Mine fire area spontaneous combustion state prediction method and system based on multi-source fusion

Abstract

The application discloses a mine fire area spontaneous combustion state prediction method and system based on multi-source fusion, relates to the technical field of mine spontaneous combustion early warning, and comprises the following steps: acquiring multi-source sensing data and geographic data information of a monitoring area; based on the geographic data information of the monitoring area, performing three-dimensional space division on the monitoring area to obtain a three-dimensional voxel distribution model of the monitoring area. The application combines multi-source sensing data with three-dimensional voxel modeling, considers the influence of multi-factor coupling in the monitoring area on spontaneous combustion risk, can output the posterior probability and comprehensive risk index of the spontaneous combustion state of the fire area voxel by voxel, finally obtains the accurate three-dimensional space spontaneous combustion risk distribution, solves the problem that the traditional monitoring method can only acquire point-line data and is difficult to comprehensively cover the complex spatial distribution of the fire area, and can realize accurate and early prediction of the spontaneous combustion state of the mine fire area.

Description

Technical Field

[0001] This invention relates to the field of mine spontaneous combustion early warning technology, and in particular to a method and system for predicting the spontaneous combustion status of mine fire zones based on multi-source fusion. Background Technology

[0002] Spontaneous combustion in mine fire zones is one of the significant safety hazards faced during mining, storage, disposal, and remediation. In goaf areas, spoil heaps, waste rock piles, gangue mountains, residual ore accumulation areas, and other mining areas containing oxidizable and exothermic media, factors such as ore body occurrence conditions, depositional structure, fracture development, ventilation and oxygen supply, rainfall infiltration, and mining disturbance can cause the internal media to undergo oxidation and exothermic processes, heat storage and temperature rise, covert combustion, or reignition and expansion. This can lead to abnormal surface heat, release of harmful gases, decreased slope stability, ecological damage, and safety risks to personnel and equipment.

[0003] Currently, spontaneous combustion prediction in mine fire zones mainly relies on single types of monitoring data, such as relying solely on temperature indicators, gas indicators, or surface thermal anomalies for early warning. However, due to the complex spatial structure of mine fire zones, heat sources are often concealed, non-uniform, and dynamically migrating. Single data sources are easily affected by environmental interference, resulting in data gaps, large monitoring errors, and insufficient spatial coverage. This makes it difficult to comprehensively reflect the true spontaneous combustion development status of the fire zone, easily leading to false alarms and missed alarms. Consequently, early warnings may be delayed or the deployment of control points may be inaccurate, failing to meet the needs for refined monitoring and early warning in mine fire zones. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies in predicting spontaneous combustion in mine fire zones with low accuracy, and to propose a method and system for predicting spontaneous combustion status in mine fire zones based on multi-source fusion.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, this invention provides a method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion, comprising: Acquire multi-source sensor data and geographic data information of the monitoring area; Based on the geographic data of the monitoring area, the monitoring area is divided into three-dimensional spaces to obtain a three-dimensional voxel distribution model of the monitoring area. Multi-source sensor data is mapped to a three-dimensional voxel distribution model of the monitoring area. Spatiotemporal alignment and missing parts are performed on each voxel in the three-dimensional voxel distribution model. At the same time, feature fusion and addition are performed on each voxel to obtain a three-dimensional voxel feature model of the monitoring area. Based on the three-dimensional voxel feature model of the monitoring area, real-time feature extraction and analysis are performed on each voxel to obtain various potential risk probability information for each voxel. Based on the physical laws of thermal evolution in the monitoring area, the state transition matrix is ​​used to correct the various potential risk probability information of each voxel at adjacent time points, thereby obtaining the various potential risk posterior probability information of each voxel. Based on the three-dimensional voxel feature model of the monitoring area, and combined with the posterior probability information of various potential risks of each voxel, the future time-dependent temperature change of each voxel is predicted, and the comprehensive risk index of each voxel is obtained. Based on the three-dimensional voxel feature model of the monitoring area and the comprehensive risk index of each voxel, multiple voxels are spliced ​​and aggregated with the same risk index to obtain at least one heat source cluster. The situation of each heat source cluster is then perceived to obtain the status and location evolution information of each heat source cluster. Based on the three-dimensional voxel feature model of the monitoring area, each heat source cluster is abstracted and marked as a dynamic entity, and the dynamic entity is tracked and simulated in real time to obtain the risk status evolution trend information of the dynamic entity.

[0006] In one feasible approach, the method for spatiotemporal alignment and missing data for each voxel in the three-dimensional voxel distribution model includes: Based on the geographical data of the monitoring area, the spatial attribute identifier of each voxel is determined. Based on the multi-source sensor data of the monitoring area, the data from different sources are mapped to each voxel separately; Obtain the distribution information of missing voxels in the three-dimensional voxel distribution model, and obtain the adjacent completed mapping voxels of the missing voxels; Based on the spatial attribute identifier of each voxel, and combined with multi-source sensor data of the monitoring area, the relevant element information of the missing voxel and the adjacent completed mapping voxel is determined. Based on the spatial attribute identifier of each voxel and the multi-source sensor data of the monitoring area, and combined with the relevant element information of the missing voxel and the adjacent completed mapping voxels, the missing voxels are completed.

[0007] In one feasible approach, the method for completing missing voxels includes: Equation 1; In Equation 1, For the first Individual elements at time missing test data completion values, The target missing voxel is numbered. The numbering of neighboring voxels, voxels The neighborhood set, As a weight for the quality of neighborhood data, voxels To voxels Normalized spatial distance inverse weights For time continuity weights, To prevent extremely small positive numbers with a denominator of zero.

[0008] In one feasible approach, the method for obtaining various potential risk probability information for each voxel includes: Based on the multi-source sensor data of the monitoring area, various features are extracted for each voxel to obtain the feature distribution information of each voxel. Based on the feature distribution information of each voxel, construct the thermal-air-water-shape coupled feature vector of each voxel; A method is used to attach the corresponding thermal-gas-water-shape coupling feature vector to each voxel to obtain various potential risk probability information for each voxel.

