Landslide displacement prediction method and system based on displacement vector angle and LSTM

Abstract

The present application discloses a landslide displacement prediction method and system based on displacement vector angle and LSTM, which relates to the field of landslide displacement monitoring. The method includes: obtaining historical monitoring data of a target reservoir; determining effective factors affecting landslide displacement; based on historical data of effective factors and landslide displacement at different times in historical monitoring data, respectively training and optimizing a first prediction model and a second prediction model under different sliding window values, and determining their respective corresponding optimal sliding window values; when the optimal sliding window values ​​of the first prediction model and the second prediction model are the same, respectively calculating the evaluation index values ​​of the first prediction model and the second prediction model after training and optimization under the optimal sliding window value, and determining the landslide displacement prediction model; collecting the current effective factor data of the target reservoir and inputting it into the landslide displacement prediction model, and determining the future landslide displacement of the target reservoir. The present application can accurately predict the future landslide displacement of the reservoir.

Description

Technical Field

[0001] The present application relates to the technical field of landslide displacement monitoring, and in particular to a landslide displacement prediction method and system based on displacement vector angle and LSTM. Background Art

[0002] Landslide is one of the common geological disasters, which is highly destructive, difficult to control, and has complex and diverse inducing factors. Among all types of landslides, reservoir landslide is the most harmful one. Reservoir landslide refers to a special type of landslide in which the accumulation layer located on the bank of the reservoir or in the reservoir area slides under the action of the reservoir water storage. Reservoir landslide is very easy to displace and deform due to its large porosity, which will not only cause direct damage to the surrounding area of ​​the reservoir, but also cause secondary disasters such as floods and mudslides, resulting in serious casualties and property losses. Therefore, it is of great significance to accurately predict the displacement of reservoir landslide.

[0003] Reservoir-type landslides are caused by the coupling of multiple factors, the main reasons of which include: the reservoir water storage increases the groundwater level in the reservoir area, resulting in an increase in the pore water pressure inside the accumulation layer and a decrease in the effective stress, thereby weakening the shear strength of the accumulation layer and making it prone to slip; when the reservoir is filled with water, the rising water level causes the slope angles of the reservoir bank to be scoured and eroded, causing the slope angles to expand and increasing the probability of landslides; after the reservoir is filled with water, the soil below the accumulation layer is affected by water pressure, and the self-weight stress changes, resulting in changes in its stability; external forces such as earthquakes and rainfall may also cause the accumulation layer to slip, exacerbating the occurrence of landslides, etc.

[0004] The existing prediction methods for reservoir-type landslide displacement mainly include physical and mechanical models, statistical models, numerical simulation and machine learning models. Most of these models only consider traditional influencing factors such as rainfall, reservoir water level, and cumulative displacement, and do not consider other evaluation factors for landslide stability. Therefore, the final prediction results of the model are inaccurate, which is not conducive to the monitoring and early warning of reservoir-type landslide displacement. Summary of the invention

[0005] The purpose of this application is to provide a landslide displacement prediction method and system based on displacement vector angle and LSTM, which can accurately predict the future landslide displacement of the reservoir and facilitate effective monitoring and early warning of reservoir-type landslide displacement.

[0006] To achieve the above objectives, this application provides the following solutions.

