Y-shaped ventilation gas extraction method for short-distance coal seam group

Abstract

The invention provides a short-distance coal seam group Y-shaped ventilation gas extraction method, which belongs to the technical field of coal mining, and comprises the following steps: firstly, arranging a main drill hole vertical to a coal seam and branch drill holes extending to upstream and downstream at an angle of 45 degrees to form a Y-shaped three-dimensional extraction network; installing a gas concentration monitoring sensor network, and setting monitoring points to collect real-time data; constructing a gas concentration adjusting buffer system and an intelligent mixed gas distribution device, and maintaining the concentration of extracted gas to be 30-45%; establishing a predictive control model based on a long short-term memory neural network, predicting a gas emission rule in advance and dynamically adjusting extraction parameters; a gas concentration grading utilization system is implemented, and gas with different concentrations is guided into the corresponding utilization system; a gas extraction intelligent monitoring platform is deployed to realize remote monitoring and fault early warning; the technical problems that in the coal seam group gas extraction process, the extraction concentration is unstable, and prediction control is difficult are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal mine mining, and specifically relates to a Y-shaped ventilation gas drainage method for close coal seam groups. Background Art

[0002] Coal mine gas drainage is a key technology to ensure the safe production of coal mines. Traditional gas drainage methods mainly use single straight boreholes or horizontal boreholes for drainage, and the gas in the coal seam is pumped out and transported to the ground through a negative pressure system. These traditional methods have achieved certain effects in single coal seam drainage, but their application scenarios are limited to simple geological conditions and stable gas emission environments.

[0003] However, under complex coal seam group conditions, the existing gas drainage technology faces problems such as large fluctuations in drainage concentration and poor stability. The traditional borehole layout cannot effectively cover the gob area and the gas enrichment areas of adjacent coal seams, resulting in limited drainage scope; at the same time, drainage parameters such as negative pressure and flow rate are mostly fixed values or manually adjusted, and cannot be adjusted in real time according to the gas emission law, resulting in low drainage efficiency and even the phenomenon of "air leakage" in drainage.

[0004] Especially during the rapid advancement of the working face, the gas emission volume and concentration change violently. The existing technology lacks the accurate prediction ability and dynamic regulation mechanism for the gas emission law, resulting in unstable drainage gas concentration and difficulty in maintaining it within the suitable utilization concentration range, seriously restricting the gas drainage efficiency and the gas resource utilization rate. That is to say, there are technical problems in the existing technology that the drainage concentration is unstable and difficult to predict and control during the gas drainage process of coal seam groups. Summary of the Invention

[0005] In view of this, the present invention provides a Y-shaped ventilation gas drainage method for close coal seam groups, which can solve the technical problems in the existing technology that the drainage concentration is unstable and difficult to predict and control during the gas drainage process of coal seam groups.

[0006] The present invention is implemented as follows: The present invention provides a Y-shaped ventilation gas drainage method for close coal seam groups, including: arranging a Y-shaped ventilation drainage borehole group; installing a gas concentration monitoring sensor network; constructing a gas concentration adjustment buffer system; installing an intelligent gas mixing device; establishing a dynamic regulation system for gas drainage parameters; applying a predictive control model, analyzing the gas emission law of the working face based on historical data, receiving data such as the working face advancement speed data, the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data, outputting the predicted value of the gas emission volume and the predicted value of the concentration change trend in the next 12 hours, and adjusting the drainage negative pressure and flow rate 12 hours in advance; implementing a gas concentration grading utilization system; deploying an intelligent monitoring platform for gas drainage; and performing an evaluation of the gas drainage effect.

[0007] Among them, the Y-shaped ventilation and drainage borehole group is a special borehole structure pre-arranged at the front of coal seam mining. It consists of a main borehole and two branch boreholes. The main borehole is drilled perpendicular to the coal seam direction and reaches the coal seam to be mined. The two branch boreholes extend from the bottom of the main borehole to the upstream and downstream directions respectively, with an inclination angle of 45°, covering the goaf and adjacent coal seams, and forming a three-dimensional drainage network.

[0008] Among them, the gas concentration monitoring sensor network specifically sets a monitoring point every 50 meters at the borehole outlet and key nodes of the main pipeline, and real-time collects the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data.

[0009] Among them, the gas concentration adjustment and buffer system is a device specifically composed of a main buffer tank, an auxiliary buffer tank, a flow regulating valve, a pressure sensor, and a concentration sensor. The main buffer tank is of a cylindrical structure with a volume of 10 3 m 3 , the volume of the auxiliary buffer tank is 5×10 2 m 3 , and electric regulating valves are installed at both ends.

[0010] Among them, the intelligent gas mixing device is a device specifically composed of a mixing chamber, multiple inlet gas pipes, an electric proportional valve, a gas analyzer, and a controller. The mixing chamber is a sealed container with a volume of 5m 3 , and a turbine stirring device is installed inside. The multiple inlet gas pipes are respectively connected to high-concentration gas sources, medium-concentration gas sources, and low-concentration gas sources.

[0011] Among them, the dynamic regulation system for gas drainage parameters specifically automatically adjusts the drainage negative pressure and flow rate according to the data of the working face advancing speed, the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data.

[0012] Among them, the predictive control model is a gas emission prediction model specifically constructed based on a long short-term memory neural network, which includes three parts: an input layer, a hidden layer, and an output layer. The input layer receives the data of the working face advancing speed, the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data. The hidden layer includes four layers of long short-term memory units, with 128 neurons in each layer.

[0013] Among them, the specific structure of the predictive control model is a multi-variable time series prediction network structure, which consists of a data preprocessing unit, a feature extraction unit, a time series learning unit, and a prediction unit. The feature extraction unit adopts a bidirectional long short-term memory network structure, and the time series learning unit adopts an attention mechanism to enhance the model's ability to capture long time series dependencies. Specifically, the time series learning unit of the predictive control model is specifically that the attention layer is composed of multi-head self-attention modules, the number of heads is set to 8, the attention dimension is 64, and the residual connection and layer normalization are combined to improve the model stability. The prediction unit is composed of two fully connected layers, and the first layer contains 64 neurons.

[0014] Among them, the steps for establishing the training data set of the predictive control model specifically include four parts: data collection stage, data cleaning stage, feature extraction stage, and data set division stage. In the data collection stage, historical monitoring data for 90 consecutive days is obtained from the gas concentration monitoring sensor network.

[0015] The present invention forms a three-dimensional drainage network through a special Y-shaped borehole structure, effectively expanding the drainage coverage range and solving the problem of insufficient coverage in traditional borehole layouts. At the same time, with the help of the gas concentration adjustment buffer system and the intelligent gas mixing device, stable control of the extracted gas concentration is achieved, effectively eliminating the concentration fluctuation problem in traditional drainage.

[0016] The predictive control model constructed based on the long short-term memory neural network can accurately analyze the correlation between the advancement of the working face and gas emission, predict the gas emission law 12 hours in advance, and provide a scientific basis for the dynamic regulation of drainage parameters. The system automatically adjusts the drainage negative pressure and flow according to the prediction results, enabling the drainage process to actively adapt to the change of gas emission, and realizing the transformation from passive drainage to active pre-control.

[0017] Through the technical solution of the present invention, the core technical problems of unstable drainage concentration and difficult prediction and control in the process of gas drainage from coal seam groups are effectively solved, the gas drainage efficiency and the utilization rate of gas resources are significantly improved, and a new technical approach is provided for coal mine safety production and efficient utilization of gas resources. Brief Description of the Drawings

[0018] Figure 1 It is a flowchart of the method of the present invention.

[0019] Figure 2 It is a schematic diagram of the overall structure of the Y-shaped ventilation gas drainage method in Embodiment 2.

[0020] Figure 3 It is a detailed drawing of the Y-shaped borehole structure in Embodiment 2.

[0021] Figure 4 It is a structural diagram of the gas concentration adjustment buffer system in Embodiment 2.

[0022] Figure 5 It is the structural diagram of the intelligent hybrid gas mixing device in Embodiment 2.

[0023] Figure 6 It is the structural diagram of the predictive control model in Embodiment 2.

[0024] Figure 7 It is the structural diagram of the hierarchical utilization system of gas concentration in Embodiment 2.

