A landslide displacement prediction method and device based on CEEMDAN-NBEATS

Abstract

This application relates to the field of landslide displacement data prediction technology, and provides a landslide displacement prediction method and apparatus based on CEEMDAN-NBEATS. The method includes: acquiring raw landslide displacement data; processing the raw landslide displacement data based on the CEEMDAN algorithm to obtain a landslide displacement time series; establishing an NBEATS model and optimizing the NBEATS model based on the landslide displacement time series; predicting landslide displacement based on the NBEATS model to obtain the prediction result; and performing inverse mean normalization on the prediction result of the NBEATS model to obtain the predicted landslide displacement value. This application enables rapid prediction with limited data, improving the accuracy and stability of the prediction.

Description

Technical Field

[0001] This application belongs to the field of landslide displacement data prediction technology, and in particular relates to a landslide displacement prediction method and device based on CEEMDAN-NBEATS. Background Technology

[0002] Currently, deep learning (DL) has been very successful in predicting landslide displacement data. Classical deep learning time series methods are mainly based on LSTM, SVM, BP, SVR, and GRU neural networks. These are prediction algorithms that uniformly model large amounts of related time series, relying heavily on large-scale landslide datasets because they train autoregressive recurrent network models on a large number of time series. However, in practice, due to limitations imposed by the natural environment or equipment conditions, we often obtain small-sample time series. Under these conditions, although classical models have good fitting properties for nonlinear data sequences, traditional fractal models also suffer from drawbacks such as low prediction accuracy, poor prediction stability, slow prediction speed, and over-reliance on data and computing power when dealing with small-sample data sequences. Summary of the Invention

[0003] To overcome the problems existing in related technologies, this application provides a landslide displacement prediction method and device based on CEEMDAN-NBEATS, which can solve the problems of low prediction accuracy and poor prediction stability of traditional fractal models under small sample data sequence conditions.

[0004] This application is achieved through the following technical solution:

[0005] In a first aspect, embodiments of this application provide a landslide displacement prediction method based on CEEMDAN-NBEATS, including: acquiring raw landslide displacement data;

[0006] The original landslide displacement data was processed using the CEEMDAN algorithm to obtain the landslide displacement time series.

[0007] Establish the NBEATS model and optimize it based on landslide displacement time series.

[0008] Landslide displacement prediction was performed based on the NBEATS model, and the prediction results were obtained.

[0009] The prediction results of the NBEATS model are subjected to inverse mean normalization to obtain the predicted values ​​of landslide displacement.

[0010] In one possible implementation of the first aspect, the raw landslide displacement data is processed based on the CEEMDAN algorithm to obtain a landslide displacement time series, including:

[0011] Outliers in the original landslide displacement data were removed based on the Laida criterion.

[0012] The original landslide displacement data after removing outliers was interpolated using spline interpolation to obtain the landslide displacement data.

[0013] The CEEMDAN algorithm was used to separate the signal and noise of the landslide displacement data to obtain the landslide displacement time series.

[0014] In one possible implementation of the first aspect, the CEEMDAN algorithm is used to separate the signal and noise of landslide displacement data to obtain a landslide displacement time series, including:

[0015] Adding white noise to the landslide displacement data yields a landslide displacement time series signal.

[0016] The landslide displacement time series signal is decomposed based on the CEEMDAN algorithm until the landslide displacement time series signal becomes monotonic, thereby obtaining the modal components of the landslide displacement time series signal; the decomposition experiment is performed multiple times.

[0017] The landslide displacement time series is calculated based on the modal components of the landslide displacement time series signal obtained from multiple decomposition experiments.

[0018] In one possible implementation of the first aspect, the NBEATS model is established, including:

[0019] Establish an NBEATS model comprising M stacks; each stack includes N basic structural blocks; each basic structural block consists of a first part and a second part; the first part is a P-layer fully connected network (FC) used to generate the basic expansion coefficients; the basic expansion coefficients include a forward prediction factor θ. f and reverse predictor θ b The second part consists of a forward base layer and a backward base layer; the forward base layer is used to receive forward predictive factors; the backward base layer is used to receive backward predictive factors.

