Estuary region light taking probability prediction model construction method and device

Abstract

The invention discloses a method and a device for constructing a light taking probability prediction model in an estuary region, which are used for solving the technical problem that the accuracy of light taking probability prediction is relatively low due to the fact that an existing light taking probability prediction technology in the estuary region is mainly used for performing empirical judgment based on historical hydrological observation data and a fixed salinity threshold value. The method comprises the following steps: carrying out binarization processing on an obtained historical salinity sequence based on a preset salinity threshold value to generate a light taking condition sequence; determining a target delay salinity sequence and a target delay water-gas coupling sequence according to a plurality of preset delay times, the historical salinity sequence and the obtained historical water-gas coupling sequence; adopting a preset model configuration item to construct a plurality of initial estuary region light taking probability prediction models according to the target time-lag salinity sequence, the target time-lag water-gas coupling sequence and the light taking condition sequence; and screening the initial estuary region light taking probability prediction models based on a preset model evaluation strategy, and determining a target estuary region light taking probability prediction model.

Description

Technical Field

[0001] The present invention relates to the field of hydrology and water resources application technology, and in particular to a method and device for constructing a freshwater extraction probability prediction model in an estuary area. Background Art

[0002] Estuaries are sensitive areas where freshwater and saltwater meet. The probability of freshwater withdrawal (i.e., the probability of obtaining usable freshwater resources) directly impacts coastal urban water supply, agricultural irrigation, and ecological balance. Due to the dynamic interaction of factors such as tides, runoff, and saltwater intrusion, the spatiotemporal distribution of freshwater resources is highly complex. Accurately predicting the probability of freshwater withdrawal is of great engineering value for water resource management, salinization prevention, and emergency response.

[0003] To meet this need, saltwater-avoidance and freshwater extraction technology has emerged. This technology involves scientifically scheduling water conservancy projects and utilizing tidal patterns and saltwater movement patterns to extract freshwater from river channels during periods of low or no saltwater impact to meet industrial, agricultural, and domestic water needs. This technology is particularly important during dry seasons or periods of active saltwater inundation, when water intakes in estuaries are susceptible to saltwater inundation and cannot draw water normally. The coordinated application of accurate prediction of the probability of freshwater extraction and saltwater-avoidance and freshwater extraction technology can significantly improve the efficiency and security of water resources in estuaries.

[0004] Existing technologies for predicting the probability of freshwater withdrawal in estuaries rely primarily on empirical judgment based on historical hydrological observations and fixed salinity thresholds (e.g., 0.5‰). These techniques estimate the period of freshwater withdrawal by statistically analyzing the correlation between tidal cycles, runoff, and salinity. However, this method requires manual determination of fixed salinity thresholds and relies on expert judgment of "acceptable freshwater" standards. This method is unable to adapt to the impact of saltwater intrusion caused by climate change, resulting in low accuracy in predicting the probability of freshwater withdrawal. Summary of the Invention

[0005] The present invention provides a method and device for constructing a prediction model for the probability of freshwater extraction in estuaries, which is used to solve the technical problem that the existing technology for predicting the probability of freshwater extraction in estuaries is mainly based on historical hydrological observation data and fixed salinity thresholds for empirical judgment, resulting in low accuracy of freshwater extraction probability prediction.

[0006] A first aspect of the present invention provides a method for constructing a prediction model for the probability of freshwater extraction in an estuary area, comprising:

[0007] Obtaining a historical salinity sequence and a historical water-gas coupling sequence, and binarizing the historical salinity sequence based on a preset salinity threshold to generate a desalination sequence;

[0008] Determining a target lag salinity sequence and a target lag water-gas coupling sequence according to a plurality of preset lag times, the historical salinity sequence and the historical water-gas coupling sequence;

[0009] Using preset model configuration items, multiple initial estuary area freshwater probability prediction models are constructed according to the target lag salinity sequence, the target lag water-air coupling sequence, and the freshwater withdrawal situation sequence;

[0010] Based on the preset model evaluation strategy, the initial estuary area freshwater probability prediction model is screened to determine the target estuary area freshwater probability prediction model.

[0011] Optionally, determining a target lag salinity sequence and a target lag water-vapor coupling sequence according to a plurality of preset lag times, the historical salinity sequence, and the historical water-vapor coupling sequence includes:

[0012] Performing time lag processing on the historical salinity sequence and the historical water-vapor coupling sequence respectively using each of the preset lag times to determine a time-lagged salinity sequence and a time-lagged water-vapor coupling sequence corresponding to each of the preset lag times;

[0013] Calculate the correlation between the historical salinity sequence and each of the lag salinity sequences, and output a first correlation value corresponding to each of the lag salinity sequences;

[0014] performing correlation calculations on the historical salinity sequence and each of the lagged water-vapor coupling sequences, and outputting second correlation values corresponding to each of the lagged water-vapor coupling sequences;

[0015] The lag salinity sequence corresponding to the largest first correlation value is selected as the optimal lag salinity sequence;

[0016] The lagged water-gas coupling sequence corresponding to the largest second correlation value is selected as the optimal lagged water-gas coupling sequence;

[0017] The optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence are preprocessed respectively to determine a target lagged salinity sequence and a target lagged water-gas coupling sequence.

[0018] Optionally, preprocessing the optimal lagged salinity sequence and the optimal lagged water-vapor coupling sequence respectively to determine a target lagged salinity sequence and a target lagged water-vapor coupling sequence includes:

[0019] performing a missing value removal operation on the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence respectively to determine an initial lagged salinity sequence and an initial lagged water-gas coupling sequence;

[0020] Normalizing the initial lagged salinity sequence and the initial lagged water-gas coupling sequence respectively to determine an intermediate lagged salinity sequence and an intermediate lagged water-gas coupling sequence;

[0021] The intermediate lagged salinity sequence and the intermediate lagged water-gas coupling sequence are standardized respectively to determine a target lagged salinity sequence and a target lagged water-gas coupling sequence.

