Three-dimensional monitoring and early warning method for natural gas leakage in coal-gas overlapping area mine laneway

Abstract

The invention discloses a three-dimensional monitoring and early warning method for natural gas leakage in a coal-gas overlapping area mine laneway. A three-dimensional monitoring network covering a laneway three-dimensional space is constructed; collecting monitoring data of each monitoring point in the three-dimensional monitoring network, and associating spatial position information of the monitoring points; gas distribution and flow characteristics under different leakage situations are simulated, and a leakage characteristic database containing the corresponding relation of leakage source position-intensity-spatial concentration distribution is established; performing training and parameter optimization on the graph structure intelligent early warning model by using the leakage feature database; and inputting monitoring data acquired in real time into the trained intelligent early warning model, and outputting a leakage judgment result, a leakage source position and leakage intensity information. According to the method, early detection of natural gas leakage in the coal-gas overlapping area mine laneway can be realized, the source of the leaked natural gas in the natural gas well is accurately judged, and spatial positioning and strength evaluation are carried out on the leakage source, so that a scientific basis is provided for emergency disposal decision making.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mine gas disaster prevention and control, in particular to a kind of coal-gas overlap area mine roadway natural gas leakage stereoscopic monitoring and early warning method. BACKGROUND

[0002] Coal and natural gas resources overlap widely in Ordos and other regions. The depth of natural gas occurrence area of gas well is 2000-3000m, and the buried depth of coal mining area is less than 1000m. There is a difference between the depth of coal mining and the depth of natural gas occurrence, but the phenomenon of natural gas well crossing coal seam is common. Although measures such as setting up protective coal pillars have been taken, high-intensity and rapid mining activities in coal mines have significantly increased the risk of leakage of far-field natural gas wells, seriously threatening safety production. At present, the gas monitoring in coal mine mainly relies on single methane sensor on roadway wall, which cannot capture the spatial distribution characteristics of leaked gas on the cross section of roadway, and it is difficult to distinguish whether it is abnormal emission of coal seam gas or leakage of deep natural gas, and it is difficult to determine the gas source. When the leaked gas diffuses to the sensor position, it only alarms, which cannot realize early identification and early warning, and misses the best disposal opportunity. Therefore, there is an urgent need for a natural gas leakage monitoring and early warning technology in coal-gas overlap area mine roadway that can realize early identification and accurate positioning. SUMMARY

[0003] The present application provides a kind of coal-gas overlap area mine roadway natural gas leakage stereoscopic monitoring and early warning method to solve the problem that the traditional method cannot realize early identification and early warning.

[0004] In an embodiment of the present application, a coal-gas overlap area mine roadway natural gas leakage stereoscopic monitoring and early warning method is provided, which comprises: In the mine roadway section to be monitored, monitoring sections and monitoring points are arranged, and monitoring sensors are integrated and installed at each monitoring point to construct a stereoscopic monitoring network covering the three-dimensional space of the roadway; Collect the monitoring data of each monitoring point in the stereoscopic monitoring network, and associate the spatial position information of the monitoring point, and transmit it to the ground data processing center through the mine special transmission network; Based on the numerical model constructed by the roadway environment parameters and the physical parameters of coal and rock mass, the gas distribution and flow characteristics under different leakage scenarios are simulated, and a leakage characteristic database containing the corresponding relationship of "leakage source position-intensity-spatial concentration distribution" is established; Based on the topological characteristics of the stereoscopic monitoring network, a graph structure intelligent early warning model is constructed, and the leakage characteristic database is used to train and optimize the graph structure intelligent early warning model; The real-time collected monitoring data is input into the trained graph structure intelligent early warning model, and a leakage judgment result, a leakage source position and leakage intensity information are output, and a graded early warning is started according to the leakage intensity information; The early warning information and the leakage key parameters are pushed to the management terminal, and the control system is linked to perform emergency disposal related operations.

