Method, device, equipment and medium for water level prediction of tidal river section

Abstract

The application provides a tidal river reach water level prediction method, device, equipment and medium, wherein the prediction method comprises the following steps: collecting historical data and current period data, the historical data and the current period data both comprise measured runoff data, measured water level data and measured tidal range data; using cross-correlation analysis method to adjust the measured runoff data and the measured tidal range data with the measured water level data as the benchmark; performing EEMD decomposition to obtain respective intrinsic mode functions and residuals; outputting the prediction values of the intrinsic mode functions and the residuals of the water level data of the analysis station in the future period through the LSTM model and superimposing them to obtain the prediction value of the water level of the tidal river reach. The application carries out joint analysis on the water level data, runoff data and tidal range data of the tidal river reach, effectively solves the nonlinear signal processing problem and the low precision problem of the water level of the tidal river reach through EEMD decomposition and the LSTM model, and greatly improves the prediction accuracy of the water level of the tidal river reach.

Description

Technical Field

[0001] This invention belongs to the field of tidal river level forecasting technology, and in particular relates to a method, device, equipment and medium for predicting tidal river levels. Background Technology

[0002] The water level changes in tidal river sections are mainly influenced by two factors: astronomical tides and river runoff. The nonlinear interaction between astronomical tides and river runoff gives the water level in tidal river sections a non-stationary characteristic, and how to accurately predict the water level of tidal river sections is a significant problem. Current non-stationary tidal harmonic analysis methods commonly used for tidal river section water level prediction can only identify a limited number of tidal constituent signals, have limited ability to analyze and process complex non-stationary signals, and rely heavily on future runoff and tidal range forecast data during the prediction phase. This means that the method incorporates errors from runoff and tidal range forecast data into the prediction stage, thus increasing the prediction bias and significantly reducing the prediction accuracy. Summary of the Invention

[0003] In view of this, the present invention aims to overcome the shortcomings of the above-mentioned problems in the prior art, solve the nonlinear signal processing problem and the low accuracy problem in the prediction of water level in tidal river sections, and propose a method, device, equipment and medium for predicting water level in tidal river sections.

[0004] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0005] The first aspect of this invention provides a method for predicting water levels in tidal river sections, comprising the following steps:

[0006] S1. Collect historical data and current time period data, including measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations;

[0007] S2. The time lag between measured water level data and measured runoff data, and between measured water level data and measured tidal range data are calculated using cross-correlation analysis. The measured runoff data and measured tidal range data are then adjusted for lag using the measured water level data as a reference.

[0008] S3. Perform EEMD decomposition on the measured water level data, measured runoff data, and measured tidal range data after hysteresis adjustment to obtain their respective intrinsic modulus functions and residuals;

[0009] S4. Construct an LSTM model. Use the intrinsic modulus function and residuals of the three sets of historical data obtained in step S3 as the training data for the LSTM model and train the LSTM model.

[0010] The intrinsic modulus function and residuals of the three sets of current time period data obtained in step S3 are used as input variables of the LSTM model, and the predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future time periods are output.

[0011] S5. The predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future periods are superimposed to obtain the predicted value of the tidal river section water level.

[0012] Furthermore, the measured runoff data, measured water level data, and measured tidal range data are time-series data of equal length with a time interval of 1 hour.

[0013] Furthermore, the cross-correlation analysis process in step S2 is as follows:

[0014] For two time series data of length J, x = {x1, x2, ..., x...} J} and y={y1,y2,...,y J The cross-correlation expression between the two is as follows:

[0015] ,

[0016] Among them, C xy ( k () is the cross-correlation coefficient between time series data x and y. k It is a time lag value, and k When k is positive, it indicates that the change in x precedes the change in y; when k is negative, it indicates that the change in y precedes the change in x. and These are the means of the time series data x and y, respectively. and These are the standard deviations of the time series data x and y, respectively; when C xy ( k When the absolute value of ) is the largest, the corresponding k is the time lag value of the time series data x and y;

[0017] Based on the cross-correlation expression, the time lag value kQ between measured water level data and measured runoff data is obtained when the time series data are measured water level data and measured runoff data respectively; and the time lag value kR between measured water level data and measured tidal range data is obtained when the time series data are measured water level data and measured tidal range data respectively.