[0009] In one feasible approach, the thermal-gas-water-shape coupling feature vector for each voxel includes: Any one or more of the following characteristics: temperature characteristics, gas oxidation characteristics, rainfall infiltration influence characteristics, oxygen supply capacity characteristics, internal heat source inversion characteristics, and auxiliary characteristics.

[0010] In one feasible approach, the thermal-gas-water-shape coupling feature vector of each voxel specifically includes: Gas oxidation characteristics: using the gas oxidation index The calculation formula is as follows: Equation 2; In Equation 2, The normalized carbon monoxide gas concentration. The normalized methane gas concentration, The normalized ethylene gas concentration, The normalized oxygen consumption concentration. The normalized concentration of sulfur dioxide gas. For normalization function, This represents the amount of oxygen consumed. to These represent the weights of the corresponding gas concentrations; Characteristics of rainfall infiltration impact: using rainfall infiltration correction factors The calculation formula is as follows: Equation 3; In Equation 3, and The first Individual elements at time The corresponding cumulative rainfall over the past 24 hours and the past 72 hours, For humidity, For slope, The degree of cracking, to These are the weights of the corresponding components; Oxygen supply capacity characteristics: using an oxygen supply correction factor The calculation formula is as follows: Equation 4; In Equation 4, For the first Individual elements at time The corresponding combined effects of wind speed and wind direction Porosity The degree of damage to the covering layer, to These are the weights of the corresponding influencing components; Internal heat source inversion characteristics: using heat source terms The calculation formula is as follows: Equation 5; In Equation 5, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume For equivalent volumetric heat capacity, voxels At any moment temperature, The time interval between two adjacent temperature monitoring times. voxels In the previous moment temperature, For the Laplace term of the temperature field, The equivalent thermal conductivity is The equivalent volumetric heat transfer coefficient is... The ambient temperature is the reference temperature for heat exchange between voxels.

[0011] In one feasible approach, the method for obtaining various potential risk probability information for each voxel further includes: Construct a random forest classifier, wherein the random forest is composed of... It consists of decision trees, each tree based on Sample training, node partitioning adopts index; The thermal-gas-water-shape coupled feature vector of each voxel is input into a random forest classifier, and the initial probability of each voxel in the four spontaneous combustion states is output through a voting method: Formula 6; In Equation 6, , Potentially flammable, For internal heating, It is burning. To mitigate the risk of afterburning / reignition, voxels At any moment eigenvectors, For the first The classification output of the decision tree, This is an indicator function.

[0012] In one feasible approach, the method for obtaining the posterior probability information of various potential risks for each voxel includes: Let the constructed state transition matrix be... for A matrix that has matrix elements , Indicates the state from the previous moment. Transition to the current state The probability of; Then the posterior probability information of each voxel for various potential risks is as follows: Formula 7; In Equation 7, voxels At any moment In state The posterior probability, voxels of the previous moment In state The posterior probability, It is a very small positive number.

[0013] In one feasible approach, the method for obtaining the comprehensive risk index for each voxel includes: Based on the thermal-gas-water-shape coupling feature vector of each voxel, the temperature change trend of each voxel at future times is predicted, i.e.: Equation 8; In Equation 8, voxels In the future The predicted temperature Where is the thermal diffusivity, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume and These are the correction factors for rainfall infiltration and oxygen supply, respectively. For ambient temperature, For voxels at time Rainfall infiltration correction factor For voxels at time Oxygen supply correction factor; Calculate the overall risk index for each voxel: Equation 9; In Equation 9, , , Each refers to a voxel At any moment The posterior probability of being in an internal heating state, an ignition state, and an after-ignition / reignition risk state. These are the non-negative weighting coefficients corresponding to each risk assessment indicator. For internal heat source items, For the heating rate, The gas oxidation index, The predicted voxel temperature after 72 hours. The vegetation degradation index, and These are the rainfall infiltration correction factor and the oxygen supply correction factor, respectively.

[0014] In a second aspect, the present invention also provides a mine fire zone spontaneous combustion state prediction system based on multi-source fusion, employing the mine fire zone spontaneous combustion state prediction method based on multi-source fusion described in any one of the first aspects above, wherein the prediction system further includes: The data acquisition module is used to collect multi-source monitoring data of the mine monitoring area; The voxel modeling module is used to map the collected multi-source monitoring data into a three-dimensional voxel grid, complete the missing voxels, and set the corresponding quality weights. The feature extraction module is used to extract the multi-dimensional coupling features of each voxel and calculate the posterior probability of the potential risk of each voxel in various spontaneous combustion states. A risk calculation module is used to calculate the comprehensive risk index of each voxel; A situational simulation module is used to mark heat source clusters and perform dynamic real-time tracking of heat source clusters. The prediction output module is used to output the spatial distribution and prediction results of the spontaneous combustion state of the mine fire zone.

