Space-time prediction model-based closed-loop mine atmosphere intelligent control system and method

Abstract

The invention discloses a closed-loop mine atmosphere intelligent control system and method based on a space-time prediction model. The closed-loop mine atmosphere intelligent control system comprises a sensing layer which is used for collecting multi-dimensional data including atmospheric environment parameters and mine operation data in real time; the artificial intelligence processing core is in data communication with the sensing layer, and the artificial intelligence processing core is configured to receive multi-dimensional data collected by the sensing layer; utilizing a space-time prediction model to predict the three-dimensional space distribution and migration track of the to-be-measured object in the atmosphere in a preset time period in the future according to the current and historical data; based on the prediction result, generating a preemptive control instruction; the execution layer is in instruction communication with the artificial intelligence processing core, the execution layer is configured to receive and execute the preemptive control instruction, and before the prediction risk of the to-be-tested object reaches a critical threshold value, physical intervention is carried out on a prediction migration path of the to-be-tested object; the problems that an existing mine atmosphere monitoring system is lagged in response, sensing and action are disjointed, and the control strategy is extensive are solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of mine safety combined with artificial intelligence application, and particularly relates to a closed-loop mine atmosphere intelligent control system and method based on a space-time prediction model. BACKGROUND

[0002] The underground operation environment of a coal mine is complex, and dust and gas are two major disaster sources that have long threatened the life safety and health of employees. With the development of technology, the mine safety monitoring system has developed from traditional manual inspection to digital and information-based intelligent monitoring.

[0003] Most current intelligent monitoring systems use various sensors to monitor environmental parameters in real time and use video monitoring systems combined with artificial intelligence algorithms to identify the existence of dangerous sources or the illegal operation of underground personnel. These systems can provide timely alarms when disasters occur and have made significant progress in dangerous source identification. However, they are essentially a passive, "after-the-fact" or "in-the-middle" response mode. That is, the system triggers an alarm or control logic only after monitoring dangerous parameters such as dust or gas concentration reaching or exceeding a preset threshold. At the same time, in the implementation of disaster prevention, such as the dust spray system or ventilation system in the mine, the operation logic is also relatively simple. The spray system is usually started and stopped according to a fixed schedule or based on the concentration exceeding a certain point, lacking prediction of the dynamic evolution trend of dust clouds or gas clouds. The adjustment of the ventilation system also relies on manual experience or preset schemes, making it difficult to respond to the rapidly changing local atmospheric environment in real time and accurately.

[0004] Therefore, how to fundamentally change the passive situation of mine atmospheric disaster prevention and build a closed-loop mine atmospheric intelligent control system and method based on a space-time prediction model that can predict risks, adaptively adjust, and preemptively intervene is a technical problem that needs to be solved in the field. SUMMARY

[0005] The present application is aimed at the above problems, fills the gaps in the prior art, and provides a closed-loop mine atmospheric intelligent control system and method based on a space-time prediction model to solve the problems of response lag, disconnection between sensing and action, and extensive control strategy of existing mine atmospheric monitoring systems, thereby achieving active prediction and preemptive intervention of atmospheric disaster risks such as dust and gas.

[0006] To achieve the above purpose, the present application adopts the following technical solutions.

[0007] A closed-loop mine atmospheric intelligent control system based on a space-time prediction model includes a sensing layer, an artificial intelligence processing core, and an execution layer. The sensing layer: configured in the key area of the mine, used for real-time collection of multi-dimensional data including at least atmospheric environmental parameters and mine operation data; The artificial intelligence processing core: deployed in the server and in data communication with the sensing layer; the artificial intelligence processing core is built-in a space-time prediction model, and the artificial intelligence processing core is configured to receive the multi-dimensional data collected by the sensing layer; The space-time prediction model is trained to receive and process real-time and historical data from the sensing layer, and by using the space-time prediction model, according to the current and historical data such as the mine operation data of the coal mining machine position, speed and the state of the fan, the three-dimensional space distribution, concentration and migration track of the to-be-measured objects such as dust cloud and gas mass in the atmosphere in a future preset time period are predicted, and based on the prediction result, the optimal preemptive control instruction is generated; The execution layer: distributed in the mine and in instruction communication with the artificial intelligence processing core; the execution layer is configured to receive and execute the preemptive control instruction from the artificial intelligence processing core, and to physically intervene in the predicted migration path of the to-be-measured object before the predicted risk of the to-be-measured object reaches the critical threshold, so as to intervene in the mine atmospheric environment.

