Dam osmotic pressure space-time early warning method and device

Abstract

The invention provides a dam osmotic pressure space-time early warning method and device. The method comprises the following steps: collecting original osmotic pressure time sequence data of a plurality of monitoring points in a dam monitoring process; preprocessing the time sequence data; performing characterization learning of a seepage pressure time sequence by using the processed time sequence data of the plurality of monitoring points, and establishing a multi-task prediction model; and establishing an early warning model on the basis of the multi-task prediction model. By means of the scheme, the relation between different seepage pressure monitoring data can be effectively processed, and dam seepage pressure space-time early warning is achieved.

Description

Technical Field

[0001] The invention relates to the field of dam seepage safety monitoring in water conservancy and hydropower projects, and in particular to a dam seepage pressure time-space early warning method and device. Background Art

[0002] The seepage safety of the dam is directly related to the stability of its structure. In order to ensure the safety of the dam, sensor equipment is usually installed inside it to monitor the seepage in real time. With the increase in the monitoring range and the number of equipment, efficient and accurate prediction of seepage pressure has become particularly important. However, in the face of a large amount of monitoring data, how to effectively handle the relationship between different seepage pressure monitoring data and establish an efficient and accurate multi-task seepage pressure prediction model for multiple monitoring points has become an urgent problem to be solved. In addition, in order for on-site operation and maintenance personnel to clearly understand the status of the current point, it is necessary to evaluate whether the prediction result exceeds the expected value. Once the predicted value exceeds the expected range, an early warning must be issued in time to improve the accuracy and real-time nature of the earth-rock dam seepage warning, and further ensure the safe and stable operation of the earth-rock dam.

[0003] At present, the research on seepage pressure time series prediction mainly focuses on the prediction of a single measuring point. However, dam seepage pressure monitoring involves multiple measuring points, and there are many sequences that need to be predicted and analyzed. If a prediction model is built for each measuring point separately, it will lead to a significant increase in computing, storage and time costs, which is difficult to meet the needs of actual engineering. Summary of the invention

[0004] The present invention provides a dam seepage pressure spatiotemporal early warning method and device, so as to effectively process the relationship between different seepage pressure monitoring data, and establish an efficient and accurate multi-task seepage pressure prediction model with multiple monitoring points to realize the dam seepage pressure spatiotemporal early warning.

[0005] To this end, the present invention provides the following technical solutions:

[0006] In one aspect, the present invention provides a spatiotemporal early warning method for dam seepage pressure, the method comprising:

[0007] Collect the original seepage pressure time series data of multiple monitoring points during the dam monitoring process;

[0008] Preprocessing the time series data;

[0009] The processed time series data of multiple monitoring points are used to characterize the seepage pressure time series and establish a multi-task prediction model.

[0010] An early warning model is established based on the multi-task prediction model.

[0011] Optionally, the preprocessing of the time series data includes:

[0012] Preprocessing and cleaning the time series data using linear interpolation;

[0013] The cleaned data were normalized using Z-score normalization.

[0014] Optionally, the method of using the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series and establish a multi-task prediction model includes:

[0015] Using a sliding window to segment the time series data of the multiple monitoring points to obtain subsequences;

[0016] Extracting the features of the subsequences and establishing labels corresponding to each subsequence to obtain a training sample set;

[0017] The training sample set is used to construct a diffusion convolutional neural network to obtain a multi-task prediction model.

[0018] Optionally, different input features in the multi-task prediction model have different weights.

[0019] Optionally, the method further includes: determining a weight of each input feature in the multi-task prediction model based on an attention mechanism.

[0020] Optionally, the diffusion convolutional neural network includes graph convolution and recursive neural network; the graph convolution is used to simulate the spatial relationship of monitoring points; and the recursive neural network simulates time dependency.

[0021] Optionally, the method further includes: constructing an adjacency matrix through Pearson correlation between multiple monitoring sequences to implement graph convolution.

[0022] Optionally, establishing an early warning model based on the multi-task prediction model includes:

[0023] Determining a loss function, wherein the loss function includes an uncertainty trade-off loss function and / or a quantile regression loss function;

[0024] An early warning model is established based on the loss function.

