CN-122020027-A - Coal mine safety risk early warning method and system

Abstract

The embodiment of the application discloses a coal mine safety risk early warning method and system, the method comprises the following steps of collecting real-time monitoring data generated in a coal mine operation process, carrying out time synchronization and feature standardization processing on the real-time monitoring data to construct a risk feature data set, carrying out risk analysis on a coal mine operation state through a pre-constructed coal mine safety risk identification model according to the risk feature data set to obtain corresponding risk grade information, risk development trend information and risk influence factors, generating coal mine safety risk early warning information according to the risk grade information, the risk development trend information and the risk influence factors, and sending the coal mine safety risk early warning information to a coal mine dispatching and control system to trigger linkage precautionary measures corresponding to the risk grade information. The embodiment of the application can identify the key time node of transition of the coal mine safety risk from the low level to the high level, thereby effectively improving the risk identification and early warning capability.

Inventors

• LIU XINYU
• LIU HENG
• ZHANG SENLANG
• LIU YABIN
• NIU YUNPENG
• SUO ZHIWEN
• Jiao Chunjin
• DING JIANMING
• WANG HUIWEI
• QU BO
• ZHOU CHAOYI
• LU PAN

Assignees

• 中国神华能源股份有限公司神东煤炭分公司
• 北京时代凌宇科技股份有限公司

Dates

Publication Date

20260512

Application Date

20260113

Claims (10)

