Geological disaster early warning method, system, electronic equipment and storage medium

Abstract

This application relates to the field of geological disaster technology and discloses a geological disaster early warning method, system, electronic device, and storage medium. The method includes: preprocessing collected multi-source data to obtain preprocessed multi-source data; constructing a multi-source dataset, fusing multi-dimensional data features extracted from the multi-source dataset through an attention mechanism, and obtaining a multi-source training dataset; constructing an initial geological disaster early warning model using a multi-layer perceptron, a residual network, and a visual model, and training the initial geological disaster early warning model using the multi-source training dataset to obtain a trained geological disaster early warning model; collecting new multi-source data in real time, and using the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data; and releasing graded early warning information through preset channels when the geological disaster early warning result data meets preset warning conditions. This method improves the accuracy and timeliness of geological disaster early warnings.

Description

Technical Field

[0001] The present application relates to the field of geological disaster technology, and in particular to a geological disaster early warning method, system, electronic equipment and storage medium. Background Art

[0002] In the field of geological disaster early warning, traditional early warning methods mostly rely on a single data source or simple monitoring indicators. For example, in the early stages, they relied on only a few data points such as rainfall and topography, which made it difficult to fully capture the complex mechanisms of geological disaster occurrence. In addition, they mostly used empirical models, which were difficult to adapt to different geological conditions and complex environmental changes. The accuracy and timeliness of early warnings were poor. In the face of sudden disasters, effective early warnings were often not possible, which easily led to casualties and property losses. Geological disaster-related data comes from a wide range of sources, including topography, stratum lithology, meteorology, and other multi-source data. The data formats are diverse and the structures are complex. Traditional geological disaster-related data processing technologies are difficult to efficiently integrate and deeply analyze them, which limits the performance improvement of early warning models. In the past, when constructing geological disaster early warning models, only a single type of feature was often considered, and problems such as gradient vanishing and overfitting were prone to occur during the training process. This resulted in poor model generalization ability, making it difficult to accurately predict geological disasters in different regions and complex situations, affecting the reliability and practicality of the early warning system. In addition, the traditional early warning and emergency response mechanisms are relatively simple, the early warning threshold is fixed and cannot be dynamically adjusted based on real-time data and geological conditions. The early warning results are inaccurate. During the emergency response phase, the allocation of rescue resources mainly relies on manual experience and lacks reasonable planning and optimization, making it difficult to achieve efficient and accurate allocation, which can easily delay the rescue opportunity. Summary of the Invention

[0003] The main technical problem solved by the implementation method of this application is that the accuracy, timeliness and response efficiency of traditional geological disaster warnings need to be improved.

[0004] To solve the above technical problems, the first technical solution adopted in the implementation mode of the present application is: providing a geological disaster early warning method, including: collecting multi-source data of preset data dimensions through a preset data collection path, performing data preprocessing on the collected multi-source data to obtain preprocessed multi-source data; using the preprocessed multi-source data to construct a multi-source data set containing positive sample data and negative sample data, and fusing the multi-dimensional data features extracted from the multi-source data set through an attention mechanism to obtain a multi-source training data set; using a multi-layer perceptron, a residual network and a visual model to construct an initial geological disaster early warning model, training the initial geological disaster early warning model through the multi-source training data set to obtain a trained geological disaster early warning model; collecting new multi-source data in real time through the data collection path, using the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data; when the geological disaster early warning result data meets the preset warning conditions, issuing graded warning information through preset channels, and using the pre-trained rescue resource intelligent allocation model to generate a rescue resource allocation plan based on the geological disaster early warning result data.

[0005] Optionally, the step of collecting multi-source data of preset data dimensions through a preset data collection approach, performing data preprocessing on the collected multi-source data, and obtaining preprocessed multi-source data includes: collecting geological factor-related data of preset data dimensions through a preset data collection approach, and sorting the geological factor-related data in chronological order; calculating the mean and standard deviation of each type of geological factor-related data arranged in chronological order, and removing the geological factor-related data if the geological factor-related data exceeds a first data range, wherein the first data range is composed of the mean and standard deviation; performing missing value processing on the geological factor-related data using a missing value processing model based on autoregressive integrated sliding average, wherein the expression of the missing value processing model is , represents the autoregressive operator, represents the moving average operator, represents the backward shift operator, represents the difference order, represents a white noise sequence, express t The geological factor correlation data at the moment; performing frequency domain analysis on the geological factor correlation data arranged in time series through Fourier transform, extracting the periodic components of the geological factor correlation data, and analyzing the amplitude and phase of different frequency components to obtain the periodic change law of the geological factor correlation data.

[0006] Optionally, the step of using the pre-processed multi-source data to construct a multi-source dataset containing positive sample data and negative sample data, and fusing the multi-dimensional data features extracted from the multi-source dataset through an attention mechanism to obtain a multi-source training dataset includes: executing a preset sliding window sampling strategy, setting a preset number of grid areas as evaluation units, determining positive sample data and negative sample data according to the evaluation units, and adding the positive sample data and the negative sample data to the multi-source dataset; using the multi-source dataset to calculate terrain dimension feature data and landform dimension feature data through a preset digital elevation model; obtaining vegetation association data in the multi-source dataset, and using a preset normalized vegetation index. The model processes the vegetation association data to obtain vegetation coverage characteristic data; processes the stratigraphic lithology data in the multi-source dataset through a pre-trained deep learning model to obtain stratigraphic lithology characteristic vector data, and performs dimensionality reduction processing on the stratigraphic lithology characteristic vector data to obtain stratigraphic lithology characteristic data; rasterizes the fracture data in the multi-source dataset, calculates the intensity value of each grid unit affected by the fracture, and generates fracture characteristic data based on the intensity value; fuses the multi-dimensional data features extracted from the multi-source dataset through an attention mechanism, and maps the multi-dimensional data features to the same dimensional space; calculates the attention weight of each multi-dimensional data feature to obtain a multi-source training dataset.

[0007] Optionally, the steps of constructing an initial geological hazard warning model using a multi-layer perceptron, a residual network and a visual model, training the initial geological hazard warning model through the multi-source training data set, and obtaining a trained geological hazard warning model include: using a multi-layer perceptron, a residual network and a visual model as sub-models to construct an initial geological hazard warning model, and inputting multi-source training data in the multi-source training data set into the initial geological hazard warning model; performing initial feature extraction and nonlinear transformation on the multi-source training data through the multi-layer hidden layers of the multi-layer perceptron to obtain first geological feature vector data; inputting the multi-source training data into the residual block of the residual network, performing data transformation through the residual block, and performing residual linking on the feature vector after data change processing. operation to obtain second geological feature vector data; dividing the multi-source training data into image block vectors, performing self-attention mechanism calculation according to the image block vectors through the self-attention mechanism module of the visual model, and processing the calculation results using a multi-layer encoder to obtain third geological feature vector data; performing feature fusion processing on the first geological feature vector data, the second geological feature vector data and the third geological feature vector data based on the attention mechanism to obtain a geological feature fusion vector; performing prediction calculation according to the geological feature fusion vector, calculating the loss value between the predicted value and the true value through a binary cross entropy loss function, and continuously optimizing the weight parameters and bias parameters of each sub-model according to the loss value until the loss value converges to obtain the trained geological disaster early warning model.

