Salt tide upstream salinity forecasting model construction method and system

Abstract

The invention relates to the technical field of hydrology and water resources, in particular to a salt tide upstream salinity forecasting model construction method and system. The method comprises the following steps: determining a salt tide monitoring section of a target estuary, and obtaining the real-time salinity of the salt tide monitoring section; when the real-time salinity exceeds a preset salinity threshold value and the continuous rising trend reaches a preset duration, judging the current time period as a high-salinity event; extracting daily average salinity, fresh water flow and tide level data of the associated hydrological monitoring station based on the high-salinity event to form high-salinity event data; and calculating the ratio of the salinity peak time difference and the salinity change amplitude of the salt tide monitoring section and the downstream tide level monitoring station. The method can accurately capture the typical propagation time-lag characteristic and the amplitude response characteristic of the salt tide upstream, overcomes the limitation based on an empirical threshold in the prior art, and remarkably improves the scientificity and the accuracy of salt tide upstream judgment.

Description

Technical Field

[0001] This invention relates to the field of hydrology and water resources technology, and in particular to a method and system for constructing a salinity forecasting model for saltwater intrusion. Background Technology

[0002] Saltwater intrusion is a common natural phenomenon in estuaries, mainly caused by the interaction of tidal dynamics and freshwater flow, leading to seawater intrusion upstream and increased water salinity. This phenomenon poses a serious threat to drinking water safety, agricultural irrigation, and the ecological environment of coastal areas. Therefore, accurate forecasting of salinity changes during saltwater intrusion is of great significance for water resource management and disaster prevention. Currently, existing methods for saltwater intrusion salinity forecasting mainly rely on traditional hydrological or statistical models. While these methods can reflect salinity trends to some extent, they generally suffer from insufficient data utilization. For example, they rely solely on salinity data from a single monitoring section, failing to effectively integrate multi-source data (such as freshwater flow and tidal levels) from upstream and downstream hydrological monitoring stations, resulting in incomplete model input information and limited forecast accuracy.

[0003] Furthermore, existing technologies rely heavily on empirical thresholds for discrimination, lacking quantitative analysis of dynamic characteristics such as peak salinity time difference, salinity variation ratio, and flow rate change. This makes it difficult to accurately capture the typical propagation time lag and amplitude response characteristics of saline intrusion. Simultaneously, the methods for screening key influencing factors also have shortcomings. For example, the analysis of factors such as flow-salinity ratio and salinity gradient often employs simple correlation analysis, failing to consider the abrupt changes and spatiotemporal consistency of these factors, resulting in poor model adaptability to complex hydrological conditions. Summary of the Invention

[0004] Based on this, it is necessary for the present invention to provide a method and system for constructing a salinity forecasting model for upstream saltwater intrusion, in order to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for constructing a salinity forecasting model for upstream saltwater intrusion includes the following steps: Step S1: Determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section; Step S2: When the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration, the current period is determined as a high salinity event; based on the high salinity event, the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations are extracted to form high salinity event data; Step S3: Calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combine the flow change rate of the upstream freshwater flow monitoring station to construct the salinity upstream intrusion discrimination criteria, which include the salinity abrupt change threshold and the flow salinity response delay threshold. Step S4: Calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section, and screen the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. The key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. Step S5: Based on the temporal variation patterns of key influencing factors and combined with the criteria for identifying upstream saltwater intrusion, construct an initial salinity forecast model; optimize the initial salinity forecast model using high salinity event data to obtain the final salinity forecast model.

[0006] Preferably, the present invention also provides a system for constructing a salinity forecasting model for upstream intrusion, used to execute the method for constructing a salinity forecasting model for upstream intrusion as described above. The system for constructing a salinity forecasting model for upstream intrusion includes: The data acquisition module is used to determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section. The event determination module is used to determine the current period as a high salinity event when the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration; based on the high salinity event, it extracts the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations to form high salinity event data; A standard construction module is used to calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combined with the flow change rate of the upstream freshwater flow monitoring station, a salinity upstream intrusion discrimination standard is constructed, which includes the salinity abrupt change threshold and the flow salinity response delay threshold. The factor screening module is used to calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section. It screens the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. Among them, the key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. The model building module is used to construct an initial salinity forecast model based on the temporal variation patterns of key influencing factors and the criteria for identifying upstream saltwater intrusion; the initial salinity forecast model is then optimized using high salinity event data to obtain the final salinity forecast model.

