Coal mine accident early warning method with learning function

Abstract

The invention relates to the technical field of coal mine accident early warning, in particular to a coal mine accident early warning method with a learning function, and the method comprises the steps: collecting the real-time environment characteristic data of a mine field sub-region, carrying out the mine field real-time environment safety analysis through the collected real-time environment characteristic data and a scientifically set mine field safety environment threshold value, and generating mine field real-time environment safety analysis data; if the mine field subareas are safe, continuing to detect until all the mine field subareas are traversed; if yes, abnormal environment characteristic data and abnormal position data are generated, data analysis is carried out in combination with the constructed coal mine accident analysis model, coal mine accident prediction data are generated, coal mine accident severity is analyzed according to the coal mine accident prediction data, and coal mine accident severity data are generated; and the information is pushed to a coal mine accident early warning platform through the Internet of Things communication network, and a corresponding early warning mode is selected to give an alarm, so that accurate prediction and intelligent early warning of the coal mine accident are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine accident early warning, and specifically to a coal mine accident early warning method with a learning function. Background Technique

[0002] Existing coal mine early warning technologies are difficult to reasonably set alarm thresholds, and at the same time, they are unable to predict the influence range when different types of coal mine accidents occur in different regions, resulting in poor quality of coal mine early warning operations.

[0003] The Chinese invention patent with the publication number CN112324506B introduces a dynamic early warning method for preventing rock bursts in coal mines based on microseismicity. It determines whether to conduct a rock burst risk prediction according to the constructed microseismic monitoring database of the working face. If a rock burst risk prediction is to be carried out, the microseismic data within the working face range is summarized, and a linear total energy trend line is generated based on the total frequency and total energy of a single working face per day. When the total energy trend line is continuously below the total energy trend line for three days, it is judged as a lack of microseismic events. Under the condition of a lack of microseismic events, the relationship between frequency and released energy is analyzed, and whether to initiate an early warning is determined according to the analysis results, achieving the effects of early warning, early prevention, and reducing the risk of rock burst accidents. However, it cannot comprehensively analyze the real-time situation in the mine by combining multiple data, resulting in an insufficiently comprehensive and accurate effect of the early warning operation. Summary of the Invention

[0004] To solve the deficiencies in the background technique, the present invention provides a coal mine accident early warning method with a learning function, realizing accurate prediction and intelligent early warning of coal mine accidents.

[0005] To solve the above technical problems, the present invention provides the following technical solution: A coal mine accident early warning method with a learning function, including the following steps:

[0006] S1. Divide the mine area into several mine sub-areas, randomly select any one mine sub-area to collect real-time environmental characteristic data, and obtain the real-time environmental characteristic data of the mine sub-area;

[0007] S2. Set the mine safety environment threshold, and analyze the real-time environmental safety of the mine sub-area to generate safety analysis data;

[0008] If it is safe, return to S1 to re-collect real-time environmental characteristic data until all mine sub-areas are traversed;

[0009] If it is abnormal, generate abnormal environmental characteristic data of the mine sub-area, and collect the position data of the corresponding sub-area to obtain the abnormal position data of the mine sub-area;

[0010] S3. Construct a coal mine accident analysis model based on the historical coal mine accident mine environmental characteristic data;

[0011] S4. Input the abnormal environmental feature data and the abnormal location data into the coal mine accident analysis model for data analysis to generate coal mine accident prediction data for the sub-region of the mine field;

[0012] S5. Analyze the severity of the coal mine accident based on the coal mine accident prediction data and generate coal mine accident severity data;

[0013] S6. Push the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network and issue an alarm.

[0014] The present invention collects real-time environmental feature data of the sub-region of the mine field and conducts real-time environmental safety analysis of the mine field by comparing with the scientifically set mine field safety environment threshold. When an abnormality occurs, abnormal environmental feature data and abnormal location data of the sub-region of the mine field are generated, and coal mine accident prediction data is generated in combination with the coal mine accident analysis model to analyze the severity of the coal mine accident. It is pushed to the coal mine accident warning platform through the Internet of Things communication network, and corresponding warning methods are selected to issue an alarm, realizing accurate prediction and intelligent warning of coal mine accidents.

[0015] Preferably, the mine field area is divided into several sub-regions of the mine field, and any one sub-region of the mine field is randomly selected for real-time environmental feature data collection. The specific steps for obtaining the real-time environmental feature data of the sub-region of the mine field are as follows:

[0016] S11. Set the mine field area, divide the mine field area into several sub-regions of the mine field to obtain a set of sub-regions of the mine field A = {a1, a2,..., a i ,..., a k}, where a i represents the i-th sub-region of the mine field, and k represents the total number of sub-regions of the mine field;

[0017] S12. Install multi-parameter sensors in each sub-region of the set of sub-regions of the mine field;

[0018] The multi-parameter sensors include, but are not limited to, temperature and humidity sensors, dust concentration sensors, wind speed and direction sensors, roof pressure sensors, gas concentration sensors, water level sensors, carbon monoxide concentration sensors, and oxygen sensors;

[0019] S13. Randomly select any one sub-region of the mine field and collect the real-time environmental feature data of the sub-region through the multi-parameter sensors to obtain a set of real-time environmental feature data B = {b1, b2,..., b i ,..., b l}, where b i represents the i-th type of real-time environmental feature data of the randomly selected sub-region of the mine field, and l represents the total number of types of real-time environmental feature data;

[0020] The real-time environmental characteristic data includes, but is not limited to, temperature data, humidity data, wind speed and direction data, gas concentration data, roof pressure data, oxygen concentration data, and water level data.

[0021] Preferably, set the safety environment threshold of the mine field, and analyze the real-time environmental safety of the sub-areas of the mine field to generate safety analysis data; if it is safe, return to S1, and re-collect the real-time environmental characteristic data until all sub-areas of the mine field are traversed; if it is abnormal, generate the abnormal environmental characteristic data of the sub-areas of the mine field, and collect the position data of the corresponding sub-areas to obtain the abnormal position data of the sub-areas of the mine field. The specific steps are as follows:

[0022] S21. Collect the environmental characteristic data of the mine field in the normal working state of p groups through the coal mine accident warning platform, and obtain the matrix C of the environmental characteristic data of the mine field in the normal working state as follows:

[0023]

[0024] Among them, c ij represents the jth type of environmental characteristic data of the ith group of the mine field in the normal working state collected through the coal mine accident warning platform;

[0025] S22. Select the minimum and maximum values of the environmental characteristic data of each column of the matrix C of the environmental characteristic data of the mine field in the normal working state by means of numerical comparison, and construct the set of environmental characteristic data intervals of the mine field in the normal working state Among them, represents the value range of the ith type of environmental characteristic data of the mine field in the normal working state, and respectively represent the maximum and minimum values of the environmental characteristic data of the ith column of the matrix C of the environmental characteristic data of the mine field in the normal working state;

