Hydrological time sequence prediction method and system, medium and equipment

Through time-series convolution and rotation matrix coding technology, the shortcomings of existing water level prediction models in prediction accuracy and training speed are solved, more efficient hydrological time series prediction is achieved, and the feature extraction range and utilization rate are expanded.

CN119004034AActive Publication Date: 2024-11-22NANJING COLLEGE OF INFORMATION TECH

Patent Information

Application Number
CN202411162961.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-22
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

The existing water level prediction model has shortcomings in prediction accuracy and training convergence speed, making it difficult to effectively analyze the correlation between different features.

Method used

Through time-sequence convolution, the receptive field is expanded, the information of multiple effective time points is extracted, and the relative position information is encoded using the rotation matrix, and the original data vector is directly rotated, the dependencies between adjacent data are explored, and the original data is fully utilized, and the original data is carried out. Prediction of hydrological time series.

Benefits of technology

It improves the prediction effect, expands the range of feature extraction, improves the prediction accuracy and training speed of the model, and can better utilize the original data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004034A_ABST
    Figure CN119004034A_ABST
Patent Text Reader

Abstract

The invention provides a hydrological time sequence prediction method and system, a medium and equipment. The method comprises the following steps: acquiring a hydrological time sequence; sequentially carrying out time sequence convolution and dimension adjustment on the hydrological time sequence to obtain hydrological features after dimension adjustment; and inputting the hydrological characteristics after dimension adjustment into a pre-trained hydrological prediction model, and outputting a prediction result of the hydrological time sequence. According to the method, the receptive field is expanded through time sequence convolution, so that the feature extraction range is expanded, and the extracted features can cover information of a plurality of effective time points; the relative position information is coded through the rotation matrix, the original data vector is directly rotated, the dependency relationship between adjacent data is excavated, the original data is fully utilized, the position vector is not added, and the prediction effect is good.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological data prediction, and in particular to a method, system, medium and equipment for predicting hydrological time series. Background Art

[0002] Pumping stations are important water conservancy infrastructure projects. Accurate water level prediction can scientifically guide dispatchers to carry out water dispatching, especially the water level prediction of the forebay of pumping stations. Hydrological data such as water level have the characteristics of fast update speed, large data volume and complex types. Mining and analyzing the complex relationships in hydrological data for use in hydrology and flood forecasting has gradually become a focus of attention.

[0003] Traditional water level prediction is based on the physical model of the hydrological environment and artificial experience, which has certain limitations and application conditions. With the continuous advancement of artificial intelligence technology, neural networks can train nonlinear mapping models from the changing patterns of historical data and are widely used in water level prediction, including water levels of water diversion projects, lake levels, and reservoir levels. Existing water level prediction models have the problems of low prediction accuracy, low training convergence speed, and difficulty in analyzing data based on the correlation between different features. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a prediction method, system, medium and equipment for hydrological time series, which expands the receptive field through time series convolution, thereby expanding the feature extraction range, encodes relative position information through a rotation matrix, directly rotates the original data vector, explores the dependency between adjacent data, makes full use of the original data, does not add position vectors, and has a good model prediction effect.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0006] In one aspect, the present invention provides a method for predicting a hydrological time series, comprising:

[0007] Obtain hydrological time series;

[0008] The hydrological time series is sequentially subjected to time series convolution and dimension adjustment to obtain dimension-adjusted hydrological features;

[0009] Inputting the dimensionally adjusted hydrological features into a pre-trained hydrological prediction model, and outputting prediction results of the hydrological time series;

[0010] The processing steps of the hydrological prediction model include:

[0011] Using a rotation matrix to perform relative position encoding on the dimensionally adjusted hydrological features to obtain hydrological features containing position information;

[0012] The encoder is used to perform sparse correlation calculation and feature compression on the hydrological features containing location information to obtain a hydrological feature map;

[0013] A decoder is used to perform masking and sparse correlation calculation on the hydrological features containing the position information to obtain the hydrological features containing the mask, and a sparse correlation calculation is performed on the hydrological features containing the mask and the hydrological feature map to obtain an output feature map;

[0014] The second fully connected layer is used to predict the output feature map to obtain the prediction result of the hydrological time series.