[0025] Optionally, the method further comprises:

[0026] Obtain current seepage pressure time series data of multiple monitoring points;

[0027] According to the current seepage pressure time series data and the early warning model, the dam is monitored and warned in real time.

[0028] On the other hand, the present invention also provides a dam seepage pressure time-space early warning device, the device comprising:

[0029] A data collection module is used to collect the original seepage pressure time series data of multiple monitoring points during the dam monitoring process;

[0030] A preprocessing module, used for preprocessing the time series data;

[0031] A multi-task prediction model building module is used to use the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series and establish a multi-task prediction model;

[0032] The early warning model building module is used to build an early warning model based on the multi-task prediction model.

[0033] Optionally, the device further comprises:

[0034] The early warning module is used to obtain the current seepage pressure time series data of multiple monitoring points, and to perform real-time monitoring and early warning of the dam based on the current seepage pressure time series data and the early warning model.

[0035] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the spatiotemporal early warning method for dam seepage pressure are executed.

[0036] The spatiotemporal early warning method and device for dam seepage pressure provided by the present invention collects original seepage pressure time series data of multiple monitoring points during dam monitoring, pre-processes the time series data, uses the processed time series data of multiple monitoring points to perform representation learning of seepage pressure time series, and establishes a multi-task prediction model; and establishes an early warning model based on the multi-task prediction model.

[0037] Compared with the existing technology, the present invention is based on a multi-task learning framework to perform spatiotemporal prediction of earth-rock dams, which has higher accuracy and reliability. By comprehensively considering the correlation between different tasks, the model can more comprehensively capture the changing trend of earth-rock dam seepage, thereby providing a more accurate prediction of seepage status.

[0038] In addition, the scheme of the present invention adopts an interval prediction method, which not only provides a single prediction value, but also gives a prediction interval at a certain confidence level. This method can better evaluate the potential range of changes in the seepage state and help on-site operation and maintenance personnel make reasonable judgments in different risk scenarios.

[0039] The prediction results not only meet the requirements of the rationality of the monitoring values, but also provide timely and accurate early warning information, which is crucial for the safe operation of earth-rock dams. By providing timely and effective early warning information to operation and maintenance personnel, the model provides an important reference for earth-rock dam operation and maintenance decision-making, helping to take timely measures and effectively ensure the safety of earth-rock dam seepage. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 It is a flow chart of the spatiotemporal early warning method for dam seepage pressure provided by the present invention;

[0042] Figure 2 It is a structural schematic diagram of the diffusion convolutional recurrent neural network provided by the present invention;

[0043] Figure 3 It is a structural schematic diagram of the dam seepage pressure time-space early warning device provided by the present invention. DETAILED DESCRIPTION

[0044] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0046] The research on the prediction of seepage pressure time series in the prior art mainly focuses on the problem that the prediction scheme of a single measuring point is difficult to meet the needs of actual projects. In addition, there are relatively few studies on multi-task spatiotemporal early warning of seepage in earth-rock dams. At the same time, in multi-task prediction, the importance of each task is different, so effectively measuring the importance of each task is an urgent problem to be solved. To this end, an embodiment of the present invention provides a spatiotemporal early warning method and device for dam seepage pressure, which establishes a multi-task prediction model through characterization learning of flow pressure time series, and establishes an early warning model based on the multi-task prediction model. Using this early warning model, the relationship between different seepage pressure monitoring data can be effectively processed, and an efficient and accurate multi-task seepage pressure prediction model with multiple monitoring points can be established to achieve spatiotemporal early warning of dam seepage pressure.

[0047] like Figure 1 As shown, it is a flow chart of the spatiotemporal early warning method for dam seepage pressure provided by the present invention, which comprises the following steps:

[0048] Step 101, collecting original seepage pressure time series data of multiple monitoring points during dam monitoring.

[0049] For example, the dam real-time monitoring technology can be used to obtain the original seepage pressure time series data of the dam, which can be based on sensor technology, Internet of Things technology and computer science technology, etc. Through these technical means, the real-time collection, transmission, storage and analysis of the seepage pressure of earth-rock dams can be realized.

[0050] Step 102: preprocess the time series data.