[0005] Further, in the mine roadway section to be monitored, monitoring sections and monitoring points are arranged, and monitoring sensors are integrated and installed at each monitoring point, and a three-dimensional monitoring network covering the three-dimensional space of the roadway is constructed, specifically including: Monitoring sections are arranged along the direction of the roadway at a predetermined interval, and monitoring points are arranged at the roof, floor, left side and right side of each monitoring section, and methane concentration sensors and wind speed and direction sensors are integrated and installed at each monitoring point.

[0006] Further, the monitoring data of each monitoring point in the three-dimensional monitoring network is collected, and the spatial position information of the monitoring point is associated, and is transmitted to the ground data processing center through a mine special transmission network, specifically including: The methane concentration data, wind speed data and wind direction data of all sensors are collected in real time or at regular intervals, the unique number of each sensor and its corresponding three-dimensional space coordinates are recorded, and all data are transmitted through a mine intrinsic safety type transmission network.

[0007] Further, a numerical model is constructed based on the roadway environment parameters and the coal and rock mass physical parameters, the gas distribution and flow characteristics under different leakage scenarios are simulated, and a leakage characteristic database containing the corresponding relationship of "leakage source position-intensity-space concentration distribution" is established, specifically including: The roadway geometric parameters, ventilation parameters and coal and rock mass permeability parameters are input, and the coal and rock mass permeability parameters are obtained through field coring experiments; A CFD numerical model of roadway ventilation and gas migration is established based on the input parameters, different simulation conditions of different preset leakage points and different leakage intensities are set, and the concentration distribution of methane gas in the roadway space and the change of flow velocity field under each condition are simulated; A leakage characteristic database containing the corresponding relationship of "leakage source position-intensity-space concentration distribution" is output.

[0008] Further, a graph structure intelligent early warning model is constructed based on the topological characteristics of the three-dimensional monitoring network, and the graph structure intelligent early warning model is trained and parameter optimized using the leakage characteristic database, specifically including: The GNN graph neural network is selected to construct the intelligent early warning model, including graph structure construction, GNN model architecture design, model training and hyperparameter optimization.

[0009] Further, the graph structure is constructed, specifically including: Obtain the monitoring point distribution data and three-dimensional coordinates of each monitoring point of the stereoscopic monitoring network; Model the entire stereoscopic monitoring network as a graph structure, with each monitoring point being considered as a node in the graph. The node features include real-time methane concentration, wind speed, wind direction, and the three-dimensional coordinates of the node. The edge connection relationship between nodes is defined based on the spatial distance of the monitoring points and the roadway topological structure. The distance threshold method or k-nearest neighbor method is used to establish the edge. The edge features include the distance between nodes and the relative azimuth angle. In the distance threshold method, the threshold is set to be less than 10 meters in the Euclidean distance between nodes. In the k-nearest neighbor method, each node is connected to the nearest 3-5 nodes. Obtain the graph structure data containing node features, edge features, and adjacency matrix.

[0010] Further, the GNN model architecture design specifically includes: Obtain the graph structure data and sample attributes of the leakage feature association database; Use a multi-layer graph convolutional network (GCN) or a graph attention network (GAT) as the core layer. The model input is the node features and adjacency matrix of the entire graph. The output layer is designed as a multi-task learning head, including a branch for regression of the leakage source location, a branch for regression of the leakage intensity, and a classification branch for determining whether a leakage has occurred. Obtain the GNN model architecture with multi-task learning capability.

[0011] Further, the model training specifically includes: Use the generated numerical simulation data as the training set. The input is the multi-parameter data of the monitoring points, and the label is the leakage source location and intensity. Use a loss function that combines mean squared error (MSE) and cross-entropy. MSE is used for regression tasks, and cross-entropy is used for classification tasks. Train using the gradient descent algorithm and introduce early stopping and dropout strategies to prevent overfitting. Obtain the preliminary trained intelligent early warning model.