[0015] Optionally, before the hydrological sequence is subjected to time series convolution, the missing values ​​of the hydrological sequence are filled using a close distance interpolation method;

[0016] When the missing value of the hydrological series is a single value, the average of the two values ​​before and after the missing value is taken as the missing value;

[0017] When the missing value of the hydrological series is multi-valued, the formula for the missing value is:

[0018] ;

[0019] In the formula, For the missing values, ; is the value preceding the missing value; is the value following the missing value; is the number of missing values; is the number of values ​​preceding the missing value.

[0020] Optionally, performing time series convolution on the hydrological time series through a time series convolution module to obtain hydrological characteristics;

[0021] The temporal convolution module includes a plurality of residual blocks connected in series, and the expansion rates of the plurality of residual blocks connected in series increase sequentially.

[0022] Optionally, a residual connection is introduced into each residual block, and the calculation formula of the hydrological characteristics is:

[0023] ;

[0024] In the formula, For hydrological characteristics; is the residual mapping; is the hydrological time series; is the first weight; is the second weight.

[0025] Optionally, the calculation formula of the rotation matrix is:

[0026] ;

[0027] In the formula, is the rotation angle.

[0028] Optionally, the calculation formula for the hydrological characteristics containing location information is:

[0029] ;

[0030] In the formula, For the A hydrological characteristic matrix containing location information; For the The hydrological characteristic matrix after dimension adjustment; For the matrix The first component of For the matrix The second component of For the The rotation matrix of the position; For the The rotation angle of the position.

[0031] Optionally, the calculation formula of the sparse correlation is:

[0032] ;

[0033] In the formula, is the query matrix; is the key matrix; is the eigenvalue matrix; is the Softmax function; for and Analyze the most active predecessors through KL divergence indivual ; is the transposed matrix of the key matrix; is the feature vector dimension.

[0034] In a second aspect, the present invention provides a hydrological time series prediction system, comprising:

[0035] Data acquisition module, used to obtain hydrological time series;

[0036] Time series convolution module, used to perform time series convolution on hydrological time series;

[0037] The first fully connected layer is used to adjust the dimension of the hydrological time series;

[0038] An encoder for sparse correlation calculation and feature compression of hydrological features containing location information;

[0039] A decoder, configured to perform sparse correlation calculation of masks on hydrological features containing location information, and to perform sparse correlation calculation of the masked features with the hydrological feature map;

[0040] The second fully connected layer is used to obtain the prediction results of the hydrological time series.

[0041] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting hydrological time series described in the first aspect.

[0042] In a fourth aspect, the present invention provides a computer device, comprising:

[0043] Memory, for storing computer instructions;

[0044] A processor is used to execute the computer instructions to implement the steps of the hydrological time series prediction method described in the first aspect.

[0045] Beneficial Effects

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention expands the receptive field through temporal convolution, thereby expanding the feature extraction range. The extracted features can cover the information of multiple effective time points. The relative position information is encoded by the rotation matrix, the original data vector is directly rotated, the dependency relationship between adjacent data is mined, the original data is fully utilized, no position vector is added, and the prediction effect is good. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 FIG. 1 is a schematic flow chart of a method for predicting hydrological time series in one embodiment of the present invention;

[0049] Figure 2 FIG. 1 is a schematic diagram of the structure of a hydrological time series prediction system in one embodiment of the present invention;

[0050] Figure 3 FIG. 1 is a schematic diagram of the structure of a temporal convolution module in an embodiment of the present invention;

[0051] Figure 4 Shown is a schematic diagram comparing prediction results of different hydrological prediction models in one embodiment of the present invention;

[0052] Figure 5 Shown is a schematic diagram of loss curve comparison of different hydrological prediction models in one embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0054] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0055] Example 1

[0056] like Figure 1 As shown, this embodiment introduces a hydrological time series prediction method, including:

[0057] Obtaining the hydrological time series: In this embodiment, continuous data of a pump station within a certain period of time is obtained.

