A weather risk intelligent early warning system for mining industry

Abstract

The present application relates to the technical field of mining safety management, and particularly relates to a weather risk intelligent early warning system for mining, which comprises the following steps: collecting multi-source heterogeneous data such as meteorology and geology of a mining area in real time by constructing a multi-field coupling monitoring network of the mining area, so as to provide rich and accurate data basis for subsequent risk analysis; dynamically synthesizing a weather threat factor of the mining area by a conditional adversarial generative network through fusion of mining scene metadata and time sequence characteristics, so as to accurately capture weather risk characteristics under different mining scenes; establishing a geologic body-meteorological field coupling risk propagation topology graph and generating a regional risk potential thermal map, which is helpful for managers to quickly understand the risk situation of the mining area; identifying a medium-high risk area through spatial clustering analysis, and generating a scene self-adaptive regulation and control instruction, so as to improve the pertinence and effectiveness of emergency response, and help to quickly control the risk and reduce the loss.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mining safety management, and in particular to a weather risk intelligent early warning system for mining. BACKGROUND

[0002] In current mining safety management, the following problems still exist: In mining production, weather changes pose a significant threat to safety operations, however, traditional monitoring methods often rely on manual observation or single sensor data, which cannot comprehensively and real-time assess weather risks, making it difficult for managers to understand the actual risk situation of the mining area in a timely manner, and thus unable to take effective preventive and response measures; The mining area environment is complex and variable, involving various geological, meteorological and hydrological factors, and traditional methods often have difficulty in accurately capturing the dynamic changes and mutual relationships of these factors when identifying weather factors affecting safety, making it difficult for managers to identify potential weather risks in a timely manner, increasing the likelihood of accidents; In mining production, although there are some risk warning mechanisms, due to incomplete data acquisition, inaccurate analysis methods or delayed transmission of warning information, the effectiveness of risk warning is often poor, making it difficult for managers to obtain accurate and timely warning information when facing potential risks, and thus unable to take effective preventive measures, increasing the probability and loss degree of accidents. Therefore, the present application proposes a weather risk intelligent early warning system for mining. SUMMARY

[0003] The purpose of the present application is to solve the problems in the background art and propose a weather risk intelligent early warning system for mining.

[0004] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0005] A weather risk intelligent early warning system for mining, comprising: a multi-field coupling monitoring network construction module, a weather threat factor acquisition module, a risk propagation and quantitative analysis module, and a scene adaptive control instruction generation module;

[0006] The multi-field coupling monitoring network construction module: for constructing a multi-field coupling monitoring system in the mining area, in the key monitoring areas of open pit, underground tunnel, dump and tailings pond, a group of Internet of Things nodes integrating micro-meteorological stations, geological deformation sensors and osmometers are laid out, forming an integrated meteorological-geological-hydrological monitoring network in the mining area; Through the network, real-time multi-source heterogeneous monitoring data is collected, and a mining area spatio-temporal meteorological-geological coupling data set is generated;

[0007] The weather threat factor acquisition module: based on the mining area spatio-temporal meteorological-geological coupling data set, integrating mining scene metadata and time series features, dynamically synthesizing mining area scenario weather threat factors through conditional adversarial generation network;

[0008] Risk propagation and quantitative analysis module: dynamic quantitative analysis is carried out on the synthesized weather threat factor, and the geological body-meteorological field coupling risk propagation topology graph is established; risk transmission is realized through real-time data driving, and then a regional risk potential thermal map is generated;

[0009] Scene adaptive regulation instruction generation module: spatial clustering analysis is carried out on the regional risk potential thermal map, and the high-risk aggregation area is identified; combined with the preset risk warning value, the corresponding scene adaptive regulation instruction is generated for the identified high-risk aggregation area.

[0010] Further, the multi-field coupling monitoring network construction module is used to construct a multi-field coupling monitoring system in the mining area. In the key monitoring areas of open pit, underground tunnel, dumping site and tailings pond, a group of Internet of Things nodes integrating micro-meteorological stations, geological body deformation sensors and osmometers are arranged to form an integrated meteorological-geological-hydrological monitoring network in the mining area. The process of collecting multi-source heterogeneous monitoring data in real time through the network to generate a mining area spatio-temporal meteorological-geological coupling dataset includes:

[0011] Based on the hazard source distribution map stored in the mining area geographic information system and database, the monitoring area is divided, and the Internet of Things nodes are arranged in the key monitoring areas of the open pit slope, the intersection or fault zone of the underground tunnel, the bottom of the dumping site and the dam body of the tailings pond.

[0012] Each node of the group of Internet of Things nodes is integrated with three types of sensors, including micro-meteorological stations, geological body deformation sensors and osmometers.

[0013] The node is built-in GPS timing sub-module, and all sensor data is transmitted to the edge computing gateway through LoRa or NB-IoT wireless transmission;

[0014] The monitored sensor data is preprocessed to generate a mining area spatio-temporal meteorological-geological coupling dataset.

