Dam seepage pressure online prediction method and device

Abstract

The invention provides a dam seepage pressure online prediction method and device, and the method comprises the steps: obtaining multi-channel monitoring data sequences of dam seepage pressure, and each monitoring data sequence corresponds to a channel; performing single-channel pressure prediction on the monitoring data sequence, and generating a first prediction result according to a single-channel prediction result; inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result; respectively calculating weights of the first prediction result and the second prediction result; and performing weighted summation on the first prediction result and the second prediction result by using the weight to obtain a seepage pressure online prediction result. By means of the scheme, the accuracy of the prediction result can be improved, online prediction of the seepage pressure of the dam is achieved, and then a theoretical guidance basis is provided for operation management of the dam.

Description

Technical Field

[0001] The invention relates to the field of dam seepage safety monitoring in water conservancy and hydropower engineering, and in particular to an online prediction method and device for dam seepage pressure. Background Art

[0002] Dam seepage safety is directly related to the stable operation of engineering structures. Seepage pressure, as a key indicator in dam seepage safety analysis, is often used to evaluate dam seepage safety. Therefore, accurate prediction of seepage pressure provides guidance for formulating dam safety operation and maintenance strategies. With the development of automation systems, seepage pressure is often obtained through real-time monitoring of automated monitoring equipment. Therefore, how to effectively realize online prediction of seepage pressure and take effective measures to deal with abnormal situations in a timely manner according to its changing trend is of great significance to ensure the safe and stable operation of dams.

[0003] However, most existing prediction methods use offline training to train models, that is, they assume that the entire seepage pressure data set is available before model training, and implicitly assume that the relationship between input and output variables remains unchanged during the learning process. However, seepage pressure monitoring data usually changes dynamically, and the seepage prevention capacity of the dam's anti-seepage system will also change during dam operation. Therefore, when the model learned from historical data is applied to the prediction of new data, the prediction results of the model are poor. Therefore, how to effectively perform online prediction of seepage pressure is an urgent problem to be solved. Summary of the invention

[0004] The present invention provides a method and device for online prediction of dam seepage pressure, so as to improve the accuracy of prediction results, realize online prediction of dam seepage pressure, and further provide a theoretical guidance basis for dam operation management.

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

[0006] A method for online prediction of dam seepage pressure, the method comprising:

[0007] Obtain a multi-channel monitoring data sequence of the dam seepage pressure, where each monitoring data sequence corresponds to one channel;

[0008] Performing single-channel pressure prediction on the monitoring data sequence, and generating a first prediction result according to the single-channel prediction result;

[0009] Inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result;

[0010] Calculating the weights of the first prediction result and the second prediction result;

[0011] The first prediction result and the second prediction result are weightedly summed using the weight to obtain an online prediction result of the seepage pressure.

[0012] Optionally, the multi-channel monitoring data sequence for obtaining the dam seepage pressure includes:

[0013] Multiple sensors are used to monitor the dam seepage pressure in real time to obtain a multi-channel monitoring data sequence of the dam seepage pressure.

[0014] Optionally, performing single-channel pressure prediction on the monitoring data sequence to obtain a first prediction result includes:

[0015] Input each monitoring data sequence into the pre-established single-channel pressure prediction model in turn for independent prediction to obtain a single-channel prediction result;

[0016] The single-channel prediction results are connected in series to obtain the first prediction result.

[0017] Optionally, the single-channel pressure prediction model is a TCN network.

[0018] Optionally, the multi-channel pressure prediction model is a STGCN network;

[0019] The input of the STGCN network includes characteristic vectors of multiple monitoring data sequences and corresponding adjacency matrices, and the output of the STGCN network is seepage pressure at multiple time steps.

[0020] Optionally, the STGCN network includes two spatiotemporal convolution blocks and an output layer.

[0021] Optionally, the method further includes: constructing the adjacency matrix by using a DTW method.

[0022] Optionally, calculating the weights of the first prediction result and the second prediction result includes:

[0023] Calculating the long-term weight and the short-term weight of the first prediction result and the second prediction result respectively;

[0024] The combined weight of the first prediction result and the second prediction result is calculated according to the long-term weight and the short-term weight.

[0025] Optionally, calculating the long-term weights of the first prediction result and the second prediction result includes: calculating the long-term weights of the first prediction result and the second prediction result using an EGD method.