[0018] Furthermore, the process of lag-adjusting the measured runoff data and measured tidal range data in step S2 is as follows:

[0019] For time series data of length T before lag adjustment, the lag adjustment method is as follows: Let the time series data of measured water level be Z, the time series data of measured runoff be Q, and the time lag between the measured water level data and the measured runoff data be kQ. Then, take the time series data of the measured runoff data as Q = {Q1, Q2, ..., Q...} T-|kQ|}, the time series data of the measured water level Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T}, as the adjusted time series data of measured runoff and measured water level data; let R be the time series data of measured tidal range data, and kR be the time lag between measured water level data and measured tidal range data, then take the time series data of measured water level data Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T}, the time series data of measured tidal range R={R |kQ|+1-|kR| ,R |kQ|+2-|kR| ,...,R T-|kR|}, which is the time series data after adjustment of measured water level data and measured tidal range data, wherein the subscript numbers of time series data Z, Q, and R increase sequentially in the corresponding time series data, and the size of the subscript number in each time series data indicates the relationship between the time of the value of a single data in the corresponding time series data;

[0020] According to the lag adjustment method, when the measured runoff data and measured tidal range data are adjusted with lag adjustment based on the measured water level data, the value range of the measured water level data remains unchanged during the adjustment process, and the measured water level data, measured runoff data and measured tidal range data after lag adjustment are time series data of equal length.

[0021] Furthermore, in step S3, the three sets of data—the hysteresis-adjusted measured water level data, the measured runoff data, and the measured tidal range data—are decomposed using EEMD. The EEMD decomposition process is as follows:

[0022] A series of N repeated trials are conducted. Random white noise is added to the time series data x(t) in the i-th trial (i=1,2,...,N), expressed as follows:

[0023] ,

[0024] Where, x i (t) represents the time series data after adding random white noise, w i (t) is a random white noise sequence;

[0025] For x i (t) Perform empirical mode decomposition to obtain a set of eigenmode functions and residuals, expressed as follows:

[0026] ,

[0027] Among them, IMF m (i) Let v be the m-th order (m=1,2,...,M) eigenmode function decomposed in the i-th trial. i (t) is the x calculated in the i-th trial. i The residual of (t);

[0028] The results of N trials are averaged, and the eigenmode functions of the same order are averaged to obtain the eigenmode function components of the EEMD decomposition.

[0029] ,

[0030] Finally, the reconstructed form of the time series data x(t) after EEMD decomposition is obtained, and the expression is:

[0031] ,

[0032] Among them, IMF m X(t) is the m-th (m=1,2,...,M) eigenmode function of x(t), and X(t) is the residual of x(t) after EEMD decomposition.

[0033] Furthermore, in step S4, the intrinsic modulus function and residuals of the three sets of current time period data obtained in step S3 are used as input variables and input into the trained LSTM model. In the LSTM model, the temporal dependencies and long-term hydrological patterns are extracted step by step through two LSTM layers, and then mapped to the prediction dimension through a fully connected layer. Finally, the predicted values ​​of the intrinsic modulus function and residuals of the water level data in the future time period are output through the output layer.

[0034] Furthermore, the upstream hydrological station is a hydrological station for observing runoff in the upstream of the tidal river section, the analysis station is a station for predicting water level in the tidal river section, and the downstream reference station is a tide gauge station close to the open sea for actual tidal range data. Actual tidal level data is obtained at the downstream reference station, and the actual tidal range data is obtained by calculating the difference between high tide and low tide in the actual tidal level data of the downstream reference station.

[0035] A second aspect of the present invention provides a device for predicting the water level of a tidal river section, used to implement the above-mentioned method, comprising:

[0036] The data acquisition module is used to collect historical data and current time period data. Both historical data and current time period data include measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations.

[0037] The first processing module is used to calculate the time lag between measured water level data and measured runoff data, as well as the time lag between measured water level data and measured tidal range data, using cross-correlation analysis, and to perform lag adjustment on the measured runoff data and measured tidal range data based on the measured water level data.

[0038] The second processing module is used to perform EEMD decomposition on the hysteresis-adjusted measured water level data, measured runoff data, and measured tidal range data to obtain their respective intrinsic modulus functions and residuals.