[0007] The beneficial effects of this invention are as follows: On the one hand, by identifying the monitoring section for saltwater intrusion and obtaining real-time salinity data, and combining the continuous upward trend with the preset duration to determine high salinity events, misjudgments caused by short-term salinity fluctuations can be effectively avoided, significantly improving the accuracy and reliability of event determination.

[0008] On the other hand, by calculating dynamic characteristics such as peak salinity time difference, salinity change ratio, and flow rate change rate, a scientific and reasonable criteria for judging saltwater intrusion was constructed. This criteria can accurately capture the typical propagation time delay characteristics and amplitude response characteristics of saltwater intrusion, overcome the limitations of existing technologies based on empirical thresholds, and significantly improve the scientificity and accuracy of saltwater intrusion judgment.

[0009] On the other hand, when screening key influencing factors, not only correlation analysis was considered, but also the mutation characteristics and spatiotemporal consistency of the factors were combined to ensure that the screened factors have stronger explanatory and predictive power for the saltwater intrusion process, thereby improving the model's adaptability to complex hydrological conditions and further enhancing the model's forecast accuracy and stability. Attached Figure Description

[0010] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings: Figure 1 A flowchart illustrating the steps of a method for constructing a salinity forecasting model for upstream saltwater intrusion, according to one embodiment, is shown.

[0011] Figure 2 A schematic diagram of the modules of a saltwater intrusion salinity forecasting model construction system according to an embodiment is shown.

[0012] Figure 3 A schematic diagram of the salinity peak time difference analysis in the construction of a saltwater intrusion discrimination criterion is shown in one embodiment.

[0013] Figure 4 A schematic diagram illustrating the calculation of salinity gradient between a saltwater intrusion monitoring section and a downstream tide level monitoring station is shown in one embodiment. Detailed Implementation

[0014] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0016] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0017] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a method for constructing a salinity forecasting model for upstream saltwater intrusion, comprising the following steps: Step S1: Determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section; Step S2: When the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration, the current period is determined as a high salinity event; based on the high salinity event, the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations are extracted to form high salinity event data; Step S3: Calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combine the flow change rate of the upstream freshwater flow monitoring station to construct the salinity upstream intrusion discrimination criteria, which include the salinity abrupt change threshold and the flow salinity response delay threshold. Step S4: Calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section, and screen the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. The key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. Step S5: Based on the temporal variation patterns of key influencing factors and combined with the criteria for identifying upstream saltwater intrusion, construct an initial salinity forecast model; optimize the initial salinity forecast model using high salinity event data to obtain the final salinity forecast model.

[0018] Preferably, the construction of the criteria for determining the upstream intrusion of saltwater in step S3 includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; In this embodiment of the invention, salinity data is simultaneously collected by a salinity sensor at both the saltwater intrusion monitoring section and the downstream tide level monitoring station, forming a synchronous salinity time series. The data collection frequency can be set to once every 10 minutes. The collected salinity data is transmitted to the data processing center in real time via a wireless transmission module.

[0019] In one implementation of this invention, it is assumed that during the period from 12:00 to 14:00 on a certain day, the salinity data of the saltwater intrusion monitoring section is a continuous measurement every 10 minutes, along with the salinity data of the downstream tide level monitoring station. These data form a synchronous salinity time series.