[0026] S23. Through the intelligent optimization algorithm, combine the environmental characteristic data intervals in the set of environmental characteristic data intervals of the mine field in the normal working state to perform the search process of the safety environment threshold of the mine field, and generate the set of safety environment thresholds of the mine field D = {d1, d2,..., d i ,..., d l}, where d i represents the safety environment threshold of the ith type of environmental characteristic data in the mine field;

[0027] S231. Construct a safety environment threshold search coot population, set the population size to N, the current iteration number to t, the maximum iteration number to t max and the dimension of the safety environment threshold search space of the mine field to l;

[0028] Take the normal working environment characteristic data interval set of the mine as the search space for the mine safety environment threshold. Randomly generate N groups of safety environment thresholds in the search space for the mine safety environment threshold. Each group of safety environment thresholds corresponds to an individual of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold.

[0029] S232. Calculate the fitness value of each individual of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold. Sort the individuals of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold from largest to smallest according to the fitness value. Select n individuals of the coot for searching the safety environment threshold with higher fitness values as leaders, and the remaining individuals of the coot for searching the safety environment threshold as followers and assign them to each leader. The fitness value calculation formula is as follows:

[0030]

[0031] Among them, Y i represents the fitness value of the i-th individual of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold. α1 and α2 respectively represent the accident false alarm weight and the accident missed alarm weight. represents the probability that the safety environment threshold corresponding to the i-th individual of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold causes safety data to be falsely reported as abnormal. θ i represents the probability that the safety environment threshold corresponding to the i-th individual of the coot for searching the safety environment threshold in the coot population for searching the safety environment threshold causes abnormal data to be judged as safe. represents the correction value.

[0032] S233. Update the behavior control factor η of the coot population for searching the safety environment threshold in the current iteration process. The update formula is as follows:

[0033]

[0034] Among them, η max represents the initial maximum value of the behavior control factor, and m represents the power exponent for controlling the attenuation speed.

[0035] S234. Generate a random number r that follows a uniform distribution between [0, 1].

[0036] S2341. If r < η, the individual of the coot for searching the safety environment threshold updates its position by randomly selecting either random movement or chain movement in the search space for the mine safety environment threshold.

[0037] If random movement is selected, the individual of the coot in the safe environment threshold will update its position to a random position in the search space of the safe environment threshold of the mine; the position update formula is as follows:

[0038]

[0039] Among them, represents the position after the i-th individual of the coot in the safe environment threshold updates its position, X i represents the current position of the i-th individual of the coot in the safe environment threshold, r1 represents a random number uniformly distributed between [0, 1], and ε represents a random position;

[0040] If chain movement is selected, the individual of the coot in the safe environment threshold will update its position according to the positions of adjacent individuals of the coot in the safe environment threshold in the search space of the safe environment threshold of the mine; the position update formula is as follows:

[0041]

[0042] Among them, X i-1 represents the current position of the (i - 1)-th individual of the coot in the safe environment threshold;

[0043] S2342. If r ≥ η, the individual of the coot in the safe environment threshold will update its position using a passive update strategy in the search space of the safe environment threshold of the mine;

[0044] The followers in the population of the coot in the safe environment threshold will be affected by the leaders assigned to them and update their positions in the search space of the safe environment threshold of the mine; the position update formula is as follows:

[0045]

[0046] Among them, represents the position after the i-th follower updates its position, represents the current position of the leader assigned to the i-th follower, M i represents the current position of the i-th follower, r2 represents a random number uniformly distributed between [0, 1], and r3 represents a random number uniformly distributed between [-1, 1];

[0047] While guiding the followers, the leaders in the population of the coot in the safe environment threshold will also search for positions with higher fitness values in the search space of the safe environment threshold of the mine; the position update formula is as follows:

[0048]

[0049] Among them, Denote the position after the $i$-th leader updates its position, $P$ i Denote the current position of the $i$-th leader, $X$ best Denote the position of the coot individual with the highest fitness value among the safe environment threshold coots, and $r_4$ and $r_5$ denote random numbers uniformly distributed between $[0, 1]$;

[0050] S235. Calculate the fitness value of each coot individual in the safe environment threshold coot population after position update. If the fitness value of a coot individual after position update is greater than the original fitness value, replace the original position with the new position; otherwise, retain the original position;

[0051] S236. Determine whether the current iteration number $t$ is greater than or equal to the maximum iteration number $t$ max If the current iteration number $t$ is greater than or equal to the maximum iteration number $t$ max then output the safe environment threshold corresponding to the coot individual with the highest fitness value and perform data identification to generate a set of mine safety environment thresholds; otherwise, increment the current iteration number by 1 and return to S233 until the current iteration number is greater than or equal to the maximum iteration number;

[0052] S237. Numerically compare the real-time environment feature data in the real-time environment feature data set $B$ with the corresponding mine safety environment thresholds in the set of mine safety environment thresholds;

[0053] If all the real-time environment feature data in the real-time environment feature data set $B$ are less than the corresponding mine safety environment thresholds in the set of mine safety environment thresholds, then output the safety analysis data $E$ as safe and return to S1 to collect real-time environment feature data again until after traversing all mine sub-areas, end this coal mine accident warning operation;

[0054] Otherwise, output the mine real-time environment safety analysis data $E$ as dangerous, perform data identification on the real-time environment feature data set to generate an abnormal environment feature data set wherein, Denote the $i$-th type of real-time environment feature data of the mine sub-area where the abnormality occurs;

[0055] S238. Online obtain the position data of the mine sub-area corresponding to the abnormal environment feature data in the abnormal environment feature data set through the coal mine accident warning platform to obtain abnormal position data

[0056] By collecting a large amount of environmental characteristic data of the mine when it is in a normal working state, and screening out the maximum and minimum values of various environmental characteristic data, an environmental characteristic data interval for the normal operation of the mine is constructed, reasonably narrowing the search range of the mine safety environment threshold, and improving the search efficiency of the mine safety environment threshold; through the coot optimization algorithm, the search process of the mine safety environment threshold is carried out in combination with the environmental characteristic data interval of the normal operation of the mine, and the optimal safety environment threshold is quickly and accurately searched, ensuring the accuracy of the search result and the stability of the search process. At the same time, a behavior control factor is introduced in the iteration process, so that the search range is expanded as much as possible in the early stage of the iteration process, and the search range is narrowed as much as possible in the later stage of the iteration process, improving the search performance of the algorithm, avoiding additional resource waste, and providing a data basis for subsequent operations.