[0025] Figure 8 It is the structural diagram of the intelligent monitoring platform for gas drainage in Embodiment 2. Specific implementation manners

[0026] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] As Figure 1 shown, it is the flow chart of a method for gas drainage in Y-shaped ventilation of close coal seam groups provided by the present invention. This method includes the following steps:

[0028] S01. Layout the Y-shaped ventilation drainage borehole group, where the main boreholes are arranged vertically to the coal seam, and the branch boreholes extend at an angle of 45° upstream and downstream to cover the goaf and coal seam groups;

[0029] S02. Install a gas concentration monitoring sensor network, and set a monitoring point every 50 meters at the borehole outlet and key nodes of the main pipeline to collect the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data in real time;

[0030] S03. Construct a gas concentration adjustment buffer system, including a main buffer tank, an auxiliary buffer tank, a flow regulating valve, and a mixing device, so that the extracted gas concentration is stabilized between 30% and 45%;

[0031] S04. Install an intelligent hybrid gas mixing device, and mix low-concentration gas and high-concentration gas in a ratio of 4:6 through a negative pressure drainage system to keep the gas concentration in the pipeline network stable;

[0032] S05. Establish a dynamic regulation system for gas drainage parameters, and automatically adjust the drainage negative pressure and flow rate according to the data of the working face advancing speed, the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data;

[0033] S06. Apply the predictive control model, analyze the gas emission law of the working face based on historical data, receive the data of the working face advance speed, the gas concentration at the drill hole outlet, the gas concentration at the key nodes of the main pipeline, and the negative pressure value, and output the predicted value of the gas emission volume and the predicted value of the concentration change trend in the next 12 hours, and adjust the drainage negative pressure and flow rate 12 hours in advance;

[0034] S07. Implement the gas concentration grading utilization system, divide the gas into high-concentration area, medium-concentration area and low-concentration area according to the concentration, and introduce them into different utilization systems respectively;

[0035] S08. Deploy an intelligent monitoring platform for gas drainage, integrate the data of drainage negative pressure, flow rate, gas concentration at the drill hole outlet, gas concentration at the key nodes of the main pipeline and system status data, and realize remote monitoring and fault warning;

[0036] S09. Conduct an evaluation of the gas drainage effect, quantify and evaluate the drainage effect through the determination of the residual gas content, the calculation of the drainage rate and the statistics of the gas utilization rate, and guide the optimization of the drainage negative pressure and flow rate.

[0037] Among them, the Y-shaped ventilation drainage borehole group is a special borehole structure arranged in advance at the front of the coal seam mining. It consists of a main borehole and two branch boreholes, which is in the shape of the English letter Y. The main borehole drills vertically in the direction of the coal seam and reaches the coal seam to be mined. The two branch boreholes extend from the bottom of the main borehole to the upstream and downstream directions respectively, with an inclination angle of 45°, covering the goaf and adjacent coal seams, and forming a three-dimensional drainage network.

[0038] Among them, the gas concentration adjustment buffer system is a device specifically composed of a main buffer tank, an auxiliary buffer tank, a flow regulating valve, a pressure sensor and a concentration sensor. The main buffer tank is a cylindrical structure with a volume of 10 3 m 3 , with multiple layers of flow guiding plates inside. The auxiliary buffer tank is connected to the main buffer tank through a pipeline, and its volume is 5×10 2 m 3 . Electric regulating valves are installed at both ends. When the gas concentration of the drainage fluctuates, the system controls the mixing ratio of different concentration gases by adjusting the valve opening to keep the output gas concentration stable.

[0039] Among them, the intelligent gas mixing device is a device specifically composed of a mixing chamber, multiple inlet gas pipes, an electric proportional valve, a gas analyzer and a controller. The mixing chamber is a closed container with a volume of 5m 3 , and a turbine stirring device is installed inside. The multiple inlet gas pipes are respectively connected to the high-concentration gas source, the medium-concentration gas source and the low-concentration gas source. An electric proportional valve is installed on each inlet gas pipe. The controller accurately controls the opening of each valve according to the real-time concentration data measured by the gas analyzer to achieve the precise mixing of different concentration gases.

[0040] Among them, the predictive control model is specifically a gas emission prediction model constructed based on a long short-term memory neural network, which includes three parts: an input layer, a hidden layer, and an output layer. It processes data through connection weights and activation functions. The input layer receives data such as the working face advancing speed, the gas concentration at the drill hole outlet, the gas concentration at the key nodes of the main pipeline, and the negative pressure value. The hidden layer contains four layers of long short-term memory units, with 128 neurons in each layer. The output layer outputs the predicted value of the gas emission volume and the predicted value of the concentration change trend in the next 12 hours.

[0041] The specific structure of the predictive control model is a multi-variable time series prediction network structure, which consists of a data preprocessing unit, a feature extraction unit, a time series learning unit, and a prediction unit. The data preprocessing unit is responsible for normalizing and denoising the raw data collected by the sensors. The feature extraction unit adopts a bidirectional long short-term memory network structure, which includes two layers of forward propagation layer and two layers of backward propagation layer, with 128 neurons in each layer, and is used to extract the deep features of the time series data. The time series learning unit uses an attention mechanism to enhance the model's ability to capture long time series dependencies. The attention layer consists of a multi-head self-attention module, with the number of heads set to 8 and the attention dimension to 64. Combining residual connection and layer normalization improves the model's stability. The prediction unit consists of two fully connected layers. The first layer contains 64 neurons, and the output dimension of the second layer is consistent with the prediction target, that is, the predicted value of the gas emission volume per hour and the predicted value of the concentration change trend in the next 12 hours. The input layer of the model is directly connected to the gas concentration monitoring sensor network and receives four key parameters: the working face advancing speed data, the gas concentration at the drill hole outlet data, the gas concentration at the key nodes of the main pipeline data, and the negative pressure value data. The working face advancing speed data comes from the working face advancing speed monitoring system. The gas concentration at the drill hole outlet data and the gas concentration at the key nodes of the main pipeline data come from the gas concentration monitoring sensor network in step S02. The negative pressure value data comes from the gas concentration monitoring sensor network in step S02. The output layer of the model is connected to the dynamic regulation system of gas extraction parameters and provides the predicted value of the gas emission volume and the predicted value of the concentration change trend in the next 12 hours for the extraction negative pressure and flow adjustment in step S05. The overall model adopts a series structure, and information is transmitted between each unit through a fully connected layer. At the same time, a dropout mechanism is set between each layer to prevent overfitting.

[0042] The steps for establishing the training dataset of the predictive control model specifically include four parts: data collection phase, data cleaning phase, feature extraction phase, and dataset division phase. In the data collection phase, historical monitoring data for 90 consecutive days is obtained from the gas concentration monitoring sensor network in step S02, including gas concentration data at the borehole outlet, gas concentration data at key nodes of the main pipeline, and negative pressure value data. At the same time, the working face advance speed data and historical gas emission data for the corresponding time period are collected from the working face advance speed monitoring system. The sampling frequency is set to once every 10 minutes to form an initial dataset. In the data cleaning phase, outlier detection and processing are performed on the initial dataset. The three-sigma method is used to identify outliers, and missing data is filled in using the forward filling method. Then, the moving average method is used to smooth the data to reduce the influence of noise. In the feature extraction phase, time series features are extracted based on the time window method. The window length is set to 24 hours, and the step size is 1 hour. For each time window, the mean value of the working face advance speed, the mean value of the gas concentration, the standard deviation of the gas concentration, the mean value of the negative pressure value, and the total gas emission are extracted as features. At the same time, the correlation coefficients between the features are calculated as additional features. In the dataset division phase, the processed dataset is divided into a training set, a validation set, and a test set according to the ratio of 7:2:1. The training set is used for model parameter learning, the validation set is used for model hyperparameter tuning and early stopping strategy implementation, and the test set is used to evaluate the performance of the final model. To enhance the generalization ability of the model, data augmentation techniques are applied to the training set, including adding Gaussian noise and time translation methods to generate additional samples.

[0043] The steps of training the predictive control model specifically include four parts: the parameter initialization stage, the model training stage, the parameter optimization stage, and the model evaluation stage. In the parameter initialization stage, the Xavier initialization method is used to initialize the network weights, so that the variances of the input and output of each layer are consistent, avoiding the problems of gradient disappearance or gradient explosion. The bias term is initialized to zero, and the forget gate bias of the long short-term memory network is initialized to 1 to promote information transmission. In the model training stage, the mini-batch stochastic gradient descent algorithm is used to update the parameters. The batch size is set to 64, the number of training epochs is 200, the initial learning rate is set to 0.001, the mean squared error is used as the loss function, and at the same time, the L2 regularization term is introduced to prevent overfitting, and the regularization coefficient is set to 0.0001. During the training process, the learning rate decay strategy is used, and the learning rate is halved every 50 epochs. In the parameter optimization stage, the Bayesian optimization method is used to tune the hyperparameters of the model. The optimization objects include the number of neurons in the hidden layer, the number of long short-term memory layers, the number of attention heads, and the dropout ratio. The optimization goal is the prediction error on the validation set. By constructing a Gaussian process regression model to estimate the relationship between the hyperparameters and the objective function, the search direction of the hyperparameters is guided. In the model evaluation stage, three indicators, namely the root mean square error, the mean absolute error, and the coefficient of determination, are used to evaluate the performance of the model on the test set, and the stability and adaptability of the model are analyzed by comparing the prediction results at different time scales. Finally, the model with the optimal comprehensive performance is selected as the core component of the predictive control system.