[0022] Optionally, the preset model configuration items include a random seed and a callback function; the use of the preset model configuration items to construct multiple initial estuary area freshwater probability prediction models according to the target lag salinity sequence, the target lag water-gas coupling sequence, and the freshwater withdrawal situation sequence includes:

[0023] Taking the target lagged salinity sequence and the target lagged water-air coupling sequence as prediction factors and the desalination situation sequence as prediction target, a desalination probability prediction model for the estuary area is constructed;

[0024] The estuary area freshwater probability prediction model is repeatedly trained based on a random seed and a callback function to determine a plurality of initial estuary area freshwater probability prediction models.

[0025] Optionally, the screening of the initial estuary area freshwater probability prediction models based on a preset model evaluation strategy to determine the target estuary area freshwater probability prediction model includes:

[0026] The preset model evaluation strategy is used to respectively evaluate the model performance of each of the initial estuary area freshwater probability prediction models, and determine the accuracy, confusion matrix statistics and model prediction consistency value corresponding to each of the initial estuary area freshwater probability prediction models;

[0027] The initial estuary area freshwater probability prediction model corresponding to the maximum accuracy, the maximum confusion matrix statistic and the maximum model prediction consistency value is selected as the target estuary area freshwater probability prediction model.

[0028] Optionally, it also includes:

[0029] When receiving a salinity sequence to be measured and a water-vapor coupling sequence to be measured, preprocessing the salinity sequence to be measured and the water-vapor coupling sequence to be measured is performed respectively to generate a target salinity sequence and a target water-vapor coupling sequence;

[0030] The target salinity sequence and the target water-gas coupling sequence are input into the target estuary area freshwater probability prediction model to generate a estuary area freshwater probability prediction result.

[0031] A second aspect of the present invention provides a device for constructing a prediction model for the probability of freshwater extraction in an estuary area, comprising:

[0032] An acquisition module is used to acquire a historical salinity sequence and a historical water-gas coupling sequence, and perform binarization processing on the historical salinity sequence based on a preset salinity threshold to generate a desalination sequence;

[0033] a determination module, configured to determine a target lag salinity sequence and a target lag water-gas coupling sequence according to a plurality of preset lag times, the historical salinity sequence and the historical water-gas coupling sequence;

[0034] A construction module, configured to construct a plurality of initial estuary area freshwater extraction probability prediction models according to the target lag salinity sequence, the target lag water-air coupling sequence, and the freshwater extraction condition sequence using preset model configuration items;

[0035] The screening module is used to screen the initial estuary area freshwater probability prediction models based on a preset model evaluation strategy to determine the target estuary area freshwater probability prediction model.

[0036] A third aspect of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for constructing a model for predicting the probability of desalination in estuary areas as described in any one of the above items.

[0037] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the steps of the method for constructing a model for predicting the probability of desalination in estuary areas as described in any one of the above items.

[0038] A fifth aspect of the present invention provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the method for constructing a model for predicting the probability of desalination in estuary areas as described in any one of the above items.

[0039] It can be seen from the above technical solutions that the present invention has the following advantages:

[0040] The above technical solution of the present invention provides a method for constructing a prediction model for the probability of fresh water extraction in an estuary area. First, a historical salinity sequence and a historical water-vapor coupling sequence are obtained, and the historical salinity sequence is binarized based on a preset salinity threshold to generate a fresh water extraction situation sequence; then, according to multiple preset lag times, historical salinity sequences and historical water-vapor coupling sequences, a target lag salinity sequence and a target lag water-vapor coupling sequence are determined; preset model configuration items are used to construct multiple initial estuary area fresh water extraction probability prediction models according to the target lag salinity sequence, the target lag water-vapor coupling sequence and the fresh water extraction situation sequence; finally, based on a preset model evaluation strategy, each initial estuary area fresh water extraction probability prediction model is screened to determine a target estuary area fresh water extraction probability prediction model. Based on the above scheme, combined with the preset lag time and preset model configuration items, the acquired historical salinity series and historical water-gas coupling series are processed to obtain multiple initial estuary area freshwater probability prediction models, and the preset model evaluation strategy is used to screen each initial estuary area freshwater probability prediction model to determine the target estuary area freshwater probability prediction model. The present invention directly uses the obtained target estuary area freshwater probability prediction model to predict the freshwater probability, and predicts the best time for freshwater in advance. There is no need to manually set a fixed salinity threshold, and it can effectively deal with the invasion of salt tides, thereby improving the accuracy of the freshwater probability prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 A flowchart of a method for constructing a prediction model for the probability of freshwater extraction in estuary areas provided in the first embodiment of the present invention;

[0043] Figure 2 A schematic diagram of a confusion matrix for short-term desalination forecasting provided by the first embodiment of the present invention;

[0044] Figure 3 A schematic diagram of the forecast results of the short-term forecast model test set with different forecast periods provided in the first embodiment of the present invention;

[0045] Figure 4 A schematic diagram of a confusion matrix for forecasting mid- to long-term desalination conditions provided in the first embodiment of the present invention;

[0046] Figure 5 A schematic diagram of the forecast results of a test set of a desalination forecast model with a medium- to long-term forecast period of 7 days provided in the first embodiment of the present invention;

[0047] Figure 6 A schematic diagram of the forecast results of the test set of the desalination forecast model with a medium- to long-term forecast period of 15 days provided in the first embodiment of the present invention;

[0048] Figure 7 A schematic diagram of the forecast results of the test set of the desalination forecast model with a medium- to long-term forecast period of 30 days provided in the first embodiment of the present invention;

[0049] Figure 8 A flowchart of the steps for predicting the desalination probability of a target estuary area using the prediction model provided in the second embodiment of the present invention;

[0050] Figure 9 This is a structural block diagram of a device for constructing a prediction model for the probability of desalination in estuary areas provided by the third embodiment of the present invention. DETAILED DESCRIPTION

[0051] The embodiment of the present invention provides a method and device for constructing a prediction model for the probability of freshwater extraction in estuary areas, which is used to solve the technical problem that the existing technology for predicting the probability of freshwater extraction in estuary areas is mainly based on historical hydrological observation data and fixed salinity thresholds for empirical judgment, resulting in low accuracy of freshwater extraction probability prediction.