[0015] Further, the weather threat factor acquisition module is based on the mining area spatio-temporal meteorological-geological coupling dataset, integrates mining scene metadata and time series features, and dynamically synthesizes the mining area scenario weather threat factor through the conditional adversarial generation network. The process includes:

[0016] Obtain the mining scene metadata, wherein the mining scene metadata includes three types of heterogeneous information, i.e. mining method, rock structure characteristics and current operation stage; wherein the mining method is a categorical variable, including open pit blasting and underground tunneling; the rock structure characteristics are continuous variables, including fracture development degree and rock mass strength; the current operation stage is a categorical variable, including drilling, transportation and heap leaching;

[0017] Through embedding layer and standardization processing, the mining scene metadata is converted into a unified scene joint feature vector ;

[0018] The generator is set to input the time sequence characteristics of the mining area space-time meteorology-geology coupled dataset and the scene joint characteristics of the mining scene metadata into the LSTM network to extract the time sequence dependence, and generate candidate weather threat factors through the full connection layer ;

[0019] The discriminator is set to input the real weather threat factors and the candidate weather threat factors output by the generator , extract spatial features through a convolutional neural network, and output a discrimination probability;

[0020] The Wasserstein distance objective function is used to optimize the game process of the generator and the discriminator, so that the generator finally outputs the mining area scenario weather threat factors consistent with the distribution of the real weather threat factors.

[0021] Further, the process of training the load balancing strategy using reinforcement learning algorithm includes:

[0022] The process of converting the mining scene metadata into a unified scene joint feature vector through embedding layer and standardization includes:

[0023] For category variables embedding, define the mining method as , and assign open pit blasting as 1 and underground excavation as 2, define the current operation stage as , and assign drilling as 1, transportation as 2, and heap leaching as 3; convert to continuous vector through embedding layer:

[0024] , where is the mining method embedding vector, is the current operation stage embedding vector, is the embedding function, is the embedding layer parameter, and the output dimensions are the embedding dimension of the mining method and the embedding dimension of the operation stage .

[0025] Standardize the fracture development degree and rock mass strength in the rock structure characteristics:

[0026] ,

[0027] In the formula, is the standardized fracture development degree, is the mean of the fracture development degree, is the standard deviation of the fracture development degree, is the standardized rock mass strength, the mean value of the rock mass strength, the standard deviation of the rock mass strength;

[0028] Concatenate the embedding vector and the normalized numerical variable as the scene joint feature : wherein , is a real number set.

[0029] Further, the generator is set, the time series features of the mining area space-time meteorology-geology coupled data set, the scene joint features of the mining scene metadata are input into the LSTM network to extract the time series dependence, and the process of generating the candidate weather threat factor through the full connection layer includes:

[0030] The change rate of meteorological data , the change rate of geological data , and the change rate of hydrological data are respectively defined as , , , wherein is the time step, is the meteorological data at time t, is the geological data at time t, is the hydrological data at time t;

[0031] Taking the past hour data as a window, a time series feature sequence is constructed:

[0032] , wherein , is the time step, and 3 is the meteorological, geological, and hydrological elements;

[0033] The generator processes the time series feature sequence through the LSTM layer, and outputs the hidden state : wherein is the LSTM layer parameter, and , is the LSTM hidden layer dimension;

[0034] After concatenating the LSTM hidden state and the scene joint feature , the generator full connection layer generates the candidate weather threat factor :

[0035] ,

[0036] , wherein is the weight matrix of the generator full connection layer; is the bias term of the generator's fully connected layer.

[0037] Furthermore, a discriminator is set up to input the real weather threat factor and the candidate weather threat factor output by the generator, and the spatial features are extracted through the convolutional neural network. The process of outputting the discrimination probability includes:

[0038] Define the spatial characteristics as :

[0039] ,

[0040] Where, Fracture development degree The region mask of For meteorological data The region mask of Rock mass strength The region mask of For geological data The region mask of is element-wise multiplication; 、 is the spatial feature weight coefficient; where the region mask is a binary matrix, is the fissure development threshold, is the meteorological data threshold, is the rock strength threshold, is the geological data threshold;

[0041] The discriminator transforms spatial features into Convert to discriminant probability :

[0042] ,

[0043] Where, is the activation function; is the weight matrix of the fully connected layer of the discriminator; is the bias term of the fully connected layer of the discriminator; is the probability of the discriminator judging the authenticity of the input sample.

[0044] Furthermore, the Wasserstein distance objective function is used to optimize the game process between the generator and the discriminator. The process of finally making the generator output mining area scenario weather threat factors that are consistent with the distribution of real weather threat factors includes:

[0045] The Wasserstein distance objective function is:

[0046] ,

[0047] Where, is a loss function for the Wasserstein generator, is an expected output of the discriminator under the real data distribution; is an expected output of the discriminator under the generated data distribution; is a distribution of real weather threat factors; is an interpolation of real weather threat factors and generated samples, , is a randomly sampled interpolation factor, and ; is a gradient penalty coefficient; is a gradient norm of the discriminator at the interpolated sample ;

[0048] By alternately training the discriminator and the generator, the objective function is optimized, eventually making the generator output consistent with the real weather threat factor distribution: fixing the generator , updating the discriminator to minimize; fixing the discriminator , updating the generator to maximize.

