Ground surface-underground combined gas extraction effect comprehensive evaluation method

Abstract

The invention discloses an earth surface-underground combined gas extraction effect comprehensive evaluation method. The method comprises the following steps: acquiring geological-engineering integrated data; acquiring sparse underground gas parameter measured data in the extraction process; constructing a geostatistics-physical information double constraint neural network evaluation model; training the model by minimizing a triple mixing loss function; and finally, generating a gas parameter space-time distribution field of the full extraction area by utilizing the trained model, and outputting an evaluation index. According to the method, through double constraints of physical information and geostatistics, the modeling problem under sparse data is solved, the model has both physical significance and geological rationality, and the evaluation precision and reliability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety engineering technology, and more specifically, to a comprehensive evaluation method for the combined surface-underground gas extraction effect. Background Technology

[0002] For a long time, coal has served as my country's basic energy source and an important industrial raw material, strongly supporting the development of the national economy. However, during coal mining, coal mine gas problems, especially coal and gas outburst accidents, have severely restricted the high-quality development of coal mining enterprises. Guizhou Province is a major coal-rich province in southern China and a coal supply center in Southwest China. Most of the existing coal mines in Guizhou are high-gas or gas outburst mines, and their coal mining is affected by "many coal seams (10-30 layers) and high gas levels (10-30 layers)". ), poor breathability ( mD), soft coal and rock ( value It is constrained by unfavorable factors such as "( )".

[0003] Among all coal mine safety threats, "coal exposure at the rock face" is the key and difficult point in the prevention and control of high-gas outburst mine disasters. According to statistics, 90% of extra-large outbursts with a capacity of over 1,000 tons occur during coal exposure at the rock face, and their average outburst intensity is 7 to 14 times that of outbursts in coal roadways.

[0004] Traditional underground coal seam uncovering technology generally employs a sequential "geophysical exploration-drilling-tunneling" process, characterized by "short exploration and slow tunneling." In Guizhou's high-gas, large-span coal seam environment, this approach suffers from serious drawbacks, including complex uncovering procedures, a large number of anti-outburst drill holes, long gas control cycles, and extremely low tunneling efficiency (an average monthly tunneling advance of only 20-30 meters).

[0005] To overcome the bottleneck of "short-term exploration and slow excavation" and achieve a "long-term exploration and fast excavation" model for "safe and rapid coal seam exposure," the industry has begun exploring integrated surface and underground gas control technologies. This technology utilizes surface horizontal wells for advanced exploration, fracturing and permeability enhancement, and pre-extraction, combined with the connection and continuous extraction of underground roadways or boreholes, to achieve efficient and comprehensive advanced control of gas in coal seam exposure areas.

[0006] However, after implementing complex surface-underground combined extraction projects, how to accurately, efficiently, and comprehensively evaluate their extraction effects to determine whether the gas level in the target area (such as in front of a rock tunnel) has dropped to a safe standard (e.g., a gas pressure below 0.74 MPa and a gas content below 8 MPa as stipulated in the "Detailed Rules for the Prevention and Control of Coal and Gas Outbursts"). This, along with guiding subsequent tunneling operations, has become a new key technical challenge.

[0007] Existing evaluation methods typically rely on drilling a limited number of inspection boreholes downhole and measuring gas content and pressure at these discrete points to infer compliance across the entire area (surface or volume). This traditional evaluation method suffers from the following technical drawbacks:

[0008] 1) Data sparsity: Due to the limitations of drilling costs, downhole operating space and construction cycle, the number of downhole measurement points (inspection holes) is inevitably sparse.

[0009] 2) Interpolation uncertainty: When drawing contour maps of gas parameters based on sparse data, commonly used methods such as inverse distance weighting (IDW) or linear interpolation lack rigorous physical basis and geostatistical foundation. In areas with sparse data and complex geological structures, the reliability of interpolation results is extremely poor, which may lead to incorrect classification of "compliant areas" and leave potential safety hazards.

