Mine fire monitoring method and system, device, medium

Abstract

The application provides a mine fire monitoring method and system, equipment and a medium, and belongs to the technical field of mine safety. The method comprises the following steps: acquiring multi-dimensional data corresponding to each target monitoring point in a target mine; inputting each multi-dimensional data into a double-flow monitoring model to output fire monitoring results corresponding to each target monitoring point; the double-flow monitoring model is a double-branch structure; the double-branch structure comprises a first branch and a second branch; the first branch and the second branch have different data processing modes; based on the fire monitoring results corresponding to each target monitoring point, an abnormal point risk cluster of the target mine is determined; the abnormal point risk cluster is a cluster in which abnormal monitoring points are concentrated and distributed in space in the target mine; and the fire monitoring result of the target mine is determined based on the abnormal point risk cluster. The application can realize accurate monitoring of mine fires.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of mine safety, and more particularly relates to a mine fire monitoring method and system, equipment and medium. BACKGROUND

[0002] As the core scene of underground resource exploitation, the safety production of a mine is directly related to the life safety of operating personnel and the development efficiency of mineral resources. Among them, mine fire is one of the high-incidence and serious safety accidents in the production process of a mine, and its causes include coal spontaneous combustion, electrical equipment failure, mechanical friction fire and other types. After the occurrence of a fire, not only toxic and harmful gases will be released, and the mine ventilation system will be damaged, but also secondary disasters such as gas explosion and roof collapse may be caused, resulting in casualties and significant economic losses. Therefore, timely and accurate monitoring of mine fire is one of the core needs to ensure the safety production of a mine.

[0003] Therefore, how to realize accurate monitoring of mine fire has become a key problem to be solved in the current mine safety field. SUMMARY

[0004] The purpose of the present application is to provide a mine fire monitoring method and system, equipment and medium, which can realize accurate monitoring of mine fire.

[0005] The first aspect of the embodiment of the present application provides a mine fire monitoring method, comprising: acquiring multi-dimensional data corresponding to each target monitoring point in a target mine; each target monitoring point is arranged in a different monitoring area in the target mine; each multi-dimensional data is obtained by aligning and splicing temperature data and gas data of the corresponding target monitoring point according to time steps;

[0006] Each multi-dimensional data is input into a double-flow monitoring model to output fire monitoring results corresponding to each target monitoring point; the double-flow monitoring model is a double-branch structure; the double-branch structure includes a first branch and a second branch; the first branch and the second branch have different data processing modes;

[0007] Based on the fire monitoring results corresponding to each target monitoring point, an abnormal point risk cluster of the target mine is determined; the abnormal point risk cluster is a cluster in which abnormal monitoring points are concentrated in space in the target mine;

[0008] The fire monitoring result of the target mine is determined based on the abnormal point risk cluster.

[0009] In a second aspect, the embodiment of the present application provides a mine fire monitoring system, comprising: a data acquisition module, configured to acquire multi-dimensional data corresponding to each target monitoring point in a target mine; each target monitoring point is arranged in a different monitoring area in the target mine; each multi-dimensional data is obtained by aligning and splicing temperature data and gas data of the corresponding target monitoring point according to time steps;

[0010] a model prediction module, configured to input each multi-dimensional data into a double-flow monitoring model, and output fire monitoring results corresponding to each target monitoring point; the double-flow monitoring model is a double-branch structure; the double-branch structure comprises a first branch and a second branch; the first branch and the second branch are different in data processing mode;

[0011] a judgment module, configured to determine an abnormal point risk cluster of the target mine based on the fire monitoring results corresponding to each target monitoring point; the abnormal point risk cluster is a cluster in which abnormal monitoring points are concentratedly distributed in space in the target mine;

[0012] a result output module, configured to determine a fire monitoring result of the target mine based on the abnormal point risk cluster.

[0013] In a third aspect, the embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor; when the processor executes the computer program, the steps of the mine fire monitoring method described above are implemented.

[0014] In a fourth aspect, the embodiment of the present application provides a computer readable storage medium, which stores a computer program; when the computer program is executed by a processor, the steps of the mine fire monitoring method described above are implemented.

[0015] In a fifth aspect, the embodiment of the present application provides a computer program product, comprising a computer program or computer executable instructions; when the computer program or computer executable instructions are executed by a processor, the steps of the mine fire monitoring method described above are implemented.

[0016] The mine fire monitoring method and system, device and medium provided by the embodiment of the present application have the following advantages:

[0017] Compared with the traditional single data collection mode, the embodiment of the application can accurately monitor the temperature and the concentration of various gases, can comprehensively reflect the real-time conditions of different areas of the mine, can avoid misjudgment caused by one-sided information, can provide a scientific basis for fire condition judgment, can greatly improve the reliability of the monitoring foundation, and can help to more accurately detect fire hazards. The embodiment of the application adopts a double-flow monitoring model with a double-branch structure, different branches process data in different ways, can analyze fire risks from multiple dimensions, takes into account the characteristics of different types of fires, and improves the comprehensiveness and accuracy of fire risk analysis. The embodiment of the application determines an abnormal point risk cluster based on the fire monitoring results of each monitoring point, considers the spatially concentrated abnormal monitoring points as a cluster, this way fully considers the spatial spread characteristics of the fire, is more forward-looking than only focusing on the abnormality of a single monitoring point, when multiple adjacent monitoring points simultaneously appear abnormal, the abnormal point risk cluster formed can more accurately indicate the area where the fire is likely to occur, can early warn potential fire hazards, and avoid missing the fire prevention opportunity due to neglecting spatial correlation. The embodiment of the application determines the mine fire monitoring result based on the abnormal point risk cluster, realizes the quantitative evaluation from the local abnormal point to the overall risk of the mine, provides comprehensive and quantitative information for decision makers, improves the overall efficiency and effect of mine fire prevention and control, and safeguards the safe production of the mine. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0019] Figure 1 The flowchart of the mine fire monitoring method provided by an embodiment of the application is shown in the figure.

[0020] Figure 2 The single-point monitoring result execution flowchart of the mine fire monitoring method provided by an embodiment of the application is shown in the figure.

[0021] Figure 3 The structural block diagram of the mine fire monitoring system provided by an embodiment of the application is shown in the figure.

[0022] Figure 4 The schematic block diagram of the electronic device provided by an embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0023] In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular architectures, techniques, etc. in order to provide a thorough understanding of the embodiments of the application. However, it will be apparent to those skilled in the art that the application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0024] In order to make the purposes, technical solutions and advantages of the present application clearer, the following will be described by specific embodiments in conjunction with the accompanying drawings.

[0025] Please refer to Figure 1 , Figure 1 The flowchart of the mine fire monitoring method provided by an embodiment of the present application can be executed by an electronic device, specifically a computer, a server or the like, and the method can include the following steps.

[0026] S101: Obtain multi-dimensional data corresponding to each target monitoring point in the target mine; each target monitoring point is arranged in a different monitoring area in the target mine; each multi-dimensional data is obtained by aligning and splicing temperature data and gas data of the corresponding target monitoring point according to time steps.

[0027] In this embodiment, the target mine refers to a specific mine targeted by the fire monitoring task, which includes all mining faces, roadways, chambers and other areas in the mine that need to be prevented and controlled from fire risks. The target monitoring point is a specific point set in the target mine for collecting environmental data, which is the direct source of monitoring data and needs to be reasonably distributed according to the fire risk characteristics of different areas in the mine to ensure coverage of key risk areas. Different monitoring areas refer to differentiated areas in the target mine according to geological conditions, mining functions, and fire risk types, such as coal mining faces, return air roadways, mechanical and electrical chambers, and goaf areas. These areas have significant differences in fire inducing factors and risk levels. Multi-dimensional data is a collection of data collected by a single target monitoring point, including multiple environmental parameters, specifically including a combination of temperature data and gas data, which can reflect the environmental status of the area where the monitoring point is located from multiple dimensions and provide a comprehensive basis for fire risk judgment. Temperature data is continuous data collected by the target monitoring point, reflecting the environmental temperature of the area where it is located, and is one of the core physical indicators for judging fire risk. Gas data is data collected by the target monitoring point, reflecting the concentration of specific gases in the air in the area where it is located. In mine fire monitoring, the focus is on gases related to combustion or spontaneous combustion, such as carbon monoxide, carbon dioxide, and methane. Time step alignment refers to matching temperature data and gas data collected by the same target monitoring point at the same time interval in the time dimension, ensuring that each set of temperature data corresponds to gas data at the same time point, avoiding analysis bias caused by data collection time difference. Splicing refers to combining temperature data and gas data at the same time point into a complete data record after time step alignment, forming multi-dimensional data for the monitoring point.

