Multi-source information fusion coal mine gas risk virtual deduction early warning method and system

Abstract

The present application provides a multi-source information fusion coal mine gas risk virtual deduction early warning method and system, belongs to the technical field of gas risk early warning. The method first collects geological, mining and environmental multi-source data, and after preprocessing and feature extraction, high-dimensional risk feature vectors are fused and constructed; then, the vector drives the virtual working face system containing mining, ventilation and gas emission agents to evolve, adopts the Monte Carlo tree search algorithm to prospectively deduce the future multi-step state, and calculates the risk entropy of each path; finally, based on the deduction result, a dynamic risk probability atlas is generated, and a Pareto optimal decision set considering safety, efficiency and economy is solved through multi-objective optimization. The present application realizes the transformation of gas risk from passive monitoring to active early warning, effectively solves the problem of early warning lag, and provides advanced and accurate scientific decision support for gas disaster prevention and control through risk visualization and spatial positioning.

Description

A Virtual Simulation and Early Warning Method and System for Coal Mine Gas Risk Based on Multi-Source Information Fusion Technical Field

[0001] This invention relates to the field of gas risk early warning technology, and in particular to a method and system for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion. Background Technology

[0002] Gas hazards have always been one of the most significant factors restricting safe production in coal mines. Gas in coal seams exists primarily in two states: free and adsorbed. These two states are in dynamic equilibrium within the coal seam, but their dominance directly determines the permeability of the coal seam and the ease of gas extraction and control. When free gas predominates, the coal seam is typically highly permeable, and the gas can be directly extracted through drainage wells. Conversely, when adsorbed gas predominates, it corresponds to a low-permeability coal seam, requiring pressure relief and permeability enhancement measures such as hydraulic fracturing and deep-hole blasting to achieve effective extraction. Accordingly, coal mines are classified into high-gas mines and low-gas mines based on the gas content found in geological surveys, with significant differences in the complexity and cost of their gas control projects.

[0003] During coal mining, the advancement of the working face disrupts the stress balance of the original coal seam, causing a large amount of adsorbed gas in the coal seam in front of and to the sides (goaf) to desorb and transform into free gas, which then flows into the mining space. The goaf, in particular, becomes a huge, hidden gas source due to the continuous leakage from residual coal pillars and broken coal bodies, exhibiting uneven gas concentrations and a tendency to accumulate locally. Furthermore, influenced by the ventilation flow characteristics of the mining area, gas can easily accumulate in the upper and lower corners of the working face, creating an explosive environment.

[0004] Currently, mine gas early warning mainly relies on a network of gas concentration sensors deployed in key locations such as roadways, working faces, and upper corners. This technology is essentially a "point-based, passive, and lagging" monitoring mode. It has the following inherent drawbacks:

[0005] Early warning lag: The sensor only triggers the alarm when gas emerges and spreads to its installation location. The time window between the gas emergence and the alarm is extremely short, leaving very limited time for emergency response, and it cannot achieve true early warning.

[0006] Ignoring the source mechanism of gas outburst: Existing technologies only monitor the result parameter of gas concentration, but fail to effectively integrate dynamic information such as geological conditions (adsorbed / free gas content, permeability) and mining activities (advance speed, roof pressure), and cannot perceive and predict the "source" process of gas outburst caused by the desorption of adsorbed gas due to mining-induced pressure relief.

[0007] Unable to predict local gas accumulation: For gas accumulation inside the goaf and in the upper and lower corners, the existing monitoring system based on fixed-point sensors is difficult to effectively cover and predict. It can only be discovered after the gas overflows into the roadway, resulting in poor risk predictability.

[0008] Information silos: Geological exploration data, mining plan data, and real-time monitoring data are usually independent of each other, lacking a platform that can integrate and intelligently analyze multi-source information, resulting in a lack of systematic and forward-looking data support for early warning decisions.