[0012] Further, the hyperparameter optimization specifically includes: Input the preliminary trained intelligent early warning model and the hyperparameter optimization range; Use the particle swarm optimization algorithm to automatically optimize the key hyperparameters of the intelligent early warning model. The hyperparameters to be optimized include: Number of graph convolution layers: optimization range 2-5 layers; Hidden layer dimension: optimization range 64-256; Learning rate: optimization range 0.001-0.01; Number of attention heads: for GAT model; Use the prediction accuracy on the validation set as the fitness function, iteratively search the hyperparameter space, and determine the optimal hyperparameter combination to obtain the hyperparameter optimized intelligent early warning model.

[0013] Furthermore, the real-time collected monitoring data is input into the trained graph structure intelligent early warning model, which outputs leakage judgment results, leakage source location, and leakage intensity information. Based on the leakage intensity information, a tiered early warning system is activated, specifically including: The real-time collected monitoring data is input into the trained intelligent early warning model. The model outputs the leakage occurrence judgment result, the possible location coordinates of the leakage source and the leakage intensity estimate. Based on the leakage intensity estimate, a graded early warning threshold is set and a graded early warning is activated. The graded early warning includes blue, yellow, orange and red warnings.

[0014] This application provides a three-dimensional monitoring and early warning method for natural gas leaks in mine roadways within coal-gas overlap zones, offering the following advantages: By strategically placing monitoring points in three dimensions, it effectively captures the diffusion path and distribution characteristics of leaking gas, providing a rich data foundation for accurate identification and location of leak sources. Combined with numerical simulation and artificial intelligence algorithms, it can keenly capture abnormal signals from complex background environments and deduce the location and intensity of the leak source, transforming passive alarms into proactive early warnings. Analyzing differences in spatial concentration distribution patterns helps distinguish between gas outbursts originating from the coal seam itself and leaks from deep natural gas wells, guiding targeted measures. A complete technical system is formed, encompassing monitoring network construction, data analysis, model building, and early warning dissemination, demonstrating a high degree of practicality. This provides a novel solution for the precise location and prediction of natural gas leaks in mine roadways within coal-gas overlap zones. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones, provided as an embodiment of the present invention; Figure 2 A flowchart illustrating the implementation of a three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones, as provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of sensor installation in a three-dimensional monitoring and early warning method for natural gas leakage in a coal-gas overlap zone mine roadway, provided as an embodiment of the present invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of this application. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to this application are not shown or described in the specification. This is to avoid obscuring the core parts of this application with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0017] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0018] The first embodiment of this invention provides a three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap areas. The following is a combination of... Figure 1 Please provide a detailed explanation.

[0019] like Figure 1 As shown, in step S100, monitoring sections and monitoring points are set up in the mine roadway section to be monitored, and monitoring sensors are integrated and installed at each monitoring point to construct a three-dimensional monitoring network covering the three-dimensional space of the roadway.

[0020] Specifically, the three-dimensional monitoring network is constructed as follows: In the mine roadway sections that require monitoring, a monitoring section is set up at regular intervals (e.g., 20-30 meters) along the roadway direction. On each monitoring section, four monitoring points are arranged on the roadway roof, floor, left side, and right side. Each monitoring point integrates a methane concentration sensor and a wind speed and direction sensor, forming a three-dimensional monitoring network covering the three-dimensional space of the roadway.

[0021] like Figure 1 As shown, in step S200, monitoring data from each monitoring point in the three-dimensional monitoring network is collected, and the spatial location information of the monitoring points is associated with them. The data is then transmitted to the ground data processing center through a dedicated mining transmission network.

[0022] Specifically, multi-parameter data is collected and transmitted synchronously: methane concentration data, wind speed data, and wind direction data from all sensors are collected in real time or at regular intervals, and the unique number of each sensor and its corresponding three-dimensional spatial coordinates (X, Y, Z) are recorded. All data is transmitted to the ground data center through a mine-use intrinsically safe transmission network.

[0023] like Figure 1 As shown, in step S300, a numerical model is constructed based on the roadway environmental parameters and the physical parameters of the coal and rock mass to simulate the gas distribution and flow characteristics under different leakage scenarios, and a leakage characteristic database containing the correspondence between "leakage source location-intensity-spatial concentration distribution" is established.