[0028] In this embodiment, multi-dimensional data is a data set for the area where the monitoring point is located within a certain period of time, including corresponding temperature data and gas data within that period of time. The size of the time period is the same as the size of the second time scale.

[0029] In this embodiment, a regional risk assessment of the target mine must first be conducted to determine the number and location of monitoring points based on the fire risk level of different monitoring areas. For example, goaf areas are high-risk areas for spontaneous combustion, and electromechanical chambers are high-risk areas for open flames. High-risk areas require denser monitoring, such as one monitoring point every 30 meters in the goaf area; low-risk areas can have a reduced density, such as one monitoring point every 100 meters in the main haulage roadway. Simultaneously, it must be ensured that monitoring points cover all critical areas without blind spots. To ensure data accuracy, each target monitoring point must be equipped with a dedicated sensor, such as a thermocouple temperature sensor or an infrared gas sensor, and a uniform acquisition frequency must be set. During the acquisition process, the corresponding acquisition timestamp must be recorded in real time to avoid data loss or time misalignment. Furthermore, the sensors need to be calibrated periodically to ensure the reliability of the acquired data. Finally, by matching timestamps, temperature and gas data at the same time point are filtered out. Following a fixed format of "time-temperature-gas concentration", the two types of data are combined into a single multidimensional data record. The multidimensional data from all time points are arranged in chronological order to form a continuous multidimensional dataset for the monitoring point, which facilitates subsequent model input and analysis.

[0030] S102: Input the various multi-dimensional data into the dual-flow monitoring model and output the fire monitoring results corresponding to each target monitoring point; the dual-flow monitoring model has a dual-branch structure; the dual-branch structure includes a first branch and a second branch; the data processing methods of the first branch and the second branch are different.

[0031] In this embodiment, the dual-flow monitoring model refers to an intelligent analysis model used to process multidimensional data from target monitoring points and output fire monitoring results. Its core feature is a dual-branch structure, which analyzes data using two different data processing methods to achieve a comprehensive assessment of fire risk. It is a key tool connecting multidimensional data with single-point fire monitoring results. The dual-branch structure is the core architecture of the dual-flow monitoring model, consisting of a first branch and a second branch with different functions and data processing methods. The two branches output corresponding monitoring results in parallel. By fusing the first and second monitoring results, the final single-point fire monitoring result is obtained, simultaneously covering different types of fire risk analysis needs. The first branch refers to one of the data processing branches in the dual-branch structure. It possesses specific data processing logic and can extract features related to a certain type of fire risk from multidimensional data, such as trend features over a longer time scale, and output corresponding monitoring results based on these features. The second branch refers to another data processing branch in the dual-branch structure. Its data processing logic differs significantly from the first branch. It can capture another type of fire risk-related characteristics from multidimensional data, such as abnormal fluctuation characteristics at a shorter time scale, and then output corresponding monitoring results. These results complement the monitoring results of the first branch, jointly supporting the assessment of single-point fire risk. The fire monitoring results refer to the assessment results output by the dual-flow monitoring model after analyzing the multidimensional data of a single target monitoring point, indicating whether there is a fire risk in the area where the monitoring point is located and the degree of risk. This data forms the basis for subsequently determining the risk clusters of abnormal points and the overall fire monitoring results of the mine.

[0032] In this embodiment, various multidimensional data are input into the first and second branches of the dual-flow monitoring model. The first branch focuses on extracting global trend features over a long time scale from the multidimensional data, such as the temperature and gas concentration variation patterns over several consecutive hours. The second branch focuses on capturing local abnormal fluctuation features over a short time scale, such as a sudden rise in temperature or a sudden increase in gas concentration within a few minutes. The data processing methods of the two branches can respectively cover the risk analysis needs of cumulative spontaneous combustion fires and sudden open flame fires. According to the preset fusion rules, the model fuses the monitoring results of the first branch and the monitoring results of the second branch to obtain the final fire monitoring result for the target monitoring point.

[0033] Furthermore, before inputting the various multidimensional data into the dual-flow monitoring model, the multidimensional data of each target monitoring point can be preprocessed to remove outliers caused by temporary sensor failures and other noise, and the data format can be converted to a format that is compatible with the input requirements of the dual-flow monitoring model.

[0034] S103: Based on the fire monitoring results corresponding to each target monitoring point, determine the abnormal point risk cluster of the target mine; the abnormal point risk cluster is a cluster in the target mine where abnormal monitoring points are spatially concentrated.

[0035] In this embodiment, an anomaly risk cluster is a group formed by the aggregation of spatially adjacent anomaly monitoring points within the target mine. Its core characteristic is the concentrated spatial distribution of these anomaly monitoring points, reflecting the regional clustering trend of fire risk underground and providing a basis for determining potential fire spread areas. An anomaly risk cluster contains at least one anomaly monitoring point.

[0036] In this embodiment, firstly, differentiated abnormal state judgment criteria are established for each target monitoring point. These criteria need to be formulated in conjunction with the geological conditions of the area where the monitoring point is located, such as whether it is a coal seam area prone to heat accumulation; mining technology, such as whether it is a working face with high mining intensity; and historical fire data, such as whether spontaneous combustion accidents have occurred. The fire monitoring results of each monitoring point are compared with the corresponding judgment criteria. If the risk level shown by the monitoring results exceeds the standard threshold, the monitoring point is marked as an abnormal monitoring point. Then, the precise spatial coordinates of all abnormal monitoring points in the target mine are obtained, and a preset spatial distance threshold is set. This threshold needs to be determined based on the width of the mine roadway and the density of the monitoring point layout. Furthermore, the straight-line distance between each abnormal monitoring point can be calculated using a spatial clustering algorithm. Abnormal monitoring points with a distance less than the preset spatial distance threshold are grouped into the same cluster, while those with a distance greater than the threshold are either grouped into a separate cluster or excluded from the existing clusters, ultimately forming risk clusters for all abnormal points in the target mine.

[0037] For example, taking a target mine as an example, the mine has obtained fire monitoring results from 22 target monitoring points. Among them, monitoring points No. 3, 4, and 5 of the 1201 working face and monitoring points No. 2 and 3 of the 1102 goaf are identified as abnormal monitoring points. First, the spatial coordinates of these abnormal monitoring points are obtained: the coordinates of No. 3 of the 1201 working face are (1200, 850), No. 4 are (1220, 855), and No. 5 are (1240, 860); the coordinates of No. 2 of the 1102 goaf are (950, 620) and No. 3 are (970, 625). A preset spatial distance threshold of 30 meters is set. Calculations revealed that the distances between monitoring points 3, 4, and 5 in the 1201 working face were 20.6 meters (distance between points 3 and 4), 20.8 meters (distance between points 4 and 5), and 41.2 meters (distance between points 3 and 5), respectively. The distances between points 3 and 4, and between points 4 and 5, were all less than 30 meters, and therefore belonged to the same cluster. The distance between monitoring points 2 and 3 in the 1102 goaf was 20.6 meters, less than 30 meters, and belonged to another cluster. The distances between the abnormal clusters in the 1201 working face and the 1102 goaf were far greater than 30 meters and were not merged. Ultimately, two abnormal risk clusters were identified for this target mine: the first cluster contained abnormal monitoring points 3, 4, and 5 in the 1201 working face, and the second cluster contained abnormal monitoring points 2 and 3 in the 1102 goaf. These two clusters corresponded to concentrated risk points in two different areas within the mine.