[0010] 3) Limitations of Geostatistical Methods: Although geostatistical methods (such as Kriging) have been applied to mapping gas pressure and content, providing optimal linear unbiased estimates and quantifying uncertainties, Kriging is essentially a purely data-driven interpolation method. It does not adhere to the physical laws governing gas seepage in porous media (such as Darcy's law and the law of conservation of mass). During extraction, the gas field is a dynamically evolving physical field, and Kriging cannot accurately reflect this physical process.

[0011] 4) Limitations of Numerical Simulation and Data Assimilation Methods: While numerical simulation methods based on physical partial differential equations (PDEs) (such as the "thermal-fluid-solid" multi-field coupling model mentioned in Project 3) can describe physical processes, their forward simulation relies on accurate descriptions of key geological parameters such as permeability fields, which is almost impossible to achieve in highly heterogeneous coal seams. Inverse simulation (data assimilation), such as the emerging Physical Information Neural Network (PINN), although theoretically able to integrate PDEs and data, often exhibits high uncertainty and non-uniqueness in solutions when measured data is extremely sparse, making training difficult to converge to the correct solution.

[0012] In summary, existing evaluation methods face a sharp technical contradiction between the sparsity of downhole measured data and the need for high precision and reliability in evaluation models. A new evaluation method is urgently needed that can fully utilize limited sparse measured data while integrating the physical laws of gas seepage and geological spatial correlations to achieve a comprehensive evaluation of extraction effectiveness with high precision and reliability. Summary of the Invention

[0013] The main objective of this invention is to overcome the shortcomings of the prior art and provide a comprehensive evaluation method for the combined surface-underground gas extraction effect, thereby solving the technical problems existing in the prior art.

[0014] To achieve the above objectives, the present invention provides the following technical solution:

[0015] A comprehensive evaluation method for the effectiveness of surface-underground combined gas extraction includes the following steps:

[0016] Step A: Obtain integrated geological and engineering data of the surface-underground combined gas extraction system. The integrated geological and engineering data includes geological structure data, coal reservoir physical property parameters, horizontal well engineering parameters, underground roadway engineering parameters, and extraction dynamic parameters.

[0017] Step B: Obtain multi-stage, sparse measured data of downhole gas parameters from the combined gas extraction system during the extraction process. The measured data includes gas pressure and gas content at predetermined three-dimensional spatial coordinates.

[0018] Step C: Construct a geostatistical-physical information dual-constraint neural network evaluation model. The evaluation model is used to characterize the nonlinear mapping relationship between the integrated geological-engineering data, the sparse measured data, and the spatiotemporal distribution field of gas parameters in the full extraction area.

[0019] Step D: Using the sparse measured data of downhole gas parameters, minimize a triple hybrid loss function. The geostatistical-physical information dual-constraint neural network evaluation model was trained.

[0020] Step E: Using the trained geostatistical-physical information dual-constraint neural network evaluation model, generate arbitrary spatiotemporal coordinate points within the entire extraction area. The distribution field of gas pressure and gas content;

[0021] Step F: Based on the distribution field, calculate and output evaluation indicators characterizing the combined surface-underground gas extraction effect. The evaluation indicators include the proportion of areas meeting extraction standards and the spatial distribution of residual gas content.

[0022] Furthermore, the triple hybrid loss function Represented as: ,

[0023] in: For data fidelity loss, it is used to minimize the difference between the predicted value and the measured value of the geostatistical-physical information dual-constraint neural network evaluation model at the sparse measured data points; The physical constraint loss term is used to minimize the residual of the predicted value of the geostatistical-physical information dual-constraint neural network evaluation model at a preset configuration point within the extraction area to the predetermined gas seepage partial differential equation. This is a geostatistical structure loss term used to minimize the difference between the experimental variogram calculated based on the sparse measured data and the model variogram calculated based on the predicted values ​​of the geostatistical-physical information dual-constraint neural network evaluation model. and These are the weighting coefficients for the physical constraint loss term and the geostatistical structure loss term.

[0024] Furthermore, the aforementioned Calculated in the following way: , , ,

[0025] in, The residual of the partial differential equation is... The number of the preset configuration points. Porosity For gas density, For gas saturation, For time, For gas seepage velocity, For source and sink items, For absolute penetration rate, The relative permeability of the gas. For gas viscosity, For gas pressure, It is the acceleration due to gravity. For divergence operators, For gradient operators, For the residual In the The value of each preset configuration point, where i is the sequence number of the preset configuration point.