[0007] In the first aspect, the present application provides a landslide displacement prediction method based on displacement vector angle and LSTM, and the landslide displacement prediction method based on displacement vector angle and LSTM includes: obtaining historical monitoring data of a target reservoir; the historical monitoring data at least includes: reservoir water level and landslide displacement at different historical moments; performing grey correlation analysis on the historical monitoring data to determine effective factors affecting landslide displacement; the effective factors at least include: displacement rate, cumulative displacement and displacement vector angle; screening the historical data of the effective factors according to the historical monitoring data; creating a first prediction model and a second prediction model based on an LSTM neural network respectively; the sliding window in the second prediction model includes historical landslide displacement; the sliding window in the first prediction model does not include historical landslide displacement; taking the historical data of the effective factor as input and the landslide displacement at the i+1th moment in the historical monitoring data as output, The first prediction model under different sliding window values ​​is trained and optimized respectively to determine the optimal sliding window value of the first prediction model; the historical data of the effective factor and the landslide displacement at the i-th moment in the historical monitoring data are used as input, and the landslide displacement at the i+1-th moment in the historical monitoring data is used as output, and the second prediction model under different sliding window values ​​is trained and optimized respectively to determine the optimal sliding window value of the second prediction model; when the optimal sliding window values ​​of the first prediction model and the second prediction model are the same, under the optimal sliding window value, the evaluation index values ​​of the first prediction model and the second prediction model after training and optimization are calculated respectively, and the final landslide displacement prediction model is determined according to the evaluation index values; the evaluation index values ​​at least include the absolute percentage error value; the current effective factor data of the target reservoir is collected and input into the landslide displacement prediction model to determine the future landslide displacement of the target reservoir.

[0008] In a second aspect, the present application also provides a computer system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the landslide displacement prediction method based on displacement vector angle and LSTM.

[0009] According to the specific embodiments provided in this application, this application discloses the following technical effects.

[0010] Compared with the existing models, in addition to considering the traditional factors that affect the landslide displacement, such as reservoir water level, rainfall, displacement rate, etc., this application also introduces the displacement vector angle as a supplementary factor, fully considering its influence on the stability of the landslide. At the same time, in order to further deepen the close relationship between each influencing factor and the landslide displacement, the grey correlation analysis method is also used to calculate the correlation degree between all influencing factors and the landslide displacement, and the influencing factors with a higher correlation degree are selected as effective factors to input into the model. In addition, this application also establishes two prediction models, one with a sliding window containing historical landslide displacement and one without, and compares and analyzes whether the sliding window contains historical landslide displacement, as well as the influence of different sliding window values ​​on the landslide displacement prediction results. Finally, the best model prediction method is selected to achieve accurate prediction of landslide displacement, which is more conducive to the monitoring and early warning of reservoir-type landslide displacement. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0012] Figure 1 A flow chart of a landslide displacement prediction method based on displacement vector angle and LSTM provided in an embodiment of the present application.

[0013] Figure 2 This is a graph of landslide displacement prediction results for Model 1 provided in an embodiment of the present application.

[0014] Figure 3 This is a graph of landslide displacement prediction results of model 2 provided in an embodiment of the present application.

[0015] Figure 4 This is a graph of landslide displacement prediction results of different models corresponding to the sliding window 2 provided in an embodiment of the present application.

[0016] Figure 5 This is a graph of landslide displacement prediction results for the test set provided in an embodiment of the present application.

[0017] Figure 6 This is a diagram of the internal structure of a computer system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] The purpose of this application is to provide a landslide displacement prediction method and system based on displacement vector angle and LSTM, which can accurately predict the future landslide displacement of the reservoir and facilitate effective monitoring and early warning of reservoir-type landslide displacement.

[0020] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0021] Example 1

[0022] This embodiment provides a landslide displacement prediction method based on displacement vector angle and LSTM. Figure 1 As shown, the landslide displacement prediction method based on displacement vector angle and LSTM is as follows.

[0023] Step S1: Obtain historical monitoring data of the target reservoir, wherein the historical monitoring data mainly include: reservoir water level, rainfall, displacement rate, displacement vector angle, cumulative displacement and landslide displacement, etc. In addition, temperature, groundwater level and geological structure also have a certain influence on landslide displacement.

[0024] In this embodiment, the reservoir water level, rainfall and landslide displacement are directly obtained, generally by using high-precision GNSS, satellite remote sensing or manual measurement. The reservoir water level is generally taken as the monthly average reservoir water level, the rainfall is taken as the monthly cumulative rainfall, and the landslide displacement is collected once a month. As for the displacement rate, displacement vector angle and cumulative displacement, they need to be obtained through further calculation.