[0020] In one possible implementation of the first aspect, the forward base layer of the basic structure block receives the forward prediction factor, projects the forward prediction factor into the base function set, and generates the forward prediction output y defined in the previous basic structure block.

[0021] The backward base layer of the basic structure block receives the backward prediction factor, projects the backward prediction factor into the base function set, and generates the backward backtracking output x defined in the previous basic structure block.

[0022] In one possible implementation of the first aspect, the NBEATS model is optimized based on the landslide displacement time series, including:

[0023] The landslide displacement time series was normalized to the mean, and the landslide displacement time series dataset was divided into training set and test set;

[0024] Set the parameters of the NBEATS model; the parameters of the NBEATS model include the prediction time length and the backtracking time length; the backtracking time length of the NBEATS model is set to k times the prediction time length;

[0025] The NBEATS model is optimized based on the training set, test set, and Adam algorithm; specifically, the NBEATS model is trained using the training set, and the parameters of the NBEATS model are optimized using the test set.

[0026] In one possible implementation of the first aspect, the NBEATS model is optimized based on the training set, the test set, and the Adam algorithm, including:

[0027] The landslide displacement time series x1, divided into training and test sets, is input as the input signal into the basic structural block Block of the NBEATS model; where, for the first basic structural block Block, x1 is the input to the entire model; x l This represents the input signal for the l-th block, where l is an integer greater than or equal to 2; for the l-th basic structural block, the input signal x l It is the residual output signal of the (l-1)th basic structural block;

[0028] The l-th basic structural block is based on the input signal x. l It outputs two signals; the two output signals include the forward predicted landslide displacement value within the predicted time period. and backward landslide displacement values ​​within the backward time period

[0029] In one possible implementation of the first aspect, landslide displacement prediction is performed based on the NBEATS model to obtain prediction results, including:

[0030] Each basic structural block outputs the forward predicted landslide displacement value. Aggregate in the corresponding Stack;

[0031] After aggregation of the corresponding stacks, a global aggregation is performed on the NBEATS model to output the prediction results; the prediction results are the sum of predictions for all basic structural blocks.

[0032] Secondly, embodiments of this application provide a landslide displacement prediction device based on CEEMDAN-NBEATS, including: an acquisition module for acquiring raw landslide displacement data;

[0033] The data processing module is used to process the raw landslide displacement data based on the CEEMDAN algorithm to obtain the landslide displacement time series.

[0034] The model building and optimization module is used to build the NBEATS model and optimize the NBEATS model based on the landslide displacement time series.

[0035] The prediction module is used to predict landslide displacement based on the NBEATS model and obtain the prediction results.

[0036] The results processing module is used to perform inverse mean normalization on the prediction results of the NBEATS model to obtain the predicted values ​​of landslide displacement.

[0037] Thirdly, embodiments of this application provide a terminal device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method as described in any of the first aspects.

[0038] The beneficial effects of the embodiments in this application compared with the prior art are:

[0039] In this embodiment, by combining the CEEMDAN algorithm and the NBEATS model, rapid prediction can be achieved with only a small amount of landslide displacement data. It does not rely on large-scale landslide datasets, enabling rapid prediction and improving the accuracy and stability of predictions when the amount of sample data is small.

