Dynamic Monitoring Method for Landslide Disasters Based on Time Series and Hybrid Model

Abstract

The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, belonging to the field of intelligent monitoring technology, including: collecting multi-source heterogeneous data of historical landslides, preprocessing the multi-source heterogeneous data to obtain first data, classifying the first data by type, and through time series encoding for each type of data, obtaining a single sample under the same time series and then obtaining a sample set, and then training a neural network model according to the sample set to obtain an initial model; verifying the model of the initial model, and combining the constructed mapping relationship between landslide deformation and monitoring dimensions of the corresponding type, optimizing to obtain a time series and hybrid model; monitoring and obtaining the current heterogeneous data of the target monitoring point, and inputting it into the time series and hybrid model, predicting the current disaster result of the target monitoring point and outputting a reminder. Through the model verification and the construction of the mapping relationship, the accuracy of the model is guaranteed, providing a convenient basis for landslide disaster prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring, and particularly relates to a dynamic monitoring method for landslide disasters based on time series and hybrid models. Background Art

[0002] A landslide refers to the natural phenomenon that soil or rock mass on a slope, under the influence of factors such as river erosion, groundwater activity, earthquake, and artificial slope cutting, slides down the slope as a whole or dispersedly along a certain weak surface or weak zone under the action of gravity.

[0003] In the research on landslides, the key roles of rainfall, reservoir water level changes, and topographic and geological conditions have been clarified in the analysis of disaster-causing factors. The application of monitoring technologies presents a situation of coexistence of multiple means and a development trend towards multi-source integration. There are various prediction methods such as time series analysis, and limit equilibrium analysis and other methods are used for stability evaluation and prevention, and corresponding engineering and non-engineering measures are implemented to ensure the safety of the reservoir area. However, there are some deficiencies in the current research. On the one hand, the prediction method is affected by the coupling of multiple factors, which affects the prediction accuracy.

[0004] Therefore, the present invention proposes a dynamic monitoring method for landslide disasters based on time series and hybrid models. Summary of the Invention

[0005] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, which is used to ensure the reliability of samples by collecting multi-source heterogeneous data and performing data preprocessing, and is convenient for realizing the training of the model through time series coding. Subsequently, through the verification of the model and the construction of the mapping relationship, the accuracy of the model is further ensured, providing a convenient basis for landslide disaster prediction.

[0006] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, including:

[0007] Step 1: Collect multi-source heterogeneous data of historical landslides, and perform preprocessing on the multi-source heterogeneous data to obtain first data, wherein the multi-source heterogeneous data involves multiple monitoring dimensions;

[0008] Step 2: Perform type division on the first data, and through time series coding on each type of data, obtain a single sample at the same time series and then obtain a sample set, and then train a neural network model according to the sample set to obtain an initial model, wherein the single sample includes monitoring values under different monitoring dimensions and the horizontal displacement of the landslide;

[0009] Step 3: Perform model verification on the initial model, and optimize to obtain a time series and hybrid model in combination with the mapping relationship between landslide deformation and monitoring dimensions corresponding to the constructed type;

[0010] Step 4: Monitor and obtain the current heterogeneous data of the target monitoring point, and input it into the time series and hybrid model to predict the current disaster result of the target monitoring point and output a reminder.

[0011] Preferably, preprocessing the multi-source heterogeneous data to obtain first data, including:

[0012] Classify the first data according to the monitoring dimension, and draw a curve smoothing for each first value in each type of data in chronological order;

[0013] Respectively determine the first absolute slope of the straight line formed by two adjacent points in the smoothed curve, and adjust the first absolute slope in combination with the curve area formed by the two adjacent points to obtain the second absolute slope;

[0014] ;

[0015] where D2 is the corresponding second absolute slope; D1 is the corresponding first absolute slope; represents the value of the highest point in the curve segment formed by the i-th point and the i+1-th point; represents the value of the corresponding point consistent with the highest point in the straight line formed by the i-th point and the i+1-th point; represents the value of the lowest point in the curve segment formed by the i-th point and the i+1-th point; represents the value of the corresponding point consistent with the lowest point in the straight line formed by the i-th point and the i+1-th point; represents the closed area of the curve segment formed by the i-th point and the i+1-th point; represents the closed area of the straight line formed by the i-th point and the i+1-th point; n represents the total number of existing first values, and is the same as the number of points for drawing the smoothed curve;

[0016] Compare the first absolute slope and the second absolute slope of the same two adjacent points with the preset slope respectively. If both are less than or equal to the preset slope, then perform a first adjustment on the latter point of the corresponding two adjacent points according to max(D1, D2), where max represents the maximum symbol;

[0017] If both are greater than or equal to the preset slope, then perform a second adjustment on the latter point of the corresponding two adjacent points according to min(D1, D2), and min represents the minimum symbol;

[0018] If there is a situation where one of D1 and D2 is greater than the preset slope and the other is less than the preset slope, at this time, randomly select a final slope according to rand((D1, D2), 1) to perform a third adjustment on the latter point of the corresponding two adjacent points, and rand represents the random symbol;

[0019] Obtain the first data based on all adjustment results.

[0020] Preferably, perform a first adjustment on the latter point of the corresponding adjacent two points according to max(D1, D2), including:

[0021]

[0022] wherein, represents the value after the first adjustment of the latter point; represents the value before the latter point is not adjusted; represents the variance of D1 - D2 based on all adjacent two points; represents the positive and negative situation of the slope between the corresponding i-th point and the (i + 1)-th point.