[0044] Among them, the gas concentration classification and utilization system specifically classifies the extracted gas into three levels according to the concentration: the gas concentration in the high-concentration area is greater than 30%, which is directly used for power generation or liquefaction utilization; the gas concentration in the medium-concentration area is between 10% and 30%, and it is converted into heat energy through a catalytic oxidation device for utilization; the gas concentration in the low-concentration area is between 5% and 10%, and it is discharged after being treated by a thermal oxidation technology.

[0045] Among them, the intelligent monitoring platform for gas extraction is a comprehensive monitoring system that integrates a data acquisition system, a transmission network, a server, and application software. The data acquisition system consists of sensors distributed at each monitoring point, and transmits the extraction negative pressure data, flow data, gas concentration data at the drilling outlet, gas concentration data at key nodes of the main pipeline, and system status data to the central server through the industrial Ethernet. The application software includes a data processing module, a visualization display module, an early warning module, and a decision support module. The system displays the extraction negative pressure data, flow data, gas concentration data at the drilling outlet, gas concentration data at key nodes of the main pipeline, and system status data in real time, and automatically triggers an early warning when the monitored data exceeds the set threshold.

[0046] Among them, the working face advancing speed monitoring system is specifically a monitoring device composed of a displacement sensor, a data acquisition unit, and a data processing unit. The displacement sensor is installed on the shearer and is used to monitor the position change of the shearer in real time. The data acquisition unit collects the signals of the displacement sensor and converts them into digital signals. The data processing unit calculates the working face advancing speed data based on the position change and time.

[0047] Among them, the evaluation of the drainage effect is specifically evaluated through three aspects: measuring the residual gas content, calculating the drainage rate, and counting the gas utilization rate. The residual gas content is measured by taking samples at different positions on the working face using the borehole sampling method. The drainage rate is calculated by dividing the difference between the original gas content and the residual gas content by the original gas content. The gas utilization rate is the ratio of the actual gas utilization volume to the total drained gas volume.

[0048] The following will describe the specific implementation manners of the above steps in detail.

[0049] The specific implementation manner of step S01 is to determine the coal seam distribution through ground penetrating radar detection technology and three-dimensional modeling technology. First, drill a main borehole above the coal seam. The main borehole is drilled perpendicular to the coal seam direction and reaches the coal seam to be mined. The borehole diameter is 120 mm, and the borehole wall is reinforced with high-strength steel pipes with a wall thickness of 10 mm. Then, drill two branch boreholes in the upstream and downstream directions respectively at the bottom of the main borehole. The inclination angle of the branch borehole with the horizontal plane is kept at 45°. The branch borehole diameter is 90 mm, and the length is determined according to the coal seam thickness, generally 3 to 5 times the coal seam thickness. Steel pipes are also laid in the branch boreholes for support. After the boreholes are completed, pressure tests are carried out to ensure the borehole tightness. When arranging the Y-shaped ventilation and drainage borehole group, the distance between adjacent main boreholes is 30 m, and a row of boreholes is arranged every 50 m along the coal seam strike to form a three-dimensional drainage network covering the entire mining area. The implementation of this step is based on the principle of fluid mechanics. By increasing the gas drainage coverage area through a three-dimensional cross-borehole network, the gas drainage efficiency is improved. The purpose of this step is to form a three-dimensional gas drainage channel network to create conditions for subsequent gas drainage.

[0050] The specific implementation of step S02 is to select catalytic combustion type gas sensors and thermal conductivity type gas sensors to construct a monitoring network. A catalytic combustion type gas sensor is installed at the outlet of the Y-shaped borehole, with a measurement range of 0 to 5% and an accuracy of ±0.1%. A thermal conductivity type gas sensor is installed every 50 meters at the key nodes of the main pipeline, with a measurement range of 0 to 100% and an accuracy of ±0.5%. At the same time, negative pressure sensors are installed at each monitoring point, with a measurement range of 0 to 25 kPa and an accuracy of ±0.2 kPa. All sensors adopt intrinsically safe design, with an explosion-proof grade of ExiaⅠ. The sensor acquisition frequency is set to once every 10 seconds. The sensors are connected to the nearest data collector through the RS485 bus, and the data collector transmits the acquired data to the monitoring center through the industrial Ethernet. In addition, a data anomaly self-check mechanism is set in the system. When the sensor readings show sudden changes or remain unchanged for a long time, the system automatically marks and issues an alarm. The implementation of this step is based on sensor network technology and the principles of industrial Internet of Things, constructing a real-time monitoring network for gas concentration and negative pressure value, providing data support for the gas drainage process. The key parameters of the sensors include measurement range, accuracy, and acquisition frequency. These parameters determine the reliability and real-time nature of the monitoring data, directly affecting the decision-making quality of the subsequent steps. The purpose of this step is to achieve real-time monitoring of gas concentration and negative pressure value, providing a data basis for subsequent dynamic adjustment.

[0051] The specific implementation of step S03 is to first establish a main buffer tank, which is made of Q345R steel, with a wall thickness of 20 mm, in a cylindrical shape, a diameter of 10 m, a height of 12.8 m, and a volume of 10 3 m 3 , with 5 layers of baffle plates inside. The baffle plates are arranged in an equally spaced spiral pattern, with an angle of 30°. The surface of the baffle plates is made of hydrophilic material, which can adsorb the moisture in the gas. The auxiliary buffer tank is also made of Q345R steel, with a diameter of 8 m, a height of 10 m, and a volume of 5×10 2 m 3, with foamed plastic filler inside, the specific surface area reaches 1000 square meters per cubic meter, and the height of the filler layer is 5 meters. An electric control valve is installed between the main buffer tank and the auxiliary buffer tank. The valve is an angular stroke electric control valve, with a maximum diameter of 800 mm, a rated pressure of 1.6 MPa, a regulation ratio of 100:1, and a switching time of less than 30 seconds. An on-line gas concentration analyzer is installed in the whole system, adopting the infrared absorption principle, with a measurement range of 0 to 100%, an accuracy of ±0.2%, and a response time of less than 10 seconds. When the monitored extracted gas concentration is lower than 30%, the control system automatically reduces the opening of the low-concentration intake valve and increases the opening of the high-concentration intake valve, and vice versa, ensuring that the mixed gas concentration is maintained between 30% and 45% through flow ratio regulation. The implementation of this step is based on the principles of fluid dynamics and automatic control, stabilizing the gas concentration through the buffer device and the automatic regulation system, and solving the problem of gas concentration fluctuation. The key parameter in this step is the stable range of gas concentration from 30% to 45%, which not only meets the concentration requirements for comprehensive gas utilization but also avoids the risk of gas entering the explosive concentration range. The purpose of this step is to stabilize the extracted gas concentration and create conditions for subsequent comprehensive utilization.

[0052] The specific implementation method of step S04 is to construct a mixing cavity, made of 304 stainless steel material, with a wall thickness of 15 mm and a volume of 5m 3 , with a turbine stirring device installed inside, the turbine diameter is 0.8 m, and the rotation speed can be adjusted within the range of 0 to 500 revolutions per minute. There are 4 gas inlet pipes set on the mixing cavity, which are respectively used to connect to the high-concentration gas source, the medium-concentration gas source and the low-concentration gas source, and there is another one for air dilution in case of emergency. An electric proportional valve is installed on each gas inlet pipe. The valve adopts a linear stroke electric control valve, with a maximum diameter of 200 mm, a rated pressure of 1.0 MPa, and a regulation ratio of 50:1. A gas analyzer is installed at the outlet of the mixing cavity, adopting the thermal conductivity principle, with a measurement range of 0 to 100%, an accuracy of ±0.3%, and a response time of less than 5 seconds. The control system adopts a fuzzy PID control algorithm, adjusts the opening of each valve according to the real-time concentration data, keeps the mixing ratio of the low-concentration gas and the high-concentration gas constant at 4:6, and makes the mixed gas concentration stable within the required range for utilization. The sampling period of the control system is 1 second, the PID parameters are adjusted adaptively, and the response time to concentration fluctuation is less than 3 seconds. The implementation of this step is based on the theory of mixed gas and fuzzy control technology, and realizes the stable control of gas concentration by precisely controlling the mixing ratio of different-concentration gases. The key parameter in this step is the mixing ratio of 4:6, which is determined based on the optimal working conditions of the gas utilization system, and can ensure the stable operation of the system and obtain the maximum energy conversion efficiency. The purpose of this step is to achieve precise control of gas concentration through intelligent mixing and gas distribution, and improve the gas utilization efficiency.