[0049] Further, the risk propagation and quantitative analysis module performs dynamic quantitative analysis on the synthesized weather threat factors, establishes a geologic body-meteorological field coupled risk propagation topology graph; through real-time data driving, the process of realizing risk conduction and further generating a regional risk potential energy thermal map includes:

[0050] Subdivide the weather threat factor types into meteorological factors, geological factors, and hydrological factors, and based on the identified factor types, construct a geologic body-meteorological field coupled risk propagation topology graph; in the risk propagation topology graph, nodes represent various factors, and edges represent potential risk association relationships between nodes;

[0051] Respectively extract real-time monitoring values corresponding to each factor, and calculate the intensity of each factor;

[0052] Take the three types of factors and their intensity values as inputs, initialize the nodes of the risk propagation topology graph, mark the meteorological nodes as Mn, the geological nodes as Gn, and the hydrological nodes as Hn; based on the historical disaster database, set the initial weights of the edges of the risk propagation topology graph; combined with the intensity of the three types of factors, calculate the initial risk value of each node = factor intensity x weighted sum of risk propagation topology graph edge weights;

[0053] The meteorological prediction data is mapped to a 10m*10m grid by Kriging interpolation combined with the injected real-time meteorological prediction data; meanwhile, the risk values of each node are updated by combining the initial risk values of each node at the current time and the influence of real-time meteorological data, to obtain the risk dynamic value: the initial risk value of the meteorological node is directly replaced by the interpolated meteorological prediction value, so as to obtain the risk dynamic value of the meteorological node, and the initial risk value of the geological or hydrological node is updated according to the risk transmission formula to obtain the risk dynamic value:

[0054]

[0055] In the formula, is the risk dynamic value of node i at t+1; is the initial risk retention coefficient, and ; is the initial risk value of node i at t; is the risk propagation influence coefficient, and ; is the set of adjacent nodes of node i; is the risk dynamic value of node j at t; is the edge weight of node i-j; is the spatial attenuation coefficient; is the Euclidean distance of node i-j; is the time step;

[0056] The risk transmission is iteratively calculated along the edges of the risk propagation topology graph starting from the meteorological node, and finally the regional risk potential energy thermal map is output; wherein the regional risk potential energy thermal map is in units of grid, and the maximum value of the risk dynamic values of all nodes in the region is selected as the thermal value of the regional risk potential energy thermal map; the coordinates of each grid and the corresponding thermal value are combined into a key-value pair to form a grid coordinate-thermal value key-value pair set.

[0057] Further, the scene adaptive control instruction generation module performs spatial clustering analysis on the regional risk potential energy thermal map to identify a medium-high risk aggregation area; in combination with a preset risk warning value, the process of generating corresponding scene adaptive control instructions for the identified medium-high risk aggregation area includes:

[0058] ​Based on the regional risk potential heat map, the medium and high risk area is identified: set the heat value threshold, traverse all the grids in the regional risk potential heat map, extract the grid whose heat value is greater than or equal to the heat value threshold; the extracted continuous medium and high risk grid is subjected to density clustering analysis; in the clustering process, the grids with similar heat values are clustered into a class according to the spatial distance and heat value similarity between the grids; each grid region in the clustering result is marked as a medium and high risk aggregation area, including the edge transition zone, and each medium and high risk aggregation area is assigned a unique identifier;

[0059] After the preset risk warning value, scene adaptive control instructions are generated according to the identified medium and high risk aggregation area: if the heat value is greater than or equal to the risk warning value A1 and less than the risk warning value A2, a first warning is generated, the device vibration warning and LED boundary projection in the region are triggered; if the heat value is greater than or equal to the risk warning value A2 and less than the risk warning value A3, a second warning is generated, and the linkage automatic broadcast system in the region is played to broadcast the risk avoidance instruction; if the heat value is greater than or equal to the risk warning value A3, a third warning is generated, the power supply in the region is forcibly cut off, and the emergency ventilation is activated.

[0060] Compared with the prior art, the beneficial effects of the present application are: by arranging the Internet of Things node group integrating multiple sensors in the key area of the mining area, monitoring and generating the mining area space-time meteorology-geology coupled data set, the real-time collection and fusion of multi-source data such as meteorology, geology and hydrology are realized, the comprehensiveness and accuracy of data acquisition are improved, and a rich data basis is provided for subsequent modules, which helps to more accurately evaluate the weather risk of the mining area; based on the generated data set, by fusing the mining scene metadata and time sequence characteristics, the weather threat factors of the mining area are dynamically synthesized, which can more accurately capture the weather risk characteristics under different mining scenes, provide more refined input for risk warning evaluation, and help to discover potential safety hazards in advance; by dynamically quantifying the synthesized weather threat factors, a geologic body-meteorological field coupled risk propagation topology graph is established, and risk transmission is realized through real-time data driving, which is beneficial to intuitively display the risk level of each area in the mining area and reduce accident losses; based on the regional risk potential heat map, spatial clustering analysis is performed to identify medium and high risk areas, and scene adaptive control instructions are generated, which can automatically adjust the warning and response measures according to the risk level of different areas, ensure the maximum personnel safety guarantee with the minimum production interruption cost, and improve the emergency response capability and safety management level of mining production. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 A module diagram of a weather risk intelligent warning system for mining industry is provided for the present application. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0063] With reference to Figure 1 A weather risk intelligent early warning system for the mining industry, the system comprising a multi-field coupling monitoring network construction module, a weather threat factor acquisition module, a risk propagation and quantitative analysis module, and a scene adaptive control instruction generation module;