[0038] S104: Determine the fire monitoring results of the target mine based on the risk cluster of anomalies.

[0039] In this embodiment, the fire monitoring results of the target mine refer to the fire risk status conclusion at the entire mine level, which is comprehensively assessed based on the characteristics and risk levels of all abnormal point risk clusters within the target mine. It covers key information such as the overall risk level of the mine, the distribution of high-risk areas, and the tendency of risk types. It is the core basis for mine managers to formulate overall fire prevention and control strategies, and is different from the local fire monitoring results of a single target monitoring point.

[0040] As can be seen from the above, this application embodiment, by deploying target monitoring points in different monitoring areas of the target mine and collecting multi-dimensional data, can comprehensively reflect the real-time status of different areas of the mine compared to the traditional single data acquisition method. This avoids misjudgments caused by incomplete information, provides a scientific basis for fire assessment, greatly improves the reliability of the monitoring foundation, and helps to more accurately detect fire hazards. This application embodiment, by adopting a dual-branch structure dual-flow monitoring model, with different branches processing data in different ways, can analyze fire risks from multiple dimensions, taking into account the characteristics of different types of fires and improving the comprehensiveness and accuracy of fire risk analysis. This application embodiment determines anomaly risk clusters based on the fire monitoring results of each monitoring point, treating spatially concentrated anomaly monitoring points as a cluster. This method fully considers the spatial spread characteristics of fire and is more forward-looking than focusing only on the anomalies of a single monitoring point. When multiple adjacent monitoring points show anomalies simultaneously, the resulting anomaly risk clusters can more accurately indicate the areas where fires may occur, providing early warning of potential fire hazards and avoiding delays in fire prevention and control due to neglecting spatial correlation. The embodiments of this application determine the mine fire monitoring results by identifying anomaly risk clusters, realizing a quantitative assessment from local anomalies to the overall mine risk, providing decision-makers with comprehensive and quantitative information, improving the overall efficiency and effectiveness of mine fire prevention and control, and ensuring safe production in the mine.

[0041] In one embodiment of this application, for each multidimensional data point, the multidimensional data is input into a dual-flow monitoring model, and the fire monitoring result of the target monitoring point corresponding to the multidimensional data is output, including:

[0042] Input the multidimensional data into the first branch to obtain the first monitoring result;

[0043] Input the multidimensional data into the second branch to obtain the second monitoring result;

[0044] The fire monitoring results of the target monitoring point corresponding to the multidimensional data are calculated based on the first and second monitoring results.

[0045] In this embodiment, the first monitoring result refers to the assessment result reflecting a certain type of fire risk at the monitoring point, output by the first branch of the dual-flow monitoring model after the multi-dimensional data of the target monitoring point is input into the model through a specific data processing method. It is an important component in the subsequent calculation of the final fire monitoring result for the target monitoring point, and its data dimensions and risk assessment dimensions differ from the second monitoring result. The second monitoring result refers to the assessment result reflecting another type of fire risk at the monitoring point, output by the second branch of the dual-flow monitoring model after the multi-dimensional data of the target monitoring point is input into the model through a differentiated data processing method. It complements the first monitoring result and together provides the basis for calculating the final fire monitoring result for the target monitoring point.

[0046] In this embodiment, multidimensional data from a single target monitoring point are simultaneously input into the first and second branches of the dual-flow monitoring model. The first branch, following a preset long-term trend analysis logic, extracts the slope of temperature and gas concentration changes over several consecutive hours, matches this data with a spontaneous combustion fire risk feature database, and generates a first monitoring result containing risk level and risk type. The second branch, following a preset short-term anomaly analysis logic, identifies fluctuations such as sudden temperature increases and gas concentration spikes within a few minutes, matches this data with an open flame fire risk feature database, and generates a second monitoring result containing risk level and risk type. Finally, based on the fire risk characteristics of the area where the target monitoring point is located, corresponding calculation weights are assigned to the first and second monitoring results to ensure that the weights are adapted to the regional risk attributes. The product of the risk level quantification value of the first monitoring result and its corresponding weight is added to the product of the risk level quantification value of the second monitoring result and its corresponding weight to obtain the comprehensive risk quantification value of the target monitoring point, thus determining the fire monitoring result for that monitoring point.

[0047] As can be seen from the above, the embodiments of this application focus on two typical mine fire types—cumulative spontaneous combustion and sudden open flame—through the first branch and the second branch, respectively. This avoids the shortcomings of traditional single models that cannot simultaneously cover different fire characteristics, ensuring the effective identification of different types of fire risks. The first monitoring result and the second monitoring result clearly correspond to the risk type and level, and the final calculation result can clearly trace the source of the risk, providing a more specific basis for subsequent anomaly analysis and prevention and control measures. This avoids the problem of ambiguous prevention and control direction caused by only outputting a single risk value, further supporting the accuracy and practicality of mine fire monitoring.

[0048] In one embodiment of this application, the first branch is a trend prediction branch; in this embodiment, inputting the multidimensional data into the first branch to obtain a first monitoring result includes:

[0049] Input multidimensional data into the trend prediction branch;

[0050] Global trend features are obtained by extracting global features over time based on multidimensional data; global trend features are used to characterize the first-time change pattern of multidimensional data at the first time scale.

[0051] Fire risk matching is performed based on global trend characteristics to obtain the first monitoring result; the first monitoring result is used to characterize the first-time fire risk trend of the target monitoring point at the first time scale.

[0052] In this embodiment, the trend prediction branch, i.e., the first branch in the dual-flow monitoring model, is a functional branch specifically designed to extract the changing trends of multidimensional data over a longer time scale and analyze cumulative fire risks. Its core function is to identify fire risks, such as coal seam spontaneous combustion, that require a long period of accumulation to manifest, by capturing the long-term changing patterns of the data, distinguishing it from the second branch which focuses on short-term anomalies. Global trend features refer to features extracted from the multidimensional data of the target monitoring point over a longer time range that reflect the overall direction and pattern of data changes. Examples include the stable increase in temperature over several consecutive hours and the gradual rate of change in gas concentration over time. These features can intuitively demonstrate the changing trends of data over a long period and are a key basis for judging cumulative fire risks. The first time scale is the time range set for the trend prediction branch to extract global trend features. This time scale is usually long, typically measured in hours or days, aiming to cover the time cycle from the emergence to the manifestation of cumulative fire risks, ensuring that the long-term changing patterns of the data can be captured. Its duration is significantly longer than the time scale corresponding to the second branch. The setting of the first time scale needs to take into account two core factors: first, the latency period of spontaneous combustion of coal seams in the target mine, ensuring that the time scale can cover this period and avoid the risk of slow accumulation due to missed detection due to too short a time; second, the rate of change of multidimensional data, ensuring that the data can show a identifiable first-time change pattern within this time scale, rather than random fluctuations. The first time scale can be set to 4-8 hours.

[0053] In this embodiment, the first branch has a memory function, enabling the fusion analysis of historical and current data. When multidimensional data is input into the first branch, it retrieves multiple segments of historical multidimensional data, concatenates them according to time sequence, ensuring that the overall data duration reaches a preset first time scale. Once the first time scale is reached, the multidimensional data corresponding to the first time scale is analyzed globally. That is, the multidimensional data corresponding to the first time scale... , where D i For multidimensional data; a = K / k, where k is D i Time scale; K is the first time scale. Specifically, when i=1, D1 is the currently input multidimensional data; when i=2, D2 is the multidimensional data input before D1; when i=3, D3 is the multidimensional data input before D2, and so on, until a=K / k, that is, the duration of each overall data reaches the preset first time scale.