[0053] The specific implementation of step S05 is to construct a dynamic regulation system for gas drainage parameters using a multi-input multi-output control system. The system receives four key parameters: the working face advancing speed data, the gas concentration data at the borehole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data. The data is preprocessed by a data processing unit, including filtering, denoising, and normalization. Then, the processed data is input into a fuzzy neural network controller, which automatically calculates the optimal drainage negative pressure and flow rate parameters according to the preset fuzzy rule base. The fuzzy rule base contains 50 inference rules, covering the combination of the working face advancing speed, gas concentration, and negative pressure value under different working conditions. The controller outputs the drainage negative pressure adjustment instruction and the flow rate adjustment instruction, and adjusts the rotation speed of the drainage pump and the valve opening degree on the pipeline through the actuator to achieve the dynamic adjustment of the drainage negative pressure and flow rate. The control period of the system is 5 minutes, and it can be shortened to 1 minute in extreme cases. The system is also equipped with a safety protection mechanism. When the gas concentration exceeds 45% or is lower than 25%, an early warning is automatically triggered and the system switches to the safe operation mode. The implementation of this step is based on the fuzzy neural network control theory, which transforms expert experience into fuzzy rules and realizes the adaptive control of the system through the self-learning ability of the neural network. The key parameters in this step include the drainage negative pressure range of 15 to 25 kPa and the flow rate adjustment range of 50 to 200 m³ / min. These parameter ranges are determined based on the actual requirements of coal mine gas drainage, which can not only ensure the drainage effect but also avoid the problem of excessive air inhalation caused by too high negative pressure. The purpose of this step is to dynamically adjust the drainage parameters according to real-time data to improve the drainage efficiency and safety.

[0054] The specific implementation of step S06 is to construct a gas emission prediction model based on a long short-term memory neural network. The model adopts a hierarchical structure, including four functional modules: a data preprocessing unit, a feature extraction unit, a time series learning unit, and a prediction unit. The data preprocessing unit standardizes the input data of the working face advance speed, the gas concentration at the borehole outlet, the gas concentration at the key nodes of the main pipeline, and the negative pressure value. The Z-score method is used to make the data mean 0 and the standard deviation 1. At the same time, wavelet transform is used for signal denoising. The feature extraction unit adopts a bidirectional long short-term memory network structure, including two layers of forward propagation layer and two layers of backward propagation layer, with 128 neurons in each layer. The tanh activation function is used to capture the context features of time series data in both forward and backward directions. The time series learning unit adopts a multi-head self-attention mechanism, with the number of heads set to 8 and the attention dimension to 64. Combining residual connection and layer normalization to improve the stability of the model, and enhancing the model's ability to capture long-term time series dependencies by calculating the correlation weights between different time steps. The prediction unit consists of two fully connected layers. The first layer contains 64 neurons and uses the ReLU activation function. The output dimension of the second layer is the same as the prediction target, and it outputs the predicted values of the gas emission per hour and the predicted values of the concentration change trend within the next 12 hours. According to the prediction results, the system adjusts the extraction negative pressure and flow rate 12 hours in advance. When the predicted gas emission increases, the extraction negative pressure and flow rate are increased in advance, and vice versa, they are appropriately reduced. The implementation of this step is based on deep learning and time series prediction theory, mining the gas emission law through historical data analysis, and realizing the early prediction of gas emission and the pre-adjustment of extraction parameters. The key parameter in this step is the prediction duration of 12 hours, which is determined based on the actual production requirements of the mine, meeting both the accuracy requirements of the prediction and leaving enough time for the adjustment of extraction parameters. The purpose of this step is to achieve the early prediction of gas emission and provide decision support for the dynamic adjustment of extraction parameters.

[0055] The specific implementation of step S07 is to establish a three - level hierarchical utilization system based on the gas concentration. First, install an on - line gas concentration analysis system. Use an infrared absorption type gas analyzer with a measurement range of 0 to 100% and an accuracy of ±0.2%, and a sampling frequency of once every 5 seconds. The system divides the gas into three grades according to the real - time measured gas concentration: the high - concentration area where the gas concentration is greater than 30%, the medium - concentration area where the gas concentration is between 10% and 30%, and the low - concentration area where the gas concentration is between 5% and 10%. Install an electric three - way diverter valve in the pipeline system. The valve is an angular stroke electric control valve with a maximum diameter of 400 mm, a rated pressure of 1.0 MPa, a regulation ratio of 50:1, and a response time of less than 5 seconds. The control system automatically adjusts the direction of the three - way valve according to the gas concentration and guides the gas with different concentrations into the corresponding utilization systems. The high - concentration area gas is directly introduced into the gas power plant or liquefaction device. The medium - concentration area gas is introduced into the catalytic oxidation device. The oxidation process is carried out at a temperature of 350 °C, using a palladium - platinum catalyst with a catalyst loading of 1.5 kg / m³ and a catalytic conversion rate of greater than 95%. The heat energy released by oxidation is used for mine heating or hot water supply. The low - concentration area gas is introduced into the thermal oxidation device and completely oxidized at a temperature of 850 °C with an oxidation efficiency of greater than 99%, and the methane content in the exhaust gas is less than 0.1%, meeting the environmental protection requirements. The implementation of this step is based on the theory of hierarchical utilization of gas resources. Different utilization methods are adopted according to the characteristics of gas with different concentrations, improving the comprehensive utilization rate of gas. The key parameters in this step are the gas concentration classification thresholds of 30% and 10%. These thresholds are determined based on the applicable conditions of different gas utilization technologies to ensure that each grade of gas can be utilized most effectively. The purpose of this step is to achieve the cascaded utilization of gas resources, improve the comprehensive utilization efficiency of gas, and reduce greenhouse gas emissions.

[0056] The specific implementation of step S08 is to build an intelligent monitoring platform for gas drainage based on the industrial Internet of Things architecture. The platform consists of four parts: the perception layer, the network layer, the platform layer, and the application layer. The perception layer is composed of sensors distributed at each monitoring point, including gas concentration sensors, negative pressure sensors, flow sensors, and status sensors, etc. The sampling frequency is once per second. The network layer adopts a redundant star network topology. The main network uses industrial Ethernet with a transmission rate of 1000 megabits per second, and the standby network uses industrial wireless network with a transmission rate of 100 megabits per second. The network latency is less than 10 milliseconds. The platform layer adopts a distributed architecture, including a data acquisition server, a data storage server, and an application server. The servers adopt a dual-machine hot backup mode. The data storage uses a time-series database, supporting a data writing speed of 10,000 points per second and a query response time in milliseconds. The application layer includes a data processing module, a visualization display module, an early warning module, and a decision support module. The data processing module adopts a stream processing architecture, supporting real-time data analysis. The visualization display module is implemented using Web technology, supporting multi-terminal access. The early warning module is implemented based on a rule engine, containing 100 early warning rules, covering situations such as abnormal gas concentration, abnormal negative pressure, and abnormal flow. The decision support module is based on case-based reasoning technology, storing 500 historical cases, and can match similar cases according to the current situation and provide decision-making suggestions. The implementation of this step is based on the industrial Internet of Things and big data analysis technologies, building an intelligent monitoring system for gas drainage, realizing real-time monitoring, analysis, and early warning of data. The key parameters in this step are that the system response time is less than 1 second and the data storage period is 1 year. These parameters ensure the real-time nature and data integrity of the system, providing guarantees for the safe operation and subsequent analysis of gas drainage. The purpose of this step is to achieve intelligent monitoring of the gas drainage process and improve the safety and reliability of system operation.

[0057] The specific implementation of step S09 is to evaluate the gas drainage effect through measurements and calculations in three aspects. First, the residual gas content is measured. The borehole sampling method is used. One measuring point is arranged every 20 meters along the strike of the working face and every 10 meters along the dip to form a measuring point network. A sampling borehole is drilled at each measuring point with a depth of 8 meters and a diameter of 42 mm. A sampler is placed in the borehole. The sampler is a sealed container with a volume of 500 cm³. The gas in the coal sample is extracted by a vacuum pump to measure the gas content. Then, the drainage rate is calculated using the formula η = (q0 - q1) / q0 × 100%, where η is the drainage rate, q0 is the original gas content in m³ / t, and q1 is the residual gas content in m³ / t. Finally, the gas utilization rate is statistically calculated using the formula ξ = V1 / V0 × 100%, where ξ is the gas utilization rate, V1 is the actual gas utilization volume in m³, and V0 is the total drained gas volume in m³. According to the evaluation results, when the drainage rate is lower than 70% or the gas utilization rate is lower than 80%, the system automatically optimizes the drainage negative pressure and flow parameters to improve the drainage effect by increasing the drainage negative pressure or adjusting the drainage time. The implementation of this step is based on the principles of scientific evaluation and closed-loop optimization. By quantitatively evaluating the drainage effect, it provides a basis for optimizing the drainage parameters. The key parameters in this step are the drainage rate target value of 70% and the gas utilization rate target value of 80%. These target values are comprehensively determined based on the actual requirements of coal mine gas control and technical feasibility, ensuring both coal mine safety production and improving the utilization rate of gas resources. The purpose of this step is to guide the optimization of drainage parameters through quantitative evaluation, form a closed-loop control, and continuously improve the gas drainage effect and utilization efficiency.