[0057] Preferably, the specific steps for constructing a coal mine accident analysis model based on the environmental characteristic data of historical coal mine accidents are as follows:

[0058] S31. Collect the characteristic data when a coal mine accident occurs in the historical data through the coal mine accident early warning platform to obtain the historical coal mine accident characteristic data set F = {f1, f2, …, f i , …, f o}, where f i represents the characteristic data when the i-th coal mine accident occurs in the historical data, and o represents the total number of coal mine accidents in the historical data;

[0059] The characteristic data when the coal mine accident occurs includes environmental characteristic data, coal mine accident occurrence location data, coal mine accident type characteristic data, and coal mine accident diffusion area location data;

[0060] S32. Set the training data ratio and the test data ratio, and divide the historical coal mine accident characteristic data set according to the training data ratio and the test data ratio to obtain a historical coal mine accident characteristic training data set and a historical coal mine accident characteristic test data set;

[0061] S33. Construct an initial LSTM model, and set the number of neurons in the input layer to α, the number of neurons in the output layer to β, and the number of hidden layers to χ;

[0062] S34. Set the training error threshold and the training times threshold, input the historical coal mine accident characteristic training data set into the initial LSTM model to train the initial LSTM model, adjust the initial weights and initial biases of the initial LSTM model according to the training results, and then continue to train the adjusted LSTM model until the training error is less than the training error threshold or the training times are greater than the training times threshold to obtain a trained LSTM model;

[0063] S35. Set the accuracy threshold, input the historical coal mine accident feature test dataset into the trained LSTM model for testing, calculate the test accuracy. If the test accuracy is greater than the accuracy threshold, the coal mine accident analysis model is obtained; otherwise, optimize the hyperparameters of the model through the grid optimization algorithm to obtain the coal mine accident analysis model.

[0064] Based on the historical coal mine accident feature dataset, continue to train and test the initial LSTM model to obtain the coal mine accident analysis model, which provides a good and reliable tool for coal mine accident early warning. And when predicting coal mine accidents, the coal mine accident analysis model not only considers the types of coal mine accidents, but also considers the diffusion range of different types of accidents in different regions, ensuring the comprehensiveness and reliability of the analysis results.

[0065] Preferably, the specific steps of inputting the abnormal environment feature data and the abnormal location data into the coal mine accident analysis model for data analysis to generate the coal mine accident prediction data of the sub-region of the mine field are as follows:

[0066] S41. Input the abnormal environment feature data in the abnormal environment feature dataset and the abnormal location data into the coal mine accident analysis model for coal mine accident analysis and processing to generate the coal mine accident prediction dataset Q = {q1, q2, q3}, where q1, q2, and q3 represent the predicted accident type, predicted accident level, and predicted accident diffusion area of the sub-region of the mine field.

[0067] Preferably, the specific steps of analyzing the severity of coal mine accidents based on the coal mine accident prediction data and generating the coal mine accident severity data are as follows:

[0068] S51. Calculate the coal mine accident severity evaluation value λ through the coal mine accident severity evaluation formula in combination with the coal mine accident prediction dataset; the coal mine accident severity evaluation formula is as follows:

[0069] λ = w1·φ1 + w2·φ2 + w3·φ3,

[0070] where w1, w2, and w3 respectively represent the accident type weight, accident level weight, and accident diffusion area weight, φ1 represents the influence factor corresponding to the predicted accident type of the sub-region of the mine field, φ2 represents the influence factor corresponding to the predicted accident level of the sub-region of the mine field, and φ3 represents the influence factor corresponding to the total area of the predicted accident diffusion area of the sub-region of the mine field;

[0071] S52. Set the first accident value as ω1 and the second accident value as ω2;

[0072] S53. Numerically compare the coal mine accident severity evaluation value λ, the first accident value ω1, and the second accident value ω2, and generate coal mine accident severity data according to the numerical comparison result;

[0073] If λ ≤ ω1, output the coal mine accident severity data G as mild;

[0074] If ω1 < λ ≤ ω2, output the coal mine accident severity data G as moderate;

[0075] If λ > ω2, output the coal mine accident severity data G as severe.

[0076] Calculate the coal mine accident severity evaluation value through the coal mine accident severity evaluation formula combined with the coal mine accident prediction data set, numerically compare it with the set first accident value and second accident value, and generate coal mine accident severity data according to the numerical comparison result. When calculating the coal mine accident severity evaluation value, comprehensively consider from multiple aspects to ensure the reliability and accuracy of the calculation result.

[0077] Preferably, the specific steps of pushing the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network and issuing an alarm are as follows:

[0078] S61. Push the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network, and the coal mine accident warning platform selects the corresponding warning method according to the coal mine accident severity data;

[0079] When the coal mine accident severity data G is mild, alarm by activating the buzzer in the sub-area of the mine field corresponding to the abnormal environment feature data in the abnormal environment feature data set, remind the on-site staff to check the surrounding environment, and push the abnormal environment feature data set, the abnormal location data, and the coal mine accident prediction data set to the patrol terminal;

[0080] When the coal mine accident severity data G is moderate, broadcast through the mine field broadcasting system in a loop, require non-essential staff to evacuate urgently, and automatically turn off the power supply system and ventilation system in the sub-area of the mine field corresponding to the abnormal environment feature data in the abnormal environment feature data set. At the same time, push the abnormal environment feature data set, the abnormal location data, and the coal mine accident prediction data set to the patrol terminal;

[0081] When the severity data G of the coal mine accident is severe, alarms are sent by activating all high-intensity acoustic and optical alarm devices in the mine, evacuation instructions for all personnel are issued in coordination with the mine broadcasting system, all power supply systems, ventilation systems, and mechanical equipment in the mine are remotely shut down, the emergency lighting system is activated, and at the same time, the abnormal environmental feature dataset, the abnormal location data, and the coal mine accident prediction dataset are pushed to the patrol terminal;

[0082] S62. After sending the alarm, return to S1 to collect real-time environmental feature data again until all sub-areas of the mine are traversed, and then end this coal mine accident early warning operation.

[0083] Select corresponding early warning methods to send alarms according to the severity data of coal mine accidents, ensuring that the possible accident situations in the mine can be understood in time and the quality of the coal mine accident early warning operation is guaranteed.