[0009] Therefore, there is an urgent need in this field for a gas risk early warning technology solution that can fundamentally overcome the above-mentioned defects. Summary of the Invention

[0010] The purpose of this invention is to provide a method and system for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion. By integrating geological, mining and environmental data, it drives the virtual working face to perform advanced risk simulation, generates dynamic risk maps and optimized decision sets, realizes the transformation from passive alarm to active early warning, thereby accurately locates risks, guides scientific prevention and control, and effectively ensures safe production in coal mines.

[0011] To achieve the above objectives, this invention provides a method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion, comprising the following steps:

[0012] Step S1: Synchronously collect static geological data, dynamic mining data and real-time environmental monitoring data. After preprocessing and feature extraction, the obtained geological features, mining features, environmental features and potential index features output by the embedded gas desorption potential prediction model are fused to construct a high-dimensional risk feature vector.

[0013] Step S2: Using the high-dimensional risk feature vector obtained in step S1 as the initial state, drive a multi-agent virtual working face system to perform simulation; the multi-agent virtual working face system includes a mining agent, a ventilation agent, and a gas emission agent; simulate the system state changes at multiple future time steps based on the Monte Carlo tree search algorithm, and calculate the terminal risk entropy of each simulation path;

[0014] Step S3: Based on all the deduced paths and their risk entropies obtained in Step S2, generate a dynamic risk probability map; construct a multi-objective optimization model with minimizing total risk, maximizing mining progress, and minimizing ventilation energy consumption as independent optimization objectives, and use a multi-objective evolutionary algorithm to solve for the Pareto optimal decision set.

[0015] Preferably, static geological data includes the gas adsorption constant of the coal seam. Value and original air permeability coefficient Dynamic mining data includes the hydraulic support working resistance sequence and the real-time advance speed of the coal mining machine; real-time environmental monitoring data includes the gas volume concentration in the return airway of the working face and the oxygen volume concentration in the upper corner.

[0016] Preferably, the preprocessing includes data cleaning, data denoising, and data standardization; wherein, the Kalman filter algorithm is used to denoise the hydraulic support working resistance sequence in the dynamic mining data; and the Z-score standardization method is used to process all feature data.

[0017] Preferably, feature extraction specifically involves directly reading the gas adsorption constant from static geological data. Value and original air permeability coefficient Geological characteristics; the variance of support working resistance and the acceleration of advance speed are calculated from dynamic mining data as mining dynamic characteristics; the oxygen concentration difference at the upper corner is calculated from environmental monitoring data as environmental characteristics.

[0018] Preferably, the embedded gas desorption potential prediction model calculates the potential index characteristics based on the following formula:

[0019] ;

[0020] in, Indicates the characteristics of the potential index. , All represent Langmuir adsorption constants. This represents the estimated gas pressure. Represents the desorption rate constant. Indicates a future time window, The effective extraction impact factor is represented by the following formula:

[0021] ;

[0022] in, Represents the empirical coefficient. Indicates the current stent resistance. This represents the average support resistance.

[0023] Preferably, the high-dimensional risk feature vector is generated by weighted concatenation. Represented as:

[0024] ;

[0025] in, The weighting coefficients representing the characteristics of the potential index. The weighting coefficients represent the geological feature vectors. Represents geological feature vectors; The weight coefficients represent the features of the sampling vector. This represents the sampling feature vector. The weight coefficients represent the environmental feature vectors. This represents the environmental feature vector.

[0026] Preferred multi-agent virtual work surface system:

[0027] The mining agent is used to update the spatial position of the working face according to the advance speed command.

[0028] A ventilation intelligent agent is used to solve for the velocity and pressure distribution of the airflow field based on updated spatial geometry parameters;

[0029] A gas emission agent is used to dynamically update the gas emission intensity based on high-dimensional risk feature vectors, pressure relief range, and wind speed field.

[0030] The preferred process for deduction and risk assessment is as follows:

[0031] Initialize the simulation state It is a high-dimensional risk feature vector;

[0032] In each future time step The three agents are based on the previous state. Parallel interaction, updating the current state ;

[0033] Monte Carlo tree search repeated inference Step 1: Generate multiple paths; the terminal risk entropy of each path is calculated using the following formula:

[0034] ;

[0035] in, Represents terminal risk entropy. Indicates a gas over-limit incident The probability, Indicates a gas over-limit incident The concentration exceeding the limit.