[0020] The peak salinity times of the saltwater intrusion monitoring section and the downstream tide level monitoring station in the synchronous salinity time series are located, and the peak salinity time difference is calculated based on the peak salinity times. When the peak salinity time difference is within the preset time difference range, the current event is judged as a typical propagation time lag characteristic of saltwater intrusion. In this embodiment of the invention, a fixed-width sliding window (e.g., 30 minutes) is used to slide across the time series in the synchronous salinity time series, with each sliding step being 10 minutes. The maximum salinity value within the window and its corresponding time point are calculated to determine the salinity peak time. Based on these peak times, the salinity peak time difference between the saltwater intrusion monitoring section and the downstream tide level monitoring station is calculated.

[0021] In one implementation of this invention, assuming the peak salinity time at the saltwater intrusion monitoring section is determined to be 13:30 using the sliding window method, and the peak salinity time at the downstream tide level monitoring station is 13:45, then the time difference between the peak salinity times is 15 minutes. If the preset time difference range is 10 to 20 minutes, then the peak salinity time difference of the current event is within the preset range, which can be determined as a typical propagation time lag characteristic of upstream saltwater intrusion.

[0022] Based on the synchronous salinity time series, the salinity variation amplitude of the saltwater intrusion monitoring section and the downstream tide level monitoring station during the same ebb tide event was calculated respectively. The salinity variation amplitude is the difference between the peak salinity and the valley salinity. In this embodiment of the invention, the start and end times of the ebb tide event are determined, and the peak and trough salinity values ​​between these two time points are calculated respectively. The salinity change range is the difference between the peak salinity and the trough salinity. Through the above steps, the degree of salinity change during the upstream intrusion of saltwater can be quantified.

[0023] In one implementation of this invention, assuming that during a certain ebb tide event, the peak salinity at the saltwater monitoring section is 3.5‰ and the trough is 1.0‰, then the salinity variation is 2.5‰. The peak salinity at the downstream tide monitoring station is 3.4‰ and the trough is 0.9‰, with a salinity variation of 2.5‰.

[0024] Calculate the ratio of the salinity change amplitude at the saltwater intrusion monitoring section to that at the downstream tide level monitoring station. When the ratio is within the preset amplitude ratio range, the current event is judged as a typical amplitude response characteristic of saltwater intrusion. In this embodiment of the invention, the salinity change amplitude of the saltwater intrusion monitoring section is divided by the salinity change amplitude of the downstream tide level monitoring station to obtain a ratio. If this ratio is within a preset amplitude ratio range, for example, 0.8 to 1.2, the current event can be determined to be a typical amplitude response characteristic of saltwater intrusion.

[0025] In one implementation of this invention, assuming the salinity change at the saltwater intrusion monitoring section is 2.5‰ and the salinity change at the downstream tide level monitoring station is 2.5‰, the ratio is 1.0. If the preset range of the amplitude ratio is 0.8 to 1.2, then the ratio of 1.0 for the current event is within the preset range and can be determined as a typical amplitude response characteristic of saltwater intrusion.

[0026] The flow rate change rate and the time lag of salinity change at the saltwater intrusion monitoring section were calculated based on the synchronous salinity time series. In this embodiment of the invention, the flow change rate of the upstream freshwater flow monitoring station is calculated, and the salinity change lag of the saltwater intrusion monitoring section is calculated. The flow change rate can be obtained by performing a first-order difference calculation on the flow data, and the salinity change lag can be obtained by calculating the time difference between the peak salinity time and the peak flow time. Through the above steps, the dynamic relationship between flow change and salinity change can be quantified.

[0027] In one implementation of this invention, it is assumed that the flow data from the upstream freshwater flow monitoring station is a continuous measurement value taken every 10 minutes, with values ​​of 100... 120 140 160 180 200 220 240 260 280 The flow rate change rate was calculated to be 20 using first-order difference. Assuming the peak salinity at the saltwater intrusion monitoring section is at 13:30 and the peak flow rate is at 13:00, then the salinity change time lag is 30 minutes.

[0028] When the salinity change lag is less than or equal to the preset lag threshold, and the decrease in the flow rate is greater than or equal to the preset decrease threshold, the current event is judged as a typical linkage feature of freshwater flow inhibiting saltwater intrusion.