[0058] The missing values ​​of the hydrological series are filled using the close-distance interpolation method.

[0059] When the missing value of the hydrological series is a single value, the average of the two values ​​before and after the missing value is taken as the missing value.

[0060] When the missing value of the hydrological series is multi-valued, the formula for the missing value is:

[0061] ;

[0062] In the formula, For the missing values, ; is the value preceding the missing value; is the value following the missing value; is the number of missing values; is the number of values ​​preceding the missing value.

[0063] The hydrological time series is sequentially subjected to time series convolution and dimension adjustment to obtain dimension-adjusted hydrological features.

[0064] The hydrological time series is subjected to temporal convolution by a temporal convolution module to obtain hydrological features. In this embodiment, the temporal convolution module may adopt a TCN (Temporal Convolutional Networks) model.

[0065] The temporal convolution module includes a plurality of residual blocks connected in series, and the expansion rates of the plurality of residual blocks connected in series increase in sequence. Figure 3 FIG. 1 is a schematic diagram of the structure of the temporal convolution module. In this embodiment, there are three residual blocks, and their expansion rates d are 1, 2, and 4 respectively.

[0066] Each residual block includes an expansion causal convolution layer, a weight normalization layer, a ReLU activation function layer, a Dropout function layer, and a residual structure layer. Different residual blocks have different expansion rates. The larger the expansion rate, the larger the receptive field and the more features are extracted.

[0067] Each residual block performs the following steps:

[0068] The hydrological time series passes through the dilated causal convolution layer, and the convolution kernels are connected across to expand the receptive field, thereby expanding the range of feature extraction.

[0069] It is then processed through a weight normalization layer to eliminate dimensions and reduce internal covariate shift, thereby accelerating the convergence of the network.

[0070] After passing through the ReLU activation function layer and the Dropout function layer, the ReLU activation function adds nonlinear properties to the residual block, and the Dropout function randomly suppresses certain features to enhance the generalization of the residual block.

[0071] A residual connection is introduced in each residual block, and the calculation formula of the hydrological characteristics is:

[0072] ;

[0073] In the formula, For hydrological characteristics; is the residual mapping; is the hydrological time series; is the first weight; is the second weight.

[0074] The hydrological time series is dimensionally adjusted through the first fully connected layer to obtain the dimensionally adjusted hydrological features.

[0075] The dimensionally adjusted hydrological features are input into the pre-trained hydrological prediction model, and the prediction results of the hydrological time series are output. The difference between the predicted results of each round and the real data in the training set is calculated through the mean absolute error (MAE) loss function and the root mean square error (RMSE) loss function. After knowing the difference, the model parameters are adjusted through the optimizer. After multiple rounds of adjustment, the prediction results are closest to the real value. The hydrological prediction model can adopt the Trpe-Informer model, that is, the Informer model is obtained by relative position encoding through the rotation matrix.

[0076] The processing steps of the hydrological prediction model include:

[0077] The rotation matrix is ​​used to perform relative position encoding on the dimensionally adjusted hydrological features to obtain hydrological features containing position information.

[0078] The rotation matrix in two-dimensional space is to obtain a new vector by rotating the vector angle. The calculation formula of the rotation matrix is:

[0079] ;

[0080] In the formula, is the rotation angle.

[0081] The calculation formula of the hydrological characteristics containing location information is:

[0082] ;

[0083] In the formula, For the A hydrological characteristic matrix containing location information; For the The hydrological characteristic matrix after dimension adjustment; For the matrix The first component of For the matrix The second component of For the The rotation matrix of the position; For the The rotation angle of the position.