[0015] The beneficial effects of this invention are as follows: This invention combines multi-source sensor data with three-dimensional voxel modeling, taking into account the impact of multiple factors coupled within the monitoring area on spontaneous combustion risk. It can also output the posterior probability and comprehensive risk index of spontaneous combustion state in the fire zone on a voxel-by-voxel basis, ultimately obtaining an accurate three-dimensional spatial distribution of spontaneous combustion risk. This solves the problem that traditional monitoring methods can only obtain point and line data and are difficult to fully cover the complex spatial distribution of the fire zone, enabling accurate early prediction of spontaneous combustion state in mine fire zones. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of the mine fire zone spontaneous combustion state prediction method based on multi-source fusion provided in the embodiments of the present invention; Figure 2 This is a schematic diagram of the three-dimensional voxel distribution model in the mine fire zone spontaneous combustion state prediction method based on multi-source fusion provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the three-dimensional voxel distribution model in the mine fire zone spontaneous combustion state prediction method based on multi-source fusion provided in this embodiment of the invention; Figure 4 This is a schematic diagram illustrating the splicing and aggregation of multiple voxels with the same risk index in the mine fire zone spontaneous combustion state prediction method based on multi-source fusion provided in this embodiment of the invention. Figure 5 This is a flowchart of the random forest classification process in the mine fire zone spontaneous combustion state prediction method based on multi-source fusion provided in this embodiment of the invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0019] Reference Figures 1 to 5 To address the shortcomings of existing technologies in predicting spontaneous combustion in mine fire zones due to low accuracy, this invention provides a multi-source fusion-based method for predicting the spontaneous combustion state of mine fire zones. This method combines multi-source sensor data with three-dimensional voxel modeling, considering the impact of multiple factors coupled within the monitoring area on spontaneous combustion risk. It also outputs the posterior probability and comprehensive risk index of the spontaneous combustion state of the fire zone on a voxel-by-voxel basis, ultimately obtaining a precise three-dimensional spatial distribution of spontaneous combustion risk. This solves the problem that traditional monitoring methods can only acquire point and line data and cannot comprehensively cover the complex spatial distribution of fire zones, enabling accurate early prediction of the spontaneous combustion state of mine fire zones. It should be noted that the monitoring area includes: goaf areas, spoil heaps, waste rock piles, gangue mountains, residual ore accumulation areas, or other mine areas containing oxidizable heat-generating media within the mine fire zone.

[0020] Reference Figure 1 and Figure 5 Specifically, a method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion includes: Step 1: Acquire multi-source sensor data and geographic data information of the monitoring area. The multi-source sensor data includes at least surface thermal infrared temperature, internal fiber optic temperature, borehole temperature, gas concentration, vegetation index, land cover change, rainfall, humidity, wind speed, wind direction, digital elevation model (DEM), fracture distribution data, and pile thickness data. The geographic data information includes at least the distribution of various areas within the monitoring area and their spatial geographic coordinates. This can be achieved by collecting data such as UAV / satellite thermal infrared temperature, distributed fiber optic temperature, borehole temperature, gas concentration, NDVI, land cover change, rainfall, humidity, wind speed, wind direction, DEM, fracture data, and pile thickness data. Each data point, once entered into the system, will be recorded as follows: ; In some feasible implementations, in order to ensure the accuracy of multi-source sensor data, zero-point correction and drift checks can be performed on temperature data; range, abrupt change, and offline detection can be performed on gas sensors; and emissivity correction, ambient temperature correction, and invalid pixel removal can be performed on thermal infrared images.

[0021] Step 2: Based on the geographic data of the monitoring area, the monitoring area can be divided into three-dimensional spaces to obtain a three-dimensional voxel distribution model of the monitoring area; that is, the monitoring area can be divided into multiple voxels in three-dimensional space, and a spatial number, depth attribute and adjacency relationship can be established for each voxel to facilitate rapid location of the early warning position in the future.

[0022] like Figure 2 and Figure 3 As shown, in the specific implementation plan, the boundary of the monitoring area is divided into three-dimensional voxels: ； in, No. A three-dimensional voxel. , , This represents the resolution of the voxel in three directions. (See reference...) Figure 3 As shown, in engineering implementation, a finer resolution can be used in shallow areas, for example... In deep, temperature-sparse regions, the following methods can be used: or Each voxel records elevation, depth, slope, pile thickness, and pore or fracture connectivity indices, and establishes 26 neighborhoods or adjacency relationships based on fracture connectivity.

[0023] Step 3: Map the multi-source sensor data to the three-dimensional voxel distribution model of the monitoring area, perform spatiotemporal alignment and missing filling on each voxel in the three-dimensional voxel distribution model, and simultaneously perform feature fusion and addition on each voxel to obtain the three-dimensional voxel feature model of the monitoring area; wherein, the feature fusion and addition on each voxel includes at least the following features: temperature features, gas oxidation features, rainfall infiltration influence features, oxygen supply capacity features, and internal heat source inversion features. In specific implementation schemes, data from different sources can be mapped to voxels: for example, surface thermal infrared temperature can be mapped to surface voxels; fiber optic temperature measurement data can be mapped to voxels traversed or adjacent to the fiber optic trajectory according to the fiber's spatial path; borehole temperature can be mapped to voxels at the corresponding depth according to the borehole depth; gas data can be interpolated to neighboring voxels based on sensor location, wind direction, fracture connectivity, and distance attenuation; NDVI, bare ground index, overburden damage, and smoke whitening information can be mapped to surface voxels; rainfall, humidity, wind speed, and wind direction can be written as time-dynamic factors into the voxel attributes at the corresponding time. More specifically: The gas space diffusion weight can be calculated using the following formula: ; in, Number the target voxels. The current gas sensor number, This is the index used to iterate through all gas sensors during the summation process in the denominator. For sensors Totozotoxin distance, Indicates fracture connectivity. Indicates the influence of wind direction. , , These are either empirical or training parameters, respectively.

[0024] Regarding voxel missing data completion, the spatial attribute identifier of each voxel can be determined based on the geographical data of the monitoring area. This involves establishing a spatial number, depth attribute, and adjacency relationship for each voxel to facilitate rapid location of the warning position. Then, based on multi-source sensor data from the monitoring area, data from different sources are mapped to each voxel. The distribution information of missing voxels in the 3D voxel distribution model is obtained, along with their adjacent mapped voxels. Missing data completion can be achieved using inverse distance weighted interpolation based on the spatial distance, depth difference, fracture connectivity, and attribute similarity of adjacent voxels. If a missing voxel has no effective adjacent data, it is filled using the statistical average value of the same depth region, while setting the quality weight of that voxel to a lower value to reduce its contribution to subsequent risk calculations.