[0008] Optionally, the step of collecting new multi-source data in real time through the data collection path, and using the geological disaster warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster warning result data includes: collecting new multi-source data in real time through the data collection path, and using wavelet transform to remove noise data in the new multi-source data; using a preset deep learning-based anomaly detection model to identify outliers in the new multi-source data, and repairing outliers through a preset machine learning-based interpolation method; sending the new multi-source data after data preprocessing to the geological disaster warning model, and performing real-time analysis and prediction through the geological disaster warning model to obtain geological disaster warning result data; and using the new multi-source data to perform online training and optimization on the geological disaster warning model through a preset stochastic gradient descent algorithm.

[0009] Optionally, before the step of calculating the attention weight of each of the multi-dimensional data features to obtain a multi-source training data set, it also includes: obtaining position information data in the multi-source data set; using the position information data to perform calculations using a preset position coding formula to obtain a position coding feature vector; and performing feature coding fusion processing on the position coding feature vector and the multi-dimensional data features.

[0010] Optionally, before the step of performing prediction calculation based on the geological feature fusion vector and calculating the loss value between the predicted value and the true value through the binary cross entropy loss function, it also includes: first training a preset first number of rounds at a first learning rate during the training process; when the training process of the first number of rounds is completed, continuing training at a second learning rate, wherein, during the training process, a regularization term is added to the loss function through a regularization method, the first learning rate is less than the second learning rate, and the second learning rate decreases according to the preset number of training rounds.

[0011] In order to solve the above technical problems, the second technical solution adopted by the embodiment of the present application is: to provide a geological disaster early warning device, including: an acquisition data preprocessing module, which is used to acquire multi-source data of preset data dimensions through a preset data acquisition path, and perform data preprocessing on the acquired multi-source data to obtain preprocessed multi-source data; a training data set module, which is used to use the preprocessed multi-source data to construct a multi-source data set containing positive sample data and negative sample data, and to fuse the multi-dimensional data features extracted from the multi-source data set through an attention mechanism to obtain a multi-source training data set; a geological disaster early warning model module, which is used to construct a geological disaster early warning model using a multi-layer perceptron, a residual network and a visual model. An initial geological disaster warning model is used to train the initial geological disaster warning model through the multi-source training data set to obtain a trained geological disaster warning model; a geological disaster prediction result module is used to collect new multi-source data in real time through the data collection path, and use the geological disaster warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster warning result data; a geological disaster release and rescue module is used to release graded warning information through preset channels when the geological disaster warning result data meets the preset warning conditions, and use the pre-trained rescue resource intelligent deployment model to generate a rescue resource deployment plan based on the geological disaster warning result data.

[0012] In order to solve the above technical problems, the third technical solution adopted in the implementation of the present application is: to provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the geological disaster warning method as described above.

[0013] In order to solve the above technical problems, the fourth technical solution adopted in the implementation of this application is: providing a non-volatile computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by an electronic device, the electronic device executes the geological disaster warning method as described above.

[0014] Different from the related technologies, this application uses multi-source data collection and preprocessing to comprehensively obtain multi-dimensional data such as topography, vegetation coverage, etc. and ensure data quality, breaking the limitations of traditional single data sources. The early warning system is constructed by integrating multi-layer perceptrons, residual networks and visual models, which greatly improves the accuracy of early warnings compared to traditional single models. New data is collected in real time and dynamically analyzed and predicted to ensure the timeliness of early warnings. When an early warning is triggered, the hierarchical early warning information release mechanism is combined with the pre-trained intelligent deployment model for rescue resources to achieve closed-loop management of the entire process from disaster warning to rescue plan generation, effectively improving the overall efficiency and effectiveness of geological disaster response. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] One or more embodiments are exemplarily illustrated by corresponding drawings, which do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0016] Figure 1 This is a schematic diagram of the operating environment of the geological disaster early warning method provided in the embodiment of the present application.

[0017] Figure 2 This is a schematic diagram of the execution flow of the geological disaster early warning method provided in the embodiment of the present application.

[0018] Figure 3 This is a schematic diagram of the execution flow of obtaining a multi-source training data set in the geological disaster early warning method provided in the embodiment of the present application.

[0019] Figure 4 It is a schematic diagram of the execution process of the geological disaster warning model trained in the geological disaster warning method provided in the embodiment of the present application.

[0020] Figure 5 This is a schematic diagram of the system structure of the geological disaster early warning device provided in the embodiment of the present application.

[0021] Figure 6 This is a schematic diagram of the hardware structure of an electronic device for executing a geological disaster early warning method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0023] It should be noted that, if there is no conflict, the various features in the embodiments of the present application can be combined with each other and are all within the scope of protection of the present application. In addition, although the functional modules are divided in the device schematics and the logical order is shown in the flow charts, in some cases, the steps shown or described can be performed in a different order than the module division in the device schematics or the order in the flow charts.

[0024] Unless otherwise defined, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification are intended only to describe specific embodiments and are not intended to limit this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the relevant listed items.

[0025] To facilitate understanding of this embodiment, first, a geological disaster early warning method disclosed in the embodiment of this application is introduced in detail. Figure 1 , Figure 1 Schematic diagram of the operating environment of the geological disaster early warning method provided in the embodiment of the present application. Figure 1 As shown, the execution subject of the geological disaster early warning method provided in the embodiment of the present application is generally an electronic device with certain computing capabilities, such as a computer device. In some possible implementations, the geological disaster early warning method can be implemented by a processor calling computer-readable instructions stored in a memory. Figure 1 The computer device in the text can be a server. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. It can be understood that Figure 1 The number of computer devices in the figure is only for reference and can be expanded to any number according to actual needs.

[0026] Please continue reading Figure 2 , Figure 2 This is a schematic diagram of the execution flow of the geological disaster early warning method provided in the embodiment of the present application, such as Figure 2 As shown, the following steps are included:

[0027] S1. Collect multi-source data of preset data dimensions through a preset data collection path, and perform data preprocessing on the collected multi-source data to obtain preprocessed multi-source data.

[0028] As an optional implementation, the above step S1 may include the following steps S11 to S14.

[0029] S11. Collect geological factor-related data of preset data dimensions through a preset data collection path, and sort the geological factor-related data in chronological order.

[0030] Step S11 acquires data through pre-defined data collection methods, which can include various sensors (such as displacement sensors and rainfall sensors), satellite remote sensing, geological survey reports, and the like. The collected data is associated with geological factors, covering various aspects such as stratum lithology, geological structure, topography, and hydrogeology, and must meet pre-set data dimension requirements to ensure that the data can provide effective information for subsequent analysis. The collected geological factor-related data is arranged in chronological order. This is done because the occurrence of geological disasters is often related to changes in data in a time series. Arranging in chronological order facilitates subsequent analysis of data evolution trends over time, such as analyzing changes in factors such as stratum pressure and rainfall over different periods, thereby better identifying potential geological disaster risk signals.

[0031] S12. Calculate the mean and standard deviation of each type of geological factor associated data arranged in time series. If the geological factor associated data exceeds a first data range, remove the geological factor associated data. The first data range is composed of the mean and the standard deviation.