[0054] In this embodiment, the multidimensional data of the target monitoring point is first converted into a time-series data format supported by the trend prediction branch (first branch) and input into the first branch. If the first time scale is set to 4 hours, the multidimensional data of the target monitoring point and the multidimensional data of historical target monitoring points are continuously combined into data segments every 4 hours. The first branch uses a time-series feature extraction algorithm to perform a global analysis of the processed multidimensional data in the time dimension. For temperature data, the average temperature, temperature change slope, and difference between the highest and lowest temperatures within the first time scale can be calculated. For gas data, the average gas concentration, cumulative concentration increment, and concentration change rate within the same time scale can be calculated. These parameters are integrated to form a global trend feature. Based on a pre-built fire risk feature library, the extracted global trend feature is compared and matched with the thresholds in the risk feature library to clarify the first-time fire risk trend of the target monitoring point at the first time scale. The fire risk feature library refers to a pre-built database that stores the mapping relationship between typical data features and risk levels corresponding to different types of fires at a specific time scale. This feature database is not fixed and will be adjusted according to the actual conditions of the target mine, such as geological conditions, coal seam type, and mining technology. For example, for mines with high-sulfur, easily spontaneously combustible coal seams, the matching threshold of the spontaneous combustion risk feature parameters will be lowered; for mines with abundant historical fire data, new case features will be continuously added to optimize the mapping relationship, ensuring that it can accurately match the fire risk characteristics of the target mine, providing a reliable basis for matching global trend features, and ultimately ensuring the accuracy of the first monitoring results.

[0055] For example, taking monitoring point No. 2 in the 1102 goaf of a target mine as an example, the first time scale is set to 4 hours. Data is collected from this monitoring point every minute, and the data collected within ten minutes is used as multidimensional data. Assuming the multidimensional data to be input into the first branch is the data from 10:00 to 10:10 of this monitoring point, after inputting it into the first branch, the first branch will extract the historical multidimensional data from 6:10 to 10:00 of this monitoring point. The multidimensional data are then combined in chronological order to generate global trend features. Further, regarding temperature, the average temperature over 4 hours (6:10-10:10) is 25.6℃, the temperature change slope is 0.4℃ / hour, and the difference between the highest and lowest temperatures is 3.2℃; regarding CO concentration, the average concentration is 10.4ppm, the cumulative concentration increase is 4.8ppm, and the concentration change rate is 0.6ppm / hour. Finally, these global trend features can be matched with different types of fires in a fire risk feature database through similarity calculation. For example, these global trend characteristics and the characteristic parameters of low-level spontaneous combustion risk in the database all meet the preset judgment threshold. Therefore, the first monitoring result is "within 4 hours, the target monitoring point has a low-level cumulative spontaneous combustion risk, and the temperature and CO concentration show a stable upward trend."

[0056] As can be seen from the above, the embodiments of this application, through the combined design of adapting the first branch to the long-term scale of cumulative fires, global trend feature extraction, and fire risk feature database matching, can accurately capture the risk of spontaneous combustion fires in mines that require long-term accumulation to manifest, effectively making up for the shortcomings of traditional monitoring methods that fail to detect slowly evolving risks due to focusing on short-term data. On the one hand, the setting of the first time scale fully combines the spontaneous combustion latency period of coal seams in the mine with the rate of data change, ensuring that it can fully cover the time cycle from the initiation to the manifestation of cumulative fires, avoiding incomplete trend extraction due to an excessively short time range. On the other hand, the global trend characteristics extract multi-dimensional long-term change parameters from temperature and gas data, rather than relying on single data or short-term fluctuations, which can more comprehensively reflect the evolution law of cumulative fires. Combined with a fire risk feature library that is personalized based on the actual situation of the mine, it can achieve accurate matching between global trend characteristics and risk levels, so that the output first monitoring results can clearly characterize the cumulative fire risk trend of the target monitoring point, providing a reliable long-term risk basis for subsequent anomaly point judgment and overall risk assessment, further improving the mine's ability to identify highly concealed fires such as spontaneous combustion, and laying the foundation for timely prevention and control measures to prevent the fire from spreading.

[0057] In one embodiment of this application, the second branch is a residual prediction branch; in this embodiment, inputting multidimensional data into the second branch to obtain a second monitoring result includes:

[0058] Input multidimensional data into the residual prediction branch;

[0059] Anomaly features are identified in multidimensional data to obtain abnormal fluctuation features; these abnormal fluctuation features are used to characterize the second time deviation pattern of multidimensional data at the second time scale.

[0060] Fire risk matching is performed based on abnormal fluctuation characteristics to obtain a second monitoring result; the second monitoring result is used to characterize the fire risk anomaly of the target monitoring point at the second time scale; the first time scale is larger than the second time scale.

[0061] In this embodiment, the residual prediction branch, the second branch in the dual-flow monitoring model, is a functional branch specifically designed to extract abnormal fluctuations in multidimensional data over shorter time scales and analyze the risk of sudden open flame fires. Its core function is to identify instantaneous fire risks, such as open flames caused by equipment malfunctions, by capturing short-term deviation patterns in the data, distinguishing it from the first branch which focuses on long-term trends. Anomaly feature identification refers to the feature analysis operation performed by the residual prediction branch on extracting abnormal features from multidimensional data. By comparing the differences between multidimensional data and the normal fluctuation range using a preset algorithm, it identifies features that deviate from the norm. This is a crucial step in capturing sudden fire risks, ensuring that the focus is on subtle changes in the data over a short period, rather than overall trends. Abnormal fluctuation features refer to characteristic parameters identified from multidimensional data that deviate from the normal range of change. They directly reflect the abnormal change patterns of the data in the short-term dimension and are the core basis for judging the risk of sudden open flame fires. The second timescale is the time range set for the residual prediction branch to identify abnormal fluctuation characteristics. This timescale is usually short, measured in minutes, and aims to cover the instantaneous period from the occurrence of a sudden open flame fire to the appearance of data, ensuring that short-term abnormal changes in the data can be captured. Its duration is significantly shorter than the first timescale corresponding to the first branch. The second timescale can be set to 5-20 minutes. The second time deviation pattern refers to the change pattern of multidimensional data within the second timescale that deviates from the normal fluctuation range. This pattern is directly related to the instantaneous occurrence process of a sudden fire. The second time fire risk anomaly refers to the abnormal risk status of the target monitoring point within the second timescale, obtained after residual prediction branch analysis, which is related to the sudden open flame fire. It can clearly reflect the instantaneous changes in the risk of sudden fires and provide a basis for real-time prevention and control.

[0062] In this embodiment, the multidimensional data of the target monitoring point is converted into a format supported by the residual prediction branch and input into the second branch. Abnormal fluctuation feature extraction can employ a residual analysis algorithm, subtracting the mean of the actual data within the multidimensional data from the mean of the historical normal data (which can be normal data consistent with the current mining conditions and equipment operating status of the mine) to obtain the residual value; alternatively, an anomaly detection algorithm can be used, based on the 3σ principle to determine whether the data exceeds the normal fluctuation range. The multidimensional data is analyzed: for temperature data, the maximum temperature increase and instantaneous change rate within the data timescale are calculated; for gas data, the maximum gas concentration increase and mutation rate within the data timescale are calculated. Parameters exceeding the normal range are integrated to form abnormal fluctuation features. Based on a pre-constructed fire risk feature library, the extracted abnormal fluctuation features are compared and matched with thresholds in the risk feature library to identify the second-time fire risk anomaly of the target monitoring point at the second timescale; if there is an abnormal output corresponding to a certain type of fire, it is taken as the second monitoring result; if there are no abnormal fluctuation features, it is a second monitoring result without the risk of a sudden open flame fire.