[0029] In this embodiment of the invention, if the salinity change lag is less than or equal to a preset lag threshold (e.g., 30 minutes), and the decrease in the flow rate change rate is greater than or equal to a preset decrease threshold (e.g., 20 minutes), then... If this is the case, then the current event can be determined to be a typical linkage characteristic of freshwater flow inhibiting saltwater intrusion.

[0030] In one implementation of this invention, it is assumed that the preset time delay threshold is 30 minutes and the preset descent threshold is 20. The calculated salinity change lag is 30 minutes, and the decrease in the flow rate change rate is 20%. Since both values ​​meet the preset conditions, the current event can be determined to be a typical linkage characteristic of freshwater flow inhibiting saltwater intrusion.

[0031] Preferably, the key influencing factors that dominate the upstream intrusion of saltwater intrusion in step S4, screened through correlation analysis, include: Obtain the synchronous flow-salinity ratio sequence of the saltwater intrusion monitoring section and the upstream freshwater flow monitoring station, as well as the salinity gradient sequence of the saltwater intrusion monitoring section and the downstream tide level monitoring station; In this embodiment of the invention, salinity and flow sensors are used to simultaneously collect data at the saltwater intrusion monitoring section, the upstream freshwater flow monitoring station, and the downstream tide level monitoring station, respectively, to generate a synchronous flow-salinity ratio sequence and a salinity gradient sequence. The data acquisition frequency can be set to once every 10 minutes.

[0032] The flow-salinity ratio sequence and the salinity gradient sequence are smoothed by a sliding window according to a preset smoothing window, resulting in a smoothed flow-salinity ratio sequence and a smoothed salinity gradient sequence. In this embodiment of the invention, a sliding window of fixed width (e.g., 30 minutes) is used to slide over the time series, with each sliding step being 10 minutes. The average value within the window is calculated to generate a smoothed series.

[0033] In one implementation of this invention, it is assumed that the preset smoothing window width is 30 minutes and the sliding step size is 10 minutes. The flow-salinity ratio sequence and the salinity gradient sequence are smoothed using the sliding window method to obtain the smoothed flow-salinity ratio sequence and salinity gradient sequence.

[0034] The correlation between the smoothed flow-salinity ratio sequence and the smoothed salinity gradient sequence was calculated separately. The significance test threshold was set to be less than the preset significance level threshold, and the absolute value of the correlation coefficient was greater than the preset correlation coefficient threshold. Factors that met the conditions were selected as initial candidate key influencing factors. In this embodiment of the invention, the Spearman rank correlation coefficient method is used to calculate the correlation between the smoothed flow-salinity ratio sequence and the smoothed salinity gradient sequence. Specifically, the rank correlation coefficient of the two sequences is calculated, and a significance test is performed. If the significance test threshold is less than a preset significance level threshold (e.g., 0.05), and the absolute value of the correlation coefficient is greater than a preset correlation coefficient threshold (e.g., 0.5), then the factor is selected as an initial candidate key influencing factor.

[0035] In one implementation of this invention, it is assumed that the preset significance level threshold is 0.05 and the correlation coefficient threshold is 0.5. The Spearman rank correlation coefficient between the smoothed flow-salinity ratio sequence and the smoothed salinity gradient sequence is calculated to be 0.7, and the significance test threshold is 0.03. Since the significance test threshold is less than 0.05 and the absolute value of the correlation coefficient is greater than 0.5, the flow-salinity ratio and salinity gradient are selected as initial candidate key influencing factors.

[0036] The first-order absolute difference between adjacent time periods is calculated based on the preset time interval for the synchronous flow salinity ratio sequence. If the absolute difference of N consecutive time periods is greater than or equal to the preset mutation threshold, the current time period is determined to be a valid mutation. The value of N ranges from 3 to 5. In this embodiment of the invention, a preset time interval (e.g., 10 minutes) is set, and the absolute value of the first-order difference for each time interval is calculated. If the absolute value of the difference for N consecutive time intervals (N ranging from 3 to 5) is greater than or equal to a preset mutation threshold (e.g., 0.1), then the current time interval is determined to be a valid mutation.