[0032] Among them, step S12 is to process the outliers of the geological factor-related data arranged in time series. Specifically, for each type of geological factor-related data (such as displacement data, rainfall data, etc.), calculate its mean and standard deviation. The mean reflects the average level of the data, and the standard deviation reflects the degree of dispersion of the data. Centered on the mean, a data range is determined in combination with the standard deviation. If the data exceeds the range, it is likely to be an outlier, because under normal distribution conditions, most of the data should be concentrated in this range. Removing outliers that are out of range can avoid the interference of abnormal data on subsequent geological disaster analysis and early warning model training, ensure data quality, and make the analysis results more accurate and reliable. For example, for a certain data feature , calculate its mean and standard deviation , will exceed Data points outside the range are considered outliers and removed.

[0033] S13. The missing value processing model based on autoregressive integrated moving average is used to process the missing values of geological factor related data. The expression of the missing value processing model is: , represents the autoregressive operator, represents the moving average operator, represents the backward shift operator, represents the difference order, represents a white noise sequence, express t Geological factor correlation data at the moment.

[0034] Among them, when actually collecting geological data, some data may be missing due to sudden sensor failure, interruption in data transmission, etc. If these missing data are not processed, it will affect the subsequent analysis and judgment of geological disasters and reduce the accuracy of the early warning model. The use of the Autoregressive Integrated Moving Average Model (ARIMA) can make use of the complete historical data that has been collected, combined with the laws and characteristics of the data over time, to reasonably infer and supplement the missing data, so that the entire geological factor-related data becomes complete and continuous, which can provide a data basis for more accurate analysis of geological disasters and the construction of reliable early warning models. For example, by analyzing the displacement data of the mountain in the past period of time, the displacement of the mountain in the future period of time can be predicted to determine whether there is a risk of landslide.

[0035] S14. Perform frequency domain analysis on the geological factor correlation data arranged in time series through Fourier transform, extract the periodic components of the geological factor correlation data, and analyze the amplitude and phase of different frequency components to obtain the periodic variation law of the geological factor correlation data.

[0036] In geological disaster research, many geological factors do not change over time in a simple linear fashion, but may exhibit periodic fluctuations. The Fourier transform can convert geological factor correlation data in the time domain into the frequency domain, decomposing the data into a combination of sine and cosine waves of different frequencies. For example, the Fourier transform can be used to extract the periodic component of the data, using the formula: ,in, is time series data, is its Fourier transform result, This operation extracts the periodic components in the geological factor-related data. The amplitude and phase of the different frequency components are then analyzed. The amplitude reflects the importance of the frequency component in the overall data, while the phase reflects its relative position in time. This analysis can reveal the cyclical patterns in the geological factor-related data. For example, rainfall in certain areas may exhibit seasonal cyclical variations, or formation pressure may fluctuate with a specific period.

[0037] S2. Use the preprocessed multi-source data to construct a multi-source dataset containing positive sample data and negative sample data, and fuse the multi-dimensional data features extracted from the multi-source dataset through the attention mechanism to obtain a multi-source training dataset.

[0038] As an optional implementation, please continue to refer to Figure 3 , Figure 3 Schematic diagram of the execution flow of obtaining a multi-source training data set in the geological disaster early warning method provided in the embodiment of the present application, such as Figure 3 As shown, the process includes the following steps S21 to S27.

[0039] S21. Execute a preset sliding window sampling strategy, set a preset number of grid areas as evaluation units, determine positive sample data and negative sample data according to the evaluation units, and add the positive sample data and negative sample data to the multi-source dataset.

[0040] For example, a sliding window sampling strategy is used to set a 16x16 grid area as an evaluation unit. For geological hazard point data, if the hazard point is located within the window unit, it is determined as a positive sample. For negative samples, they are randomly selected from the entire area, and the distance between the center point of the negative sample and all known hazard points is guaranteed to be greater than two evaluation units, thereby constructing a multi-source dataset.

[0041] S22. Using the multi-source dataset, calculate the terrain dimension feature data and the landform dimension feature data through a preset digital elevation model.

[0042] A Digital Elevation Model (DEM) is a digital representation of the Earth's surface topography. It is constructed by measuring and collecting elevation data from a series of points on the ground, then processing and interpolating the data. The data is typically stored in a regular grid format, with each grid cell corresponding to a specific ground location, and its value representing the elevation at that location. This accurately reflects the undulations of the terrain, including various features such as mountains, valleys, plains, and rivers. By analyzing and processing DEM data, a variety of topographic information can be extracted, such as slope, aspect, elevation difference, and terrain roughness.

[0043] For example, the slope calculation formula is as follows:

[0044] ;

[0045] in, Indicates the elevation value, and is the plane coordinate direction. This formula can accurately quantify the terrain steepness of each evaluation unit, providing key terrain basis for geological hazard risk assessment.

[0046] S23. Obtain vegetation-related data from the multi-source dataset, process the vegetation-related data using a preset normalized vegetation index model, and obtain vegetation coverage characteristic data.

[0047] The Normalized Difference Vegetation Index (NDVI) model calculates vegetation coverage characteristics based on the reflectance characteristics of vegetation in different wavelengths. Vegetation coverage is a key factor influencing the occurrence of geological disasters. Higher vegetation coverage generally reduces soil erosion and lowers the risk of geological disasters. Therefore, this characteristic data is of great significance for geological disaster early warning.

[0048] For example, the vegetation index is calculated using the following normalized vegetation index calculation formula:

[0049] ;

[0050] Among them, NIR represents the reflectivity of the near-infrared band, and RED represents the reflectivity of the red light band.

[0051] S24. Processing the stratigraphic lithology data in the multi-source data set through a pre-trained deep learning model to obtain stratigraphic lithology characteristic vector data, and performing dimensionality reduction processing on the stratigraphic lithology characteristic vector data to obtain stratigraphic lithology characteristic data.

[0052] The pre-trained deep learning model is first used to analyze the lithologic data to obtain lithologic feature vectors. However, the initial vectors contain complex lithologic feature information, and their high dimensionality may be detrimental to subsequent calculations and analysis. Therefore, dimensionality reduction is performed on the lithologic feature vectors to remove redundant information and retain the most critical features. Ultimately, the lithologic feature data is obtained, which can be more effectively used in geological hazard early warning models.

[0053] For example, for the text description information in the stratum lithology data, the pre-trained Bert model is used for vectorization processing. T Input the Bert model fine-tuned with a large amount of geological data to obtain the vector representation of each stratigraphic unit V(T) , whose dimensions are d (d It is determined by the Bert model framework and the characteristics of geological data), that is, At the same time, in order to improve the computational efficiency and model performance, the 768-dimensional vector (if the original Bert model output is 768-dimensional) is subjected to principal component analysis (PCA) dimensionality reduction. Suppose the original vector is V(T) , the low-dimensional vector after PCA transformation is V'(T) , the transformation matrix is U , then: V'(T)=U T V(T), in, U It is the eigenvector matrix obtained by PCA analysis of all formation vectors. In actual implementation, the top three can also be selected based on the principle that the cumulative contribution rate reaches a certain threshold (such as 95%). k principal components ( k <768) to form U, ensuring that key information is retained while reducing the data dimension and reducing the computational burden.