[0025] Specifically, according to the formula Calculate displacement velocity ; In the formula, , They are i and i The horizontal displacement at time -1, , They are i and i The vertical displacement at time -1, t for i and i -1 moment in time.

[0026] According to the formula Calculate the displacement vector angle ; In the formula, is the vertical displacement, is the horizontal displacement.

[0027] In addition, n The cumulative displacement can be obtained by summing up the displacements of the previous month.

[0028] Step S2: Perform grey correlation analysis on historical monitoring data to determine the effective factors affecting landslide displacement.

[0029] In this embodiment, step S2 is specifically as follows.

[0030] The first step is to classify the historical monitoring data and determine the multiple influencing factors of landslide displacement.

[0031] In the second step, the grey correlation analysis method is used to calculate the correlation between each influencing factor and landslide displacement, and the grey correlation coefficients of different influencing factors are obtained.

[0032] The third step is to determine the influencing factors whose grey correlation coefficient is greater than or equal to the set threshold as effective factors. The set threshold in this embodiment is 0.5, that is, the influencing factors whose grey correlation coefficient is greater than or equal to 0.5 are determined as effective factors. The effective factors finally determined in this embodiment include: reservoir water level, rainfall, displacement rate, cumulative displacement and displacement vector angle.

[0033] Step S3: Filter historical data of effective factors according to historical monitoring data.

[0034] In this embodiment, since the historical monitoring data also contain other influencing factors of landslide displacement, it is necessary to filter out the historical data corresponding to the effective factors from the historical monitoring data based on the effective factors, and divide the filtered historical data into training set, test set and validation set in proportion to facilitate subsequent model training and optimization.

[0035] Step S4: creating a first prediction model and a second prediction model based on the LSTM neural network respectively; the sliding window in the second prediction model includes the historical landslide displacement; the sliding window in the first prediction model does not include the historical landslide displacement.

[0036] In this embodiment, step S4 is specifically as follows.

[0037] The first step is to build two basic models with the same LSTM neural network structure.

[0038] In the second step, a base model is used as the first prediction model.

[0039] In the third step, the historical landslide displacement is added to the sliding window of another basic model to obtain the second prediction model.

[0040] As a preferred implementation, whether the sliding window in the LSTM neural network contains historical landslide displacements affects the accuracy of the model prediction results to a certain extent. When the sliding window contains historical landslide displacements, it can provide more complete context information for the prediction of future landslide displacements. The LSTM neural network captures long-term dependencies through memory units and gating mechanisms, allowing the model to better understand long-term dependencies in time series, thereby improving the prediction accuracy of the model.

[0041] Step S5: Taking the historical data of effective factors as input and the landslide displacement at the i+1th moment in the historical monitoring data as output, the first prediction model under different sliding window values ​​(i.e., sliding windows of different sizes) is trained and optimized to determine the optimal sliding window value of the first prediction model; wherein, i>0.

[0042] In this embodiment, step S5 is specifically as follows.

[0043] In the first step, the sliding window value (the size of the sliding window) of the first prediction model is set to 1, 2, and 3, respectively, that is, one prediction is made every 1 test, one prediction is made every 2 tests, and one prediction is made every 3 tests.

[0044] In the second step, based on the historical data of effective factors and the landslide displacement at the i+1th moment in the historical monitoring data, the first prediction model under different sliding window values ​​is trained and optimized, and the absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value of the first prediction model under different sliding window values ​​are calculated.

[0045] The third step is to use the sliding window value corresponding to the minimum absolute percentage error value, mean square error value and root mean square error value and the maximum correlation coefficient value as the optimal sliding window value (optimal sliding window size) of the first prediction model.