[0024] Specifically, a numerical simulation and feature database for leakage scenarios is established: based on roadway geometric parameters, ventilation parameters, and coal and rock permeability parameters obtained from on-site core sampling experiments, a numerical model of roadway ventilation and gas transport is established. By simulating the concentration distribution and velocity field changes of methane gas in the roadway space at different preset leakage points (such as surrounding rock at different locations in the roadway) and under different leakage intensities, a feature database containing the correspondence between "leakage source location - intensity - spatial concentration distribution" is generated.

[0025] like Figure 1 As shown, in step S400, a graph structure intelligent early warning model is constructed based on the topological characteristics of the three-dimensional monitoring network, and the graph structure intelligent early warning model is trained and its parameters are optimized using a leakage feature database.

[0026] Specifically, the intelligent early warning model is constructed and trained using a Graph Neural Network (GNN) as the base model because it excels at handling non-Euclidean spatial data (such as sensor networks) and can effectively capture the spatial dependencies and topological structures between monitoring points. Details are as follows: Graph Structure Construction: The entire monitoring network is modeled as a graph structure, where each monitoring point is considered a node. Node features include real-time methane concentration, wind speed, wind direction, and the node's three-dimensional coordinates (X, Y, Z). Edges represent the connections between nodes and are defined based on the spatial distance between monitoring points and the roadway topology: Edges are established using a distance threshold method (e.g., the Euclidean distance between nodes is less than 10 meters) or the k-nearest neighbor method (each node connects to its 3-5 nearest neighbors). Edge features may include the distance between nodes, relative azimuth angle, etc.

[0027] The GNN model architecture employs either a multi-layer Graph Convolutional Network (GCN) or Graph Attention Network (GAT) as its core layers. GCN updates node representations by aggregating neighbor node information, thereby capturing local spatial patterns; GAT introduces an attention mechanism to dynamically learn the importance weights of different neighbor nodes, handling complex diffusion scenarios more flexibly. The model input consists of node features and an adjacency matrix for the entire graph. The output layer is designed as a multi-task learning head: one branch regresses the location of the leakage source (3D coordinates), another branch regresses the leakage intensity (e.g., leakage flow rate L / s), and a classification branch can be added to determine whether a leakage has occurred.

[0028] Training process: Numerical simulation data generated by the S300 was used as the training set, with inputs consisting of multi-parameter data (concentration field, wind speed field) from monitoring points, and labels representing the location and intensity of the leakage source. The loss function combined with mean squared error (MSE) was used for regression tasks, and cross-entropy for classification tasks. Gradient descent algorithms (such as the Adam optimizer) were employed during training, and early stopping and dropout strategies were introduced to prevent overfitting.

[0029] Hyperparameter optimization: Particle swarm optimization (PSO) is used to automatically optimize key hyperparameters of the GNN, including the number of graph convolutional layers (2-5 layers), hidden layer dimension (64-256), learning rate (0.001-0.01), and number of attention heads (for GAT). PSO iteratively searches the hyperparameter space, using the prediction accuracy on the validation set (such as position error and intensity error) as the fitness function, and finally determines the optimal combination of hyperparameters to improve the model's generalization ability.

[0030] Model validation: In addition to simulated data, historical real-world leak data (if available) can be used for fine-tuning to enhance the model's robustness in real-world scenarios.

[0031] like Figure 1 As shown, in step S500, the real-time collected monitoring data is input into the trained graph structure intelligent early warning model, and the leakage judgment result, leakage source location and leakage intensity information are output. Based on the leakage intensity information, a graded early warning is initiated.

[0032] Specifically, real-time monitoring and intelligent early warning: The three-dimensional monitoring data collected in real time in the S200 is input into the trained intelligent early warning model. The model output includes: ① a judgment on whether a natural gas leak has occurred; ② the possible location coordinates of the leak source; ③ an estimated value of the leak intensity. Based on the estimated leak intensity, a threshold is set to activate a graded early warning system (such as blue, yellow, orange, and red warnings).