[0084] When performing sparse correlation calculation, the correlation results of two tokens depend on nm, that is, they are affected by their relative positions, which can improve the utilization of data relationships. For example, the formula:

[0085] ;

[0086] In the formula, For the A hydrological characteristic matrix containing location information; For the The rotation matrix of the position.

[0087] The hydrological features containing location information can use the location information to determine which data are most closely correlated, or which data are most correlated with the data to be predicted, so as to facilitate subsequent sparse correlation calculations.

[0088] The encoder is used to perform sparse correlation calculation and feature compression on the hydrological features containing location information to obtain a hydrological feature map. Specifically:

[0089] The hydrological features containing location information are added to the sparse attention mechanism of the encoder, and sparse correlation calculation is performed on them. The final result is a feature containing location information, which makes full use of the correlation between different features.

[0090] The features containing location information are compressed through distillation operation to obtain a hydrological feature map.

[0091] The encoder improves its robustness by stacking the above two steps.

[0092] A decoder is used to perform masking operation and sparse correlation calculation on the hydrological features containing position information to obtain hydrological features containing masks, and sparse correlation calculation is performed on the hydrological features containing masks and the hydrological feature map to obtain an output feature map.

[0093] Through masking operation, some feature values ​​are concealed so that they have no effect when the parameters are updated. For a sequence, our decoding output can only depend on the output before time t, but not on the output after time t. Therefore, we need to hide the information after time t through masking operation.

[0094] A sparse correlation calculation is performed between the hydrological features containing the mask and the hydrological feature map to obtain an output feature map.

[0095] The calculation formula of the sparse correlation is:

[0096] ;

[0097] In the formula, is the query matrix; is the key matrix; is the eigenvalue matrix; is the Softmax function; for and Analyze the most active predecessors through KL divergence indivual ; is the transposed matrix of the key matrix; is the feature vector dimension.

[0098] The second fully connected layer is used to predict the output feature map to obtain the prediction result of the hydrological time series.

[0099] Example 2

[0100] like Figure 2 As shown, this embodiment introduces a hydrological time series prediction system, including:

[0101] Data acquisition module, used to obtain hydrological time series.

[0102] The time series convolution module is used to perform time series convolution on hydrological time series.

[0103] The first fully connected layer is used to adjust the dimension of the hydrological time series.

[0104] The hydrological prediction model is used to obtain the prediction results of the hydrological time series. The hydrological prediction model includes an encoder, a decoder and a second fully connected layer.

[0105] The encoder is used to perform sparse correlation calculation and feature compression on the hydrological features containing position information.

[0106] The decoder is used to perform masked sparse correlation calculation on the hydrological features containing location information, and to perform sparse correlation calculation on the masked features and the hydrological feature map.

[0107] The second fully connected layer is used to obtain the prediction result of the hydrological time series.

[0108] Example 3

[0109] like Figure 4 , 5 As shown, this embodiment provides an experimental design of a hydrological time series prediction method, which specifically includes the following steps:

[0110] Continuous data of a pumping station within a certain period of time is obtained to form a hydrological time series. The unit power, pumping station unit speed, flow, outlet pressure, main engine bearing temperature, and upstream water level data of Jinhu Station of the East Line of the South-to-North Water Diversion Project are selected to predict the water level of the pumping station.

[0111] First, use the close-distance interpolation method for preprocessing to ensure that there is data at each time point.

[0112] The data set of this embodiment selects the minute data of the power of the unit of Huai'an Jinhu Station, the speed of the unit of the pump station, the flow rate, the outlet pressure, the main engine bearing temperature, and the upstream water level of the pump station from May 6, 2017 to May 9, 2017, a total of 10,000 sets of data, to train the model and predict the upstream water level of the pump station. The normalization process uses the maximum and minimum normalization method, as follows:

[0113] ;

[0114] In the formula, The data is normalized. is the original data; is the maximum value in the original data; is the minimum value in the original data.