[0026] Furthermore, the aforementioned Calculated in the following way: , , ,

[0027] in, Let be the experimental variability function. The model's mutation function is... Spatial lag distance, Lag distance The corresponding number of measured data point pairs For in position The measured gas pressure value, Lag distance The corresponding number of model sampling point pairs, The evaluation model of the geostatistical-physical information dual-constraint neural network at the sampling location The predicted gas pressure value, Lag distance used for comparison The total number.

[0028] Furthermore, the surface-underground combined gas extraction system is implemented using an "L"-shaped horizontal well, with a horizontal section length of not less than 650 meters, and segmented fracturing and permeability enhancement are performed on the preferred target coal seams, which are no less than 10 layers.

[0029] Furthermore, the surface-underground combined gas extraction system includes a surface horizontal well shaft and an underground connecting roadway or cross-layer borehole. The underground connecting roadway or cross-layer borehole achieves safe and precise docking with the surface horizontal well shaft with a positioning accuracy at the centimeter level.

[0030] Furthermore, the combined gas extraction system adopts a combination of surface extraction and underground negative pressure continuous extraction. The underground negative pressure continuous extraction is connected to the underground gas extraction system through the underground connecting roadway or through-layer borehole.

[0031] Furthermore, the sparse downhole gas parameter measured data in step B are obtained by collecting gas parameter test points from multiple cross-layer boreholes deployed within the target rock passage excavation footage and the comparison rock passage excavation footage.

[0032] Furthermore, the evaluation index in step F also includes assessing the percentage increase in tunneling efficiency brought about by the combined gas extraction by comparing the average tunneling efficiency of the target tunnel with that of the comparison tunnel.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1) The present invention The term (geostatistical structure loss) provides powerful geostatistical prior knowledge (i.e., spatial correlation structure). The physical constraint loss term provides physical priors. The combination of these two (G-PINN) allows the model to generate a high-precision, highly stable spatiotemporal distribution field with only a small amount of sparse data. Unlike Kriging, this invention... This ensures that the dynamic evolution of evaluation results (such as the gas pressure field) conforms to the physical laws of gas seepage; unlike the standard PINN, the present invention... This method ensures that the evaluation results conform to the true geological spatial correlation structure revealed by the measured data. Through the joint constraints of the triple loss function of data fidelity, physical laws, and geostatistical structure, the gas parameter distribution field generated by this method is closer to the real situation than any single method, which greatly improves the accuracy of the evaluation of the extraction effect and provides a reliable scientific basis for safe and rapid coal seam exposure in the rock gate.

[0035] 2) Because this method has a strong modeling capability for sparse data, it can significantly reduce the number and frequency of downhole inspection boreholes while maintaining the same evaluation accuracy, saving a lot of drilling costs and time, which is in line with the economic benefit goal of "long exploration and fast excavation". Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating the comprehensive evaluation of gas extraction effectiveness in an embodiment of the present invention;

[0038] Figure 2 This is a structural diagram of the G-PINN evaluation model in an embodiment of the present invention;

[0039] Figure 3 This is a computer-generated graph of the triple loss function in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to specific embodiments. It should be noted that these embodiments are only used to illustrate the invention and are not intended to limit the scope of the invention.

[0041] Example 1

[0042] refer to Figure 1-3 This embodiment provides a best practice for the comprehensive evaluation method of surface-underground gas extraction. Using the 241 stone gate of the 24th mining area of ​​the Huoshaopu Mine of Guizhou Panjiang Refined Coal Co., Ltd. as an engineering application scenario, this embodiment details the complete implementation process of the method of the present invention.

[0043] 1. Acquiring integrated geological and engineering data

[0044] 1.1 Engineering Geological Background and Optimization of Geological Targets

[0045] The engineering application scenario in this embodiment has complex geological conditions and poses a risk of coal and gas outbursts.

[0046] (1) Geological data collection and database construction: Collect coalfield exploration data, coalbed methane exploration data, mine production data, seismic data and adjacent parameter well data in the region, and establish a digital basic database.