[0023] Preferably, perform a second adjustment on the latter point of the corresponding adjacent two points according to min(D1, D2), including:

[0024]

[0025] wherein, represents the value after the second adjustment of the latter point.

[0026] Preferably, randomly select a final slope according to rand((D1, D2), 1) to perform a third adjustment on the latter point of the corresponding adjacent two points, including:

[0027]

[0028]

[0029] wherein, represents the value after the third adjustment of the latter point; represents the adjustment function for the corresponding adjacent two points; rand((D1, D2), 1) represents the function of randomly selecting a final slope from D1 and D2 of the corresponding adjacent two points; represents the preset slope.

[0030] Preferably, optimize to obtain a time series and a hybrid model, including:

[0031] Extract the first multi-source data and the second source data at two adjacent moments from the sample set, wherein the first multi-source data corresponds to the first historical moment, the second multi-source data corresponds to the second historical moment, and the first historical moment is before the second historical moment;

[0032] Input the first multi-source data into the initial model to obtain the verification type and the first verification coefficients under different monitoring dimensions;

[0033] According to the historical displacements corresponding to the disaster types and the mapping relationship, respectively determine the first disaster coefficients of each monitoring dimension in the same single sample corresponding to the disaster types;

[0034] Obtain all the first disaster coefficients of different monitoring dimensions involved in the same disaster type, and construct the first coefficient vector of each monitoring dimension under the same disaster type;

[0035] Extract the first verification coefficients and verification types of the first multi-source data corresponding to the first coefficient vector to obtain the verification coefficient vector and the verification type vector, and combine with the first coefficient vector to obtain the first matrix;

[0036] Based on the first matrix, obtain the difference factors of each monitoring dimension in the same disaster type determined by historical practice;

[0037] Rely on the first historical difference between the second multi-source data and the first multi-source data, the type difference between the historical actual disaster type under the second multi-source data and the verification type of the first multi-source data, and combine the difference factors of each monitoring dimension based on the corresponding historical actual disaster type to construct the optimization vector corresponding to the first historical moment;

[0038] Optimize the initial model based on all the optimization vectors to obtain the time series and the hybrid model.

[0039] Preferably, according to the historical displacements corresponding to the disaster types and the mapping relationship, respectively determine the first disaster coefficients of each monitoring dimension in the same single sample corresponding to the disaster types, including:

[0040] Respectively obtain the historical displacements of different disaster types from the time series - type - displacement comparison table;

[0041] Determine N1 first samples having a time series association with the single sample, sort the monitoring values of each monitoring dimension in the N1 + 1 samples based on the time series order, and determine whether there are differences on both sides of the corresponding single sample;

[0042] If there are differences on both sides, then according to and the size relationship of and the size relationship of, determine the important coefficient of the corresponding single sample under the corresponding monitoring dimension, where, represents the left difference of the corresponding single sample; represents the right difference of the corresponding single sample; represents the difference threshold under the corresponding monitoring dimension; represents the monitoring value of the corresponding single sample under the corresponding monitoring dimension; 、 respectively represent the left monitoring value and the right monitoring value based on the corresponding single sample under the corresponding monitoring dimension; , respectively represent constants;

[0043] If there is only a difference on one side, then according to determine the importance coefficient of the corresponding single sample under the corresponding monitoring dimension, where represents the average value of the monitoring values corresponding to N1 + 1 samples under the corresponding monitoring dimension; represents the maximum value among the monitoring values corresponding to N1 + 1 samples; represents the variance of the monitoring values corresponding to N1 + 1 samples, where , determine that the corresponding importance coefficient is 0; otherwise, determine that the corresponding importance coefficient is ;

[0044] Based on the mapping relationship of the corresponding type and all historical displacements under the corresponding type, determine the set weight of each monitoring dimension in the corresponding disaster type, and combine the importance coefficient to determine the first disaster coefficient of the corresponding monitoring dimension in the corresponding disaster type;

[0045]

[0046] where Ui represents the first disaster coefficient of the i-th monitoring dimension in the corresponding disaster type; represents the historical disaster value of the corresponding disaster type; represents the set weight of the i-th monitoring dimension in the corresponding disaster type; represents the importance coefficient of the i-th monitoring dimension in the corresponding disaster type; represents the sum of the set weights of all monitoring dimensions under the corresponding disaster type, with a value of 1; represents the sum of the importance coefficients of all monitoring dimensions under the corresponding disaster type; m represents the number of monitoring dimensions.

[0047] Preferably, based on the first matrix, obtain the difference factors based on each monitoring dimension in the same disaster type determined by historical reality, including:

[0048] If all the verification types involved in the first matrix are consistent with the corresponding disaster type, at this time, calculate the absolute value of the difference between the first disaster coefficient and the first verification coefficient in each column vector, and obtain the coefficient variance and combine the corresponding disaster type to obtain the difference factor;

[0049] If there is a type among all the verification types involved in the first matrix that is inconsistent with the corresponding disaster type, the occurrence times of each inconsistent type are extracted separately, and the absolute value of the difference between all the first disaster coefficients and the first verification coefficients under the inconsistent type is combined to obtain the current factor of each inconsistent type, thereby constituting the difference factor.

[0050] Compared with the prior art, the beneficial effects of the present application are as follows:

[0051] By collecting multi-source heterogeneous data and performing data preprocessing, the reliability of the samples is ensured, and through time series coding, it is convenient to implement the training of the model. Subsequently, through model verification and the construction of mapping relationships, the accuracy of the model is further ensured, providing a convenient basis for landslide disaster prediction.