[0058] Specifically, the principle of the present invention is as follows: The technical principle of the present invention to solve the problem of unstable gas drainage concentration in coal seam groups is mainly reflected in three aspects: spatial layout optimization, concentration regulation mechanism, and intelligent prediction control. From the perspective of spatial layout, the Y-shaped borehole structure innovatively combines the main borehole and branch boreholes to form a three-dimensional drainage network. The main borehole is perpendicular to the coal seam to provide a stable drainage channel, and the two branch boreholes at a 45° angle extend upstream and downstream respectively, effectively covering the gob area and the gas enrichment areas of adjacent coal seams, solving the problem of insufficient coverage of traditional single boreholes.

[0059] In terms of the concentration regulation mechanism, the present invention designs a buffer system composed of a main buffer tank, an auxiliary buffer tank, and flow regulating valves. The physical mixing principle is used to buffer and regulate gas with different concentrations. When high-concentration gas enters the system, an appropriate amount of low-concentration gas is introduced through the regulating valve for dilution; when low-concentration gas enters, high-concentration gas is introduced for enrichment. Through this dynamic balance mechanism, the output gas concentration is maintained within a stable range of 30% to 45%, meeting the concentration requirements for resource utilization.

[0060] Intelligent predictive control is the core innovation of the present invention. The prediction model based on the long short-term memory neural network can capture the complex non-linear relationship between factors such as the advancing speed of the working face and the drainage negative pressure and the gas emission volume. The model adopts a multi-head self-attention mechanism to enhance the learning ability of long-term time-dependent relationships. By analyzing the time patterns and correlations in historical data, it can accurately predict the gas emission trend in the next 12 hours. The prediction results directly drive the dynamic regulation system of drainage parameters, realizing the advance adjustment of drainage negative pressure and flow rate, so that the drainage process can actively adapt to the upcoming gas emission changes, rather than the lagged response in traditional technologies, fundamentally improving the adaptability and control accuracy of the drainage system to gas emission changes.

[0061] The organic combination of these three technical principles constitutes a closed-loop gas drainage control system, comprehensively solving the technical problem of unstable gas drainage concentration in coal seam groups from aspects such as spatial coverage, concentration adjustment to predictive control, and realizing efficient, stable and intelligent gas drainage.

[0062] The following provides a specific Embodiment 1 of the present invention, and the specific implementation manners of each step in this Embodiment 1 are described in detail as follows.

[0063] The specific implementation manner of step S01 is to determine the coal seam distribution through ground penetrating radar detection technology and three-dimensional modeling technology. First, drill a main borehole above the coal seam. The main borehole is drilled perpendicular to the coal seam direction and reaches the coal seam to be mined. The borehole diameter is 120 mm, and the borehole wall is reinforced with high-strength steel pipes with a wall thickness of 10 mm. Then, drill two branch boreholes in the upstream and downstream directions respectively at the bottom of the main borehole. The inclination angle of the branch borehole with the horizontal plane is kept at 45°. The diameter of the branch borehole is 90 mm, and the length is determined according to the coal seam thickness, generally 3 to 5 times the coal seam thickness. Steel pipes are also laid in the branch boreholes for support. After the boreholes are completed, pressure tests are carried out to ensure the borehole sealing. When arranging the Y-shaped ventilation and drainage borehole group, the distance between adjacent main boreholes is 30 m, and a row of boreholes is arranged every 50 m along the coal seam strike to form a three-dimensional drainage network covering the entire mining area. The calculation formula for the length of the branch borehole is: L = k × H, where L is the length of the branch borehole in meters; k is the length coefficient with a value range of 3 - 5; H is the coal seam thickness in meters. The calculation formula for the coverage area of the Y-shaped borehole is: S = π × (L × sin(α)) 2 , where S is the coverage area of a single Y-shaped borehole in square meters; L is the length of the branch borehole in meters; α is the angle between the branch borehole and the horizontal plane, with a value of 45°. The implementation of this step is based on the principle of fluid mechanics, increasing the gas drainage coverage area through a three-dimensional cross borehole network and improving the gas drainage efficiency. The purpose of this step is to form a three-dimensional gas drainage channel network to create conditions for subsequent gas drainage.

[0064] The specific implementation of step S02 is to select catalytic combustion type gas sensors and thermal conductivity type gas sensors to construct a monitoring network. Install a catalytic combustion type gas sensor at the outlet of the Y-shaped borehole, with a measurement range of 0 to 5% and an accuracy of ±0.1%. Install thermal conductivity type gas sensors at key nodes of the main pipeline every 50 meters, with a measurement range of 0 to 100% and an accuracy of ±0.5%. At the same time, install negative pressure sensors at each monitoring point, with a measurement range of 0 to 25 kPa and an accuracy of ±0.2 kPa. All sensors adopt intrinsically safe design, with an explosion-proof grade of ExiaⅠ. The sensor acquisition frequency is set to once every 10 seconds. The sensors are connected to the nearest data collector through the RS485 bus, and the data collector transmits the acquired data to the monitoring center through the industrial Ethernet. During the sensor data acquisition process, the original data is filtered, and the digital low-pass filtering algorithm is adopted. The filtering function is: y(n) = α × x(n) + (1 - α) × y(n - 1), where y(n) is the current output value; x(n) is the current input value; y(n - 1) is the previous output value; α is the filtering coefficient, and its value range is 0 to 1. In this embodiment, α is taken as 0.3. During the data transmission process, the CRC check algorithm is adopted to ensure data integrity. The check function is: CRC(M) = R(M × x r modG(x)), where M is the data to be transmitted; G(x) is the generating polynomial; r is the order of the generating polynomial; R represents the remainder operation. The implementation of this step is based on sensor network technology and the principle of industrial Internet of Things, constructing a real-time monitoring network for gas concentration and negative pressure value, providing data support for the gas drainage process. The purpose of this step is to realize the real-time monitoring of gas concentration and negative pressure value, providing a data basis for subsequent dynamic adjustment.

[0065] The specific implementation of step S03 is to first establish a main buffer tank, which is made of Q345R steel, with a wall thickness of 20 mm, in a cylindrical shape, a diameter of 10 m, a height of 12.8 m, and a volume of 10 3 m 3 , with 5 layers of flow guiding plates installed inside. The flow guiding plates are arranged in an equally spaced spiral pattern, with an angle of 30°. The surface of the flow guiding plates is a hydrophilic material, which can adsorb the moisture in the gas. The auxiliary buffer tank also adopts Q345R steel, with a diameter of 8 m, a height of 10 m, and a volume of 5 × 10 2 m 3, with foamed plastic filler inside, the specific surface area reaches 1000 square meters per cubic meter, and the height of the filler layer is 5 meters. An electric control valve is installed between the main buffer tank and the auxiliary buffer tank. The valve is an angular travel electric control valve, with a maximum diameter of 800 mm, a rated pressure of 1.6 MPa, a regulation ratio of 100:1, and a switching time of less than 30 seconds. An on-line gas concentration analyzer is installed in the whole system, adopting the infrared absorption principle, with a measurement range of 0 to 100%, an accuracy of ±0.2%, and a response time of less than 10 seconds. When the monitored extracted gas concentration is lower than 30%, the control system automatically reduces the opening of the low-concentration intake valve and increases the opening of the high-concentration intake valve, and vice versa. The gas concentration after mixing is ensured to be maintained between 30% and 45% through flow ratio regulation. The regulation of the valve opening adopts the proportional integral derivative control algorithm, and the control function is: where u(t) is the control output; e(t) is the error signal, that is, the difference between the target concentration and the actual concentration; K p is the proportionality coefficient, with a value of 1.5; K i is the integral coefficient, with a value of 0.3; K d is the derivative coefficient, with a value of 0.05; t is the time variable; τ is the integral variable. The implementation of this step is based on the principles of fluid dynamics and automatic control. The gas concentration is stabilized through the buffer device and the automatic regulation system, solving the problem of gas concentration fluctuation. The purpose of this step is to stabilize the extracted gas concentration and create conditions for subsequent comprehensive utilization.