[0052] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions 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 embodiments described below 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 making creative work are within the scope of protection of the present invention.

[0053] See also Figure 1 , Figure 1 This is a flowchart of the steps of a method for constructing a prediction model for the probability of desalination in estuary areas provided in Example 1 of the present invention.

[0054] The present invention provides a method for constructing a prediction model for the probability of freshwater extraction in an estuary area, comprising:

[0055] Step 101: Obtain a historical salinity sequence and a historical water-gas coupling sequence, and perform binarization processing on the historical salinity sequence based on a preset salinity threshold to generate a desalination condition sequence.

[0056] It should be noted that the historical hourly salinity series (historical salinity series) during the dry season at the main water intakes or water plants in the target estuary area are collected and classified into two categories according to national drinking water safety standards: exceeding the standard and not exceeding the standard. That is, salinity below the preset salinity threshold (250 mg / L) is recorded as not exceeding the standard (fresh water can be taken), and salinity at or above 250 mg / L is recorded as exceeding the standard (not fresh water can be taken). This is converted into a binary classification problem, forming a new series, recorded as the fresh water withdrawal sequence. Specifically, the salinity data in the historical salinity series is binarized, that is, salinity exceeding 250 mg / L is recorded as 1, and salinity below 250 mg / L is recorded as 0, resulting in a fresh water withdrawal sequence consisting of 0 and 1.

[0057] For example, the Modaomen waterway was selected as the target estuary for research. Hourly salinity data (historical salinity series) were collected from the Pinggang Station during the dry season from 2019 to 2023. This salinity series was categorized as exceeding or not exceeding the national drinking water safety standard. Salinity values below 250 mg / L were considered acceptable for freshwater extraction (not exceeding the standard), while salinity values of 250 mg / L or higher were considered unacceptable (exceeding the standard). This generated a freshwater extraction sequence. An example of this sequence is shown in Table 1.

[0058] Table 1 Example of a sequence for taking a light-off condition

[0059]

[0060] Furthermore, we collected sequences of other factors related to salinity exceeding the target area during the same period. According to relevant research, estuaries frequently experience saltwater upwelling due to the combined influence of multiple factors, such as runoff, tidal currents, and wind speed and direction, which leads to salinity exceeding the standard in estuaries. Therefore, we collected historical water-gas coupling sequences from the same period as the historical salinity sequences. These historical water-gas coupling sequences include historical runoff sequences, historical tidal level sequences, and historical wind speed and direction sequences.

[0061] For example, hourly discharge data from Ma Kou Station and hourly tide levels from San Zao Station were collected for the same period from 2019 to 2023, along with wind speed data in the u and v directions at 10 meters for the Macao region from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5-Land hourly data from 1950 to the present dataset. The correlation between the salinity series and the collected discharge, tide, and u and v wind speed series was analyzed. The optimal lag sequence with the highest correlation between the different sequences and the salinity series was selected. The selected prediction factors for the model with a forecast period of 1-6 hours are shown in Table 2. The optimal lag series required for the target forecast period model were compiled, missing values removed, and normalized. Where S is the hourly average salinity at Pinggang Station (mg / L); F is the hourly average flow at Makou Station (m³ / s); T is the hourly tide level at Sanzao Station (m); Wu is the east-west projection of the wind speed in Macao (m / s); Wv is the north-south projection of the wind speed in Macao (m / s); S(t-1) represents the salinity one hour ahead of the salinity forecast value, and the same applies to other variables.

[0062] Table 2 Prediction factors of the lightening probability model at different forecast periods

[0063]

[0064] Step 102: Determine a target lag salinity sequence and a target lag water-vapor coupling sequence according to a plurality of preset lag times, historical salinity sequences, and historical water-vapor coupling sequences.

[0065] Specifically, step 102 may include the following sub-steps S21-S:

[0066] Step S21: performing time lag processing on the historical salinity sequence and the historical water-vapor coupling sequence using each preset lag time, and determining the lag salinity sequence and lag water-vapor coupling sequence corresponding to each preset lag time;

[0067] Step S22: performing correlation calculation on the historical salinity sequence and each lag salinity sequence, and outputting a first correlation value corresponding to each lag salinity sequence;

[0068] Step S23: performing correlation calculations on the historical salinity sequence and each lagged water-vapor coupling sequence, and outputting a second correlation value corresponding to each lagged water-vapor coupling sequence;

[0069] Step S24: selecting the lag salinity sequence corresponding to the largest first correlation value as the optimal lag salinity sequence;

[0070] Step S25: selecting the time-lagged water-gas coupling sequence corresponding to the largest second correlation value as the optimal time-lagged water-gas coupling sequence;

[0071] The time-delayed water-air coupling series includes the time-delayed runoff series, the time-delayed tide level series, and the time-delayed wind speed and direction series.

[0072] The optimal time-lag water-air coupling sequence includes the optimal time-lag runoff sequence, the optimal time-lag tide sequence, and the optimal time-lag wind speed and direction sequence.

[0073] It should be noted that, first, the historical salinity sequence, historical runoff sequence, historical tide level sequence, and historical wind speed and direction sequence are delayed using multiple preset lag times to obtain the delayed salinity sequence, delayed runoff sequence, delayed tide level sequence, and delayed wind speed and direction sequence corresponding to each preset lag time. For example, if the preset lag time is 3h or 5h, the delayed salinity sequence, delayed runoff sequence, delayed tide level sequence, and delayed wind speed and direction sequence with a 3-hour lag, and the delayed salinity sequence, delayed runoff sequence, delayed tide level sequence, and delayed wind speed and direction sequence with a 5-hour lag can be obtained.