[0008] Further, the sensing layer at least includes one or more combinations of the following: an optical particle counter for monitoring dust concentration, a tunable diode laser absorption spectrum sensor for measuring gas concentration, and a camera network for acquiring live video stream; wherein the measured gas is methane.

[0009] Further, the space-time prediction model is a GNN graph neural network model or a 3D ConvLSTM three-dimensional convolutional long short-term memory network model.

[0010] Further, the mine operation data at least includes the position, speed of the coal mining machine, and the running state of the fan.

[0011] Further, the execution layer at least includes one or more combinations of the following: a plurality of independently controllable spray nozzles with variable flow and variable atomization form, a plurality of automatically adjustable ventilation regulating valves, and a plurality of local fans.

[0012] A closed-loop mine atmospheric intelligent control method based on a space-time prediction model, which is implemented by using the closed-loop mine atmospheric intelligent control system based on the space-time prediction model described above, and includes the following steps: Step 1, through the sensing layer, continuously collecting real-time dust concentration, gas concentration, atmospheric environmental parameters of live video images, and mine operation data of coal mining machine position, speed and fan state from each monitoring point in the mine.

[0013] Step 2, the artificial intelligence processing core receives the data collected in step 1 and inputs it into a pre-trained spatiotemporal prediction model; the spatiotemporal prediction model performs fusion analysis on the received data to predict the three-dimensional spatial distribution and migration trajectory of the object to be measured in the atmosphere in a future preset time period, such as the concentration distribution, diffusion path and evolution trend of a dust cloud or a gas cloud in the three-dimensional space of a mine.

[0014] Step 3, the artificial intelligence processing core calculates an optimal preemptive intervention strategy in real time according to the prediction result of step 2, in combination with a mine ventilation dynamics model and a preset prevention and control target; the optimal preemptive intervention strategy includes the execution units that need to be started, the action parameters such as the spray flow, the atomization angle, the ventilation valve opening degree of each unit and the execution time.

[0015] Step 4, the artificial intelligence processing core converts the optimal preemptive intervention strategy generated in step 3 into specific control instructions and sends them to the execution layer, which executes the control instructions before the predicted risk of the object to be measured reaches a critical threshold; specifically, the execution layer starts and accurately adjusts the corresponding spray nozzles, ventilation valves or local fans in advance according to the received control instructions before the predicted disaster risk reaches a critical state, to accurately intervene in the predicted path of the dust cloud or gas cloud.

[0016] Further, the future preset time period predicted is 30 to 60 seconds.

[0017] Further, the step of calculating the optimal preemptive intervention strategy includes calculating with the optimization objectives of minimizing the prediction risk and minimizing resource consumption.

[0018] Further, the closed-loop mine atmosphere intelligent control method based on the spatiotemporal prediction model further includes a feedback optimization step 5, in which the system continuously repeats steps 1 to 4 to form a dynamic closed-loop control; after executing the control instructions, the feedback data of the intervention effect is collected by the sensing layer, and the feedback data is used to iteratively optimize the spatiotemporal prediction model and the control strategy.

[0019] Further, in the execution step of step 4, the physical intervention method for the object to be measured includes at least one of the following: deploying a water mist curtain on its predicted migration path, or adjusting the local ventilation of its predicted influence area to form an air curtain for guidance or blocking.

[0020] Compared with the prior art, the present application has the following remarkable beneficial effects: Transforming passive response into proactive prevention: By introducing a spatiotemporal prediction model, this invention elevates its core capability from "monitoring the current situation" to "predicting the future." It can predict the development trend of dust or gas 30-60 seconds before a high concentration of danger is formed, thereby transforming traditional delayed and remedial prevention and control into forward-looking and preventive intervention, fundamentally improving safety redundancy.

[0021] In summary, the present invention has at least one of the following beneficial technical effects: (1) Realizing an intelligent closed loop of sensing and action: This invention deeply integrates three links: distributed sensing, artificial intelligence prediction, and precise execution, to construct a complete "perception-cognition-decision-action" closed loop. The system can autonomously and collaboratively mobilize multiple hardware systems based on real-time dynamics and future predictions, thus resolving the problem of the separation between monitoring and prevention in traditional systems.