[0026] An online prediction device for dam seepage pressure, comprising:

[0027] A data acquisition module is used to acquire a multi-channel monitoring data sequence of the dam seepage pressure, each monitoring data sequence corresponds to one channel;

[0028] A single-channel prediction module, used to perform single-channel pressure prediction on the monitoring data sequence, and generate a first prediction result according to the single-channel prediction result;

[0029] A multi-channel prediction module, used for inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result;

[0030] A weight determination module, used to calculate the weights of the first prediction result and the second prediction result;

[0031] A calculation module is used to perform weighted summation on the first prediction result and the second prediction result using the weight to obtain an online prediction result of seepage pressure.

[0032] A computer-readable storage medium stores a computer program, which executes the steps of the online prediction method for dam seepage pressure when the computer program is executed by a processor.

[0033] The online prediction method and device for dam seepage pressure provided by the present invention overcomes the limitation of a single model in online time series prediction by introducing a set of models that share different data deviations, and each model is independently trained and updated online, so that the best performance can be obtained from each online model. Then, by dynamically combining the predictions of each model, the overall prediction effect can be effectively improved.

[0034] Furthermore, the TCN model is used to process time dependencies, and the STGCN model is used to process cross-variable dependencies. Each branch focuses on modeling time and cross-variable dependencies, and the reinforcement learning method is used to improve the slow response problem in the exponential gradient descent method to achieve more accurate online prediction of dam seepage pressure. Compared with classic online learning methods (such as exponential gradient descent), the reinforcement learning-based method is more effective in adapting to concept changes / drifts, thereby achieving better performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] 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.

[0036] Figure 1 It is a flow chart of the online prediction method of dam seepage pressure provided by the present invention;

[0037] Figure 2 Schematic diagram of the process of single-channel prediction using TCN network in the embodiment of the method of the present invention;

[0038] Figure 3 Schematic diagram of a structure of the STGCN network in the embodiment of the method of the present invention;

[0039] Figure 4 Schematic diagram of a short-term weight determination process based on reinforcement learning in an embodiment of the method of the present invention;

[0040] Figure 5 It is a structural schematic diagram of the online prediction device for dam seepage pressure provided by the present invention. DETAILED DESCRIPTION

[0041] 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.

[0042] 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.

[0043] In actual dam projects, there are many monitoring sequences. At present, the main ideas for time series prediction can be divided into two categories: some focus on modeling time dependencies, while others focus on modeling the relationship between cross variables. In terms of time dependency modeling, some prediction algorithms focus on variable-independent methods. These algorithms only use information from a single channel / variable as input each time, treat multivariate time series as multi-channel signals, process each time series separately, and concatenate the results to obtain the final prediction results. Studies have shown that variable independence is crucial to improving the robustness of the model under concept drift, which can alleviate noise and distribution drift problems and improve robustness. Although variable independence enhances the robustness of the model, the dependency between cross variables is equally important for prediction. For a specific variable, information from other related sequences may improve the prediction results. In the current study, the use of the HST (hydrostatic-seasonal-time, water pressure-temperature change-time) model to consider environmental factors can effectively improve the prediction accuracy. However, the cross variable method is prone to overfitting and performance degradation. Therefore, how to effectively combine the advantages of time dependency and cross variables in the prediction of dam seepage pressure to further improve the prediction accuracy is an urgent problem to be solved.

[0044] Establishing an accurate online prediction model for seepage pressure is one of the effective means to ensure the safe operation of dams. However, most of the dam seepage pressure prediction models in the existing technology are established in an offline environment. These models are difficult to adapt to complex and changeable real-time monitoring data. Moreover, these models are only based on time dependence or cross-variable dependence, which has the problem of poor accuracy.

[0045] To this end, an embodiment of the present invention provides a method and device for online prediction of dam seepage pressure, which performs single-channel pressure prediction on each channel monitoring data sequence of a multi-channel monitoring data sequence of the dam seepage pressure, generates a first prediction result based on the single-channel prediction result, and performs multi-channel pressure prediction on the multi-channel monitoring data sequence together to obtain a second prediction result, then calculates the weights of the first prediction result and the second prediction result, and uses the corresponding weights to perform weighted summation on the first prediction result and the second prediction result to obtain an online prediction result of the seepage pressure.

[0046] like Figure 1 FIG. 1 is a flow chart of an online prediction method for dam seepage pressure provided by the present invention, comprising the following steps:

[0047] Step 101, obtaining a multi-channel monitoring data sequence of the dam seepage pressure, each monitoring data sequence corresponds to a channel.