1. 1. The coal mine safety risk early warning method is characterized by comprising the following steps of: S1, collecting real-time monitoring data generated in the coal mine operation process, performing time synchronization and feature standardization processing on the real-time monitoring data, and constructing a risk feature data set; The real-time monitoring data comprises environment monitoring data, equipment operation data and personnel operation data, wherein the environment monitoring data comprises gas concentration, carbon monoxide concentration, ventilation wind speed and dust concentration, the equipment operation data comprises equipment current, voltage, vibration signals and operation states, the personnel operation data comprises personnel positioning information and operation behavior event information, and the risk characteristic data set comprises the current value, first-order change rate and second-order change trend of each real-time monitoring data; The characteristic standardization processing comprises the steps of respectively calculating a first-order change rate and a second-order change trend for each real-time monitoring data in a sliding time window with the length of a preset duration, wherein the first-order change rate and the second-order change trend are used for representing dynamic evolution characteristics of the coal mine running state; s2, performing risk analysis on the coal mine running state through a pre-constructed coal mine safety risk identification model according to the risk characteristic dataset to obtain corresponding risk grade information, risk development trend information and risk influence factors; The coal mine safety risk identification model is a multi-module collaborative model and comprises a space-time diagram neural network model for describing the time evolution and propagation relationship of risks under the constraint of a mine space structure and a risk analysis model for judging a risk state and identifying risk transition; S3, generating coal mine safety risk early warning information according to the risk level information, the risk development trend information and the risk influence factors, and sending the coal mine safety risk early warning information to a coal mine dispatching and control system so as to trigger linkage precautionary measures corresponding to the risk level information; the coal mine safety risk early warning information comprises risk grade and risk evolution trend indication information.
2. 2. The method according to claim 1, wherein the risk analysis is performed on the coal mine operation state according to the risk characteristic data set through a pre-constructed coal mine safety risk identification model, specifically including: inputting the risk characteristic dataset into a pre-constructed risk analysis model, and carrying out risk analysis on the coal mine running state; The risk analysis model is used for constructing coupling risk features between gas concentration, ventilation state and personnel operation behaviors aiming at gas explosion risk, and the coupling risk features are used for identifying superposition states of abnormal enrichment and induction conditions of gas, and comprise correlation features between gas concentration change rate and ventilation wind speed fluctuation, and can be calculated according to the following formula in a sliding time window: Wherein, the As a feature of the risk of said coupling, Is the concentration of the gas, and the concentration of the gas is the concentration of the gas, Is the ventilation wind speed; when the coupling risk index exceeds a preset threshold value in a plurality of continuous time windows, judging that the coal mine gas risk is transited from a stable accumulation state to a high explosion risk state which rapidly evolves; The risk analysis model is also used for constructing a ground pressure evolution feature based on surrounding rock stress, displacement change and microseismic data aiming at the rock burst risk, and identifying a risk transition trend of transition from continuous microseismic to high-energy emergency according to the ground pressure evolution feature, wherein the ground pressure evolution feature comprises an association relation between a stress accumulation rate and a microseismic event frequency and can be calculated by the following formula: Wherein, the For the characteristics of the ground pressure evolution, In order for the stress of the surrounding rock, The occurrence frequency of the microseismic event in unit time; When the ground pressure evolution characteristic is in a continuous rising trend, judging that the coal mine is in a risk transition stage of transition from a hidden high-risk state to a rock burst state; The risk analysis model is also used for constructing water damage risk characteristics aiming at the coal mine water damage risk based on water level, water pressure and surrounding rock seepage parameters, and identifying a gradual destabilization process of transition from a slow seepage state to a water bursting risk state according to the water damage risk characteristics, wherein the water damage risk characteristics comprise the association relation between a water pressure change trend and the surrounding rock seepage characteristics, and can be calculated by the following formula: Wherein, the As a feature of the risk of water damage, Is the water pressure, the water pressure is the water pressure, Is the seepage coefficient of surrounding rock; When the water damage risk characteristics show monotonous increasing trend in a plurality of continuous time windows, judging that the coal mine water damage risk enters a progressive unsteady evolution stage; Before the risk characteristic data set is input into the pre-constructed risk analysis model, the method further comprises the following steps: The method comprises the steps of constructing a risk analysis model based on pre-acquired training data, wherein the training data are derived from historical monitoring data acquired in a coal mine operation process, accident states, risk levels and risk evolution stages corresponding to the historical monitoring data, the historical monitoring data comprise at least one complete evolution process data before and after a known safety accident, the risk evolution stages comprise a normal operation stage, a risk accumulation stage, a risk acceleration stage and an accident critical stage, and training targets of the risk analysis model comprise the steps of minimizing differences between predicted risk levels and actual accident risk levels and maximizing the prediction accuracy of transition time points of risks from low levels to high levels.