[0033] like Figure 1As shown, in step S600, the early warning information and key leakage parameters are pushed to the management terminal, and the control system is linked to perform relevant emergency response operations.

[0034] Specifically, the system links early warning information dissemination with emergency response. It automatically pushes early warning levels and leak source location information to relevant management personnel and can also link with the control system to prompt adjustments to ventilation plans or the initiation of emergency evacuation procedures.

[0035] Application example: Taking a coal mine return airway with a natural gas well within a certain mining rights area as an example, such as Figure 2 As shown, the specific steps are as follows: Step 1: In a 500-meter-long tunnel section potentially affected by a natural gas leak, 25 monitoring sections are deployed at a density of one section every 20 meters. For example... Figure 3 As shown, five multi-functional sensor nodes are installed on the top plate, bottom plate, two sides and center of each section.

[0036] Step 2: The system collects methane concentration and wind speed and direction data from all 125 monitoring points every minute, and runs continuously.

[0037] Step 3: Using the permeability parameters of the coal and rock mass obtained through experiments in this project, a CFD numerical model of the roadway is established. The model simulates the spatial distribution of methane concentration within the roadway when the leak point is located at the roadway floor (simulating a natural gas well leaking from bottom to top) and the leakage intensities are 0.1 L / s, 0.5 L / s, and 1.0 L / s, respectively, thus constructing a training sample library.

[0038] Step 4: Select a Graph Neural Network (GNN) model, treating each monitoring point as a node in a graph. Node features include concentration, wind speed, and coordinates, while edge features can consider the distance between nodes. Train the model using the simulated data generated in Step 3, and optimize the GNN's hyperparameters using the PSO algorithm.

[0039] Step 5: System Operation. One day, based on real-time data, the model determined a suspected leak in the floor slab near a section 300 meters from the tunnel entrance, with a yellow warning level (estimated leak intensity 0.3 L / s). The platform immediately issued an alarm and displayed the location of the leak source.

[0040] Step Six: Inspection personnel went to the designated area based on the location information and indeed found gas escaping from a crack in the base plate, confirming the accuracy of the early warning. Immediately, measures such as increased ventilation and preparation of sealing materials were taken, effectively preventing the leak from escalating.

[0041] Corresponding to the above-disclosed method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap areas, this invention also discloses a three-dimensional monitoring and early warning system for natural gas leakage in mine roadways in coal-gas overlap areas, which specifically includes: The three-dimensional monitoring network construction module is used to deploy monitoring sections and monitoring points in the mine roadway section to be monitored, and to integrate and install monitoring sensors at each monitoring point to construct a three-dimensional monitoring network covering the three-dimensional space of the roadway. The multi-parameter data synchronous acquisition and transmission module is used to acquire monitoring data from each monitoring point in the three-dimensional monitoring network, associate the spatial location information of the monitoring points, and transmit the data to the ground data processing center through a dedicated mining transmission network. The module for numerical simulation of leakage scenarios and construction of a feature database is used to build numerical models based on roadway environmental parameters and coal and rock physical parameters, simulate the gas distribution and flow characteristics under different leakage scenarios, and establish a leakage feature database containing the correspondence between "leakage source location-intensity-spatial concentration distribution". The intelligent early warning model construction and training module is used to construct a graph structure intelligent early warning model based on the topological characteristics of a three-dimensional monitoring network, and to train and optimize the parameters of the graph structure intelligent early warning model using a leakage feature database. The real-time monitoring and intelligent early warning module is used to input the real-time collected monitoring data into the trained graph structure intelligent early warning model, output the leakage judgment result, leakage source location and leakage intensity information, and initiate graded early warning based on the leakage intensity information.

[0042] The early warning information release and emergency response linkage is used to push early warning information and key leakage parameters to the management terminal and link the control system to perform relevant emergency response operations.

[0043] It should be noted that for a detailed description of the three-dimensional monitoring and early warning system for natural gas leakage in mine roadways in coal-gas overlap areas provided by the embodiments of the present invention, please refer to the relevant description of the three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap areas provided by the embodiments of this application, which will not be repeated here.