[0084] The present invention also includes a coal mine accident early warning system with a learning function, including a real-time environmental feature data acquisition module, a mine safety environment threshold setting module, a mine real-time environmental safety analysis module, a coal mine accident analysis model construction module, a coal mine accident prediction module, a coal mine accident severity direction module, and a coal mine accident early warning module;

[0085] The real-time environmental feature data acquisition module divides the mine area into several sub-areas by setting the mine area, randomly selects any one sub-area of the mine, and collects the real-time environmental feature data of this sub-area through the installed multi-parameter sensors to obtain the real-time environmental feature data;

[0086] The mine safety environment threshold setting module performs mine safety environment threshold search processing through an intelligent optimization algorithm in combination with the constructed interval of the normal working environment feature data of the mine to generate the mine safety environment threshold;

[0087] The mine real-time environmental safety analysis module numerically compares the real-time environmental feature data with the mine safety environment threshold, and generates mine real-time environmental safety analysis data according to the numerical comparison result; if it is safe, return to S1 to collect real-time environmental feature data again until all sub-areas of the mine are traversed; if it is abnormal, generate abnormal environmental feature data and collect the location data of the corresponding sub-area to obtain the abnormal location data;

[0088] The coal mine accident analysis model construction module collects the feature data when coal mine accidents occur in the historical data through the coal mine accident early warning platform to obtain the historical coal mine accident feature data, and constructs a coal mine accident analysis model based on the historical coal mine accident feature data;

[0089] The coal mine accident prediction module inputs the abnormal environment feature data in the abnormal environment feature data set and the abnormal location data into a coal mine accident analysis model for coal mine accident analysis and processing, and generates coal mine accident prediction data;

[0090] The coal mine accident severity direction module calculates the coal mine accident severity evaluation value by combining the coal mine accident severity evaluation formula with the coal mine accident prediction data set, and numerically compares the coal mine accident severity evaluation value with the set first accident value and second accident value, and generates coal mine accident severity data according to the numerical comparison result;

[0091] The coal mine accident warning module pushes the coal mine accident severity data to a coal mine accident warning platform through an Internet of Things communication network, and the coal mine accident warning platform selects a corresponding warning method according to the coal mine accident severity data.

[0092] By means of the above technical solution, the present invention provides a coal mine accident warning method with a learning function, which at least has the following beneficial effects:

[0093] 1. The present invention analyzes the real-time environmental safety of the mine sub-region by collecting the real-time environmental feature data of the mine sub-region and comparing it with the scientifically set mine safety environment threshold. When abnormal, it generates the abnormal environmental feature data and abnormal location data of the mine sub-region, and combines a coal mine accident analysis model to generate coal mine accident prediction data to analyze the severity of coal mine accidents. It is pushed to a coal mine accident warning platform through an Internet of Things communication network, and a corresponding warning method is selected to issue an alarm, realizing the accurate prediction and intelligent warning of coal mine accidents.

[0094] 2. The present invention collects a large amount of environmental feature data when the mine is in a normal working state, and screens out the maximum and minimum values of various environmental feature data to construct the normal working environmental feature data interval of the mine, reasonably narrowing the search range of the mine safety environment threshold and improving the search efficiency of the mine safety environment threshold.

[0095] 3. The present invention performs mine safety environment threshold search processing by combining the coot optimization algorithm with the normal working environmental feature data interval of the mine, quickly and accurately searches for the optimal safety environment threshold, ensures the accuracy of the search result and the stability of the search process. At the same time, a behavior control factor is introduced in the iteration process, so that the search range is expanded as much as possible in the early stage of the iteration process, and the search range is narrowed as much as possible in the later stage of the iteration process, improving the search performance of the algorithm, avoiding additional resource waste, and providing a data basis for subsequent operations.

[0096] 4. The present invention continues to train and test the initial LSTM model based on the historical coal mine accident feature dataset to obtain a coal mine accident analysis model, which provides a good and reliable tool for coal mine accident early warning. When predicting coal mine accidents, the coal mine accident analysis model not only considers the types of coal mine accidents, but also takes into account the diffusion range of different types of accidents in different regions, ensuring the comprehensiveness and reliability of the analysis results.

[0097] 5. The present invention calculates the severity evaluation value of coal mine accidents by combining the coal mine accident severity evaluation formula with the coal mine accident prediction dataset, and generates coal mine accident severity data according to the numerical comparison results. When calculating the severity evaluation value of coal mine accidents, it comprehensively considers from multiple aspects to ensure the reliability and accuracy of the calculation results; at the same time, it selects the corresponding early warning method to issue an alarm according to the coal mine accident severity data, ensuring that the possible accident situations in the mine can be understood in time and guaranteeing the quality of coal mine accident early warning operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0099] Figure 1 It is a flowchart of the coal mine accident early warning method provided by the present invention;

[0100] Figure 2 It is a schematic diagram of the modules of the coal mine accident early warning system provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0101] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0102] Embodiment 1 is as follows:

[0103] When faced with the problem of insufficient comprehensiveness and accuracy in early warning operations, traditional methods cannot comprehensively analyze the real-time situation in the mine by combining multiple data. This embodiment proposes a coal mine accident early warning method with a learning function, which can generate abnormal environment feature data and abnormal location data of mine sub-areas when the mine environment is abnormal, and generate coal mine accident prediction data in combination with the coal mine accident analysis model to analyze the severity of coal mine accidents, push it to the coal mine accident early warning platform through the Internet of Things communication network, and select corresponding early warning methods to issue alarms, realizing accurate prediction and intelligent early warning of coal mine accidents. As Figure 1 shown, the method includes the following steps:

[0104] S1. Divide the mine area into several mine sub-areas, randomly select any one mine sub-area to collect real-time environment feature data, and obtain the real-time environment feature data of the mine sub-area. As a specific implementation plan of this method, the detailed plan of this step is as follows:

[0105] S11. Set the mine area, divide the mine area into several mine sub-areas, and obtain the set of mine sub-areas A = {a1, a2,..., a i ,…, a k}, where a i represents the i-th mine sub-area, and k represents the total number of mine sub-areas;

[0106] S12. Install multi-parameter sensors in each mine sub-area in the set of mine sub-areas; the multi-parameter sensors include but are not limited to temperature and humidity sensors, dust concentration sensors, wind speed and direction sensors, roof pressure sensors, gas concentration sensors, water level sensors, carbon monoxide concentration sensors, and oxygen sensors;

[0107] S13. Randomly select any one mine sub-area, and collect the real-time environment feature data of this mine sub-area through the multi-parameter sensors, and obtain the real-time environment feature data set B = {b1, b2,..., b i ,…, b l}, where b i represents the i-th type of real-time environment feature data of the randomly selected mine sub-area, and l represents the total number of types of real-time environment feature data; the real-time environment feature data includes but is not limited to temperature data, humidity data, wind speed and direction data, gas concentration data, roof pressure data, oxygen concentration data, and water level data.