[0063] For example, taking monitoring point No. 1 of the central substation of a target mine as an example, the second time scale is set to 10 minutes, and the multidimensional data is input into the residual prediction branch. The temperature data is 32℃→37℃, and the CO concentration data is 10ppm→27ppm. Anomaly feature identification is performed on this multidimensional data, calculating that the temperature increases by 5℃ and the CO concentration increases by 17ppm within 10 minutes, forming an abnormal fluctuation feature [temperature increase of 5℃, CO concentration increase of 17ppm]. Finally, these abnormal fluctuation features are matched with different types of fires in the fire risk feature database through similarity calculation. For example, if these abnormal fluctuation features and the feature parameters of low-level open flame risk in the database all meet the preset judgment thresholds, then the second monitoring result is "the second time-related fire risk anomaly is a low-level sudden open flame risk, manifested as a short-term sudden increase in temperature and a sudden increase in CO concentration."

[0064] As can be seen from the above, the embodiments of this application, through the combined design of adapting to the short-term time scale of sudden fires and identifying abnormal fluctuation characteristics, can achieve rapid identification of sudden fire risks. The short-term setting of the second time scale can accurately capture the instantaneous abnormal changes in data when an open flame occurs, avoiding the abnormal features being masked by the overall trend due to an excessively long time scale, thus solving the problem that traditional long-term monitoring is unable to detect short-term sudden risks. The embodiments of this application can improve the timeliness and positioning accuracy of abnormal risks. Through the second time scale, the specific time period of the anomaly can be quickly located, providing accurate guidance for on-site personnel to promptly investigate hidden dangers and reducing investigation time costs. The embodiments of this application can focus on the core characteristics of short-term abnormal fluctuations in sudden open flame fires through residual prediction branches. Combined with a personalized sudden risk feature library, it can effectively distinguish between normal data fluctuations and abnormal fire fluctuations, reduce the misjudgment rate of sudden fire risks, enhance the pertinence of risk identification, further improve the coverage of different types of fires in the mine fire monitoring system, and provide reliable technical support for the real-time prevention and control of sudden fires in mines.

[0065] In one embodiment of this application, before calculating the fire monitoring result of the target monitoring point based on the first monitoring result and the second monitoring result, the method further includes:

[0066] The first weight and the second weight are determined based on the fire risk characteristics of each target monitoring point; the first weight is the weighted weight calculated according to the first monitoring result; the second weight is the weighted weight calculated according to the second monitoring result; the fire risk characteristics include fire type, which includes cumulative spontaneous combustion fire and sudden open flame fire;

[0067] The fire monitoring results of the target monitoring point are calculated based on the first and second monitoring results, including:

[0068] The fire monitoring results of the target monitoring point are obtained by weighting and fusing the first and second monitoring results based on the first and second weights.

[0069] In this embodiment, the first weight is a weighted calculation coefficient assigned to the first monitoring result when fusing the first and second monitoring results. Its value is adapted to the cumulative fire risk characteristics of the area where the target monitoring point is located, and is used to reflect the contribution of the first monitoring result to the final fire monitoring result. The second weight is a weighted calculation coefficient assigned to the second monitoring result when fusing the first and second monitoring results. Its value is adapted to the sudden fire risk characteristics of the area where the target monitoring point is located, and is used to reflect the contribution of the second monitoring result to the final fire monitoring result. Fire risk characteristics refer to the inherent attributes and potential risk tendencies of the area where the target monitoring point is located related to fire, specifically including the types of fires that are prone to occur in the area and the main inducing causes of fires. It is the core basis for determining the first and second weights. Fire types can be cumulative spontaneous combustion or sudden open flame. The inducing causes of different fire types are different. Inducing causes can be coal seam oxidation in goaf areas, short circuits in electrical equipment in electromechanical chambers, frictional heat generation in transport roadways, etc.

[0070] In this embodiment, the dominant fire type in the area where the monitoring point is located is determined in advance based on the actual working conditions of the mine area. For example, in the goaf, the dominant fire type is cumulative spontaneous combustion, while in the electromechanical chamber, the dominant fire type is sudden open flame. The main inducing causes are also identified, for example, the inducing cause in the goaf is residual coal oxidation, and the inducing cause in the electromechanical chamber is cable short circuit. If the area where the target monitoring point is located is dominated by cumulative spontaneous combustion, the first weight is set to 0.7-0.8, and the second weight is set to 0.2-0.3; if the dominant fire type is sudden open flame, the first weight is set to 0.3-0.4, and the second weight is set to 0.6-0.7; if the two types of risks are balanced, such as in the return airway, the weights can be set to 0.5:0.5. The first monitoring results and the second monitoring results are converted into corresponding quantitative values ​​and weighted and fused to determine the final fire monitoring result of the target monitoring point. The fire monitoring result of each target monitoring point includes the risk type and risk level corresponding to the target monitoring point. The risk level can be divided into low risk, medium risk, and high risk based on the actual working conditions and preset threshold range.

[0071] Furthermore, by Figure 2The specific process of obtaining fire monitoring results for a single target monitoring point based on the dual-flow monitoring model in this application is as follows: First, the multi-dimensional time-series data collected from the target monitoring point is synchronously input into two independent branches of the dual-flow monitoring model. The first branch is the trend prediction branch, whose core function is to capture the evolution pattern of potential risks over a long time scale. This branch first extracts global features of the input data in the time dimension based on a preset first time scale. These global trend features are then matched with a pre-constructed fire risk feature library, outputting a first monitoring result characterizing the fire risk trend of the target monitoring point over a long period. At the same time, the second branch, the residual prediction branch, processes the same data input in parallel, and its design goal is to identify short-term sudden anomalies. This branch uses a residual analysis algorithm to identify the deviation between the input data and historical normal benchmarks. These features are then matched with the fire risk feature library to generate a second monitoring result characterizing the fire risk anomalies of the target monitoring point in the short time domain. After the two branches produce intermediate results, the risk quantification values ​​output by the two branches are fused using a weighted calculation formula to obtain the final fire monitoring result for the monitoring point.

[0072] As can be seen from the above, the first branch adapts to the evolutionary pattern of cumulative spontaneous combustion by extracting long-term global features, while the second branch adapts to the explosive characteristics of sudden open flames by identifying short-term anomalies. The combination of the two achieves accurate coverage of different fire types. Furthermore, by designing a risk feature-adapted weight and quantified integration of results, the system improves the scenario adaptability of single-point risk assessment on the one hand. The weight allocation is directly related to the fire risk characteristics of the area where the monitoring point is located, avoiding the one-size-fits-all problem caused by traditional fixed weights, making the final result more in line with the actual risk characteristics of the area. On the other hand, it optimizes the accuracy of risk assessment. The weighted calculation retains the independent judgment of the two types of fire risks, and the weight reflects their importance in a specific area, so that the comprehensive result can balance the contribution of long-term trends and short-term anomalies, reducing misjudgments caused by the over-amplification or neglect of a certain type of risk, and further supporting the upgrade of mine fire monitoring from comprehensive coverage to precise focus.

[0073] In one embodiment of this application, based on the fire monitoring results corresponding to each target monitoring point, an anomaly risk cluster of the target mine is determined, including:

[0074] The status of each target monitoring point is determined based on the fire monitoring results of each target monitoring point and the abnormal status judgment criteria corresponding to each target monitoring point. The abnormal status judgment criteria are different for different target monitoring points. The abnormal status judgment criteria are determined based on the geological conditions, mining technology and historical fire data of the area where the target monitoring point is located. The status includes abnormal status or non-abnormal status.

[0075] The target monitoring points in abnormal states are designated as abnormal monitoring points, and abnormal monitoring points whose distance from each other is less than a preset spatial distance threshold are classified into the same abnormal point risk cluster.