[0047] (2) Three-dimensional geological modeling: Using geological modeling software (such as Petrel), we can perform detailed structural and lithofacies modeling of the coal-bearing rock series, construct a three-dimensional geological model of the 241 Shimen coal seam exposure area, and finely depict the geological structures such as faults and folds in the target area, as well as the spatial distribution characteristics of the main mineable coal seams (such as No. 12, No. 17, No. 19, etc.).

[0048] (3) Optimization of the four properties of the target strata: Based on the three-dimensional geological model, a comprehensive study of the four properties of the coal seam group in the target area is conducted:

[0049] Coal-bearing properties, including analysis of coal seam thickness, continuity, structure, and interbedded rock distribution.

[0050] Gas content, combined with measured gas content data and well logging interpretation, is used to assess the gas content and gas pressure of each coal seam.

[0051] Compressionability is assessed by analyzing the coal body structure, coal and rock mechanical parameters, and geostress field to determine the ease or difficulty of hydraulic fracturing for permeability enhancement.

[0052] Mining feasibility is assessed by evaluating the coal seam spacing, roof and floor lithology, and determining the compatibility of combined seam extraction.

[0053] (4) Output of optimization results: Based on the evaluation results of the "four properties", within the influence range of the 241 stone gate trajectory, 10 to 15 coal seams were selected as the main target coal seams for segmented fracturing and permeability enhancement and gas extraction in surface horizontal wells. The data of these selected target permeability enhancement coal seams constitute part of the integrated geological-engineering data.

[0054] 1.2 Construction of Surface Horizontal Well Cross-Sectional Pumping System

[0055] Based on the above optimization results, the surface component of the surface-underground integrated gas extraction system is constructed as follows:

[0056] (1) Drilling and completion of surface horizontal well: Based on the design trajectory direction of 241 stone gate, deploy an "L"-shaped surface horizontal well at a favorable location on the ground.

[0057] The wellbore structure design employs an optimized wellbore structure and drilling fluid formulation to address coal and rock hydration and wellbore instability issues.

[0058] The trajectory control employs high-precision geological steering technology and a rotary steering drilling system to ensure that the horizontal well trajectory accurately passes through multiple target coal seams selected in Section 1.1, in accordance with the design scheme.

[0059] Technical specifications stipulate that the length of the horizontal section should be no less than 650 meters.

[0060] (2) Adaptive fracturing permeability enhancement:

[0061] Segmented fracturing employs casing segmentation, hydraulic jetting, or open-hole packer techniques to segment a 650-meter horizontal section into multiple stages (e.g., 10-20 stages).

[0062] Fracturing parameters were optimized by adopting adaptive fracturing construction parameters to address the heterogeneity and differences in mechanical properties of coal seams.

[0063] The technical specifications aim to achieve effective fracturing and permeability enhancement of at least 10 target coal seams.

[0064] (3) Surface extraction: After fracturing, extraction equipment is installed on the ground to drain and depressurize and extract gas, and to obtain dynamic extraction parameters.

[0065] 1.3 Pre-drainage for shaft-tunnel connection and surface-to-surface connection

[0066] After a period of extraction from the surface horizontal well (e.g., 6 months), when the underground 241 stone gate has been excavated to the predetermined position, the shaft and tunnel connection and subsequent pre-extraction will be carried out:

[0067] (1) Efficient and safe docking of shafts and tunnels:

[0068] The docking plan involves constructing one or more large-diameter (e.g., diameter) connectors within the 241 stone gate tunnel. () Through-layer drilling.

[0069] The docking technology utilizes high-precision directional drilling technology (such as gyroscope inclination measurement and magnetic positioning system) to monitor and adjust the trajectory of the downhole borehole in real time, so that it can be safely and accurately docked to the end of the horizontal section of the surface horizontal well.

[0070] The docking accuracy is improved to achieve centimeter-level positioning accuracy between the downhole borehole and the surface wellbore.

[0071] (2) Rapid pre-drainage between the well and the surface:

[0072] After the docking is completed, the ground extraction system will be shut down.

[0073] Connect the underground borehole (or roadway) to the underground gas drainage system (negative pressure system) through pipelines.