[0064] The multi-field coupling monitoring network construction module: for constructing a multi-field coupling monitoring system in a mining area, in key monitoring areas of open-pit mines, underground tunnels, dumping sites and tailing ponds, a group of Internet of Things nodes integrating micro-meteorological stations, geological body deformation sensors and osmometers are arranged to form an integrated meteorological-geological-hydrological monitoring network in the mining area; through the network, multi-source heterogeneous monitoring data are collected in real time to generate a spatiotemporal meteorological-geological coupling data set of the mining area;

[0065] The weather threat factor acquisition module: based on the spatiotemporal meteorological-geological coupling data set of the mining area, mining scene metadata and time sequence characteristics are fused, and a conditional adversarial generative network is used to dynamically synthesize a scene-based weather threat factor of the mining area;

[0066] The risk propagation and quantitative analysis module: dynamic quantitative analysis is performed on the synthesized weather threat factor, a geological body-meteorological field coupling risk propagation topology graph is established, risk conduction is realized through real-time data driving, and then a regional risk potential thermal map is generated;

[0067] The scene adaptive control instruction generation module: spatial clustering analysis is performed on the regional risk potential thermal map, a medium-high risk aggregation area is identified, a preset risk warning value is combined, scene adaptive control instructions corresponding to the identified medium-high risk aggregation area are generated, and maximum personnel safety protection is ensured at the minimum production interruption cost.

[0068] It needs to be further explained that, in the specific implementation process, the multi-field coupling monitoring network construction module is used to construct a multi-field coupling monitoring system in a mining area, in key monitoring areas of open-pit mines, underground tunnels, dumping sites and tailing ponds, a group of Internet of Things nodes integrating micro-meteorological stations, geological body deformation sensors and osmometers are arranged to form an integrated meteorological-geological-hydrological monitoring network in the mining area; through the network, multi-source heterogeneous monitoring data are collected in real time to generate a spatiotemporal meteorological-geological coupling data set of the mining area, and the process comprises:

[0069] Based on the hazard source distribution map stored in the mining area's geographic information system and database, monitoring areas are divided and IoT nodes are densely deployed in key monitoring areas, such as open-pit slopes (one deployed every 50m×50m), underground tunnel intersections or fault zones (one deployed every 30m×30m), the bottom of the spoil dump (one deployed every 100m×100m), and the tailings dam (one deployed every 20m×20m).

[0070] Each node in the IoT node cluster integrates three types of sensors: a micro-meteorological station, a geological deformation sensor, and an osmometer. The micro-meteorological station is installed on a 3-meter-high lightning protection bracket to monitor meteorological data such as wind speed. The geological deformation sensor uses a fiber Bragg grating (FBG) and is buried in the rock fracture zone or potential slip surface of the slope to monitor geological data such as rock displacement rate. The osmometer is drilled and buried 1-3 meters below the groundwater level to monitor hydrological data such as groundwater osmotic pressure.

[0071] The node has a built-in GPS timing submodule, and all sensor data is wirelessly transmitted to the edge computing gateway via LoRa or NB-IoT;

[0072] The monitored sensor data is preprocessed to generate a spatiotemporal meteorological-geological coupled data set for the mining area, wherein the data preprocessing includes data cleaning (outlier removal and missing value filling) of the sensor data. It should be noted that, in the embodiment of the present invention, a representative parameter is selected as the monitoring object for each of the meteorological data, geological data, and hydrological data. Specifically, the wind speed is selected for monitoring of the meteorological data, the rock displacement rate is selected for monitoring of the geological data, and the groundwater seepage pressure is selected for monitoring of the hydrological data, thereby constructing a multi-source data monitoring system.

[0073] It should be further explained that, in the specific implementation process, the weather threat factor acquisition module is based on the mining area's spatiotemporal meteorological-geological coupled dataset, integrating mining scenario metadata with time series features, and dynamically synthesizing mining area scenario-specific weather threat factors through a conditional adversarial generative network. The process includes:

[0074] Obtain mining scenario metadata, which includes three types of heterogeneous information: mining method, rock formation structural characteristics, and current operation stage. Mining method is a categorical variable, including open-pit blasting and underground tunneling; rock formation structural characteristics are continuous variables, including fracture development and rock mass strength; and current operation stage is a categorical variable, including drilling, transportation, and heap leaching.

[0075] Through the embedding layer and standardization processing, the mining scene metadata is converted into a unified scene joint feature vector: for the categorical variable embedding, the mining method is defined as And assign open-pit blasting to 1 and underground tunneling to 2, defining the current operation stage as , and assign the drilling value as 1, the transportation value as 2, and the heap leaching value as 3; convert to a continuous vector through an embedding layer:

[0076] , wherein is the mining method embedding vector, is the current operation stage embedding vector, is the embedding function, is the embedding layer parameter, and the output dimensions are the mining method embedding dimension , the operation stage embedding dimension , respectively;

[0077] Standardize the fracture development degree in the rock formation structure characteristics and the rock mass strength :

[0078] ,

[0079] In the formula, is the standardized fracture development degree, is the mean fracture development degree, is the standard deviation of the fracture development degree, is the standardized rock mass strength, is the mean rock mass strength, is the standard deviation of the rock mass strength;

[0080] Concatenate the embedding vector and the standardized numerical variable to form the scene joint feature : , wherein (the total dimension is the sum of the embedding dimension and the standardized numerical dimension), is a real number set;