[0039] The third processing module is used to construct the LSTM model. It uses the intrinsic modulus functions and residuals of the three sets of historical data after lag adjustment and decomposition by EEMD as the training data for the LSTM model. It also uses the intrinsic modulus functions and residuals of the three sets of current time data after lag adjustment and decomposition by EEMD as the input variables of the LSTM model, and outputs the predicted values ​​of the intrinsic modulus functions and residuals of the water level data of the analysis station in future time periods.

[0040] The prediction module is used to superimpose the predicted values ​​of the intrinsic modulus function and residuals of the water level data from the analysis station for future periods to obtain the predicted value of the tidal river section water level.

[0041] A third aspect of the present invention provides an electronic device, comprising:

[0042] One or more processors;

[0043] Memory, used to store one or more programs;

[0044] When one or more programs are executed by one or more processors, the one or more processors perform the above methods.

[0045] A fourth aspect of the present invention provides a computer-readable storage medium comprising:

[0046] It stores computer-executable instructions that, when executed, are used to implement the methods described above.

[0047] Compared with the prior art, the present invention has the following advantages:

[0048] The tidal river level prediction method of this invention uses cross-correlation analysis to calculate the time lag between measured runoff data and measured runoff data, and between measured water level data and measured tidal range data, respectively. Then, it performs lag adjustment on the measured runoff data and measured tidal range data based on the measured water level data. Finally, it performs EEMD decomposition on the lag-adjusted data to obtain their respective intrinsic modulus functions and residuals, which are then input into an LSTM model to predict water level data for future periods. The method predicts the intrinsic modulus function and residuals of the tidal range, and finally superimposes the predicted values ​​of the intrinsic modulus function and residuals of the water level data for future periods to obtain the predicted value of the tidal range water level. This method focuses on the tidal range water level and the core factors affecting water level changes, namely runoff data and tidal range data. By adopting the EEMD decomposition method with high decomposition stability and good mode confusion suppression, and the LSTM model with high prediction accuracy, it effectively solves the nonlinear signal processing problem and the low accuracy problem of tidal range water level, thereby greatly improving the prediction accuracy of tidal range water level. Attached Figure Description

[0049] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0050] Figure 1 This is a flowchart of the tidal river level prediction method according to Embodiment 1 of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0052] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0053] Example 1

[0054] like Figure 1 As shown, the method for predicting water levels in tidal river sections includes the following steps:

[0055] S1. Collect historical data and current time period data, where both historical data and current time period data include measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations. In this embodiment, historical data and current time period data are strictly speaking historical data that have already occurred. Here, historical data occurs earlier than current time period data, that is, historical data precedes current time period data. For example, historical data is data within the previous 365 days, and current time period data is data within the previous day. Among them, the upstream hydrological station is the hydrological station for observing runoff in the upstream of the tidal section, the analysis station is the station for predicting water level in the tidal section, and the downstream reference station is the tide gauge station near the open sea for measuring tidal range data. Measured tidal level data is obtained at the downstream reference station, and the measured tidal range data is obtained by calculating the difference between the high tide level and the low tide level in the measured tidal level data of the downstream reference station.

[0056] S2. The time lag between measured water level data and measured runoff data, and between measured water level data and measured tidal range data are calculated using cross-correlation analysis. The measured runoff data and measured tidal range data are then adjusted for lag using the measured water level data as a reference.

[0057] S3. Perform EEMD decomposition (i.e., ensemble empirical mode decomposition) on the hysteresis-adjusted measured water level data, measured runoff data, and measured tidal range data to obtain their respective intrinsic mode functions and residuals;

[0058] S4. Construct an LSTM model. Use the intrinsic modulus function and residuals of the three sets of historical data obtained in step S3 as the training data for the LSTM model and train the LSTM model. Use the intrinsic modulus function and residuals of the three sets of current time data obtained in step S3 as the input variables of the LSTM model and output the predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future time periods.

[0059] S5. The predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future periods are superimposed to obtain the predicted value of the tidal river section water level.

[0060] The upstream hydrological station is a hydrological station for observing runoff in the upstream section of the tidal river, the analysis station is a station for predicting water level in the tidal river section, and the downstream reference station is a tide gauge station for measuring tidal range data near the open sea.