[0025] In a specific implementation, missing voxels can be filled in based on neighborhoods at the same depth, adjacent time points, and voxels of the same type. The filled-in values ​​are not directly used as high-confidence observations, but rather assigned a lower quality weight. The filling formula is: Equation 1; In Equation 1, For the first Individual elements at time missing test data completion values, The target missing voxel is numbered. The numbering of neighboring voxels, voxels The neighborhood set, As a weight for the quality of neighborhood data, voxels To voxels Normalized spatial distance inverse weights As a weight for time continuity, To prevent extremely small positive numbers with a denominator of zero. For target voxels The most recent valid monitoring time before the current time, and the same type of monitoring value that has passed data quality checks. , , and All are dimensionless quantities.

[0026] Step 4: Based on the three-dimensional voxel feature model of the monitoring area, perform real-time feature extraction and analysis on each voxel to obtain various potential risk probability information for each voxel. The various potential risk probability information for each voxel includes: outputting the initial probability of each voxel in four types of spontaneous combustion states: potential flammability, internal heating, ignition, and post-ignition or reignition risk. Specifically, based on the three-dimensional voxel feature model, extract the thermal-gas-water-shape coupling feature vector for each voxel. The thermal-gas-water-shape coupling feature vector for each voxel includes any one or more combinations of temperature features, gas oxidation features, rainfall infiltration influence features, oxygen supply capacity features, internal heat source inversion features, and auxiliary features.

[0027] Then, random forest classification is used to obtain the likelihood probability of each voxel for each type of spontaneous combustion state. Combined with the state transition prior information, the posterior probability information of each type of potential risk is calculated. Among them, the spontaneous combustion state is divided into four categories: potential flammable state, internal heating state, ignition state, and post-ignition / reignition risk state.

[0028] In practical implementation, methods for extracting the thermal-gas-water-shape coupling feature vector of each voxel include: 1. Extraction of temperature characteristics (including: surface temperature anomalies, internal temperature gradients, and heating rates): Extract surface temperature anomalies and calculate local temperature anomalies for each surface voxel: ; in, The surface temperature is abnormal. voxels At any moment Surface temperature, This represents the local background area. This indicator is used to exclude overall temperature shifts caused by differences in season, sunlight, slope aspect, and material.

[0029] Extracting internal temperature gradient and heating rate The internal temperature vertical gradient is: ; The heating rate over time is: ; like It remains positive over multiple consecutive time periods, and If heat is transferred from the depths to the surface, then the voxel may have an internal heating risk or a hidden heat source.

[0030] 2. Regarding the extraction of gas oxidation characteristics: the gas oxidation index was used. The characterization, and its calculation formula is: Equation 2; In Equation 2, The normalized carbon monoxide gas concentration. The normalized methane gas concentration, The normalized ethylene gas concentration, The normalized oxygen consumption concentration. The normalized concentration of sulfur dioxide gas. For normalization function, This represents the amount of oxygen consumed. to These represent the weights of the corresponding gas concentrations; This represents the oxygen deficit relative to the normal reference oxygen concentration. When the current oxygen concentration is not lower than the reference value, Take 0. to These represent the weights of the corresponding gas characteristic indicators. to The weighting coefficients are dimensionless and non-negative, and the sum of all weighting coefficients is 1. Carbon monoxide is the most typical indicator gas in the spontaneous combustion and exothermic process of coal oxidation in mines. When an increase in carbon monoxide is detected, accompanied by abnormal sulfur dioxide concentration, it can be highly inferred that it is spontaneous combustion of high-sulfur coal. Light hydrocarbon gases such as methane and ethylene will be gradually released as the spontaneous combustion temperature rises. Therefore, by weighting and combining multiple gas components, the oxidation degree of the coal seam within the current voxel range can be more comprehensively reflected, avoiding misjudgments caused by environmental disturbances in single gas detection.

[0031] 3. Regarding the extraction of characteristics related to rainfall infiltration: a rainfall infiltration correction factor was used. The characterization is performed, and its calculation formula is as follows: Equation 3; In Equation 3, to simplify the formula representation, , , and Voxel numbers are omitted. and time stamp , and The first Individual elements at time The corresponding cumulative rainfall over the past 24 hours and the past 72 hours, For humidity, For slope, The degree of cracking, to These are the dimensionless weights of the corresponding components, and the sum of the weight coefficients is 1. That is, rainfall will infiltrate into the interior of the coal seam in the monitoring area through surface fissures. On the one hand, it will have a cooling effect on the coal seam temperature, and on the other hand, it will promote the oxidation reaction of the coal seam. Therefore, this feature is characterized by cumulative rainfall, infiltration depth and current voxel depth. The greater the cumulative rainfall and the shallower the voxel burial depth, the higher the weight of rainfall influence.

[0032] 4. Regarding the extraction of oxygen supply capacity characteristics: an oxygen supply correction factor was used. The characterization, and its calculation formula is: Equation 4; In Equation 4, to simplify the formula representation, and Voxel numbers are omitted. and time stamp , For the first Individual elements at time The corresponding combined effects of wind speed and wind direction Porosity The degree of damage to the covering layer, to These are the dimensionless non-negative weighting coefficients for the corresponding influencing components, and the sum of each weighting coefficient is 1; that is, based on the comprehensive calculation of the fissure connectivity, porosity, slope and surface cover of the voxel's region, the better the fissure connectivity, the higher the porosity, the higher the surface exposure, the stronger the oxygen supply capacity, and the higher the risk of spontaneous combustion.