[0074] Furthermore, the correlation between different lag sequences (lagged salinity sequence, lagged runoff sequence, lagged tide sequence, and lagged wind speed and direction sequence) and the historical salinity sequence is analyzed, and the cross-correlation analysis method combined with Gaussian process is used to determine the optimal lag of salinity, runoff, tide sequence, wind speed and direction sequences required for different forecast period models. For example, assuming that there is a lagged runoff sequence with a 3-hour lag and a lagged runoff sequence with a 5-hour lag, the correlation value between the lagged runoff sequence with a 3-hour lag and the historical salinity sequence is calculated, as is the correlation value between the lagged runoff sequence with a 5-hour lag and the historical salinity sequence. If the correlation value between the lagged runoff sequence with a 3-hour lag and the historical salinity sequence is the largest, the lagged runoff sequence with a 3-hour lag is selected as the optimal lagged runoff sequence. Similarly, the optimal lagged salinity sequence, the optimal lagged tide sequence, and the optimal lagged wind speed and direction sequence can be obtained.

[0075] Furthermore, by introducing Gaussian weighting factors to enhance the ability to capture nonlinear relationships, the weight term automatically attenuates outliers far from the mean and focuses on the main distribution area of the data, thereby more accurately reflecting the true correlation pattern and significantly improving the robustness and explanatory power of traditional cross-correlation analysis under nonlinear and noisy data. The correlation calculation process is specifically as follows:

[0076] ;

[0077] in, For historical salinity series and preset lag time The correlation values between the corresponding lag series include the first correlation value and the second correlation value, and N is the total length of the time series; is the characteristic value of the historical salinity series; is a time-delayed series, which includes a time-delayed salinity series, a time-delayed runoff series, a time-delayed tide level series, and a time-delayed wind speed and direction series; is the standard deviation of the historical salinity series, used for normalization and Gaussian weight calculation; is the standard deviation of the lagged series, used for normalization and Gaussian weight calculation; is the mean of the historical salinity series; is the mean of the lagged series; is the lag period, i.e. the preset lag time; is the Gaussian weight term, which assigns a Gaussian weight to the cross-covariance at each time point. The weight is given by and The degree of deviation from its mean.

[0078] Step S26: pre-process the optimal lagged salinity sequence and the optimal lagged water-vapor coupling sequence respectively to determine the target lagged salinity sequence and the target lagged water-vapor coupling sequence.

[0079] The target time-lag water-air coupling sequence includes the target time-lag runoff sequence, the target time-lag tide level sequence, and the target time-lag wind speed and direction sequence.

[0080] Furthermore, step S26 may include the following sub-steps S261-S263:

[0081] Step S261: performing a missing value removal operation on the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence respectively to determine an initial lagged salinity sequence and an initial lagged water-gas coupling sequence;

[0082] Step S262: normalize the initial lagged salinity sequence and the initial lagged water-vapor coupling sequence respectively to determine an intermediate lagged salinity sequence and an intermediate lagged water-vapor coupling sequence;

[0083] Step S263: Standardize the intermediate lagged salinity sequence and the intermediate lagged water-vapor coupling sequence respectively to determine the target lagged salinity sequence and the target lagged water-vapor coupling sequence.

[0084] It should be noted that the different sequences with time lags relative to the salinity sequence are organized into a data frame, that is, the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence are preprocessed separately. This process includes removing missing values and normalizing them. By calculating the mean and standard deviation of each feature, the feature values are converted into standardized values with a mean of 0 and a standard deviation of 1, eliminating the impact of the differences in dimensions and numerical ranges between different features on the model performance.

[0085] Step 103: Using preset model configuration items, multiple initial estuary area freshwater probability prediction models are constructed according to the target lag salinity sequence, the target lag water-air coupling sequence, and the freshwater withdrawal situation sequence.

[0086] Preset model configuration items include random seeds and callback functions.

[0087] It should be noted that after determining the characteristic variable X and the target variable Y, the above-collected salinity, runoff, tide level, wind direction and speed series (i.e., target lagged salinity series, target lagged runoff series, target lagged tide level series, and target lagged wind speed and direction series) are all used as the prediction factor, i.e., X, and the fresh water situation is taken as the prediction target, i.e., Y. The training set and test set are divided into 6:4. The method of fusing multiple time series features can better capture the dynamic laws of salinity changes and the correlation with other variables, provide richer information for the model, and help improve the accuracy of the prediction.

[0088] Furthermore, the model for predicting the probability of freshwater extraction in estuaries established by the present invention is a machine learning model with a structure similar to that of a gated recurrent unit (GRU). Specifically, the above-mentioned feature variables and target variables are input into the model to train and construct a freshwater extraction prediction model (the initial freshwater extraction probability prediction model for estuaries). The model employs a two-layer GRU structure: the first layer has 256 units and returns a sequence, while the second layer has 128 units and does not return a sequence. This structural design gradually extracts feature information from time series while using a dropout layer to prevent overfitting. Finally, a fully connected layer with a sigmoid activation function outputs the prediction results, making it suitable for binary classification problems. During model training, an early stopping callback function is used to monitor the validation set loss (val_loss). If the validation set loss stops decreasing for five consecutive epochs, training is stopped early and the model's best weights are restored (restore_best_weights=True). This mechanism prevents overtraining, conserves computing resources, and ensures good performance on the test set. Before training each model, set the random seed (seed(s)) to ensure the repeatability of the model training process, facilitate debugging and verification of model performance, and subsequent optimal model selection.

[0089] Specifically, step 103 may include the following sub-steps S31-S32:

[0090] Step S31, using the target lag salinity sequence and the target lag water-gas coupling sequence as prediction factors and the freshwater extraction sequence as the prediction target, constructing a freshwater extraction probability prediction model for the estuary area;

[0091] Step S32: Repeat model training of the estuary area freshwater probability prediction model based on the random seed and the callback function to determine multiple initial estuary area freshwater probability prediction models.