[0048] In specific implementation, modern monitoring methods can be used to obtain real-time monitoring data sequences of seepage pressure. For example, multiple sensors are used to monitor the seepage pressure of the dam in real time to obtain a multi-channel monitoring data sequence of the seepage pressure of the dam. Each sensor corresponds to a channel, and the monitoring data is transmitted back to the server in real time, and the server combines the monitoring data sequences of different channels into a multi-channel monitoring data sequence.

[0049] The monitoring data sequence of each channel is a time series of a single variable, and the multi-channel monitoring data sequence is a data matrix composed of time series of multiple variables.

[0050] It should be noted that the position of each sensor can be set as needed, and this embodiment of the present invention does not limit this.

[0051] Step 102: Perform single-channel pressure prediction on the monitoring data sequence to obtain a first prediction result.

[0052] Specifically, multiple monitoring data sequences are separated into individual monitoring data sequences, and then each individual monitoring data sequence is input into a pre-established single-channel pressure prediction model in turn, and each monitoring data sequence is independently predicted to obtain a single-channel prediction result, and then multiple single-channel prediction results are connected in series to obtain a multi-channel prediction result, which is referred to as the first prediction result for the sake of ease of description.

[0053] The single-channel pressure prediction model may adopt a temporal convolutional network (TCN).

[0054] like Figure 2 FIG. 1 is a schematic diagram of a process of using TCN to predict single-channel pressure in an embodiment of the present invention.

[0055] Figure 2 The TCN in can be expressed as:

[0056] TCN=1DFCN+causalconvolutions (1)

[0057] In the formula, 1DFCN is a one-dimensional fully convolutional network, and causal convolutions is a causal convolution, that is, the value of the previous layer at time t only depends on the value of the next layer at time t and before.

[0058] Figure 2 In the above equation, X represents the original multivariate time series (i.e., multi-channel monitoring data series) matrix; N is the number of variables, i.e., the number of monitoring data series, i.e., the number of monitoring channels; T is the time step, i.e., the length of the time series contained in each variable. irepresents the i-th time series separated from the original data matrix X; Represents the single-channel prediction result of the TCN network output corresponding to the i-th time series.

[0059] It should be noted that the single sequence prediction model may adopt an existing related model, or may be obtained by training based on historical monitoring data, which is not limited in the embodiment of the present invention.

[0060] Step 103: input the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result.

[0061] The multi-channel pressure prediction model can adopt a spatio-temporal graph convolutional network (STGCN) to model the spatio-temporal relationship between multiple monitoring data sequences through the STGCN network and determine the cross-variable dependencies of different monitoring data sequences.

[0062] A structural diagram of the STGCN network is shown in Figure 3 As shown, the network input is the feature vector X∈R of M time steps of N monitoring data sequences M×N And the corresponding adjacency matrix A∈R N×N , after two spatiotemporal convolution blocks and an output layer output To predict the seepage pressure in the next H time steps. Represents the output result of the output layer of the model, R n Indicates that the prediction result is an n-dimensional real number vector.

[0063] The adjacency matrix (A) is used to represent the relationship or connection between nodes. For a graph with N nodes, the adjacency matrix is ​​an N×N matrix, and the elements A in the matrix are ij Represents whether there is a connection between node i and node j, and the weight of the connection. Usually, the adjacency matrix is ​​used to capture the spatial dependencies in the graph structure.

[0064] Different from the conventional adjacency matrix establishment method, in the embodiment of the present invention, the dynamic time warping (DTW) technology is introduced to dynamically adjust the connection weights between nodes based on the similarity of time series data, so that the adjacency matrix can simultaneously reflect the temporal and spatial dependencies. Compared with the traditional method that only relies on spatial distance, the embodiment of the present invention can more accurately capture the nonlinear changes in time series data through DTW, thereby dynamically modeling the temporal association between nodes.

[0065] The specific steps of constructing the adjacency matrix using the DTW method are as follows:

[0066] (1) Determine the nodes in the graph;

[0067] (2) Determine the association between nodes.

[0068] Specifically, it can be determined which nodes are connected based on specific metrics (such as distance, correlation, etc.);

[0069] (3) Use dynamic time warping to adjust the connection relationship between nodes.

[0070] Specifically, DTW measures the degree of alignment of different time series data and calculates a dynamic distance matrix to reflect the similarity or difference between the time series data of different nodes.