[0115] The training set data of Jinhu Station is input into the temporal convolution module (TCN) and three residual blocks are connected in series, such as Figure 3 As shown in the figure, x represents input, y represents output, and the expansion rates are d=1, 2, and 4 respectively. By expanding the receptive field and performing multi-layer convolution, the feature extraction range is expanded so that the final data can cover multiple valid time point information. Each residual block performs the following steps:

[0116] After receiving the data, the residual block undergoes dilated causal convolution, and the convolution kernel is connected across to increase the receptive field, expand the feature extraction range, and make full use of historical data at different times.

[0117] By normalizing the weights, the dimension is eliminated and the internal covariate shift is reduced, thus accelerating the convergence of the network.

[0118] Through the ReLU activation function and the Dropout function. The activation function adds nonlinear properties to the neuron, and the Dropout function randomly suppresses certain features to enhance the generalization of the model.

[0119] Introducing residual connection, the calculation formula of the hydrological characteristics is:

[0120] ;

[0121] In the formula, For hydrological characteristics; is the residual mapping; is the hydrological time series; is the first weight; is the second weight.

[0122] The hydrological features are input into the fully connected layer, and the output dimension is adjusted to obtain the hydrological features after dimension adjustment.

[0123] The dimensionally adjusted hydrological features are input into a pre-trained hydrological prediction model to output the prediction results of the hydrological time series. The hydrological prediction model continuously trains the back propagation optimization parameters. The hydrological prediction model in this embodiment can be a Trpe-Informer model.

[0124] In this embodiment, the Trpe-Informer model uses the Adam optimizer, the initial learning rate is 0.002, and the learning rate is halved in each epoch. The expansion rate d of the TCN model is 1, 2, and 4 respectively. The input size of the encoder and decoder is set to 6 to receive the six key features of unit power, pump station unit speed, flow, outlet pressure, main engine bearing temperature, and upstream water level of the pump station; the output scale of the encoder and decoder is 1, that is, the upstream water level of the pump station is output.

[0125] In order to verify the prediction effect of the present invention, the Jinhu Pumping Station data was selected to compare with the LSTM, Transformer, and Informer models through the hydrological prediction model of the present invention, namely the Trpe-Informer model. The evaluation index of the model prediction results adopted the MAE loss function and the RMSE loss function. The results are as follows:

[0126] Table 1 Performance comparison of different models

[0127]

[0128] It is not difficult to see from the data comparison in Table 1 that the hydrological prediction model of the technical solution of the present invention has relatively small errors compared with the other three models. Overall, the prediction accuracy of the present invention is higher. Figure 4 The figure shows the comparison of prediction results of different models.

[0129] like Figure 5 As shown in the figure, by comparing the Informer and Trpe-Informer models, the loss value gradually tends to be stable as the number of training times increases during the training process, and the loss curve of the Trpe-Informer model converges faster. This is because the rotation matrix used in the Trpe-Informer model performs relative position encoding, and each position in the sequence is represented by a rotation in the embedding space. Compared with sinusoidal embedding, its training time is faster.

[0130] Example 4

[0131] This embodiment introduces a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the hydrological time series prediction method described in Embodiment 1 or 3 are implemented.

[0132] Example 5

[0133] This embodiment introduces a computer device, including:

[0134] Memory, for storing computer instructions;

[0135] A processor is used to execute the computer instructions to implement the steps of the hydrological time series prediction method described in Example 1 or 3.