[0044] In addition, embodiments of the present invention also provide an electronic device, the device comprising: a processor and a memory; the memory being used to store one or more program instructions; the processor being used to execute one or more program instructions to perform the steps of a three-dimensional monitoring and early warning method for natural gas leakage in a coal-gas overlap zone mine roadway as described in any of the preceding embodiments.

[0045] It should be noted that for a detailed description of the electronic device provided in the embodiments of the present invention, please refer to the relevant description of the three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap areas provided in the embodiments of this application, which will not be repeated here.

[0046] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones as described in any of the preceding claims.

[0047] It should be noted that for a detailed description of the computer-readable storage medium provided in the embodiments of the present invention, please refer to the relevant description of the three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap areas provided in the embodiments of this application, which will not be repeated here.

[0048] Those skilled in the art will understand that all or part of the functions of the various methods in the above embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, which may include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to achieve the above functions. For example, the program can be stored in the memory of a device, and when the program in the memory is executed by the processor, all or part of the above functions can be achieved. In addition, when all or part of the functions in the above embodiments are implemented by computer programs, the program can also be stored in a server, another computer, disk, optical disk, flash drive, or external hard drive, etc., and can be downloaded or copied to the memory of a local device, or the system of the local device can be updated. When the program in the memory is executed by the processor, all or part of the functions in the above embodiments can be achieved.

[0049] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones, characterized in that, The method includes: In the section of mine roadway to be monitored, monitoring sections and monitoring points are set up, and monitoring sensors are integrated and installed at each monitoring point to construct a three-dimensional monitoring network covering the three-dimensional space of the roadway. The monitoring data from each monitoring point in the three-dimensional monitoring network is collected and correlated with the spatial location information of the monitoring points, and then transmitted to the ground data processing center through a dedicated mining transmission network. Numerical models were constructed based on roadway environmental parameters and coal and rock mass physical parameters to simulate gas distribution and flow characteristics under different leakage scenarios, and a leakage characteristic database containing the correspondence between "leakage source location-intensity-spatial concentration distribution" was established. A graph-structured intelligent early warning model is constructed based on the topological characteristics of a three-dimensional monitoring network. The model is then trained and its parameters are optimized using a leakage feature database. The real-time collected monitoring data is input into the trained graph structure intelligent early warning model, which outputs the leakage judgment result, leakage source location and leakage intensity information, and initiates graded early warning based on the leakage intensity information. The system pushes early warning information and key leakage parameters to the management terminal and links the control system to perform relevant emergency response operations.

2. The method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 1, characterized in that, In the mine roadway section to be monitored, monitoring sections and monitoring points are set up, and monitoring sensors are integrated and installed at each monitoring point to construct a three-dimensional monitoring network covering the three-dimensional space of the roadway, specifically including: Monitoring sections are set up at preset intervals along the tunnel direction. Monitoring points are set up on the tunnel roof, floor, left side and right side of each monitoring section. A methane concentration sensor and a wind speed and direction sensor are integrated at each monitoring point.

3. The method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 1, characterized in that, The monitoring data from each monitoring point in the three-dimensional monitoring network is collected, and the spatial location information of the monitoring points is correlated. This data is then transmitted to the ground data processing center via a dedicated mining transmission network. Specifically, this includes: Data on methane concentration, wind speed, and wind direction from all sensors are collected in real-time or timed mode. The unique number of each sensor and its corresponding three-dimensional spatial coordinates are recorded. All data is transmitted through an intrinsically safe transmission network for mining.

4. The method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 1, characterized in that, Numerical models were constructed based on roadway environmental parameters and coal and rock mass physical parameters to simulate gas distribution and flow characteristics under different leakage scenarios. A leakage characteristic database containing the correspondence between "leakage source location - intensity - spatial concentration distribution" was established, specifically including: Input the roadway geometric parameters, ventilation parameters, and coal and rock mass permeability parameters, wherein the coal and rock mass permeability parameters are obtained through on-site core sampling experiments; A CFD numerical model of roadway ventilation and gas transport was established based on the input parameters. Simulation conditions with different preset leak points and different leak intensities were set to simulate the concentration distribution and velocity field changes of methane gas in the roadway space under each condition. The output contains a leak feature database that includes the correspondence between "leak source location - intensity - spatial concentration distribution".