[0108] S2. Set the safety environment threshold of the mine field, and analyze the real-time environmental safety of the sub-areas of the mine field to generate safety analysis data. If it is safe, return to S1 to collect real-time environmental characteristic data again until all sub-areas of the mine field are traversed. If it is abnormal, generate abnormal environmental characteristic data of the sub-areas of the mine field, and collect the position data of the corresponding sub-areas to obtain the abnormal position data of the sub-areas of the mine field. As a specific implementation plan of this method, the detailed plan of this step is as follows:

[0109] S21. Collect p groups of environmental characteristic data of the mine field in the normal working state through the coal mine accident warning platform to obtain the matrix C of the environmental characteristic data of the mine field in the normal working state as follows:

[0110]

[0111] Among them, c ij represents the jth type of environmental characteristic data of the ith group of the mine field in the normal working state collected through the coal mine accident warning platform;

[0112] S22. Select the minimum and maximum values of the environmental characteristic data of each column of the environmental characteristic data matrix C of the mine field in the normal working state through numerical comparison, and construct the set of environmental characteristic data intervals of the mine field in the normal working state Among them, represents the value interval of the ith type of environmental characteristic data when the mine field is in the normal working state, and respectively represent the maximum and minimum values of the environmental characteristic data of the ith column in the environmental characteristic data matrix C of the mine field in the normal working state;

[0113] S23. Perform mine safety environment threshold search processing on the environmental characteristic data intervals in the set of environmental characteristic data intervals of the mine field in the normal working state through an intelligent optimization algorithm, and generate the set D of mine safety environment thresholds = {d1, d2,..., d i ,..., d l}, where d i represents the safety environment threshold of the ith type of environmental characteristic data in the mine field;

[0114] S231. Construct a safety environment threshold search coot population, set the population size as N, the current iteration number as t, the maximum iteration number as t max and the dimension of the mine safety environment threshold search space as l;

[0115] Take the interval set of the characteristic data of the normal working environment of the mine as the search space for the safety environment threshold of the mine. Randomly generate N groups of safety environment thresholds in the search space for the safety environment threshold of the mine. Each group of safety environment thresholds corresponds to an individual of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold.

[0116] S232. Calculate the fitness values of each individual of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold. Sort the individuals of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold from largest to smallest according to the fitness values, and select the n individuals of the coot for searching the safety environment threshold with higher fitness values as leaders, and the remaining individuals of the coot for searching the safety environment threshold as followers and assign them to each leader. The fitness value calculation formula is as follows:

[0117]

[0118] Among them, Y i represents the fitness value of the i-th individual of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold. α1 and α2 respectively represent the weights of false alarms and missed alarms of accidents. represents the probability that the safety environment threshold corresponding to the i-th individual of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold causes safety data to be misreported as abnormal. θ i represents the probability that the safety environment threshold corresponding to the i-th individual of the coot for searching the safety environment threshold in the population of coots for searching the safety environment threshold causes abnormal data to be judged as safe. represents the correction value;

[0119] S233. Update the behavior control factor η of the population of coots for searching the safety environment threshold in the current iteration process. The update formula is as follows:

[0120]

[0121] Among them, η max represents the initial maximum value of the behavior control factor, and m represents the power exponent for controlling the attenuation speed;

[0122] S234. Generate a random number r that follows a uniform distribution between [0, 1];

[0123] S2341. If r < η, the individual of the coot for searching the safety environment threshold updates its position by choosing either random movement or chain movement in the search space for the safety environment threshold of the mine.

[0124] If random movement is selected, the individual of the coot for searching the safety environment threshold updates its position to a random position in the search space for the safety environment threshold of the mine. The position update formula is as follows:

[0125]

[0126] Among them, represents the position after the position update of the i-th coot individual with the safety environment threshold, X i represents the current position of the i-th coot individual with the safety environment threshold, r1 represents a random number uniformly distributed between [0, 1], and ε represents a random position;

[0127] If chain movement is selected, the coot individual with the safety environment threshold updates its position according to the positions of adjacent coot individuals with the safety environment threshold in the search space of the mine safety environment threshold; the position update formula is as follows:

[0128]

[0129] Among them, X i-1 represents the current position of the (i - 1)-th coot individual with the safety environment threshold;

[0130] S2342. If r ≥ η, the coot individual with the safety environment threshold updates its position using a passive update strategy in the search space of the mine safety environment threshold;

[0131] The followers in the coot population with the safety environment threshold are affected by the leaders assigned to them and update their positions in the search space of the mine safety environment threshold; the position update formula is as follows:

[0132]

[0133] Among them, represents the position after the position update of the i-th follower, represents the current position of the leader assigned to the i-th follower, M i represents the current position of the i followers, r2 represents a random number uniformly distributed between [0, 1], and r3 represents a random number uniformly distributed between [-1, 1];

[0134] While guiding the followers, the leaders in the coot population with the safety environment threshold also search for positions with higher fitness values in the search space of the mine safety environment threshold; the position update formula is as follows:

[0135]

[0136] Among them, represents the position after the position update of the i-th leader, P i represents the current position of the i-th leader, X bestDenote the position of the coot individual with the highest fitness value as the safety environment threshold, and r4 and r5 denote random numbers uniformly distributed between [0, 1];

[0137] S235. Calculate the fitness values of each coot individual in the coot population with the safety environment threshold after position update. If the fitness value of a coot individual after position update is greater than the original fitness value, replace the original position with the new position; otherwise, retain the original position;

[0138] S236. Determine whether the current iteration number t is greater than or equal to the maximum iteration number t max If the current iteration number t is greater than or equal to the maximum iteration number t max then output the safety environment threshold corresponding to the coot individual with the highest fitness value and perform data identification to generate a set of mine safety environment thresholds; otherwise, increment the current iteration number by 1 and return to S233 until the current iteration number is greater than or equal to the maximum iteration number;

[0139] S237. Numerically compare the real-time environment feature data in the real-time environment feature data set B with the corresponding mine safety environment thresholds in the set of mine safety environment thresholds;

[0140] If all the real-time environment feature data in the real-time environment feature data set B are less than the corresponding mine safety environment thresholds in the set of mine safety environment thresholds, then output the safety analysis data E as safe and return to S1 to collect real-time environment feature data again until all mine sub-areas are traversed and this coal mine accident warning operation ends;

[0141] Otherwise, output the mine real-time environment safety analysis data E as dangerous, perform data identification on the real-time environment feature data set to generate an abnormal environment feature data set where, represents the i-th type of real-time environment feature data of the mine sub-area where an abnormality occurs;

[0142] S238. Online obtain the position data of the mine sub-area corresponding to the abnormal environment feature data in the abnormal environment feature data set through the coal mine accident warning platform to obtain abnormal position data

