A dam displacement prediction method considering time lag characteristics

Abstract

The application relates to a dam displacement prediction method considering time lag characteristics, comprising the following steps: obtaining displacement observation data of a target dam and preprocessing the displacement observation data to form a displacement observation sequence and determine an upper limit of time lag in the displacement observation sequence; constructing a displacement prediction data set of the target dam based on the displacement observation sequence and environmental variable factors affecting the displacement of the dam, and time-paneling data in the displacement prediction data set according to the upper limit of time lag; constructing a deep neural network based on a gated recurrent unit, and establishing a time lag parameter considering the upper limit of time lag; traversing the time lag parameter, and selecting an optimal time lag parameter; training the deep neural network based on the gated recurrent unit by using the time-panelized displacement prediction data set according to the optimal time lag parameter, so as to obtain a trained dam displacement prediction model; and using the trained dam displacement prediction model to predict the displacement of the target dam.

Description

TECHNICAL FIELD

[0001] The present application relates to a dam displacement prediction method considering time lag characteristics, belonging to the technical field of dam monitoring. BACKGROUND

[0002] As a hydraulic engineering system integrating irrigation, power generation, flood control and other functions, the reservoir dam has important significance for promoting regional economic prosperity and maintaining people's livelihood. Under the combined action of complex external dynamic and static loads and internal material aging, the structural safety factor of the dam decreases with the increase of the operation time. In order to prevent dam break from causing disasters to the environment and human beings, applying a reasonable method to monitor the safety of various indexes of the dam has always been the focus of attention in the field of hydraulic safety. In most cases, the prediction model with displacement as the main target can identify the difference between the actual behavior and the predicted behavior of the dam structure, so as to confirm the potential abnormal phenomena in the dam system. Therefore, developing a reliable and accurate displacement prediction model to explain the possible future structural behavior of the dam is of great importance to prolong the safe service life of the dam.

[0003] At present, the displacement prediction model based on computer technology has gradually become one of the main safety monitoring means in this field. For example, the Chinese invention patent with publication number "CN117540265A" discloses a concrete dam deformation prediction method based on IPSO-Transformer model, which constructs an IPSO-Transformer dam displacement deformation prediction model to predict the dam displacement deformation.

[0004] However, the above monitoring method cannot consider the inherent time lag characteristics of dam displacement due to its static nature, that is, the displacement condition at the current time will have a lasting impact on the degree of dam displacement in the future period of time, and this impact will show significant variability under the action of complex environmental factors. Therefore, considering the needs of actual engineering, a dam displacement prediction method considering the lag characteristics is needed to achieve more efficient and reliable dam safety state evaluation. SUMMARY

[0005] In order to solve the problems existing in the prior art, the present application proposes a dam displacement prediction method considering time lag characteristics.

[0006] The technical scheme of the present application is as follows:

[0007] On the one hand, the present application proposes a dam displacement prediction method considering time lag characteristics, comprising the following steps:

[0008] Obtain the displacement observation data of the target dam and preprocess it to form a displacement observation sequence, and determine the upper limit of time lag in the displacement observation sequence;

[0009] Based on the displacement observation sequence and the environmental variable factor affecting the displacement of the dam, a displacement prediction data set of the target dam is constructed, and the data in the displacement prediction data set is time-panelized according to the upper limit of time lag;

[0010] A deep neural network based on a gated recurrent unit is constructed, and a time lag parameter considering the upper limit of time lag is established; the time-panelized displacement prediction data set is taken as input, the learning error waveform and the evaluation error waveform of the network are drawn based on the loss between the predicted displacement and the actual displacement output by the network, the path similarity distance between the two waveforms is obtained, the time lag parameter is traversed, and the optimal time lag parameter with the minimum path similarity distance is selected;

[0011] According to the optimal time lag parameter, the deep neural network based on the gated recurrent unit is trained by using the time-panelized displacement prediction data set, and a trained dam displacement prediction model is obtained;

[0012] The trained dam displacement prediction model is used for displacement prediction of the target dam.