[0037] In one implementation of this invention, it is assumed that the preset time interval is 10 minutes and the mutation threshold is 0.1. The calculated absolute values ​​of the first-order differences in the synchronous flow-salinity ratio sequence are 0.2, 0.1, 0.3, 0.1, 0.2, 0.1, 0.3, 0.1, 0.2, and 0.1, respectively. Assuming N is 3, starting from the first time interval, the absolute values ​​of the differences for three consecutive time intervals are 0.2, 0.1, and 0.3, all greater than or equal to 0.1. Therefore, the first to third time intervals are determined to be valid mutations.

[0038] Candidate key impact factors are determined based on the spatiotemporal consistency between the effective mutation period and the initial candidate key impact factors.

[0039] In this embodiment of the invention, the temporal and spatial matching degree between the effective mutation period and the initial candidate key impact factor is examined. If the two are highly consistent in time and space, the factor is determined as a candidate key impact factor.

[0040] In one implementation of this invention, it is assumed that the effective mutation period is from period 1 to period 3, and the initial candidate key influencing factors are the flow-to-salinity ratio and the salinity gradient. Analysis revealed that both the flow-to-salinity ratio and the salinity gradient exhibited significant mutation characteristics during periods 1 to 3, and showed high consistency in both time and space. Therefore, the flow-to-salinity ratio and the salinity gradient were selected as candidate key influencing factors.

[0041] Preferably, the calculation of the salinity gradient in step S4 includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; For each sampling time, the difference between the salinity value at the saltwater intrusion monitoring section and the salinity value at the downstream tide level monitoring station is calculated based on the synchronous salinity time series. In this embodiment of the invention, for each time point, the corresponding salinity value is extracted from the synchronous salinity time series, and the difference between the two is calculated.

[0042] Obtain the distance between the river centerline of the saltwater intrusion monitoring section and the downstream tide level monitoring station; In one implementation of this invention, it is assumed that the distance between the river centerline of the salinity monitoring section and the downstream tide level monitoring station is measured to be 10 kilometers using a GIS system. This distance value will be used for subsequent salinity gradient calculations.

[0043] The salinity gradient is calculated based on the difference and the distance from the river centerline.

[0044] In this embodiment of the invention, the salinity gradient is calculated based on the salinity difference and the distance to the river centerline at each sampling time. Specifically, the salinity gradient is defined as the salinity difference divided by the distance to the river centerline. Through the above steps, the spatial rate of change of salinity during the upstream intrusion of saltwater can be quantified.

[0045] Preferably, the calculation of the peak salinity time difference in step S3 includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; Based on the preset analysis window, the sliding window method is used to scan the synchronous salinity time series of the saltwater intrusion monitoring section and the downstream tide level monitoring station to locate the local salinity peak time. In this embodiment of the invention, a preset analysis window (e.g., 30 minutes) is set, and the window slides on the time series with a sliding step of 10 minutes each time. The maximum salinity value within the window and its corresponding time point are calculated to determine the salinity peak time.

[0046] In one implementation of this invention, it is assumed that the preset analysis window is 30 minutes and the sliding step is 10 minutes. The salinity time series of the saltwater intrusion monitoring section is scanned using the sliding window method, and the local salinity peak time is located at 13:30, with a peak value of 3.5‰. Similarly, the salinity time series of the downstream tide level monitoring station is scanned, and the local salinity peak time is located at 13:45, with a peak value of 3.4‰.

[0047] Calculate the time lag between the synchronous salinity time series of the saltwater intrusion monitoring section and the downstream tide level monitoring station, and calibrate the timestamp deviation between the saltwater intrusion monitoring section and the downstream tide level monitoring station based on the time lag to obtain the calibrated synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station. In one implementation of this invention, it is assumed that the peak salinity time at the saltwater intrusion monitoring section is 13:30, and the peak salinity time at the downstream tide level monitoring station is 13:45, with a calculated time lag of 15 minutes. The timestamp of the downstream tide level monitoring station is adjusted based on this time lag to align with that of the saltwater intrusion monitoring section. After adjustment, a calibrated synchronous salinity time series is obtained.

