A coal seam mining face gas emission amount prediction method and system based on a digital twin model

Abstract

The application discloses a coal seam mining face gas emission amount prediction method and system based on a digital twin model, relates to the gas prevention and control field of a mining face, and comprises the following steps: acquiring the gas concentration of a historical moment of a coal seam mining face, and performing delay processing to obtain processed historical gas concentration; constructing a gas emission digital twin prediction model based on the nonlinear relationship among multi-source data of the historical moment, the processed historical gas concentration and the gas emission amount predicted by a trained double-layer LSTM neural network; collecting real-time multi-source data and real-time gas concentration of a coal mining environment, and performing delay processing on the real-time gas concentration; and based on the real-time multi-source data and the processed real-time gas concentration, the gas emission digital twin prediction model is used to predict the future gas emission amount, so that the gas emission amount at a future moment is obtained. The application improves the prediction accuracy of the gas emission amount.

Description

Technical Field

[0001] This application relates to the field of gas control in longwall mining faces, and in particular to a method and system for predicting gas emission from longwall mining faces based on a digital twin model. Background Technology

[0002] Gas outburst at longwall mining faces has always been a core and challenging aspect of gas disaster prevention and control in coal mining. The high stress, high gas content, high gas pressure, and low permeability of deep coal seams result in gas outburst patterns at longwall mining faces exhibiting significant characteristics such as multi-factor coupling and delayed response. This greatly increases the risk of gas exceeding limits, seriously threatening the safe and efficient mining of coal.

[0003] Accurate prediction of gas emission is a core prerequisite for achieving a balance between coal mine gas safety control and efficient coal mine production. Existing technologies mostly rely on traditional methods based on empirical formulas and statistical analysis of field measurement data. These methods are ill-suited to addressing the multi-factor coupling characteristics and lag effects of coal seams, resulting in low accuracy in gas emission prediction. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for predicting gas emission in coal seam longwall mining faces based on a digital twin model, which can improve the prediction accuracy of gas emission.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for predicting gas emission in coal seam longwall faces based on a digital twin model, including: The historical gas concentration at the coal seam longwall face is obtained and then delayed to obtain the processed historical gas concentration. Using multi-source historical data and processed historical gas concentration as input, and historical gas emission as label, a two-layer LSTM neural network is trained to obtain the trained two-layer LSTM neural network. A digital twin prediction model for gas outbursts is constructed based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outbursts predicted by a trained two-layer LSTM neural network. Collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and perform delay processing on the real-time gas concentration; Based on real-time multi-source data and processed real-time gas concentration, the gas emission digital twin prediction model is used to predict the future gas emission amount, thus obtaining the gas emission amount at future times.

[0006] Secondly, this application provides a gas emission prediction system for coal seam longwall faces based on a digital twin model, including: The data processing module is used to obtain the historical gas concentration of the coal seam mining face and perform delay processing to obtain the processed historical gas concentration. The training module is used to train a two-layer LSTM neural network with multi-source data of historical moments and processed historical gas concentration as input and historical gas emission as label, so as to obtain the trained two-layer LSTM neural network. The model building module is used to construct a digital twin prediction model for gas outburst based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outburst volume predicted by a trained two-layer LSTM neural network. The data acquisition module is used to collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and to perform delay processing on the real-time gas concentration. The gas prediction module is used to predict the future gas emission amount based on real-time multi-source data and processed real-time gas concentration, using the gas emission digital twin prediction model, to obtain the gas emission amount at future times.

[0007] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method and system for predicting gas emission in coal seam longwall mining faces based on a digital twin model. The method utilizes multi-source data and the nonlinear relationship between processed gas concentration and gas emission to construct a digital twin prediction model for gas emission. It considers the coupling characteristics of multiple factors in the coal seam. This process involves delaying the processing of the gas concentration used to account for the hysteresis of the gas response. Based on this, the digital twin prediction model for gas emission is used to predict future gas emission, significantly improving the prediction accuracy compared to traditional methods. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is an application environment diagram of a method for predicting gas emission from a coal seam longwall face based on a digital twin model, as described in one embodiment of this application.

[0010] Figure 2 This is a flowchart illustrating a method for predicting gas emission from a coal seam longwall face based on a digital twin model, provided as an embodiment of this application.