[0013] As a preferred embodiment, the method for obtaining displacement observation data of the target dam and preprocessing the displacement observation data is specifically:

[0014] The horizontal displacement observation data of the target dam perpendicular to the flow direction of the water flow at each displacement monitoring node is obtained;

[0015] The abnormal data in the horizontal displacement observation data is identified and removed;

[0016] The horizontal displacement observation data after the abnormal data identification and removal is subjected to missing data filling processing, and the displacement observation sequence is obtained.

[0017] As a preferred embodiment, the method for determining the upper limit of time lag in the displacement observation sequence is specifically:

[0018] The autoregressive correlation coefficient at each displacement monitoring node is calculated, which is specifically as follows:

[0019] μ=E(V t );

[0020] σ 2 =D(V t )=E((V t -μ) 2 );

[0021]

[0022] Wherein, V t =(v1,v2,…v t|t∈[1,N]) represents the displacement observation sequence of the corresponding displacement monitoring node, where N represents the total length of the sequence, and μ and σ 2 These are the mean and variance of the displacement observation sequence, respectively; k Here, is the autocovariance, k is the penalty coefficient for autonomous selection, and AC(k) is the autoregressive correlation coefficient of the corresponding displacement observation sequence.

[0023] Traverse the displacement monitoring nodes to obtain the corresponding autoregressive correlation coefficient sequence, calculate the confidence interval of the autoregressive correlation coefficient sequence with a confidence level greater than a set threshold, and use the number of autoregressive correlation coefficients falling outside the confidence interval as the upper limit of time delay.

[0024] As a preferred embodiment, the method for temporal panelization of the data in the displacement prediction dataset based on the upper limit of time delay is specifically as follows:

[0025] Any data in the displacement prediction dataset is represented by a two-dimensional matrix;

[0026] For each two-dimensional matrix, time panelization is performed based on the upper limit of time delay:

[0027]

[0028] Where X is the input two-dimensional matrix, X∈R n×m n and m represent the number of samples and the feature dimension, respectively; i∈[1,n-T+1] represents the starting position of each sliding window; T is the sample lag order;

[0029] We obtain a three-dimensional tensor after time panelization:

[0030]

[0031] Among them, X 3D This is the three-dimensional tensor corresponding to the input two-dimensional matrix X.

[0032] In a preferred embodiment, the environmental variable factors affecting dam displacement include the water pressure component δ. H Temperature component δ T and the time-dependent component δ θ ;

[0033] Among them, the water pressure component denoted as z-th power of the average daily reservoir water level upstream of the target dam, where a1, a2, a3, and c are the regression coefficients of the preset water pressure components;

[0034] Temperature component δ T =b1T 1-2 +b2T 3-7 +…b 13 T301-365 ;T p-q This represents the segmented average temperature from day p to day q prior to the current observation date; b1, b2, and b3 are the regression coefficients of the preset temperature components.

[0035] Time-dependent component δ θ =c1θ+c2lnθ; θ=t / 100, where t represents the total number of days from the initial monitoring date to the current monitoring date; c1 and c2 are the preset regression coefficients for the deformation of the time-dependent components.

[0036] In a preferred embodiment, in the step of plotting the learning error waveform and evaluation error waveform of the network based on the loss between the predicted displacement and the actual displacement output by the network, the loss function is specifically:

[0037]

[0038] Where, f(X) 3D ;T max ) i Y is the predicted displacement of the i-th sample output by the network. i The true displacement of the i-th sample, T max This represents the upper limit of the time delay.

[0039] In a preferred embodiment, the formula for calculating the path similarity distance in the step of obtaining the path similarity distance between two waveforms is as follows:

[0040]

[0041] Where α is a non-decreasing continuous function from [0,1] to [a,b], β is a non-decreasing continuous function from [0,1] to [a',b'], and d(·) represents the Euclidean distance.