[0052] Other features and advantages of the present invention will be described in the following specification, and part of them will become obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification and the drawings.

[0053] The technical solutions of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0054] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0055] Figure 1 is a flowchart of a landslide disaster dynamic monitoring method based on time series and hybrid model in an embodiment of the present invention;

[0056] Figure 2 is a comparison chart of the actual and predicted values of the present invention. Detailed Embodiments

[0057] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0058] The present invention provides a landslide disaster dynamic monitoring method based on time series and hybrid model, as Figure 1 shown, including:

[0059] Step 1: Collect multi-source heterogeneous data of historical landslides, and preprocess the multi-source heterogeneous data to obtain first data, wherein the multi-source heterogeneous data involves multiple monitoring dimensions;

[0060] Step 2: Classify the first data by type, and through time series encoding for each type of data, obtain a single sample at the same time series, and then obtain a sample set. Then, train a neural network model according to the sample set to obtain an initial model, where the single sample includes monitoring values under different monitoring dimensions and landslide horizontal displacement;

[0061] Step 3: Verify the initial model, and combine the established mapping relationship between landslide deformation and monitoring dimensions of the corresponding type to optimize and obtain a time series and hybrid model;

[0062] Step 4: Monitor and obtain the current heterogeneous data of the target monitoring point, and input it into the time series and hybrid model to predict the current disaster result of the target monitoring point and output a reminder.

[0063] In this embodiment, the neural network model includes: an input layer, an embedding layer, an encoder layer, a bidirectional long short-term memory network, a fully connected layer, and an output layer. Specifically:

[0064] Input layer: The input dimension is set to 4, which will be elaborated in detail in the parameter settings.

[0065] Embedding layer: The model first includes an embedding layer whose function is to perform dimensionality conversion on the input features. Here, the input feature dimension is 4. After being processed by this linear layer, the output feature dimension becomes 64 and has a bias. The role of this embedding layer is to perform a mapping on the original input data so that it can better adapt to the processing requirements of subsequent model layers.

[0066] Encoder layer: It contains multiple important components inside:

[0067] Multi-head Self-Attention: Through its internal output projection layer, the input feature dimension is processed from 64 and the output is still 64 dimensions, which is used to capture the correlation information between different positions of the input sequence, enabling the model to pay attention to different parts of the data and their mutual relationships.

[0068] Linear transformation layers (Linear1 and Linear2): Linear layer 1 expands the input 64-dimensional features to 2048 dimensions, and then undergoes a random inactivation Dropout operation (randomly discarding some neurons with a probability of 0.1 to prevent overfitting). Then, linear layer 2 changes the dimension from 2048 dimensions back to 64 dimensions. Through such a transformation, the features of the data are extracted and integrated.

[0069] Normalization Layers (Norm1 and Norm2): There are two layer normalization operations respectively, with the eps in their parameters set to 1e-05. Their function is to normalize the input data of each layer, which helps to stabilize the gradients during model training, accelerate convergence and improve model performance.

[0070] Additional Dropout Layers (Dropout1 and Dropout2): These two Dropout layers also randomly discard neurons with a probability of 0.1, further assisting in preventing overfitting.

[0071] Bidirectional Long Short-Term Memory Network (BiLSTM):

[0072] This part is a bidirectional LSTM that receives a 64-dimensional feature input processed previously. The output of each direction of the LSTM is 32-dimensional, but since it is bidirectional, the final output dimension is 32×2 = 64 dimensions. The bidirectional structure is adopted in this layer, which means it can consider the information of the sequence data in both the forward and reverse directions simultaneously, capture the temporal dependencies in the sequence more comprehensively, and at the same time, a Dropout operation is set with a probability of 0.1 to avoid model overfitting.

[0073] Fully Connected Layer:

[0074] Then there is a fully connected layer that receives the features from the output of the bidirectional LSTM. Since the BiLSTM is bidirectional, the output dimension should be 512 dimensions after operations such as merging (corresponding dimension integration is carried out according to the characteristics of the bidirectional LSTM), and then it is converted into a 1-dimensional output through this linear layer (Linear), and this layer has a bias, which plays a role in appropriately adjusting the output result.

[0075] Output Layer:

[0076] This is the last layer of the model, which outputs the processing result of the prediction model for the 4-dimensional input data in 1 dimension.

[0077] Set its core parameters and training parameters, and then introduce the sample set for model training. The parameter setting table is as follows:

[0078]

[0079] In parameter settings, since the input data includes four features: horizontal displacement, elevation, rainfall, and reservoir water level, the input dimension is set to 4. The purpose of this model is to obtain the disaster situation, so the output dimension is 1. The length of the input sequence is adjusted to 8 according to the scale of the dataset. To improve the generalization ability of the model and reduce overfitting, the dropout technique is adopted, and the dropout rate is usually set between 0.1 and 0.2.

[0080] In the optimization of training parameters, through repeated experiments and adjustments, the training configuration most suitable for the current model and dataset is found. The number of epochs is set to 25, which not only ensures that the model has enough epochs for learning but also avoids the problem of overfitting. Since the Adam optimization algorithm is used, the learning rate is set to 0.001. The number of training samples per batch is set to 32, which can achieve a balance between memory usage and computational efficiency and helps to enhance the generalization performance of the model. The proportion of the validation set is set to 20%, which helps to reasonably allocate training and validation data and thus more accurately evaluate the effect of the model.