[0051] For example, linear interpolation may be used to clean the time series data, and then the cleaned data may be standardized using Z-score normalization to convert the original data into high-quality data that can be used for modeling and analysis.

[0052] The above preprocessing process can effectively reduce the impact of noise, extract effective information, and ensure data consistency, thus providing a reliable foundation for subsequent time series analysis and prediction.

[0053] Step 103, using the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series, and establish a multi-task prediction model.

[0054] In an embodiment of the present invention, the multi-task prediction model may adopt a diffuse convolutional neural network, and a diffuse convolutional loop layer may be constructed by combining graph convolution and recurrent neural network to learn spatiotemporal representation. Graph convolution is used to simulate the spatial relationship of monitoring points; recurrent neural network simulates time dependency.

[0055] Specifically, the sliding window technology can be used to divide the time series data of multiple monitoring points into subsequences of fixed length, extract the features of each subsequence and establish labels corresponding to each subsequence to obtain a training sample set. The training sample set is used to construct a diffusion convolutional neural network to obtain a multi-task prediction model.

[0056] During the training process, the model learns how to predict the future values ​​of the pressure time series based on the input features; the performance of the model is evaluated using the validation set or test set, and the model parameters are adjusted based on the evaluation results to optimize the model performance.

[0057] In order to simultaneously consider the relationship between different tasks, effectively integrate the information of different tasks, and improve the accuracy and comprehensiveness of feature extraction. In a non-limiting embodiment, the improved diffusion convolutional recurrent neural network can be used to comprehensively consider the relationship between time and space, effectively learn the spatiotemporal representation of seepage pressure, and then perform multi-task prediction of dam seepage pressure. Each prediction task corresponds to a monitoring point, that is, for each monitoring point, there is a prediction result after learning the spatiotemporal representation, and the prediction result is used to represent the seepage pressure of the corresponding monitoring point within a certain period of time in the future.

[0058] It should be noted that the input of the multi-task prediction model is the time series data of the permeability pressure of multiple monitoring points in a certain historical time period, and the output is the time series data of the permeability pressure of each monitoring point in a certain future time period.

[0059] Multivariate time series forecasting based on multi-task learning aims to learn a set of functions for m variables Using this function, the output A good approximation of the actual label y i Since the function F(x) is embedded in the multi-task learning model, the learning task associated with the variable m is Execute simultaneously, using each task Effective knowledge in the model can effectively improve the performance of the model.

[0060] Furthermore, taking into account the importance of different time points to the prediction results and enabling the multi-task prediction model to pay more attention to the key information in the input data, in an embodiment of the present invention, the weight of each input feature in the multi-task prediction model can also be determined, that is, different input features can have different weights.

[0061] For example, in a non-limiting embodiment, the weight of each input feature in the multi-task prediction model can be determined based on an attention mechanism. The attention mechanism learns the weight of each input element and dynamically assigns weights to emphasize different parts of the input data, so that the model can selectively process and utilize information to form an attention distribution, thereby more accurately predicting the spatiotemporal seepage pressure.

[0062] For example, the following temporal attention formula can be used to determine the weight of each input feature:

[0063]

[0064] Q (Query): query matrix, used to represent the input that needs attention; K (Key): key matrix, which can be some representation of the input sequence; V (Value): value matrix, which stores the actual value information of each position in the sequence; K Tis the transpose of the key matrix, and its purpose is to calculate the similarity score between each query and the key by multiplying it with the query matrix Q; is a scaling factor, d is the dimension of query and key; the softmax operation converts the similarity score into a probability distribution, indicating how much attention each position pays to all other positions in the input sequence (attention weight).

[0065] It should be noted that the above formula is only an exemplary description. In specific implementation, other calculation formulas based on the attention mechanism may also be used, and the present invention does not limit this.

[0066] In an embodiment of the present invention, a diffusion convolutional recurrent neural network can be used to model the spatiotemporal relationship, such as Figure 2 The figure shows a structural diagram of the diffusion convolutional recurrent neural network. The input data is processed by the diffusion convolutional recurrent layer, and the convolution expansion is used to sense and capture the spatiotemporal dependency. After being processed by the ReLU activation function, a compressed feature representation is generated. The final features are assigned to different tasks to achieve multi-task learning, and the performance of multiple tasks is improved by sharing features.