[0074] Working principle: By utilizing the underground drainage system (usually -10kPa to -30kPa), which has a much greater negative pressure difference than surface drainage, the system can achieve "rapid continuation" and "deep depressurization" of the fracturing and transformation area. The aim is to significantly shorten the gas control cycle and facilitate rapid tunneling in the rock passage.

[0075] At this point, all the integrated geological and engineering data required for step A have been obtained.

[0076] 2. Obtain sparse measured data

[0077] In accordance with the requirements of the "Detailed Rules for the Prevention and Control of Coal and Gas Outbursts", multiple sets of cross-layer inspection boreholes were constructed in stages (e.g., before extraction, on the 30th day, the 60th day, and the 90th day of extraction) in front of the 241 stone gate (target stone gate) and in front of another stone gate with similar geological conditions (comparison stone gate) where the joint extraction method of this invention was not implemented.

[0078] Parameters were collected using methods such as core drilling, coal sample desorption, and borehole pressure measurement to obtain data in specific three-dimensional coordinates. ) and specific time ( ) gas pressure ( ) and gas content ( Actual measured data.

[0079] Data characteristics: Due to the high cost and long cycle of borehole inspection, the obtained measured dataset ( It must be highly sparse in space and time.

[0080] 3. Constructing and training the G-PINN evaluation model

[0081] To resolve the technical contradiction between "sparse data" and "physical / geological dual constraints", this embodiment constructs and trains a geostatistical-physical information dual-constraint neural network (G-PINN) evaluation model.

[0082] 3.1 Model Construction (corresponding to step C)

[0083] The neural network architecture adopts a fully connected neural network (FNN) architecture. The input layer uses 4 neurons, corresponding to the normalized spatiotemporal coordinates. The hidden layer consists of 8 layers, each with 64 neurons, and the activation function is the hyperbolic tangent (Tanh). The output layer uses 2 neurons, corresponding to the predicted gas pressure. and gas content .

[0084] 3.2 Model Training (corresponding to step D)

[0085] Using the sparse measured data obtained in step B, the result is obtained by minimizing a triple mixture loss function. Train the model: ,

[0086] in, and These are the weighting coefficients (e.g., 0.1 and 0.05 respectively).

[0087] (1) (Data fidelity loss): This is the mean squared error (MSE), which ensures that the network is accurate on sparse measured points and serves as an "anchor" for model training. ,

[0088] in For sparse measured data points (e.g.) The quantity of ).

[0089] (2) (Physical Constraint Loss): This term is based on the governing equations of gas (assumed to be a single-phase gas) seeping through porous media (combining Darcy's law and the law of conservation of mass in a PDE). Its calculation method is as follows: Define the PDE residual. : ,

[0090] Among them, seepage velocity (Darcy velocity) Given by Darcy's Law: ,

[0091] Loss Items Defined as in Preset configuration points (e.g.) The sum of the root mean square of the residuals on: ,

[0092] in, For residuals, Porosity For gas density, For gas saturation, For time, For gas seepage velocity, For source and sink items, For absolute penetration rate, The relative permeability of the gas. For gas viscosity, For gas pressure, It is the acceleration due to gravity. For divergence operators, This is the gradient operator. This loss term forces the network to use solutions that are approximate solutions to the PDE across the entire time and space domain.

[0093] (3) (Geostatistical Structure Loss): This item penalizes the discrepancy between the "spatial structure" implied by the neural network predictions and the "true spatial structure" revealed by the measured data. First, it utilizes the sparse dataset from step B (e.g., (Initial gas pressure data at time), calculate the experimental variability function. : ,

[0094] Then, in each iteration of training, the current prediction value of G-PINN is... Calculate the model variability function : ,

[0095] Finally, calculate Loss (in) lag distance (mean square error) ,

[0096] in, Spatial lag distance, Lag distance The corresponding number of measured data point pairs To measure the gas pressure value, The number of sample point pairs in the model. To predict gas pressure values ​​for the model, Lag distance used for comparison The total number.