[0066] The specific implementation method of step S04 is to construct a mixing cavity, which is made of 304 stainless steel material, with a wall thickness of 15 mm and a volume of 5m 3 , and a turbine stirring device is installed inside. The diameter of the turbine is 0.8 m, and the rotation speed can be adjusted in the range of 0 to 500 revolutions per minute. 4 gas inlet pipes are arranged on the mixing cavity, which are respectively used to connect to the high-concentration gas source, the medium-concentration gas source and the low-concentration gas source, and there is also one for air dilution in case of emergency. An electric proportional valve is installed on each gas inlet pipe. The valve adopts a linear travel electric control valve, with a maximum diameter of 200 mm, a rated pressure of 1.0 MPa, and a regulation ratio of 50:1. A gas analyzer is installed at the outlet of the mixing cavity, adopting the thermal conductivity principle, with a measurement range of 0 to 100%, an accuracy of ±0.3%, and a response time of less than 5 seconds. The control system adopts the fuzzy PID control algorithm, adjusts the opening of each valve according to the real-time concentration data, keeps the mixing ratio of the low-concentration gas and the high-concentration gas constant at 4:6, and stabilizes the gas concentration after mixing within the required range for utilization. The calculation formula for the gas concentration after mixing is: where C mC is the gas concentration after mixing, in percentage; C1 is the concentration of low-concentration gas, in percentage; V1 is the volume flow rate of low-concentration gas, in cubic meters per hour; C2 is the concentration of high-concentration gas, in percentage; V2 is the volume flow rate of high-concentration gas, in cubic meters per hour. The fuzzy control rules adopt the IF-THEN structure. For example: IF (the current concentration is "low") AND (the concentration change rate is "negative") THEN (the opening of the high-concentration valve is "increased") AND (the opening of the low-concentration valve is "decreased"). The membership function of the fuzzy set adopts a combination of triangle and trapezoid. The fuzzy inference adopts the Mamdani algorithm, and the defuzzification adopts the centroid method. The calculation formula is: where u is the exact value after defuzzification; μ(x i ) is the membership degree; x i is the corresponding domain value; n is the number of discrete points. The implementation of this step is based on the mixed gas theory and fuzzy control technology. By precisely controlling the mixing ratio of gases with different concentrations, the stable control of gas concentration is achieved. The purpose of this step is to precisely control the gas concentration through intelligent mixing and gas distribution, and improve the gas utilization efficiency.

[0067] The specific implementation method of step S05 is to construct a dynamic regulation system for gas drainage parameters using a multi-input multi-output control system. The system receives four key parameters: the working face advancement speed data, the gas concentration data at the drill hole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data. The data is preprocessed by the data processing unit, including filtering, denoising, and normalization. Then the processed data is input into the fuzzy neural network controller, and the controller automatically calculates the optimal drainage negative pressure and flow rate parameters according to the preset fuzzy rule base. The normalization process adopts the maximum-minimum normalization method, and the calculation formula is: where x norm is the normalized value; x is the original value; x min is the minimum value of this parameter; x max is the maximum value of this parameter. The structure of the fuzzy neural network includes five parts: the input layer, the fuzzy layer, the rule layer, the defuzzification layer, and the output layer. Among them, the input layer has 4 neurons, corresponding to the four input parameters; the fuzzy layer performs fuzzy processing on the input parameters, converting the exact value into a fuzzy set. Each input parameter is divided into three fuzzy sets: "low", "medium", and "high"; the rule layer contains 3 4 = 81 neurons, corresponding to 81 IF-THEN rules; the defuzzification layer performs weighted averaging on the output of the rule layer to obtain an exact control quantity; the output layer has 2 neurons, respectively outputting the adjustment value of the drainage negative pressure and the adjustment value of the flow rate. The weight update of the fuzzy neural network adopts the backpropagation algorithm, the learning rate is set to 0.05, the momentum factor is set to 0.8, and the error function is the mean square error. The calculation formula is: where E is the error; y i is the expected output; is the actual output; m is the number of samples. The weight update formula is: where w(t + 1) is the updated weight; w(t) is the current weight; η is the learning rate; is the partial derivative of the error with respect to the weight; α is the momentum factor; Δw(t) is the current weight change. The implementation of this step is based on the fuzzy neural network control theory, which transforms expert experience into fuzzy rules and realizes the adaptive control of the system through the self-learning ability of the neural network. The purpose of this step is to dynamically adjust the extraction parameters according to real-time data and improve the extraction efficiency and safety.

[0068] The specific implementation of step S06 is to construct a gas emission prediction model based on the long short-term memory neural network. The model adopts a hierarchical structure, including four functional modules: a data preprocessing unit, a feature extraction unit, a time series learning unit, and a prediction unit. The data preprocessing unit normalizes the input data of the working face advance speed, the gas concentration at the drill hole outlet, the gas concentration at the key nodes of the main pipeline, and the negative pressure value. The Z-score method is used to make the data mean 0 and the standard deviation 1. The calculation formula is: where z is the normalized value; x is the original value; μ is the mean of this parameter; σ is the standard deviation of this parameter. At the same time, wavelet transform is used for signal denoising. The wavelet transform formula is: where W f (a, b) are the wavelet transform coefficients; f(t) is the original signal; ψ * is the conjugate of the wavelet function; a is the scale parameter; b is the translation parameter; t is the time variable. The feature extraction unit adopts a bidirectional long short-term memory network structure, including two layers each of the forward propagation layer and the backward propagation layer, with 128 neurons in each layer. The tanh activation function is used to capture the context features of the time series data in both the forward and backward directions. The core of the long short-term memory unit is three gate structures: the forget gate, the input gate, and the output gate. Their calculation formulas are:

[0069] f t = σ(W f · [h t-1 , x t + b f )

[0070] i t = σ(W i · [h t-1 , x t + b i )

[0071]

[0072] o t = σ(W o · [h t-1 , x t + b o )

[0073] h t = o t × tanh(C t )

[0074] where f t is the output of the forget gate; i t is the output of the input gate; is the candidate memory cell; C t is the current memory cell; o t is the output of the output gate; h t is the hidden state; x t is the current input; h t-1 is the hidden state at the previous time step; C t-1 is the memory cell at the previous time step; W f , w i , W C , W o are weight matrices; b f , b i , b C , b o are bias terms; σ is the sigmoid activation function; tanh is the hyperbolic tangent activation function. The time series learning unit adopts a multi-head self-attention mechanism with the number of heads set to 8 and the attention dimension to 64. Combining residual connections and layer normalization improves the model stability. By calculating the correlation weights between different time steps, the model's ability to capture long time series dependencies is enhanced. The calculation formula of the self-attention mechanism is:

[0075] where Q is the query matrix; K is the key matrix; V is the value matrix; d k is the dimension of the key vector; softmax is the softmax normalization function. The prediction unit consists of two fully connected layers. The first layer contains 64 neurons and uses the ReLU activation function. The output dimension of the second layer is the same as the prediction target, and it outputs the predicted values of the gas emission volume per hour and the predicted values of the concentration change trend within the next 12 hours. The training of the model uses the backpropagation algorithm, the loss function is the mean squared error, the optimization algorithm is the Adam optimizer, and the initial value of the learning rate is 0.001, which decays exponentially as the number of training rounds increases, with a decay rate of 0.95. The implementation of this step is based on deep learning and time series prediction theory. By analyzing historical data to mine the gas emission pattern, the early prediction of gas emission and the pre-adjustment of extraction parameters are realized. The purpose of this step is to achieve the early prediction of gas emission and provide decision support for the dynamic adjustment of extraction parameters.

[0076] The specific implementation of step S07 is to establish a three - level hierarchical utilization system based on the gas concentration. First, install an on - line gas concentration analysis system. Use an infrared absorption type gas analyzer with a measurement range of 0 to 100% and an accuracy of ±0.2%, and a sampling frequency of once every 5 seconds. The system divides the gas into three grades according to the real - time measured gas concentration: the gas concentration in the high - concentration area is greater than 30%, the gas concentration in the medium - concentration area is between 10% and 30%, and the gas concentration in the low - concentration area is between 5% and 10%. Install an electric three - way diverter valve in the pipeline system. The valve is an angular travel electric control valve with a maximum diameter of 400 mm, a rated pressure of 1.0 MPa, a regulation ratio of 50:1, and a response time of less than 5 seconds. The control system automatically adjusts the direction of the three - way valve according to the gas concentration and guides the gas with different concentrations into the corresponding utilization systems. The calculation formula for the energy conversion efficiency of hierarchical utilization is: where η i is the energy conversion efficiency of the i - th level utilization system; E out，i is the output energy, with the unit of megajoule; E in，i is the input energy, with the unit of megajoule. The calculation formula for the input energy is: E in，i = V i ×C i ×Q, where V i is the volume of the i - th level gas, with the unit of cubic meter; C i is the concentration of the i - th level gas, with the unit of percentage; Q is the calorific value of pure methane, taking 35.9 megajoules per cubic meter. The gas in the high - concentration area is directly guided into a gas power plant or a liquefaction device, and the power generation efficiency is not less than 40%. The gas in the medium - concentration area is guided into a catalytic oxidation device. The oxidation process is carried out under the condition of a temperature of 350°C, using a palladium - platinum catalyst with a catalyst loading of 1.5 kg / m³, and the catalytic conversion rate is greater than 95%. The heat energy released by oxidation is used for mine heating or hot water supply, and the heat energy recovery efficiency is not less than 70%. The gas in the low - concentration area is guided into a thermal oxidation device, and complete oxidation is carried out under the condition of a temperature of 850°C, with an oxidation efficiency greater than 99%, and the methane content in the exhaust gas is less than 0.1%, meeting the environmental protection requirements, and the heat energy recovery efficiency is not less than 60%. The implementation of this step is based on the theory of hierarchical utilization of gas resources. Different utilization methods are adopted according to the characteristics of gas with different concentrations, improving the comprehensive utilization rate of gas. The purpose of this step is to achieve the cascade utilization of gas resources, improve the comprehensive utilization efficiency of gas, and reduce greenhouse gas emissions.