[0067] In an embodiment of the present invention, the diffusion convolutional neural network may include a graph convolution and a recursive neural network; the graph convolution is used to simulate the spatial relationship of monitoring points; and the recursive neural network simulates time dependence.

[0068] Specifically, the adjacency matrix can be constructed through the Pearson correlation between multiple monitoring sequences to implement graph convolution. The specific process is as follows:

[0069] First, the adjacency matrix is ​​constructed through the Pearson correlation between each monitoring sequence, as follows:

[0070]

[0071] The meanings of the parameters in the formula are as follows:

[0072] r: Pearson correlation coefficient, which reflects the degree of linear correlation between two variables.

[0073] X i : The i-th data point of variable X.

[0074] Y i : The i-th data point of variable Y.

[0075] The mean of the variable X.

[0076] The mean of the variable Y.

[0077] The stationary distribution of the diffusion process in the diffusion convolutional loop layer can be expressed as a weighted combination of infinite random walks on a graph structure composed of multiple measurement points and calculated in a closed form as follows:

[0078]

[0079] Among them, α is a parameter that controls the diffusion rate and determines the decay rate of each diffusion step; k is the diffusion step; (1-α) k It is an exponential decay factor, and the contribution gradually decreases as k increases; It is the inverse matrix of the degree matrix of the graph, which represents the normalized weight of the node; W is the adjacency matrix of the graph, which represents the connection relationship between the nodes.

[0080] In practice, one can use a finite number of k steps to truncate the diffusion process and assign a trainable weight to each step.

[0081] For the graph signal X∈R N×P and filter f θ The diffusion convolution operation is defined as:

[0082]

[0083] Where X represents the input signal matrix, which contains N nodes and P features; X :,p represents the p-th column of the input matrix X, i.e. the p-th eigenvector; represents the diffusion convolution operation, θ∈R k*2 is the parameter of the filter; the maximum step size of K diffusion, that is, the maximum number of diffusion steps considered by the convolution layer; θ k,1 ,θ k,2 are the parameters of the convolution filter; and are the transition matrices of the diffusion process and the reverse process, respectively.

[0084] Let the parameter tensor be Θ∈R Q*P*K*2 =[θ] q,p , where Θ q,p,:,: ∈R K*2 represents a convolutional filter parameterized with the Pth input and the qth output.

[0085]

[0086] Where X∈R N×P is the input, H∈R N×Q For output, is the filter, and a is the activation function (such as ReLU).

[0087] Diffused convolutional layers learn representations of graph-structured data and can be trained using stochastic gradient-based methods.

[0088] Afterwards, recurrent neural networks (RNNs) are used to model the temporal dependencies.

[0089] For example, we can use the Gated Recurrent Unit (GRU), which is a simple yet powerful variant of RNN. We replace the matrix multiplication in the GRU with a diffuse convolution, resulting in a diffuse convolutional gated recurrent unit (DCGRU).

[0090]

[0091] H (t) =u (t) ⊙H (t-1) +(1-u (t) )⊙C (t)

[0092] Reset Gate (t) It determines how much information of the hidden state of the previous moment needs to be discarded in order to introduce new information for the current moment; σ is the Sigmoid activation function, which is used to limit the output value between [0,1]; Θ r The parameters of the diffusion convolution filter are used to calculate the reset gate; * diffusion convolution operator, indicating that the diffusion convolution is applied to the input data; the current input X (t) and the hidden state H at the previous moment (t-1) Splicing; b r is the bias term for the reset gate.

[0093] Update Gate (t) Determines how much information in the hidden state at the current moment needs to be retained from the hidden state at the previous moment. Θ u is the parameter of the diffusion convolution filter, used to calculate the update gate; b u is the bias term for the update gate.

[0094] Candidate hidden state C (t) It is the candidate information based on the current input and part of the previous hidden state, used to generate a new hidden state. tanh is the hyperbolic tangent function, and the output value is between (-1,1)(-1,1)(-1,1); Θ C is the parameter of the diffusion convolution filter, which is used to calculate the candidate hidden state; r (t) ⊙H (t-1) To reset the gate, multiply the hidden state of the previous moment by element to adjust the influence of the hidden state of the previous moment; b c is the bias term of the candidate hidden state.