[0048] The peak salinity time difference between the salinity monitoring section and the downstream tide level monitoring station was calculated based on the synchronous salinity time series of the calibrated salinity monitoring section and the downstream tide level monitoring station. In one implementation of this invention, it is assumed that the peak salinity time at the calibrated saltwater intrusion monitoring section is 13:30, and the peak salinity time at the downstream tide level monitoring station is 13:45, with a calculated peak salinity time difference of 15 minutes. This time difference is used to assess the propagation time lag characteristics of upstream saltwater intrusion.

[0049] If the peak salinity time difference exceeds the preset time difference threshold or is negative, then the data synchronization and peak detection parameters should be rechecked.

[0050] In this embodiment of the invention, if the calculated peak salinity time difference exceeds a preset time difference threshold (e.g., 30 minutes) or is negative, it indicates a data synchronization problem or improper peak detection parameter settings. In this case, it is necessary to re-check the data synchronization and peak detection parameters.

[0051] In one implementation of this invention, assuming a preset time difference threshold of 30 minutes, the calculated peak salinity time difference is 15 minutes, which is within a reasonable range. However, if the time difference is 35 minutes or negative, such as -10 minutes, it is necessary to recheck the data synchronization and peak detection parameters. By recalibrating data synchronization and adjusting the peak detection parameters, the calculation result of the peak salinity time difference is ensured to be accurate and reliable.

[0052] Preferably, the calculation of the flow-to-salinity ratio in step S4 includes: The daily average salinity and daily average flow rate of the saltwater intrusion monitoring section and the upstream freshwater flow monitoring station were obtained synchronously. In this embodiment of the invention, daily average salinity and daily average flow data are simultaneously collected at the saltwater intrusion monitoring section and the upstream freshwater flow monitoring station using a salinity sensor and a flow sensor, respectively. The data collection frequency can be set to once a day.

[0053] Data from days with missing or excessive salinity in the daily average salinity are removed to obtain the effective daily average salinity sequence. In this embodiment of the invention, the collected daily average salinity data is preprocessed to remove data from days with missing salinity or salinity exceeding a preset reasonable range. The preset reasonable range can be determined based on historical data and actual monitoring conditions; for example, the reasonable range for salinity can be set to 0.5‰ to 8.5‰.

[0054] Remove date data with a flow rate of 0 or exceeding the historical maximum flow rate from the daily average flow rate to obtain the effective daily average flow rate sequence; In this embodiment of the invention, the collected daily average flow data is preprocessed to remove data from days with a flow rate of 0 or exceeding the historical maximum flow rate. The historical maximum flow rate can be determined based on historical data, for example, a historical maximum flow rate of 700 m³ / s can be set.

[0055] In one implementation of this invention, it is assumed that the historical maximum traffic value is 700. Within a certain month, the average daily flow data from the upstream freshwater flow monitoring station showed a flow of 720 cubic meters per second on the 28th day. The data for that date was removed because it exceeded the historical maximum traffic volume. Meanwhile, the traffic volume for day 15 was 0. They were also removed.

[0056] The ratio of the daily average flow rate of the upstream freshwater flow monitoring station to the daily average salinity of the saltwater intrusion monitoring section is calculated based on the effective daily average salinity sequence and the effective daily average flow rate sequence, and a flow-salinity ratio sequence is generated. In this embodiment of the invention, for each valid date, the ratio of daily average flow rate to daily average salinity is calculated to form a flow rate-salinity ratio sequence.

[0057] If the absolute value of the flow-salinity ratio in the flow-salinity ratio sequence is greater than a preset absolute value threshold, then the meteorological data for the corresponding date is obtained; In this embodiment of the invention, the flow-salinity ratio sequence is analyzed to identify outliers. If the absolute value of the flow-salinity ratio is greater than a preset absolute value threshold (e.g., 100), meteorological data for the corresponding date is obtained. Through the above steps, it can be determined whether the anomaly is caused by natural factors.