[0076] In this embodiment, the abnormal state judgment criteria are individually set for each target monitoring point to determine whether it is in an abnormal state. These criteria vary depending on the region where the monitoring point is located, and the core basis includes regional geological conditions, mining technology, and historical fire data, directly determining whether a certain risk level is judged as abnormal. An abnormal state refers to the state where the fire monitoring results of a target monitoring point exceed its corresponding abnormal state judgment criteria, indicating that there is a fire risk in the area where the monitoring point is located that requires attention, and forming the basis for subsequent formation of anomaly risk clusters. A non-abnormal state refers to the state where the fire monitoring results of a target monitoring point do not exceed its corresponding abnormal state judgment criteria, indicating that the fire risk in the area where the monitoring point is located is within an acceptable range and does not need to be included in the anomaly risk cluster analysis. The preset spatial distance threshold is a critical value for spatial distance used to divide anomaly risk clusters. It is set based on factors such as mine roadway width, monitoring point layout density, mine working face size, and fire spread rate. When the straight-line distance between two abnormal monitoring points is less than this threshold, they are judged to be spatially adjacent and can be classified into the same risk cluster. Anomaly risk clusters refer to risk groups formed by aggregating anomaly monitoring points in a target mine scenario based on fire monitoring results and specific anomaly judgment criteria of each target monitoring point. These anomaly monitoring points are selected and their spatial distance is less than a preset spatial distance threshold. They also include single-point clusters formed by the absence of neighboring anomaly monitoring points. Their core function is to intuitively present the regional aggregation characteristics and spatial distribution patterns of fire risks underground, providing regionalized risk units for judging the potential spread of fire and formulating targeted prevention and control strategies.

[0077] In this embodiment, the geological conditions of the goaf monitoring point are characterized by easily ignited coal seams and frequent historical spontaneous combustion accidents; the monitoring point of the electromechanical chamber involves a large number of electrical equipment in the mining process and has a history of numerous open flame accidents; the criteria for judging the abnormal state of each monitoring point in different areas are different. The fire monitoring results of each monitoring point are compared with its own standards. If the results exceed the standards, they are marked as abnormal; otherwise, they are not abnormal. The three-dimensional spatial coordinates of all abnormal monitoring points are collected, and the actual distance between any two abnormal monitoring points is calculated using a spatial distance calculation algorithm. Abnormal monitoring points with a distance less than a preset spatial distance threshold are aggregated into the same cluster, while those with a distance greater than the threshold are either grouped separately or not included in the existing clusters, ultimately forming a risk cluster set of abnormal points covering the entire mine.

[0078] As can be seen from the above, the embodiments of this application improve the accuracy of anomaly detection through a combination of differentiated judgment criteria and spatial clustering. These embodiments formulate specific anomaly criteria for monitoring points in different areas, avoiding the omission of high-risk areas or the misjudgment of low-risk areas caused by uniform standards, ensuring that the selection of anomaly monitoring points is more closely aligned with the actual risk level of the region. These embodiments aggregate adjacent anomaly points into clusters by setting a preset spatial distance threshold, overcoming the limitations of isolated analysis of single monitoring points, and can intuitively reflect the regional clustering trend of fire risk, providing a key basis for judging the potential spread range of fire. Furthermore, each risk cluster corresponds to concentrated anomalies in a specific area, facilitating managers to formulate differentiated prevention and control strategies by cluster, avoiding the inefficiency of prevention and control caused by the average allocation of resources, and significantly improving the overall management efficiency of mine fire risk.

[0079] In one embodiment of this application, the fire monitoring results of the target mine are determined based on various anomaly risk clusters, including:

[0080] A risk cluster feature matrix is ​​constructed based on outlier risk clusters; the row vectors of the risk cluster feature matrix correspond to individual outlier risk clusters; the column vectors of the risk cluster feature matrix correspond to the features of the corresponding outlier risk clusters; the risk cluster feature matrix is ​​as follows. The matrix; where m is the total number of anomaly risk clusters within the target mine; n represents the features of different dimensions of the risk clusters;

[0081] Construct a weight matrix, which is a 1×n row matrix; each element of the weight matrix corresponds one-to-one with a feature of the outlier risk cluster.

[0082] The risk value matrix is ​​calculated based on the risk cluster feature matrix and weight matrix; the risk value matrix is ​​shown below:

[0083] V=M×W^T

[0084] Where V is the risk value matrix, and the risk value matrix is ​​m The risk value matrix V is a column matrix of 1. Each element of the risk value matrix V represents the risk value of a single outlier risk cluster. The elements of the risk value matrix correspond one-to-one with the outlier risk clusters. M is the risk cluster feature matrix. W is the weight matrix. W^T is the transpose of the weight matrix W.

[0085] The target mine fire monitoring results are determined based on the risk value matrix.

[0086] In this embodiment, the risk cluster feature matrix is ​​a matrix used to integrate the core features of all anomaly risk clusters within the target mine. Its number of rows equals the total number of anomaly risk clusters (m), and its number of columns (n) corresponds to the pre-selected risk cluster feature dimensions. Each element represents a specific value for a risk cluster under a certain feature dimension. When n=3, the risk cluster feature dimensions are risk importance, the proportion of high-risk monitoring points, and the distance coefficient of the critical area. Risk importance is the sum of the importance weights of all anomaly monitoring points within the anomaly risk cluster, reflecting the strategic impact of the cluster on mine safety. Different monitoring points have varying impacts on the overall mine safety; therefore, importance weights must first be assigned to individual anomaly monitoring points before summing them to obtain the cluster's risk importance. The proportion of high-risk monitoring points is the ratio of the number of high-risk points within the anomaly risk cluster to the total number of anomalies, reflecting the severity of the risk within the cluster. The distance coefficient of the critical area is a standardized value based on the distance transformation from the cluster center to the critical area of ​​the mine, reflecting the threat level of the cluster to critical facilities.

[0087] Furthermore, critical areas in a mine refer to areas that play an irreplaceable supporting role in the overall operation of the mine. These can include the main ventilation room, central substation, and underground refuge chambers, and must be clearly defined in advance by the mine safety department. The distance to a critical area can be calculated using a three-dimensional spatial distance formula, representing the straight-line distance from the center of the risk cluster to the nearest critical area. To eliminate the influence of distance units, a reverse linear standardization method can be used to convert the straight-line distance into a standardized value in the 0-1 interval, which serves as the distance coefficient for the critical area. The reverse linear standardization method is as follows:

[0088]

[0089] Where F is the distance coefficient; dmax is the upper limit of the distance without security threats; dmin is the lower limit of the distance without security threats; and d is the straight-line distance from the center of the risk cluster to the nearest critical area.

[0090] In this embodiment, the weight matrix is ​​a 1×3 row matrix used to characterize the importance of each feature of the risk cluster, and each column element corresponds one-to-one with the column feature of the risk cluster feature matrix.

[0091]

[0092] in, The weights of the risk importance features are used to adjust the contribution ratio of the feature in the risk value calculation; The weight of the high-risk monitoring point proportion feature is used to adjust the contribution ratio of this feature; The weights of the distance coefficient feature for key regions are used to adjust the contribution ratio of this feature. .

[0093] In this embodiment, the risk value matrix is ​​an m×1 column matrix obtained by multiplying the risk cluster feature matrix (M) with the weight matrix transpose (W^T). Each element represents the comprehensive risk quantification value of the corresponding anomaly risk cluster. The value directly reflects the risk level of the risk cluster and is the core quantitative basis for determining the overall fire monitoring results of the mine.

[0094] For example, consider a target mine with three risk clusters of abnormal points, i.e., m=3. Cluster 1 [goaf, containing 2 abnormal monitoring points with weights of 0.6 and 0.5 respectively, 1 high-risk point, distance from cluster center to main ventilation room 80 meters], Cluster 2 [mechanical and electrical chamber, containing 3 abnormal monitoring points, each with a weight of 0.8, 2 high-risk points, distance from cluster center to main ventilation room 40 meters], and Cluster 3 [transport roadway, containing 1 abnormal monitoring point with a weight of 0.5, 0 high-risk points, distance from cluster center to main ventilation room 120 meters]. The upper limit of the safe distance is 200 meters, and the lower limit of the safe distance is 50 meters.