[0081] Set the generator, input the time series features of the mining area space-time meteorology-geology coupling dataset and the scene joint features of the mining scene metadata into the LSTM network to extract the time series dependence, and generate candidate weather threat factors through a fully connected layer: define the meteorological data change rate , the geological data change rate , and the hydrological data change rate as , , respectively, in the formula, is the time step, is the meteorological data at time t, is the geological data at time t, is the hydrological data at time t;

[0082] Take the past hour data as a window to construct a time series feature sequence :

[0083] wherein, , is the time step, and 3 is the meteorology, geology, and hydrology three elements;

[0084] The generator processes the time series of features through an LSTM layer , and outputs the hidden state : wherein, is the LSTM layer parameter, and , is the LSTM hidden layer dimension;

[0085] The LSTM hidden state is concatenated with the scene joint feature , and then the generator fully connected layer generates the candidate weather threat factor :

[0086] ,

[0087] wherein, is the weight matrix of the generator fully connected layer, is used to map the concatenation vector of the LSTM hidden state and the scene joint feature to the output space of the candidate weather threat factor; is the dimension of the candidate weather threat factor; is the dimension of the scene joint feature, ; is the bias term of the generator fully connected layer, used to adjust the offset of the output;

[0088] The discriminator is set, and the real weather threat factor and the candidate weather threat factor output by the generator are input, and the spatial features are extracted through a convolutional neural network, and the discrimination probability is output: the spatial feature is defined as :

[0089] ,

[0090] wherein, is the regional mask of the fissure development degree ; is the regional mask of the meteorological data ; is the regional mask of the rock mass strength ; is the regional mask of the geological data ; for element-wise multiplication, used to represent spatial superposition; , is a spatial feature weight coefficient, used to control the contribution degree of different spatial patterns; wherein the region mask is a binary matrix, is a fissure development threshold, is a meteorological data threshold, is a rock mass strength threshold, is a geological data threshold;

[0091] The discriminator converts the spatial features into a discrimination probability :

[0092] ,

[0093] In the formula, is an activation function; is a weight matrix of the discriminator full connection layer, used to map the spatial features extracted by the convolutional neural network to the discrimination probability space; is a bias term of the discriminator full connection layer, used to adjust the offset of the probability output; is the authenticity judgment probability of the input sample by the discriminator, indicating the probability that the input (real weather threat factor or candidate weather threat factor) is a real weather threat factor associated with a mine disaster event, wherein when the input is a real weather threat factor, the discriminator should output a high probability when the input is a candidate factor generated by the generator, the discriminator should output a low probability ; it can be understood that the real mine disaster event is obtained by collecting and statistically analyzing mine disaster events (such as slope landslide, rock burst, and equipment failure) in the past H years;

[0094] The Wasserstein distance objective function is used to optimize the game process between the generator and the discriminator, so that the generator finally outputs a scenario weather threat factor consistent with the distribution of the real weather threat factor:

[0095] ,

[0096] In the formula, is the loss function of the Wasserstein generator, is the expected output of the discriminator under the real data distribution; is the expected output of the discriminator under the generated data distribution; is the distribution of the real weather threat factor (from the historical disaster database); is the interpolation of the real weather threat factor and the generated sample, used to cover all possible transition areas between the real data and the generated data, , is an interpolation factor of random sampling, and ; is a gradient penalty coefficient; is an interpolation sample gradient norm of the discriminator at

[0097] By alternately training the discriminator and the generator, the objective function , so that the generator output is consistent with the real weather threat factor distribution:

[0098] Fix the generator , update the discriminator to minimize:

[0099] ,

[0100] wherein, is the loss function of the discriminator, indicating the target that the discriminator needs to minimize in training;

[0101] Fix the discriminator , update the generator to maximize:

[0102] ,

[0103] wherein, is the loss function of the generator, indicating the target that the generator needs to maximize in training.

[0104] It should be further pointed out that in the specific implementation process, the risk propagation and quantitative analysis module performs dynamic quantitative analysis on the synthesized weather threat factor, establishes a geologic body-meteorological field coupled risk propagation topology graph; through real-time data driving, risk transmission is realized, and then the specific process of generating a regional risk potential energy thermal map includes:

[0105] Subdivide the weather threat factor type into meteorological factors, geological factors, and hydrological factors, and based on the identified factor types, construct a geologic body-meteorological field coupled risk propagation topology graph; in the risk propagation topology graph, the nodes represent various factors, and the edges represent the potential risk association relationship between the nodes;

[0106] Respectively extract the real-time monitoring values corresponding to each factor, and calculate the intensity of each factor; wherein the meteorological factor intensity = real-time monitoring meteorological value × scenario coefficient (such as rainstorm intensity × open-air blasting area coefficient), the geological factor intensity = real-time sensor reading × rock structure coefficient (such as displacement rate × fissure development degree), and the hydrological factor intensity = osmotic pressure gradient × hydrogeological coefficient (such as water level change × tailings pond permeability) ;

[0107] The three types of factors and their intensity values are taken as inputs to initialize the nodes of the risk propagation topology graph, with meteorological nodes marked as Mn, geological nodes marked as Gn, and hydrological nodes marked as Hn. The initial weights of the edges of the risk propagation topology graph are set based on the historical disaster database (the weight of rainstorm triggering landslide from meteorological node Mn to geological node Gn = 0.85); the initial risk values of each node are calculated in combination with the intensity of the three types of factors = factor intensity x risk propagation topology graph edge weight weighted sum;