[0097] 4. Generate the spatiotemporal distribution field of gas parameters (corresponding to step E)

[0098] G-PINN was trained using optimizers such as Adam or L-BFGS until... Convergence. After training, the G-PINN model becomes a continuous function. Input any grid point within the sampling area. and at any time (For example, on the 30th, 60th, and 90th day after extraction), the G-PINN model can instantly output the gas pressure at that point at that moment. and gas content This generates a high-resolution spatiotemporal dynamic evolution distribution map of gas parameters in the entire extraction area.

[0099] 5. Evaluation of sampling effectiveness and output of indicators (corresponding to step F)

[0100] A quantitative evaluation is performed based on the spatiotemporal distribution field generated in step E:

[0101] (1) Percentage of areas meeting extraction standards: According to the "Detailed Rules for Prevention and Control of Coal and Gas Outbursts", a gas pressure threshold for safe coal seam exposure is set (e.g., ) and gas content threshold (e.g. ).exist At a given time (e.g., the planned coal seam exposure time), within the 241-gate tunneling profile predicted by the calculation model, the following conditions are met: and Voxel volume Calculate the percentage of areas that meet the standards: .when At that time, the area is determined to be a safe and compliant zone.

[0102] (2) Spatial distribution of residual gas content: Output Residual gas content at any given time The three-dimensional visualization map intuitively shows the specific spatial location of the high gas residue zone ("hard bone"), guiding the downhole to carry out supplementary drilling or local enhanced extraction.

[0103] (3) Comparison of tunneling efficiency: Compare the average monthly tunneling footage of the target rock passage 241 (using the combined extraction and evaluation method of this invention) with the comparison rock passage mentioned in step B (using the traditional "short exploration and slow tunneling" method). Calculate the percentage increase in efficiency: Expected goals: .

[0104] 6. Model Validation and Beneficial Effects

[0105] To verify the superiority of the G-PINN method in this embodiment, a comparative experiment was conducted. Sparse measured data obtained from step B... In this process, 10% of the data is randomly reserved as a "validation set". The remaining 90% of the "training set" is modeled using three different methods (traditional Kriging, standard PINN, and the G-PINN of this invention), and the root mean square error (RMSE) of the predictions is tested on the "validation set".

[0106] G-PINN and contrast model hyperparameter settings

[0107]

[0108] Comparison of RMSE of different methods on the validation set

[0109]

[0110] Results Analysis: Table 2 shows that Method 2 (standard PINN) performs the worst with the highest RMSE under sparse data. Method 1 (Kriging) has moderate error, but its interpolation results do not guarantee physical conservation. Method 3 (the G-PINN of this invention) has a significantly lower RMSE than the other two and the highest prediction accuracy. The fundamental reason for this is... The introduction of (geostatistical structural loss). and The combined effect transforms the ill-conditioned problem under sparse data into a well-conditioned problem, allowing the model to converge to a unique, high-precision solution that simultaneously satisfies physical laws and geological statistics.

[0111] This embodiment fully demonstrates the non-obvious beneficial effect of the G-PINN method of the present invention in performing high-precision evaluation under sparse data conditions.

[0112] Example 2

[0113] This embodiment is a variation of Embodiment 1. The method in this embodiment is basically the same as that in Embodiment 1, except that: in Section 3.2(2) of Embodiment 1, the method described... (Physical constraint loss) is constructed based on the partial differential equation (PDE) of single-phase gas seepage.

[0114] In this embodiment, for coal seams with high water content, in order to further improve the model accuracy, the... (Physical constraint loss) is constructed based on the partial differential equations (PDEs) of the gas-water two-phase flow. At this point, the PDE residuals... This will include continuity equations for both the gas phase and the aqueous phase, as well as the seepage velocity. (Gas) and (Water) will be simultaneously affected by relative permeability ( and ) and capillary pressure ( The impact of ).

