Coal mine multi-disaster prediction method, system, medium and equipment
By combining the LSTM-Transformer-GResNet model with multi-source disaster data, the accuracy problem of multi-hazard prediction in coal mines was solved, achieving high-precision prediction and real-time linkage prevention and control of multiple hazards, thus improving the intelligence level of coal mines.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNIV OF MINING & TECH
- Filing Date
- 2026-01-15
- Publication Date
- 2026-04-28
AI Technical Summary
The existing coal mine disaster monitoring system is deployed in a decentralized manner, which cannot achieve data sharing and integrated analysis of multiple disasters and lacks the ability to link and control disasters in real time. This results in insufficient accuracy of multi-hazard prediction and affects the construction of smart mines and the level of safety.
The LSTM-Transformer-GResNet model is combined with multi-source disaster data. Data such as mine water inflow, gas concentration, dust concentration, underground temperature and roof pressure are acquired by sensors to perform multi-hazard prediction. The LSTM layer is used to capture local temporal features, the Transformer is used for location encoding and feature extraction, and GResNet is used for adaptive weighted fusion to achieve multi-hazard prediction.
It achieves high-precision prediction of multiple disasters, supports real-time coordinated prevention and control, provides comprehensive safety decision-making basis, and promotes the intelligent construction of coal mines.
Smart Images
Figure CN121936727A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of mine disaster prevention and control, and in particular to a method, system, medium and equipment for predicting multiple hazards in coal mines. Background Technology
[0002] As my country's shallow coal resources become increasingly depleted, coal mining is shifting towards deeper areas, exacerbating the risks of disasters such as gas, rock bursts, roof falls, water hazards, and heat hazards. Furthermore, dynamic disasters like rock bursts, gas outbursts, and water inrushes are exhibiting coupled and catastrophic characteristics, further increasing the difficulty of advanced prediction, early warning, and prevention of mine disasters. Although my country has achieved significant results in the prevention and control of various coal mine disasters, the overall prevention and control technologies and systems still urgently need improvement.
[0003] Most current coal mine disaster information monitoring sensors are discretely installed and perform single-point measurements. However, disasters such as rock bursts, gas anomalies, and water hazards often have coupled evolution mechanisms. Traditional disaster monitoring systems are mostly deployed in a decentralized manner, meaning that independent monitoring systems are developed for different types of disasters (such as gas outbursts, water hazards, and roof collapses), with each system monitoring only one type of disaster. This decentralized deployment makes it difficult for data to be shared and integrated for analysis, lacks real-time linkage and control capabilities between different types of disasters, and cannot meet the needs of accurate prediction of multiple disasters. In addition, when using deep learning models for prediction, existing technologies typically train deep learning models separately for each type of disaster, with the input being historical data for a single disaster and the output being the future predicted value for that disaster. It is clear that none of the above methods can achieve collaborative prediction of multiple disasters. For example, when gas exceedances are predicted, the roof support equipment cannot be automatically triggered, making it difficult to provide accurate basis for monitoring and early warning, thus affecting the accuracy of multi-source disaster prediction and greatly restricting the construction of smart mines and the improvement of inherent safety levels. Summary of the Invention
[0004] This invention provides a method, system, medium, and equipment for predicting multiple hazards in coal mines, to solve the aforementioned problems existing in the prior art, namely, how to achieve accurate prediction of multiple hazards in coal mines in the prior art. This invention provides a method for predicting multiple hazards in coal mines, which includes: Multiple sensors are deployed in different areas of the test area to acquire multi-source disaster data for different areas. The multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination. Multi-source disaster data is input into a pre-trained LSTM-Transformer-GResNet model to obtain coal mine multi-source disaster prediction results for each time point in the future period. The LSTM-Transformer-GResNet model includes LSTM layers, a location encoding module, a Transformer module, a gated residual connection network GResNet, and fully connected layers. The LSTM layers determine the local temporal features corresponding to each time point in the multi-source disaster data. The location encoding module inputs the local temporal features into the temporal data to obtain the prediction results for the local temporal features. The system performs location encoding to generate location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses local and global temporal features to determine the fused features; and the fully connected layer obtains the prediction results of multiple coal mine disasters at various time points in the future period based on the fused features. The prediction results of multiple coal mine disasters include prediction results for mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
[0005] Optionally, the acquisition of local temporal features corresponding to each time point in the multi-source disaster data specifically includes: The hidden states of multi-source disaster data at adjacent time steps are dynamically filtered using input gates, forget gates, and output gates. h t The computation depends on the hidden state of the previous time step. h t−1 and current input x t ; Retrieve the previous output through the forget gate. h t−1 and current input x t Combining the nonlinear mapping of the Sigmoid function, the following formula is used to obtain local temporal features. f t : ; In the formula, σ is the Sigmoid activation function; W f and b f These represent the weights and biases of the neurons, respectively.