[0143] S3. Construct a coal mine accident analysis model based on the historical coal mine accident mine environment feature data. As a specific implementation plan of this method, the detailed plan for this step is as follows:

[0144] S31. Collect the feature data when coal mine accidents occurred in the historical data through the coal mine accident warning platform to obtain the historical coal mine accident feature data set F = {f1, f2, …, f i , …, fo}, where f i represents the characteristic data at the i-th occurrence of a coal mine accident in the historical data, and o represents the total number of coal mine accidents in the historical data;

[0145] The characteristic data at the time of a coal mine accident includes environmental characteristic data, coal mine accident occurrence location data, coal mine accident type characteristic data, and coal mine accident diffusion area location data;

[0146] S32. Set the training data ratio and the test data ratio, and partition the historical coal mine accident characteristic data set according to the training data ratio and the test data ratio to obtain a historical coal mine accident characteristic training data set and a historical coal mine accident characteristic test data set;

[0147] S33. Construct an initial LSTM model, and set the number of neurons in the input layer to α, the number of neurons in the output layer to β, and the number of hidden layers to χ;

[0148] S34. Set the training error threshold and the training times threshold, input the historical coal mine accident characteristic training data set into the initial LSTM model to train the initial LSTM model, adjust the initial weights and initial biases of the initial LSTM model according to the training results, and then continue to train the adjusted LSTM model until the training error is less than the training error threshold or the training times are greater than the training times threshold to obtain a trained LSTM model;

[0149] S35. Set the accuracy threshold, input the historical coal mine accident characteristic test data set into the trained LSTM model for testing, calculate the test accuracy. If the test accuracy is greater than the accuracy threshold, obtain a coal mine accident analysis model; otherwise, optimize the hyperparameters of the model through a grid optimization algorithm to obtain a coal mine accident analysis model.

[0150] S4. Input the abnormal environmental characteristic data and abnormal location data into the coal mine accident analysis model for data analysis to generate coal mine accident prediction data for the sub-region of the coal mine. S41. Input the abnormal environmental characteristic data and abnormal location data in the abnormal environmental characteristic data set into the coal mine accident analysis model for coal mine accident analysis and processing to generate a coal mine accident prediction data set Q = {q1, q2, q3}, where q1, q2, and q3 represent the predicted accident type, predicted accident level, and predicted accident diffusion area of the sub-region of the coal mine.

[0151] S5. Analyze the severity of the coal mine accident based on the coal mine accident prediction data and generate coal mine accident severity data as a specific implementation plan of this method. The detailed plan for this step is as follows:

[0152] S51. Calculate the severity evaluation value λ of coal mine accidents by combining the coal mine accident severity evaluation formula with the coal mine accident prediction data set. The coal mine accident severity evaluation formula is as follows:

[0153] λ = w1·φ1 + w2·φ2 + w3·φ3,

[0154] where w1, w2, and w3 represent the accident type weight, accident level weight, and accident diffusion area weight respectively, φ1 represents the influence factor corresponding to the predicted accident type in the sub - area of the mine field, φ2 represents the influence factor corresponding to the predicted accident level in the sub - area of the mine field, and φ3 represents the influence factor corresponding to the total area of the predicted accident diffusion area in the sub - area of the mine field;

[0155] S52. Set the first accident value as ω1 and the second accident value as ω2;

[0156] S53. Compare the numerical values of the coal mine accident severity evaluation value λ, the first accident value ω1, and the second accident value ω2, and generate coal mine accident severity data according to the numerical comparison result;

[0157] If λ ≤ ω1, output the coal mine accident severity data G as mild;

[0158] If ω1 < λ ≤ ω2, output the coal mine accident severity data G as moderate;

[0159] If λ > ω2, output the coal mine accident severity data G as severe.

[0160] S6. Push the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network and issue an alarm. As a specific implementation plan of this method, the detailed plan for this step is as follows:

[0161] S61. Push the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network, and the coal mine accident warning platform selects the corresponding warning method according to the coal mine accident severity data;

[0162] When the coal mine accident severity data G is mild, start the buzzer in the sub - area of the mine field corresponding to the abnormal environment feature data in the abnormal environment feature data set to give an alarm, remind the on - site staff to check the surrounding environment, and push the abnormal environment feature data set, abnormal location data, and coal mine accident prediction data set to the patrol terminal;

[0163] When the severity data G of the coal mine accident is moderate, it is broadcast cyclically through the mine broadcasting system, requiring non-essential staff to evacuate urgently, and automatically shutting down the power supply system and ventilation system in the mine sub-region corresponding to the abnormal environment feature data in the abnormal environment feature dataset. At the same time, the abnormal environment feature dataset, abnormal location data, and coal mine accident prediction dataset are pushed to the patrol terminal;

[0164] When the severity data G of the coal mine accident is severe, all high-intensity audible and visual alarm devices in the mine are activated for alarm, and an evacuation order for all personnel is issued in cooperation with the mine broadcasting system. All power supply systems, ventilation systems, and mechanical equipment in the mine are remotely shut down, and the emergency lighting system is activated. At the same time, the abnormal environment feature dataset, abnormal location data, and coal mine accident prediction dataset are pushed to the patrol terminal;

[0165] S62. After issuing the alarm, return to S1 to collect real-time environment feature data again until all mine sub-regions are traversed, and then end this coal mine accident early warning operation.

[0166] The second embodiment is as follows:

[0167] Please refer to Figure 2 , a coal mine accident early warning system with a learning function, including a real-time environment feature data acquisition module, a mine safety environment threshold setting module, a mine real-time environment safety analysis module, a coal mine accident analysis model construction module, a coal mine accident prediction module, a coal mine accident severity direction module, and a coal mine accident early warning module;

[0168] The real-time environment feature data acquisition module divides the mine area into several mine sub-regions by setting the mine area, randomly selects any one mine sub-region, and collects the real-time environment feature data of this mine sub-region through the installed multi-parameter sensors to obtain the real-time environment feature data;

[0169] The mine safety environment threshold setting module performs mine safety environment threshold search processing through an intelligent optimization algorithm combined with the constructed normal working environment feature data interval of the mine to generate the mine safety environment threshold,

[0170] The mine real-time environment safety analysis module compares the real-time environment feature data with the mine safety environment threshold numerically, and generates mine real-time environment safety analysis data according to the numerical comparison result; if it is safe, return to S\(1\) to collect real-time environment feature data again until all mine sub-regions are traversed; if it is abnormal, generate abnormal environment feature data, and collect the location data of the corresponding sub-region to obtain the abnormal location data;

[0171] The coal mine accident analysis model construction module collects the characteristic data when coal mine accidents occur from the historical data through the coal mine accident early warning platform to obtain the historical coal mine accident characteristic data, and constructs a coal mine accident analysis model based on the historical coal mine accident characteristic data;

[0172] The coal mine accident prediction module inputs the abnormal environment characteristic data and abnormal location data in the abnormal environment characteristic data set into the coal mine accident analysis model for coal mine accident analysis and processing, and generates coal mine accident prediction data;

[0173] The coal mine accident severity direction module calculates the coal mine accident severity evaluation value by combining the coal mine accident severity evaluation formula with the coal mine accident prediction data set, and numerically compares the coal mine accident severity evaluation value with the set first accident value and second accident value, and generates coal mine accident severity data according to the numerical comparison result;

[0174] The coal mine accident early warning module pushes the coal mine accident severity data to the coal mine accident early warning platform through the Internet of Things communication network, and the coal mine accident early warning platform selects the corresponding early warning method according to the coal mine accident severity data.