[0095] The first step is to calculate the characteristics of each anomaly risk cluster: Cluster 1 risk importance = 0.6 + 0.5 = 1.1, high-risk monitoring point proportion = 1 / 2 = 0.5, distance coefficient = 0.8; Cluster 2 risk importance = 0.8 × 3 = 2.4, high-risk monitoring point proportion = 2 / 3 ≈ 0.67, distance coefficient = 1; Cluster 3 risk importance = 0.5, high-risk monitoring point proportion = 0, distance coefficient = 0.53. The risk cluster feature matrix M is:

[0096]

[0097] The second step is to establish a weight matrix. The critical areas of the target mine are the lifeline for ensuring normal mine operation and personnel safety. Therefore, the distance coefficient of the critical areas should be given the highest priority to ensure these areas are not affected by the fire. The proportion of high-risk monitoring points directly affects the fire spread rate; if the proportion of high-risk points within a cluster is high, the fire is likely to escalate rapidly, so this should be given a second priority. The importance of the risk level should be the basic priority. Furthermore, the weights are adjusted based on historical data to ensure they align with the actual risk patterns of the mine. The final weight matrix is ​​then determined. .

[0098] Step 3: Calculate the risk value matrix .

[0099] Finally, the magnitude of each risk value in the risk value matrix directly reflects the fire risk level of the corresponding anomaly cluster. Classification criteria can be pre-established based on mine standards; for example, a risk value < 0.5 indicates low risk, 0.5 ≤ risk value < 1.0 indicates medium risk, and a risk value ≥ 1.0 indicates high risk. The final monitoring results are: "Cluster 2's risk value 1.181 ≥ 1.0, belonging to the high-risk level, indicating a high probability of fire occurrence in this cluster; professional personnel should be organized for on-site investigation, and fire extinguishing equipment should be prepared, etc."; "Cluster 1's risk value 0.77 is in the 0.5 ≤ 0.77 < 1.0 range, belonging to the medium-risk level; monitoring and early warning should be strengthened, and fire prevention and control preparations should be made."; "Cluster 3's risk value 0.365 < 0.5, belonging to the low-risk level; routine monitoring should be maintained."

[0100] As can be seen from the above, the embodiments of this application achieve comprehensive risk assessment through multi-feature matrixing and weight adaptation. The risk cluster feature matrix of this application embodiment covers the strategic impact, severity, and threat scope of the risk clusters respectively, avoiding the one-sidedness caused by single-dimensional assessment. The weight matrix can be dynamically adjusted according to the actual needs of the mine, making the risk value calculation more in line with the on-site safety requirements, and facilitating the accurate identification of high-priority risk clusters with significant impact, severe severity, and close threat. Based on the quantitative results of the risk value matrix, a graded disposal strategy can be formulated, while clarifying the key points of prevention and control, avoiding resource waste, and significantly improving the overall management efficiency of mine fire risk and the pertinence of emergency response.

[0101] Corresponding to the mine fire monitoring method in the above embodiments, Figure 3 This is a structural block diagram of a mine fire monitoring system provided in one embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 3 The mine fire monitoring system 20 includes: a data acquisition module 21, a model prediction module 22, a judgment module 23, and a result output module 24.

[0102] Among them, the data acquisition module 21 is used to acquire the multi-dimensional data corresponding to each target monitoring point in the target mine; each target monitoring point is arranged in a different monitoring area in the target mine; each multi-dimensional data is obtained by aligning and stitching the temperature data and gas data of the corresponding target monitoring point according to the time step;

[0103] Model prediction module 22 is used to input various multi-dimensional data into the dual-flow monitoring model and output the fire monitoring results corresponding to each target monitoring point; the dual-flow monitoring model has a dual-branch structure; the dual-branch structure includes a first branch and a second branch; the data processing methods of the first branch and the second branch are different;

[0104] The judgment module 23 is used to determine the abnormal point risk clusters of the target mine based on the fire monitoring results corresponding to each target monitoring point; the abnormal point risk clusters are clusters in the target mine where abnormal monitoring points are spatially concentrated.

[0105] The results output module 24 is used to determine the fire monitoring results of the target mine based on the anomaly risk cluster.

[0106] In one embodiment of this application, when the model prediction module 22 inputs the multi-dimensional data into the dual-flow monitoring model and outputs the fire monitoring result of the target monitoring point corresponding to the multi-dimensional data for each multi-dimensional data, it is specifically used for:

[0107] Input the multidimensional data into the first branch to obtain the first monitoring result;

[0108] Input the multidimensional data into the second branch to obtain the second monitoring result;

[0109] The fire monitoring results of the target monitoring point corresponding to the multidimensional data are calculated based on the first and second monitoring results.

[0110] In one embodiment of this application, when the model prediction module 22 inputs multidimensional data into the second branch to obtain the second monitoring result, it is specifically used for:

[0111] Input multidimensional data into the trend prediction branch;

[0112] Global trend features are obtained by extracting global features over time based on multidimensional data; global trend features are used to characterize the first-time change pattern of multidimensional data at the first time scale.

[0113] Fire risk matching is performed based on global trend characteristics to obtain the first monitoring result; the first monitoring result is used to characterize the first-time fire risk trend of the target monitoring point at the first time scale.

[0114] In one embodiment of this application, when the model prediction module 22 inputs the multidimensional data into the first branch to obtain the first monitoring result, it is specifically used for:

[0115] Input multidimensional data into the residual prediction branch;

[0116] Anomaly features are identified in multidimensional data to obtain abnormal fluctuation features; these abnormal fluctuation features are used to characterize the second time deviation pattern of multidimensional data at the second time scale.

[0117] Fire risk matching is performed based on abnormal fluctuation characteristics to obtain a second monitoring result; the second monitoring result is used to characterize the fire risk anomaly of the target monitoring point at the second time scale; the first time scale is larger than the second time scale.

[0118] In one embodiment of this application, when the model prediction module 22 calculates the fire monitoring result of the target monitoring point based on the first monitoring result and the second monitoring result, it is specifically used for:

[0119] The fire monitoring results of the target monitoring point are obtained by weighting and fusing the first and second monitoring results based on the first and second weights.

[0120] In one embodiment of this application, when determining the abnormal risk cluster of the target mine based on the fire monitoring results corresponding to each target monitoring point, the judgment module 23 is specifically used for:

[0121] The status of each target monitoring point is determined based on the fire monitoring results of each target monitoring point and the abnormal status judgment criteria corresponding to each target monitoring point. The abnormal status judgment criteria are different for different target monitoring points. The abnormal status judgment criteria are determined based on the geological conditions, mining technology and historical fire data of the area where the target monitoring point is located. The status includes abnormal status or non-abnormal status.

[0122] The target monitoring points in abnormal states are designated as abnormal monitoring points, and abnormal monitoring points whose distance from each other is less than a preset spatial distance threshold are classified into the same abnormal point risk cluster.

[0123] In one embodiment of this application, when determining the fire monitoring results of the target mine based on each anomaly risk cluster, the result output module 24 is specifically used for:

[0124] A risk cluster feature matrix is ​​constructed based on outlier risk clusters; the row vectors of the risk cluster feature matrix correspond to individual outlier risk clusters; the column vectors of the risk cluster feature matrix correspond to the features of the corresponding outlier risk clusters; the risk cluster feature matrix is ​​as follows. The matrix; where m is the total number of anomaly risk clusters within the target mine; n represents the features of different dimensions of the risk clusters;

[0125] Construct a weight matrix, which is a 1×n row matrix; each element of the weight matrix corresponds one-to-one with a feature of the outlier risk cluster.

[0126] The risk value matrix is ​​calculated based on the risk cluster feature matrix and weight matrix; the risk value matrix is ​​shown below:

[0127] V=M×W^T

[0128] Where V is the risk value matrix, and the risk value matrix is ​​m The risk value matrix V is a column matrix of 1. Each element of the risk value matrix V represents the risk value of a single outlier risk cluster. The elements of the risk value matrix correspond one-to-one with the outlier risk clusters. M is the risk cluster feature matrix. W is the weight matrix. W^T is the transpose of the weight matrix W.

[0129] The target mine fire monitoring results are determined based on the risk value matrix.