[0081] By using the provided model architecture and parameter configuration, the training dataset is used for the training process. The purpose of training is to enable the model to identify and understand the patterns and features in the data so that it can make relatively accurate predictions. Figure 2 As shown, it presents the prediction results of the horizontal displacement of the GNSS monitoring point No. 02. Among them, the red line represents the actually observed horizontal displacement data, while the blue line represents the predicted values generated by the model.

[0082] In this embodiment, a historical landslide refers to an area where there is a whole or scattered downward slope movement along a certain weak surface or weak zone, and all kinds of data related to the historical landslide are known, including: surface displacement monitoring, rainfall monitoring, reservoir water level change monitoring, etc. And the surface displacement monitoring uses GNSS real-time monitoring to obtain the surface displacement change of the landslide to master the deformation dynamics of the landslide; rainfall monitoring is obtained in real time by a rain gauge station; reservoir water level change monitoring is obtained through data collection, providing basic data for landslide deformation analysis and early warning.

[0083] In this embodiment, according to the basic characteristics and deformation characteristics of the landslide body, combined with the existing exploration data, the main monitoring content of the landslide is determined as surface displacement monitoring, supplemented by rainfall monitoring and reservoir water level change monitoring.

[0084] In this embodiment, multi-source heterogeneous data refers to data under five dimensions. Among them, there are 4 input dimensions, namely: horizontal displacement, elevation, rainfall, and reservoir water level, and the output dimension is the disaster situation. It should be noted that the disaster situation is the horizontal displacement of the landslide.

[0085] That is, a single sample = {horizontal displacement, elevation, rainfall, reservoir water level at corresponding time series, horizontal displacement of landslide at corresponding time series}.

[0086] In this embodiment, preprocessing refers to adjusting the result values involved in multi-source heterogeneous data. Therefore, there may be certain measurement errors during the monitoring process. Therefore, the values are finely adjusted based on continuous time series to ensure the reliability of the data and avoid reducing the accuracy of the model due to errors.

[0087] In this embodiment, type division is the division of the first data according to five dimensions, and the data under each dimension is the corresponding type of data.

[0088] In this embodiment, constructing the mapping relationship between landslide deformation and influencing factors of the corresponding type means determining the influence of the data under the corresponding dimension on the landslide. The greater the influence, the greater the weight corresponding to this dimension, and the influencing factor refers to the influence of the corresponding dimension.

[0089] In this embodiment, time series coding refers to setting the collection time for different data to facilitate subsequent sample construction.

[0090] In this embodiment, the current heterogeneous data of the target monitoring point refers to the data under the four input dimensions obtained by measurement.

[0091] In this embodiment, the current disaster is related to the horizontal displacement result of the predicted target monitoring point. If the horizontal displacement result exceeds the set displacement, a reminder is required; otherwise, no reminder is required. And the reminder refers to reminding the relevant guardians. If the predicted result is that the horizontal displacement under the current disaster is 10 cm, but the set displacement is 8 cm, at this time, a warning is required.

[0092] The beneficial effects of the above technical solutions are: By collecting multi-source heterogeneous data and performing data preprocessing, the reliability of the samples is ensured, and through time series coding, it is convenient to implement the training of the model. Subsequently, through model verification and the construction of the mapping relationship, the accuracy of the model is further ensured, providing a convenient basis for landslide disaster prediction.

[0093] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, which preprocesses the multi-source heterogeneous data to obtain first data, including:

[0094] Dividing the multi-source heterogeneous data into types according to the monitoring dimension, and smoothly plotting a curve for each first value in each type of data according to the time sequence;

[0095] Determine the first absolute slope of the straight line formed by two adjacent points in the smooth curve respectively, and adjust the first absolute slope in combination with the curve area formed by the two adjacent points to obtain the second absolute slope;

[0096]

[0097] wherein, D2 is the corresponding second absolute slope; D1 is the corresponding first absolute slope; represents the value of the highest point in the curve segment formed by the i-th point and the (i + 1)-th point; represents the value of the corresponding point consistent with the highest point in the straight line formed by the i-th point and the (i + 1)-th point; represents the value of the lowest point in the curve segment formed by the i-th point and the (i + 1)-th point; represents the value of the corresponding point consistent with the lowest point in the straight line formed by the i-th point and the (i + 1)-th point; represents the closed area of the curve segment formed by the i-th point and the (i + 1)-th point; represents the closed area of the straight line formed by the i-th point and the (i + 1)-th point; n represents the total number of existing first values, and is consistent with the number of points for drawing the smooth curve;

[0098] Compare the first absolute slope and the second absolute slope of the same two adjacent points with a preset slope respectively. If both are less than or equal to the preset slope, then perform a first adjustment on the latter point of the corresponding two adjacent points according to max(D1, D2), where max represents the maximum symbol;

[0099] If both are greater than or equal to the preset slope, then perform a second adjustment on the latter point of the corresponding two adjacent points according to min(D1, D2), and min represents the minimum symbol;

[0100] If there is a situation where one of D1 and D2 is greater than the preset slope and the other is less than the preset slope, at this time, randomly select a final slope according to rand((D1, D2), 1) to perform a third adjustment on the latter point of the corresponding two adjacent points, and rand represents the random symbol;

[0101] Obtain the first data based on all adjustment results.

[0102] In this embodiment, the adjusted data in each dimension is statistically analyzed to obtain the first data.