[0033] 5. Regarding the inversion characteristics of internal heat sources: using heat source terms The characterization, and its calculation formula is: Equation 5; In Equation 5, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume For equivalent volumetric heat capacity, voxels At any moment temperature, The time interval between two adjacent temperature monitoring times. voxels In the previous moment temperature, For the Laplace term of the temperature field, The equivalent thermal conductivity is The equivalent volumetric heat transfer coefficient, This is the ambient temperature at which heat exchange occurs in the voxels. It should be noted that... , and All are the first The local equivalent thermophysical properties of the medium corresponding to each voxel are omitted for the sake of simplifying the formula representation. For voxels with the same medium composition and thermal properties, or the same homogeneous zone, the same parameter values ​​can be used. The equivalent volumetric heat capacity can be calculated based on the medium density and specific heat capacity, or determined by weighting the proportions of each component. The equivalent thermal conductivity can be determined through conventional thermal property testing or an equivalent medium model. The equivalent volumetric heat transfer coefficient can be determined based on… Calibration can be performed by calculation or by using on-site temperature response data.

[0034] Additionally, auxiliary features can be provided, including fracture connectivity, porosity, depth, and overall quality weights. These feature vectors can then be appended to the corresponding voxels to form a three-dimensional voxel feature model.

[0035] Step 5: Refer to Figure 5 As shown, based on the three-dimensional voxel feature model of the monitoring area, real-time feature extraction and analysis are performed on each voxel to obtain various potential risk probability information for each voxel. Specifically, after completing all feature extraction, the above features are combined to obtain the coupled feature vector of the voxel, which is then input into a pre-trained random forest classification model to output the likelihood probability of the voxel under four spontaneous combustion states. By fusing the prior probabilities of state transitions between different times, the posterior probability of each state of each voxel at the current time is calculated.

[0036] In a specific implementation plan, the extracted features can be combined to obtain the coupled feature vector of the voxel, that is, the feature vector of each voxel at time t. The feature vectors are: ; in, The overall quality weight is used to avoid low-quality thermal infrared or offline sensor data dominating model judgments.

[0037] Reference Figure 4 As shown, a random forest is then used to identify spontaneous combustion states. First, four states are defined: , Potentially flammable, For internal heating, It is burning. Risk of afterburning / reignition; A random forest classifier was constructed using samples obtained from on-site drilling, fiber optic temperature verification, gas verification, thermal infrared inspection, and historical management records. The random forest consisted of... It consists of decision trees, each tree based on Sample training, node partitioning adopts index; ; The thermal-gas-water-shape coupled feature vector of each voxel is input into a random forest classifier, and the initial probability of each voxel in the four spontaneous combustion states is output through a voting method: Formula 6; In Equation 6, , Potentially flammable, For internal heating, It is burning. To mitigate the risk of afterburning / reignition, voxels At any moment eigenvectors, For the first The classification output of the decision tree, This is an indicator function. When the condition within the parentheses is true (i.e., the first...), the function is considered an indicator function. The predicted state of the trees is The value is 1 when the decision tree is selected, and 0 otherwise. By using the voting results of multiple decision trees, the classification error caused by sample bias in a single decision tree can be effectively reduced, thus improving the stability and robustness of state recognition.

[0038] Step Six: Based on the physical laws of thermal evolution in the monitoring area, the potential risk probability information of each voxel at adjacent time points is corrected using the state transition matrix to obtain the posterior probability information of various potential risks for each voxel. This allows for the construction of the state transition matrix. for Matrix, example below: ; It has matrix elements , Indicates the state from the previous moment. Transition to the current state The probability; it should be noted that in the state transition matrix The non-zero element setting satisfies the physical laws of thermal evolution in mine fire zones: allowing the gradual evolution of the state from potentially flammable → internal heating → ignition → risk of reignition, and prohibiting or restricting direct jumps in the reverse physical process; Then the posterior probability information of each voxel for various potential risks is as follows: Formula 7; In Equation 7, voxels At any moment In state The posterior probability, voxels of the previous moment In state The posterior probability, The result is a very small positive number. Since the system includes four spontaneous combustion states, the denominator is calculated accordingly. Probability normalization equilibrium is performed. It should be noted that the state transition matrix is ​​not used for probabilistic smoothing, but rather to constrain the physically reachable state paths in the thermal evolution process of the fire zone in the monitoring area, so that the voxel state changes satisfy the continuous evolution law of heat accumulation, oxygen supply, reignition and cooling processes.

[0039] Step 7: Based on the three-dimensional voxel feature model of the monitoring area and combined with the posterior probability information of various potential risks for each voxel, predict the future time-dependent temperature change of each voxel to obtain the comprehensive risk index of each voxel. That is, based on the current posterior probability distribution and actual heat generation rate of each voxel, combined with the heat conduction and convection equation, predict the voxel temperature change of the monitoring area over the future time period to obtain the predicted temperature field distribution.

[0040] In a specific implementation plan, based on the thermal-gas-water-shape coupling feature vector of each voxel, the temperature change trend of each voxel at future times is predicted, i.e.: Equation 8; In Equation 8, voxels In the future The predicted temperature Where is the thermal diffusivity, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume and These are the correction factors for rainfall infiltration and oxygen supply, respectively. For ambient temperature, For voxels at time Rainfall infiltration correction factor For voxels at time The oxygen supply correction factor. To avoid a negative value in the internal heat generation rate after correction, when When the heat source correction coefficient is set to 0, Equation 8 is solved using a time-stepping method. At each prediction time step, based on the predicted temperature of the previous time step and the corresponding internal heat generation rate, rainfall infiltration correction factor, oxygen supply correction factor, and environmental reference temperature, the voxel temperature for the next time step is calculated, and the predicted temperatures for the next 24, 48, and 72 hours are gradually obtained. Relevant parameters within the prediction time interval are determined based on meteorological forecast data, real-time monitoring data, or preset operating conditions; when future data is lacking, the parameter values ​​at the current moment are used for continued calculation. Thus, Equation 8 can be used to quantitatively predict the temperature evolution of each voxel over different time periods in the future. Combined with the posterior probability of the current state and the predicted temperature rise, the comprehensive spontaneous combustion risk index of each voxel is calculated. Specifically, calculate the overall risk index for each voxel: Equation 9; In Equation 9, , , Each refers to a voxel At any moment The posterior probability of being in an internal heating state, an ignition state, and an after-ignition / reignition risk state. These are the non-negative weight coefficients corresponding to each risk assessment indicator, and they satisfy the following conditions: , For internal heat source items, For the heating rate, The gas oxidation index, The predicted voxel temperature after 72 hours. The vegetation degradation index, and These are rainfall infiltration correction factors and oxygen supply correction factors, respectively. Based on this comprehensive risk index, the spontaneous combustion status of mine fire zones can be predicted; the higher the risk index, the greater the likelihood of spontaneous combustion or continued spread of the voxel.