[0036] Preferably, the specific process of step S3 is as follows:

[0037] Discretize the working surface space into a grid. Future time is discretized into a sequence ;

[0038] Statistical analysis of each grid cell in all simulation paths At each time step Probability of gas exceeding the limit ;

[0039] Combining terminal risk entropy The dynamic risk probability map is generated using the following formula:

[0040] ;

[0041] in, Represents grid cells in a dynamic risk probability map At time step The risk probability value; Represents grid cells At time step Importance weights;

[0042] by As input, construct a multi-objective optimization problem:

[0043] Minimize the total risk objective function:

[0044] ;

[0045] in, Represents the total risk value. This indicates that for all grid cells and all time steps corresponding Perform double accumulation calculation;

[0046] Objective function to maximize mining progress:

[0047] ;

[0048] in, Indicates the overall mining progress. Indicates time step The advance speed of the coal mining machine;

[0049] Minimize ventilation energy consumption objective function:

[0050] ;

[0051] in, Indicates total ventilation energy consumption. Indicates time step The power of the ventilation fan;

[0052] The NSGA-II algorithm is used to solve the above multi-objective optimization problem, and the Pareto optimal solution set is obtained.

[0053] This invention also provides a virtual simulation and early warning system for coal mine gas risk based on multi-source information fusion, comprising:

[0054] The data acquisition module is used to simultaneously collect static geological data, dynamic mining data, and real-time environmental monitoring data;

[0055] The data processing module is used to preprocess and extract features from the collected data. It integrates the extracted geological features, mining features, and environmental features with the potential index features output by the embedded gas desorption potential prediction model to construct a high-dimensional risk feature vector.

[0056] The virtual simulation module is used to drive the multi-agent virtual working surface system to perform simulations with a high-dimensional risk feature vector as the initial state. It simulates the changes in the system state over multiple future time steps based on the Monte Carlo tree search algorithm and calculates the terminal risk entropy for each simulation path.

[0057] The risk assessment module is used to generate a dynamic risk probability map based on the simulation results, construct a multi-objective optimization model, and use a multi-objective evolutionary algorithm to solve for the Pareto optimal decision set.

[0058] The decision output module is used to output the sequence of agent action instructions corresponding to the optimal decision set.

[0059] Therefore, the present invention employs the above-mentioned multi-source information fusion method and system for virtual simulation and early warning of coal mine gas risk, and the beneficial technical effects are as follows:

[0060] (1) A fundamental shift from "delayed alarm" to "advanced early warning": This invention, by constructing a "gas desorption potential prediction model" and conducting "multi-step dynamic simulation," advances the entry point for early warning from monitoring gas concentration results to predicting the source potential of gas outbursts. The system can simulate the risk evolution over multiple future time steps, thereby issuing early warnings before gas accumulation or even outbursts, providing a valuable time window for taking control measures and greatly improving the mine's emergency response capabilities.

[0061] (2) Improved accuracy and reliability of early warning: This invention breaks through the limitations of traditional "point-based" monitoring. By integrating static geological data, dynamic mining data, and real-time environmental data, a high-dimensional risk feature vector reflecting coal seam gas occurrence, mining impact, and ventilation conditions is constructed. This method of deep integration of multi-source information enables the early warning model to perceive the state of the complex underground environment more comprehensively and accurately, significantly reducing the probability of false alarms and missed alarms.

[0062] (3) It realizes the visualization and spatial positioning of risks: By generating a "dynamic risk probability map", this invention can intuitively display the abstract gas risks in the three-dimensional space model of the mine in a visual way, and clearly identify high-risk areas (such as the deep part of the goaf and the upper corner of the working face) and their changing trends over time. This enables safety managers to "see" the risks and realize the transformation from passive response to proactive defense.