[0022] (2) Improve intervention efficiency and resource utilization: Based on the accurate prediction of the spatiotemporal evolution path of disaster sources, the system can calculate the optimal intervention plan, such as accurately deploying water mist on the inevitable path of dust clouds, or using ventilation to form an air curtain to suppress gas diffusion, avoiding the traditional extensive, long-term, and high-intensity extensive prevention and control measures, significantly saving water, electricity and other resources, and achieving the maximum prevention and control effect.

[0023] (3) Promoting a new stage of autonomous mine safety: The predictive safety system proposed in this invention transforms the role of human operators from real-time controllers to system supervisors, greatly reducing reliance on manpower and the risk of human error. It is one of the indispensable core technologies for realizing full automation and unmanned operation in mines, marking a leap in mine safety concepts from "digital monitoring" to "autonomous and predictive safety". Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the architecture of a closed-loop intelligent control system for mine atmosphere based on a spatiotemporal prediction model provided in an embodiment of the present invention.

[0025] Figure 2 This is a detailed schematic diagram of a closed-loop intelligent control system for mine atmosphere based on a spatiotemporal prediction model provided in an embodiment of the present invention.

[0026] Figure 3 This is a flowchart of a closed-loop intelligent atmospheric control method for mines based on a spatiotemporal prediction model, provided by an embodiment of the present invention.

[0027] Marked in the figure: 100, sensing layer; 200, artificial intelligence processing core; 300, execution layer; 101, optical particle counter; 102, TDLAS tunable diode laser absorption spectrum methane sensor; 103, camera network; 104, operation data interface; 201, space-time prediction model; 202, optimal control strategy generation module; 301, intelligent spray nozzle; 302, automatic ventilation adjusting device. DETAILED DESCRIPTION

[0028] In order to make the technical problems solved by the present application, technical solutions and beneficial effects more clearly understood, the present application will be further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0029] Please refer to Figure 1 and Figure 2 The embodiment provides a closed-loop mine atmosphere intelligent control system based on a space-time prediction model; the control system is deployed at a coal mine underground working face, especially an excavation or fully mechanized working face, and aims to predictively and cooperatively control dust and gas.

[0030] The control system comprises a sensing layer 100, an artificial intelligence processing core 200 and an execution layer 300.

[0031] Sensing layer 100: distributed grid deployment along the roadway and working face.

[0032] Optical particle counter 101: used for real-time monitoring of dust concentration in different particle sizes in the air, such as PM2.5 and PM10. Its high time resolution reading provides an instant data source for dust cloud generation for the model.

[0033] TDLAS tunable diode laser absorption spectrum methane sensor 102: used for high-precision and high-selectivity measurement of methane concentration, which is not interfered by other gases and dust, and provides reliable gas mass data input for the space-time prediction model 201.

[0034] Camera network 103: covers key areas such as coal winning machines, transfer points, etc. Its video stream can be used for traditional visual monitoring on the one hand, and its image data can be processed and used as one of the inputs of the model, for example: through image analysis to identify the source, form and initial diffusion speed of dust.

[0035] In addition, the sensing layer 100 also integrates an operation data interface 104 for real-time acquisition of data such as the position of the coal winning machine, the coal cutting speed, the cutting direction, and the running state of the main and auxiliary ventilators. These data are crucial for predicting the generation and migration of disaster sources.

[0036] AI processing core 200: can be deployed on the central server of the mine ground dispatch center or the high-performance edge computing unit underground to ensure low-latency response.

[0037] The key of this core is the spatio-temporal prediction model 201. In this embodiment, a 3D ConvLSTM three-dimensional convolutional long short-term memory network can be used. The mine roadway space is virtually divided into a three-dimensional grid, and the parameters such as dust, gas concentration, and wind speed in each grid constitute a time series tensor. The 3D ConvLSTM network can effectively learn the complex dynamic relationship in this spatio-temporal data, that is, not only can it learn the change rule of the concentration of each grid point over time through the ability of LSTM, but also can learn the spatial mutual influence between adjacent grid points through the ability of 3D convolution.

[0038] The training data of the spatio-temporal prediction model 201 comes from the historical sensor data, operation data accumulated for a long time in the mine, and the corresponding disaster evolution process record. Through supervised learning, the model learns to predict the dust and gas concentration values of each grid in the entire three-dimensional space within 30-60 seconds in the future according to the current sensor readings, coal mining machine state, etc.