3. 3. The method according to claim 2, wherein the constructing a risk analysis model based on pre-acquired training data, in particular comprises: Aiming at the gas explosion risk, constructing a plurality of gas risk training samples from historical monitoring data in a time sequence mode, and extracting the following characteristics from a continuous time window corresponding to each gas risk training sample as joint characteristic vectors of the gas risk training samples, wherein the joint characteristic vectors comprise absolute value, change rate and change trend of gas concentration, fluctuation amplitude of ventilation wind speed and ventilation direction, running state change of a gas extraction system, personnel operation density and operation time sequence distribution; training a risk analysis model based on the plurality of gas risk training samples and the corresponding gas risk levels and risk evolution stages, so that the risk analysis model learns a gas risk evolution mode in the training process, wherein the gas accumulation evolution mode is formed when the gas concentration is continuously increased and the ventilation compensation is insufficient, the high explosion risk evolution mode is formed by the superposition of the rapid fluctuation of the gas concentration and the abnormal ventilation state, and the evolution mode of the amplified gas risk under the intervention condition of personnel operation activities; Aiming at rock burst risks, constructing a plurality of rock burst risk training samples from historical monitoring data according to spatial positions and time sequences, and extracting the following characteristics from continuous time windows corresponding to each rock burst risk training sample as joint characteristic vectors of the rock burst risk training samples, wherein the joint characteristic vectors comprise surrounding rock stress values, stress accumulation rates, variation amplitude and variation trend of roadway displacement, occurrence frequency, energy distribution and spatial aggregation characteristics of microseismic events; based on the ground pressure risk training samples and the corresponding ground pressure risk levels and risk evolution stages, training a risk analysis model to enable the risk analysis model to learn ground pressure risk characteristics in the training process, wherein the ground pressure risk characteristics are that stress is continuously accumulated but corresponding hidden high risk states are generated when microseismic activities are weaker, corresponding risk transition trends are generated when microseismic events are converted from space dispersion to space concentration, and a precursor mode that microseismic energy is abnormally concentrated before rock burst occurs; Aiming at the coal mine water damage risk, constructing a plurality of water damage risk training samples from historical monitoring data according to a water damage evolution process, extracting the following characteristics from continuous or quasi-continuous time windows corresponding to each water damage risk training sample as joint characteristic vectors of the water damage risk training samples, namely, the water level of a water-bearing stratum, the variation amplitude and the variation trend of water pressure, the evolution characteristics of surrounding rock seepage parameters and seepage rates and the variation situation of water inflow of a roadway or a working surface, carrying out water damage risk grade and risk evolution stage labeling on the water damage risk training samples according to water damage accident occurrence records, and training a risk analysis model based on the water damage risk training samples and the corresponding water damage risk grade and risk evolution stage, so that the risk analysis model learns the hidden water damage risk formed by the water pressure slow-rise interaction with the seepage parameter variation, the gradual instability process of the seepage state from stable to unstable transition, and the long-term water damage evolution mode before the occurrence of the water bursting accident.
4. 4. The method according to claim 2, wherein the constructing a risk analysis model based on pre-acquired training data, in particular comprises: dividing a training sample into a low-risk stage sample and a high-risk stage sample according to a risk evolution stage marked in historical monitoring data, wherein the low-risk stage sample comprises a normal operation stage sample and a risk accumulation stage sample, and the high-risk stage sample comprises a risk acceleration stage sample and an accident critical stage sample; Training the risk analysis model in sequence according to the sequence from the low risk stage sample to the high risk stage sample, so that the risk analysis model gradually learns the stepwise characteristics of the evolution of the coal mine safety risk from a steady state to an accident state; after training of a low-risk stage sample is completed, freezing or partially freezing converged parameters in the risk analysis model, and training other parameters except the converged parameters in the risk analysis model based on a high-risk stage sample so as to prevent the risk analysis model from forgetting evolution features of a low-risk stage in the high-risk stage training process; Aiming at a training sample pair in an adjacent risk evolution stage, constraining the risk level change direction output by the risk analysis model to an adjacent time window to be consistent with the historical accident evolution direction so as to reduce the prediction result of risk level reverse change or discontinuous jump of the risk analysis model in the risk evolution process; Setting initial weights or weight constraint intervals for input features according to disaster mechanisms of different coal mine safety risk types in an initial model training stage, wherein the weights of gas concentration change rates and ventilation wind speed fluctuation features are improved in a gas explosion risk training process, the weights of surrounding rock stress accumulation rates and microseismic energy features are improved in a rock burst risk training process, the weights of water pressure change trends and surrounding rock seepage parameter features are improved in a coal mine water disaster risk training process, and only the feature weights are allowed to be adaptively adjusted in the weight constraint intervals in a subsequent training process; The method comprises the steps of setting independent risk prediction tasks for gas explosion risks, rock burst risks and coal mine water damage risks respectively, sharing common characteristic parameters used for representing the overall operation state of the coal mine in a bottom network of a risk analysis model, and independently updating parameters of different risk types in a high-level network of the risk analysis model to realize collaborative learning and differentiated convergence among various coal mine safety risks; and carrying out time deviation correction on a risk transition discrimination result output by the risk analysis model by taking the risk transition time position as a supervision signal so as to improve the prediction consistency and accuracy of the risk analysis model on the occurrence time of the coal mine safety risk critical state.