[0115] Although this (two-phase flow) physical constraint loss term is computationally more complex, in high water-cut regions, it can more accurately reflect the competitive seepage physical process of gas (free gas) and water, thus providing stronger physical constraints for the G-PINN model and making its output gas pressure ( ) and gas saturation ( The distribution field has higher physical fidelity.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A comprehensive evaluation method for the effectiveness of surface-underground combined gas extraction, characterized in that, Includes the following steps: Step A: Obtain integrated geological and engineering data of the surface-underground combined gas extraction system. The integrated geological and engineering data includes geological structure data, coal reservoir physical property parameters, horizontal well engineering parameters, underground roadway engineering parameters, and extraction dynamic parameters. Step B: Obtain multi-stage, sparse measured data of downhole gas parameters from the combined gas extraction system during the extraction process. The measured data includes gas pressure and gas content at predetermined three-dimensional spatial coordinates. Step C: Construct a geostatistical-physical information dual-constraint neural network evaluation model. The evaluation model is used to characterize the nonlinear mapping relationship between the integrated geological-engineering data, the sparse measured data, and the spatiotemporal distribution field of gas parameters in the full extraction area. Step D: Using the sparse measured data of downhole gas parameters, minimize a triple hybrid loss function. The geostatistical-physical information dual-constraint neural network evaluation model was trained. Step E: Using the trained geostatistical-physical information dual-constraint neural network evaluation model, generate arbitrary spatiotemporal coordinate points within the entire extraction area. The distribution field of gas pressure and gas content; Step F: Based on the distribution field, calculate and output evaluation indicators characterizing the combined surface-underground gas extraction effect. The evaluation indicators include the proportion of areas meeting extraction standards and the spatial distribution of residual gas content.

2. The method according to claim 1, characterized in that, The triple hybrid loss function Represented as: , in: For data fidelity loss, it is used to minimize the difference between the predicted value and the measured value of the geostatistical-physical information dual-constraint neural network evaluation model at the sparse measured data points; The physical constraint loss term is used to minimize the residual of the predicted value of the geostatistical-physical information dual-constraint neural network evaluation model at a preset configuration point within the extraction area to the predetermined gas seepage partial differential equation. This is a geostatistical structure loss term used to minimize the difference between the experimental variogram calculated based on the sparse measured data and the model variogram calculated based on the predicted values ​​of the geostatistical-physical information dual-constraint neural network evaluation model. and These are the weighting coefficients for the physical constraint loss term and the geostatistical structure loss term.

3. The method according to claim 2, characterized in that, The Calculated in the following way: ， ， ， in, The residual of the partial differential equation is... The number of the preset configuration points. Porosity For gas density, For gas saturation, For time, For gas seepage velocity, For source and sink items, For absolute penetration rate, The relative permeability of the gas. For gas viscosity, For gas pressure, It is the acceleration due to gravity. For divergence operators, For gradient operators, For the residual In the The value of each preset configuration point, where i is the sequence number of the preset configuration point.

4. The method according to claim 2, characterized in that, The Calculated in the following way: ， ， ， in, Let be the experimental variability function. The model's mutation function is... Spatial lag distance, Lag distance The corresponding number of measured data point pairs For in position The measured gas pressure value, Lag distance The corresponding number of model sampling point pairs, The evaluation model of the geostatistical-physical information dual-constraint neural network at the sampling location The predicted gas pressure value, Lag distance used for comparison The total number.

5. The method according to claim 1, characterized in that, The surface-underground combined gas extraction system is implemented using an "L"-shaped horizontal well. The horizontal section of the well is no less than 650 meters long, and segmented fracturing and permeability enhancement are performed on the preferred target coal seams, which are no less than 10 layers.

6. The method according to claim 1, characterized in that, The surface-underground combined gas extraction system includes a surface horizontal well shaft and an underground connecting roadway or cross-layer borehole. The underground connecting roadway or cross-layer borehole achieves safe and precise docking with the surface horizontal well shaft with a positioning accuracy at the centimeter level.

7. The method according to claim 6, characterized in that, The combined gas extraction system adopts a combination of surface extraction and underground negative pressure continuous extraction. The underground negative pressure continuous extraction is connected to the underground gas extraction system through the underground connecting roadway or through-layer borehole.

8. The method according to claim 1, characterized in that, The sparse downhole gas parameter measured data in step B are obtained by collecting gas parameter test points from multiple cross-layer boreholes deployed within the target rock passage excavation footage and the comparison rock passage excavation footage.

9. The method according to claim 1, characterized in that, The evaluation index in step F also includes assessing the percentage increase in tunneling efficiency brought about by the combined gas extraction by comparing the average tunneling efficiency of the target tunnel with that of the comparison tunnel.

Images