[0063] (4) Provides scientific and optimal decision support, balancing safety and efficiency: This invention is the first to construct the gas management problem as a multi-objective optimization problem. By solving the "Pareto optimal decision set", it provides decision-makers with a series of solutions that achieve the best balance among the three objectives of "safety, efficiency, and economy". This changes the blindness of previous decisions based solely on experience, and realizes scientific, quantitative, and optimal disaster prevention and control command. Under the premise of ensuring safety, it helps to optimize production arrangements and reduce operating costs. Attached Figure Description

[0064] Figure 1 is a flowchart of the virtual simulation and early warning method for coal mine gas risk based on multi-source information fusion of the present invention.

[0065] Figure 2 is a flowchart of multi-agent interaction;

[0066] Figure 3 is an architecture diagram of the coal mine gas risk virtual simulation and early warning system based on multi-source information fusion of the present invention. Detailed Implementation

[0067] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0068] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0069] Example 1

[0070] This embodiment provides a method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion. This method is implemented in a coal mine working face (600 meters deep, 3.5 meters thick coal seam, belonging to a high-gas mine). Its core process is shown in Figure 1, and specifically includes the following steps:

[0071] Step S1: Multi-source data fusion and construction of high-dimensional risk features.

[0072] This step aims to integrate scattered, multi-type data into a digital feature vector that can comprehensively characterize the current risk status of the working face.

[0073] 1.1 Data Acquisition.

[0074] The system obtains static geological data of the working face from the mine geological database, including the gas adsorption constant of the coal seam. Value and original air permeability coefficient Simultaneously, dynamic mining data is collected in real time through the underground industrial ring network, including the hydraulic support working resistance sequence recorded every 10 seconds and the real-time advance speed of the coal mining machine. Environmental monitoring data, including methane and oxygen volume concentrations, is also collected in real time through sensors deployed in the return airway and upper corner of the working face.

[0075] 1.2 Data preprocessing and feature extraction.

[0076] Preprocessing: First, the collected raw data was cleaned to remove abnormal missing values ​​caused by momentary communication interruptions, and then linear interpolation was used to complete the data. Next, to address the random fluctuations in the stent's working resistance sequence, a Kalman filter algorithm was used for smoothing and denoising to extract the true pressure trend. Finally, all feature data were processed using Z-score normalization to eliminate the influence of dimensions.

[0077] Feature extraction:

[0078] Geological characteristics: Adsorption constants were read directly from static data. Value and original air permeability coefficient .

[0079] Mining characteristics: Calculate the variance of the support working resistance over the most recent 10 sampling periods to characterize the stability of the roof pressure; at the same time, calculate the first derivative of the coal mining machine's advance speed as an acceleration characteristic.

[0080] Environmental characteristics: The difference between the oxygen concentration in the upper corner and the standard oxygen concentration (20.8%) in the intake airway is calculated as the eddy current retention characteristic of the ventilation dead corner.

[0081] 1.3 Potential Index Feature Calculation and Feature Fusion.

[0082] The above pre-processed geological and mining characteristics (such as current support resistance) =35MPa, average resistance =32MPa) Input to the embedded gas desorption potential prediction model. The model is calculated according to the following formula:

[0083] ;

[0084] in, Indicates the characteristics of the potential index. , All represent Langmuir adsorption constants. Parameters characterizing the maximum adsorption capacity of coal seams for methane Parameters characterizing the affinity of coal seams for gas adsorption. This represents the estimated gas pressure. Represents the desorption rate constant. Indicates a future time window, The effective mining impact factor, considering the influence of mining on coal seam permeability, is calculated using the following formula:

[0085] ;

[0086] in, This represents an empirical coefficient (obtained by fitting based on actual field conditions and historical data, used to correct the degree of influence of mining on air permeability; in this embodiment, it is set to 0.5). Indicates the current stent resistance. This represents the average support resistance.