[0006] Optionally, when the predicted results of multiple disasters in the coal mine exceed a preset threshold set based on geological conditions, a disaster warning is issued.
[0007] Optionally, based on the multi-source disaster data, spatial interpolation and structural analysis methods are used to construct a three-dimensional geological model for disaster location and risk area identification.
[0008] Optionally, the multi-source disaster data may be preprocessed, specifically including: The filtering algorithm is used to perform noise reduction, normalization, outlier removal, spatiotemporal synchronization, and data interpolation on multi-source disaster data.
[0009] This invention provides a multi-hazard prediction system for coal mines, comprising: The multi-source sensing module is used to deploy multiple sensors in the area to be measured, and to acquire multi-source disaster data for different areas through the multiple sensors; wherein, the multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure and roof delamination. The prediction module is used to input multi-source disaster data into a pre-trained LSTM-Transformer-GResNet model to obtain the prediction results of coal mine multi-source disasters at various time points within a future period. The LSTM-Transformer-GResNet model includes LSTM layers, a location encoding module, a Transformer module, a gated residual connection network GResNet, and fully connected layers. Specifically, the LSTM layers determine the local temporal features corresponding to each time point in the multi-source disaster data; the location encoding module inputs the temporal data into the local temporal features to predict the local... Temporal features are position-encoded to generate location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses local and global temporal features to determine the fused features; the fully connected layer obtains the prediction results of multiple coal mine disasters at various time points in the future period based on the fused features; wherein, the prediction results of multiple coal mine disasters include the prediction results corresponding to mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
[0010] Also includes: The 3D geological modeling and visualization module is used to construct a 3D geological model for disaster location and risk area identification based on multi-source disaster data and using spatial interpolation and structural analysis methods; the multi-source disaster data includes methane concentration, dust concentration, mine water inflow, and temperature; The intelligent early warning module is used to issue a disaster warning when the predicted results of multiple disasters in the coal mine exceed a preset threshold set based on geological conditions. The preprocessing module is used to preprocess the multi-source disaster data, specifically including: denoising the multi-source disaster data using filtering algorithms, removing outliers from the multi-source disaster data, and performing spatiotemporal synchronization and data interpolation on the multi-source disaster data.
[0011] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting multiple hazards in coal mines.
[0012] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned coal mine multi-hazard prediction method.
[0013] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a method for predicting multiple hazards in coal mines. This method acquires multi-source hazard data from different zones of the test area, such as methane concentration, dust concentration, and temperature. By using an LSTM network to dynamically filter multi-source hazard data at adjacent time steps, it captures the local dependencies of each hazard over time. Then, it calculates the correlation weights of all time steps through the self-attention mechanism of the Transformer, extracting global temporal features between hazards. Based on the local and global temporal features in the acquired multi-source hazard data, it can provide key feature parameters for subsequent multi-hazard prediction. By using a gated residual connection network GResNet to adjust the weights of local and global temporal features and performing multi-scale fusion, it determines the fused features. By inputting the fused features into a fully connected layer, it outputs corresponding data predicting a future time period, such as mine water inflow and gas concentration. This method can comprehensively consider the coupling relationship between multiple hazards, achieve real-time linkage prevention and control, realize high-precision prediction of multiple hazards underground, provide a comprehensive basis for scheduling and safety decisions, and promote the intelligent construction of coal mines. Attached Figure Description
[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0015] Figure 1 A flowchart of a multi-hazard prediction method for coal mines provided in an embodiment of the present invention; Figure 2 A schematic diagram of a coal mine multi-hazard prediction system provided in an embodiment of the present invention; Figure 3 A schematic diagram of a computer device for a coal mine multi-hazard prediction system method provided in an embodiment of the present invention. Detailed Implementation
[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0017] Figure 1 This is a flowchart of a multi-hazard prediction method for coal mines provided in an embodiment of the present invention, such as... Figure 1 As shown in this embodiment, a multi-hazard prediction method for coal mines includes: S1: Deploy multiple sensors in the area to be tested, and acquire multi-source disaster data for different areas through the multiple sensors; wherein, the multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure and roof delamination.