[0046] Step S6: Taking the historical data of the effective factor and the landslide displacement at the i-th moment in the historical monitoring data as input, and taking the landslide displacement at the i+1-th moment in the historical monitoring data as output, respectively train and optimize the second prediction model under different sliding window values ​​(i.e., setting sliding windows of different sizes), and determine the optimal sliding window value of the second prediction model. Among them, because the sliding window in the second prediction model includes the historical landslide displacement, when training and optimizing the second prediction model, in addition to inputting the historical data of the effective factor, the input of the landslide displacement at the i-th moment should also be added.

[0047] In this embodiment, step S6 is specifically as follows.

[0048] In the first step, the sliding window value (the size of the sliding window) of the second prediction model is set to 1, 2, and 3, respectively, that is, one prediction is made for every 1 test, one prediction is made for every 2 tests, and one prediction is made for every 3 tests.

[0049] In the second step, based on the historical data of effective factors and the landslide displacement at the i-th moment and the i+1-th moment in the historical monitoring data, the second prediction model under different sliding window values ​​is trained and optimized, and the absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value of the first prediction model under different sliding window values ​​are calculated.

[0050] The third step is to use the sliding window value corresponding to the minimum absolute percentage error value, mean square error value and root mean square error value and the maximum correlation coefficient value as the optimal sliding window value (optimal sliding window size) of the second prediction model.

[0051] Step S7: When the optimal sliding window values ​​of the first prediction model and the second prediction model are the same, under the optimal sliding window value, the evaluation index values ​​of the first prediction model and the second prediction model after training and optimization are calculated respectively, and the final landslide displacement prediction model is determined according to the evaluation index values. The evaluation index values ​​include: absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value.

[0052] In this embodiment, step S7 is specifically as follows.

[0053] In the first step, the sliding window values ​​of the first prediction model and the second prediction model are both set to the optimal sliding window values.

[0054] In the second step, the first prediction model and the second prediction model are trained and optimized respectively under the optimal sliding window value.

[0055] The third step is to calculate the absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value of the first prediction model and the second prediction model after training and optimization.

[0056] In the fourth step, the prediction model with the smallest absolute percentage error value, mean square error value and root mean square error value and the largest correlation coefficient value is determined as the landslide displacement prediction model. Among them, the sliding window in the landslide displacement prediction model finally determined contains the historical landslide displacement, and the sliding window value of the landslide displacement prediction model is the optimal sliding window value (that is, the size of the sliding window in the landslide displacement prediction model is set to the above-mentioned optimal sliding window size).

[0057] Step S8: Collect the current effective factor data of the target reservoir and input it into the landslide displacement prediction model to determine the future landslide displacement of the target reservoir.

[0058] Based on the above analysis, the landslide displacement prediction method based on displacement vector angle and LSTM provided in this embodiment has a high accuracy in predicting reservoir-type landslide displacement, can dynamically predict future changes in landslide displacement in real time, and can provide efficient and accurate technical support for landslide support and geological disaster prevention and control.

[0059] Example 2

[0060] This embodiment provides a practical application scenario of a landslide displacement prediction method based on displacement vector angle and LSTM, taking the Bazimen landslide as an example, the details are as follows.

[0061] The Bajimen landslide is located near the dam in the Three Gorges Reservoir area. The landslide is high in the west and low in the east, in a stepped shape, with an elevation of 55~280m. The total length of the landslide is 520m, with a low slope in the front and high slope in the back. The front edge is 20°~25°, and the middle and rear parts are 35°~40°. The upper part of the slope is 100~110m wide, and the lower part is 350m wide. The overall shape is a dustpan and extends to the river. The front edge of 55~156m is submerged by water. The main sliding direction of the landslide is NW112°, the front edge is facing the Xiangxi River, the shear outlet elevation is 65m, the total area is about 117,800 m², and the total volume is about 2 million m³.

[0062] There are 10 deformation monitoring points in Bajimen landslide, namely ZG110, ZG111, GSC1~GSC4, GSC5, GSC7~GSC9. The ZG series monitoring points were deployed and used first, with ZG110 and ZG111 located at the front and middle of the landslide respectively; the GSC series monitoring points are widely distributed, covering the back to the front of the landslide, and the north to the south.