[0092] For example, the collected salinity, runoff, tidal level, and wind direction and speed sequences (i.e., target lagged salinity sequence, target lagged runoff sequence, target lagged tidal level sequence, and target lagged wind speed and direction sequence) are used as predictors (X), and the freshwater situation is used as the prediction target (Y). The training and test sets are divided into a 6:4 ratio. These feature variables and target variables are input into the model using a gated recurrent unit (GRU) to train and construct a freshwater situation prediction model (an initial freshwater situation probability prediction model for estuaries). The model uses a two-layer GRU architecture. The first layer has 256 units and returns a sequence, while the second layer has 128 units and does not return a sequence. This structural design gradually extracts feature information from the time series while using a dropout layer to prevent overfitting. Finally, a fully connected layer with a sigmoid activation function outputs the prediction results, making it suitable for binary classification problems. During model training, the EarlyStopping callback function monitors the validation set loss (val_loss). If the validation set loss stops decreasing for five consecutive epochs, training is stopped early and the model's best weights are restored (restore_best_weights=True). This mechanism prevents overtraining, conserves computing resources, and ensures good performance on the test set. Before training each model, 10 random seeds (seed(s)) are set. Ten relatively stable models are trained for each forecast period and temporarily saved.

[0093] Step 104 : Screen the initial estuary area freshwater probability prediction models based on the preset model evaluation strategy to determine the target estuary area freshwater probability prediction model.

[0094] Specifically, step 104 may include the following sub-steps S41-S42:

[0095] Step S41: using a preset model evaluation strategy to evaluate the performance of each initial estuary region freshwater probability prediction model, and determining the accuracy, confusion matrix statistics, and model prediction consistency value corresponding to each initial estuary region freshwater probability prediction model;

[0096] Step S42: Select the initial estuary area freshwater probability prediction model corresponding to the maximum accuracy, the maximum confusion matrix statistic, and the maximum model prediction consistency value as the target estuary area freshwater probability prediction model.

[0097] It should be noted that to more comprehensively evaluate the model's performance in predicting the probability of a dropout, we use a combination of accuracy, MCC, and κ to comprehensively measure model performance. Accuracy directly reflects the model's prediction accuracy on the overall data. However, in the evaluation of binary classification models, while accuracy is intuitive, it performs poorly in scenarios with class imbalance or where more rigorous evaluation is required. To this end, this paper utilizes a pre-built model evaluation strategy, which includes accuracy calculation and utilizes the Matthews Correlation Coefficient (MCC) and Cohen's Kappa (κ) as high-order confusion matrix-derived metrics, providing a more comprehensive and robust method for evaluating model performance.

[0098] Furthermore, Accuracy: Accuracy is one of the most intuitive performance indicators, defined as the ratio of the number of samples correctly predicted by the model to the total number of samples. Its calculation formula is as follows:

[0099] ;

[0100] Among them, TP (True Positives) is the number of samples whose salinity does not exceed the standard and the model correctly predicts that the salinity does not exceed the standard; TN (True Negatives) is the number of samples whose salinity exceeds the standard and the model correctly predicts that the salinity exceeds the standard; FP (False Positives) is the number of samples whose salinity exceeds the standard and the model incorrectly predicts that the salinity does not exceed the standard; FN (False Negatives) is the number of samples whose salinity does not exceed the standard and the model incorrectly predicts that the salinity exceeds the standard.

[0101] Furthermore, MCC (Matthews Correlation Coefficient): MCC is a statistic based on the confusion matrix. It focuses on evaluating the overall predictive ability of a binary classification model by comprehensively considering four categories of results: true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN). Its core idea is to measure the consistency between model predictions and actual results through the concept of covariance. The calculation formula for MCC (confusion matrix statistic) is:

[0102] ;

[0103] The metric in this formula is the same as that used in the accuracy calculation formula. The numerator reflects the difference between correct and incorrect classifications, and the denominator is a normalization factor to eliminate the impact of class imbalance. The MCC value ranges from [-1 to 1]. An MCC value of 1 indicates that the model makes perfect predictions (all samples are correctly classified), an MCC value of 0 indicates that the model is equivalent to random guessing, and an MCC value of -1 indicates that the model makes completely inverse predictions (all predictions are opposite to the true labels).

[0104] Furthermore, κ (Cohen's Kappa): Cohen's Kappa, the model prediction consistency value, is used to measure the consistency between the model prediction and the true label. It is particularly suitable for evaluating the consistency between annotators or eliminating the influence of random guesses in model predictions. Its core is to compare the observed consistency rate (p0, that is, the accuracy of the model) with the random consistency rate (p e , assuming that the model predictions are independent of the true labels, calculate the expected agreement rate of random classification) to quantify the model's ability to surpass random guessing. The formula for calculating the model prediction agreement value is:

[0105] ;

[0106] ;

[0107] Among them, κ can standardize the difference between the observed agreement rate and the random agreement rate, and is extremely robust to imbalanced data. The κ value range is [-1, 1]. If the MCC value is 1, it indicates perfect agreement (the model prediction and the true label match exactly), if the MCC value is 0, it indicates the same as random guessing, and if the MCC value is -1, it indicates complete disagreement (the model prediction and the true label are completely opposite).

[0108] Furthermore, from each model with a set number of random seeds, that is, each initial estuary area freshwater probability prediction model, the initial estuary area freshwater probability prediction model with the highest test set accuracy and the highest MCC value (confusion matrix statistic) and κ value (model prediction consistency value) is selected as the target estuary area freshwater probability prediction model and saved in pb format (Protocol Buffer) or h5 format (Hierarchical Data Format version 5).

[0109] For example, 10 models with different forecast periods are verified respectively, and the accuracy and AUC statistics of the optimal model are shown in Table 3.

[0110] Table 3. Accuracy evaluation of short-term desalination probability forecast

[0111]

[0112] Further, see Figure 2 Taking the model with a forecast period of 1h, 6h, 12h, and 24h as an example, the prediction confusion matrix of the model is as follows: Figure 2 As shown in Figure 2, the confusion matrix can show the correct and incorrect predictions of salinity exceeding and not exceeding the standard in the training set and test set. It can be clearly observed that only a very small number of samples are inaccurately predicted.