[0061] The measured runoff data, measured water level data, and measured tidal range data are time-series data of equal length with a time interval of 1 hour. That is, the set of runoff data measured every 1 hour is the measured runoff data, the set of water level data is the measured water level data, and the set of tidal range data is the measured tidal range data, and all datasets have the same length. For example, measuring the runoff and water level data at each hour (1, 2, 3, etc.) constitutes the measured runoff data and measured water level data, respectively. Measuring the tidal range data for each hour (0 to 1, 1 to 2, 2 to 3, etc.) constitutes the measured tidal range data.

[0062] The cross-correlation analysis process in step S2 is as follows:

[0063] For two time series data of length J, x = {x1, x2, ..., x...} J} and y={y1,y2,...,y J The cross-correlation expression between the two is as follows:

[0064] ,

[0065] Among them, C xy ( k () is the cross-correlation coefficient between time series data x and y. k It is a time lag value, and k When k is positive, it indicates that the change in x precedes the change in y; when k is negative, it indicates that the change in y precedes the change in x. and These are the means of the time series data x and y, respectively. and These are the standard deviations of the time series data x and y, respectively; when C xy ( k When the absolute value of ) is the largest, the corresponding k is the time lag value of the time series data x and y;

[0066] Based on the cross-correlation expression, the time lag value kQ between measured water level data and measured runoff data is obtained when the time series data are measured water level data and measured runoff data, respectively; and the time lag value kR between measured water level data and measured tidal range data is obtained when the time series data are measured water level data and measured tidal range data, respectively. That is, when the time series data x and y are measured water level data and measured runoff data, respectively, C is obtained according to the cross-correlation expression. xy ( k When the absolute value of ) is at its maximum, the corresponding k is the time lag value kQ between the measured water level data and the measured runoff data; when the time series data x and y are the measured water level data and the measured tidal range data respectively, C is obtained according to the cross-correlation expression. xy ( kWhen the absolute value of ) is at its maximum, the corresponding k is the time lag value kR between the measured water level data and the measured tidal range data.

[0067] The process of lag-adjusting the measured runoff data and measured tidal range data in step S2 is as follows:

[0068] For time series data of length T before lag adjustment, the lag adjustment method is as follows: Let the time series data of measured water level be Z, the time series data of measured runoff be Q, and the time lag between the measured water level data and the measured runoff data be kQ. Then, take the time series data of the measured runoff data as Q = {Q1, Q2, ..., Q...} T-|kQ|}, the time series data of the measured water level Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T}, as the adjusted time series data of measured runoff and measured water level data; let R be the time series data of measured tidal range data, and kR be the time lag between measured water level data and measured tidal range data, then take the time series data of measured water level data Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T}, the time series data of measured tidal range R={R |kQ|+1-|kR| ,R |kQ|+2-|kR| ,...,R T-|kR| The time series data, adjusted from measured water level and measured tidal range data, consists of Z, Q, and R. The subscripts of these subscripts increase sequentially within their respective time series, with each subscript indicating its position relative to the time of a single data point. For example, T=4, kQ=1, KR=-3 indicates that the current measured water level data is influenced by the measured runoff data from the previous hour and the measured tidal range data from the previous three hours. Specifically, the time series data for measured water level is Z={Z2,Z3,Z4}, the time series data for measured runoff is Q={Q1,Q2,Q3}, and the time series data for measured tidal range is R={R... -1 Let Z2 be a variable R1, where Z2 is influenced by Q1 at the previous hour and R1 at the previous three hours. -1 The influences of Q2 (1 hour prior) and R0 (3 hours prior) on Z3 and Z4 respectively are as follows: Z3 is influenced by Q3 (1 hour prior) and R1 (3 hours prior); and the value time corresponding to Z2 is one hour earlier than the value time corresponding to Z3. That is, Z2 represents data obtained one hour before the value time corresponding to Z3, and similarly, Z3 represents data obtained one hour before the value time corresponding to Z4. Q1 represents data obtained one hour before the value time corresponding to Q2, and Q3 represents data obtained one hour after the value time corresponding to Q2. R... -1R0 represents data obtained one hour before the time corresponding to the value of R0, and R1 represents data obtained one hour after the time corresponding to the value of R0.