[0054] S25. Rasterize the fracture data in the multi-source dataset, calculate the intensity value of each grid cell affected by the fracture, and generate fracture feature data based on the intensity value.

[0055] For example, an exponential decay function is used to describe the impact of a fracture on the surrounding area, as follows:

[0056] ;

[0057] Where, I(d) Indicates distance break d The impact strength at the fracture is the initial impact strength at the fracture, λ is the attenuation coefficient ( λ It can be determined based on geological conditions and historical disaster data fitting analysis, and its value can be continuously optimized to more accurately reflect the impact of the fault. By processing the fault data with this formula, the intensity value of each grid unit affected by the fault can be obtained as an important feature of the comprehensive data set.

[0058] S26. The multi-dimensional data features extracted from multi-source data sets are fused through the attention mechanism, and the multi-dimensional data features are mapped to the same dimensional space.

[0059] Among them, because data features from different sources may have different dimensions and scales, it is not conducive to direct comprehensive analysis. Therefore, through the attention mechanism, these multi-dimensional data features are mapped to the same dimensional space, making different features comparable, facilitating the subsequent unified processing and analysis of all features, and thus better exploring the relationship between each feature and its comprehensive impact on geological disaster warning.

[0060] For example, suppose the data feature vectors of different modes are F 1 , F 2 ,…, F k , firstly map the features to the same dimensional space through their respective embedding layers, and obtain E 1 , E 2 ,…, E k . Calculate the attention weight matrix A :

[0061] ;

[0062] in, Q and K are the query matrix and key matrix calculated based on the eigenvectors, d k is the dimension of the key matrix. Then, the feature vectors are weighted and summed by the attention weights to obtain the fused feature vector: .

[0063] S27. Calculate the attention weights of each multi-dimensional data feature to obtain a multi-source training dataset.

[0064] Among them, through the above steps S21 to S27, the sliding window sampling strategy is used to determine the positive and negative sample data and expand them to the multi-source data set, which increases the richness and representativeness of the data. Feature data are extracted from multiple dimensions such as topography, vegetation cover, stratum lithology, and fault structure, and various factors affecting geological disasters are comprehensively considered. The digital elevation model is used to calculate the topographic and geomorphological features, the normalized vegetation index model is used to process the vegetation data, the pre-trained deep learning model is used to process the stratum lithology data and reduce the dimension, the fracture data is rasterized and the impact intensity is calculated, and the value of multi-source data is fully and deeply excavated. The multi-dimensional features are mapped to the same space through the attention mechanism, and the attention weight of each feature is calculated, which highlights the role of important features and reduces the interference of secondary features, so that the multi-source training data set can more accurately reflect the relevant information of geological disasters, thereby improving the accuracy and reliability of the geological disaster early warning model.

[0065] As another optional implementation, before calculating the attention weights for each multi-dimensional data feature to obtain a multi-source training dataset, the method may further include obtaining position information data from the multi-source dataset. Then, the position information data is used to perform calculations using a preset position encoding formula to obtain a position encoding feature vector. Preferably, feature encoding fusion processing is performed on the position encoding feature vector and the multi-dimensional data features.

[0066] For example, in a visual model, the input data is divided into multiple image blocks, each image block is regarded as a vector (token), and position encoding is added to retain the position information of the data. The position encoding adopts sine cosine position encoding, and the formula is as follows:

[0067]

[0068] in, POS Indicates location, i is the dimension index, d model is the dimension of the input vector. Through this position encoding method, the model can distinguish features at different positions and improve its understanding of the spatial structure of the data.

[0069] S3. Use multi-layer perceptron, residual network and visual model to build an initial geological disaster warning model, train the initial geological disaster warning model through multi-source training data set, and obtain a trained geological disaster warning model.

[0070] As an optional implementation, please continue to refer to Figure 4 , Figure 4 Schematic diagram of the execution flow of the geological disaster warning model trained in the geological disaster warning method provided in the embodiment of the present application, such as Figure 4 As shown, the process includes the following steps S31 to S36.

[0071] S31. Use a multi-layer perceptron, a residual network, and a visual model as sub-models to construct an initial geological disaster warning model, and input multi-source training data in a multi-source training data set into the initial geological disaster warning model.

[0072] A multi-layer perceptron, residual network, and visual model were selected as sub-models and combined to form a comprehensive model. Multi-source training data from the multi-source training dataset was then fed into this initial model, providing a data foundation for subsequent feature extraction and model training. This model leverages the strengths of the different sub-models to analyze and process data from multiple perspectives, ultimately improving the accuracy of geological hazard warnings.

[0073] S32. Performing initial feature extraction and nonlinear transformation on the multi-source training data through the multi-layer hidden layers of the multi-layer perceptron to obtain first geological feature vector data.

[0074] The multilayer perceptron (MLP) has multiple hidden layers, which perform initial feature extraction and nonlinear transformations on the data. This multi-layer processing can uncover deep-level feature information within the data, converting multi-source training data into primary geological feature vectors. Nonlinear transformations enable the model to learn complex patterns and relationships within the data, providing more valuable feature representations for subsequent analysis and prediction.

[0075] For example, the hidden layer neurons are calculated as:

[0076]

[0077] in, h (i) It is i The output of the hidden layer, W (i) It is i The weight matrix of the layer, b (i) is the bias vector, As the activation function, the activation function can use the Relu function, that is .

[0078] S33, inputting multi-source training data into the residual block of the residual network, performing data transformation through the residual block, and performing residual link operation on the feature vector after data change processing to obtain second geological feature vector data.

[0079] The residual block is the core component of the residual network, enabling effective data transformation. After the data is transformed, the processed feature vector is subjected to a residual link operation, which alleviates the vanishing gradient problem and enables deeper network training, thereby learning more complex features. Ultimately, the second geological feature vector data is obtained, which contains important geological information extracted by the residual network.

[0080] For example, the output formula in the residual block in the residual network is:

[0081] output=x+F(x) ;

[0082] in, x is the input, F(x) It is a transformation or mapping of the input.

[0083] S34. Divide the multi-source training data into image block vectors, perform self-attention mechanism calculation based on the image block vectors through the self-attention mechanism module of the visual model, and process the calculation results using a multi-layer encoder to obtain third geological feature vector data.

[0084] The aforementioned visual model can be a Transformer-based visual model. In a Transformer-based visual model, multi-source training data is divided into image block vectors, each of which contains geological feature information for a specific area. The self-attention mechanism module conducts in-depth analysis of the image block vectors, calculating the degree of correlation between each image block vector and all other image block vectors, i.e., the attention weight. In this way, the model can automatically focus on important parts of the data and ignore irrelevant information. For example, in geological image data, it can accurately identify key feature areas related to geological hazards, such as faults and cracks.

[0085] For example, in order to better capture long-range dependencies and global features in the data, the model incorporates a Transformer-based visual model. The core of the Transformer is the self-attention mechanism, which is calculated as follows:

[0086] ;

[0087] in, Q, K, V represent query vector, key vector and value vector respectively, d k Represents the key vector K dimension.

[0088] S35. Perform feature fusion processing on the first geological feature vector data, the second geological feature vector data, and the third geological feature vector data based on the attention mechanism to obtain a geological feature fusion vector.