[0042] On the other hand, the present invention also proposes a dam displacement prediction system considering time delay characteristics, comprising:

[0043] The preprocessing module is used to acquire and preprocess the displacement observation data of the target dam to form a displacement observation sequence and determine the upper limit of the time delay in the displacement observation sequence.

[0044] The time feature insertion module constructs a displacement prediction dataset for the target dam based on displacement observation sequences and environmental variable factors affecting dam displacement, and performs time panelization on the data in the displacement prediction dataset according to the upper limit of time delay.

[0045] The model building module is used to build a deep neural network based on gated recurrent units and establish time delay parameters that consider the upper limit of time delay. The displacement prediction dataset after time panelization is used as input. The learning error waveform and evaluation error waveform of the network are plotted based on the loss between the predicted displacement and the actual displacement output by the network. The path similarity distance between the two waveforms is obtained. The time delay parameters are traversed and the optimal time delay parameter with the minimum path similarity distance is selected.

[0046] The prediction model training module is used to train a deep neural network based on gated recurrent units using the displacement prediction dataset after time panelization, based on the optimal time delay parameters, to obtain a trained dam displacement prediction model.

[0047] The dam displacement prediction module uses a trained dam displacement prediction model to predict the displacement of the target dam.

[0048] In another aspect, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the program to implement the dam displacement prediction method considering time delay characteristics as described in any embodiment of the present invention.

[0049] In another aspect, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements the dam displacement prediction method considering time delay characteristics as described in any embodiment of the present invention.

[0050] The present invention has the following beneficial effects:

[0051] This invention proposes a dam displacement prediction method that considers time lag characteristics. This method can take into account and capture the potential impact of time lag on dam displacement development, greatly improving the robustness and generalization ability of current dam displacement prediction models due to their static nature. By using artificial intelligence and mathematical statistics, the method determines the dam displacement lag condition that best reflects actual engineering conditions and uses this as an important parameter in the construction and training of the prediction model, thereby improving the accuracy of dam displacement prediction. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the method of an embodiment of the present invention;

[0053] Figure 2 This is an example diagram of the preprocessed displacement observation sequence in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram comparing the predicted displacement sequence waveform output by the prediction model and the actual displacement sequence waveform in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0057] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0058] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0059] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0060] Example 1:

[0061] See Figure 1 This embodiment provides a dam displacement prediction method that considers time delay characteristics, including the following steps:

[0062] S100. Obtain displacement observation data of the target dam and preprocess it to form a displacement observation sequence, and determine the upper limit of time delay in the displacement observation sequence.

[0063] In a preferred embodiment of this invention, the method for obtaining and preprocessing the displacement observation data of the target dam in step S100 is as follows:

[0064] S110. Obtain horizontal displacement data of the target dam at each monitoring node, perpendicular to the direction of water flow, through automatic observation equipment and manual observation methods;

[0065] S111. Use the ISODEPTH method to calculate and filter out abnormal data in the horizontal displacement observation data;

[0066] S112. Using linear interpolation, missing data filling is performed on the horizontal displacement observation data after the abnormal data identification and removal are completed to obtain the displacement observation sequence.

[0067] An example of the displacement observation sequence obtained after the above preprocessing is shown in the figure below. Figure 2 As shown.

[0068] In a preferred embodiment of this invention, step S100, the method for determining the upper limit of the time delay in the displacement observation sequence, specifically comprises:

[0069] S120. Calculate the autoregressive correlation coefficient for each displacement monitoring node, as follows:

[0070] μ=E(V t );

[0071] σ 2 =D(V) t )=E((V t -μ) 2 );

[0072]

[0073] Among them, V t =(v1,v2,…v t |t∈[1,N]) represents the displacement observation sequence of the corresponding displacement monitoring node, where N represents the total length of the sequence, and μ and σ 2 These are the mean and variance of the displacement observation sequence, respectively; k Here, is the autocovariance, k is the self-selected penalty coefficient (set to 50 in this embodiment); AC(k) is the autoregressive correlation coefficient of the corresponding displacement observation sequence;

[0074] S121. Execute the above step S120 until the displacement monitoring nodes are traversed to obtain the corresponding autoregressive correlation coefficient sequence. Calculate the confidence interval of the autoregressive correlation coefficient sequence whose confidence level is greater than a set threshold. In this embodiment, the threshold is set to 95%, and the number of autoregressive correlation coefficients falling outside the confidence interval is used as the upper limit of the time delay.