[0069] According to the lag adjustment method, when lag adjustment is performed on the measured runoff data and measured tidal range data based on the measured water level data, the value range of the measured water level data remains unchanged during the adjustment process. The value range of the measured runoff data and measured tidal range data is adjusted according to the time lag value kQ between the measured water level data and the measured runoff data, and the time lag value kR between the measured water level data and the measured tidal range data. After lag adjustment, the measured water level data, measured runoff data, and measured tidal range data are time series data of equal length.

[0070] In step S3, the measured water level data, measured runoff data, and measured tidal range data after hysteresis adjustment are decomposed using EEMD. The EEMD decomposition process is as follows:

[0071] Perform N repeated trials. For the time series data x(t) (where x(t) remains constant in each trial), add random white noise in the i-th trial (i=1,2,...,N), expressed as follows:

[0072] ,

[0073] Where, x i (t) represents the time series data after adding random white noise, w i (t) is a random white noise sequence;

[0074] For x i (t) Perform Empirical Mode Decomposition (EMD) to obtain a set of intrinsic mode functions and residuals, expressed as follows:

[0075] ,

[0076] Among them, IMF m (i) Let v be the m-th order (m=1,2,...,M) eigenmode function decomposed in the i-th trial. i (t) is the x calculated in the i-th trial. i The residual of (t);

[0077] The results of N trials are averaged, and the eigenmode functions of the same order are averaged to obtain the eigenmode function components of the EEMD decomposition.

[0078] ,

[0079] Finally, the reconstructed form of the time series data x(t) after EEMD decomposition is obtained, and the expression is:

[0080] ,

[0081] Among them, IMF m (t) is the m-th (m=1,2,...,M) eigenmode function of x(t), and X(t) is the residual of x(t) after EEMD decomposition;

[0082] The measured water level data, measured runoff data, and measured tidal range data obtained after hysteresis adjustment are decomposed according to the EEMD decomposition method to obtain their respective intrinsic modulus functions and residuals.

[0083] In step S4, the intrinsic modulus function and residuals of the three sets of current time period data obtained in step S3 are used as input variables and input into the trained LSTM model. In the LSTM model, the temporal dependencies and long-term hydrological patterns are extracted step by step through two LSTM layers, and then mapped to the prediction dimension through a fully connected layer. Finally, the predicted values ​​of the intrinsic modulus function and residuals of the water level data in the future time period are output through the output layer. In this embodiment, the LSTM model includes one input layer, two LSTM layers, two Dropout layers, one fully connected layer, and one output layer. The input layer receives the intrinsic modulus function and residual data of 24-hour water level data and converts them into a three-dimensional tensor format that the LSTM model can process. The first LSTM layer extracts basic temporal features from the input layer data. The first Dropout layer randomly discards 10% of the neurons in the output of the first LSTM layer to prevent the model from over-relying on local features (such as outliers in a few hours) and reduce the risk of overfitting. The second LSTM layer captures deep temporal dependencies and long-term hydrological patterns based on the first LSTM layer. The second Dropout layer randomly discards 10% of the neurons in the output of the second LSTM layer to further enhance the model's generalization ability. The fully connected layer maps the high-dimensional features output by the second LSTM layer to the prediction dimension (in this embodiment, the number of components for the next 12 hours) and then passes them to the output layer. The output layer outputs the predicted values ​​of the intrinsic modulus function and residuals of the water level data for future periods. In this embodiment, the time window for the input variables in the input layer is set to 24 hours. This is because past water level conditions affect future water level conditions, exhibiting temporal dependence, and a 24-hour window can capture the complete fluctuation pattern of water level conditions within a day. The prediction window length is 12 hours, meaning it predicts data for the next 12 hours. In this embodiment, the Adam optimizer is used for gradient dimensionality reduction in the fully connected layer, with a learning rate of 0.0001. This is because hydrological data often contains a lot of noise, and a smaller learning rate can prevent excessive noise fluctuations during model training, resulting in more stable parameter updates.

[0084] Example 2

[0085] A tidal river level prediction device, used to implement the method described in Embodiment 1, includes:

[0086] The data acquisition module is used to collect historical data and current time period data. Both historical data and current time period data include measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations.