[0040] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of a landslide displacement prediction method based on CEEMDAN-NBEATS provided in an embodiment of this application;

[0043] Figure 2 This is a schematic diagram of the branching architecture of a basic building block (Block) provided in an embodiment of this application;

[0044] Figure 3 This is a schematic diagram of the branching architecture of a stack provided in an embodiment of this application;

[0045] Figure 4 This is a schematic diagram of a branch architecture based on the NBEATS model provided in an embodiment of this application;

[0046] Figure 5 This is a schematic diagram of the landslide displacement prediction device based on CEEMDAN-NBEATS provided in one embodiment of this application;

[0047] Figure 6 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0048] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0049] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0050] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0051] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0052] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0053] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0054] my country has diverse topography and complex geological conditions, resulting in widespread and severe landslides, particularly concentrated in the mountainous southwest, the Three Gorges Reservoir area of ​​the Yangtze River, and the middle and upper reaches of the Yellow River. Crustal uplift and river valley incision have created steep, mountainous terrain, making landslides and collapses frequent and prone to occur due to active tectonic movements and external forces. Reliable landslide prediction and early warning methods are a reasonable way to reduce landslide risk, and the successful implementation of accurate landslide displacement prediction methods will have significant social value and economic benefits.

[0055] The slope system is a highly complex nonlinear system with numerous influencing factors. The high robustness, learning and associative memory functions, and data mining characteristics of neural networks give them a significant advantage in predicting time-series landslide displacement data.

[0056] Currently, deep learning has been very successful in predicting landslide displacement data. Classical deep learning time series methods are mainly based on LSTM, SVM, BP, SVR, and GRU neural networks. These are prediction algorithms that uniformly model a large number of related time series. By training autoregressive recurrent network models on a large number of time series, they largely depend on large-scale landslide datasets. However, in reality, due to limitations in natural environment or equipment conditions, the number of data samples available for landslide early warning is relatively small. Under these conditions, although classical models have good fitting properties for nonlinear data sequences, traditional fractal models also suffer from drawbacks such as low prediction accuracy, poor prediction stability, slow prediction speed, and over-reliance on data and computing power when dealing with small sample data sequences.

[0057] Therefore, this invention proposes a landslide displacement prediction method based on CEEMDAN-NBEATS. Compared with traditional models, the CEEMDAN-NBEATS model performs a meta-learning form, which can achieve rapid prediction with only a small amount of data, and has higher accuracy and stability.

[0058] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0059] Reference Figure 1 The landslide displacement prediction method based on CEEMDAN-NBEATS is described in detail below:

[0060] Step 101: Obtain the original landslide displacement data.

[0061] For example, the raw landslide displacement data is the landslide displacement data detected by the monitoring instrument.

[0062] Step 102: Process the original landslide displacement data based on the CEEMDAN algorithm to obtain the landslide displacement time series.

[0063] Specifically, step 102 includes steps 1021 to 1023:

[0064] Step 1021: Remove outliers from the original landslide displacement data based on the Laida criterion.

[0065] In step 1021, the original landslide displacement data X is measured with equal precision according to the Laida criterion to obtain x1, x2, ..., x n , where n is the length of the detection dataset. Calculate the arithmetic mean x and residual error v of the total raw landslide displacement data X. i =x i -x(i=1,2,…,n), and calculate the standard error σ of the original landslide displacement data according to Bessel's formula. If a certain measured value x b The residual error v b (1≤b≤n), satisfying |v b |=|x b If -x|>3σ, then x is considered b These are bad values ​​containing gross errors, i.e., outliers in the original landslide displacement data, and should be removed.

[0066] The Bessel formula is as follows:

[0067]

[0068] Step 1022: Use spline interpolation to interpolate the original landslide displacement data after removing outliers to obtain landslide displacement data.

[0069] For example, cubic spline interpolation can be used to interpolate the removed outliers to obtain landslide displacement data x(t).

[0070] Step 1023: Use the CEEMDAN algorithm to separate the signal and noise of the landslide displacement data to obtain the landslide displacement time series.

[0071] In step 1023, white noise is added to the landslide displacement data to obtain a landslide displacement time series signal. The landslide displacement time series signal is decomposed using the CEEMDAN algorithm until it becomes monotonic, obtaining the modal components of the signal; the decomposition is performed multiple times. Based on the modal components of the landslide displacement time series signal obtained from the multiple decomposition experiments, the landslide displacement time series is calculated.