5. 5. The method of claim 1, wherein the performing the time synchronization and feature normalization process on the real-time monitoring data, after constructing the risk feature data set, further comprises: Performing multi-mode feature decomposition processing on the risk feature dataset, and respectively extracting feature subspaces representing environment evolution, structural response, energy release and personnel activity states; based on cross-modal feature consistency constraint and event triggering discrimination criteria, carrying out joint modeling on the feature subspace, and identifying to obtain a semantic event feature sequence containing event types, event occurrence time, event space positions and event intensity parameters; Constructing a mine space topological graph according to a mine tunnel structure, a ventilation system structure, equipment arrangement relations and monitoring point space distribution, mapping the semantic event feature sequence to nodes or edges of the mine space topological graph according to space positions of the mine space topological graph to obtain a mine space-time feature graph sequence evolving along with time, wherein each mine space-time feature graph comprises the mine space topological graph and feature vectors mapped to the nodes or edges of the mine space topological graph, and the feature vectors are composed of semantic event features extracted in corresponding time windows; According to a time sequence, a mine space-time characteristic diagram corresponding to a plurality of continuous time windows is input into the space-time diagram neural network model, and a risk prediction result output by the space-time diagram neural network model is obtained, wherein the risk prediction result comprises occurrence probability distribution of coal mine safety risks in a future preset time window and a corresponding space influence region; Inputting the risk prediction result to a causal reasoning correction module constructed based on a structured causal model, performing inverse fact reasoning analysis on the risk prediction result by introducing a ventilation adjustment state, an equipment operation state and personnel operation behaviors as controllable intervention variables, and outputting a real risk assessment result subjected to causal correction; The method for generating coal mine safety risk early warning information according to the risk grade information, the risk development trend information and the risk influence factors specifically comprises the following steps: Generating coal mine safety risk early warning information according to the real risk assessment result, the risk grade information, the risk development trend information and the risk influence factors; The space-time diagram neural network model is pre-established before the operation phase through a training phase, wherein the training phase comprises the following steps: Constructing a historical time sequence characteristic data set based on multi-source historical monitoring data collected in the historical operation process of the coal mine and corresponding accident records; Based on a mine space topological structure and the historical time sequence characteristic data set, constructing a space-time diagram training sample, taking a historical accident occurrence result or a historical risk level as a supervision signal, and carrying out parameter training on the space-time diagram neural network model by using the space-time diagram training sample; After training is completed and verification is passed, the structural parameters of the space-time diagram neural network model are solidified and stored, and the solidified model is only used for risk prediction calculation and parameter updating is not performed any more in the operation stage.
6. 6. A coal mine safety risk early warning system, comprising: The acquisition module is used for acquiring real-time monitoring data generated in the coal mine operation process, performing time synchronization and feature standardization processing on the real-time monitoring data, and constructing a risk feature data set; The real-time monitoring data comprises environment monitoring data, equipment operation data and personnel operation data, wherein the environment monitoring data comprises gas concentration, carbon monoxide concentration, ventilation wind speed and dust concentration, the equipment operation data comprises equipment current, voltage, vibration signals and operation states, the personnel operation data comprises personnel positioning information and operation behavior event information, and the risk characteristic data set comprises the current value, first-order change rate and second-order change trend of each real-time monitoring data; The characteristic standardization processing comprises the steps of respectively calculating a first-order change rate and a second-order change trend for each real-time monitoring data in a sliding time window with the length of a preset duration, wherein the first-order change rate and the second-order change trend are used for representing dynamic evolution characteristics of the coal mine running state; The analysis module is used for carrying out risk analysis on the coal mine running state through a pre-constructed coal mine safety risk identification model according to the risk characteristic data set to obtain corresponding risk grade information, risk development trend information and risk influence factors; The coal mine safety risk identification model is a multi-module collaborative model and comprises a space-time diagram neural network model for describing the time evolution and propagation relationship of risks under the constraint of a mine space structure and a risk analysis model for judging a risk state and identifying risk transition; The generation module is used for generating coal mine safety risk early warning information according to the risk level information, the risk development trend information and the risk influence factors, and sending the coal mine safety risk early warning information to a coal mine dispatching and control system so as to trigger linkage precautionary measures corresponding to the risk level information; the coal mine safety risk early warning information comprises risk grade and risk evolution trend indication information.