[0108] In combination with the injected real-time meteorological forecast data, the meteorological forecast data is mapped to 10m x 10m grids through Kriging interpolation, wherein the real-time meteorological forecast data covers the meteorological information of the mining area and the surrounding area within a future period of time; at the same time, in combination with the initial risk values of each node at the current time and the influence of real-time meteorological data, the risk values of each node are updated to obtain the dynamic risk values: the initial risk values of meteorological nodes are directly replaced by the interpolated meteorological forecast values, thereby obtaining the dynamic risk values of meteorological nodes, and the initial risk values of geological or hydrological nodes are updated according to the risk transmission formula to obtain their dynamic risk values:

[0109] ,

[0110] In the formula, is the dynamic risk value of node i at time t+1; is the initial risk retention coefficient, and ; is the initial risk value of node i at time t; is the risk propagation influence coefficient, and ; is the set of adjacent nodes of node i; is the dynamic risk value of node j at time t; is the edge weight of node i-j; is the spatial decay coefficient; is the Euclidean distance of node i-j; is the time step;

[0111] Taking the meteorological node as the starting point, the risk transmission is iteratively calculated along the edges of the risk propagation topology graph, and finally a regional risk potential thermal map is output; wherein the regional risk potential thermal map is in units of grids, and the maximum value of the dynamic risk values of all nodes in the region is selected as the thermal value of the regional risk potential thermal map; the coordinates of each grid and its corresponding thermal value are combined into a key-value pair to form a grid coordinate-thermal value key-value pair set, which is taken as the input data for spatial clustering analysis.

[0112] It needs to be further explained that, in the specific implementation process, the scene self-adaptive regulation instruction generation module performs spatial clustering analysis on the regional risk potential heat map to identify a high-risk aggregation area; in combination with a preset risk warning value, a corresponding scene self-adaptive regulation instruction is generated for the identified high-risk aggregation area, and the process is as follows:

[0113] Based on the regional risk potential heat map, a high-risk area is identified: a heat value threshold is set, all grids in the regional risk potential heat map are traversed, and grids with a heat value greater than or equal to the heat value threshold in the region are extracted; density clustering analysis is performed on the extracted continuous high-risk grids; in the clustering process, grids with similar heat values are clustered into a class according to the spatial distance and heat value similarity between the grids; each grid region in the clustering result is marked as a high-risk aggregation area, including an edge transition zone, and each high-risk aggregation area is assigned a unique identifier;

[0114] After the preset risk warning value, a scene self-adaptive regulation instruction is generated according to the identified high-risk aggregation area: if the heat value is greater than or equal to the risk warning value A1 and less than the risk warning value A2, a first warning (blue) is generated, the device vibration warning and LED boundary projection (marking a safe evacuation path) are triggered for the area; if the heat value is greater than or equal to the risk warning value A2 and less than the risk warning value A3, a second warning (yellow) is generated, the area is broadcasted with an automatic broadcasting system to report risk avoidance instructions (such as stopping work immediately in a tunnel); if the heat value is greater than or equal to the risk warning value A3, a third warning (red) is generated, the area is forced to cut off the power supply and activate the emergency ventilation; wherein the above instructions need to meet the minimum production interruption cost and the maximum personnel safety redundancy.

[0115] In addition, the formulas involved in the above are calculated by removing the dimension and taking the numerical value, which is obtained by collecting a large amount of data to simulate the closest real situation. The proportion coefficient in the formula and each preset threshold in the analysis process are set by a person skilled in the art according to the actual situation or obtained by a large amount of data simulation; the size of the proportion coefficient is to obtain a specific numerical value of each parameter for subsequent comparison, and the size of the proportion coefficient depends on the number of sample data and the corresponding processing coefficient initially set by a person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameters and the quantized numerical value.

[0116] Each embodiment in the specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it basically relies on the method embodiment, the description is relatively simple, and the relevant part can be referred to the part of the method embodiment.

[0117] For convenience of description, the above apparatus is described in various units by function for separate description. Of course, the functions of the units can be implemented in one or more software and / or hardware in implementing the present application.

[0118] Those skilled in the art should understand that embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0119] The present application is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0120] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the functions specified in the flowcharts and / or block diagrams.

[0122] Secondly, in the drawings of the embodiments disclosed in the present application, only the structures related to the embodiments disclosed in the present application are involved, and other structures can be referred to the general design. In the case of no conflict, the same embodiments and different embodiments of the present application can be combined with each other.

[0123] Finally: the above is only the preferred embodiment of the present application, but the scope of protection of the present application is not limited to this, any skilled in the technical field within the scope of the present application disclosed by the technical range, according to the technical solution of the present application and its inventive concept to make equivalent replacement or change, should be covered within the scope of protection of the present application.