[0011] Figure 3This is a detailed flowchart illustrating a method for predicting gas emission from a coal seam longwall face based on a digital twin model, provided as an embodiment of this application.

[0012] Figure 4 This is a schematic diagram showing the arrangement of a gas sensor, a wind speed sensor, and a temperature sensor according to an embodiment of this application.

[0013] Figure 5 This is a schematic diagram illustrating the construction process of a digital twin prediction model for gas outbursts, provided as an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] To make the objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] The method for predicting gas emission in coal seam longwall faces based on digital twin models provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a mining communication network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send real-time multi-source data and real-time gas concentration of the coal mining working environment to server 102. Server 102 receives the real-time multi-source data and the processed real-time gas concentration, and performs delayed processing on the real-time gas concentration. Based on the real-time multi-source data and the processed real-time gas concentration, the server uses the gas emission digital twin prediction model to predict the future gas emission amount, obtaining the gas emission amount at the future time. Server 102 can feed back the obtained gas emission amount to terminal 101. In addition, in some embodiments, the method for predicting gas emission from coal seam mining faces based on digital twin models can also be implemented by either server 102 or terminal 101. For example, terminal 101 can directly predict gas emission from real-time multi-source data and real-time gas concentration to be processed, or server 102 can obtain real-time multi-source data and real-time gas concentration from the data storage system and predict gas emission.

[0017] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, coal mining machines, underground monitoring equipment, and portable wearable devices. Underground monitoring equipment can include gas sensors, wind speed sensors, temperature sensors, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, portable data acquisition devices, etc. The server 102 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0018] In one exemplary embodiment, such as Figure 2 and Figure 3 As shown, a method for predicting gas emission in coal seam longwall faces based on a digital twin model is provided, including the following steps 201 to 205. Wherein: Step 201: Obtain the gas concentration at historical moments of the coal seam longwall face and perform delay processing to obtain the processed historical gas concentration.

[0019] Step 202: Using multi-source data from historical moments and processed historical gas concentrations as inputs, and historical gas outburst volume as labels, train a two-layer LSTM neural network to obtain the trained two-layer LSTM neural network.

[0020] Step 203: Construct a digital twin prediction model for gas outburst based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outburst volume predicted by a trained two-layer LSTM neural network.

[0021] Step 204: Collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and perform delay processing on the real-time gas concentration.

[0022] Step 205: Based on real-time multi-source data and processed real-time gas concentration, the gas emission digital twin prediction model is used to predict the future gas emission amount to obtain the gas emission amount at future time.

[0023] Implementing steps 201 to 205 above can improve the accuracy of gas outburst prediction.

[0024] In an exemplary embodiment, step 201 involves obtaining the lag time by monitoring the mining operations at historical moments and adjusting the time offset with the time of sudden changes in gas concentration. t, The gas concentration at historical moments is delayed by using the coordinate delay method, with t = t1 - t2, to obtain the processed gas concentration. .

[0025] In one exemplary embodiment, step 202 specifically includes: Time-series data on coal mining machine travel speed (x1), coal cutting depth (x2), coal seam thickness (h), roadway cross-sectional area (S), gas concentration (c), roadway wind speed (V), and roadway temperature (K) are collected, sorted by time nodes, and maintained at a consistent step size to form raw data. Multi-source data and gas concentrations for historical moments are then extracted from the raw data.

[0026] The gas concentration at historical moments is subjected to the above-described delay processing to obtain the processed gas concentration. The gas emission rate was calculated using the treated gas concentration.

[0027] The formula for calculating the actual gas emission rate is as follows: ; Where Q is the gas emission rate, m 3 / min; V is the wind speed in the tunnel, m / s; S is the cross-sectional area of ​​the tunnel, m². 2 c represents the gas concentration, %.

[0028] A time-series dataset was established using multi-source data from historical moments and processed gas concentrations. The time-series dataset is labeled with historical gas emission volume. The multi-source data includes coal mining machine travel speed, coal cutting depth, coal seam thickness, roadway cross-sectional area, roadway wind speed, and roadway temperature.