[0095] Hidden state H (t)It is the result of updating the gate control, and its hidden state H at the previous moment (t-1) and the current candidate hidden state C (t) strike a balance between.

[0096] In the above formula, X (t) and H (t) Represent the input and output at time t, r (t) and u (t) They are the reset gate and update gate at time t respectively. Represented as diffuse convolution, Θ r , Θ u ,Θ C For the corresponding parameters.

[0097] The feature extraction of spatial dimension is performed through Graph Convolutional Networks (GCN), and the extracted feature Z can be expressed as:

[0098]

[0099] X is the input of the GCN module; X∈R N×C ; Z is the output of the GCN module; Z∈R N×F ; A is the adjacency matrix of the graph; A∈R N ×N ; N is the number of nodes; C is the input dimension hyperparameter set during network training; F is the output dimension hyperparameter set during network training; D is a diagonal matrix containing the number of connections for each node in the graph; Θ is the network parameter that needs to be learned during model training; b is the network parameter of the model bias term; I N is the identity matrix.

[0100] Similar to GRU, DCGRU (Diffusion convolution + GRU) can be used to build recurrent neural network layers and train them through time back propagation.

[0101] Finally, the encoder part of the prediction model is predicted using an improved scattered convolutional recurrent neural network to effectively learn the representation of the sequence, and the learned representation is used for multi-task prediction.

[0102] Step 104: establishing an early warning model based on the multi-task prediction model.

[0103] The early warning model is established by modifying the loss function based on the multi-task prediction model. The input of the expected model includes: the prediction results of each monitoring point and the residual with the true value, and the output is an interval. For example, if the result of the previous point prediction is 1, then the result of the early warning model is [0.5, 1.5] which is an interval.

[0104] In some embodiments, an early warning model may be established based on an uncertainty trade-off loss function and / or a quantile regression loss function.

[0105] By introducing the uncertainty loss function, the confidence is adjusted dynamically during the prediction process to balance the prediction error and uncertainty; quantile regression is used to capture data trends under different risk conditions and predict results at different quantiles; the total loss function of the model can use any of these functions or integrate the two parts, and the model is trained by minimizing the loss function through the optimizer. The trained model can not only provide accurate predictions, but also provide risk warnings based on the uncertainty of the prediction results and the outputs of different quantiles, which is suitable for decision-making and response in different risk and uncertainty scenarios.

[0106] Since each monitoring point is regarded as a prediction task, each prediction task has different importance, and a loss function improved by uncertainty trade-off loss is used for prediction.

[0107] Due to different functions, each monitoring point may have a different demand distribution pattern. If these differences are ignored and the prediction of all monitoring points is regarded as single-task learning, local errors may accumulate. On the contrary, if monitoring points with different distribution patterns are subjected to multi-task learning, the predictions of monitoring points with different distribution patterns can share knowledge and improve generalization performance.

[0108] In the embodiment of the present invention, the total number of monitoring points is assumed to be M, that is, the total number of learning tasks. For each learning task, a fully connected neural network (FNN) can be used to complete the final prediction. For any monitoring point m = 1, 2, ... M, the prediction can be performed by the following formula:

[0109]

[0110] Among them, H m represents the spatiotemporal learning component output of the mth monitoring point, Represents the prediction result. For the convenience of annotation, let W m and b m Denote all learnable weight parameters and bias parameters in the network, respectively. A straightforward approach to multi-task learning is to combine the losses of each learning task through a simple weighted linear sum.

[0111] However, the performance of this naive approach is dependent on the weight W m The choice of is highly sensitive, and adjustment of these weights is difficult and expensive.

[0112] In order to solve the above problems, the loss function improved by uncertainty trade-off loss can be used for prediction.

[0113] The following formula is used to express the prediction results of all M monitoring points:

[0114]

[0115] The model has M outputs, corresponding to M learning tasks.

[0116] Assume the input is x and the model output is f w,b (X), Y is the model output f w,b (X) is the mean, σ 2 is a normal distribution with a variance of .