[0103] In this embodiment, the multi-source heterogeneous data under each monitoring dimension is classified by type to obtain the data under different dimensions. The abscissa of the drawn smooth curve is time, and the ordinate is the value monitored under the corresponding dimension, and the values of all monitored values are greater than or equal to 0.

[0104] In this embodiment, for example, there are points 1, 2, and 3 arranged in chronological order on the smooth curve. At this time, point 1 and point 2, and point 2 and point 3 are adjacent points respectively.

[0105] In this embodiment, the first absolute slope = ∣(value of the latter point - value of the former point) / time interval∣.

[0106] In this embodiment, for example, the curve area between point 1 and point 2 is the area of the closed region enclosed by the smooth curve vertically cut with point 1 and point 2 as the reference points and the time axis.

[0107] In this embodiment, the purpose of adjusting the slope is to reduce the area difference between the curve and the straight line, so as to minimize the inaccuracy of subsequent disaster situation analysis caused by measurement errors.

[0108] In this embodiment, the preset slope value is generally 0.1.

[0109] In this embodiment, by comparing the two absolute slopes with the preset slope, the reliability of the corresponding value adjustment is further ensured to reduce data errors.

[0110] The beneficial effect of the above technical solution is: drawing a smooth curve based on the data in each dimension, and determining the reliability of the data by comparing the slopes before and after adjustment between two points with the preset slope, providing an accurate basis for subsequent training of the model.

[0111] The present invention provides a landslide disaster dynamic monitoring method based on time series and hybrid model. According to max(D1, D2), the latter point of the corresponding adjacent two points is subjected to the first adjustment, including:

[0112]

[0113] Among them, represents the value after the first adjustment of the latter point; represents the value before the latter point is not adjusted; represents the variance of D1 - D2 based on all adjacent two points; represents the positive and negative situation of the slope between the corresponding i-th point and the (i + 1)-th point.

[0114] In this embodiment, the positive and negative situation of the slope refers to the situation of being greater than 0, less than 0, or equal to 0 before the absolute value of the slope is added.

[0115] The beneficial effect of the above technical solution is: making the first adjustment to the relevant values based on the positive and negative of the slope.

[0116] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, which performs a second adjustment on the latter point of the corresponding adjacent two points according to min(D1, D2), including:

[0117]

[0118] Among them, represents the value after the second adjustment of the latter point.

[0119] The beneficial effect of the above technical solution is that the relevant values are second-adjusted based on the positive and negative of the slope.

[0120] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, which randomly selects a final slope according to rand((D1, D2), 1) to perform a third adjustment on the latter point of the corresponding adjacent two points, including:

[0121]

[0122]

[0123] Among them, represents the value after the third adjustment of the latter point; D3 represents the adjustment function for the corresponding adjacent two points; rand((D1, D2), 1) represents the function of randomly selecting a final slope from D1 and D2 of the corresponding adjacent two points; represents the preset slope.

[0124] The beneficial effect of the above technical solution is that the relevant values are third-adjusted based on the random function.

[0125] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models, which optimizes the time series and hybrid models, including:

[0126] Extract the first multi-source data and the second source data at two adjacent moments from the sample set, where the first multi-source data corresponds to the first historical moment, the second multi-source data corresponds to the second historical moment, and the first historical moment is before the second historical moment;

[0127] Input the first multi-source data into the initial model to obtain the verification type and the first verification coefficients under different monitoring dimensions;

[0128] According to the historical displacement and the mapping relationship of the corresponding disaster type, respectively determine the first disaster coefficients of each monitoring dimension and the corresponding disaster type in the same single sample;

[0129] Obtain all the first disaster coefficients for different monitoring dimensions involved in the same disaster type, and construct the first coefficient vector for each monitoring dimension under the same disaster type;

[0130] Extract the first verification coefficient and verification type of the first multi-source data corresponding to the first coefficient vector to obtain the verification coefficient vector and verification type vector, and combine with the first coefficient vector to obtain the first matrix;

[0131] Based on the first matrix, obtain the difference factors for each monitoring dimension in the same disaster type determined historically and actually;

[0132] Rely on the first historical difference between the second multi-source data and the first multi-source data, the type difference between the historically actual disaster type under the second multi-source data and the verification type of the first multi-source data, and combine the difference factors for each monitoring dimension corresponding to the historical actual disaster type to construct the optimization vector corresponding to the first historical moment;

[0133] Optimize the initial model based on all the optimization vectors to obtain the time series and the hybrid model.

[0134] In this embodiment, the first disaster coefficients are all known. That is, the monitoring values in different dimensions are matched from the dimension-monitoring-disaster comparison table, and the dimension-monitoring-disaster comparison table contains different monitoring values under different input dimensions and the disaster coefficients corresponding to these values. The value range of the disaster coefficient is from 0 to 1. Generally, the closer the monitoring value in the corresponding dimension is to the preset warning value, the larger the corresponding disaster coefficient. That is, before determining the horizontal displacement, it is necessary to perform disaster analysis on the monitoring values in each monitoring dimension, and then comprehensively obtain the horizontal displacement to ensure the accuracy of subsequent model prediction. Among them, in the case of the measurement data and horizontal displacement in the known four dimensions, the corresponding disaster type already exists and is pre-evaluated by experts. Therefore, there will be relevant verification coefficients and verification types in different dimensions during the operation of the model. Finally, only the horizontal displacement is output.