[0063] (1) Landslide displacement data collection and processing.

[0064] The measured data of monitoring point ZG110 were taken as the research object. The monitoring period was from January 2006 to December 2012, and the data were collected once a month, with a total of 84 measurement data. The reservoir water level adopted the monthly average reservoir water level, and the rainfall adopted the monthly cumulative rainfall. The monthly displacement rate, displacement vector angle and cumulative displacement were calculated, and the landslide displacement was collected once a month.

[0065] Matlab program is used to realize data screening and calculation.

[0066] (2) Determine the number of input features.

[0067] Reservoir water level and rainfall are two important influencing factors that induce displacement of reservoir-type accumulation layer landslides. When the landslide displacement changes, its displacement rate, cumulative displacement, displacement vector angle, etc. will change significantly. These variables can more accurately describe the change law of the slope in different stable stages.

[0068] In order to verify the rationality of the selection of the above influencing factors, the correlation between the landslide displacement and each influencing factor must be considered. Specifically, the grey correlation analysis method is used to analyze the above influencing factors and landslide displacement, and the calculation results are shown in Table 1. The correlation between each influencing factor and landslide displacement is: cumulative displacement> rainfall> displacement vector angle> reservoir water level> displacement rate. The calculation results of the correlation between each influencing factor and landslide displacement are all close to 1. Therefore, it is determined that the cumulative displacement, rainfall, displacement vector angle, reservoir water level and displacement rate are used as effective factors of landslide displacement.

[0069] Table 1 Correlation analysis between landslide displacement and various influencing factors

[0070]

[0071] (3) Set the sliding mode and sliding window.

[0072] Considering that the change of landslide displacement is affected by the displacement of the previous landslide, two prediction models are established, one with the sliding window excluding historical landslide displacement and the other with the sliding window including historical landslide displacement, named Model 1 and Model 2, to analyze the impact of whether the sliding window includes historical landslide displacement on the prediction results.

[0073] In order to determine the impact of the sliding window size on the prediction results, three sliding windows of different sizes are constructed, namely: predicting one test for every 1 test, predicting one test for every 2 tests, and predicting one test for every 3 tests. The training distribution under different models and sliding windows is shown in Table 2.

[0074] Table 2 Landslide displacement prediction methods and sliding window size distribution

[0075]

[0076] Taking reservoir water level, rainfall, displacement rate, cumulative displacement and displacement vector angle as model input and landslide displacement at the i+1th moment as model output, model 1 with different sliding window sizes is trained and optimized. Finally, the prediction results of model 1 after training and optimization are shown in Figure 2 shown.

[0077] according to Figure 2The results in Table 3 show that during the training of the three sliding windows, the absolute percentage error (MAPE), mean square error (MSE) and root mean square error (RMSE) of sliding window 2 are all minimized, and the correlation coefficient R2 reaches a maximum of 0.92. Sliding window 1 may have fewer training tests, resulting in the model only focusing on local information in the time series and lacking a global perspective. It may not be able to capture long-term dependencies and overall trends in the sequence. In addition, overfitting of model 1 may reduce generalization ability, which may also be one of the reasons. Sliding window 3 may have too many training tests, resulting in model 1 being unable to capture local patterns and short-term trends in the time series, and training information is lost, thus affecting prediction accuracy.

[0078] Table 3 Landslide displacement prediction evaluation index values ​​for model 1

[0079]

[0080] The reservoir water level, rainfall, displacement rate, cumulative displacement, displacement vector angle and landslide displacement at the i-th moment are used as model inputs, and the landslide displacement at the i+1-th moment is used as model output. Model 2 with different sliding window sizes is trained and optimized. Finally, the prediction results of the trained and optimized model 2 are shown in Figure 3 shown.