[0087] The first processing module is used to calculate the time lag between measured water level data and measured runoff data, as well as the time lag between measured water level data and measured tidal range data, using cross-correlation analysis, and to perform lag adjustment on the measured runoff data and measured tidal range data based on the measured water level data.

[0088] The second processing module is used to perform EEMD decomposition on the hysteresis-adjusted measured water level data, measured runoff data, and measured tidal range data to obtain their respective intrinsic modulus functions and residuals.

[0089] The third processing module is used to construct the LSTM model. It uses the intrinsic modulus functions and residuals of the three sets of historical data after lag adjustment and decomposition by EEMD as the training data for the LSTM model. It also uses the intrinsic modulus functions and residuals of the three sets of current time data after lag adjustment and decomposition by EEMD as the input variables of the LSTM model, and outputs the predicted values ​​of the intrinsic modulus functions and residuals of the water level data of the analysis station in future time periods.

[0090] The prediction module is used to superimpose the predicted values ​​of the intrinsic modulus function and residuals of the water level data from the analysis station for future periods to obtain the predicted value of the tidal river section water level.

[0091] Example 3

[0092] Electronic devices, including:

[0093] One or more processors;

[0094] Memory, used to store one or more programs;

[0095] When one or more programs are executed by one or more processors, the one or more processors implement the method described in Embodiment 1.

[0096] Example 4

[0097] Computer-readable storage media, including:

[0098] It stores computer-executable instructions, which, when executed, are used to implement the method described in Embodiment 1.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting water levels in tidal river sections, characterized in that, Includes the following steps: S1. Collect historical data and current time period data, including measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations; S2. The time lag between measured water level data and measured runoff data, and between measured water level data and measured tidal range data are calculated using cross-correlation analysis. The measured runoff data and measured tidal range data are then adjusted for lag using the measured water level data as a reference. For time series data of length T before lag adjustment, the lag adjustment method is as follows: Let the time series data of measured water level be Z, the time series data of measured runoff be Q, and the time lag between the measured water level data and the measured runoff data be kQ. Then, take the time series data of the measured runoff data as Q = {Q1, Q2, ..., Q...} T-|kQ| }, the time series data of the measured water level Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T }, as the adjusted time series data of measured runoff and measured water level data; let R be the time series data of measured tidal range data, and kR be the time lag between measured water level data and measured tidal range data, then take the time series data of measured water level data Z={Z |kQ|+1 Z |kQ|+2 ,...,Z T }, the time series data of measured tidal range R={R |kQ|+1-|kR| ,R |kQ|+2-|kR| ,...,R T-|kR| The data consists of time series data adjusted from measured water level and measured tidal range data. The subscripts of time series data Z, Q, and R increase sequentially within their respective time series data, with each subscript indicating the relative position of a single data point within that time series. According to the lag adjustment method, when lag-adjusting the measured runoff and tidal range data based on the measured water level data, the range of measured water level data remains unchanged during the adjustment process. Furthermore, the adjusted measured water level, runoff, and tidal range data are time series data of equal length. S3. Perform EEMD decomposition on the measured water level data, measured runoff data, and measured tidal range data after hysteresis adjustment to obtain their respective intrinsic modulus functions and residuals; S4. Construct an LSTM model. Use the intrinsic modulus function and residuals of the three sets of historical data obtained in step S3 as the training data for the LSTM model and train the LSTM model. The intrinsic modulus function and residuals of the three sets of current time period data obtained in step S3 are used as input variables of the LSTM model, and the predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future time periods are output. S5. The predicted values ​​of the intrinsic modulus function and residuals of the water level data of the analysis station in future periods are superimposed to obtain the predicted value of the tidal river section water level.

2. The method for predicting water levels in tidal river sections according to claim 1, characterized in that: The measured runoff data, measured water level data, and measured tidal range data are time-series data of equal length with a time interval of 1 hour.