7. 7. The system of claim 6, wherein the system further comprises a controller configured to control the controller, The analysis module is specifically used for inputting the risk characteristic data set into a pre-constructed risk analysis model to perform risk analysis on the coal mine running state; The risk analysis model is used for constructing coupling risk features between gas concentration, ventilation state and personnel operation behaviors aiming at gas explosion risk, and the coupling risk features are used for identifying superposition states of abnormal enrichment and induction conditions of gas, and comprise correlation features between gas concentration change rate and ventilation wind speed fluctuation, and can be calculated according to the following formula in a sliding time window: Wherein, the As a feature of the risk of said coupling, Is the concentration of the gas, and the concentration of the gas is the concentration of the gas, Is the ventilation wind speed; when the coupling risk index exceeds a preset threshold value in a plurality of continuous time windows, judging that the coal mine gas risk is transited from a stable accumulation state to a high explosion risk state which rapidly evolves; The risk analysis model is also used for constructing a ground pressure evolution feature based on surrounding rock stress, displacement change and microseismic data aiming at the rock burst risk, and identifying a risk transition trend of transition from continuous microseismic to high-energy emergency according to the ground pressure evolution feature, wherein the ground pressure evolution feature comprises an association relation between a stress accumulation rate and a microseismic event frequency and can be calculated by the following formula: Wherein, the For the characteristics of the ground pressure evolution, In order for the stress of the surrounding rock, The occurrence frequency of the microseismic event in unit time; When the ground pressure evolution characteristic is in a continuous rising trend, judging that the coal mine is in a risk transition stage of transition from a hidden high-risk state to a rock burst state; The risk analysis model is also used for constructing water damage risk characteristics aiming at the coal mine water damage risk based on water level, water pressure and surrounding rock seepage parameters, and identifying a gradual destabilization process of transition from a slow seepage state to a water bursting risk state according to the water damage risk characteristics, wherein the water damage risk characteristics comprise the association relation between a water pressure change trend and the surrounding rock seepage characteristics, and can be calculated by the following formula: Wherein, the As a feature of the risk of water damage, Is the water pressure, the water pressure is the water pressure, Is the seepage coefficient of surrounding rock; When the water damage risk characteristics show monotonous increasing trend in a plurality of continuous time windows, judging that the coal mine water damage risk enters a progressive unsteady evolution stage; the system further comprises: the system comprises a building module, a risk analysis module and a risk analysis module, wherein the building module is used for building a risk analysis model based on pre-acquired training data, the training data is derived from historical monitoring data acquired in a coal mine operation process and accident states, risk grades and risk evolution stages corresponding to the historical monitoring data, the historical monitoring data comprise at least one complete evolution process data before and after a known safety accident occurs, the risk evolution stages comprise a normal operation stage, a risk accumulation stage, a risk acceleration stage and an accident critical stage, and the training target of the risk analysis model comprises the steps of minimizing the difference between a predicted risk grade and an actual accident risk grade and maximizing the prediction accuracy of a transition time point of risks from a low grade to a high grade.
8. 8. The system of claim 7, wherein the system further comprises a controller configured to control the controller, The system comprises a construction module, a risk analysis module and a risk analysis module, wherein the construction module is particularly used for constructing a plurality of gas risk training samples from historical monitoring data according to a time sequence mode aiming at gas explosion risks, extracting the following characteristics from a continuous time window corresponding to each gas risk training sample as joint characteristic vectors of the gas risk training samples, wherein the absolute value, the change rate and the change trend of gas concentration, the fluctuation range of ventilation wind speed and ventilation direction, the running state change of a gas extraction system, personnel operation intensity and operation time sequence distribution, the risk grade and the risk evolution stage marking are carried out on each gas risk training sample according to gas accident occurrence records, and the risk analysis module is trained based on the plurality of gas risk training samples and the corresponding gas risk grade and risk evolution stage, so that the gas risk analysis model learns a gas risk evolution mode formed when the gas concentration is continuously increased and ventilation compensation is insufficient, a high explosion evolution mode formed by the rapid fluctuation of the gas concentration and the ventilation abnormal state superposition, and a gas risk amplification mode formed under the operation activity condition; Aiming at rock burst risks, constructing a plurality of rock burst risk training samples from historical monitoring data according to spatial positions and time sequences, and extracting the following characteristics from continuous time windows corresponding to each rock burst risk training sample as joint characteristic vectors of the rock burst risk training samples, wherein the joint characteristic vectors comprise surrounding rock stress values, stress accumulation rates, variation amplitude and variation trend of roadway displacement, occurrence frequency, energy distribution and spatial aggregation characteristics of microseismic events; based on the ground pressure risk training samples and the corresponding ground pressure risk levels and risk evolution stages, training a risk analysis model to enable the risk analysis model to learn ground pressure risk characteristics in the training process, wherein the ground pressure risk characteristics are that stress is continuously accumulated but corresponding hidden high risk states are generated when microseismic activities are weaker, corresponding risk transition trends are generated when microseismic events are converted from space dispersion to space concentration, and a precursor mode that microseismic energy is abnormally concentrated before rock burst occurs; Aiming at the coal mine water damage risk, constructing a plurality of water damage risk training samples from historical monitoring data according to a water damage evolution process, extracting the following characteristics from continuous or quasi-continuous time windows corresponding to each water damage risk training sample as joint characteristic vectors of the water damage risk training samples, namely, the water level of a water-bearing stratum, the variation amplitude and the variation trend of water pressure, the evolution characteristics of surrounding rock seepage parameters and seepage rates and the variation situation of water inflow of a roadway or a working surface, carrying out water damage risk grade and risk evolution stage labeling on the water damage risk training samples according to water damage accident occurrence records, and training a risk analysis model based on the water damage risk training samples and the corresponding water damage risk grade and risk evolution stage, so that the risk analysis model learns the hidden water damage risk formed by the water pressure slow-rise interaction with the seepage parameter variation, the gradual instability process of the seepage state from stable to unstable transition, and the long-term water damage evolution mode before the occurrence of the water bursting accident.