[0117] The multi-tasking possibilities are described as:

[0118]

[0119] Among them, σ m is the observation noise parameter of the model, which represents the uncertainty of the mth task.

[0120] The log-likelihood value of the model parameters and the observed noise parameters is maximized to obtain the minimization target, that is, the multi-task loss function, which is expressed as:

[0121]

[0122] in, Represents the loss of the mth task. The coefficient term It can be regarded as the relative weight of the mth task. m The increase of The weight of is reduced. In addition, logσ1...σ m acts as a regularizer to prevent σ from increasing too much. In practice, the network can be trained to predict the log variance, s:=logσ.

[0123] In addition, the loss function plays a key role in the result generation of the machine learning model. In the embodiment of the present invention, the conditional expectation or median output can be generated based on the classic square error loss of least squares regression. The loss function of the interval prediction model is constructed using the idea of ​​quantile regression. The conditional τth quantile of the target distribution is composed of the weighted sum of multiple individual predictions. Quantile loss The weight coefficients are estimated by minimizing the derivative of the absolute residual.

[0124]

[0125] Where τ∈(0,1) is the target quantile. The larger the τ quantile, the greater the underestimation loss and the smaller the overestimation loss in terms of quantile loss. Represents the predicted value, y i Represents the true value.

[0126] As can be seen from the definition, the loss function is asymmetrically weighted. Greater than the measured value y i , the penalty is multiplied by (1-τ); otherwise, it is multiplied by τ.

[0127] The upper and lower bounds of the quartiles correspond to the confidence intervals, and the quantile loss is used instead of the original square error loss for interval prediction. Specifically, the quantiles of the upper and lower bounds are set to τ u =1-α / 2 and τ l = α / 2, making it equivalent to the confidence level. However, when the prediction error is zero, the original quantile loss is not differentiable, which means that Therefore, the Huber norm combining L1 and L2 norms is used to correct the loss function.

[0128]

[0129] In the formula, Represents the predicted value, y i represents the true value, δ>0, and determines whether the Huber norm tends to the L1 norm or the L2 norm. Compared with the original quantile loss, the redefined loss function is differentiable everywhere (that is, the new loss function is smooth over its entire domain). Combining the above loss function with the proposed deep learning point prediction model, an interval prediction model based on combined deep learning can be constructed. In this way, the accuracy and stability of the prediction results can be more comprehensively evaluated, providing more reliable information for decision makers.

[0130] Accordingly, the above-mentioned early warning model can be used to implement monitoring and early warning of the dam. Specifically, the current seepage pressure time series data of multiple monitoring points are obtained; according to the current seepage pressure time series data and the early warning model, when the prediction result exceeds the early warning interval, the dam is monitored and warned in real time.

[0131] In the embodiment of the present invention, by introducing the uncertainty trade-off loss function, not only the accuracy of the predicted value is focused on, but also the confidence of the model in the result is considered. By dynamically adjusting the uncertainty parameter, the model makes a trade-off between prediction accuracy and uncertainty to avoid overfitting unstable data. By introducing the quantile regression loss function, the values ​​of different quantiles can be predicted to capture the data trends under different risk levels, which enables the model to generate corresponding prediction results under different risk scenarios, thereby better adapting to diverse actual situations.

[0132] When building the model, the uncertainty loss and quantile loss are combined into a multi-objective loss function, which is minimized by training the model and updating the weight parameters. After evaluation and testing, the model can issue warnings based on the prediction results in actual applications. If the model detects a quantile result with high uncertainty or high risk, it will issue a warning. During the deployment process, the model is continuously iterated and optimized, and continuously updated based on new data to ensure the real-time and accuracy of the warning system.

[0133] Accordingly, the embodiment of the present invention also provides a dam seepage pressure time-space early warning device, such as Figure 3 Shown is a structural schematic diagram of the device.

[0134] The dam seepage pressure time-space early warning device 300 includes the following modules:

[0135] The data collection module 301 is used to collect the original seepage pressure time series data of multiple monitoring points during the dam monitoring process;

[0136] A preprocessing module 302, used for preprocessing the time series data;

[0137] A multi-task prediction model building module 303 is used to use the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series and establish a multi-task prediction model;

[0138] The early warning model building module 304 is used to build an early warning model based on the multi-task prediction model.