[0072] For example, firstly, I sets of adaptive white noise ω with a mean of 0 are added to the landslide displacement data x(t). i (t) yields the landslide displacement time series signal, and the i-th signal can be represented as:

[0073] x i (t)=x(t)+ω i (t) (2)

[0074] Where i represents the number of experiments 1, 2, ..., I.

[0075] The CEEMDAN algorithm was used to analyze x. i (t) is decomposed to obtain the first modal component, and then immediately summed and averaged to obtain:

[0076]

[0077] The residual component obtained by subtracting the first modal component from the landslide displacement data signal:

[0078] r1=x(t)-IMF1 (4)

[0079] Then, solve for the second-order modal component IMF2, and continue to add white noise ω to the residual component r1. i (t), forming a new signal to be decomposed:

[0080] R1(t)=r1(t)+ω i (t) (5)

[0081] Perform i experiments (i = 1, 2, ..., I), then perform EMD decomposition on R1(t) to obtain the second modal component:

[0082]

[0083] The residual component obtained by subtracting the second modal component from the landslide displacement data signal can be expressed as:

[0084] r2=x(t)-IMF2 (7)

[0085] Continue to repeat the execution of the k-th modal component IMF kThis continues until the landslide displacement time series signal after outlier processing can no longer be decomposed, i.e., the signal is monotonic, thus obtaining k IMFs. The signal x(t) can be expressed as:

[0086]

[0087] The CEEMDAN algorithm described above adds adaptive Gaussian white noise at each decomposition stage and immediately performs summation and averaging after obtaining the modal components. The same operation is performed in subsequent decompositions, which can ensure that the reconstruction error is 0 with fewer averaging operations and effectively avoid the noise propagation problem.

[0088] Step 103: Establish the NBEATS model and optimize the NBEATS model based on the landslide displacement time series.

[0089] Specifically, such as Figure 2-4 As shown, the NBEATS model is established, including: establishing an NBEATS model comprising M stacks; each stack comprising N basic structural blocks; each basic structural block consisting of a first part and a second part; the first part is a P-layer fully connected network (FC) used to generate the basic expansion coefficient; the basic expansion coefficient includes a forward prediction factor θ. f and reverse predictor θ b The second part consists of a forward base layer and a backward base layer; the forward base layer is used to receive forward predictive factors; the backward base layer is used to receive backward predictive factors.

[0090] Specifically, the forward base layer of the basic structure block receives the forward prediction factor, projects the forward prediction factor into the base function set, and generates the forward prediction output y defined in the previous basic structure block; the backward base layer of the basic structure block receives the backward prediction factor, projects the backward prediction factor into the base function set, and generates the backward backtracking output x defined in the previous basic structure block.

[0091] For example, the NBEATS model includes three general stacks, each containing three blocks. A block is the basic structural unit of NBEATS, composed of multiple fully connected layers, and mainly consists of two parts. The first part outputs the basal inflation coefficient (RBC) of the input time series data: the forward predictor θ. f and reverse predictor θ b The second part returns the basic inflation coefficient to the time series. Each block processes the input sequence and outputs a forward-predicted sequence and a backward-predicted sequence. The input of each block is the input of the previous block minus the output of the previous block.

[0092] The prediction result of each stack is the sum of the prediction results of its internal blocks, and the sum of the three stacks is the final prediction result of the model.

[0093] Specifically, based on landslide displacement time series, the NBEATS model is optimized, including: normalizing the mean of the landslide displacement time series and dividing the dataset into training and test sets. The parameters of the NBEATS model are set; these parameters include the prediction time length and the backtracking time length; the backtracking time length is set to k times the prediction time length. The NBEATS model is then optimized using the training and test sets and the Adam algorithm; specifically, the NBEATS model is trained using the training set and its parameters are optimized using the test set.