[0087] The fusion process employs a weighted concatenation method to generate a high-dimensional risk feature vector. Represented as:

[0088] ;

[0089] in, The weighting coefficients representing the potential index characteristics are set according to the importance of the potential index characteristics to gas risk warning and are given the highest weight; in this embodiment, they are set to 0.5. The weighting coefficient of the geological feature vector is set according to the importance of the geological features to gas risk warning; in this embodiment, it is set to 0.2. Geological feature vectors (composed of features extracted from static geological data, such as adsorption constants) represent geological feature vectors. Value, original air permeability coefficient wait); The weighting coefficient of the mining feature vector is set according to the importance of the mining feature to the gas risk warning; in this embodiment, it is set to 0.2. It represents the mining feature vector (composed of features extracted from dynamic mining data, such as the variance of support working resistance, advance speed acceleration, etc.). The weight coefficient of the environmental feature vector is set according to the importance of the environmental features to the gas risk warning; in this embodiment, it is set to 0.1. This represents an environmental feature vector (composed of features extracted from environmental monitoring data, such as the oxygen concentration difference in the upper corner).

[0090] Step S2: Multi-step dynamic simulation and risk assessment of the virtual working surface.

[0091] This step uses the feature vector generated in step S1 to simulate the future state of the working face in virtual space and assess the risk.

[0092] 2.1 System Initialization: Initialize the high-dimensional risk feature vector... As the initial state of the multi-agent virtual working surface system .

[0093] 2.2 Multi-agent collaborative inference (as shown in Figure 2): The system sets the inference time step to 1 minute, and the total number of inference steps is... This is a 30-step process (i.e., predicting risk within the next 30 minutes). At each time step... Inside:

[0094] (1) The mining agent updates the spatial position of the working face according to the preset advance speed command.

[0095] This intelligent agent is essentially a rule-based position updater, which continuously maintains a core state variable: the three-dimensional spatial coordinates of the working surface. It is used to characterize the spatial position of the working surface in real time.

[0096] Input: Receive the expected pace of progress from the production scheduling instructions. and the current simulation time step .

[0097] Processing: Update the working surface position based on kinematic principles. The update formula is as follows:

[0098] ;

[0099] in, Indicates the updated working face position. This indicates the position of the working face before the update. This represents a predefined working face advance direction vector (such as a unit vector along the coal seam strike), used to determine the spatial orientation for position updates.

[0100] Output: The updated working face position It broadcasts to other agents within the system (especially providing key input to ventilation agents). Its core function is to dynamically change the geometric boundaries of the virtual working surface, providing a real-time spatial reference for subsequent ventilation network calculations and depressurization range determination.

[0101] (2) Based on the new working face geometric parameters, the ventilation intelligent agent re-solves the mine ventilation network and calculates the new wind speed field and wind pressure distribution.

[0102] The agent is a simplified computational fluid dynamics (CFD) model or a solver based on a wind resistance network. Internally, it maintains a complete set of mine ventilation network data structures, covering core parameters such as network nodes, roadway branches, branch wind resistance, and fan characteristic curves.

[0103] Input: Receive the updated working face position from the mining agent. Changes in the working face location will directly lead to dynamic adjustments in the ventilation network topology (such as the length of related roadways and node connectivity).

[0104] Solution: Based on the updated ventilation network topology, a steady-state ventilation equations solving algorithm (such as the Scott-Hinsley algorithm) is used for solution. The core is to solve for the following key parameters:

[0105] Air volume in each branch of the ventilation network ;

[0106] Pressure at each node in the ventilation network Finally, the three-dimensional wind speed field distribution of the entire working area is obtained through interpolation or field calculation. With three-dimensional wind pressure field distribution .

[0107] Output: The calculated three-dimensional wind speed field distribution With three-dimensional wind pressure field distribution The broadcast is sent to the system. Its core function is to simulate the flow state of mine airflow, providing a fluid dynamic environment for the subsequent gas dilution and migration process.

[0108] (3) The gas emission agent determines the gas emission location based on the new mining location (pressure relief range), wind speed field, and high-dimensional risk feature vector. The intensity of gas outburst in the goaf and coal face is dynamically updated.

[0109] The agent is a gas emission simulator based on the source term model. Its internal state variables are the gas content of the coal body in front of the working face and the discrete units in the goaf, which are used to characterize the material basis of gas generation.