[0041] Step 8: Refer to Figure 4As shown, after obtaining the comprehensive risk index of each voxel in the monitoring area, to facilitate intuitive and clear observation of the fire situation in the monitoring area, multiple voxels can be spliced ​​and aggregated with the same risk index based on the three-dimensional voxel feature model of the monitoring area and the obtained comprehensive risk index of each voxel, to obtain at least one heat source cluster. The situation of each heat source cluster is then assessed to obtain the status and location evolution information of each heat source cluster. The status and location evolution information of each heat source cluster includes: the center position, combustion depth, volume, and diffusion direction of each heat source cluster, and remediation suggestions are generated. In one feasible scheme, to facilitate the splicing and aggregation of multiple voxels with the same risk index, color coding can be used according to the comprehensive risk index of each voxel. For example, each voxel can be color-coded according to the comprehensive risk index, such as outputting graded warnings in blue, yellow, orange, and red.

[0042] Example warning levels are as follows: ; High-risk orange and red voxels can then be aggregated into heat source clusters based on their spatial adjacency. Calculate the center of the heat source cluster Cluster depth, cluster volume, highest predicted temperature, and diffusion direction: ; in, For the first A cluster of heat sources in Risk centroid coordinate vector at any given moment; It describes a set of heat source clusters formed by the aggregation of multiple spatially adjacent high-risk voxels; For the heat source cluster The voxel unit numbers contained within; voxels At any moment Comprehensive risk index voxels The spatial geometric center coordinate vector, which is composed of voxels Three-dimensional coordinate system in the mining geospatial coordinate system The formula, by using the comprehensive risk index of each voxel as the spatial coordinate weight for weighted centroid calculation, achieves the technical effect of shifting the finally determined heat source cluster center towards the high-risk, high-heat-producing core area. This provides a baseline trajectory starting point for the accurate spatial deduction and positioning of dynamic entities in the subsequent time-series knowledge graph. Furthermore, it should be noted that the aforementioned warning level classification thresholds are only a preferred embodiment of this invention in a specific coal mine scenario. In actual mine engineering implementation, the graded warning thresholds are not fixed static empirical values ​​but can be dynamically calibrated and adaptively adjusted based on historical disaster data and on-site management requirements of different mining areas.

[0043] Step Nine: To facilitate rapid location and control of risk areas within the monitoring region, each heat source cluster can be abstracted and labeled as a dynamic entity based on the three-dimensional voxel feature model of the monitoring region. Real-time status tracking and simulation of these dynamic entities can then be performed to obtain information on their risk status evolution trends. Specifically, each aggregated heat source cluster can be abstracted as a dynamic entity in a temporal knowledge graph. The entity is identified and tracked using its name and dynamically drifting three-dimensional coordinates over time. Furthermore, based on the temperature gradient between voxels and the connectivity of fissures, heat conduction relationships and oxygen supply channels between entities are constructed, allowing for the simulation of the entity's spatial displacement trajectory and evolution trend in three-dimensional space.

[0044] The specific steps are as follows: (1) Graph entities ( Abstraction: This involves abstracting each high-risk heat source cluster. Defined as a dynamic entity Its characteristic tuples are expressed as: ； Among them, among them, The name assigned to the entity by the system (e.g., "No. 1 reignition source"); For the entity in The potential three-dimensional coordinates that drift with evolution are the values ​​of the center of the heat source cluster. ; Cluster volume; The highest predicted temperature; This is the current dominant spontaneous combustion state.

[0045] (2) Dynamic relation edge construction: Using the extracted voxel-level coupling feature vectors, establish topological relation edges between entities and between entities and the environment: Characterizing the heat source entity point to The heat conduction propagation trend, its directed topological edge weight The specific calculation formula is as follows: in, To the heat source entity point to The relative weight of the heat spread trend The real-time maximum temperature difference between the two heat source clusters. A heat transfer limiting operator is used to ensure that heat flows unidirectionally only from high-temperature entities to low-temperature entities, completely eliminating the interference of negative weights on the graph inference algorithm. The fracture / pore connectivity coefficient along the connection path between two heat source clusters serves as a dynamic gain adjustment factor for air leakage heat convection, characterizing the mass transfer promoting effect of fracture development on heat spread. Its value range is [value missing]. , The spatial Euclidean distance between the centers of the two heat source clusters. To prevent extremely small positive numbers with a denominator of zero, the unit of measurement is consistent with the spatial distance. The relative weight of the heat spread trend is used to characterize the direction of heat spread and the strength of the correlation between heat source entities, providing a basis for the construction and evolution of the spectral relationship, but is not used to directly calculate the actual heat flow or heat transfer between two heat source entities.

[0046] 3) Evolution and Diffusion Tracking: Over time, the system tracks the potential coordinates of entities. By analyzing the spatial displacement trajectory and combining it with the weight changes of the heat conduction edge and the oxygen supply channel edge, the diffusion direction of the fire zone entity and its merging evolution trend with other heat sources can be deduced.

