules is dynamically adjusted based on the formation water salinity obtained or predicted in real time.

3. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 1, characterized in that, During the spatiotemporal calibration and quality control process, real-time requirements must be met: the cuttings depth identification delay is <5 meters; the delay of logging signals transmitted from downhole to the surface processing system is <30 seconds.

4. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 3, characterized in that, In the quality control process, the clay content is corrected using neutron porosity (CNL) logging data; when the corrected density (DEN) and neutron porosity (CNL) data meet the preset gas reservoir response characteristic conditions, the well section is marked as a potential gas reservoir; the potential gas reservoir is a high-quality reservoir.

5. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 1, characterized in that, The machine learning model input feature parameters for the reservoir matching index SMI include: gamma GR curve variation features, resistivity RT gradient features, acoustic transit time DT variation features, and neutron-density intersection features.

6. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 1, characterized in that, When the reservoir matching index (SMI) is lower than a preset threshold, the following operations are automatically performed: merging seismic inversion data and real-time logging data to generate a spatial probability distribution cloud map of sand bodies around the wellbore; the cloud map is generated by merging seismic wave attribute volumes and logging inversion results, and the predicted reservoir boundaries and their uncertainty range are marked in a visual manner.

7. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 6, characterized in that, The spatial probability distribution cloud map of the sand body is input into the machine learning model in step S4 to update the calculation results of the reservoir matching index SMI; at the same time, the real-time drilling trajectory and the predicted trajectory after error calibration are superimposed on the design trajectory map.

8. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 1, characterized in that, The specific constraints of the drilling project include: the maximum single-meter build-up rate for trajectory adjustment is limited to 5° / meter; the target for horizontal drilling is a high-quality reservoir contact rate of ≥90%; and the drilling fluid equivalent circulation density must be less than the specified safety factor of the formation fracture pressure gradient.

9. The geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 1, characterized in that, The trajectory dynamic optimization includes: generating at least 3 candidate trajectory adjustment schemes based on the latest data every 10-20 meters of drilling unit length; evaluating each candidate scheme according to the drilling engineering constraints, and quantifying the engineering risk level through a weighted algorithm. The weighting factors include: trajectory adjustment range (weight 0.4), formation uncertainty coefficient (weight 0.3), and drilling fluid safety window margin (weight 0.3).

10. A geological steering method for a low-permeability sandstone reservoir gas horizontal well according to claim 9, characterized in that, The trajectory adjustment operation specifically includes: upward obstacle avoidance operation: when mudstone interlayers are identified, upward adjustment is performed at a build-up rate of ≤3° / m; downward tracking operation: when a high-permeability sand body region is identified, downward adjustment is performed at a build-up rate of ≤5° / m; wherein: the criteria for determining the high-permeability sand body region are: sonic transit time >280μs / m and bulk density <2.4 g / cm³. 3 The criteria for determining the mudstone interlayer are: gamma GR value > 100 API and the intersection value of neutron porosity CNL and density DEN is within the preset mudstone response range.

Images