[0077] The specific implementation of step S08 is to build an intelligent monitoring platform for gas drainage based on the industrial Internet of Things architecture. The platform consists of four parts: the perception layer, the network layer, the platform layer, and the application layer. The perception layer is composed of sensors distributed at each monitoring point, including gas concentration sensors, negative pressure sensors, flow sensors, and status sensors, etc. The sampling frequency is once per second. The network layer adopts a redundant star network topology. The main network uses industrial Ethernet with a transmission rate of 1000 megabits per second, and the backup network uses industrial wireless network with a transmission rate of 100 megabits per second. The network latency is less than 10 milliseconds. The platform layer adopts a distributed architecture, including a data acquisition server, a data storage server, and an application server. The servers adopt a dual-machine hot backup mode. The data storage uses a time-series database, supporting a data writing speed of 10,000 points per second and a query response time in milliseconds. The application layer includes a data processing module, a visualization display module, an early warning module, and a decision support module. The data processing module adopts a stream processing architecture, supporting real-time data analysis. The visualization display module is implemented using Web technology, supporting multi-terminal access. The early warning module is implemented based on a rule engine, including 100 early warning rules, covering situations such as abnormal gas concentration, abnormal negative pressure, and abnormal flow. The decision support module is based on case-based reasoning technology, storing 500 historical cases, and can match similar cases according to the current situation and provide decision-making suggestions. The system reliability calculation formula is: R(t) = e -λt , where R(t) is the reliability of the system within time t; λ is the system failure rate, with the unit of times per hour; t is the running time, with the unit of hours. The system availability calculation formula is: where A is the system availability; MTBF is the mean time between failures, with the unit of hours; MTTR is the mean time to repair, with the unit of hours. The MTBF of this system is not less than 10,000 hours, the MTTR is not higher than 2 hours, and the availability is not less than 99.98%. The implementation of this step is based on the industrial Internet of Things and big data analysis technologies, building an intelligent monitoring system for gas drainage, and realizing real-time monitoring, analysis, and early warning of data. The purpose of this step is to achieve intelligent monitoring of the gas drainage process and improve the safety and reliability of system operation.

[0078] The specific implementation of step S09 is to evaluate the gas drainage effect through three aspects of measurement and calculation. First, measure the residual gas content. Using the borehole sampling method, a measuring point is arranged every 20 meters along the strike of the working face and every 10 meters along the dip to form a measuring point network. A sampling borehole is drilled at each measuring point, with a depth of 8 meters and a diameter of 42 millimeters. A sampler is placed in the borehole. The sampler is a sealed container with a volume of 500 cubic centimeters. The gas in the coal sample is extracted by a vacuum pump to measure the gas content. Then calculate the drainage rate, using the formula where η is the extraction rate; q0 is the original gas content in cubic meters per ton; q1 is the residual gas content in cubic meters per ton. Finally, the gas utilization rate is calculated using the formula where ξ is the gas utilization rate; V1 is the actual gas utilization volume in cubic meters; V0 is the total extracted gas volume in cubic meters. The comprehensive evaluation of the extraction effect adopts the weighted scoring method, and the calculation formula is: S = w1×η + w2×ξ + w3×γ, where S is the comprehensive score; η is the extraction rate; ξ is the gas utilization rate; γ is the safety index, and the calculation formula is N is the number of gas overlimit times after extraction implementation, and N0 is the number of gas overlimit times before extraction; w1, w2, and w3 are the weight coefficients, taking 0.4, 0.3, and 0.3 respectively. According to the evaluation results, when the extraction rate is lower than 70% or the gas utilization rate is lower than 80%, the system automatically optimizes the extraction negative pressure and flow parameters, and improves the extraction effect by increasing the extraction negative pressure or adjusting the extraction time. The implementation of this step is based on the scientific evaluation and closed-loop optimization principle. By quantitatively evaluating the extraction effect, it provides a basis for the optimization of extraction parameters. The purpose of this step is to guide the optimization of extraction parameters through quantitative evaluation, form a closed-loop control, and continuously improve the gas extraction effect and utilization efficiency.

[0079] Optionally, in the prior art, the Y-shaped ventilation extraction borehole group is a special borehole structure pre-arranged along the front of coal seam mining. It consists of a main borehole and two branch boreholes, resembling the English letter Y. The main borehole drills vertically in the coal seam direction and reaches the coal seam to be mined. The two branch boreholes extend from the bottom of the main borehole in the upstream and downstream directions respectively, with an inclination angle of 45°, covering the goaf and adjacent coal seams, forming a three-dimensional extraction network. The advantage of the Y-shaped borehole is that its three-dimensional structure can increase the contact area with the coal seam, improve the gas extraction efficiency, and at the same time reduce the number of boreholes and lower the construction cost. The calculation formula for the penetration radius of the Y-shaped borehole is: where R is the penetration radius in meters; k is the coal seam permeability in millidarcies; ΔP is the pressure difference in Pascals; t is the extraction time in seconds.

[0080] The following provides Example 2 of a specific application scenario of the present invention: The application practice of the Y-shaped ventilation gas extraction method was carried out in a large coal mine working face. The coal mine is located in the deep mining area, with an average coal seam thickness of 3.5 meters and an original gas content as high as 15.8 cubic meters per ton, belonging to a high-gas mine. The working face length is 240 meters, and the advancing speed is 5.2 meters per day on average. Due to the large and unstable gas emission in the working face, which poses a serious threat to the safety of coal mining operations, the researchers decided to implement the Y-shaped ventilation gas extraction method, as Figure 2 shown, to solve the gas control problem through systematic design and implementation.

[0081] During the implementation process, first, according toFigure 3 The shown design arranges a Y-shaped ventilation and drainage borehole group. A total of 8 main boreholes are arranged, one every 30 meters along the working face strike. The main boreholes are drilled perpendicular to the coal seam direction, with a depth of 150 meters, a diameter of 120 mm, and the borehole wall is reinforced with high-strength steel pipes with a wall thickness of 10 mm. Two branch boreholes are drilled from the bottom of each main borehole in the upstream and downstream directions respectively, with an inclination angle of 45°, a branch borehole diameter of 90 mm, and a length of 15 meters, covering the goaf and adjacent coal seams. The borehole layout parameters are shown in Table 1:

[0082] Table 1 Y-shaped Borehole Layout Parameter Table

[0083]

[0084]

[0085] Gas concentration monitoring sensors and negative pressure sensors are installed at the borehole outlets and key nodes of the main pipeline every 50 meters. A total of 16 gas concentration sensors and 8 negative pressure sensors are installed. The gas concentration sensors combine catalytic combustion type and thermal conductivity type, with a measurement range of 0 to 100%, an accuracy of ±0.2%, and a collection frequency of once every 10 seconds. The negative pressure sensor has a measurement range of 0 to 25 kPa and an accuracy of ±0.2 kPa. The actual monitoring data are shown in Table 2:

[0086] Table 2 Gas Concentration and Negative Pressure Monitoring Data at Borehole Outlets (Partial)

[0087] Time Drilling number Gas concentration (%) Negative pressure value (kPa) <![CDATA[Flow rate (m 3 / min)]]> 2025-01-10 08:00 YZK-01 42.5 18.6 85.3 2025-01-10 08:00 YZK-02 38.7 17.8 78.4 2025-01-10 08:00 YZK-03 45.2 19.2 92.7 2025-01-10 08:00 YZK-04 40.1 18.5 83.2 2025-01-10 14:00 YZK-01 36.8 17.5 76.3 2025-01-10 14:00 YZK-02 33.2 16.9 72.8 2025-01-10 14:00 YZK-03 39.6 18.3 85.1 2025-01-10 14:00 YZK-04 35.4 17.2 77.5 2025-01-10 20:00 YZK-01 44.3 19.4 89.7 2025-01-10 20:00 YZK-02 41.5 18.7 84.6 2025-01-10 20:00 YZK-03 47.8 20.2 96.4 2025-01-10 20:00 YZK-04 43.2 19.1 87.5

[0088] According to Figure 4 shown, a gas concentration regulation and buffer system is constructed, including 1 main buffer tank and 1 auxiliary buffer tank. The main buffer tank is made of Q345R steel, with a wall thickness of 20 mm, a diameter of 10 meters, a height of 12.8 meters, and a volume of 10 3 m 3 . The auxiliary buffer tank is also made of Q345R steel, with a diameter of 8 meters, a height of 10 meters, and a volume of 5×10 2 m 3 . The key parameters of the buffer system are shown in Table 3:

[0089] Table 3 Gas Concentration Regulation and Buffer System Parameter Table

[0090] Parameter name Main buffer tank Auxiliary buffer tank Material Q345R steel Q345R steel Wall thickness (mm) 20 18 Diameter (m) 10 8 Height (m) 12.8 10 <![CDATA[Volume (m 3 )]]> <![CDATA[10 3 > <![CDATA[5×10 2 > Internal structure 5 layers of flow guiding plates Foamed plastic filler <![CDATA[Specific surface area of filler (m 2 / m 3 )]]> - 1000