[0135] In this embodiment, the first multi-source data involves the monitoring values in four dimensions, the first disaster coefficient of each monitoring value, the corresponding disaster type, and the horizontal displacement. Therefore, inputting this data into the model for verification can obtain: the first verification coefficients, verification types, and verification displacements in four dimensions.

[0136] In this embodiment, the difference factor refers to the coefficients and types with differences.

[0137] In this embodiment, the first historical difference = {differences in monitoring values in four monitoring dimensions, differences in disaster coefficients corresponding to each monitoring value, differences in disaster types, differences in horizontal displacements}.

[0138] In this embodiment, the type difference is the difference between the historical actual disaster type at the second historical moment and the verification type at the first historical moment.

[0139] In this embodiment, the optimization vector = {the first historical difference, the type difference, the difference factor}.

[0140] In this embodiment, the time sequence can be 1 s, that is, determined through a short time. The result obtained by inputting the information of the known moment is used as the prediction result of the next moment. Furthermore, the optimization vector can be constructed through two comparisons to ensure the effective output of the displacement corresponding to the monitoring under the four dimensions of the current monitoring point.

[0141] In this embodiment, the first coefficient vector = {the first disaster coefficient of each time sequence under the same monitoring dimension for the corresponding disaster type}.

[0142] In this embodiment, the verification coefficient vector = {the first verification coefficient of each time sequence under the same monitoring dimension for the corresponding disaster type}.

[0143] In this embodiment, the verification type vector = {the verification type at each time sequence}.

[0144] In this embodiment, .

[0145] The beneficial effects of the above technical solution are as follows: By using the verification of the previous moment for the next moment through two consecutive moments, specifically constructing a matrix through the vectors composed of the disaster coefficient, the verification coefficient, and the disaster type to obtain the difference factor, and combining the historical difference, the type difference, etc. of multi-source data to construct the optimization vector, the reliable optimization of the model is realized, and the model accuracy is guaranteed.

[0146] The present invention provides a landslide disaster dynamic monitoring method based on time series and hybrid model. According to the historical displacement corresponding to the disaster type and the mapping relationship, the first disaster coefficient of each monitoring dimension and the corresponding disaster type in the same single sample are determined respectively, including:

[0147] Respectively obtain the historical displacements of different disaster types from the time sequence - type - displacement comparison table;

[0148] Determine N1 first samples having a time sequence association with the single sample, sort the monitoring values of each monitoring dimension in the N1 + 1 samples based on the time sequence order, and determine whether there is a difference between both sides of the corresponding single sample;

[0149] If there is a difference between both sides, then according to and the magnitude relationship of and Determine the importance coefficient of the corresponding single sample under the corresponding monitoring dimension according to the size relationship. Among them, represents the left difference of the corresponding single sample; represents the right difference of the corresponding single sample; represents the difference threshold under the corresponding monitoring dimension; represents the monitoring value of the corresponding single sample under the corresponding monitoring dimension; 、 respectively represent the left monitoring value and the right monitoring value based on the corresponding single sample under the corresponding monitoring dimension; 、 respectively represent constants;

[0150] If there is only a difference on one side, then according to determine the importance coefficient of the corresponding single sample under the corresponding monitoring dimension. Among them, represents the average value of the monitoring values corresponding to N1 + 1 samples under the corresponding monitoring dimension; represents the maximum value among the monitoring values corresponding to N1 + 1 samples; represents the variance of the monitoring values corresponding to N1 + 1 samples. Among them, , determine that the corresponding importance coefficient is 0; otherwise, determine that the corresponding importance coefficient is ;

[0151] Based on the mapping relationship of the corresponding type and all historical displacements under the corresponding type, determine the set weight of each monitoring dimension in the corresponding disaster type, and combine the importance coefficient to determine the first disaster coefficient of the corresponding monitoring dimension in the corresponding disaster type;

[0152]

[0153] Among them, Ui represents the first disaster coefficient of the i-th monitoring dimension in the corresponding disaster type; represents the historical disaster value of the corresponding disaster type; represents the set weight of the i-th monitoring dimension in the corresponding disaster type; represents the importance coefficient of the i-th monitoring dimension in the corresponding disaster type; represents the sum of the set weights of all monitoring dimensions under the corresponding disaster type, with a value of 1; represents the sum of the importance coefficients of all monitoring dimensions under the corresponding disaster type; m represents the number of monitoring dimensions.

[0154] In this embodiment, the set weight = , where Yc represents the influence degree corresponding to the mapping relationship of the corresponding type; sum1 represents the sum of all influence degrees of the corresponding type; represents the average value of all historical displacements of the corresponding type, Represents the maximum value among all historical positions under the corresponding type; Represents the sum of under all types, and the calculation results of under each type are different.

[0155] In this embodiment, the time series - type - displacement comparison table contains different historical actual disaster types and the corresponding historical displacements, all of which are stored for convenient direct retrieval and use. Moreover, the historical displacement under each type is unique because the time series is unique. Among them, the disaster type can be a step - type wading landslide type, a translational landslide, etc., and the uniqueness under each type is obtained through analysis after the landslide occurs.

[0156] In this embodiment, N1 is greater than or equal to 10.

[0157] In this embodiment, time - series association means that as long as the corresponding single sample is included in the continuous time - series order.

[0158] In this embodiment, = the monitoring value of the corresponding single sample in the corresponding dimension - the monitoring value of the left sample of the single sample in the corresponding dimension, = the monitoring value of the corresponding single sample in the corresponding dimension - the monitoring value of the right sample of the single sample in the corresponding dimension.