[0175] Those of ordinary skill in the art can understand that all or part of the steps in implementing the methods of the above embodiments can be completed by instructing relevant hardware through a program. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0176] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0177] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. The present specification selects and specifically describes these embodiments in order to better explain the principle and practical application of the invention, so that those skilled in the art in the technical field can well understand and utilize the invention.

Claims

1. A coal mine accident early warning method with a learning function, characterized in that, It includes the following steps: S1. Divide the mine area into several mine sub-areas, randomly select any one mine sub-area to collect real-time environmental characteristic data, and obtain the real-time environmental characteristic data of the mine sub-area; S2. Set the mine safety environment threshold, and analyze the real-time environmental safety of the mine sub-area to generate safety analysis data; If it is safe, return to S1 to re-collect real-time environmental characteristic data until all mine sub-areas are traversed; If it is abnormal, generate abnormal environmental characteristic data of the mine sub-area, and collect the position data of the corresponding sub-area to obtain the abnormal position data of the mine sub-area; S3. Build a coal mine accident analysis model based on the historical coal mine accident mine environmental characteristic data; S4. Input the abnormal environmental characteristic data and the abnormal position data into the coal mine accident analysis model for data analysis, and generate coal mine accident prediction data of the mine sub-area; S5. Analyze the severity of the coal mine accident based on the coal mine accident prediction data, and generate coal mine accident severity data; S6. Push the coal mine accident severity data to the coal mine accident warning platform through the Internet of Things communication network and issue an alarm.

2. The coal mine accident early warning method according to claim 1, characterized in that The S1 includes the following steps: S11. Set the mining area and divide the mining area into several sub-mining areas to obtain a set of sub-mining areas A = {a1, a2, …, a i , …, a k}, where a i represents the i-th sub-mining area, and k represents the total number of sub-mining areas; S12. Install multi-parameter sensors in each mine sub-area where the mine sub-areas are concentrated; S13. Randomly select any mine sub-region, and collect the real-time environmental characteristic data of this mine sub-region through a multi-parameter sensor to obtain a real-time environmental characteristic data set B = {b1, b2, …, b i , …, b l}, where b i represents the i-th type of real-time environmental characteristic data of the randomly selected mine sub-region, and l represents the total number of types of real-time environmental characteristic data; The real-time environmental characteristic data includes but is not limited to temperature data, humidity data, wind speed and direction data, gas concentration data, roof pressure data, oxygen concentration data, and water level data.

3. The coal mine accident warning method according to claim 2, characterized in that, The S2 includes the following steps: S21. Collect environmental characteristic data of the mine when it is in a normal working state in p groups through the coal mine accident warning platform to obtain the mine normal working environmental characteristic data matrix C; S22. Select the minimum and maximum values of the characteristic data of the normal working environment of each column of mines in the C through numerical comparison, and construct a set of intervals of the characteristic data of the normal working environment of the mines Among them, represents the value range of the i-th type of environmental characteristic data when the mine is in a normal working state, and respectively represent the maximum and minimum values of the characteristic data of the normal working environment of the i-th column of mines in the C; S23. Perform a search process for the mine safety environment threshold by combining the intelligent optimization algorithm with the mine normal working environment characteristic data intervals in the mine normal working environment characteristic data interval set, and generate a mine safety environment threshold set D = {d1, d2, …, d i , …, d l}, where d i represents the safety environment threshold of the i-th type of environmental characteristic data in the mine.