[0071] (4) The distance matrix is ​​generated by the DTW method. This distance matrix can be used as the basis of the adjacency matrix to represent the temporal similarity between nodes, that is, the element X in the adjacency matrix ij Xi represents the similarity between the i-th node and the j-th node.

[0072] Assume two seepage pressure time series C and Q, C n ={c1,c2,…c n}, Q n ={q1,q2,....q n}, the distance dp(i,j) between the two is calculated through the state transfer equation, that is:

[0073] dp(i,j)=min(dp(i-1,j-1),dp(i-1,j),dp(i,j-1))+d(i,j) (2)

[0074] For graph convolution, Chebyshev polynomial approximation is widely used. Graph convolution can be written as:

[0075]

[0076] In the formula, Representation signal Through the convolution of the graph convolution kernel Θ, x is the signal on the graph (i.e., the node feature vector), θ k are the learnable parameters of the k-th order polynomial; represents the Chebyshev polynomial, is a Chebyshev polynomial of order k, is the normalized graph Laplacian matrix, usually in the form Where L is the Laplacian matrix of the graph, λ max is its maximum eigenvalue, I n is the identity matrix.

[0077] The hierarchical linear formulation can be defined by stacking multiple local graph convolutional layers with a first-order approximation of the graph Laplacian. Therefore, deeper architectures can be constructed to deeply recover spatial information without being restricted by the explicit parameterization given by the polynomial. Based on the scaling and normalization of the neural network, it can be further assumed that λ max ≈2. Therefore, the above graph convolution can be simplified to:

[0078]

[0079] Where θ0 and θ1 are two shared parameters of the kernel function.

[0080] In order to constrain parameters and stabilize numerical performance, θ0 and θ1 are replaced by a single parameter θ, θ0 = -θ1, and A and D are reshaped into and Where A is the adjacency matrix, is the adjacency matrix with self-connection added, D and are the degree matrices before and after the corresponding changes, representing the degree of the nodes in the graph. Then, the graph convolution can be expressed as:

[0081]

[0082] Defined in x∈R n The graph convolution operator G on can be generalized to multi-dimensional tensors. i The signal of the channel Graph convolution can be generalized as:

[0083]

[0084] The temporal convolution layer contains a 1-D causal convolution with a width of k. t kernel, followed by a gated linear unit (GLU) as a nonlinearity. For each node in the graph G, the temporal convolution explores K elements of the input without padding. t neighbors, which results in shortening the length of the sequence by k each time t -1. Therefore, the temporal convolution input of each node can be regarded as a sequence of length M, where C i The channels are Convolution Kernel Aims to map the input Y to a single output element (P,Q are divided into two halves, and the channel size is the same.) Therefore, the time domain gated convolution can be defined as:

[0085]

[0086] Where P and Q are the inputs of each gate in GLU respectively; ⊙ represents the element-by-element Hadamard product.

[0087] Afterwards, the spatial graph convolution and temporal convolution layers form a spatiotemporal convolution block (ST-Conv block) to jointly process the time series of the graph structure and effectively integrate the features of the spatial and temporal domains. Figure 3 As shown in Figure 1, the middle spatial layer is a bridge connecting the two temporal layers, and can achieve fast propagation from graph convolution to spatial state through temporal convolution. Channel C is downscaled and upscaled through the graph convolution layer to achieve scale compression and feature compression. In addition, layer normalization can be used in each ST-Conv block to prevent overfitting.

[0088] The input and output of the ST-Conv block are both 3D tensors. For the input of block l Output The calculation formula is:

[0089]

[0090] In the formula, are the upper and lower temporal kernels in block l, respectively; Θ l is the spectral kernel of graph convolution; ReLU(·) is the activation function.

[0091] After stacking two ST-Conv blocks, an additional temporal gated convolutional layer is added, and finally a fully connected layer is used as the output layer. The temporal convolutional layer maps the output of the last ST-Conv block to a single-step prediction. Then, the final output Z∈R can be obtained from the model n×c , and apply linear transformation to calculate the prediction results of seepage pressure at n nodes (different monitoring points or locations).

[0092] Step 104: Calculate the weights of the first prediction result and the second prediction result.

[0093] In a non-limiting embodiment, the long-term weight w and the short-term weight b of the first prediction result and the second prediction result are calculated respectively, and then the combined weight is calculated according to the long-term weight w and the short-term weight b.