[0029] Specifically, after the longwall face of the mine is completed, gas sensors, wind speed sensors, and temperature sensors are installed in the longwall face, intake airway, and return airway to monitor the gas concentration c, wind speed V, and temperature K in the airway, respectively. Figure 4 As shown, the sensor locations are as follows: The gas sensors are arranged as follows: gas sensor T0 is placed at the corner of the return airway; gas sensor T1 is placed 15m away from gas sensor T0 in the return airway; gas sensor T2 is placed at the end of the return airway; between gas sensors T1 and T2, a gas sensor is placed every 50-300m as needed according to the working conditions; gas sensors are placed 15m away from the entrance of the intake airway and every 200-500m in the intake airway as needed according to the working conditions; and gas sensors are placed every 30-100m in the longwall face as needed according to the working conditions.

[0030] The wind speed sensors are arranged as follows: one wind speed sensor is placed 15m from the entrance of the intake airway and 10m from the coal face of the working face; one wind speed sensor is placed in the middle of the longwall face; one wind speed sensor is placed 15m from the corner of the return airway; and one wind speed sensor is also placed in the middle of the return airway.

[0031] The temperature sensors are arranged as follows: one temperature sensor is placed 15m from the entrance of the intake airway, one temperature sensor is placed in the middle of the longwall face, one temperature sensor is placed at the corner of the return airway, and one temperature sensor is placed 15m from the corner of the return airway.

[0032] Based on the collected data such as gas concentration, wind speed and temperature in the roadway, the data is analyzed using interpolation methods to construct a spatial dynamic distribution cloud map of gas concentration in the longwall face and roadway, providing a data foundation for subsequent model verification.

[0033] Partitioning of time series datasets: The time series dataset was divided in a 7:2:1 ratio: 70% was selected as the training set for learning and training the two-layer LSTM neural network; 20% was selected as the validation set for optimizing the parameters of the two-layer LSTM neural network; and 10% was selected as the test set for evaluating the prediction ability of the two-layer LSTM neural network.

[0034] In this embodiment, the gas concentration in the multi-source data can be gas concentration distribution data, which is the gas concentration collected at multiple locations.

[0035] Training a two-layer LSTM neural network: The two-layer LSTM neural network is trained using the training set to obtain the trained two-layer LSTM neural network, such as... Figure 5 As shown, the input layer of the two-layer LSTM neural network receives training data and feeds it into the memory layer to train the LSTM neural network. The first LSTM memory layer is configured with 32 neurons, and the forget gate, update gate, and output gate work in parallel to analyze the coal mining machine's travel speed x1, coal cutting depth x2, coal seam thickness h, roadway cross-sectional area S, and processed gas concentration in the training set. The changes in wind speed V and temperature K within the tunnel at each time step are recorded to preserve the temporal characteristics of the data. Simultaneously, a Dropout mechanism is added during model training, with Dropout=0.3, which disables 30% of neurons during training to prevent overfitting.

[0036] Each neuron is equipped with a combination of three gates: a forget gate, an update gate, and an output gate. The forget gate filters out useless historical data, such as random noise from the sensor, and weakens the weights of data that have not changed significantly in the early stages, thus preventing interference with subsequent predictions. The update gate is used by the memory layer to learn new information, recording the coal mining machine's travel speed x1, coal cutting depth x2, coal seam thickness h, roadway cross-sectional area S, and the processed gas concentration. Changes in parameters such as wind speed V and temperature K within the tunnel update the content of long-term memory; the output gate, based on the updated long-term memory, determines the content of the output information for use by the next neuron or for final prediction.

[0037] The second LSTM memory layer is configured with 16 neurons to analyze the correlation changes of the above parameters over a long period of time in the training set, providing support for the prediction of gas outflow at the final moment.

[0038] A two-layer LSTM neural network was used to fit the coal mining machine travel speed x1, coal cutting depth x2, coal seam thickness h, roadway cross-sectional area S, and processed gas concentration through a fully connected layer. The nonlinear relationship between wind speed V, temperature K, and gas emission Q in the tunnel is investigated, and the predicted value of gas emission Q is finally calculated and output.

[0039] Evaluate the predictive power of the model: Introducing mean square error (MSE) and coefficient of determination (R²) 2 This method measures the overall accuracy and pattern learning ability of a two-layer LSTM neural network during training, evaluates the model's predictive ability, and validates the model using a test set that was not used in training. A lower MSE value indicates higher model prediction accuracy; R... 2 The closer the value is to 1, the higher the model fit, and the more effectively it can capture the correlation between multi-source parameters and gas emission.