[0130] See Figure 4 , Figure 4 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 4 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of each module / unit in the above-described device embodiments, for example... Figure 3 The functions of the data acquisition module 21, model prediction module 22, judgment module 23, and result output module 24 are shown.

[0131] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0132] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0133] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0134] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the mine fire monitoring method provided in the embodiments of this application, or they can execute the implementation method of the electronic device described in the embodiments of this application, which will not be repeated here.

[0135] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0136] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0137] This application provides a computer program product, which includes computer-executable instructions or a computer program. The computer-executable instructions or computer program are stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the mine fire monitoring method described in this application embodiment.

[0138] Those skilled in the art will recognize that the modules / units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules, units, or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or modules / units, or it may be an electrical, mechanical, or other form of connection.

[0141] The modules / units described as separate components may or may not be physically separate. Similarly, the components shown as modules / units may or may not be physical modules / units; they may be located in one place or distributed across multiple network modules / units. Some or all of the modules / units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0142] Furthermore, the functional modules / units in the various embodiments of this application can be integrated into one processing module / unit, or each module / unit can exist physically separately, or two or more modules / units can be integrated into one module / unit. The integrated modules / units described above can be implemented in hardware or in the form of software functional modules / units.

[0143] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for monitoring mine fires, characterized in that, include: Acquire multidimensional data corresponding to each target monitoring point within the target mine; The target monitoring points are arranged in different monitoring areas within the target mine; Each multidimensional data point is obtained by aligning and stitching the temperature and gas data of the corresponding target monitoring point according to time steps; The multi-dimensional data are input into the dual-flow monitoring model, and the fire monitoring results corresponding to each target monitoring point are output. The dual-flow monitoring model has a dual-branch structure, which includes a first branch and a second branch. The data processing methods of the first branch and the second branch are different. Based on the fire monitoring results corresponding to each target monitoring point, the abnormal point risk cluster of the target mine is determined; The anomaly risk cluster is a cluster in the target mine where anomaly monitoring points are spatially concentrated. The fire monitoring results of the target mine are determined based on the anomaly risk clusters; Wherein, the first branch is the trend prediction branch; the second branch is the residual prediction branch; For each multidimensional data point, the multidimensional data is input into the dual-flow monitoring model, and the fire monitoring results of the target monitoring point corresponding to the multidimensional data are output, including: The multidimensional data is input into the trend prediction branch; global features of the time dimension are extracted based on the multidimensional data to obtain global trend features; the global trend features are used to characterize the first time change pattern of the multidimensional data at the first time scale; fire risk matching is performed based on the global trend features to obtain the first monitoring result; the first monitoring result is used to characterize the first time fire risk trend of the target monitoring point at the first time scale. The multidimensional data is input into the residual prediction branch; abnormal features are identified in the multidimensional data to obtain abnormal fluctuation features; the abnormal fluctuation features are used to characterize the second time deviation pattern of the multidimensional data at the second time scale; fire risk matching is performed based on the abnormal fluctuation features to obtain a second monitoring result; the second monitoring result is used to characterize the second time fire risk anomaly of the target monitoring point at the second time scale; the first time scale is larger than the second time scale; The fire monitoring results of the target monitoring point corresponding to the multidimensional data are calculated based on the first monitoring results and the second monitoring results.

2. The mine fire monitoring method as described in claim 1, characterized in that, Before calculating the fire monitoring result of the target monitoring point based on the first monitoring result and the second monitoring result, the method further includes: A first weight and a second weight are determined based on the fire risk characteristics of each target monitoring point; the first weight is the weighted weight corresponding to the first monitoring result; the second weight is the weighted weight corresponding to the second monitoring result; the fire risk characteristics include fire types, which include cumulative spontaneous combustion fires and sudden open flame fires; The calculation of the fire monitoring results of the target monitoring point based on the first monitoring results and the second monitoring results includes: Based on the first weight and the second weight, the first monitoring result and the second monitoring result are fused by weighted calculation to obtain the fire monitoring result of the target monitoring point.

3. The mine fire monitoring method as described in claim 1, characterized in that, The determination of the abnormal risk clusters of the target mine based on the fire monitoring results corresponding to each target monitoring point includes: The status of each target monitoring point is determined based on the fire monitoring results of each target monitoring point and the abnormal status judgment criteria corresponding to each target monitoring point; wherein, the abnormal status judgment criteria are different for different target monitoring points; the abnormal status judgment criteria are determined based on the geological conditions, mining technology and historical fire data of the area where the target monitoring point is located, and the status includes abnormal status or non-abnormal status; The target monitoring points in abnormal states are designated as abnormal monitoring points, and abnormal monitoring points whose distance from each other is less than a preset spatial distance threshold are classified into the same abnormal point risk cluster.

4. The mine fire monitoring method as described in claim 1, characterized in that, The determination of the fire monitoring results of the target mine based on the anomaly risk cluster includes: A risk cluster feature matrix is ​​constructed based on the anomaly risk clusters; the row vectors of the risk cluster feature matrix correspond to a single anomaly risk cluster; the column vectors of the risk cluster feature matrix correspond to the features of the corresponding anomaly risk cluster; the risk cluster feature matrix is... The matrix; where m is the total number of anomaly risk clusters within the target mine; n represents the features of different dimensions of the risk clusters; Construct a weight matrix, which is a 1×n row matrix; each element of the weight matrix corresponds one-to-one with a feature of the outlier risk cluster. Based on the risk cluster feature matrix and the weight matrix, a risk value matrix is ​​calculated; the risk value matrix is ​​shown below: V=M×W^T Where V is the risk value matrix, and the risk value matrix is ​​m The risk value matrix V is a column matrix of 1, where each element represents the risk value of a single outlier risk cluster, and the elements of the risk value matrix correspond one-to-one with the outlier risk clusters; M is the risk cluster feature matrix; W is the weight matrix; and W^T is the transpose of the weight matrix W. The fire monitoring results of the target mine are determined based on the risk value matrix.

5. A mine fire monitoring system, characterized in that, include: The data acquisition module is used to acquire multidimensional data corresponding to each target monitoring point within the target mine. The target monitoring points are arranged in different monitoring areas within the target mine; Each multidimensional data point is obtained by aligning and stitching the temperature and gas data of the corresponding target monitoring point according to time steps; The model prediction module is used to input various multi-dimensional data into the dual-flow monitoring model and output the fire monitoring results corresponding to each target monitoring point; the dual-flow monitoring model has a dual-branch structure; the dual-branch structure includes a first branch and a second branch; the data processing methods of the first branch and the second branch are different; The judgment module is used to determine the abnormal risk clusters of the target mine based on the fire monitoring results corresponding to each target monitoring point. The anomaly risk cluster is a cluster in the target mine where anomaly monitoring points are spatially concentrated. The result output module is used to determine the fire monitoring results of the target mine based on the anomaly risk cluster; Wherein, the first branch is the trend prediction branch; the second branch is the residual prediction branch; When the model prediction module takes each multidimensional data point, inputs it into the dual-flow monitoring model, and outputs the fire monitoring results for the corresponding target monitoring point, it is specifically used for: The multidimensional data is input into the trend prediction branch; global features of the time dimension are extracted based on the multidimensional data to obtain global trend features; the global trend features are used to characterize the first time change pattern of the multidimensional data at the first time scale; fire risk matching is performed based on the global trend features to obtain the first monitoring result; the first monitoring result is used to characterize the first time fire risk trend of the target monitoring point at the first time scale. The multidimensional data is input into the residual prediction branch; abnormal features are identified in the multidimensional data to obtain abnormal fluctuation features; the abnormal fluctuation features are used to characterize the second time deviation pattern of the multidimensional data at the second time scale; fire risk matching is performed based on the abnormal fluctuation features to obtain a second monitoring result; the second monitoring result is used to characterize the second time fire risk anomaly of the target monitoring point at the second time scale; the first time scale is larger than the second time scale; The fire monitoring results of the target monitoring point corresponding to the multidimensional data are calculated based on the first monitoring results and the second monitoring results.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 4.

Images