[0081] according to Figure 3 The results in Table 4 show that the absolute percentage error, mean square error, and root mean square error of sliding window 2 are all minimized, and the correlation coefficient R2 reaches a maximum of 0.94. Sliding window 1 may have fewer training tests, resulting in model 2 only focusing on local information in the time series, lacking a global perspective, and may not be able to capture long-term dependencies and overall trends in the series. Sliding window 3 may have too many training tests, resulting in model 2 being unable to capture local patterns and short-term trends in the time series, and training information is lost, thus affecting prediction accuracy.

[0082] Table 4 Landslide displacement prediction evaluation index values ​​for model 2

[0083]

[0084] The above analysis shows that regardless of model 1 or model 2, sliding window 2 shows higher prediction accuracy. Figure 4The results in Table 5 show that the absolute percentage error, mean square error and root mean square error corresponding to the prediction results when the sliding window includes the historical landslide displacement are smaller, and the correlation coefficient R2 is larger. When the sliding window includes the historical landslide displacement, it can provide more complete context information of the landslide displacement. The LSTM neural network captures long-term dependencies through memory units and gating mechanisms, allowing the model to better understand the long-term dependencies in the time series, thereby improving the accuracy of the prediction. It can be seen that the prediction result of model 2 when the sliding window is set to 2 is the optimal prediction result.

[0085] Table 5 Displacement prediction evaluation index values ​​corresponding to different models of sliding window 2

[0086]

[0087] (4) Validation of prediction model.

[0088] According to the above landslide displacement prediction method based on displacement vector angle and LSTM, the model 2 rolling prediction test set data corresponding to sliding window 2 is used, and the calculation results are shown in Figure 5 .according to Figure 5 It can be seen that the prediction results under the model conditions are close to the actual landslide displacement and the prediction accuracy is high, which verifies the effectiveness of the landslide displacement prediction method based on displacement vector angle and LSTM in this application.

[0089] Example 3

[0090] This embodiment provides a computer system, which may be a server or a terminal, and its internal structure diagram may be as follows: Figure 6 As shown. The computer system includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer system is used to provide computing and control capabilities. The memory of the computer system includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer system is used to store video tag processing data. The input / output interface of the computer system is used to exchange information between the processor and an external device. The communication interface of the computer system is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a landslide displacement prediction method based on a displacement vector angle and LSTM is implemented.

[0091] Those skilled in the art will understand that Figure 6The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer system to which the scheme of the present application is applied. The specific computer system may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0093] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0094] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0095] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A landslide displacement prediction method based on displacement vector angle and LSTM, characterized in that: The landslide displacement prediction method based on displacement vector angle and LSTM includes: Acquiring historical monitoring data of the target reservoir; the historical monitoring data at least includes: reservoir water level and landslide displacement at different historical moments; Performing grey correlation analysis on the historical monitoring data to determine effective factors affecting landslide displacement; the effective factors include at least displacement rate, cumulative displacement and displacement vector angle; Filtering the historical data of the effective factor according to the historical monitoring data; Creating a first prediction model and a second prediction model based on the LSTM neural network; wherein the sliding window in the second prediction model includes historical landslide displacements; and the sliding window in the first prediction model does not include historical landslide displacements; Taking the historical data of the effective factor as input and the landslide displacement at the i+1th moment in the historical monitoring data as output, respectively training and optimizing the first prediction model under different sliding window values ​​to determine the optimal sliding window value of the first prediction model; Taking the historical data of the effective factor and the landslide displacement at the i-th moment in the historical monitoring data as input, and taking the landslide displacement at the i+1-th moment in the historical monitoring data as output, respectively training and optimizing the second prediction model under different sliding window values, and determining the optimal sliding window value of the second prediction model; When the optimal sliding window values ​​of the first prediction model and the second prediction model are the same, under the optimal sliding window values, respectively calculating evaluation index values ​​of the first prediction model and the second prediction model after training and optimization, and determining a final landslide displacement prediction model according to the evaluation index values; the evaluation index values ​​at least include an absolute percentage error value; Collecting current effective factor data of the target reservoir and inputting the data into the landslide displacement prediction model to determine the future landslide displacement of the target reservoir; The first prediction model and the second prediction model are created based on the LSTM neural network, specifically including: Build two basic models with the same LSTM neural network structure; Using one of the basic models as a first prediction model; The historical landslide displacement is added to the sliding window of another basic model to obtain a second prediction model.

2. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 1 is characterized in that: The sliding window in the landslide displacement prediction model includes historical landslide displacements, and the sliding window value of the landslide displacement prediction model is the optimal sliding window value.

3. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 1 is characterized in that: Grey correlation analysis was performed on the historical monitoring data to determine the effective factors affecting landslide displacement, including: Classifying the historical monitoring data into categories and determining multiple influencing factors of landslide displacement; The grey correlation analysis method is used to calculate the correlation between each influencing factor and landslide displacement, and the grey correlation coefficients of different influencing factors are obtained. The influencing factors whose grey relational coefficient is greater than or equal to a set threshold are determined as effective factors.

4. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 3 is characterized in that: The threshold value is set to 0.

5.

5. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 1 is characterized in that: The evaluation index values ​​also include mean square error value, root mean square error value and correlation coefficient value.

6. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 5 is characterized in that: Taking the historical data of the effective factor as input and the landslide displacement at the i+1th moment in the historical monitoring data as output, respectively training and optimizing the first prediction model under different sliding window values ​​to determine the optimal sliding window value of the first prediction model, specifically including: The sliding window values ​​of the first prediction model are set to 1, 2 and 3 respectively; Based on the historical data of the effective factor and the landslide displacement at the i+1th moment in the historical monitoring data, respectively, training and optimizing the first prediction model under different sliding window values, and calculating the absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value of the first prediction model under different sliding window values; The sliding window value corresponding to the time when the absolute percentage error value, the mean square error value, and the root mean square error value are minimized and the correlation coefficient value is maximized is taken as the optimal sliding window value of the first prediction model.

7. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 5 is characterized in that: Taking the historical data of the effective factor and the landslide displacement at the i-th moment in the historical monitoring data as input, and taking the landslide displacement at the i+1-th moment in the historical monitoring data as output, respectively training and optimizing the second prediction model under different sliding window values, and determining the optimal sliding window value of the second prediction model, specifically comprising: The sliding window values ​​of the second prediction model are set to 1, 2 and 3 respectively; Based on the historical data of the effective factor and the landslide displacement at the i-th moment and the i+1-th moment in the historical monitoring data, respectively, training and optimizing the second prediction model under different sliding window values, and calculating the absolute percentage error value, mean square error value, root mean square error value and correlation coefficient value of the second prediction model under different sliding window values; The sliding window value corresponding to the time when the absolute percentage error value, the mean square error value and the root mean square error value are minimized and the correlation coefficient value is maximized is used as the optimal sliding window value of the second prediction model.

8. The landslide displacement prediction method based on displacement vector angle and LSTM according to claim 5 is characterized in that: When the optimal sliding window values ​​of the first prediction model and the second prediction model are the same, under the optimal sliding window values, respectively calculating evaluation index values ​​of the first prediction model and the second prediction model after training and optimization, and determining a final landslide displacement prediction model according to the evaluation index values, specifically including: Setting the sliding window values ​​of the first prediction model and the second prediction model to the optimal sliding window value; Under the optimal sliding window value, respectively training and optimizing the first prediction model and the second prediction model; Calculate the absolute percentage error, mean square error, root mean square error and correlation coefficient of the first prediction model and the second prediction model after training and optimization respectively; The prediction model with the smallest absolute percentage error value, the mean square error value and the root mean square error value and the largest correlation coefficient value is determined as the landslide displacement prediction model.

9. A computer system comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the landslide displacement prediction method based on displacement vector angle and LSTM according to any one of claims 1 to 8.

Images