[0094] For example, the exponential gradient descent (EGD) method can be used to calculate the long-term weight w of the results predicted by TCN and STGCN, and reinforcement learning (RL) can be introduced to calculate a set of different short-term weights b to dynamically capture short-term changes in the environment; then, the combined weight is calculated by combining the long-term weight w and the short-term weight b to effectively combine long-term historical information and dynamic changes in the environment.

[0095] EGD is a commonly used method. Specifically, the decision space △ is a d-dimensional simplex, that is, △ = {wt|wt,i≥0, ∥wt∥1=1}, where t is the time step index.

[0096] Given an online data stream x, its prediction target y, and d prediction experts with different parameters f i (x) is the prediction value of the ith expert, and the goal is to minimize the prediction error, that is:

[0097]

[0098] Among them, ω i is the weight of the i-th expert at time step t.

[0099] According to the EGD algorithm, select As the center point of the simplex, is the loss of i at time step t, then each ω i The update rule is:

[0100]

[0101] in, is the normalizer.

[0102] The algorithm has a regret bound, which is the gap between the current strategy and the optimal strategy. For T>2log(d), the regret bound of time step t=1,...T is denoted as R(T), let Then the update strategy has external regret.

[0103]

[0104] in, Indicates that at time step t, the current strategy w t The loss function of It means finding the optimal fixed policy u and calculating the minimum total loss from time step 1 to T; w t,i represents the weight of the i-th expert at time step t; f i (x) represents the prediction result of the i-th expert for the input data x; y represents the true target value of input x; It represents the upper bound of the external regret, which is calculated by the time step T and the number of experts d. It is usually used in online learning algorithms to measure the gap between the algorithm and the optimal strategy.

[0105] By considering short-term information, lower regret can be obtained in a short time interval. In an embodiment of the present invention, this challenge of online learning can be solved by offline reinforcement learning.

[0106] In addition, a set of different weights b are introduced, which are called short-term weights for the convenience of description, so as to better capture the latest performance of a single model.

[0107] Specifically, you can use Figure 4 The framework shown in the figure implements reinforcement learning through supervised learning. Specifically, the results of different network training are weighted and summed to generate prediction results, and a supervised learning framework is constructed in combination with real monitoring results. At time step t, the goal is to learn a short-term weight based on the long-term weight w and the historical performance of the expert in a short period of time I = [1, t]. The process can be expressed as minimizing the prediction error through supervised learning, and the prediction result optimizes the weighted sum of the prediction results of each prediction network. The weighting process includes a long-term weight ω i and a short-term weight b i ,Right now:

[0108]

[0109] In order to simplify the calculation and improve efficiency, l can be set to t-1. The agent can use the rl (θ rl Parameters representing the policy function) parameterized policy Select an operation. is a parameter θ rl The policy function in reinforcement learning is used to select actions based on the current state. The parameter θ of the policy function here is rl Can be obtained through training, representing the parameters in the reinforcement learning algorithm; b t The state represents a function that represents the state at time step t. The state function can be expressed by the historical weight w t,i Experts’ predictions and historical time period I; Y represents the historical true label or target value.

[0110] During training, each prediction can be compared with the expert weights The product between and is connected with the result y as the conditional input. Then the short-term weight and the final combination weight are:

[0111] and

[0112] The network is trained by minimizing the prediction error caused by the new weights, i.e.

[0113]

[0114] In the actual online prediction process, as the concept drift gradually changes, using wt-1 +b t-1 to generate predictions, and train the network after observing the true value results. The multivariate data of the time series input is input to two independent predictors, a temporal dependency predictor f1 and a cross-variable dependency predictor f2. Each predictor contains an encoder and a prediction head, which produce different but complementary inductive biases for the prediction task. Then, the optimal combination weights are learned using EGD improved by reinforcement learning. Specifically, EGD is used to update the long-term weight w of each prediction i , and use offline reinforcement learning to learn additional short-term weights b i , then the final combined weight is w i ←w i +b i .

[0115] Step 105: perform weighted summation of the first prediction result and the second prediction result using the weight to obtain an online prediction result of the seepage pressure.

[0116] Specifically, by weighted summing the combined weight with the prediction results of the two independent models, TCN and STGCN, the online prediction results of the seepage pressure can be obtained, that is, the prediction results of the seepage pressure of multiple monitoring sequences.

[0117] The online prediction method for dam seepage pressure provided by the embodiment of the present invention overcomes the limitation of a single model in online time series prediction by introducing a set of models that share different data deviations, and each model is independently trained and updated online, so that the best performance can be obtained from each online model. Then, by dynamically combining the predictions of each model, the overall prediction effect can be effectively improved.