5. The method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 1, characterized in that, A graph-structured intelligent early warning model is constructed based on the topological characteristics of a three-dimensional monitoring network. The model is then trained and its parameters optimized using a leakage feature database. Specifically, this includes: A graph neural network (GNN) was selected to construct an intelligent early warning model, including graph structure construction, GNN model architecture design, model training, and hyperparameter optimization.

6. A three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 5, characterized in that, Graph structure construction, specifically including: Obtain the distribution data of monitoring points in the three-dimensional monitoring network and the three-dimensional coordinates of each monitoring point; The entire three-dimensional monitoring network is modeled as a graph structure, with each monitoring point considered as a node in the graph. Node features include real-time methane concentration, wind speed, wind direction, and the node's three-dimensional coordinates. The edge connection relationship between nodes is defined based on the spatial distance between monitoring points and the roadway topology. Edges are established using either the distance threshold method or the k-nearest neighbor method, with edge features including the distance between nodes and their relative azimuth angle. The threshold for the distance threshold method is set to a Euclidean distance of less than 10 meters between nodes, while in the k-nearest neighbor method, each node connects to its 3-5 nearest neighbors. Obtain graph structure data containing node features, edge features, and adjacency matrix.

7. A three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 5, characterized in that, GNN model architecture design, specifically including: Obtain sample attributes from the database that correlate graph structure data and leakage characteristics; The core layer is a multi-layer graph convolutional network (GCN) or a graph attention network (GAT). The model input is the node features and adjacency matrix of the entire graph. The output layer is designed as a multi-task learning head, including a branch for regressing the location of the leakage source, a branch for regressing the leakage intensity, and a classification branch for determining whether a leakage has occurred. A GNN model architecture with multi-task learning capabilities was obtained.

8. A three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 5, characterized in that, Model training specifically includes: The generated numerical simulation data is used as the training set, and the input is the multi-parameter data of the monitoring points. The labels are the location and intensity of the leakage source. A loss function combining mean squared error (MSE) and cross-entropy is adopted, where MSE is used for regression tasks and cross-entropy is used for classification tasks. The model is trained using gradient descent algorithm, and early stopping and dropping strategies are introduced to prevent overfitting. A preliminary intelligent early warning model is obtained.

9. A three-dimensional monitoring and early warning method for natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 8, characterized in that, Hyperparameter optimization, specifically including: Input the preliminarily trained intelligent early warning model and the hyperparameter optimization range; The particle swarm optimization algorithm is used to automatically optimize the key hyperparameters of the intelligent early warning model. The hyperparameters to be optimized include: Number of graph convolutional layers: Optimization range 2-5 layers; Hidden layer dimensions: Optimization range 64-256; Learning rate: Optimization range 0.001-0.01; Attention head count: for GAT models; Using the prediction accuracy on the training set as the fitness function, the hyperparameter space is iteratively searched and the optimal combination of hyperparameters is determined to obtain the intelligent early warning model with optimized hyperparameters.

10. A method for three-dimensional monitoring and early warning of natural gas leakage in mine roadways in coal-gas overlap zones as described in claim 1, characterized in that, The real-time collected monitoring data is input into the trained graph structure intelligent early warning model, which outputs leakage judgment results, leakage source location and leakage intensity information, and initiates graded early warning based on the leakage intensity information, specifically including: The real-time collected monitoring data is input into the trained intelligent early warning model. The model outputs the leakage occurrence judgment result, the possible location coordinates of the leakage source and the leakage intensity estimate. Based on the leakage intensity estimate, a graded early warning threshold is set and a graded early warning is activated. The graded early warning includes blue, yellow, orange and red warnings.

Images