[0040] The formula for calculating the mean square error (MSE) is: ; Coefficient of determination R 2 The calculation formula is: ; Where Q1 is the predicted gas emission rate, n is the total number of training samples, and Q0 is the actual gas emission rate. This represents the average of the actual gas emission rate.

[0041] Set model performance standards: when MSE ≤ 0.9 and R0.9 2 When the value is ≥0.92, the trained two-layer LSTM neural network is deemed to meet the engineering requirements of the longwall mining face, and the training is terminated.

[0042] In one exemplary embodiment, step 203 includes: A digital twin prediction model for gas outbursts is obtained by utilizing the nonlinear relationship between multi-source data at historical moments in step 202, the processed historical gas concentration, and the gas outburst predicted by the trained two-layer LSTM neural network.

[0043] The expression for the digital twin prediction model of gas outbursts is: Q=f(x1, x2, h, S, (V, K); Where Q is the gas emission rate, f(·) is a function representing the nonlinear relationship between multi-source data, processed historical gas concentration, and gas emission rate, and x1, x2, h, S, V and K represent the coal mining machine travel speed, coal cutting depth, coal seam thickness, roadway cross-sectional area, gas concentration, roadway wind speed, and roadway temperature, respectively.

[0044] This digital twin prediction model for gas emission is used to accurately predict the gas emission volume of the longwall face, while also providing support for subsequent coal cutting parameter optimization and sensor diagnosis.

[0045] In an exemplary embodiment, the digital twin prediction method for gas outburst at a coal seam mining face further includes steps 210-213: Step 210: Determine the range of operating parameters based on the upper limit of the technical parameters of the coal mining machine.

[0046] The operating parameters include the coal mining machine's travel speed and cutting depth. Specifically, the range of parameters for the coal mining machine's travel speed and cutting depth is set based on the upper limit of the coal mining machine's technical parameters.

[0047] Step 211: Based on the parameter value range, the parameter values ​​of the operation parameters are inverted using the gas emission digital twin prediction model to obtain the time series change data of the operation parameter values.

[0048] In an exemplary embodiment, step 211 specifically includes: determining the inversion parameter space, determining the adjustable range and division interval of the coal mining machine travel speed, coal cutting depth parameters and wind speed based on the upper limit of the coal mining machine technical parameters and the minimum adjustment accuracy of the equipment, and dividing the parameter grid according to the above interval.

[0049] By using a grid search method to traverse all feasible combinations of operational parameters within the aforementioned parameter space, the gas emission rate of each combination of operational parameters is predicted using a digital twin prediction model for gas emission. The predicted gas emission rate is then used to calculate the corresponding gas concentration. With the gas concentration not exceeding the limit as a constraint, operational parameter combinations that meet the constraint are selected, resulting in multiple sets of operational parameter combinations. This yields the time-series variation data of the operational parameter values.

[0050] Step 212: Calculate the coal cutting efficiency of the coal mining machine using the time-series change data of the operation parameters, and determine the parameter value of the operation parameters corresponding to the maximum coal cutting efficiency as the target parameter value.

[0051] Specifically, the coal cutting efficiency corresponding to each set of operating parameters is calculated, the highest coal cutting efficiency under the current geological conditions is determined by comparison, and the parameter values ​​of the coal mining machine travel speed and coal cutting depth corresponding to the highest coal cutting efficiency are used as target parameter values.

[0052] The formula for calculating coal cutting efficiency is: ; in, η For coal cutting efficiency, m 3 / h; x1 is the coal mining machine's travel speed, m / h; x2 is the coal cutting depth, m; B is the coal cutting height, a fixed value, m.

[0053] Step 213: Adjust the values ​​of the operation parameters based on the target parameter values.

[0054] By executing steps 210-213, the operation parameters are dynamically adjusted, thereby achieving a balance between safety and efficiency.

[0055] As the actual mining operations at the longwall face progress, the digital twin prediction model for gas emission continuously records the gas emission volume and parameter changes during the process. When the accumulated new data reaches 2000 records, the model will automatically call upon the new data to continue training the trained two-layer LSTM neural network, continuously optimizing the coal mining machine travel speed x1, coal cutting depth x2, coal seam thickness h, roadway cross-sectional area S, and processed gas concentration. The mapping relationship between wind speed V, temperature K, and gas emission Q is established to avoid the problem of prediction results failing due to changes in geological conditions, realize dynamic iteration of the model, and ensure the real-time reliability of the digital twin prediction model for gas emission.