[0094] For example, the landslide displacement time series dataset can be divided into an 80% training set and a 20% test set. It should be noted that the training and test sets are not limited to any other ratio. The NBEATS model is trained using the training set, and its parameters are tuned using the loss from the test set, thereby improving the accuracy of landslide displacement prediction.

[0095] For example, the model parameters are set to predict the landslide displacement value in the next 5 days. The model backtracking time is set to 3 times the prediction length. The Adam algorithm is used as the optimizer to optimize the model. Data is used to train the model, and the parameters of the NBEATS model are adjusted using the loss of the test set to improve the accuracy of landslide displacement prediction.

[0096] The optimization of the NBEATS model based on the training and test sets and the Adam algorithm includes: inputting the landslide displacement time series x1, divided into training and test sets, as input signals into the basic structural block of the NBEATS model; where x1 is the input of the entire model for the first basic structural block; x l This represents the input signal for the l-th block, where l is an integer greater than or equal to 2; for the l-th basic structural block, the input signal x l This is the residual output signal of the (l-1)th basic structural block. The l-th basic structural block is based on the input signal x. l It outputs two signals; the two output signals include the forward predicted landslide displacement value within the predicted time period. and backward landslide displacement values ​​within the backward time period

[0097] To more clearly illustrate the structure of the NBEATS model and the process of optimizing it, detailed examples are provided:

[0098] The NBEATS model consists of 3 stacks, and each stack contains 3 blocks. l This indicates that the l-th block accepts its input and outputs two vectors. and For the first block in the model, its corresponding x1 is the entire model input, and the length of the history review window is a specific length ending with the last measured observation.

[0099] The input window is set to contain the raw landslide displacement data for the past 15 days, and the prediction range is the raw landslide displacement data for the next 5 days, with a total window length of 20. For the remaining blocks, their input x... l This is the residual output of the previous block. Each block has two outputs: These are the forward predicted landslide displacement values ​​over a 5-day period. The values ​​represent the backward landslide displacement over a period of 15 days.

[0100] The basic building block, Block, has a branching structure, consisting of two parts, such as... Figure 2 As shown.

[0101] The first part is a 4-layer fully-connected network (FC) that generates the basic inflation coefficient: the forward predictor θ. f and reverse predictor θ b The operation is described by the following equation:

[0102]

[0103] The LINEAR layer is simply a linear projection layer, i.e. An FC layer is a standard fully connected layer with ReLU nonlinearity, therefore for FC... l, We have, for example: h l, =RELU(W l,1 x l +b l,1 One task of this part of the architecture is to predict the forward expansion coefficient. The ultimate goal is to achieve this through proper mixing. The provided basis vectors are used to optimize the accuracy of partial landslide displacement prediction. Furthermore, this sub-network predicts... Used to generate x l Backward expansion coefficient of the estimated value Its ultimate goal is to help downstream blocks by removing input components that do not contribute to landslide displacement prediction.

[0104] Part Two is backwards and forward The base layer consists of layers that each receive its own forward pass. and backward The expansion coefficients are projected onto the base function set and used to generate the backtracking output x and the forward prediction output y defined in the previous section. This is achieved through... and Expanding factor and This is mapped to the output of the base layer. Its operation is described by the following equation:

[0105]

[0106] In the formula, and It is the basis vector for landslide displacement prediction and backtracking. yes The i-th element. and Its function is to provide a sufficiently rich collection. and So that their respective outputs can be expanded by varying coefficients. and To fully express.

[0107] Stack hierarchical double-residual topology, such as Figure 3 and Figure 4 As shown, this Stack architecture has two residual branches, running on the backtracking prediction and prediction branches of each Block, respectively. The backtracking branch is represented by "-", and the prediction branch by "+". This architecture consists of M Stacks, and each Stack contains N Blocks. Its operation is described by the following equation:

[0108]

[0109] The input x1 of the first block is the input of the entire model. For all other blocks, backtrack the remaining branches x. l This can be viewed as a sequential analysis of the input signal. The input x of each block... l , is the input x of the previous block. l-1 Subtract the output of the previous block Processed signal x lIt can approximate the data well, making the prediction of downstream blocks easier. In this way, the data processed by each block in the model is the residual that the previous blocks could not fit correctly, which also plays the role of decomposing and predicting the time series layer by layer.