[0075] S200. Based on displacement observation sequences and environmental variable factors affecting dam displacement, a displacement prediction dataset for the target dam is constructed. The dataset is divided into training, validation, and test sets in an 8:1:1 ratio. The data in the displacement prediction dataset is then time-panelized according to the upper limit of time delay.

[0076] In a preferred embodiment of this example, in step S200, the method of temporal panelization of the data in the displacement prediction dataset based on the upper limit of time delay is specifically as follows:

[0077] Any data in the displacement prediction dataset is represented by a two-dimensional matrix X;

[0078] For each two-dimensional matrix X, it can be processed by time panelization (X 3D For each sliding window, at the starting position i∈[1,n-T+1], construct a structure belonging to X. 3D Each slice:

[0079]

[0080] Where X is the input two-dimensional matrix, X∈R n×m n and m represent the number of samples and the feature dimension, respectively; T is the sample lag order.

[0081] Finally, we obtain the three-dimensional tensor X after time panelization. 3D :

[0082]

[0083] Among them, X 3D Let X be the three-dimensional tensor corresponding to the input two-dimensional matrix X. 3D ∈R (n-T+1)×T×m .

[0084] In a preferred embodiment of this invention, in step S200, the environmental variable factors affecting the dam displacement include the water pressure component δ. H Temperature component δ T and the time-dependent component δ θ ;

[0085] The water pressure component is expressed as a cubic polynomial of the upstream reservoir water level, namely: denoted as z-th power of the average daily reservoir water level upstream of the target dam, where a1, a2, a3, and c are the regression coefficients of the preset water pressure components;

[0086] The temperature component is chosen using the expression for the piecewise mean air temperature to simulate the temperature field changes inside the dam, namely: δ T =b1T 1-2 +b2T 3-7 +…b 13 T 301-365 ;T p-q This represents the segmented average temperature from day p to day q prior to the current observation date; b1, b2, and b3 are the regression coefficients of the preset temperature components.

[0087] The time-dependent component is chosen from a multinomial expression containing a logarithmic function, namely: δ θ=c1θ+c2lnθ; θ=t / 100, where t represents the total number of days from the initial monitoring date to the current monitoring date; c1 and c2 are the preset regression coefficients for the deformation of the time-dependent components.

[0088] S300. Constructing a Deep Neural Network Based on Gated Recurrent Units (GRU): GRU is a widely used model in natural language processing and other sequence data processing tasks. By introducing a gating mechanism, GRU effectively solves the gradient vanishing and gradient exploding problems that traditional recurrent neural networks (RNNs) easily encounter when processing long sequences. Specifically, GRU contains two main gating structures: a reset gate and an update gate. The reset gate determines how much past information should be discarded at the current time step, while the update gate controls how much past information should be retained at the current time step. These two gating mechanisms work together to enable GRU to effectively capture short-term dynamic changes while maintaining long-term dependencies. In this way, GRU performs excellently in various complex sequence data tasks, such as language modeling, machine translation, speech recognition, and time series prediction.

[0089] And establish a time delay parameter W∈[1,T] that considers the upper limit of time delay. max Using the time-panelized displacement prediction dataset as input, and based on the loss between the predicted displacement and the actual displacement output by the network, the learning error waveform and evaluation error waveform of the network are plotted. The path similarity distance between the two waveforms is obtained, and the time delay parameter W∈[1,T] is traversed. max The optimal time delay parameter with the minimum path similarity distance is selected as the time delay order T' that best reflects the actual displacement. max ;

[0090] In a preferred embodiment of this invention, in step S300, the step of plotting the learning error waveform and evaluation error waveform of the network based on the loss between the predicted displacement and the actual displacement output by the network, specifically includes the following loss function:

[0091]

[0092] Where, f(X) 3D ;T max ) i Y is the predicted displacement of the i-th sample output by the network. i The true displacement of the i-th sample, T max This represents the upper limit of the time delay.