Claims

1. A weather risk intelligent early warning system for mining industry, characterized in that: The method comprises a multi-field coupling monitoring network construction module, a weather threat factor acquisition module, a risk propagation and quantitative analysis module, and a scene adaptive regulation instruction generation module. The multi-field coupling monitoring network construction module is used for constructing a multi-field coupling monitoring system in a mining area. The weather threat factor acquisition module is used for dynamically synthesizing a weather threat factor of a mining area scene based on a mining area space-time meteorology-geology coupling data set and by fusing mining scene metadata and time sequence characteristics through a conditional adversarial generation network. The risk propagation and quantitative analysis module is used for dynamically quantitatively analyzing the synthesized weather threat factor, and establishing a geology body-meteorology field coupling risk propagation topology graph. The scene adaptive regulation instruction generation module is used for performing spatial clustering analysis on the regional risk potential thermal map, identifying a medium-high risk aggregation area, and generating a corresponding scene adaptive regulation instruction for the identified medium-high risk aggregation area in combination with a preset risk warning value. The multi-field coupling monitoring network construction module is used for constructing a multi-field coupling monitoring system in a mining area. The process of generating a mining area space-time meteorology-geology coupling data set through real-time data driving of the multi-field coupling monitoring network comprises the following steps.

2. The weather risk intelligent early warning system for mining industry according to claim 1, characterized in that: Based on a dangerous source distribution map stored in a mining area geographic information system and a database, a monitoring area is divided, and the Internet of Things nodes are densely arranged in key monitoring areas of a pit slope of an open pit, an intersection or a fault zone of an underground tunnel, a bottom of a dump, and a dam body of a tailings pond. Each node of the Internet of Things node group is integrated with three types of sensors, including a micro meteorological station, a geology body deformation sensor, and a osmometer. A GPS time granting submodule is built in the node, and all sensor data is wirelessly transmitted to an edge computing gateway through LoRa or NB-IoT. The monitored sensor data is preprocessed to generate a mining area space-time meteorology-geology coupling data set. The process of dynamically synthesizing a weather threat factor of a mining area scene by the weather threat factor acquisition module based on a mining area space-time meteorology-geology coupling data set and by fusing mining scene metadata and time sequence characteristics through a conditional adversarial generation network comprises the following steps. S1, obtain mining scene metadata, wherein the mining scene metadata comprises three types of heterogeneous information, i.e., mining method, rock stratum structure characteristics, and current operation stage.

3. The weather risk intelligent early warning system for mining industry according to claim 1, characterized in that: The mining method is a categorical variable, including open pit blasting and underground tunneling. The rock stratum structure characteristics are continuous variables, including fracture development degree and rock mass strength. The current operation stage is a categorical variable, including drilling, transportation, and heap leaching. S2, converting the mining scene metadata into a unified scene joint feature vector by embedding layer and standardization processing ; S3, set the generator, the time series characteristics of the mining area space-time meteorology-geology coupling dataset, the scene joint characteristics of the mining scene metadata are input into the LSTM network to extract the time series dependence, and the candidate weather threat factors are generated through the full connection layer ; S4, setting a discriminator, inputting real weather threat factors candidate weather threat factors output by the generator , extracting spatial features through a convolutional neural network, and outputting a discrimination probability; S5, the game process of optimizing the generator and the discriminator by using the Wasserstein distance objective function, finally making the generator output the mine scene weather threat factor consistent with the distribution of the real weather threat factor.

4. The weather risk intelligent early warning system for mining industry according to claim 3, characterized in that: The process of converting the mining scene metadata into a unified scene joint feature vector through embedding layer and standardization processing includes: For categorical variable embedding, define the mining method as and assign open pit blasting as 1, underground excavation as 2, define the current operation stage as and assign drilling as 1, transportation as 2, heap leaching as 3; convert to continuous vector through embedding layer: wherein is an extraction mode embedding vector, is a current job phase embedding vector, is an embedding function, are embedding layer parameters with output dimensions of extraction mode embedding dimension , job phase embedding dimension , respectively; fracture development degree in the rock formation structure characteristics with the rock mass strength standardization: ， wherein, is the normalized fracture intensity, is the mean fracture intensity, is the standard deviation of the fracture intensity, is the normalized rock mass strength, is the mean rock mass strength, is the standard deviation of the rock mass strength; Concatenating embedding vectors with normalized numerical variables as scene joint features : where , is the set of real numbers.

5. The weather risk intelligent early warning system for mining industry according to claim 3, characterized in that: The process of setting the generator, inputting the time sequence features of the mine space-time meteorology-geology coupling dataset and the scene joint features of the mining scene metadata into the LSTM network to extract time sequence dependence, and generating the candidate weather threat factor through the full connection layer includes: a rate of change of meteorological data a rate of change of geological data and a rate of change of hydrological data are defined as , , respectively, where is a time step, is meteorological data at time t, is geological data at time t, is hydrological data at time t. In the past hour data as a window, construct a time series feature sequence : , wherein, , is the time step, and 3 is the meteorology, geology, and hydrology elements. The generator processes the sequence of time-series features through an LSTM layer , outputting a hidden state : where are the LSTM layer parameters, and , is the LSTM hidden layer dimension; concatenate the LSTM hidden states contextual features After concatenation, the candidate weather threat factors are generated by the generator fully connected layer : ， wherein is the weight matrix for the generator fully connected layer; is the bias term for the generator fully connected layer.