[0058] In one implementation of this invention, it is assumed that the preset threshold for the absolute value of the flow-salinity ratio is 100. In the flow-salinity ratio sequence, the flow-salinity ratio on the 10th day is 120, exceeding the preset threshold. Therefore, meteorological data for that date, including rainfall, wind speed, and temperature, are obtained.

[0059] Determine whether the current anomaly is a natural anomaly based on the meteorological data for the corresponding date; otherwise, discard the corresponding data to obtain the flow-salinity ratio.

[0060] In this embodiment of the invention, based on the acquired meteorological data, it is determined whether the current anomaly is a natural anomaly. If the anomaly is caused by natural factors, such as a significant increase in rainfall or abnormal changes in wind speed, the data is retained; otherwise, the data is discarded.

[0061] In one implementation of this invention, it is assumed that meteorological data for the 10th day shows rainfall of 100 mm, significantly higher than the average rainfall for other days (10 mm). Therefore, this anomaly is determined to be a natural anomaly, and the flow-salinity ratio data for that date is retained. If the meteorological data does not show significant changes in natural factors, the data for that date is removed, resulting in the final flow-salinity ratio sequence.

[0062] Preferably, the setting of the preset salinity threshold in step S2 includes: The historical salinity of the saltwater intrusion monitoring section during a preset historical period is obtained, including the highest daily salinity value and the corresponding collection time. In this embodiment of the invention, historical salinity data is collected at a salinity monitoring section using a salinity sensor to obtain the daily highest salinity value and its corresponding collection time within a preset historical period. The data collection frequency is once per day.

[0063] In one implementation of this invention, assuming a preset historical period of five years, the historical salinity data of the saltwater intrusion monitoring section includes the daily maximum salinity value and its collection time. For example, the maximum salinity on January 1st of a certain year is 1.2‰, and the collection time is 12:00; the maximum salinity on January 2nd is 1.5‰, and the collection time is 13:00, and so on. This data is completely recorded and stored in a database.

[0064] Data that shows daily salinity fluctuations exceeding a preset threshold and for which there are no associated meteorological or hydrological events during that period are removed from historical salinity data to generate cleaned historical salinity data. In this embodiment of the invention, historical salinity data is preprocessed to remove data whose daily salinity fluctuation exceeds a preset threshold (e.g., 0.5‰). Simultaneously, it is checked whether there are any relevant meteorological or hydrological events (such as heavy rain or typhoons) supporting the data within the corresponding time period. If no relevant events support the data, it is considered abnormal and is removed.

[0065] In one implementation of this invention, it is assumed that the preset salinity fluctuation threshold is 0.5‰. In historical salinity data, the highest salinity on a certain day is 3.0‰, while the highest salinity the previous day was 2.0‰, a fluctuation of 1.0‰, exceeding the preset threshold. Checking the meteorological and hydrological data for this period reveals no related heavy rain or typhoon events, therefore the salinity data for that day is removed. After preprocessing, cleaned historical salinity data is generated.

[0066] The historical salinity of the cleaning process is divided into years, and the Kth percentile of salinity during the flood season and non-flood season of each year is calculated. The average value of the Kth percentile during the flood season of each year is taken as the preset salinity threshold, where the value of K ranges from 70% to 90%.

[0067] In this embodiment of the invention, the value of K is set to range from 70% to 90%, and the Kth percentile of salinity during the flood season and non-flood season of each year is calculated respectively. The average value of the Kth percentile of the flood season of each year is taken as the preset salinity threshold.

[0068] In one implementation of this invention, it is assumed that K is 80%. The cleaned historical salinity data is divided by year, and the 80th percentile of salinity for each year's flood season and non-flood season is calculated. For example, the 80th percentile of salinity during the flood season in one year is 2.5‰, and during the non-flood season it is 1.8‰; in another year, it is 2.6‰ during the flood season and 1.9‰ during the non-flood season. The average of the 80th percentiles for each year's flood season, i.e., (2.5‰ + 2.6‰) / 2 = 2.55‰, is taken as the preset salinity threshold.