[0018] For example, multiple sensors deployed in different zones acquire data on multiple disaster sources in real time, continuously, and accurately. Its core task is to provide high-quality, multi-source, heterogeneous raw data support for backend disaster identification, trend analysis, and intelligent early warning. It focuses on key areas such as longwall faces, goafs, fault zones, and water hazard channels, employing a layered distributed architecture to ensure the comprehensiveness and accuracy of data perception. The system integrates multiple types of intrinsically safe mining sensors to collect physical and chemical parameters related to disasters such as rock bursts, gas, water hazards, fires, and roof collapses, constructing a multi-source, heterogeneous data acquisition network for multiple disasters.
[0019] Optionally, the multi-source disaster data may be preprocessed, specifically including: The filtering algorithm is used to denoise multi-source disaster data, remove outliers, and perform spatiotemporal synchronization and data interpolation on the multi-source disaster data.
[0020] The first step is data preprocessing and normalization. The Z-Score algorithm is used to calculate the standard deviation of each value from the mean, thereby identifying outliers in the data. The specific calculation method is shown below: In the formula, X is any single value in the time series data, μ is the mean of the overall time series data, and σ is the standard deviation of the overall time series data.
[0021] After outlier removal and missing values filled using linear interpolation, the selected pressure data undergoes linear normalization to reduce error. This linear transformation maps the mine pressure sample data to the [0,1] interval, ensuring that the numerical ranges of different variables are the same, reducing bias and weighting of the data, and guaranteeing the accuracy and stability of the model's predictions. The calculation formula is shown below.
[0022] In the formula, X is the normalized value of the i-th time series data; min The minimum value of the time series data; X max This represents the maximum value of the time-series data.
[0023] S2: Input the multi-source disaster data into the pre-trained LSTM-Transformer-GResNet model to obtain the coal mine multi-source disaster prediction results for each time point in the future period.
[0024] Optionally, the LSTM-Transformer-GResNet model includes an LSTM layer, a location encoding module, a Transformer module, a gated residual connection network GResNet, and a fully connected layer. Specifically, the LSTM layer determines the local temporal features corresponding to each time point in the multi-source disaster data; the location encoding module inputs the local temporal features into the temporal data to perform location encoding, generating location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses the local and global temporal features to determine the fused features; and the fully connected layer, based on the fused features, obtains the prediction results of multi-source disasters in the coal mine corresponding to each time point in the future period. The prediction results of multi-source disasters in the coal mine include prediction results corresponding to mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
[0025] For example, the present invention performs temporal feature analysis on multi-source disaster data through the LSTM layer in the LSTM-Transformer-GResNet model to determine local temporal features; performs location encoding on multi-source disaster data through the Transformer module in the disaster multi-factor coupling relationship model to determine location-aware temporal data; and extracts features from the location-aware temporal data to determine global temporal features.
[0026] For example, preprocessed data can be input into an LSTM layer for local feature extraction. The LSTM dynamically filters information from adjacent time steps through gating mechanisms (input gate, forget gate, output gate), and the calculation of its hidden state ht depends on the hidden state h of the previous time step. t−1 and the current input x t This recursive structure makes it adept at capturing local temporal dependencies. Its main innovation lies in the forget gate, which reads the previous output ht-1 and the current input xt, performs a sigmoid non-linear mapping, and thus outputs a vector. f t Its expression is as follows:
[0027] In the formula, σ is the Sigmoid activation function; W f b f These represent the weights and biases of the neurons.
[0028] The Sigmoid activation function maps the read vector to a value between 0 and 1, where 0 represents forgetting and 1 represents remembering, thus updating or forgetting information. Its mathematical expression is as follows: Where x is the input vector.
[0029] For example, location encoding can be performed on multi-source disaster data. A differentiable location encoding layer encodes the spatiotemporal information of the corresponding time-series data of the input multi-source disaster data, generating a location-aware embedding representation. Then, the location-aware time-series data is input into the Transformer Encoder layer for global feature extraction. The Transformer calculates the association weight between any two time steps in the sequence using a self-attention mechanism, as shown in the following formula:
[0030] The input time-series data consists of Query values, Key values, and Value values. The dot product of the Query and all Key values is calculated, and then the weights of the Values are obtained through the softmax function.
[0031] For example, the present invention uses a gated residual connection network GResNet to adaptively weight and fuse local and global temporal features to determine the fused features.