[0039] The AI processing core 200 also includes an optimal control strategy generation module 202. This module receives the output of the spatio-temporal prediction model 201, that is, the future disaster spatio-temporal distribution map, and takes it as input. It runs an optimization algorithm such as reinforcement learning or operations research algorithm, aiming to minimize the consumption of water, electricity, etc. under the premise of meeting the safety constraints that the predicted concentration of all areas is below the threshold. Its output is a set of specific, time-stamped control instructions.

[0040] Execution layer 300: includes intelligent spray nozzles 301 and automatic ventilation adjustment devices 302.

[0041] Intelligent spray nozzles 301: These nozzles are different from traditional on-off nozzles. They are built-in with electronic control modules and precision valves, and can receive digital signals from the AI processing core 200. The instructions can accurately control their opening / closing, spray flow, and spray cone angle and direction.

[0042] Automatic ventilation adjustment devices 302: include automatic air windows, air doors in the roadway, and local fans in key branches. These devices can also be remotely digitally controlled to change the air flow path or wind speed in a specific area for guiding, diluting, or blocking gas clusters.

[0043] Reference Figure 3 The method provided in this embodiment has the following process flow: Step 1: Continuous data acquisition; The system continuously collects data from the sensing layer 100. For example, when the coal mining machine starts cutting coal, the optical particle counter 101 near it begins to read the data, the TDLAS tunable diode laser absorption spectrum methane sensor 102 monitors the gas emission, the camera network 103 captures the picture of dust, and the operation data interface 104 transmits the data that the coal mining machine is moving at a speed of 2 meters per minute in a certain direction.

[0044] Step 2: spatiotemporal dynamic prediction; The artificial intelligence processing core 200 receives the above data stream. The spatiotemporal prediction model 201 immediately performs calculation. For example, the model predicts that a high-concentration dust cloud will reach a crossroad 30 meters downstream in 45 seconds, and that the gas mass will gather in the gob under the action of ventilation, based on the initial concentration of dust, the moving direction and speed of the coal mining machine, and the ventilation model of the current roadway.

[0045] Step 3: optimal strategy generation; Based on the above prediction results, the optimal control strategy generation module 202 starts to calculate the countermeasures. It finds that if the spray is directly started at the crossroad, although it can reduce dust, it will disturb the airflow and may cause gas to spread to unintended areas. After calculation, the optimal strategy is: At T=0 seconds, two groups of intelligent spray nozzles 301 located 15 meters downstream of the coal mining machine are opened in medium flow and large cone angle mode to form a preset water mist curtain.

[0046] At T=5 seconds, a local fan near the crossroad is instructed to adjust the speed to change the airflow direction and form an air curtain to guide the gas flow to the predetermined gob.

[0047] At T=30 seconds, the intelligent spray nozzles 301 at the crossroad are instructed to open in high flow and small cone angle mode to finally suppress the approaching dust cloud.

[0048] Step 4: preemptive closed-loop control; The artificial intelligence processing core 200 decomposes the above strategy into specific instructions, such as: "Nozzle ID: A05, Flow: 5L / min, Mode: Cone-60°, Execution Time: T=0s", and sends them to the execution layer 300 through industrial Ethernet. The corresponding nozzles and fans have completed the deployment in advance before the dust cloud and gas pose an actual threat.

[0049] Step 5: cyclic feedback optimization; After the intervention is performed, the sensing layer 100 continues to monitor the intervention effect: for example, whether the dust concentration peak at the cross lane is successfully inhibited below the threshold value. These real feedback data are transmitted back to the artificial intelligence processing core 200, which is used to verify and fine-tune the accuracy of the spatiotemporal prediction model 201 and the effectiveness of the optimal control strategy generation module 202, so that the system has the ability of self-learning and adaptive optimization.

[0050] In the above manner, the present application constructs a new safety paradigm from passive response to active prevention, significantly improving the prevention and control level of mine atmospheric disasters.