[0089] The attention mechanism assigns different weights to each feature vector based on its importance, effectively fusing feature vectors from different sub-models to produce a fusion vector of geological features. This fusion approach fully utilizes the feature information extracted by each sub-model, comprehensively considers multiple factors, and improves the accuracy and reliability of geological hazard warnings.

[0090] For example, suppose the features extracted by the multilayer perceptron are F MLP , the features extracted by the residual network are F ResNet , the features extracted by the visual model are F ViT , first map these features to the same dimensional space through their respective embedding layers, and obtain EMLP 、 E ResNet 、 E ViT . Then calculate the attention weight matrix A :

[0091] ;

[0092] in, Q, K Finally, the fusion feature vector is calculated by weighted summing of the feature vectors using the attention weights to obtain the fusion feature vector F fusion : F fusion =A 1 E MLP +A 2 E ResNet +A 3 E ViT .

[0093] S36. Perform prediction calculations based on the geological feature fusion vector, calculate the loss value between the predicted value and the true value through the binary cross entropy loss function, and continuously optimize the weight parameters and bias parameters of each sub-model based on the loss value until the loss value converges to obtain a trained geological disaster warning model.

[0094] The loss value reflects the degree of deviation between the model's predictions and the actual situation. Based on this loss value, the weight and bias parameters of each sub-model are continuously optimized to bring the model's predictions closer to the true value. The optimization process continues iteratively until the loss value converges, meaning the model's performance reaches a stable state. At this point, the trained geological disaster warning model is obtained.

[0095] For example, the training process uses the error back propagation algorithm to update the model parameters to minimize the difference between the model prediction value and the true label. The binary cross entropy is used as the loss function, and the calculation formula is:

[0096] ;

[0097] ;

[0098] in, y i is the true label, p i is the predicted probability, Nis the number of samples. By continuously adjusting the weights and biases of the model, the value of the loss function is gradually reduced, thereby optimizing the performance of the model.

[0099] As another optional embodiment, before the step of calculating the loss between the predicted value and the true value using the binary cross entropy loss function, the method may further include: first training a preset first number of rounds at a first learning rate. Then, when the training process for the first number of rounds is completed, continuing training at a second learning rate, wherein a regularization term is added to the loss function in a regularization manner during the training process, the first learning rate is less than the second learning rate, and the second learning rate decreases according to the preset number of training rounds.

[0100] For example, during the training process, the learning rate is adjusted using a warm-up strategy. The model is first optimized for several rounds (such as 6 rounds) with a lower first learning rate (such as 0.001). After the model is relatively stable, it is trained using a pre-set second learning rate (such as 0.01). In addition, during the training process, the second learning rate is decayed by a certain multiple (such as 0.1 times) according to certain rounds (such as the 12th round and the 18th round) to promote faster and more stable model convergence. To prevent overfitting of the model, the L2 regularization technology is used to add a regularization term to the loss function: .in, λ is the regularization coefficient, W is the set of all parameters of the model.

[0101] S4. New multi-source data are collected in real time through data collection channels, and the geological disaster early warning model is used to analyze and predict the new multi-source data in real time to obtain geological disaster early warning result data.

[0102] As an optional implementation, the process of obtaining geological disaster warning result data in the above step S4 may include the following steps S41 to S44.

[0103] S41. Collect new multi-source data in real time through a data collection method, and use a wavelet transform method to remove noise data in the new multi-source data.

[0104] For example, wavelet transform technology is used to remove noise in data. The wavelet transform formula is:

[0105]

[0106] in, f(t) is the original data signal, ψ is the wavelet basis function, a is the scale parameter, b is the translation parameter.

[0107] S42. Use a preset deep learning-based anomaly detection model to identify outliers in new multi-source data, and repair outliers using a preset machine learning-based interpolation method.

[0108] Deep learning models possess powerful feature learning capabilities, enabling them to identify data points that do not conform to normal patterns, known as outliers, from complex data. Outliers can be caused by sensor failure, data transmission errors, or sudden changes in geological conditions. After identifying outliers, they are corrected using a pre-defined machine learning-based interpolation method. This method can reasonably estimate and replace outliers based on data distribution patterns and information about surrounding normal data points. This ensures data integrity and continuity, preventing outliers from negatively impacting subsequent geological hazard early warning model analysis and predictions. This allows the model to operate based on more accurate data, improving the reliability of early warnings.

[0109] For example, construct an autoencoder network whose input is monitoring data x , and obtain low-dimensional representation through the encoding layer , and then reconstruct the data through the decoding layer . The autoencoder is trained to minimize its reconstruction error, i.e. ( n is the data dimension). In the actual implementation process, the reconstruction error is calculated. If the reconstruction error exceeds the preset threshold, the data point is determined to be an outlier. For outliers, an interpolation method based on machine learning is used to repair them, such as interpolation based on the K-nearest neighbor (KNN) algorithm. For missing values or outlier points, x , whose repair value is :

[0110]

[0111] in, x i yes x of K nearest neighbor data points.

[0112] S43. Send the new multi-source data after data preprocessing to the geological disaster early warning model, and perform real-time analysis and prediction through the geological disaster early warning model to obtain geological disaster early warning result data.

[0113] Among them, after the above step S43, it is also possible to combine the threshold-based warning rules. When the predicted probability of a disaster exceeds the preset threshold, or certain key monitoring indicators (such as the displacement rate exceeds the warning value, the effective rainfall reaches a specific threshold, etc.) trigger the warning conditions, the warning process is immediately started, and the warning information is issued to relevant departments and personnel through various channels (such as SMS, APP push, broadcast, etc.).

[0114] S44. Use the preset stochastic gradient descent algorithm to perform online training and optimization on the geological disaster warning model using new multi-source data.

[0115] In order to enable the geological disaster early warning model to adapt to the ever-changing geological environment and emerging data patterns, step S44 adopts a preset stochastic gradient descent algorithm to train and optimize the model online using new multi-source data. The stochastic gradient descent algorithm calculates the gradient by randomly selecting a portion of the data in each iteration, thereby updating the model parameters. By continuously training with new data, the model can learn new features and patterns and adjust its own parameters to improve its ability to predict geological disasters. The online training and optimization method can enable the model to maintain good performance. As time goes by and new data accumulates, the accuracy and timeliness of geological disaster early warnings are continuously improved, and it can better cope with complex and changing geological disaster situations.

[0116] For example, using a variant algorithm of stochastic gradient descent, such as the Adadelta algorithm, its parameter update formula is:

[0117] ;

[0118] ;

[0119] in, It is t The model parameters at the iteration, gt It is t The gradient at iteration , and are the accumulation of squares of past parameter updates and the accumulation of squares of current gradients, ε is a small constant that prevents the denominator from being zero.

[0120] S5. When the geological disaster warning result data reaches the preset warning conditions, the graded warning information is released through the preset channels, and the pre-trained rescue resource intelligent allocation model is used to generate a rescue resource allocation plan based on the geological disaster warning result data.