[0110] Input: Receive three types of key inputs:

[0111] Initial high-dimensional risk feature vector (Specifically, the characteristics of the gas outburst potential index) );

[0112] The mining intelligent agent determines the coal body decompression range based on the working face location;

[0113] Three-dimensional wind speed field distribution provided by the ventilation intelligent agent .

[0114] Process: The dynamic calculation of gas distribution is completed in two steps:

[0115] Outflow calculation:

[0116] Decompression zone ahead of the working face: Near the location of the updated working face in the mining intelligence system, the gas emission intensity from the coal face is calculated based on the spatial extent of the decompression zone. :

[0117] ;

[0118] in, This indicates the gas permeability of the coal seam in the decompression zone. This represents the gas pressure gradient between the interior of the coal seam and the working face space. This indicates the area of ​​the coal face exposed at the working face;

[0119] Goaf: Based on the degree of mining impact (such as overlying strata movement angle) and the distribution of residual coal, simulate the gas release intensity of the residual coal in the goaf. :

[0120] ;

[0121] in, This indicates the desorption efficiency of gas from residual coal in the goaf. Indicates the density of residual coal. Indicates the effective volume of the goaf. Indicates the original gas content of the residual coal. This indicates the residual gas content in the coal seam.

[0122] Gas source term integration: finally obtaining the total gas source term .

[0123] Migration and diffusion simulation: Total gas source term Substituting the gas convection-diffusion control equations, the three-dimensional wind speed field distribution provided by the ventilation agent is analyzed. As the driving condition for convection, solving the equations yields the spatial distribution and evolution of gas, ultimately outputting a three-dimensional gas concentration field. .

[0124] Output: Update the three-dimensional gas concentration field of the entire virtual working face space. Its core function is to dynamically predict the generation, migration, and distribution of gas at the working face, providing data support for gas risk early warning.

[0125] 2.3 Multipath Simulation and Risk Entropy Calculation: The Monte Carlo Tree Search (MCTS) algorithm was used to repeat the above simulation process 500 times to simulate multiple possible paths under the influence of different random factors (such as small fluctuations in roof pressure). For each simulated path, the event of gas concentration exceeding the limit (>0.8%) in the terminal state (30 minutes later) was counted, and calculated according to the formula:

[0126] ;

[0127] Calculate the terminal risk entropy of this path.

[0128] in, Represents terminal risk entropy. Indicates a gas over-limit incident The probability, Indicates a gas over-limit incident The concentration exceeding the limit.

[0129] Step S3: Risk map generation and optimization decision.

[0130] This step transforms the simulation results into an intuitive risk map and actionable decision recommendations.

[0131] 3.1 Generate a dynamic risk probability map.

[0132] The working face and goaf area are discretized in three-dimensional space into a set of 1m×1m grid cells. Discretize the next 30 minutes into a time series at 1-minute intervals. .

[0133] In the statistics of all 500 simulation paths, each grid cell At each time step Calculate the probability of gas exceeding the limit based on the number of times the gas exceedance occurs. .

[0134] Combine the terminal risk entropy of each path, and consider the importance weights of key areas such as the upper and lower corners. (The weight of the upper corner is set to 1.5, and the weight of other areas is 1.0). Generate a dynamic risk probability map using the following formula (this map visually displays the risk level at different times and locations in the future in the form of a heat map):

[0135] ;

[0136] in, Represents grid cells in a dynamic risk probability map At time step The risk probability value; Represents grid cells At time step Importance weights.

[0137] 3.2 Multi-objective optimization to solve for the Pareto optimal decision set.

[0138] The generated risk map For the input, construct three independent objective functions:

[0139] Security objective: Minimize total risk. ;

[0140] Efficiency objective: Maximize overall mining progress. ;

[0141] Economic objective: Minimize total ventilation energy consumption. ;

[0142] in, This represents the total risk value (used to quantitatively assess the overall gas risk level of the entire working face over all future time steps). This indicates that for all grid cells and all time steps corresponding Perform double accumulation calculation. It represents the total mining progress (used to quantitatively assess the total mining advance distance over all future time steps). Indicates time step The speed of the coal mining machine's advance, This represents total ventilation energy consumption (used to quantitatively assess the total energy consumed by the ventilation system over all future time steps). Indicates time step The power of the ventilation fan.