[0113] Further, see Figure 3 , Figure 3 The accuracy of freshwater withdrawal forecasts for different forecast periods (1 hour, 6 hours, 12 hours, and 24 hours) is shown. Red dots represent incorrect predictions, while green dots represent correct predictions. As can be seen from the figure, correct predictions (green) account for the vast majority, demonstrating that the model maintains high accuracy across all forecast periods. For longer forecast periods (e.g., 24 hours), the forecasts maintain high accuracy, demonstrating the model's robustness and stability. Furthermore, the distribution of prediction errors is relatively dispersed, with no tendency for concentrated errors. This indicates that the model adapts well to data changes across different time periods and possesses strong generalization capabilities. The table shows that the model performs well across all forecast periods. In particular, for the 1-hour forecast, the accuracy is 0.9701, and the MCC and κ values reach 0.9395. These results demonstrate that the model has extremely high classification accuracy in the short term and is able to effectively identify whether salinity exceeds the standard within the forecast period, thereby determining whether water withdrawal is permitted. Even when the forecast horizon was extended to 24 hours, the model's test set accuracy reached 0.8758, demonstrating that the model remains effective in identifying salinity violations even when forecasting over longer timeframes. Furthermore, the accuracy on the training set was consistently slightly higher than that on the test set, demonstrating that the model was able to effectively learn patterns in the training data and exhibit good fitting capabilities. Overall, the model demonstrated stable and reliable performance across forecasting tasks across various timeframes.

[0114] Furthermore, a medium- to long-term binary classification forecast model is constructed, such as Figure 4 The figure shows the confusion matrix of whether the salinity exceeds the standard for the forecast period of 7 days (168h), 15 days (360h), and 30 days (720h). The number of samples for four different situations (predicted to exceed the standard but actually exceeded the standard, predicted to exceed the standard but actually did not exceed the standard, predicted not to exceed the standard but actually exceeded the standard, and predicted not to exceed the standard but actually did not exceed the standard) were counted. The number of samples counted on the main diagonal is the case where the prediction is accurate. It can be seen from the figure that in the models with different forecast periods, the number of samples with accurate predictions in the training set and the test set is much larger than the number of samples with incorrect predictions.

[0115] Furthermore, Figure 5-Figure 7The model demonstrates its performance in forecasting bearish scenarios for medium- to long-term forecast horizons (7, 15, and 30 days). Analysis shows that the model maintains high forecast accuracy even for forecast horizons up to 30 days, demonstrating its excellent long-term forecasting capability and stability. Across all forecast horizons, the number of red dots with incorrect predictions is relatively small and evenly distributed, with no concentrated errors. This demonstrates that the model effectively captures data characteristics and reduces errors when processing long-term data series. Furthermore, the model's consistent performance across different forecast horizons demonstrates its adaptability to forecasts at varying timescales, making it highly valuable for practical applications.

[0116] Furthermore, the accuracy evaluation of models for different forecast periods is shown in Table 4. All models demonstrate good predictive capabilities. The present invention uses hourly data for long-term forecasts, achieving relatively accurate and refined long-term forecasts. In particular, the accuracy of the test set reached 0.719 for a 15-day forecast period, a relatively high value, demonstrating that the model has a good ability to distinguish whether a sample's salinity exceeds the standard in medium-term forecasts.

[0117] Table 4 Accuracy evaluation of the medium- and long-term desalination forecast model

[0118]

[0119] From the above, it can be seen that, whether in the short term (within 24 hours) or in the medium and long term (7 days, 15 days, 30 days), the method for predicting the probability of freshwater extraction in estuary areas proposed in this invention, which integrates binary classification and machine learning algorithms, can be used to construct a prediction model to achieve accurate prediction. When the latest data is collected, the best model selected can be called to predict whether freshwater extraction in the target estuary area will be possible in the future.

[0120] As a comparison of technical effects, we can refer to existing technologies. Avoiding salt water and drawing fresh water refers to a water resource allocation technology in estuaries that uses scientific scheduling of water conservancy projects and the use of tidal patterns and saltwater movement patterns to extract fresh water from rivers when the impact of saltwater is small or not. During the dry season or when saltwater is active, water intakes in estuaries are easily affected by saltwater, making it impossible to draw water normally. Currently, there are many studies on saltwater upstream forecasting, but fewer studies on the prediction of the probability of drawing fresh water. Therefore, developing a prediction method that can accurately predict the probability of drawing fresh water is extremely important for water security in estuaries.

[0121] To address the above issues, the present invention proposes a method for constructing a model to predict the probability of freshwater extraction in estuaries. Specifically, the method collects historical salinity sequences in the target area and classifies them into two categories: exceeding the standard and not exceeding the standard, forming a sequence of freshwater extraction conditions. Sequences of other factors related to salinity exceeding the standard in the target area during the same period are further collected and preprocessed accordingly. All of the above-collected sequences, except for the freshwater extraction conditions, are input into a machine learning model to construct a freshwater extraction prediction model, where the freshwater extraction conditions are the prediction targets. The model accuracy is verified and the preferred model is saved. Newly collected data is input, and the saved model is called to predict the probability of freshwater extraction in the target area in the future. This method predicts the optimal time for freshwater extraction in advance, scientifically dispatches water conservancy projects, and optimizes water resource allocation, effectively responding to saltwater intrusions and ensuring stable and sustainable water supply in estuaries.

[0122] In an embodiment of the present invention, the present invention provides a method for constructing a freshwater probability prediction model in an estuary area. First, a historical salinity sequence and a historical water-vapor coupling sequence are obtained, and the historical salinity sequence is binarized based on a preset salinity threshold to generate a freshwater situation sequence; then, according to multiple preset lag times, historical salinity sequences and historical water-vapor coupling sequences, a target lag salinity sequence and a target lag water-vapor coupling sequence are determined; preset model configuration items are used to construct multiple initial estuary area freshwater probability prediction models according to the target lag salinity sequence, the target lag water-vapor coupling sequence and the freshwater situation sequence; finally, based on the preset model evaluation strategy, each initial estuary area freshwater probability prediction model is screened to determine the target estuary area freshwater probability prediction model. Based on the above scheme, combined with the preset lag time and preset model configuration items, the acquired historical salinity series and historical water-gas coupling series are processed to obtain multiple initial estuary area freshwater probability prediction models, and the preset model evaluation strategy is used to screen each initial estuary area freshwater probability prediction model to determine the target estuary area freshwater probability prediction model. The present invention directly uses the obtained target estuary area freshwater probability prediction model to predict the freshwater probability, and predicts the best time for freshwater in advance. There is no need to manually set a fixed salinity threshold, and it can effectively deal with the invasion of salt tides, thereby improving the accuracy of the freshwater probability prediction.