[0110] Step 104: Based on the NBEATS model, landslide displacement is predicted, and the prediction results are obtained.

[0111] Specifically, landslide displacement prediction is performed based on the NBEATS model, and the prediction results include: the forward predicted landslide displacement value output by each basic structural block. Aggregate in the corresponding Stack;

[0112] After aggregation of the corresponding stacks, a global aggregation is performed on the NBEATS model to output the prediction results; the prediction results are the sum of predictions for all basic structural blocks.

[0113] Step 105: Perform inverse mean normalization on the prediction results of the NBEATS model to obtain the predicted landslide displacement value.

[0114] In one embodiment, the prediction performance of the above-mentioned NBEATS model is verified and evaluated. The mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), and symmetric mean absolute percentage error (SMAPE) are used to calculate the error between the final predicted landslide displacement value and the actual value of the NBEATS model for model evaluation. The evaluation index formula is as follows:

[0115] Mean Square Error (MSE):

[0116]

[0117] Root Mean Square Error (RMSE):

[0118]

[0119] Mean Absolute Percentage Error (MAPE):

[0120]

[0121] Symmetric Mean Absolute Percentage Error (SMAPE):

[0122]

[0123] Among them, y i This represents the actual landslide displacement. These are predicted values.

[0124] The NBEATS model with set parameters was trained and validated, and the experimental results were compared with those of the traditional LSTM model, as shown in Table 1.

[0125] Table 1. Comparison of Evaluation Metrics between NBEATS Model and Traditional LSTM Model

[0126]

[0127] Analysis of the results shows that the RMSE of the landslide displacement predicted by NBEATS is 1.6879, which is 5.5969 lower than the 7.2848 of the traditional LSTM model; the prediction time is 35.16 s, which is 347.97 s faster than the 383.13 s of the traditional LSTM model, thus outperforming the traditional LSTM model. Therefore, the NBEATS model has faster speed and higher accuracy and stability than the LSTM model.

[0128] As can be seen, the landslide displacement prediction method based on CEEMDAN-NBEATS provided in this embodiment of the invention, by combining the CEEMDAN algorithm and the NBEATS model, can achieve rapid prediction even with a small amount of time-series landslide displacement data. Compared with the traditional LSTM model, this method significantly reduces prediction bias and improves accuracy. Compared with traditional landslide displacement prediction algorithms, CEEMDAN-NBEATS performs a meta-learning approach, which can quickly adapt to new landslide displacement prediction tasks.

[0129] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0130] Corresponding to the landslide displacement prediction method based on CEEMDAN-NBEATS in the above embodiments, Figure 5 The diagram shows a structural block diagram of a landslide displacement prediction device based on CEEMDAN-NBEATS provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown.

[0131] See Figure 5 The landslide displacement prediction device based on CEEMDAN-NBEATS in this application embodiment may include an acquisition module 201, a data processing module 202, a model building and optimization module 203, a prediction module 204, and a result processing module 205.

[0132] The acquisition module 201 is used to acquire raw landslide displacement data.

[0133] The data processing module 202 is used to process the raw landslide displacement data based on the CEEMDAN algorithm to obtain the landslide displacement time series.

[0134] The model building and optimization module 203 is used to build the NBEATS model and optimize the NBEATS model based on the landslide displacement time series.

[0135] Prediction module 204 is used to predict landslide displacement based on the NBEATS model and obtain prediction results.

[0136] The result processing module 205 is used to perform inverse mean normalization on the prediction results of the NBEATS model to obtain the predicted values ​​of landslide displacement.