[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0137] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0138] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0140] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A prediction method for hydrological time series, It is characterized in that include: Obtain hydrological time series; The hydrological time series is sequentially subjected to time series convolution and dimension adjustment to obtain dimension-adjusted hydrological features; Inputting the dimensionally adjusted hydrological features into a pre-trained hydrological prediction model, and outputting prediction results of the hydrological time series; The processing steps of the hydrological prediction model include: Using a rotation matrix to perform relative position encoding on the dimensionally adjusted hydrological features to obtain hydrological features containing position information; The encoder is used to perform sparse correlation calculation and feature compression on the hydrological features containing location information to obtain a hydrological feature map; A decoder is used to perform masking and sparse correlation calculation on the hydrological features containing the position information to obtain the hydrological features containing the mask, and a sparse correlation calculation is performed on the hydrological features containing the mask and the hydrological feature map to obtain an output feature map; The second fully connected layer is used to predict the output feature map to obtain the prediction result of the hydrological time series.

2. The method for predicting hydrological time series according to claim 1, It is characterized in that Before performing time series convolution on the hydrological sequence, using a close distance interpolation method to fill missing values ​​in the hydrological sequence; When the missing value of the hydrological series is a single value, the average of the two values ​​before and after the missing value is taken as the missing value; When the missing value of the hydrological series is multi-valued, the formula for the missing value is: ; In the formula, For the missing values, ; is the value preceding the missing value; is the value following the missing value; is the number of missing values; is the number of values ​​preceding the missing value.

3. The method for predicting hydrological time series according to claim 1, It is characterized in that Performing time series convolution on the hydrological time series through a time series convolution module to obtain hydrological characteristics; The temporal convolution module includes a plurality of residual blocks connected in series, and the expansion rates of the plurality of residual blocks connected in series increase sequentially.

4. The method for predicting hydrological time series according to claim 3, It is characterized in that A residual connection is introduced in each residual block, and the calculation formula of the hydrological characteristics is: ; In the formula, For hydrological characteristics; is the residual mapping; is the hydrological time series; is the first weight; is the second weight.

5. The method for predicting hydrological time series according to claim 1, It is characterized in that The calculation formula of the rotation matrix is: ; In the formula, is the rotation angle.

6. The method for predicting hydrological time series according to claim 1, It is characterized in that The calculation formula of the hydrological characteristics containing location information is: ; In the formula, For the A hydrological feature matrix containing location information; For the The hydrological characteristic matrix after the dimensions are adjusted; For the matrix The first component of For the matrix The second component of For the The rotation matrix of the position; For the The rotation angle of the position.

7. The method for predicting hydrological time series according to claim 6, It is characterized in that The calculation formula of the sparse correlation is: ; In the formula, is the query matrix; is the key matrix; is the eigenvalue matrix; is the Softmax function; for and Analyze the most active predecessors through KL divergence indivual ; is the transposed matrix of the key matrix; is the feature vector dimension.

8. A hydrological time series prediction system, It is characterized in that include: Data acquisition module, used to obtain hydrological time series; Time series convolution module, used to perform time series convolution on hydrological time series; The first fully connected layer is used to adjust the dimension of the hydrological time series; An encoder for sparse correlation calculation and feature compression of hydrological features containing location information; A decoder, configured to perform sparse correlation calculation of masks on hydrological features containing location information, and to perform sparse correlation calculation of the masked features with the hydrological feature map; The second fully connected layer is used to obtain the prediction results of the hydrological time series.

9. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the hydrological time series prediction method described in any one of claims 1 to 7 are implemented.

10. A computer device, It is characterized in that include: Memory, for storing computer instructions; A processor, configured to execute the computer instructions to implement the steps of the method for predicting a hydrological time series according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Construction method of multivariable long-sequence time sequence prediction model based on Transform framework

    CN114519469A

  • Water quality prediction method based on Transform sparse attention mechanism

    CN116050571A

  • Ship traffic flow long sequence space-time prediction method based on ST-Informer

    CN116821784A

  • Sequenced data processing method and device, and text processing method and device

    US20230244704A1

Cited By

  • Basin flood forecasting method and system based on artificial intelligence

    CN121234163A