[0123] For better explanation, refer to Figure 8 , shows a flowchart of the steps of using the target estuary area desalination probability prediction model provided by the second embodiment of the present invention to perform prediction. The process may include the following steps:

[0124] Step 801: When receiving a salinity sequence to be measured and a water-vapor coupling sequence to be measured, preprocess the salinity sequence to be measured and the water-vapor coupling sequence to be measured to generate a target salinity sequence and a target water-vapor coupling sequence;

[0125] Step 802: Input the target salinity sequence and the target water-vapor coupling sequence into the target estuary area freshwater probability prediction model to generate a estuary area freshwater probability prediction result.

[0126] The water-air coupling sequence to be measured includes a runoff sequence to be measured, a tide level sequence to be measured, and a wind speed and direction sequence to be measured.

[0127] It should be noted that the latest salinity sequence to be measured, the runoff sequence to be measured, the tide level sequence to be measured, and the wind speed and direction sequence to be measured are collected and organized into the corresponding format according to the processing principles mentioned above, that is, missing values are removed, normalized, and standardized, and the optimal model saved in the required forecast period (the freshwater probability prediction model for the target estuary area) is called to predict the freshwater probability, thereby providing a reference for the allocation of regional water resources.

[0128] In an embodiment of the present invention, a freshwater extraction probability prediction model for a target estuary area is called to predict the probability of freshwater extraction, so that the best time for freshwater extraction can be predicted in advance, water conservancy projects can be scientifically dispatched, and water resource allocation can be optimized, which can effectively respond to saltwater intrusion and ensure stable and sustainable water supply in the estuary area.

[0129] See also Figure 9 , Figure 9 This is a structural block diagram of a device for constructing a prediction model for the probability of desalination in estuary areas provided by the third embodiment of the present invention.

[0130] The present invention provides a device for constructing a prediction model for the probability of freshwater extraction in an estuary area, comprising:

[0131] The acquisition module 901 is used to acquire a historical salinity sequence and a historical water-air coupling sequence, and perform binarization processing on the historical salinity sequence based on a preset salinity threshold to generate a desalination sequence;

[0132] A determination module 902 is configured to determine a target lag salinity sequence and a target lag water-gas coupling sequence based on a plurality of preset lag times, historical salinity sequences, and historical water-gas coupling sequences;

[0133] A construction module 903 is used to construct multiple initial estuary area freshwater extraction probability prediction models using preset model configuration items according to a target lag salinity sequence, a target lag water-air coupling sequence, and a freshwater extraction situation sequence;

[0134] The screening module 904 is used to screen the initial estuary area freshwater probability prediction models based on a preset model evaluation strategy to determine the target estuary area freshwater probability prediction model.

[0135] Furthermore, the acquisition module 901 includes:

[0136] The first submodule is used to perform time lag processing on the historical salinity series and the historical water-gas coupling series using each preset lag time, and determine the lag salinity series and lag water-gas coupling series corresponding to each preset lag time;

[0137] The second submodule is used to calculate the correlation between the historical salinity sequence and each lagged salinity sequence, and output a first correlation value corresponding to each lagged salinity sequence;

[0138] The third submodule is used to calculate the correlation between the historical salinity series and each lagged water-vapor coupling series, and output the second correlation value corresponding to each lagged water-vapor coupling series;

[0139] A fourth submodule is configured to select a lag salinity sequence corresponding to the largest first correlation value as an optimal lag salinity sequence;

[0140] A fifth submodule is configured to select a lagged water-gas coupling sequence corresponding to a maximum second correlation value as an optimal lagged water-gas coupling sequence;

[0141] The sixth submodule is used to preprocess the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence respectively, and determine the target lagged salinity sequence and the target lagged water-gas coupling sequence.

[0142] Furthermore, the sixth submodule is specifically configured to:

[0143] The missing values of the optimal lagged salinity series and the optimal lagged water-gas coupling series are removed respectively to determine the initial lagged salinity series and the initial lagged water-gas coupling series.

[0144] Normalize the initial lag salinity series and the initial lag water-air coupling series respectively to determine the intermediate lag salinity series and the intermediate lag water-air coupling series;

[0145] The intermediate lagged salinity series and the intermediate lagged water-air coupling series were standardized respectively to determine the target lagged salinity series and the target lagged water-air coupling series.

[0146] Furthermore, the preset model configuration items include a random seed and a callback function; the construction module 903 is specifically used to:

[0147] Taking the target lagged salinity series and target lagged water-air coupling series as prediction factors and the freshwater extraction sequence as the prediction target, a freshwater extraction probability prediction model for the estuary area was constructed.

[0148] Based on random seeds and callback functions, the estuary area freshwater probability prediction model is repeatedly trained to determine multiple initial estuary area freshwater probability prediction models.

[0149] Furthermore, the screening module 904 is specifically configured to:

[0150] The preset model evaluation strategy was used to evaluate the performance of each initial estuary region freshwater probability prediction model, and the accuracy, confusion matrix statistics and model prediction consistency value corresponding to each initial estuary region freshwater probability prediction model were determined;

[0151] The initial estuary area freshwater probability prediction model corresponding to the maximum accuracy, the maximum confusion matrix statistic and the maximum model prediction consistency value is selected as the target estuary area freshwater probability prediction model.

[0152] In an optional embodiment of the device, the device further comprises:

[0153] When receiving the salinity sequence to be measured and the water-gas coupling sequence to be measured, preprocessing the salinity sequence to be measured and the water-gas coupling sequence to be measured is performed respectively to generate a target salinity sequence and a target water-gas coupling sequence;

[0154] The target salinity sequence and the target water-gas coupling sequence are input into the target estuary area freshwater probability prediction model to generate the estuary area freshwater probability prediction results.