[0051] In summary, the system of the present application aims to solve the problem of response lag, disconnection between sensing and action of the existing mine atmospheric monitoring system. The system is composed of a sensing layer 100, an artificial intelligence processing core 200 and an execution layer 300: the sensing layer 100 collects dust concentration, gas concentration and video stream data in real time; the artificial intelligence processing core 200 is deployed on a server and has a built-in spatiotemporal prediction model 201, which is used to fuse historical and real-time multi-source data, accurately predict the three-dimensional spatial distribution, concentration dynamics and migration trajectory of dust clouds or gas clusters within 30-60 seconds in the future, and generate preventive control instructions accordingly; the execution layer 300 dynamically implements intervention through variable flow spray nozzles and automatically adjusted ventilation valves. The method of the present application includes continuous data acquisition, spatiotemporal dynamic prediction, generation of optimal intervention strategy, and preemptive execution of closed-loop control. The present application changes passive response to active prevention, through the closed loop of intelligent prediction and accurate execution, realizes the prospective, automated and efficient intervention of mine atmospheric disaster risk, and lays a technical foundation for realizing autonomous mine safety.

[0052] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A closed-loop mine atmosphere intelligent control system based on a space-time prediction model, characterized in that: Comprise a sensing layer, an artificial intelligence processing core, and an execution layer; The sensing layer is configured to collect multi-dimensional data including at least atmospheric environmental parameters and mine operation data in real time; The artificial intelligence processing core is in data communication with the sensing layer, and is configured to receive the multi-dimensional data collected by the sensing layer, and comprises a space-time prediction model; The space-time prediction model is configured to predict the three-dimensional spatial distribution and migration path of the object to be measured in the atmosphere within a preset time period according to current and historical data, and generate a preemptive control instruction based on the prediction result; The execution layer is in instruction communication with the artificial intelligence processing core, and is configured to receive and execute the preemptive control instruction, and physically intervene in the predicted migration path of the object to be measured before the predicted risk of the object to be measured reaches a critical threshold.

2. The system of claim 1, wherein: The sensing layer comprises at least one or a combination of the following: an optical particle counter for monitoring dust concentration, a tunable diode laser absorption spectrum sensor for measuring gas concentration, and a camera network for acquiring live video streams.

3. The system of claim 1, wherein: The space-time prediction model is a GNN graph neural network model or a 3D ConvLSTM three-dimensional convolutional long short-term memory network model.

4. The system of claim 1, wherein: The mine operation data at least includes the position and speed of a coal mining machine and the operating state of a ventilator.

5. The system of claim 1, wherein: The execution layer at least comprises one or a combination of the following: a spray nozzle with independently controllable variable flow and variable atomization form, an automatically adjustable ventilation regulating valve, and a local fan.

6. The closed-loop mine atmosphere intelligent control method based on a space-time prediction model, characterized in that: The closed-loop mine atmosphere intelligent control system based on the space-time prediction model according to any one of claims 1 to 5 comprises the following steps: Step 1: continuously collecting atmospheric environmental parameters and mine operation data by the sensing layer deployed in the mine; Step 2: inputting the collected data into the space-time prediction model built in the artificial intelligence processing core to predict the three-dimensional spatial distribution and migration path of the object to be measured in the atmosphere within a preset time period; Step 3: calculating, by the artificial intelligence processing core, an optimal preemptive intervention strategy including the execution unit to be started and the action parameters and execution time of each unit based on the prediction result of step 2; Step 4: converting the optimal preemptive intervention strategy generated in step 3 into a control instruction by the artificial intelligence processing core and sending it to the execution layer, and executing the control instruction by the execution layer before the predicted risk of the object to be measured reaches a critical threshold.

7. The closed-loop mine atmosphere intelligent control method based on a spatiotemporal prediction model according to claim 6, characterized in that: The predicted preset time period is 30 to 60 seconds.

8. The closed-loop mine atmosphere intelligent control method based on a space-time prediction model according to claim 6, characterized in that: The step of calculating the optimal preemptive intervention strategy further comprises: calculating with the optimization objectives of minimizing the prediction risk and minimizing the resource consumption.

9. The closed-loop mine atmosphere intelligent control method based on a space-time prediction model according to claim 6, characterized in that: Further comprising a feedback optimization step 5, the system continuously repeats steps 1 to 4 to form a dynamic closed-loop control; after executing the instruction, the sensing layer collects feedback data of the intervention effect, and uses the feedback data to iteratively optimize the space-time prediction model and the control strategy.

10. The closed-loop mine atmosphere intelligent control method based on a spatiotemporal prediction model according to claim 6, characterized in that: In the execution step of step 4, the physical intervention means on the object under test include at least one of the following: deploying a water mist curtain on its predicted path of migration, or adjusting the local ventilation forming an air curtain to its predicted area of influence for guidance or blocking.

Images