[0139] Furthermore, the dam seepage pressure spatiotemporal warning device 300 also includes: an early warning module (not shown in the figure), which is used to obtain current seepage pressure time series data of multiple monitoring points, and perform real-time monitoring and early warning of the dam based on the current seepage pressure time series data and the early warning model.

[0140] The specific implementation of each module in the dam seepage pressure spatiotemporal warning device 300 can refer to the description in the previous embodiment of the method of the present invention, and will not be repeated here.

[0141] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0142] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0143] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways.

[0144] The present invention also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed when it is run. Figure 1 The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The storage medium may also include a non-volatile memory (non-volatile) or a non-transitory memory, etc.

[0145] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless means.

[0146] The embodiments of the present invention are described in detail above. The present invention is described in detail using specific implementation methods herein. The description of the above embodiments is only used to help understand the method and system of the present invention. It is only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should belong to the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A spatiotemporal early warning method for dam seepage pressure, characterized in that: The method comprises: Collect the original seepage pressure time series data of multiple monitoring points during the dam monitoring process; Preprocessing the time series data; The processed time series data of multiple monitoring points are used to characterize the seepage pressure time series and establish a multi-task prediction model. An early warning model is established based on the multi-task prediction model.

2. The dam seepage pressure time-space early warning method according to claim 1, characterized in that: The preprocessing of the time series data comprises: Preprocessing and cleaning the time series data using linear interpolation; The cleaned data were normalized using Z-score normalization.

3. The dam seepage pressure spatiotemporal early warning method according to claim 1, characterized in that: The method of using the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series and establish a multi-task prediction model includes: Using a sliding window to segment the time series data of the multiple monitoring points to obtain subsequences; Extracting the features of the subsequences and establishing labels corresponding to each subsequence to obtain a training sample set; The training sample set is used to construct a diffusion convolutional neural network to obtain a multi-task prediction model.

4. The dam seepage pressure spatiotemporal early warning method according to claim 3 is characterized in that: Different input features in the multi-task prediction model have different weights.

5. The dam seepage pressure spatiotemporal early warning method according to claim 4, characterized in that: The method further comprises: The weight of each input feature in the multi-task prediction model is determined based on the attention mechanism.

6. The dam seepage pressure time-space early warning method according to claim 3 is characterized in that: The diffusion convolutional neural network includes graph convolution and recursive neural network; the graph convolution is used to simulate the spatial relationship of monitoring points; the recursive neural network simulates time dependency.

7. The dam seepage pressure spatiotemporal early warning method according to claim 6, characterized in that: The method further comprises: Through the Pearson correlation between multiple monitoring sequences, an adjacency matrix is ​​constructed to achieve graph convolution.

8. The dam seepage pressure spatiotemporal early warning method according to claim 1, characterized in that: The establishing of the early warning model based on the multi-task prediction model comprises: Determining a loss function, wherein the loss function includes an uncertainty trade-off loss function and / or a quantile regression loss function; An early warning model is established based on the loss function.

9. The dam seepage pressure time-space early warning method according to any one of claims 1 to 8, characterized in that: The method further comprises: Obtain current seepage pressure time series data of multiple monitoring points; According to the current seepage pressure time series data and the early warning model, the dam is monitored and warned in real time.

10. A dam seepage pressure time-space early warning device, characterized in that: The device comprises: A data collection module is used to collect the original seepage pressure time series data of multiple monitoring points during the dam monitoring process; A preprocessing module, used for preprocessing the time series data; A multi-task prediction model building module is used to use the processed time series data of multiple monitoring points to perform characterization learning of the seepage pressure time series and establish a multi-task prediction model; The early warning model building module is used to build an early warning model based on the multi-task prediction model.

11. The dam seepage pressure time-space early warning device according to claim 10, characterized in that: The device also includes: The early warning module is used to obtain the current seepage pressure time series data of multiple monitoring points, and to perform real-time monitoring and early warning of the dam based on the current seepage pressure time series data and the early warning model.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the spatiotemporal early warning method for dam seepage pressure described in any one of claims 1 to 9 are executed.

Images