[0155] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, modules and sub-modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0156] An embodiment of the present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory; when the computer program is executed by the processor, the processor executes the steps of the method for constructing a model for predicting the probability of desalination in estuary areas as described in any of the above embodiments.

[0157] An embodiment of the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the steps of the method for constructing a model for predicting the probability of desalination in an estuary area as described in any of the above embodiments are implemented.

[0158] An embodiment of the present invention further provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for constructing a model for predicting the probability of desalination in estuary areas as described in any of the above embodiments.

[0159] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0160] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0161] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a prediction model for the probability of freshwater extraction in estuary areas, characterized in that: include: Obtaining a historical salinity sequence and a historical water-gas coupling sequence, and binarizing the historical salinity sequence based on a preset salinity threshold to generate a desalination sequence; Determining a target lag salinity sequence and a target lag water-gas coupling sequence according to a plurality of preset lag times, the historical salinity sequence and the historical water-gas coupling sequence; Using preset model configuration items, multiple initial estuary area freshwater probability prediction models are constructed according to the target lag salinity sequence, the target lag water-air coupling sequence, and the freshwater withdrawal situation sequence; Based on the preset model evaluation strategy, the initial estuary area freshwater probability prediction model is screened to determine the target estuary area freshwater probability prediction model.

2. The method for constructing a prediction model for the probability of desalination in estuary areas according to claim 1, wherein: The step of determining a target lag salinity sequence and a target lag water-gas coupling sequence according to a plurality of preset lag times, the historical salinity sequence, and the historical water-gas coupling sequence comprises: Performing time lag processing on the historical salinity sequence and the historical water-vapor coupling sequence respectively using each of the preset lag times to determine a time-lagged salinity sequence and a time-lagged water-vapor coupling sequence corresponding to each of the preset lag times; Calculate the correlation between the historical salinity sequence and each of the lag salinity sequences, and output a first correlation value corresponding to each of the lag salinity sequences; performing correlation calculations on the historical salinity sequence and each of the lagged water-vapor coupling sequences, and outputting second correlation values corresponding to each of the lagged water-vapor coupling sequences; The lag salinity sequence corresponding to the largest first correlation value is selected as the optimal lag salinity sequence; The lagged water-gas coupling sequence corresponding to the largest second correlation value is selected as the optimal lagged water-gas coupling sequence; The optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence are preprocessed respectively to determine a target lagged salinity sequence and a target lagged water-gas coupling sequence.

3. The method for constructing a prediction model for the probability of desalination in estuary areas according to claim 2, wherein: The preprocessing of the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence respectively to determine a target lagged salinity sequence and a target lagged water-gas coupling sequence includes: performing a missing value removal operation on the optimal lagged salinity sequence and the optimal lagged water-gas coupling sequence respectively to determine an initial lagged salinity sequence and an initial lagged water-gas coupling sequence; Normalizing the initial lagged salinity sequence and the initial lagged water-gas coupling sequence respectively to determine an intermediate lagged salinity sequence and an intermediate lagged water-gas coupling sequence; The intermediate lagged salinity sequence and the intermediate lagged water-gas coupling sequence are standardized respectively to determine a target lagged salinity sequence and a target lagged water-gas coupling sequence.

4. The method for constructing a prediction model for the probability of desalination in estuary areas according to claim 1, wherein: The preset model configuration items include a random seed and a callback function; the preset model configuration items are used to construct multiple initial estuary area freshwater probability prediction models based on the target lag salinity sequence, the target lag water-gas coupling sequence, and the freshwater situation sequence, including: Taking the target lagged salinity sequence and the target lagged water-air coupling sequence as prediction factors and the desalination situation sequence as prediction target, a desalination probability prediction model for the estuary area is constructed; The estuary area freshwater probability prediction model is repeatedly trained based on a random seed and a callback function to determine a plurality of initial estuary area freshwater probability prediction models.

5. The method for constructing a prediction model for the probability of desalination in estuary areas according to claim 1, characterized in that: The method of screening the initial estuary area freshwater probability prediction models based on the preset model evaluation strategy to determine the target estuary area freshwater probability prediction model includes: The preset model evaluation strategy is used to respectively evaluate the model performance of each of the initial estuary area freshwater probability prediction models, and determine the accuracy, confusion matrix statistics and model prediction consistency value corresponding to each of the initial estuary area freshwater probability prediction models; The initial estuary area freshwater probability prediction model corresponding to the maximum accuracy, the maximum confusion matrix statistic and the maximum model prediction consistency value is selected as the target estuary area freshwater probability prediction model.

6. The method for constructing a prediction model for the probability of desalination in estuary areas according to claim 1, characterized in that: Also includes: When receiving a salinity sequence to be measured and a water-vapor coupling sequence to be measured, preprocessing the salinity sequence to be measured and the water-vapor coupling sequence to be measured is performed respectively to generate a target salinity sequence and a target water-vapor coupling sequence; The target salinity sequence and the target water-gas coupling sequence are input into the target estuary area freshwater probability prediction model to generate a estuary area freshwater probability prediction result.

7. A device for constructing a prediction model for the probability of freshwater extraction in estuary areas, characterized in that: include: An acquisition module is used to acquire a historical salinity sequence and a historical water-gas coupling sequence, and perform binarization processing on the historical salinity sequence based on a preset salinity threshold to generate a desalination sequence; a determination module, configured to determine a target lag salinity sequence and a target lag water-gas coupling sequence according to a plurality of preset lag times, the historical salinity sequence and the historical water-gas coupling sequence; A construction module, configured to construct a plurality of initial estuary area freshwater extraction probability prediction models according to the target lag salinity sequence, the target lag water-air coupling sequence, and the freshwater extraction condition sequence using preset model configuration items; The screening module is used to screen the initial estuary area freshwater probability prediction models based on a preset model evaluation strategy to determine the target estuary area freshwater probability prediction model.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for constructing a model for predicting the probability of desalination in an estuary area as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for constructing a model for predicting the probability of desalination in an estuary area according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer is caused to execute the method for constructing a model for predicting the probability of desalination in an estuary area as described in any one of claims 1 to 6.

Images