[0056] The digital representation of the gas outburst prediction from the longwall mining face is as follows: ; X = {C, D}, C = {θ} LSTM1 θ LSTM2 Dropout}; D={MSE, R 2}; ; Y = {y1, y2, y3…y} n}; ; ; Data foundation layer: A is the parameter monitoring set, and B is the time-series dataset established using the coordinate delay method, where x1 is the coal mining machine travel speed, x2 is the coal cutting depth, h is the coal seam thickness, S is the roadway cross-sectional area, c is the gas concentration, V is the wind speed in the roadway, and K is the temperature in the roadway. The gas concentration after processing is y0, and the data preprocessing constraints include uniform step size and determining lag time.

[0057] Core training layer: X is the model training constraint set, C is the model training parameter constraint, and the model training parameters include the first layer LSTM parameters θ. LSTM1 The second LSTM layer parameters θ LSTM2 Dropout coefficients, where D is the set of model performance parameter constraints, including mean squared error (MSE) and coefficient of determination (R²). 2 E represents the core algorithm module of the model, which is manifested as the mapping relationship between multi-source parameters and gas emission. Interactive application layer: Y is the reasonable threshold for model parameter data in the system, allowing the coal mining machine's travel speed, coal cutting depth, wind speed, temperature, and gas concentration to not exceed the limits; F is the generated prediction-optimization scheme, including the generated gas emission prediction data and the parameter data corresponding to the highest coal cutting efficiency; Iterative optimization layer: B new For the new dataset being built as mining operations progress, E new To maintain the mapping relationship as new datasets are continuously updated, The model is continuously and automatically iterated.

[0058] In an exemplary embodiment, the digital twin prediction method for gas outburst at a coal seam longwall face further includes establishing a sensor status diagnosis mechanism, with steps 220-222 as follows: Step 220: Collect the real-time gas concentration in the roadway using a gas sensor.

[0059] Step 221: Determine the difference between the gas concentration at the future time and the real-time gas concentration.

[0060] Specifically, the predicted gas emission value Q1 is compared with the actual gas emission value Q0 calculated through gas concentration monitoring to determine the difference between the two.

[0061] Step 222: If the absolute value of the difference exceeds a preset value, it is determined that the gas sensor has malfunctioned and an early warning is triggered; otherwise, it is determined that the gas sensor has not malfunctioned.

[0062] Specifically, if the absolute value of the difference exceeds the threshold, i.e. If the value is greater than 0.1, the gas sensor is considered to be faulty, triggering an audible and visual warning in the well. The gas sensor at the monitoring point needs to be repaired, recalibrated, and Q0 calculated. Normal data acquisition can be resumed after confirming that the difference is within the normal range.

[0063] This application has the following effects: This application uses multiple parameters, including gas, production equipment parameters, and ventilation parameters, as input data for the model, revealing complex patterns that cannot be reflected by a single parameter, enabling the model to more comprehensively perceive the working face conditions. Secondly, this application analyzes the mapping relationship between multiple parameters and gas emission volume based on a two-layer LSTM neural network, constructing a digital twin prediction model for gas emission, achieving accurate prediction of gas emission volume. Then, based on the digital twin prediction model, the optimal coal cutting parameters are inverted with the constraint of not exceeding the gas concentration limit, achieving a balance between safety and efficiency. Simultaneously, a sensor status diagnosis mechanism is established to further reduce the cost of manual inspection. Furthermore, this application designs a dynamic iteration mechanism for the digital twin, continuously optimizing the mapping relationship between multiple parameters and gas emission volume using newly added datasets, avoiding the problem of prediction results failing due to changes in geological conditions, and ensuring the real-time reliability of the digital twin prediction model for gas emission. Finally, the real-time linkage between the digital twin prediction model for gas emission and the physical working face is achieved, solving the shortcomings of existing coal seam gas emission prediction technologies in terms of accuracy, adaptability, and functional synergy, providing technical support for safe and efficient coal mining.

[0064] Based on the same inventive concept, this application also provides a system for implementing the above-mentioned method for predicting gas emission from coal seam longwall faces based on digital twin models. The solution provided by this system is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the system for predicting gas emission from coal seam longwall faces based on digital twin models provided below can be found in the limitations of the method for predicting gas emission from coal seam longwall faces based on digital twin models described above, and will not be repeated here.