[0091] As Figure 5 shown, an intelligent gas mixing device is installed, which mixes low-concentration gas and high-concentration gas in a ratio of 4:6 through the negative pressure drainage system to keep the gas concentration in the pipeline network stable between 30% and 45%. The mixing chamber is made of 304 stainless steel, with a wall thickness of 15 mm and a volume of 5m3 , with a turbine stirring device installed inside, whose rotation speed can be adjusted within the range of 0 to 500 revolutions per minute. The intelligent mixing and gas distribution effect is shown in Table 4:

[0092] Table 4 Comparison Table of Gas Concentration before and after Mixing

[0093]

[0094]

[0095] As Figure 6 shown, a gas emission prediction model based on long short-term memory neural network is established, and the gas emission volume and concentration change trend in the working face in the next 12 hours are predicted by analyzing historical data. The model training uses 90 days of historical data, including the working face advance speed data, the gas concentration data at the drill hole outlet, the gas concentration data at the key nodes of the main pipeline, and the negative pressure value data. The performance evaluation results of the prediction model are shown in Table 5:

[0096] Table 5 Performance Evaluation Table of Prediction Model

[0097] Evaluation index Training set Validation set Test set Root mean square error (RMSE) 1.35 1.62 1.78 Mean absolute error (MAE) 0.94 1.15 1.23 <![CDATA[Coefficient of determination (R 2 )]]> 0.921 0.893 0.875 1-hour prediction accuracy rate (%) 95.8 93.6 92.1 6-hour prediction accuracy rate (%) 87.5 84.2 82.6 12-hour prediction accuracy rate (%) 78.3 75.4 73.8

[0098] As Figure 7 shown, a gas concentration classification utilization system is implemented, and the gas is classified into three levels according to the concentration for utilization. The gas in the high concentration area (>30%) is directly used for power generation; the gas in the medium concentration area (10% - 30%) is converted into heat energy through a catalytic oxidation device; the gas in the low concentration area (5% - 10%) is discharged after being treated by a thermal oxidation technology. The utilization effects of the three different concentration gases are shown in Table 6:

[0099] Table 6 Utilization Effect Table of Gases with Different Concentrations

[0100]

[0101] As Figure 8 shown, a gas drainage intelligent monitoring platform is deployed, integrating drainage negative pressure data, flow data, gas concentration data at the drill hole outlet, gas concentration data at the key nodes of the main pipeline, and system status data to achieve remote monitoring and fault warning. The platform includes four parts: the perception layer, the network layer, the platform layer, and the application layer, realizing real-time monitoring and intelligent warning of the whole system. The key performance indicators of this monitoring platform are shown in Table 7:

[0102] Table 7 Performance Index Table of Intelligent Monitoring Platform

[0103] Index name Index value Technical standard System reliability (MTBF, hours) 12580 >10000 System availability (%) 99.986 >99.98 Response time (ms) 356 <1000 Data storage period (days) 365 >180 Number of early warning rules (pieces) 128 >100 Capacity of historical case library (pieces) 625 >500

[0104] In the link of evaluating the effect of gas drainage, the residual gas content is measured by the borehole sampling method, and the drainage rate and gas utilization rate are calculated. The residual gas content at 8 measuring points in the working face is tested, and the calculated drainage effect is shown in Table 8:

[0105] Table 8 Evaluation Table of Gas Drainage Effect

[0106]

[0107]

[0108] Traditional gas drainage methods usually adopt single vertical borehole or horizontal borehole drainage technology. The distribution of drainage boreholes is uneven, and the drainage rate is low, generally only reaching 45% - 55%. At the same time, traditional methods lack intelligent prediction and control means, it is difficult to cope with the change of gas emission volume, and there are potential safety hazards. The unstable gas concentration also leads to low comprehensive utilization efficiency, and the gas utilization rate is usually only about 60%.

[0109] The Y-shaped ventilation gas drainage method of the present invention has significant progress compared with traditional methods: First, the Y-shaped borehole structure increases the contact area with the coal seam, improves the drainage coverage rate, and the drainage rate is increased from about 50% of the traditional method to 77.5%, an increase of about 55%. Second, the gas concentration adjustment buffer system and intelligent gas mixing device achieve precise control of the gas concentration, maintaining it within the range of 30% - 45% required for utilization, and greatly improving the gas utilization rate to 89.3%, an increase of nearly 50% compared with traditional methods. Third, the prediction control model based on the long short-term memory neural network realizes the advance prediction of gas emission, and the 12-hour prediction accuracy reaches 73.8%, providing a scientific basis for the dynamic adjustment of drainage parameters and significantly improving the system safety.

[0110] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 9 and 10 below.

[0111] Table 9 Explanation Table of Variables (First Part)

[0112]

[0113]

[0114] Table 10 Explanation Table of Variables (Second Part)

[0115]

[0116] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A Y-type ventilation gas extraction method for a close-range coal seam group, characterized in that: include: Lay out Y-shaped ventilation and extraction borehole groups; install a gas concentration monitoring sensor network; build a gas concentration adjustment buffer system; install an intelligent mixed gas distribution device; establish a dynamic control system for gas extraction parameters; apply a predictive control model, analyze the gas outburst law of the working face based on historical data, receive working face advancement speed data, borehole outlet gas concentration data, main pipeline key node gas concentration data and negative pressure value data, output the predicted value of gas outburst volume and concentration change trend in the next 12 hours, and adjust the extraction negative pressure and flow rate 12 hours in advance; implement a gas concentration graded utilization system; Deploy a gas extraction intelligent monitoring platform; perform gas extraction effect evaluation.

2. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 1 is characterized in that: The Y-shaped ventilation and extraction drilling hole group is a special drilling hole structure pre-arranged at the front of coal seam mining, consisting of a main drilling hole and two branch drilling holes. The main drilling hole is drilled vertically in the direction of the coal seam to a depth reaching the coal seam to be mined. The two branch drilling holes extend from the bottom of the main drilling hole in the upstream and downstream directions respectively, with an inclination of 45°, covering the goaf and adjacent coal seams to form a three-dimensional extraction network.

3. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 2 is characterized in that: The gas concentration monitoring sensor network specifically sets a monitoring point every 50 meters at the borehole outlet and the key nodes of the main pipeline to collect real-time gas concentration data at the borehole outlet, gas concentration data at the key nodes of the main pipeline and negative pressure value data.

4. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 3 is characterized in that: The gas concentration regulating buffer system is specifically composed of a main buffer tank, an auxiliary buffer tank, a flow regulating valve, a pressure sensor and a concentration sensor. The main buffer tank is a cylindrical structure with a volume of 10 3 m 3 , the volume of the auxiliary buffer tank is 5×10 2 m 3 , electric regulating valves are installed at both ends.

5. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 4 is characterized in that: The intelligent mixing and distributing device is specifically composed of a mixing chamber, a multi-way air inlet pipe, an electric proportional valve, a gas analyzer and a controller. The mixing chamber is a closed container with a volume of 5m 3 A turbine stirring device is installed inside, and multiple air inlet pipes are respectively connected to high-concentration gas sources, medium-concentration gas sources and low-concentration gas sources.

6. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 5 is characterized in that: The gas extraction parameter dynamic control system automatically adjusts the extraction negative pressure and flow rate based on the working face advancement speed data, the borehole outlet gas concentration data, the main pipeline key node gas concentration data and the negative pressure value data.

7. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 6 is characterized in that: The predictive control model is specifically a gas outburst prediction model constructed based on a long short-term memory neural network, which includes three parts: an input layer, a hidden layer and an output layer. The input layer receives working face advancement speed data, borehole outlet gas concentration data, main pipeline key node gas concentration data and negative pressure value data, and the hidden layer includes four layers of long short-term memory units, with 128 neurons in each layer.

8. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 7 is characterized in that: The specific structure of the predictive control model is a multivariable time series prediction network structure, which consists of a data preprocessing unit, a feature extraction unit, a time series learning unit and a prediction unit. The feature extraction unit adopts a bidirectional long short-term memory network structure, and the time series learning unit adopts an attention mechanism to enhance the model's ability to capture long time series dependencies.

9. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 8 is characterized in that: The steps of establishing the training data set of the predictive control model specifically include four parts: data collection stage, data cleaning stage, feature extraction stage and data set division stage. The data collection stage obtains 90 consecutive days of historical monitoring data from the gas concentration monitoring sensor network.

10. The Y-type ventilation gas extraction method for close-range coal seam groups according to claim 9 is characterized in that: The temporal learning unit of the predictive control model, specifically the attention layer, is composed of a multi-head self-attention module, the number of heads is set to 8, the attention dimension is 64, and the residual connection and layer normalization are combined to improve the stability of the model. The prediction unit is composed of two fully connected layers, and the first layer contains 64 neurons.

Images