[0121] As an example, in the graded warning based on dynamic thresholds, combined with real-time monitored rainfall data, the improved effective rainfall calculation formula is used: (in, R c is the effective rainfall, R 0 The rainfall for the day, R i For the past n The rainfall on the day of the day, αis the rainfall coefficient determined according to local geological and meteorological conditions, γ is the rainfall change rate influence coefficient, is the time interval, is the rainfall change rate), calculate the effective rainfall. Based on the effective rainfall, the geological disaster susceptibility level (refer to the previous geological disaster susceptibility assessment results, divided into three levels: low, medium and high) and real-time geological monitoring data (such as displacement rate, ground stress change, etc.), the warning level is dynamically delineated (such as blue, yellow, orange and red four-level warning). A blue warning means there is a certain risk and attention is required. A yellow warning indicates an increased risk and precautionary preparations must be made. An orange warning means that the possibility of a disaster is high and some emergency response measures are activated. A red warning indicates that a disaster is imminent and a full emergency response is activated. The frequency and scope of information release are adjusted for different warning levels. When a red warning is issued, it is released frequently through all channels to ensure that information coverage is comprehensive.

[0122] As another example, in efficient and intelligent emergency response, a rescue resource intelligent deployment model is constructed based on big data analysis and operations research principles. m A relief material storage point, n disaster-affected areas, x ij Indicates that from i Reserve points are deployed to j The amount of supplies in the affected areas c ij Indicates that from i Reserve points are deployed to j The comprehensive transportation cost of the affected area (taking into account the transportation distance d ij , road conditions r ij , type of transportation t ij , urgency e ij Factors such as cost function c ij =d ij × r ij × t ij × e ij calculate), s i For the i The amount of material reserves at each storage point, d j For the j The material demand in the disaster-stricken areas. A mixed integer programming model is established, and the objective function is represents minimizing the total transportation cost. The constraints include: (Ensure that the materials transferred out from each storage point do not exceed the reserve quantity). (Meet the basic material needs of each disaster-stricken area). x ij ≥ 0 and part x ij The number of materials to be allocated is an integer (whether it is an integer depends on the actual situation of material allocation, such as the allocation quantity of large equipment). This model, combined with intelligent algorithms (such as genetic algorithms and particle swarm optimization algorithms), determines the optimal material allocation plan to achieve rapid and efficient distribution of relief supplies. At the same time, considering the timeliness and importance of the materials, priority is given to the allocation of urgently needed and critical materials, such as medical supplies, food, and emergency rescue equipment. During the emergency response process, on-site feedback data is continuously collected, such as the progress of the rescue team, emerging geological hazards in the affected area, casualties, and changes in material demand. This data is fed back to the early warning and emergency decision-making model in real time, allowing the model to be dynamically adjusted.

[0123] The geological disaster early warning method provided in the embodiments of the present application effectively addresses the problem of traditional geological disaster early warning accuracy, timeliness, and response efficiency that need to be improved through a series of technical means such as multi-dimensional data collection and preprocessing, multi-source training data set construction, multi-model fusion training, and real-time dynamic analysis and early warning. In terms of data processing, wavelet transform is used to remove noise, and deep learning models are used to identify and repair outliers, ensuring data reliability and avoiding early warning bias caused by data quality issues. When constructing the multi-source training data set, multi-dimensional features such as topography and vegetation cover are comprehensively extracted, and feature fusion is performed using an attention mechanism, significantly improving the model's ability to analyze complex geological conditions. In the model construction phase, a multi-layer perceptron, residual network, and visual model are combined to fully utilize the advantages of each model and further improve the accuracy of early warning. In practical applications, timely early warning of geological disasters is achieved by collecting new data in real time and rapidly processing and analyzing it, combined with threshold-based hierarchical early warning rules. At the same time, the model is trained and optimized online using a stochastic gradient descent algorithm, enabling it to adapt to the ever-changing geological environment and maintain high early warning accuracy. In addition, the intelligent allocation model of rescue resources built based on big data and operations research principles can quickly generate rescue plans based on early warning results, greatly improving the efficiency of emergency response and forming a complete closed loop from data collection and processing to early warning response. It comprehensively improves the accuracy, timeliness and response efficiency of geological disaster early warnings, and provides technical support for the effective prevention and control of geological disasters.

[0124] Please continue reading Figure 5 , Figure 5 This is a schematic diagram of the system structure of the geological disaster early warning device provided in the embodiment of the present application. Figure 5As shown, the geological disaster early warning device 50 includes: a data collection preprocessing module 51, a training data set module 52, a geological disaster early warning model module 53, a geological disaster prediction result module 54 and a geological disaster release and rescue module 55.

[0125] The collected data preprocessing module 51 is specifically used to collect multi-source data of preset data dimensions through a preset data collection path, and perform data preprocessing on the collected multi-source data to obtain preprocessed multi-source data.

[0126] The training dataset module 52 is specifically used to use the preprocessed multi-source data to construct a multi-source dataset containing positive sample data and negative sample data, and to fuse the multi-dimensional data features extracted from the multi-source dataset through the attention mechanism to obtain a multi-source training dataset.

[0127] The geological disaster warning model module 53 is specifically used to use a multi-layer perceptron, a residual network and a visual model to construct an initial geological disaster warning model, and train the initial geological disaster warning model through the multi-source training data set to obtain a trained geological disaster warning model.

[0128] The geological disaster prediction result module 54 is specifically used to collect the new multi-source data in real time through the data collection path, and use the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data.

[0129] The geological disaster release and rescue module 55 is specifically used to release graded warning information through preset channels when the geological disaster warning result data reaches the preset warning conditions, and use the pre-trained rescue resource intelligent deployment model to generate a rescue resource deployment plan based on the geological disaster warning result data.

[0130] As an optional implementation, the data acquisition preprocessing module 51 is further specifically used to acquire geological factor-related data of preset data dimensions through a preset data acquisition path, and sort the geological factor-related data in chronological order; calculate the mean and standard deviation of each type of geological factor-related data arranged in chronological order, and remove the geological factor-related data if the geological factor-related data exceeds a first data range, wherein the first data range is composed of the mean and standard deviation; and perform missing value processing on the geological factor-related data using a missing value processing model based on autoregressive integrated sliding average, wherein the expression of the missing value processing model is: , represents the autoregressive operator, represents the moving average operator, represents the backward shift operator, represents the difference order, represents a white noise sequence, express t The geological factor correlation data at the moment; performing frequency domain analysis on the geological factor correlation data arranged in time series through Fourier transform, extracting the periodic components of the geological factor correlation data, and analyzing the amplitude and phase of different frequency components to obtain the periodic change law of the geological factor correlation data.

[0131] As an optional implementation, the training data set module 52 is further specifically used to execute a preset sliding window sampling strategy, set a preset number of grid areas as evaluation units, determine positive sample data and negative sample data according to the evaluation units, and add the positive sample data and the negative sample data to the multi-source data set; use the multi-source data set to calculate terrain dimension feature data and landform dimension feature data through a preset digital elevation model; obtain vegetation association data in the multi-source data set, process the vegetation association data through a preset normalized vegetation index model, and obtain vegetation coverage feature data; and calculate the vegetation coverage feature data through a preset normalized vegetation index model. The trained deep learning model processes the stratigraphic lithology data in the multi-source dataset to obtain stratigraphic lithology feature vector data, and performs dimensionality reduction processing on the stratigraphic lithology feature vector data to obtain stratigraphic lithology feature data; rasterizes the fracture data in the multi-source dataset, calculates the intensity value of each grid unit affected by the fracture, and generates fracture feature data based on the intensity value; fuses the multi-dimensional data features extracted from the multi-source dataset through the attention mechanism, and maps the multi-dimensional data features to the same dimensional space; calculates the attention weight of each multi-dimensional data feature to obtain a multi-source training dataset.