[0047] In a second aspect, the present invention also provides a mine fire zone spontaneous combustion state prediction system based on multi-source fusion, which adopts the mine fire zone spontaneous combustion state prediction method based on multi-source fusion as described in any one of the first aspects above. The prediction system further includes: a data acquisition module, a voxel modeling module, a feature extraction module, a risk calculation module, a situation inference module, and a prediction output module. The data acquisition module is used to collect multi-source monitoring data of the mine monitoring area, including temperature monitoring data, gas concentration monitoring data, meteorological monitoring data, three-dimensional terrain and fracture distribution data, and remote sensing monitoring data. The voxel modeling module is used to map the collected multi-source monitoring data into a three-dimensional voxel grid, complete the missing voxels, and set corresponding quality weights. The feature extraction module is used to extract the multi-dimensional coupling features of each voxel and calculate the potential risk posterior probability of each voxel for various spontaneous combustion states. The risk calculation module is used to calculate the comprehensive risk index of each voxel by combining the potential risk posterior probability of each voxel for various spontaneous combustion states. The situation inference module is used to mark heat source clusters and perform dynamic real-time tracking of heat source clusters. The prediction output module is used to output the spatial distribution and prediction results of spontaneous combustion states in the mine fire area. In this embodiment, the prediction system accesses multi-source heterogeneous sensing devices and remote sensing image data through a data acquisition module. A voxel modeling module unifies the spatial coordinate system, accurately mapping data from different sources onto a preset three-dimensional voxel grid. Spatial interpolation is used to complete missing voxels, and corresponding quality weights are set for data sources with different precision and reliability to reduce the interference of low-reliability data on the overall prediction results. The feature extraction module then extracts heat-gas-water-shape coupling features from each voxel, performs state probability transitions based on a hidden Markov model, and calculates the posterior probability of potential risk for each voxel in different spontaneous combustion states. Subsequently, a risk calculation module combines multi-dimensional features to calculate the comprehensive spontaneous combustion risk index for each voxel. A situational deduction module then aggregates high-risk voxels, abstracts and marks heat source clusters, and performs real-time tracking and deduction of dynamic entities. Finally, a prediction output module outputs graded early warning results, spatial distribution of fire zones, and evolution trends, achieving accurate prediction of spontaneous combustion states in mine fire zones.

[0048] In some implementations, the prediction system can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0049] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), etc.

[0050] A third aspect of the present invention provides a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method for predicting spontaneous combustion status in mine fire zones based on multi-source fusion as described in any one of the first aspects. The computer-readable medium in this embodiment can be written in one or more programming languages ​​or a combination thereof to perform computer program code for carrying out operations of some embodiments of the present disclosure. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on a user's computer, partially on a user's computer, as a standalone software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0051] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0052] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts.

[0053] A fourth aspect of the present invention provides an electronic device, comprising: one or more processors; a storage device storing one or more programs thereon; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the multi-source fusion-based method for predicting the spontaneous combustion state of a mine fire zone as described in the first aspect. The computer-readable medium may be included in the electronic device or may exist independently, i.e., not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, enable the electronic device to implement the multi-source fusion-based method for predicting the spontaneous combustion state of a mine fire zone as described in the first aspect.

[0054] The fifth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method for predicting spontaneous combustion status of mine fire zones based on multi-source fusion as described in the first aspect.

[0055] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion, characterized in that, include: Acquire multi-source sensor data and geographic data information of the monitoring area; Based on the geographic data of the monitoring area, the monitoring area is divided into three-dimensional spaces to obtain a three-dimensional voxel distribution model of the monitoring area. Multi-source sensor data is mapped to a three-dimensional voxel distribution model of the monitoring area. Spatiotemporal alignment and missing parts are performed on each voxel in the three-dimensional voxel distribution model. At the same time, feature fusion and addition are performed on each voxel to obtain a three-dimensional voxel feature model of the monitoring area. Based on the three-dimensional voxel feature model of the monitoring area, real-time feature extraction and analysis are performed on each voxel to obtain various potential risk probability information for each voxel. Based on the physical laws of thermal evolution in the monitoring area, namely the physical laws of thermal evolution of potential flammability → internal heating → combustion → reignition risk, the state transition matrix is ​​used to correct the probability information of various potential risks of each voxel at adjacent time points, so as to obtain the posterior probability information of various potential risks of each voxel. The method for obtaining the posterior probability information of various potential risks for each voxel includes: Let the constructed state transition matrix be... for A matrix that has matrix elements , Indicates the state from the previous moment. Transition to the current state The probability of; Then the posterior probability information of each voxel for various potential risks is as follows: Formula 7; In Equation 7, voxels At any moment In state The posterior probability, voxels of the previous moment In state The posterior probability, It is a very small positive number; Based on the three-dimensional voxel feature model of the monitoring area, and combined with the posterior probability information of various potential risks of each voxel, the future time-dependent temperature change of each voxel is predicted, and the comprehensive risk index of each voxel is obtained. Based on the three-dimensional voxel feature model of the monitoring area and the comprehensive risk index of each voxel, multiple voxels are spliced ​​and aggregated with the same risk index to obtain at least one heat source cluster. The situation of each heat source cluster is then perceived to obtain the status and location evolution information of each heat source cluster. Based on the three-dimensional voxel feature model of the monitoring area, each heat source cluster is abstracted and marked as a dynamic entity, and the dynamic entity is tracked and simulated in real time to obtain the risk status evolution trend information of the dynamic entity.

2. The method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion according to claim 1, characterized in that, The method for spatiotemporal alignment and missing correction of each voxel in the three-dimensional voxel distribution model includes: Based on the geographical data of the monitoring area, the spatial attribute identifier of each voxel is determined. Based on the multi-source sensor data of the monitoring area, the data from different sources are mapped to each voxel separately; Obtain the distribution information of missing voxels in the three-dimensional voxel distribution model, and obtain the adjacent completed mapping voxels of the missing voxels; Based on the spatial attribute identifier of each voxel, and combined with multi-source sensor data of the monitoring area, the relevant element information of the missing voxel and the adjacent completed mapping voxel is determined. Based on the spatial attribute identifier of each voxel and the multi-source sensor data of the monitoring area, and combined with the relevant element information of the missing voxel and the adjacent completed mapping voxels, the missing voxels are completed.