[0032] For example, a gated residual connection network GResNet is established to fuse extracted local and global temporal features. First, to ensure that the output feature dimensions of both are the same, a residual linear transformation is performed on the output feature matrix of the LSTM while extracting features at the transformer encoding layer, so that it remains consistent with the feature dimension output by the transformer encoding layer, facilitating subsequent feature fusion. Then, a dynamic gating structure is established to adjust the contribution weights of global and local temporal features through dynamic gating, thereby dynamically fusing multi-scale global and local temporal features.
[0033] For example, by selecting the last time step of the fused features, the model's prediction results are mapped through a fully connected layer based on the weight matrix and weight bias of the output layer calculated during the training process. The time-series prediction results are compared with a threshold set based on geological conditions. An alarm is immediately triggered once the prediction results exceed the threshold within a specified time.
[0034] For example, based on the fused features, the present invention obtains the prediction results of multiple sources of coal mine disasters in the future through a fully connected layer.
[0035] For example, this application can construct three-dimensional geological models of coal seams, faults, and surrounding rocks based on various geological data such as borehole data, seismic data, geophysical data, and CAD drawings, using spatial interpolation and structural analysis methods, thus achieving a three-dimensional transformation from two-dimensional maps to real space. Through TIN modeling, fault reconstruction, and coal seam layering algorithms, a realistic underground geological structure is constructed. Combined with real-time data such as tunneling footage and drilling feedback, the model supports dynamic repair and updates, achieving "visible, measurable, and controllable" geological information. The model has three-dimensional rendering and interactive functions, supporting scaling, rotation, sectioning, transparent display, and data overlay. It can be linked with multi-hazard monitoring data for precise spatial hazard location and intuitive risk area identification.
[0036] For example, this invention supports multiple mining data communication methods, including wired and wireless modes, to adapt to different underground environments. To achieve high-precision data calibration, the system is equipped with a GPS clock synchronization module or an IEEE 1588 protocol module to ensure that data from different acquisition points have a unified timestamp. Simultaneously, the acquisition equipment integrates underground positioning technology to achieve precise sensor calibration and geographic binding in three-dimensional space, ensuring the correspondence between data and the three-dimensional geological model. The acquired data is uniformly formatted and transmitted to the geological transparency platform via an industrial communication protocol. The platform can automatically identify the equipment type, data format, and spatial location, and integrate various types of acquired data into the three-dimensional digital geological model. This model supports dynamic overlay, multi-source fusion, and risk identification, supporting the platform's subsequent disaster prediction and early warning.
[0037] The above are one or more embodiments of the coal mine multi-hazard prediction method provided in this specification. Based on the same idea, such as... Figure 2 As shown, this manual also provides a corresponding multi-hazard prediction system for coal mines, including: The multi-source sensing module is used to deploy multiple sensors in the area to be measured, and to acquire multi-source disaster data for different areas through the multiple sensors; wherein, the multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure and roof delamination. The prediction module is used to input multi-source disaster data into a pre-trained LSTM-Transformer-GResNet model to obtain the prediction results of coal mine multi-source disasters at various time points within a future period. The LSTM-Transformer-GResNet model includes LSTM layers, a location encoding module, a Transformer module, a gated residual connection network GResNet, and fully connected layers. Specifically, the LSTM layers determine the local temporal features corresponding to each time point in the multi-source disaster data; the location encoding module inputs the temporal data into the local temporal features to predict the local... Temporal features are position-encoded to generate location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses local and global temporal features to determine the fused features; the fully connected layer obtains the prediction results of multiple coal mine disasters at various time points in the future period based on the fused features; wherein, the prediction results of multiple coal mine disasters include the prediction results corresponding to mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
[0038] Also includes: The 3D geological modeling and visualization module is used to construct a 3D geological model for disaster location and risk area identification based on multi-source disaster data and using spatial interpolation and structural analysis methods; the multi-source disaster data includes methane concentration, dust concentration, mine water inflow, and temperature; The intelligent early warning module is used to issue a disaster warning when the predicted results of multiple disasters in the coal mine exceed a preset threshold set based on geological conditions. The preprocessing module is used to preprocess the multi-source disaster data, specifically including: denoising the multi-source disaster data using filtering algorithms, removing outliers from the multi-source disaster data, and performing spatiotemporal synchronization and data interpolation on the multi-source disaster data.
[0039] Specific limitations regarding the coal mine multi-hazard prediction system can be found in the limitations of the coal mine multi-hazard prediction method described above, and will not be repeated here. Each module in the aforementioned coal mine multi-hazard prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.
[0040] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described coal mine multi-hazard prediction method.