[0065] In one exemplary embodiment, a gas emission prediction system for coal seam longwall mining faces based on a digital twin model is provided, comprising: The data processing module is used to obtain the historical gas concentration of the coal seam mining face and perform delay processing to obtain the processed historical gas concentration.

[0066] The training module is used to train a two-layer LSTM neural network by taking multi-source data of historical moments and processed historical gas concentration as input and historical gas emission as label, so as to obtain the trained two-layer LSTM neural network.

[0067] The model building module is used to construct a digital twin prediction model for gas outbursts based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outbursts predicted by a trained two-layer LSTM neural network.

[0068] The data acquisition module is used to collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and to perform delay processing on the real-time gas concentration.

[0069] The gas prediction module is used to predict the future gas emission amount based on real-time multi-source data and processed real-time gas concentration, using the gas emission digital twin prediction model, to obtain the gas emission amount at future times.

[0070] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting gas emission in coal seam longwall faces based on a digital twin model, characterized in that, include: The historical gas concentration at the coal seam longwall face is obtained and then delayed to obtain the processed historical gas concentration. Using multi-source historical data and processed historical gas concentration as input, and historical gas emission as label, a two-layer LSTM neural network is trained to obtain the trained two-layer LSTM neural network. A digital twin prediction model for gas outbursts is constructed based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outbursts predicted by a trained two-layer LSTM neural network. Collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and perform delay processing on the real-time gas concentration; Based on real-time multi-source data and processed real-time gas concentration, the gas emission digital twin prediction model is used to predict the future gas emission amount, thus obtaining the gas emission amount at future times.

2. The method for predicting gas emission in coal seam longwall faces based on a digital twin model according to claim 1, characterized in that, The multi-source data includes the coal mining machine's travel speed, coal cutting depth, coal seam thickness, roadway cross-sectional area, roadway wind speed, and roadway temperature.

3. The method for predicting gas emission in coal seam longwall faces based on a digital twin model according to claim 1, characterized in that, The coordinate delay method is used to delay the gas concentration at historical moments.

4. The method for predicting gas emission in coal seam longwall faces based on a digital twin model according to claim 1, characterized in that, The digital twin prediction method for gas outburst at coal seam longwall faces also includes: The range of operating parameters is determined based on the upper limit of the technical parameters of the coal mining machine; Based on the range of the parameter values, the parameter values ​​of the operation parameters are inverted using the gas emission digital twin prediction model to obtain the time-series change data of the operation parameter values; The coal cutting efficiency of the coal mining machine is calculated using the time-series variation data of the operation parameters, and the parameter value of the operation parameters corresponding to the maximum coal cutting efficiency is determined as the target parameter value. The values ​​of the operation parameters are adjusted based on the target parameter values.

5. The method for predicting gas emission in coal seam longwall faces based on a digital twin model according to claim 4, characterized in that, The operating parameters include the coal mining machine's travel speed and the coal cutting depth.

6. The method for predicting gas emission in coal seam longwall faces based on a digital twin model according to claim 1, characterized in that, The digital twin prediction method for gas outburst at coal seam longwall faces also includes: Real-time gas concentration in the tunnel is collected using a gas sensor; Determine the difference between the future gas concentration and the real-time gas concentration; If the absolute value of the difference exceeds a preset value, the gas sensor is determined to be faulty and an early warning is triggered; otherwise, the gas sensor is determined not to be faulty.

7. A gas emission prediction system for coal seam longwall faces based on a digital twin model, characterized in that, include: The data processing module is used to obtain the historical gas concentration of the coal seam mining face and perform delay processing to obtain the processed historical gas concentration. The training module is used to train a two-layer LSTM neural network with multi-source data of historical moments and processed historical gas concentration as input and historical gas emission as label, so as to obtain the trained two-layer LSTM neural network. The model building module is used to construct a digital twin prediction model for gas outburst based on the nonlinear relationship between multi-source data at historical moments, processed historical gas concentrations, and gas outburst volume predicted by a trained two-layer LSTM neural network. The data acquisition module is used to collect real-time multi-source data and real-time gas concentration of the coal mining working environment, and to perform delay processing on the real-time gas concentration. The gas prediction module is used to predict the future gas emission amount based on real-time multi-source data and processed real-time gas concentration, using the gas emission digital twin prediction model, to obtain the gas emission amount at future times.

Images