[0093] In a preferred embodiment of this invention, in step S300, the formula for calculating the path similarity distance between the two waveforms is as follows:

[0094]

[0095] Where α is a non-decreasing continuous function from [0,1] to [a,b], β is a non-decreasing continuous function from [0,1] to [a',b'], and d(·) represents the Euclidean distance.

[0096] S400, based on the optimal time delay parameter T' max A deep neural network based on gated recurrent units was reconstructed and trained using a time-panelized displacement prediction dataset. The network was then tested using a test set, which allowed for comparison of the visualized results of the predicted displacement sequence waveform with the actual displacement sequence waveform. Figure 3 As shown; with R 2 The model performance is evaluated using MAE and RMSE as the main indicators. Once the model performance meets the standards, a well-trained dam displacement prediction model is obtained.

[0097] Specifically, MAE, RMSE, and R 2 The specific calculation process is as follows:

[0098]

[0099] S500: Input the current observation data of the target dam into the trained dam displacement prediction model, and use the trained dam displacement prediction model to predict the displacement of the target dam.

[0100] To verify the effectiveness and superiority of the method proposed in this embodiment, the following table compares the accuracy metrics of the prediction model proposed in this embodiment with those of traditional models:

[0101] Table 1. Comparison of Model Accuracy

[0102]

[0103] Compared with traditional statistical models and machine learning models, the R of the GRU constructed in this embodiment is superior. 2 The highest index was 0.996, while the lowest were MAE and RMSE, at 0.041 and 0.055 respectively. This phenomenon means that the dam displacement prediction model constructed based on this embodiment performs significantly better than the statistical and machine learning models commonly used in current dam safety monitoring in terms of overall displacement trend and local error control, thus outputting high-precision displacement prediction results.

[0104] Example 2:

[0105] This embodiment also proposes a dam displacement prediction system that considers time delay characteristics, including:

[0106] The preprocessing module is used to acquire and preprocess the displacement observation data of the target dam to form a displacement observation sequence and determine the upper limit of the time delay in the displacement observation sequence. This module is used to implement the function of step S100 in Embodiment 1, which will not be described in detail here.

[0107] The time feature insertion module constructs a displacement prediction dataset for the target dam based on the displacement observation sequence and environmental variable factors affecting the dam displacement, and performs time panelization on the data in the displacement prediction dataset according to the upper limit of time delay; this module is used to implement the function of step S200 in Example 1, and will not be described in detail here.

[0108] The model building module is used to construct a deep neural network based on gated recurrent units and establish a time delay parameter that considers the upper limit of time delay. It takes the displacement prediction dataset after time panelization as input, and plots the learning error waveform and evaluation error waveform of the network based on the loss between the predicted displacement and the actual displacement output by the network. It obtains the path similarity distance between the two waveforms, traverses the time delay parameters, and selects the optimal time delay parameter with the minimum path similarity distance. This module implements the function of step S300 in Example 1, which will not be elaborated here.

[0109] The prediction model training module is used to train a deep neural network based on gated recurrent units using the displacement prediction dataset after time panelization, based on the optimal time delay parameters, to obtain a trained dam displacement prediction model; this module is used to implement the function of step S400 in Example 1, which will not be described in detail here.

[0110] The dam displacement prediction module uses a trained dam displacement prediction model to predict the displacement of the target dam; this module is used to implement the function of step S500 in Example 1, and will not be described in detail here.

[0111] Example 3:

[0112] This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the dam displacement prediction method considering time delay characteristics as described in any embodiment of the present invention.