[0159] In this embodiment, takes a value of 0.5. takes a value of 0.4.

[0160] If , and , at this time, it is determined that the importance coefficient of the corresponding single sample in the corresponding monitoring dimension is 1;

[0161] If , and , at this time, it is determined that the importance coefficient of the corresponding single sample in the corresponding monitoring dimension is 0.5;

[0162] If , and , at this time, it is determined that the importance coefficient of the corresponding single sample in the corresponding monitoring dimension is 0.5.

[0163] The beneficial effects of the above - mentioned technical solution are as follows: By obtaining samples under continuous time series to determine the importance coefficients in the case of one - side difference and two - side differences of a single sample, and combining with the mapping relationship and historical displacement to determine the difference types in different dimensions, it provides convenience for obtaining the optimization vector.

[0164] The present invention provides a dynamic monitoring method for landslide disasters based on time series and hybrid models. Based on the first matrix, the difference factors based on each monitoring dimension in the same disaster type determined historically are obtained, including:

[0165] If all the verification types involved in the first matrix are consistent with the corresponding disaster types, at this time, the absolute value of the difference between the first disaster coefficient and the first verification coefficient in each column vector is calculated respectively, and the coefficient variance is obtained and combined with the corresponding disaster type to obtain the difference factor;

[0166] If there are types among all the verification types involved in the first matrix that are inconsistent with the corresponding disaster types, the occurrence times of each inconsistent type are extracted respectively, and the absolute value of the difference between all the first disaster coefficients and the first verification coefficients under the inconsistent type is combined to obtain the current factor of each inconsistent type, and then the difference factor is formed.

[0167] In this embodiment, if the row vectors corresponding to the verification types in the first matrix are all consistent with the disaster types, at this time, the difference factor: coefficient variance - disaster type.

[0168] In this embodiment, if they are different, at this time, the average value of the absolute value of the difference under the inconsistent type / the corresponding occurrence times is obtained, and then combined with the inconsistent type to obtain the difference factor.

[0169] The beneficial effects of the above technical solutions are: By means of type comparison and analysis, the difference factors in the cases of consistent types and inconsistent types are determined, providing a basis for constructing an optimized vector.

[0170] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A landslide disaster dynamic monitoring method based on time series and hybrid model, characterized in that, Including: Step 1: Collect multi-source heterogeneous data of historical landslides, and preprocess the multi-source heterogeneous data to obtain first data, where the multi-source heterogeneous data involves multiple monitoring dimensions; Step 2: Perform type division on the first data, and through time series encoding for each type of data, obtain a single sample at the same time series and then obtain a sample set, and then train a neural network model according to the sample set to obtain an initial model, where the single sample includes monitoring values under different monitoring dimensions and landslide horizontal displacement; Step 3: Perform model verification on the initial model, and combine the constructed mapping relationship between landslide deformation and monitoring dimensions of the corresponding type to optimize and obtain a time series and hybrid model; Step 4: Monitor and obtain the current heterogeneous data of the target monitoring point, and input it into the time series and hybrid model to predict the current disaster result of the target monitoring point and output a reminder; Among them, optimizing to obtain a time series and hybrid model includes: Extract the first multi-source data and the second source data at two adjacent time moments from the sample set, where the first multi-source data corresponds to the first historical moment, the second multi-source data corresponds to the second historical moment, and the first historical moment is before the second historical moment; Input the first multi-source data into the initial model to obtain a verification type and first verification coefficients under different monitoring dimensions; According to the historical displacement of the corresponding disaster type and the mapping relationship, respectively determine the first disaster coefficients of each monitoring dimension and the corresponding disaster type in the same single sample; Obtain all the first disaster coefficients of different monitoring dimensions involved in the same disaster type, and construct a first coefficient vector for each monitoring dimension in the same disaster type; Extract the first verification coefficients and verification type of the first multi-source data corresponding to the first coefficient vector to obtain a verification coefficient vector and a verification type vector, and combine the first coefficient vector to obtain a first matrix; Obtain the difference factors based on each monitoring dimension in the same disaster type determined by historical reality based on the first matrix; Rely on the first historical difference between the second multi-source data and the first multi-source data, the type difference between the historical actual disaster type under the second multi-source data and the verification type of the first multi-source data, and combine the difference factors based on each monitoring dimension of the corresponding historical actual disaster type to construct an optimization vector corresponding to the first historical moment; Optimize the initial model based on all the optimization vectors to obtain a time series and hybrid model.