[0132] As an optional implementation, the geological disaster warning model module 53 is further specifically used to use a multi-layer perceptron, a residual network and a visual model as sub-models to construct an initial geological disaster warning model, and input the multi-source training data in the multi-source training data set into the initial geological disaster warning model; perform initial feature extraction and nonlinear transformation on the multi-source training data through the multi-layer hidden layers of the multi-layer perceptron to obtain first geological feature vector data; input the multi-source training data into the residual block of the residual network, perform data transformation through the residual block, and perform residual link operation on the feature vector after data change processing to obtain second geological feature vector data; divide the multi-source training data into An image block vector is obtained by performing a self-attention mechanism calculation on the image block vector through the self-attention mechanism module of the visual model, and the calculation result is processed using a multi-layer encoder to obtain third geological feature vector data; based on the attention mechanism, feature fusion processing is performed on the first geological feature vector data, the second geological feature vector data and the third geological feature vector data to obtain a geological feature fusion vector; a prediction calculation is performed based on the geological feature fusion vector, and the loss value between the predicted value and the true value is calculated through a binary cross entropy loss function, and the weight parameters and bias parameters of each sub-model are continuously optimized according to the loss value until the loss value converges to obtain the trained geological disaster warning model.

[0133] As an optional implementation, the geological disaster prediction result module 54 is also specifically used to collect new multi-source data in real time through the data collection path, and use wavelet transform to remove noise data in the new multi-source data; use a preset deep learning-based anomaly detection model to identify outliers in the new multi-source data, and repair outliers through a preset machine learning-based interpolation method; send the new multi-source data after data preprocessing to the geological disaster warning model, and perform real-time analysis and prediction through the geological disaster warning model to obtain geological disaster warning result data; use the new multi-source data to perform online training and optimization on the geological disaster warning model through a preset stochastic gradient descent algorithm.

[0134] As an optional implementation, the training data set module 52 is also specifically used to obtain position information data in the multi-source data set; use the position information data to perform calculations using a preset position coding formula to obtain a position coding feature vector; and perform feature coding fusion processing on the position coding feature vector and the multi-dimensional data features.

[0135] As an optional implementation, the geological disaster warning model module 53 is also specifically used to first train a preset first number of rounds at a first learning rate during the training process; when the training process of the first number of rounds is completed, continue training at a second learning rate, wherein, during the training process, a regularization term is added to the loss function through a regularization method, the first learning rate is less than the second learning rate, and the second learning rate decreases according to the preset number of training rounds.

[0136] It should be noted that the above-mentioned geological disaster early warning device can execute the geological disaster early warning method provided in the embodiment of this application, and has the corresponding functional modules and beneficial effects of the execution method. For technical details not fully described in the embodiment of the geological disaster early warning device, please refer to the geological disaster early warning method provided in the embodiment of this application.

[0137] Figure 6 Schematic diagram of the hardware structure of the electronic device 600 for executing the geological disaster early warning method provided in the embodiment of the present application. Figure 6 As shown, the electronic device 600 includes:

[0138] One or more processors 610 and memory 620, Figure 6 A processor 610 is taken as an example.

[0139] The processor 610 and the memory 620 may be connected via a bus or other means. Figure 6 The bus connection is taken as an example.

[0140] Memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the geological disaster early warning method in the embodiments of the present application. Processor 610 executes the non-volatile software programs, instructions, and modules stored in memory 620 to execute various server functional applications and data processing, thereby implementing the geological disaster early warning method in the above-mentioned method embodiment.

[0141] Memory 620 may include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function; the data storage area may store data generated based on the use of the geological disaster early warning device. Furthermore, memory 620 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state memory device. In some embodiments, memory 620 may optionally include memory remotely located relative to processor 610. Such remote memory may be connected to the geological disaster early warning device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0142] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, execute the geological disaster early warning method in any of the above method embodiments, for example, execute the above described Figure 2 Steps S1 to S5 of the method, Figure 3 Steps S21 to S27 of the method, Figure 4 Steps S31 to S36 of the method are implemented Figure 5 The functions of modules 51-55 in.

[0143] The above-mentioned product can execute the method provided in the embodiment of this application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of this application.

[0144] An embodiment of the present application provides a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by one or more processors, for example Figure 6 A processor 610 in the embodiment may enable the one or more processors to execute the geological disaster early warning method in any of the above method embodiments, for example, executing the above described Figure 2 Steps S1 to S5 of the method, Figure 3 Steps S21 to S27 of the method, Figure 4 Steps S31 to S36 of the method are implemented Figure 5 The functions of modules 51-55 in.

[0145] The present application provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by the electronic device, the electronic device is enabled to perform the electronic device method in any of the above method embodiments, for example, the above described Figure 2 Steps S1 to S5 of the method, Figure 3 Steps S21 to S27 of the method, Figure 4 Steps S31 to S36 of the method are implemented Figure 5 The functions of modules 51-55 in.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0147] Through the description of the above embodiments, it is clear to those skilled in the art that each embodiment can be implemented by means of software plus a general hardware platform, or of course by hardware. It is understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Based on the concept of the present application, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the present application as described above. For the sake of simplicity, they are not provided in detail. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A geological disaster early warning method, characterized in that: include: Collecting multi-source data of preset data dimensions through a preset data collection path, and performing data preprocessing on the collected multi-source data to obtain preprocessed multi-source data; Using the preprocessed multi-source data to construct a multi-source dataset including positive sample data and negative sample data, and fusing multi-dimensional data features extracted from the multi-source dataset through an attention mechanism to obtain a multi-source training dataset; Using a multi-layer perceptron, a residual network and a visual model to construct an initial geological disaster warning model, and training the initial geological disaster warning model with the multi-source training data set to obtain a trained geological disaster warning model; Collect the new multi-source data in real time through the data collection approach, and use the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data; When the geological disaster warning result data meets the preset warning conditions, the graded warning information is released through the preset channels, and the pre-trained rescue resource intelligent deployment model is used to generate a rescue resource deployment plan based on the geological disaster warning result data; Among them, the steps of using a multi-layer perceptron, a residual network and a visual model to construct an initial geological hazard warning model, training the initial geological hazard warning model through the multi-source training data set, and obtaining a trained geological hazard warning model include: using a multi-layer perceptron, a residual network and a visual model as sub-models to construct an initial geological hazard warning model, and inputting multi-source training data in the multi-source training data set into the initial geological hazard warning model; performing initial feature extraction and nonlinear transformation on the multi-source training data through the multi-layer hidden layers of the multi-layer perceptron to obtain first geological feature vector data; inputting the multi-source training data into the residual block of the residual network, performing data transformation through the residual block, and performing residual link operation on the feature vector after data change processing. Calculate to obtain second geological feature vector data; divide the multi-source training data into image block vectors, perform self-attention mechanism calculation according to the image block vectors through the self-attention mechanism module of the visual model, and process the calculation results using a multi-layer encoder to obtain third geological feature vector data; perform feature fusion processing on the first geological feature vector data, the second geological feature vector data and the third geological feature vector data based on the attention mechanism to obtain a geological feature fusion vector; perform prediction calculation according to the geological feature fusion vector, calculate the loss value between the predicted value and the true value through the binary cross entropy loss function, and continuously optimize the weight parameters and bias parameters of each sub-model according to the loss value until the loss value converges to obtain the trained geological disaster early warning model.