6. A weather risk intelligent early warning system for mining industry according to claim 3, characterized in that: The process of setting the discriminator, inputting the real weather threat factor and the candidate weather threat factor output by the generator, extracting spatial features through the convolutional neural network, and outputting the discrimination probability includes: Defining spatial features as : ， In the formula, is a fissure development degree of a region mask; is a meteorological data of a region mask; is a rock mass strength of a region mask; is a geological data of a region mask; is an element-by-element multiplication; , is a spatial feature weight coefficient; wherein the region mask is a binary matrix, is a fissure development degree threshold value, is a meteorological data threshold value, is a rock mass strength threshold value, is a geological data threshold value; The discriminator converts the spatial features to discriminative probabilities through a fully connected layer ， wherein is an activation function; is a weight matrix of the discriminator fully connected layer; is a bias term of the discriminator fully connected layer; is the probability of the authenticity of the input sample judged by the discriminator.

7. The weather risk intelligent early warning system for mining industry according to claim 3, characterized in that: The process of optimizing the generator and the discriminator by using the Wasserstein distance objective function, finally making the generator output the mine scene weather threat factor consistent with the distribution of the real weather threat factor includes: The Wasserstein distance objective function is: ， wherein is the loss function for the Wasserstein generator, is the expected output of the discriminator under the real data distribution; is the expected output of the discriminator under the generated data distribution; is the distribution of real weather threat factors; is the interpolation of real weather threat factors and generated samples, , is a randomly sampled interpolation factor, and ; is a gradient penalty coefficient; is the gradient norm of the discriminator at the interpolated sample . By alternating training the discriminator and the generator, the objective function is optimized , eventually making the generator output consistent with the real weather threat factor distribution: fixing the generator , updating the discriminator to minimize; fixing the discriminator , updating the generator to maximize.

8. The weather risk intelligent early warning system for mining industry according to claim 1, characterized in that: The risk propagation and quantitative analysis module performs dynamic quantitative analysis on the synthesized weather threat factor, and establishes a geologic body-meteorological field coupling risk propagation topology graph; The process of realizing risk conduction through real-time data driving and then generating a regional risk potential thermal map includes: Subdivide the weather threat factor type into meteorological factor, geological factor and hydrological factor, and based on the identified factor type, construct a geologic body-meteorological field coupling risk propagation topology graph; in the risk propagation topology graph, the nodes represent various factors, and the edges represent the potential risk association relationship between the nodes; Respectively extract the real-time monitoring values of each factor and calculate the intensity of each factor; The three types of factors and intensity values thereof are taken as inputs to initialize nodes of the risk propagation topology graph, meteorological nodes are marked as Mn, geological nodes are marked as Gn, and hydrological nodes are marked as Hn; initial weights of edges of the risk propagation topology graph are set based on a historical disaster database; and initial risk values of the nodes are calculated in combination with the intensity of the three types of factors = factor intensity x weighted sum of risk propagation topology graph edge weights Combine the injected real-time meteorological forecast data, map the meteorological forecast data to 10m*10m grid through Kriging interpolation; at the same time, combine the risk initial value of each node at the current time and the influence of real-time meteorological data, update the risk value of each node to obtain the risk dynamic value: replace the initial risk value of the meteorological node with the interpolated meteorological forecast value, thereby obtaining the risk dynamic value of the meteorological node, update the initial risk value of the geological or hydrological node according to the risk conduction formula to obtain its risk dynamic value: ， wherein, is the risk dynamic value of node i at time t+1; is the initial risk reserve coefficient, and ; is the initial risk value of node i at time t; is the risk propagation influence coefficient, and ; is the set of adjacent nodes of node i; is the risk dynamic value of node j at time t; is the edge weight of node i-j; is the spatial decay coefficient; is the Euclidean distance of node i-j; is the time step. Take the meteorological node as the starting point, iteratively calculate the risk conduction along the edges of the risk propagation topology graph, and finally output the regional risk potential thermal map; wherein, the regional risk potential thermal map is in grid unit, and the maximum value of the risk dynamic value of all nodes in the region is selected as the thermal value of the regional risk potential thermal map; combine the coordinates of each grid with its corresponding thermal value to form a key-value pair set of grid coordinates-thermal value key-value pairs.

9. The weather risk intelligent early warning system for mining industry according to claim 1, characterized in that: The scene adaptive regulation instruction generation module performs spatial clustering analysis on the regional risk potential thermal map, and identifies the medium and high risk aggregation area; The process of generating the corresponding scene adaptive regulation instruction for the identified medium and high risk aggregation area in combination with the preset risk warning value includes: Based on the sub-regional risk potential heat map, the high-risk area is identified: set the threshold value of heat value, traverse all the grids in the sub-regional risk potential heat map, extract the grid whose heat value is greater than or equal to the threshold value of heat value; carry out density clustering analysis on the extracted continuous high-risk grid; in the clustering process, according to the spatial distance between the grids and the similarity of the heat value, the grids with similar heat value are clustered into a class; each grid region in the clustering result is marked as a high-risk aggregation area, including the edge transition zone, and each high-risk aggregation area is assigned a unique identifier; After presetting the risk warning value, scene adaptive control instructions are generated according to the identified high-risk aggregation area: if the heat value is greater than or equal to the risk warning value A1 and less than the risk warning value A2, a first-level warning is generated, the device vibration warning and LED boundary projection in the region are triggered; if the heat value is greater than or equal to the risk warning value A2 and less than the risk warning value A3, a second-level warning is generated, the region is broadcasted with the automatic broadcasting system to broadcast the risk avoidance instruction; if the heat value is greater than or equal to the risk warning value A3, a third-level warning is generated, the power supply in the region is forcibly cut off and the emergency ventilation is activated.

Images