3. The method for predicting water levels in tidal river sections according to claim 1, characterized in that, The cross-correlation analysis process in step S2 is as follows: For two time series data of length J, x = {x1, x2, ..., x...} J } and y={y1,y2,...,y J The cross-correlation expression between the two is as follows: ， Among them, C xy ( k () is the cross-correlation coefficient between time series data x and y. k It is a time lag value, and k When k is positive, it indicates that the change in x precedes the change in y; when k is negative, it indicates that the change in y precedes the change in x. and These are the means of the time series data x and y, respectively. and These are the standard deviations of the time series data x and y, respectively; when C xy ( k When the absolute value of ) is the largest, the corresponding k is the time lag value of the time series data x and y; Based on the cross-correlation expression, the time lag value kQ between measured water level data and measured runoff data is obtained when the time series data are measured water level data and measured runoff data respectively; and the time lag value kR between measured water level data and measured tidal range data is obtained when the time series data are measured water level data and measured tidal range data respectively.

4. The method for predicting water levels in tidal river sections according to claim 1, characterized in that: In step S3, the measured water level data, measured runoff data, and measured tidal range data after hysteresis adjustment are decomposed using EEMD. The EEMD decomposition process is as follows: A series of N repeated trials are conducted. Random white noise is added to the time series data x(t) in the i-th trial (i=1,2,...,N), expressed as follows: ， Where, x i (t) represents the time series data after adding random white noise, w i (t) is a random white noise sequence; For x i (t) Perform empirical mode decomposition to obtain a set of eigenmode functions and residuals, expressed as follows: ， Among them, IMF m (i) Let v be the m-th order (m=1,2,...,M) eigenmode function decomposed in the i-th trial. i (t) is the x calculated in the i-th trial. i The residual of (t); The results of N trials are averaged, and the eigenmode functions of the same order are averaged to obtain the eigenmode function components of the EEMD decomposition. ， Finally, the reconstructed form of the time series data x(t) after EEMD decomposition is obtained, and the expression is: ， Among them, IMF m X(t) is the m-th (m=1,2,...,M) eigenmode function of x(t), and X(t) is the residual of x(t) after EEMD decomposition.

5. The method for predicting water levels in tidal river sections according to claim 1, characterized in that: In step S4, the intrinsic modulus function and residuals of the three sets of current time period data obtained in step S3 are used as input variables and input into the trained LSTM model. In the LSTM model, the temporal dependencies and long-term hydrological patterns are extracted step by step through two LSTM layers, and then mapped to the prediction dimension through a fully connected layer. Finally, the predicted values ​​of the intrinsic modulus function and residuals of the water level data in the future time period are output through the output layer.

6. The method for predicting water levels in tidal river sections according to claim 1, characterized in that: The upstream hydrological station is a hydrological station for observing runoff in the upstream of the tidal river section. The analysis station is a station for predicting water level in the tidal river section. The downstream reference station is a tide gauge station close to the open sea for measuring tidal range data. The measured tidal level data is obtained at the downstream reference station. The measured tidal range data is obtained by calculating the difference between the high tide level and the low tide level in the measured tidal level data of the downstream reference station.

7. A device for predicting the water level of a tidal river section, used to implement the method for predicting the water level of a tidal river section as described in any one of claims 1-6, characterized in that, include: The data acquisition module is used to collect historical data and current time period data. Both historical data and current time period data include measured runoff data from upstream hydrological stations, measured water level data from analysis stations, and measured tidal range data from downstream reference stations. The first processing module is used to calculate the time lag between measured water level data and measured runoff data, as well as the time lag between measured water level data and measured tidal range data, using cross-correlation analysis, and to perform lag adjustment on the measured runoff data and measured tidal range data based on the measured water level data. The second processing module is used to perform EEMD decomposition on the hysteresis-adjusted measured water level data, measured runoff data, and measured tidal range data to obtain their respective intrinsic modulus functions and residuals. The third processing module is used to construct the LSTM model. It uses the intrinsic modulus functions and residuals of the three sets of historical data after lag adjustment and decomposition by EEMD as the training data for the LSTM model. It also uses the intrinsic modulus functions and residuals of the three sets of current time data after lag adjustment and decomposition by EEMD as the input variables of the LSTM model, and outputs the predicted values ​​of the intrinsic modulus functions and residuals of the water level data of the analysis station in future time periods. The prediction module is used to superimpose the predicted values ​​of the intrinsic modulus function and residuals of the water level data from the analysis station for future periods to obtain the predicted value of the tidal river section water level.

8. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by one or more processors, the one or more processors implement the tidal river level prediction method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, include: It stores computer-executable instructions, which, when executed, are used to implement the tidal river level prediction method as described in any one of claims 1-6.

Images