[0143] The NSGA-II multi-objective optimization algorithm is employed, using the advance speed of the mining agent and the main fan frequency of the ventilation agent as decision variables. The algorithm ultimately outputs a Pareto optimal solution set, which contains multiple non-dominated solutions. Each solution corresponds to a sequence of action instructions for the next 30 minutes (e.g., "reduce the coal mining machine speed to 3.5 m / min for the first 10 minutes, and simultaneously increase the frequency of main fan No. 1 to 45 Hz"). This is existing technology and will not be elaborated further.

[0144] 3.3 Decision Output: The system recommends the Pareto optimal solution set to the mine dispatch center. Decision-makers can select the most suitable solution from the solution set based on the current production safety priorities, thereby achieving the optimal balance between safety and efficiency.

[0145] Example 2

[0146] As shown in Figure 3, the multi-source information fusion-based coal mine gas risk virtual simulation and early warning system includes:

[0147] The data acquisition module is used to simultaneously collect static geological data, dynamic mining data, and real-time environmental monitoring data;

[0148] The data processing module is used to preprocess and extract features from the collected data. It integrates the extracted geological features, mining features, and environmental features with the potential index features output by the embedded gas desorption potential prediction model to construct a high-dimensional risk feature vector.

[0149] The virtual simulation module is used to drive the multi-agent virtual working surface system to perform simulations with a high-dimensional risk feature vector as the initial state. It simulates the changes in the system state over multiple future time steps based on the Monte Carlo tree search algorithm and calculates the terminal risk entropy for each simulation path.

[0150] The risk assessment module is used to generate a dynamic risk probability map based on the simulation results, construct a multi-objective optimization model, and use a multi-objective evolutionary algorithm to solve for the Pareto optimal decision set.

[0151] The decision output module is used to output the sequence of agent action instructions corresponding to the optimal decision set.

[0152] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0153] Therefore, the present invention adopts the above-mentioned multi-source information fusion method and system for virtual simulation and early warning of coal mine gas risk, realizing the transformation of gas risk from passive monitoring to active early warning, effectively solving the problem of early warning lag, and providing advanced and accurate scientific decision support for gas disaster prevention and control through risk visualization and spatial positioning.

[0154] 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion, characterized in that, Includes the following steps: Step S1: Synchronously collect static geological data, dynamic mining data and real-time environmental monitoring data. After preprocessing and feature extraction, the obtained geological features, mining features, environmental features and potential index features output by the embedded gas desorption potential prediction model are fused to construct a high-dimensional risk feature vector. The embedded gas desorption potential prediction model calculates the potential index characteristics based on the following formula: ;in, Indicates the characteristics of the potential index. 、 All represent Langmuir adsorption constants. This represents the estimated gas pressure. Represents the desorption rate constant. Indicates a future time window, The effective extraction impact factor is represented by the following formula: ;in, Represents the empirical coefficient. Indicates the current stent resistance. The average support resistance is represented in step S2. The high-dimensional risk feature vector obtained in step S1 is used as the initial state to drive a multi-agent virtual working face system for simulation. The multi-agent virtual working face system includes a mining agent, a ventilation agent, and a gas emission agent. The system state changes over multiple future time steps are simulated using a Monte Carlo tree search algorithm, and the terminal risk entropy of each simulation path is calculated. In the multi-agent virtual working face system: the mining agent updates the working face spatial position according to the advance speed command; the ventilation agent solves for the velocity and pressure distribution of the airflow field based on the updated spatial geometric parameters; the gas emission agent dynamically updates the gas emission intensity based on the high-dimensional risk feature vector, pressure relief range, and wind speed field. Step S3: Based on all simulation paths and their risk entropies obtained in step S2, a dynamic risk probability map is generated. A multi-objective optimization model is constructed with minimizing total risk, maximizing mining progress, and minimizing ventilation energy consumption as independent optimization objectives, and a multi-objective evolutionary algorithm is used to solve for the Pareto optimal decision set.

2. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, Static geological data includes the gas adsorption constant of coal seams. Value and original air permeability coefficient Dynamic mining data includes the hydraulic support working resistance sequence and the real-time advance speed of the coal mining machine; real-time environmental monitoring data includes the gas volume concentration in the return airway of the working face and the oxygen volume concentration in the upper corner.

3. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, Preprocessing includes data cleaning, data denoising, and data standardization; among them, the Kalman filter algorithm is used to denoise the hydraulic support working resistance sequence in the dynamic mining data; and the Z-score standardization method is used to process all feature data.

4. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, Feature extraction specifically involves directly reading the gas adsorption constant from static geological data. Value and original air permeability coefficient As a geological feature; The working resistance variance and advance speed acceleration of the support are calculated from dynamic mining data as mining characteristics; The oxygen concentration difference at the upper corner is calculated from environmental monitoring data as an environmental characteristic.

5. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, The fusion process employs a weighted concatenation method to generate a high-dimensional risk feature vector. Represented as: ;in, The weighting coefficients representing the characteristics of the potential index. The weighting coefficients represent the geological feature vectors. Represents geological feature vectors; The weight coefficients represent the features of the sampling vector. This represents the sampling feature vector. The weight coefficients represent the environmental feature vectors. This represents the environmental feature vector.

6. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, The specific process of simulation and risk assessment is as follows: Initialize the simulation state For high-dimensional risk feature vectors; at each future time step The three agents are based on the previous state. Parallel interaction, updating the current state Monte Carlo tree search with repeated inferences. Step 1: Generate multiple paths; the terminal risk entropy of each path is calculated using the following formula: ;in, Represents terminal risk entropy. Indicates a gas over-limit incident The probability, Indicates a gas over-limit incident The concentration exceeding the limit.

7. The method for virtual simulation and early warning of coal mine gas risk based on multi-source information fusion according to claim 1, characterized in that, The specific process of step S3 is as follows: Discretize the working surface space into a grid. Future time is discretized into a sequence ； Statistical analysis of each grid cell in all simulation paths At each time step Probability of gas exceeding the limit ； Combining terminal risk entropy The dynamic risk probability map is generated using the following formula: ;in, Represents grid cells in a dynamic risk probability map At time step The risk probability value; Represents grid cells At time step Importance weights; As input, construct a multi-objective optimization problem: minimize the total risk objective function: ;in, Represents the total risk value. This indicates that for all grid cells and all time steps corresponding Perform double-accumulation calculation; maximize the mining progress objective function: ;in, Indicates the overall mining progress. Indicates time step The coal mining machine's advance speed; the objective function for minimizing ventilation energy consumption: ;in, Indicates total ventilation energy consumption. Indicates time step The power of the ventilation fan was determined; the NSGA-II algorithm was used to solve the above multi-objective optimization problem, and the Pareto optimal solution set was obtained.

8. A virtual simulation and early warning system for coal mine gas risk based on multi-source information fusion, characterized in that: The method for virtual simulation and early warning of coal mine gas risk, which integrates multi-source information as described in any one of claims 1-7, comprises: a data acquisition module for simultaneously acquiring static geological data, dynamic mining data, and real-time environmental monitoring data; a data processing module for preprocessing and feature extraction of the acquired data, fusing the extracted geological features, mining characteristics, environmental features, and potential index features output by the embedded gas desorption potential prediction model to construct a high-dimensional risk feature vector; a virtual simulation module for driving a multi-agent virtual working face system to perform simulations using the high-dimensional risk feature vector as the initial state, simulating system state changes over multiple future time steps based on the Monte Carlo tree search algorithm, and calculating the terminal risk entropy of each simulation path; a risk assessment module for generating a dynamic risk probability map based on the simulation results, constructing a multi-objective optimization model, and using a multi-objective evolutionary algorithm to solve for the Pareto optimal decision set; and a decision output module for outputting the agent action instruction sequence corresponding to the optimal decision set.

Images