[0118] Accordingly, the present invention also provides an online prediction device for dam seepage pressure, such as Figure 5 Shown is a structural schematic diagram of the device.

[0119] The dam seepage pressure online prediction device 500 includes the following modules:

[0120] The data acquisition module 501 is used to acquire a multi-channel monitoring data sequence of the dam seepage pressure, each monitoring data sequence corresponds to one channel;

[0121] A single-channel prediction module 502 is used to perform single-channel pressure prediction on the monitoring data sequence and generate a first prediction result according to the single-channel prediction result;

[0122] A multi-channel prediction module 503, used for inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result;

[0123] A weight determination module 504, configured to calculate the weights of the first prediction result and the second prediction result;

[0124] The calculation module 505 is used to perform weighted summation on the first prediction result and the second prediction result using the weight to obtain an online prediction result of the seepage pressure.

[0125] The specific implementation of each of the above modules can refer to the description in the previous embodiment of the method of the present invention, which will not be repeated here.

[0126] 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.

[0127] 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.

[0128] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways.

[0129] 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 or 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.

[0130] 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.

[0131] 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 method for online prediction of dam seepage pressure, characterized in that: The method comprises: Obtain a multi-channel monitoring data sequence of the dam seepage pressure, where each monitoring data sequence corresponds to one channel; Performing single-channel pressure prediction on the monitoring data sequence, and generating a first prediction result according to the single-channel prediction result; Inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result; Calculating the weights of the first prediction result and the second prediction result; The first prediction result and the second prediction result are weightedly summed using the weight to obtain an online prediction result of the seepage pressure.

2. The online prediction method for dam seepage pressure according to claim 1, characterized in that: The multi-channel monitoring data sequence for obtaining the dam seepage pressure includes: Multiple sensors are used to monitor the dam seepage pressure in real time to obtain a multi-channel monitoring data sequence of the dam seepage pressure.

3. The online prediction method for dam seepage pressure according to claim 1, characterized in that: The performing single-channel pressure prediction on the monitoring data sequence to obtain a first prediction result comprises: Input each monitoring data sequence into the pre-established single-channel pressure prediction model in turn for independent prediction to obtain a single-channel prediction result; The single-channel prediction results are connected in series to obtain the first prediction result.

4. The online prediction method for dam seepage pressure according to claim 3 is characterized in that: The single-channel pressure prediction model is a TCN network.

5. The online prediction method for dam seepage pressure according to claim 1, characterized in that: The multi-channel pressure prediction model is a STGCN network; The input of the STGCN network includes characteristic vectors of multiple monitoring data sequences and corresponding adjacency matrices, and the output of the STGCN network is seepage pressure at multiple time steps.

6. The online prediction method for dam seepage pressure according to claim 5, characterized in that: The STGCN network consists of two spatiotemporal convolutional blocks and an output layer.

7. The online prediction method for dam seepage pressure according to claim 5, characterized in that: The method further comprises: The adjacency matrix is ​​constructed by the DTW method.

8. The online prediction method for dam seepage pressure according to any one of claims 1 to 7, characterized in that: Calculating the weights of the first prediction result and the second prediction result includes: Calculating the long-term weight and the short-term weight of the first prediction result and the second prediction result respectively; The combined weight of the first prediction result and the second prediction result is calculated according to the long-term weight and the short-term weight.

9. The online prediction method for dam seepage pressure according to claim 8, characterized in that: The calculating the long-term weights of the first prediction result and the second prediction result comprises: The long-term weights of the first prediction result and the second prediction result are calculated using the EGD method.

10. An online prediction device for dam seepage pressure, characterized in that the device comprises: A data acquisition module is used to acquire a multi-channel monitoring data sequence of the dam seepage pressure, each monitoring data sequence corresponds to one channel; A single-channel prediction module, used to perform single-channel pressure prediction on the monitoring data sequence, and generate a first prediction result according to the single-channel prediction result; A multi-channel prediction module, used for inputting the multi-channel monitoring data sequence into a pre-established multi-channel pressure prediction model to obtain a second prediction result; A weight determination module, used to calculate the weights of the first prediction result and the second prediction result; A calculation module is used to perform weighted summation on the first prediction result and the second prediction result using the weight to obtain an online prediction result of seepage pressure.

11. 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 method for online prediction of dam seepage pressure according to any one of claims 1 to 9 are executed.

Images