[0137] This application also provides a terminal device, see [link to relevant documentation] Figure 6 The terminal device 300 may include at least one processor 310 and a memory 320. The memory 320 stores a computer program that can run on the at least one processor 310. When the processor 310 executes the computer program, it implements the steps in any of the above-described method embodiments, for example... Figure 1 Steps 101 to 105 in the illustrated embodiment. Alternatively, when the processor 310 executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 5 The functions of modules 201 to 205 are shown.

[0138] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 320 and executed by processor 310 to complete this application. The one or more modules / units may be a series of computer program segments capable of performing specific functions, which describe the execution process of the computer program in terminal device 300.

[0139] Those skilled in the art will understand that Figure 6 This is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0140] The processor 310 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0141] The memory 320 can be an internal storage unit of the terminal device or an external storage device, such as a plug-in hard drive, a smart media card (SMC), a secure digital card (SD), or a flash card. The memory 320 is used to store the computer program and other programs and data required by the terminal device. The memory 320 can also be used to temporarily store data that has been output or will be output.

[0142] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0143] The landslide displacement prediction method based on CEEMDAN-NBEATS provided in this application embodiment can be applied to terminal devices such as computers, tablets, laptops, netbooks, and personal digital assistants (PDAs). This application embodiment does not impose any restrictions on the specific type of terminal device.

[0144] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various embodiments of the landslide displacement prediction method based on CEEMDAN-NBEATS.

[0145] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the various embodiments of the landslide displacement prediction method based on CEEMDAN-NBEATS.

[0146] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0147] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0148] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0150] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A landslide displacement prediction method based on CEEMDAN-NBEATS, characterized in that, The method comprises the following steps: obtaining original landslide displacement data; processing the original landslide displacement data based on a CEEMDAN algorithm to obtain a landslide displacement time series; establishing an NBEATS model and optimizing the NBEATS model based on the landslide displacement time series; performing landslide displacement prediction based on the optimized NBEATS model to obtain a prediction result; performing inverse mean normalization processing on the prediction result to obtain a landslide displacement prediction value; the step of establishing the NBEATS model comprises: Establishing an NBEATS model comprising M stacks Stack; each of the stacks Stack comprises N basic structural blocks Block; each of the basic structural blocks Block consists of a first part and a second part; the first part is a P-layer fully connected network FC, used for generating a basic expansion coefficient; the basic expansion coefficient comprises a forward prediction factor and a reverse prediction factor ; the second part consists of a forward basic layer and a backward basic layer; the forward basic layer is used for receiving the forward prediction factor; the backward basic layer is used for receiving the reverse prediction factor; the step of optimizing the NBEATS model based on the landslide displacement time series comprises: performing mean normalization processing on the landslide displacement time series, and dividing a data set of the landslide displacement time series into a training set and a test set; setting parameters of the NBEATS model; the parameters of the NBEATS model comprise a prediction time length and a backtracking time length; the backtracking time length of the NBEATS model is set to k times of the prediction time length; optimizing the NBEATS model based on the training set and the test set and an Adam algorithm; wherein the NBEATS model is trained by using the training set, and the NBEATS model is parameter-optimized by using the test set; the step of performing landslide displacement prediction based on the optimized NBEATS model to obtain a prediction result comprises: a forward predicted landslide displacement value outputted by each of the infrastructure blocks Block aggregated in a corresponding stack Stack; After the corresponding stack Stack is aggregated, global aggregation is performed on the NBEATS model, and a prediction result is output; the prediction result is the sum of all basic structure blocks Block predictions .

2. The landslide displacement prediction method based on CEEMDAN-NBEATS of claim 1, wherein, the step of processing the original landslide displacement data based on the CEEMDAN algorithm to obtain a landslide displacement time series comprises: removing abnormal values of the original landslide displacement data based on a Laplace criterion; performing interpolation on the original landslide displacement data from which the abnormal values are removed by using spline interpolation to obtain landslide displacement data; performing signal-noise separation on the landslide displacement data by using the CEEMDAN algorithm to obtain a landslide displacement time series.