3. The method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion according to claim 2, characterized in that, The method for completing missing voxels includes: Equation 1; In Equation 1, For the first Individual elements at time missing test data completion values, The target missing voxel is numbered. The numbering of neighboring voxels, voxels The neighborhood set, As a weight for the quality of neighborhood data, voxels To voxels Normalized spatial distance inverse weights As a weight for time continuity, To prevent extremely small positive numbers with a denominator of zero.

4. A method for predicting the spontaneous combustion state of a mine fire zone based on multi-source fusion according to any one of claims 1 to 3, characterized in that, The methods for obtaining various potential risk probability information for each voxel include: Based on the multi-source sensor data of the monitoring area, various features are extracted for each voxel to obtain the feature distribution information of each voxel. Based on the feature distribution information of each voxel, construct the thermal-air-water-shape coupled feature vector of each voxel; A method is used to attach the corresponding thermal-gas-water-shape coupling feature vector to each voxel to obtain various potential risk probability information for each voxel.

5. The method for predicting the spontaneous combustion state of a mine fire zone based on multi-source fusion according to claim 4, characterized in that, The thermal-air-water-shape coupling feature vector of each voxel includes: Any one or more of the following characteristics: temperature characteristics, gas oxidation characteristics, rainfall infiltration influence characteristics, oxygen supply capacity characteristics, internal heat source inversion characteristics, and auxiliary characteristics.

6. The method for predicting the spontaneous combustion state of a mine fire zone based on multi-source fusion according to claim 5, characterized in that, The thermal-air-water-shape coupling feature vector of each voxel specifically includes: Gas oxidation characteristics: using the gas oxidation index The calculation formula is as follows: Equation 2; In Equation 2, The normalized carbon monoxide gas concentration. The normalized methane gas concentration, The normalized ethylene gas concentration, The normalized oxygen consumption concentration. The normalized concentration of sulfur dioxide gas. For normalization function, This represents the amount of oxygen consumed. to These represent the weights of the corresponding gas concentrations; Characteristics of rainfall infiltration impact: using rainfall infiltration correction factor The calculation formula is as follows: Equation 3; In Equation 3, and The first Individual elements at time The corresponding cumulative rainfall over the past 24 hours and the past 72 hours, For humidity, For slope, The degree of cracking, to These are the weights of the corresponding components; Oxygen supply capacity characteristics: using an oxygen supply correction factor The calculation formula is as follows: Equation 4; In Equation 4, For the first Individual elements at time The corresponding combined effects of wind speed and wind direction Porosity The degree of damage to the covering layer, to These are the weights of the corresponding influencing components; Internal heat source inversion characteristics: using heat source terms The calculation formula is as follows: Equation 5; In Equation 5, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume For equivalent volumetric heat capacity, voxels At any moment temperature, The time interval between two adjacent temperature monitoring times. voxels In the previous moment temperature, For the Laplace term of the temperature field, The equivalent thermal conductivity is The equivalent volumetric heat transfer coefficient, The ambient temperature is the reference temperature for heat exchange between voxels.

7. The method for predicting the spontaneous combustion state of a mine fire zone based on multi-source fusion according to claim 6, characterized in that, The method for obtaining various potential risk probability information for each voxel also includes: Construct a random forest classifier, wherein the random forest is composed of... It consists of decision trees, each tree based on Sample training, node partitioning adopts index; The thermal-gas-water-shape coupled feature vector of each voxel is input into a random forest classifier, and the initial probability of each voxel in the four spontaneous combustion states is output through a voting method: Formula 6; In Equation 6, , Potentially flammable, For internal heating, It is burning. To mitigate the risk of afterburning / reignition, voxels At any moment eigenvectors, For the first The classification output of the decision tree, This is an indicator function.

8. The method for predicting the spontaneous combustion state of mine fire zones based on multi-source fusion according to claim 1, characterized in that, The method for obtaining the comprehensive risk index for each voxel includes: Based on the thermal-gas-water-shape coupling feature vector of each voxel, the temperature change trend of each voxel at future times is predicted, i.e.: Equation 8; In Equation 8, voxels In the future The predicted temperature Where is the thermal diffusivity, For the first Individual elements at time The internal heat generation rate per unit volume of the equivalent unit volume and These are the correction factors for rainfall infiltration and oxygen supply, respectively. For ambient temperature, For voxels at time Rainfall infiltration correction factor For voxels at time Oxygen supply correction factor; Calculate the overall risk index for each voxel: Equation 9; In Equation 9, , , Each refers to a voxel At any moment The posterior probability of being in an internal heating state, an ignition state, and an after-ignition / reignition risk state. These are the non-negative weighting coefficients corresponding to each risk assessment indicator. For internal heat source items, For the heating rate, The gas oxidation index, The predicted voxel temperature after 72 hours. The vegetation degradation index, and These are the rainfall infiltration correction factor and the oxygen supply correction factor, respectively.

9. A mine fire zone spontaneous combustion state prediction system based on multi-source fusion, characterized in that, The method for predicting the spontaneous combustion state of a mine fire zone based on multi-source fusion, as described in any one of claims 1 to 8, further includes: The data acquisition module is used to collect multi-source monitoring data of the mine monitoring area; The voxel modeling module is used to map the collected multi-source monitoring data into a three-dimensional voxel grid, complete the missing voxels, and set the corresponding quality weights. The feature extraction module is used to extract the multi-dimensional coupling features of each voxel and calculate the posterior probability of the potential risk of each voxel in various spontaneous combustion states. A risk calculation module is used to calculate the comprehensive risk index of each voxel; A situational simulation module is used to mark heat source clusters and perform dynamic real-time tracking of heat source clusters. The prediction output module is used to output the spatial distribution and prediction results of the spontaneous combustion state of the mine fire zone.

Images