9. 9. The system of claim 7, wherein the system further comprises a controller configured to control the controller, The construction module is specifically configured to divide a training sample into a low-risk stage sample and a high-risk stage sample according to a risk evolution stage marked in historical monitoring data, where the low-risk stage sample includes a normal operation stage sample and a risk accumulation stage sample, the high-risk stage sample includes a risk acceleration stage sample and an accident critical stage sample, train the risk analysis model sequentially according to a sequence from the low-risk stage sample to the high-risk stage sample, so that the risk analysis model gradually learns stage characteristics of evolution of a coal mine safety risk from a stable state to an accident state, freeze or partially freeze parameters converged in the risk analysis model after training of the low-risk stage sample is completed, train other parameters except the converged parameters based on the high-risk stage sample to prevent the risk analysis model from forgetting the evolution characteristics of the low-risk stage in the high-risk stage training process, constrain a risk level change direction output by the risk analysis model to an adjacent time window to keep the accident level change direction consistent with the accident level change direction in the adjacent time window in the training sample pair, continuously change in the risk level change direction or the accident risk level change direction, and continuously change the risk level change in the risk level change direction, and the risk change in the initial risk change direction is not predicted, set up in the initial risk change factor, and the risk change is continuously in the risk change is triggered in the risk change, and the risk change is triggered in the risk change is triggered, the risk is triggered, and the risk is continuously is triggered, and the risk is triggered, the method comprises the steps of improving the weight of surrounding rock stress accumulation rate and microseismic energy characteristics, improving the weight of water pressure change trend and surrounding rock seepage parameter characteristics in the coal mine water disaster risk training process, only allowing the characteristic weight to be adaptively adjusted in the weight constraint interval in the subsequent training process, setting independent risk prediction tasks for gas explosion risks, rock burst risks and coal mine water disaster risks, sharing common characteristic parameters used for representing the whole running state of a coal mine in a bottom network of a risk analysis model, independently updating parameters of different risk types in a high-level network of the risk analysis model to realize collaborative learning and differentiated convergence among various coal mine safety risks, carrying out alignment processing on transition time positions of coal mine safety risks from low level to high level in different training samples according to historical accident occurrence time points, and carrying out time deviation correction on risk discrimination results output by the risk analysis model by taking the transition time positions as supervision signals so as to improve the prediction consistency and the accuracy of the occurrence time of the coal mine safety critical state of the risk analysis model.
10. 10. The system of claim 6, further comprising: the extraction module is used for executing multi-mode feature decomposition processing on the risk feature data set and respectively extracting feature subspaces for representing environment evolution, structural response, energy release and personnel activity states; The identification module is used for carrying out joint modeling on the feature subspace based on cross-modal feature consistency constraint and event triggering discrimination criteria, and identifying to obtain a semantic event feature sequence containing event types, event occurrence time, event space positions and event intensity parameters; The mapping module is used for constructing a mine space topological graph according to a mine tunnel structure, a ventilation system structure, equipment arrangement relations and monitoring point space distribution, mapping the semantic event feature sequence to nodes or edges of the mine space topological graph according to space positions of the mine space topological graph to obtain a mine space-time feature graph sequence evolving with time, wherein each mine space-time feature graph comprises the mine space topological graph and feature vectors mapped to the nodes or edges of the mine space topological graph, and the feature vectors are composed of semantic event features extracted in corresponding time windows; The prediction module is used for inputting mine space-time characteristic diagrams corresponding to a plurality of continuous time windows into the space-time diagram neural network model according to a time sequence to obtain a risk prediction result output by the space-time diagram neural network model, wherein the risk prediction result comprises occurrence probability distribution of coal mine safety risks in a future preset time window and a corresponding space influence region; The correction module is used for inputting the risk prediction result to a causal reasoning correction module constructed based on a structured causal model, carrying out inverse fact reasoning analysis on the risk prediction result by introducing a ventilation adjustment state, a device running state and personnel operation behaviors as controllable intervention variables, and outputting a real risk assessment result after causal correction; The generation module is specifically configured to generate coal mine safety risk early warning information according to the real risk assessment result, the risk level information, the risk development trend information and the risk influence factors; The space-time diagram neural network model is pre-established before the operation phase through a training phase, wherein the training phase comprises the following steps: Constructing a historical time sequence characteristic data set based on multi-source historical monitoring data collected in the historical operation process of the coal mine and corresponding accident records; Based on a mine space topological structure and the historical time sequence characteristic data set, constructing a space-time diagram training sample, taking a historical accident occurrence result or a historical risk level as a supervision signal, and carrying out parameter training on the space-time diagram neural network model by using the space-time diagram training sample; After training is completed and verification is passed, the structural parameters of the space-time diagram neural network model are solidified and stored, and the solidified model is only used for risk prediction calculation and parameter updating is not performed any more in the operation stage.