[0041] The present invention also provides Figure 3 The schematic diagram of the computer device shown is as follows: Figure 3 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the coal mine multi-hazard prediction method provided in the above embodiments.
[0042] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.
Claims
1. A method for predicting multiple hazards in coal mines, characterized in that, include: Multiple sensors are deployed in different areas of the test area to acquire multi-source disaster data for different areas. The multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination. Multi-source disaster data is input into a pre-trained LSTM-Transformer-GResNet model to obtain coal mine multi-source disaster prediction results for each time point in the future period. The LSTM-Transformer-GResNet model includes LSTM layers, a location encoding module, a Transformer module, a gated residual connection network GResNet, and fully connected layers. The LSTM layers determine the local temporal features corresponding to each time point in the multi-source disaster data. The location encoding module inputs the local temporal features into the temporal data to obtain the prediction results for the local temporal features. The system performs location encoding to generate location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses local and global temporal features to determine the fused features; and the fully connected layer obtains the prediction results of multiple coal mine disasters at various time points in the future period based on the fused features. The prediction results of multiple coal mine disasters include prediction results for mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
2. The coal mine multi-hazard prediction method as described in claim 1, characterized in that, The acquisition of local temporal features corresponding to each time point in the multi-source disaster data specifically includes: The hidden states of multi-source disaster data at adjacent time steps are dynamically filtered using input gates, forget gates, and output gates. h t The computation depends on the hidden state of the previous time step. h t−1 and current input x t ; Retrieve the previous output through the forget gate. h t−1 and current input x t Combining the nonlinear mapping of the Sigmoid function, the following formula is used to obtain local temporal features. f t : ； In the formula, σ is the Sigmoid activation function; W f and b f These represent the weights and biases of the neurons, respectively.
3. The coal mine multi-hazard prediction method as described in claim 1, characterized in that, When the predicted results of multiple disasters in the coal mine exceed the preset threshold set based on geological conditions, a disaster warning is issued.
4. The coal mine multi-hazard prediction method as described in claim 1, characterized in that, Based on the multi-source disaster data, spatial interpolation and structural analysis methods are used to construct a three-dimensional geological model for disaster location and risk area identification.
5. The coal mine multi-hazard prediction method as described in claim 1, characterized in that, The preprocessing of the multi-source disaster data specifically includes: The filtering algorithm is used to perform noise reduction, normalization, outlier removal, spatiotemporal synchronization, and data interpolation on multi-source disaster data.
6. A multi-hazard prediction system for coal mines, characterized in that, include: The multi-source sensing module is used to deploy multiple sensors in the area to be measured, and to acquire multi-source disaster data for different areas through the multiple sensors; wherein, the multi-source disaster data includes mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure and roof delamination. The prediction module is used to input multi-source disaster data into a pre-trained LSTM-Transformer-GResNet model to obtain the prediction results of coal mine multi-source disasters at various time points within a future period. The LSTM-Transformer-GResNet model includes LSTM layers, a location encoding module, a Transformer module, a gated residual connection network GResNet, and fully connected layers. Specifically, the LSTM layers determine the local temporal features corresponding to each time point in the multi-source disaster data; the location encoding module inputs the temporal data into the local temporal features to predict the local... Temporal features are position-encoded to generate location-aware temporal data; the Transformer extracts features from the location-aware temporal data to determine global temporal features; the gated residual connection network GResNet adaptively weights and fuses local and global temporal features to determine the fused features; the fully connected layer obtains the prediction results of multiple coal mine disasters at various time points in the future period based on the fused features; wherein, the prediction results of multiple coal mine disasters include the prediction results corresponding to mine water inflow, gas concentration, dust concentration, underground temperature, roof pressure, and roof delamination.
7. The coal mine multi-hazard prediction system as described in claim 6, characterized in that, Also includes: The 3D geological modeling and visualization module is used to construct 3D geological models for disaster location and risk area identification based on multi-source disaster data and using spatial interpolation and structural analysis methods.
8. An intelligent early warning module, used to issue a disaster warning when the predicted results of multiple sources of disasters in the coal mine exceed a preset threshold set based on geological conditions; The preprocessing module is used to preprocess the multi-source disaster data, specifically including: The filtering algorithm is used to denoise multi-source disaster data, remove outliers, and perform spatiotemporal synchronization and data interpolation on the multi-source disaster data.
9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the coal mine multi-hazard prediction method according to any one of claims 1-5.
10. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the coal mine multi-hazard prediction method according to any one of claims 1-5.