2. The geological disaster early warning method according to claim 1, characterized in that: The step of collecting multi-source data of preset data dimensions through a preset data collection approach, and performing data preprocessing on the collected multi-source data to obtain preprocessed multi-source data includes: Collecting geological factor-related data of preset data dimensions through a preset data collection path, and sorting the geological factor-related data in chronological order; Calculating a mean and a standard deviation of each type of geological factor-related data arranged in time series, and removing the geological factor-related data if the geological factor-related data exceeds a first data range, wherein the first data range is composed of the mean and the standard deviation; The missing value processing model based on autoregressive integrated moving average is used to process the missing values of the geological factor correlation data, wherein the expression of the missing value processing model is: , represents the autoregressive operator, represents the moving average operator, represents the backward shift operator, represents the difference order, represents a white noise sequence, express t The geological factor related data at the time; Frequency domain analysis is performed on the geological factor associated data arranged in time series by Fourier transform, the periodic components of the geological factor associated data are extracted, and the amplitudes and phases of different frequency components are analyzed to obtain the periodic variation law of the geological factor associated data.

3. The geological disaster early warning method according to claim 1, characterized in that: The step of using the preprocessed multi-source data to construct a multi-source dataset including positive sample data and negative sample data, and fusing multi-dimensional data features extracted from the multi-source dataset through an attention mechanism to obtain a multi-source training dataset includes: Executing a preset sliding window sampling strategy, setting a preset number of grid areas as evaluation units, determining positive sample data and negative sample data according to the evaluation units, and adding the positive sample data and the negative sample data to the multi-source data set; Using the multi-source dataset, calculating terrain dimension feature data and landform dimension feature data through a preset digital elevation model; Acquiring vegetation-related data from the multi-source dataset, and processing the vegetation-related data using a preset normalized vegetation index model to obtain vegetation coverage characteristic data; Processing the formation lithology data in the multi-source data set through a pre-trained deep learning model to obtain formation lithology feature vector data, and performing dimensionality reduction processing on the formation lithology feature vector data to obtain formation lithology feature data; rasterizing the fracture data in the multi-source dataset, calculating the intensity value of each grid cell affected by the fracture, and generating fracture feature data according to the intensity value; fusing multi-dimensional data features extracted from the multi-source datasets through an attention mechanism, and mapping the multi-dimensional data features into the same dimensional space; Calculate the attention weight of each of the multi-dimensional data features to obtain a multi-source training data set.

4. The geological disaster early warning method according to claim 1, characterized in that: The step of collecting the new multi-source data in real time through the data collection approach, and using the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data includes: Collecting new multi-source data in real time through the data collection approach, and removing noise data from the new multi-source data using a wavelet transform method; Identify outliers in the new multi-source data using a preset deep learning-based anomaly detection model, and repair the outliers using a preset machine learning-based interpolation method; Sending the new multi-source data after data preprocessing to the geological disaster early warning model, and performing real-time analysis and prediction through the geological disaster early warning model to obtain geological disaster early warning result data; The geological disaster early warning model is trained and optimized online using the new multi-source data through a preset stochastic gradient descent algorithm.

5. The geological disaster early warning method according to claim 3, characterized in that: Before the step of calculating the attention weights of the multi-dimensional data features to obtain a multi-source training data set, the method further includes: Acquiring location information data from the multi-source data set; The position information data is used to calculate using a preset position coding formula to obtain a position coding feature vector; Perform feature coding fusion processing on the position coding feature vector and the multi-dimensional data feature.

6. The geological disaster early warning method according to claim 1, characterized in that: Before the step of performing prediction calculation based on the geological feature fusion vector and calculating the loss value between the predicted value and the true value using a binary cross entropy loss function, the method further includes: During the training process, a first learning rate is used to train a preset first number of rounds; When the training process of the first number of rounds is completed, the training is continued at a second learning rate, wherein a regularization term is added to the loss function by a regularization method during the training process, the first learning rate is less than the second learning rate, and the second learning rate decreases according to a preset number of training rounds.

7. A geological disaster early warning device, characterized in that: include: A data collection preprocessing module is used to collect multi-source data of preset data dimensions through a preset data collection path, and perform data preprocessing on the collected multi-source data to obtain preprocessed multi-source data; A training dataset module is configured to construct a multi-source dataset comprising positive sample data and negative sample data using the pre-processed multi-source data, and to fuse multi-dimensional data features extracted from the multi-source dataset through an attention mechanism to obtain a multi-source training dataset; A geological disaster warning model module is used to construct an initial geological disaster warning model using a multi-layer perceptron, a residual network and a visual model, and train the initial geological disaster warning model using the multi-source training data set to obtain a trained geological disaster warning model; A geological disaster prediction result module is used to collect new multi-source data in real time through the data collection path, and use the geological disaster early warning model to perform real-time analysis and prediction on the new multi-source data to obtain geological disaster early warning result data; A geological disaster release and rescue module is used to release graded warning information through preset channels when the geological disaster warning result data meets the preset warning conditions, and use a pre-trained rescue resource intelligent deployment model to generate a rescue resource deployment plan based on the geological disaster warning result data; Among them, the geological disaster warning model module is also specifically used to use a multi-layer perceptron, a residual network and a visual model as sub-models to construct an initial geological disaster warning model, and input the multi-source training data in the multi-source training data set into the initial geological disaster warning model; the multi-source training data is subjected to initial feature extraction and nonlinear transformation through the multi-layer hidden layers of the multi-layer perceptron to obtain first geological feature vector data; the multi-source training data is input into the residual block of the residual network, data transformation is performed through the residual block, and the feature vector after data change processing is subjected to residual link operation to obtain second geological feature vector data; the multi-source training data is divided into image block vectors, and the self-attention mechanism module of the visual model is used to perform self-attention mechanism calculation according to the image block vector, and the calculation result is processed using a multi-layer encoder to obtain third geological feature vector data; Based on the attention mechanism, feature fusion processing is performed on the first geological feature vector data, the second geological feature vector data and the third geological feature vector data to obtain a geological feature fusion vector; prediction calculation is performed based on the geological feature fusion vector, and the loss value between the predicted value and the true value is calculated through the binary cross entropy loss function, and the weight parameters and bias parameters of each sub-model are continuously optimized according to the loss value until the loss value converges to obtain the trained geological disaster early warning model.

8. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the geological disaster early warning method described in any one of claims 1-6.

9. A non-volatile computer-readable storage medium, characterized in that: The non-volatile computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are executed by an electronic device, the electronic device executes the geological disaster early warning method described in any one of claims 1 to 6.

Images