[0113] Example 4:

[0114] This embodiment proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements a dam displacement prediction method considering time delay characteristics as described in any embodiment of the present invention.

[0115] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0116] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. 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.

[0117] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting dam displacement considering time delay characteristics, characterized in that, Includes the following steps: The displacement observation data of the target dam is acquired and preprocessed to form a displacement observation sequence, and the upper limit of the time delay in the displacement observation sequence is determined. A displacement prediction dataset for the target dam is constructed based on displacement observation sequences and environmental variables that affect dam displacement. The data in the displacement prediction dataset are then time-panelized according to the upper limit of time delay. Construct a deep neural network based on gated recurrent units and establish a time delay parameter that considers the upper limit of time delay. Take the displacement prediction dataset after time panelization as input, and draw the learning error waveform and evaluation error waveform of the network based on the loss between the predicted displacement and the actual displacement output by the network. Obtain the path similarity distance between the two waveforms, traverse the time delay parameters, and select the optimal time delay parameter with the minimum path similarity distance. Based on the optimal time delay parameters, a deep neural network based on gated recurrent units is trained using the displacement prediction dataset after time panelization to obtain a trained dam displacement prediction model. The displacement of the target dam is predicted using a trained dam displacement prediction model. Specifically, the method for determining the upper limit of time delay in the displacement observation sequence is as follows: The autoregressive correlation coefficients at each displacement monitoring node were calculated as follows: μ=E(V t ); s 2 =D(V t )=E((V t -m) 2 ); Among them, V t =(v1,v2,…v t |t∈[1,N]) represents the displacement observation sequence of the corresponding displacement monitoring node, where N represents the total length of the sequence, and μ and σ 2 These are the mean and variance of the displacement observation sequence, respectively; k Here, is the autocovariance, k is the penalty coefficient for autonomous selection, and AC(k) is the autoregressive correlation coefficient of the corresponding displacement observation sequence. Traverse the displacement monitoring nodes to obtain the corresponding autoregressive correlation coefficient sequence, calculate the confidence interval of the autoregressive correlation coefficient sequence with a confidence level greater than a set threshold, and use the number of autoregressive correlation coefficients falling outside the confidence interval as the upper limit of time delay. Among them, the environmental variable factors affecting dam displacement include the water pressure component δ H Temperature component δ T and the time-dependent component δ θ ; Among them, the water pressure component denoted as z-th power of the average daily reservoir water level upstream of the target dam, where a1, a2, a3, and c are the regression coefficients of the preset water pressure components; Temperature component δ T =b1T 1-2 +b2T 3-7 +…b 13 T 301-365 ;T p-q This represents the segmented average temperature from day p to day q prior to the current observation date; b1, b2, b 13 The regression coefficients for the preset temperature components; Time-dependent component δ θ =c1θ+c2lnθ; θ=t / 100, where t represents the total number of days from the initial monitoring date to the current monitoring date; c1 and c2 are the preset regression coefficients for the deformation of the time-dependent components.

2. The dam displacement prediction method considering time delay characteristics according to claim 1, characterized in that, The method for obtaining and preprocessing the displacement observation data of the target dam is as follows: Acquire the horizontal displacement observation data of the target dam at each monitoring node, which is perpendicular to the direction of water flow; Identify and remove outlier data from horizontal displacement observation data; After identifying and removing outlier data, missing data filling processing is performed on the horizontal displacement observation data to obtain the displacement observation sequence.

3. The dam displacement prediction method considering time delay characteristics according to claim 1, characterized in that, The method for temporal panelization of the data in the displacement prediction dataset based on the upper limit of time delay is as follows: Any data in the displacement prediction dataset is represented by a two-dimensional matrix; For each two-dimensional matrix, time panelization is performed based on the upper limit of time delay: Where X is the input two-dimensional matrix, X∈R n×m n and m represent the number of samples and the feature dimension, respectively; i∈[1,n-T+1] represents the starting position of each sliding window; T is the sample lag order; We obtain a three-dimensional tensor after time panelization: Among them, X 3D This is the three-dimensional tensor corresponding to the input two-dimensional matrix X.