3. The landslide displacement prediction method based on CEEMDAN-NBEATS of claim 2, wherein, the step of performing signal-noise separation on the landslide displacement data by using the CEEMDAN algorithm to obtain a landslide displacement time series comprises: adding a white noise signal to the landslide displacement data to obtain a landslide displacement time series signal; decomposing the landslide displacement time series signal based on the CEEMDAN algorithm until the landslide displacement time series signal is monotonic to obtain modal components of the landslide displacement time series signal; wherein the number of decomposition experiments is multiple; calculating a landslide displacement time series based on the modal components of the landslide displacement time series signal obtained through multiple decomposition experiments.

4. The landslide displacement prediction method based on CEEMDAN-NBEATS of claim 1, wherein, The forward base layer of the base block receives the forward prediction factor, projects the forward prediction factor inside a base function set, and generates a forward prediction output y defined in a previous base block; The backward base layer of the base block receives the backward prediction factor, projects the backward prediction factor inside a base function set, and generates a backward backtracking output x defined in a previous base block.

5. The landslide displacement prediction method based on CEEMDAN-NBEATS of claim 1, wherein, The step of optimizing the NBEATS model based on the training set and the test set and the Adam algorithm comprises: sliding displacement time series divided into a training set and a test set is input into a base block Block of the NBEATS model; wherein for a first base block Block, is the entire model input; denotes the input signal of the th Block, is an integer greater than or equal to 2; for the th base block Block, the input signal is the residual output signal of the th base block Block; The first infrastructure block Block outputs two output signals based on the input signals ; the two output signals include a forward prediction landslide displacement value for the prediction time length and a backward trace landslide displacement value for the trace time length.

6. A landslide displacement prediction device based on CEEMDAN-NBEATS, characterized by, The method comprises the following steps: an acquisition module is configured to acquire original landslide displacement data; The data processing module is configured to process the original landslide displacement data based on a CEEMDAN algorithm to obtain a landslide displacement time series. The model establishing and optimizing module is configured to establish an NBEATS model and optimize the NBEATS model based on the landslide displacement time series. The prediction module is configured to perform landslide displacement prediction based on the NBEATS model to obtain a prediction result. The result processing module is configured to perform inverse mean normalization processing on the prediction result of the NBEATS model to obtain a landslide displacement prediction value. The model establishing and optimizing module is specifically configured to: Establishing an NBEATS model comprising M stacks Stack; each of the stacks Stack comprises N basic structural blocks Block; each of the basic structural blocks Block consists of a first part and a second part; the first part is a P-layer fully connected network FC, used for generating a basic expansion coefficient; the basic expansion coefficient comprises a forward prediction factor and a reverse prediction factor ; the second part consists of a forward basic layer and a backward basic layer; the forward basic layer is used for receiving the forward prediction factor; the backward basic layer is used for receiving the reverse prediction factor; The model establishing and optimizing module is further specifically configured to: perform mean normalization processing on the landslide displacement time series, and divide a data set of the landslide displacement time series into a training set and a test set; set parameters of the NBEATS model, wherein the parameters of the NBEATS model include a prediction time length and a backtracking time length, and the backtracking time length of the NBEATS model is set to k times of the prediction time length; optimize the NBEATS model based on the training set, the test set, and an Adam algorithm, wherein the NBEATS model is trained by using the training set, and the NBEATS model is parameter-optimized by using the test set; The result processing module is specifically configured to: a forward predicted landslide displacement value outputted by each of the infrastructure blocks Block aggregated in a corresponding stack Stack; After the corresponding stack Stack is aggregated, global aggregation is performed on the NBEATS model, and a prediction result is output; the prediction result is the sum of all basic structure blocks Block predictions . 7.A terminal device, comprising a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the computer program comprises the following steps of: The processor implements the CEEMDAN-NBEATS-based landslide displacement prediction method in any one of claims 1 to 5 when executing the computer program.

Images