2. The landslide disaster dynamic monitoring method based on time series and hybrid model according to claim 1, characterized in that, Preprocessing the multi-source heterogeneous data to obtain first data includes: Perform type division on the first data according to the monitoring dimension, and draw a curve smoothing for each first value in each type of data in chronological order; Respectively determine the first absolute slope of the straight line formed by two adjacent points in the smoothed curve, and adjust the first absolute slope in combination with the curve area formed by the two adjacent points to obtain a second absolute slope; ； wherein, D2 is the corresponding second absolute slope; D1 is the corresponding first absolute slope; represents the value of the highest point in the curve segment formed by the i-th point and the (i + 1)-th point; represents the value of the corresponding point that coincides with the highest point in the straight line formed by the i-th point and the (i + 1)-th point; represents the value of the lowest point in the curve segment formed by the i-th point and the (i + 1)-th point; represents the value of the corresponding point that coincides with the lowest point in the straight line formed by the i-th point and the (i + 1)-th point; represents the closed area of the curve of the curve segment formed by the i-th point and the (i + 1)-th point; represents the closed area of the straight line formed by the i-th point and the (i + 1)-th point; n represents the total number of the existing first values and is consistent with the number of points for drawing the smooth curve; Compare the first absolute slope and the second absolute slope at the same adjacent two points with a preset slope respectively. If both are less than or equal to the preset slope, perform a first adjustment on the subsequent point of the corresponding adjacent two points according to max(D1, D2), where max represents the maximum symbol; If both are greater than or equal to the preset slope, perform a second adjustment on the subsequent point of the corresponding adjacent two points according to min(D1, D2), where min represents the minimum symbol; If there is a situation where one of D1 and D2 is greater than the preset slope and the other is less than the preset slope, at this time, randomly select a final slope according to rand((D1, D2), 1) to perform a third adjustment on the subsequent point of the corresponding adjacent two points, where rand represents the random symbol; Obtain the first data based on all adjustment results.

3. The landslide disaster dynamic monitoring method based on time series and hybrid model according to claim 2, wherein Performing a first adjustment on the subsequent point of the corresponding adjacent two points according to max(D1, D2) includes: ； Among them, represents the value after the first adjustment to the subsequent point; represents the value before the subsequent point is not adjusted; represents the variance of D1 - D2 based on all adjacent two points; represents the positive or negative situation of the slope between the corresponding i-th point and the (i + 1)-th point.

4. The landslide disaster dynamic monitoring method based on time series and hybrid model according to claim 3, characterized in that Performing a second adjustment on the subsequent point of the corresponding adjacent two points according to min(D1, D2) includes: ； Among them, represents the value after the second adjustment to the subsequent point.

5. The landslide disaster dynamic monitoring method based on time series and hybrid model according to claim 3, characterized in that, Randomly selecting a final slope according to rand((D1, D2), 1) to perform a third adjustment on the subsequent point of the corresponding adjacent two points includes: ； ； Among them, represents the value after the third adjustment to the next point; D3 represents the adjustment function for the corresponding adjacent two points; rand((D1, D2), 1) represents the function of randomly selecting a final slope from D1 and D2 of the corresponding adjacent two points; represents the preset slope.

6. The landslide disaster dynamic monitoring method based on time series and hybrid model according to claim 1, characterized in that According to the historical displacements of the corresponding disaster types and the mapping relationships, respectively determine the first disaster coefficients of each monitoring dimension in the same single sample corresponding to the disaster types, including: Respectively obtain the historical displacements of different disaster types from the time series - type - displacement comparison table; Determine N1 first samples that have a time series association with the single sample. Sort the monitoring values under each monitoring dimension in the N1 + 1 samples based on the time series order, and determine whether there is a difference between both sides of the corresponding single sample; If there is a difference between the two sides, then according to and the magnitude relationship, as well as and the magnitude relationship, determine the importance coefficient of the corresponding single sample under the corresponding monitoring dimension. Among them, represents the left-side difference of the corresponding single sample; represents the right-side difference of the corresponding single sample; represents the difference threshold under the corresponding monitoring dimension; represents the monitoring value of the corresponding single sample under the corresponding monitoring dimension; and respectively represent the left-side monitoring value and the right-side monitoring value based on the corresponding single sample under the corresponding monitoring dimension; and respectively represent constants; If there is only a difference on one side, then according to determine the importance coefficient of the corresponding single sample under the corresponding monitoring dimension, where represents the average value of the monitoring values corresponding to N1 + 1 samples under the corresponding monitoring dimension; represents the maximum value among the monitoring values corresponding to N1 + 1 samples; represents the variance of the monitoring values corresponding to N1 + 1 samples, where , determine that the corresponding importance coefficient is 0; otherwise, determine that the corresponding importance coefficient is ; Based on the mapping relationships of the corresponding types and all the historical displacements of the corresponding types, determine the set weights of each monitoring dimension in the corresponding disaster types, and combine the importance coefficients to determine the first disaster coefficients of the corresponding monitoring dimensions in the corresponding disaster types; ； Among them, \(U_i\) represents the first disaster coefficient of the \(i\)-th monitoring dimension for the corresponding disaster type; represents the historical disaster value for the corresponding disaster type; represents the set weight of the \(i\)-th monitoring dimension in the corresponding disaster type; represents the importance coefficient of the \(i\)-th monitoring dimension in the corresponding disaster type; represents the sum of the set weights of all monitoring dimensions under the corresponding disaster type, with a value of 1; represents the sum of the importance coefficients of all monitoring dimensions under the corresponding disaster type; \(m\) represents the number of monitoring dimensions.

7. The dynamic landslide disaster monitoring method based on time series and hybrid model according to claim 6, characterized in that Obtain the difference factors based on each monitoring dimension in the same disaster type determined by the historical actual situation based on the first matrix, including: If all the verification types involved in the first matrix are consistent with the corresponding disaster types, at this time, respectively calculate the absolute value of the difference between the first disaster coefficient and the first verification coefficient in each column vector, and obtain the coefficient variance and combine it with the corresponding disaster type to obtain the difference factor; If there are types among all the verification types involved in the first matrix that are inconsistent with the corresponding disaster types, then respectively extract the occurrence times of each inconsistent type, and combine the absolute value of the difference between all the first disaster coefficients and the first verification coefficients under the inconsistent types to obtain the current factors of each inconsistent type and thus form the difference factor.

Images