4. The dam displacement prediction method considering time delay characteristics according to claim 1, characterized in that, In the step of plotting the learning error waveform and evaluation error waveform of the network based on the loss between the predicted displacement and the actual displacement output by the network, the loss function is specifically as follows: Where, f(X) 3D ;T max ) i Y is the predicted displacement of the i-th sample output by the network. i The true displacement of the i-th sample, T max This represents the upper limit of the time delay.

5. The dam displacement prediction method considering time delay characteristics according to claim 1, characterized in that, In the step of obtaining the path similarity distance between two waveforms, the formula for calculating the path similarity distance is as follows: Where α is a non-decreasing continuous function from [0,1] to [a,b], β is a non-decreasing continuous function from [0,1] to [a',b'], and d(·) represents the Euclidean distance.

6. A dam displacement prediction system considering time delay characteristics, characterized in that, include: The preprocessing module is used to acquire and preprocess the displacement observation data of the target dam to form a displacement observation sequence and determine the upper limit of the time delay in the displacement observation sequence. The time feature insertion module constructs a displacement prediction dataset for the target dam based on displacement observation sequences and environmental variable factors affecting dam displacement, and performs time panelization on the data in the displacement prediction dataset according to the upper limit of time delay. The model building module is used to build a deep neural network based on gated recurrent units and establish time delay parameters that consider the upper limit of time delay. The displacement prediction dataset after time panelization is used as input. The learning error waveform and evaluation error waveform of the network are plotted based on the loss between the predicted displacement and the actual displacement output by the network. The path similarity distance between the two waveforms is obtained. The time delay parameters are traversed and the optimal time delay parameter with the minimum path similarity distance is selected. The prediction model training module is used to train a deep neural network based on gated recurrent units using the displacement prediction dataset after time panelization, based on the optimal time delay parameters, to obtain a trained dam displacement prediction model. The dam displacement prediction module uses a trained dam displacement prediction model to predict the displacement of the target dam. Specifically, the method for determining the upper limit of time delay in the displacement observation sequence is as follows: The autoregressive correlation coefficients at each displacement monitoring node were calculated as follows: μ=E(V t ); s 2 =D(V t )=E((V t -m) 2 ); Among them, V t =(v1,v2,…v t |t∈[1,N]) represents the displacement observation sequence of the corresponding displacement monitoring node, where N represents the total length of the sequence, and μ and σ 2 These are the mean and variance of the displacement observation sequence, respectively; k Here, is the autocovariance, k is the penalty coefficient for autonomous selection, and AC(k) is the autoregressive correlation coefficient of the corresponding displacement observation sequence. Traverse the displacement monitoring nodes to obtain the corresponding autoregressive correlation coefficient sequence, calculate the confidence interval of the autoregressive correlation coefficient sequence with a confidence level greater than a set threshold, and use the number of autoregressive correlation coefficients falling outside the confidence interval as the upper limit of time delay. Among them, the environmental variable factors affecting dam displacement include the water pressure component δ H Temperature component δ T and the time-dependent component δ θ ; Among them, the water pressure component denoted as z-th power of the average daily reservoir water level upstream of the target dam, where a1, a2, a3, and c are the regression coefficients of the preset water pressure components; Temperature component δ T =b1T 1-2 +b2T 3-7 +…b 13 T 301-365 ;T p-q This represents the segmented average temperature from day p to day q prior to the current observation date; b1, b2, b 13 The regression coefficients for the preset temperature components; Time-dependent component δ θ =c1θ+c2lnθ; θ=t / 100, where t represents the total number of days from the initial monitoring date to the current monitoring date; c1 and c2 are the preset regression coefficients for the deformation of the time-dependent components.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the dam displacement prediction method considering time delay characteristics as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the dam displacement prediction method considering time delay characteristics as described in any one of claims 1 to 5.

Images