Description

Coal mine safety risk early warning method and system Technical Field The application belongs to the technical field of coal mine safety monitoring, and particularly relates to a coal mine safety risk early warning method and system. Background The method is limited and influenced by geological environment, the working environment of the coal mining industry is bad and complex, and the safety management level in the coal mining industry is insufficient, so that the overall quality of staff is different, and the coal industry is always a high-incidence field of accidents. The high-risk safety risk in the coal mine production process often has the characteristics of long inoculation time, hidden evolution process, strong burst property and the like, and the existing coal mine safety early warning technology mainly depends on a single sensor threshold value alarm or a risk assessment method based on historical statistical characteristics, so that the recognition capability of risks is limited, and safety risk early warning cannot be effectively realized. Disclosure of Invention The embodiment of the application aims to provide a coal mine safety risk early warning method and system, which are used for solving the defect that the safety risk early warning cannot be effectively realized in the prior art. In order to solve the technical problems, the application is realized as follows: In a first aspect, a coal mine safety risk early warning method is provided, including the following steps: S1, collecting real-time monitoring data generated in the coal mine operation process, performing time synchronization and feature standardization processing on the real-time monitoring data, and constructing a risk feature data set; The real-time monitoring data comprises environment monitoring data, equipment operation data and personnel operation data, wherein the environment monitoring data comprises gas concentration, carbon monoxide concentration, ventilation wind speed and dust concentration, the equipment operation data comprises equipment current, voltage, vibration signals and operation states, the personnel operation data comprises personnel positioning information and operation behavior event information, and the risk characteristic data set comprises the current value, first-order change rate and second-order change trend of each real-time monitoring data; The characteristic standardization processing comprises the steps of respectively calculating a first-order change rate and a second-order change trend for each real-time monitoring data in a sliding time window with the length of a preset duration, wherein the first-order change rate and the second-order change trend are used for representing dynamic evolution characteristics of the coal mine running state; s2, performing risk analysis on the coal mine running state through a pre-constructed coal mine safety risk identification model according to the risk characteristic dataset to obtain corresponding risk grade information, risk development trend information and risk influence factors; The coal mine safety risk identification model is a multi-module collaborative model and comprises a space-time diagram neural network model for describing the time evolution and propagation relationship of risks under the constraint of a mine space structure and a risk analysis model for judging a risk state and identifying risk transition; S3, generating coal mine safety risk early warning information according to the risk level information, the risk development trend information and the risk influence factors, and sending the coal mine safety risk early warning information to a coal mine dispatching and control system so as to trigger linkage precautionary measures corresponding to the risk level information; the coal mine safety risk early warning information comprises risk grade and risk evolution trend indication information. In a second aspect, a coal mine safety risk early warning system is provided, including: The acquisition module is used for acquiring real-time monitoring data generated in the coal mine operation process, performing time synchronization and feature standardization processing on the real-time monitoring data, and constructing a risk feature data set; The real-time monitoring data comprises environment monitoring data, equipment operation data and personnel operation data, wherein the environment monitoring data comprises gas concentration, carbon monoxide concentration, ventilation wind speed and dust concentration, the equipment operation data comprises equipment current, voltage, vibration signals and operation states, the personnel operation data comprises personnel positioning information and operation behavior event information, and the risk characteristic data set comprises the current value, first-order change rate and second-order change trend of each real-time monitoring data; The characteristic standardization processing comprises the steps