4. The coal mine accident early warning method according to claim 3, characterized in that, The S23 includes the following steps: S231. Construct a search for the safety environment threshold of the coot population, set the population size as N, the current iteration number as t, and the maximum iteration number as t max and the dimension of the search space for the safety environment threshold of the mine as l; Taking the as the search space of the safety environment threshold for the mine, randomly generate N groups of safety environment thresholds in the search space of the safety environment threshold for the mine, and each group of safety environment thresholds corresponds to an individual of the safety environment threshold search coot. S232. Calculate the fitness values of each individual of the safety environment threshold coots, sort each individual of the safety environment threshold coots from large to small according to the fitness value, select the n individuals of the safety environment threshold coots with higher fitness values as leaders, and the remaining individuals of the safety environment threshold coots as followers and assign them to each leader. The fitness value calculation formula is as follows: Among them, Y i represents the fitness value of the i-th safety environment threshold coot individual, and α1 and α2 respectively represent the accident false alarm weight and the accident missed alarm weight. represents the probability that the safety environment threshold corresponding to the i-th safety environment threshold coot individual causes safety data to be falsely reported as abnormal, and θ i represents the probability that the safety environment threshold corresponding to the i-th safety environment threshold coot individual causes abnormal data to be judged as safe. represents the correction value; S233. Update the behavior control factor η of the safety environment threshold coot population in the current iteration process; the update formula is as follows: where η max represents the initial maximum value of the behavior control factor, and m represents the power exponent for controlling the attenuation rate; S234. Generate a random number r that follows a uniform distribution between [0,1]; S235. Calculate the fitness values of each individual of the safety environment threshold coot population after position update. If the fitness value of an individual of the safety environment threshold coot after position update is greater than the original fitness value, replace the original position with the new position; otherwise, retain the original position; S236. Determine whether t is greater than or equal to t max , if t is greater than or equal to t max , then output the safety environment threshold corresponding to the coot individual with the highest fitness value and perform data identification to generate a set of safety environment thresholds for the mine; otherwise, increment t by 1 and return to S233; S237. Numerically compare the real-time environmental characteristic data in B with the corresponding mine safety environment threshold in D; If all the real-time environmental characteristic data in B are less than the corresponding mine safety environment threshold in D, output the safety analysis data E as safe, return to S1, re-collect real-time environmental characteristic data, and end this coal mine accident warning operation after traversing all mine sub-areas; Otherwise, the output of the real-time environmental safety analysis data E of the mine is dangerous, and the B is data-identified to generate an abnormal environmental feature data set Among them, represents the i-th type of real-time environmental feature data of the sub-region of the mine where the abnormality occurs; S238. Online obtain the location data of the sub-area of the mine field corresponding to the abnormal environmental characteristic data in to obtain the abnormal location data 5. The coal mine accident early warning method according to claim 4, characterized in that, The S234 includes the following steps: S2341. If r < η, the coot individuals with the safety environment threshold perform position update by randomly selecting any one of the random movement and chain movement in the search space of the safety environment threshold of the mine; If random movement is selected, the coot individuals with the safety environment threshold update their positions to a random position in the search space of the safety environment threshold of the mine; the position update formula is as follows: Among them, represents the position of the i-th coot individual after position update with the safety environment threshold, X i represents the current position of the i-th coot individual with the safety environment threshold, r1 represents a random number uniformly distributed between [0, 1], and ε represents a random position; If chain movement is selected, the coot individuals with the safety environment threshold update their positions according to the positions of adjacent coot individuals with the safety environment threshold in the search space of the safety environment threshold of the mine; the position update formula is as follows: where X i-1 represents the current position of the i-1st individual of the coot with the safety environment threshold; S2342. If r ≥ η, the coot individuals with the safety environment threshold perform position update by adopting a passive update strategy in the search space of the safety environment threshold of the mine; The followers in the coot population with the safety environment threshold are affected by the leaders assigned to them and perform position update in the search space of the safety environment threshold of the mine; the position update formula is as follows: Among them, represents the position after the $i$-th follower updates its position, represents the current position of the leader assigned to the $i$-th follower, $M$ i represents the current positions of $i$ followers, $r_2$ represents a random number uniformly distributed between $[0, 1]$, and $r_3$ represents a random number uniformly distributed between $[-1, 1]$; While guiding the followers, the leaders in the coot population with the safety environment threshold also search for positions with higher fitness values in the search space of the safety environment threshold of the mine; the position update formula is as follows: Among them, P i new represents the position after the i-th leader updates its position, and P i represents the current position of the i-th leader. X best represents the position of the individual of the coot with the highest fitness value among the safety environment thresholds, and r4 and r5 represent random numbers uniformly distributed between [0, 1].

6. The coal mine accident early warning method according to claim 5, wherein The S3 includes the following steps: S31. Collect the characteristic data at the time of coal mine accidents in the historical data through the coal mine accident early warning platform, and obtain the historical coal mine accident characteristic data set F = {f1, f2, …, f i , …, f o}, where f i represents the characteristic data at the i-th occurrence of coal mine accidents in the historical data, and o represents the total number of coal mine accidents in the historical data; The characteristic data at the time of coal mine accident includes environmental characteristic data, coal mine accident occurrence location data, coal mine accident type characteristic data, and coal mine accident diffusion area location data; S32. Set the training data ratio and the test data ratio, and divide the data of F according to the training data ratio and the test data ratio to obtain a historical coal mine accident characteristic training data set and a historical coal mine accident characteristic test data set; S33. Construct an initial LSTM model, and set the number of neurons in the input layer to α, the number of neurons in the output layer to β, and the number of hidden layers to χ; S34. Set the training error threshold and the training times threshold, input the historical coal mine accident characteristic training data set into the initial LSTM model to train the initial LSTM model, adjust the initial weights and initial biases of the initial LSTM model according to the training results, and then continue to train the adjusted LSTM model until the training error is less than the training error threshold or the training times are greater than the training times threshold to obtain a trained LSTM model; S35. Set the accuracy threshold, input the historical coal mine accident characteristic test data set into the trained LSTM model for testing, calculate the test accuracy. If the test accuracy is greater than the accuracy threshold, a coal mine accident analysis model is obtained; otherwise, optimize the hyperparameters of the model through a grid optimization algorithm to obtain a coal mine accident analysis model.

7. The coal mine accident warning method according to claim 6, wherein The S4 includes the following steps: Input the abnormal environmental characteristic data in and the into the coal mine accident analysis model for coal mine accident analysis and processing, generating a coal mine accident prediction data set Q = {q1, q2, q3}, where q1, q2, and q3 represent the predicted accident types, predicted accident levels, and predicted accident diffusion areas of the mine sub-regions.

8. The coal mine accident early warning method according to claim 7, characterized in that The S5 includes the following steps: S51. Calculate the coal mine accident severity evaluation value λ by combining the coal mine accident prediction data set with the coal mine accident severity evaluation formula; the coal mine accident severity evaluation formula is as follows: λ = w1·φ1 + w2·φ2 + w3·φ3, Among them, w1, w2, and w3 respectively represent the accident type weight, accident level weight, and accident diffusion area weight, φ1 represents the influence factor corresponding to the predicted accident type in the sub-region of the coal mine, φ2 represents the influence factor corresponding to the predicted accident level in the sub-region of the coal mine, and φ3 represents the influence factor corresponding to the total area of the predicted accident diffusion region in the sub-region of the coal mine; S52. Set the first accident value as ω1 and the second accident value as ω2; S53. Numerically compare the λ, the ω1, and the ω2, and generate the coal mine accident severity data G according to the numerical comparison result; If λ ≤ ω1, then output the G as mild; If ω1 < λ ≤ ω2, then output the G as moderate; If λ > ω2, then output the G as severe.

9. The coal mine accident warning method according to claim 8, characterized in that, The S6 includes the following steps: Push the G to the coal mine accident early warning platform through the Internet of Things communication network, and the coal mine accident early warning platform selects the corresponding early warning method according to the G; When G is at a mild level, the buzzer in the mine sub-region corresponding to the abnormal environmental feature data in the abnormal environmental feature dataset is activated for alarm to remind on-site workers to check the surrounding environment, and the the and the Q are pushed to the inspection terminal; When G is at a medium level, the mine broadcast system will broadcast in a loop, requiring non-essential staff to evacuate urgently, and automatically shut down the power supply system and ventilation system in the mine sub-region corresponding to the abnormal environmental feature data in the abnormal environmental feature dataset. At the same time, push the the and the Q to the patrol terminal; When G is at a severe level, alarms are given by activating all high-intensity acoustic and optical alarm devices in the mine, an evacuation order for all personnel is issued in coordination with the mine's broadcasting system, all power supply systems, ventilation systems, and mechanical equipment in the mine are remotely shut down, the emergency lighting system is activated, and at the same time, the the and Q are pushed to the patrol terminals.

10. A system for implementing the coal mine accident early warning method according to any one of claims 1-9.

Images