[0069] Preferably, the construction of high salinity event data in step S2 further includes: Obtain daily salinity data from saltwater intrusion monitoring sections, and calculate salinity fluctuation values ​​between adjacent sampling times based on the daily salinity data; In this embodiment of the invention, daily salinity data is collected at the salinity monitoring section using a salinity sensor, with a data collection frequency of once per hour. Based on the collected daily salinity data, the salinity fluctuation value between adjacent sampling times is calculated, that is, the absolute value of the difference between the salinity values ​​at two adjacent sampling times.

[0070] If the daily salinity fluctuation value exceeds the preset fluctuation threshold, and there is no related meteorological or hydrological event to support the current event, the sampled value at the corresponding time in the daily salinity data will be removed to obtain the salinity data after removing outliers. In this embodiment of the invention, a preset fluctuation threshold (e.g., 0.5‰) is set to identify abnormal salinity fluctuations. If the salinity fluctuation value at a certain sampling time exceeds the preset fluctuation threshold, and there are no related meteorological or hydrological events to support it during that period, the sampled value is considered an abnormal value and is discarded.

[0071] In one implementation of this invention, a preset fluctuation threshold is assumed to be 0.5‰. In the salinity fluctuation values ​​of a certain day, the fluctuation value in the 12th hour is 0.6‰, exceeding the preset threshold. Checking the meteorological and hydrological data for this period reveals no related events such as heavy rain, typhoons, or sudden changes in flow rate; therefore, the salinity sample value of 3.8‰ in the 12th hour is discarded.

[0072] After removing outliers, check for consecutive missing periods in the salinity data. If missing periods exist, obtain daily salinity data from adjacent hydrological monitoring stations for the same period and use linear interpolation to fill in the missing values ​​to obtain a complete salinity sequence. In this embodiment of the invention, the salinity data after removing outliers is examined to identify whether there are consecutive missing periods. If missing periods exist, daily salinity data from adjacent hydrological monitoring stations for the same period are obtained, and the missing values ​​are filled in using linear interpolation.

[0073] In one implementation of this invention, it is assumed that after removing outliers, salinity data for hours 12 to 14 is missing. Daily salinity data from adjacent hydrological monitoring stations for the same period are obtained, which are 4.1‰ and 4.3‰, respectively. Linear interpolation is used to fill in the missing values, calculating the salinity value for hour 12 to be 4.1‰, for hour 13 to be 4.2‰, and for hour 14 to be 4.3‰. Through these steps, a complete salinity sequence is obtained.

[0074] The complete salinity sequence is integrated with the corresponding freshwater flow and tidal level data in a time series manner to generate high salinity event data.

[0075] In this embodiment of the invention, the complete salinity sequence is aligned with the freshwater flow and tidal data of the same period to form a comprehensive dataset containing salinity, flow, and tidal levels.

[0076] Preferably, the optimization of the initial salinity prediction model in step S5 includes: Step S51: Divide the high salinity event data into a training set and a validation set in chronological order, with the training set accounting for more than or equal to 70% and the validation set accounting for less than or equal to 30%. In one implementation of this invention, it is assumed that the high salinity event data contains 100 days of records. The data from the first 70 days is used as the training set, and the data from the last 30 days is used as the validation set.

[0077] Step S52: Optimize the learning rate, regularization coefficient and number of hidden layer nodes of the initial salinity prediction model, with the goal of minimizing the root mean square error of the salinity prediction value. The root mean square error is calculated based on the validation set data. In this embodiment of the invention, a grid search method is used to optimize the hyperparameters of the initial salinity prediction model. Specifically, the optimization learning rate ranges from 0.001, 0.01, to 0.1, the regularization coefficient ranges from 0.01, 0.1, to 1, and the number of hidden layer nodes ranges from 10, 50, to 100. The grid search method iterates through all hyperparameter combinations and calculates the root mean square error (RMSE) on the validation set to find the optimal hyperparameter combination.

[0078] In one implementation of this invention, it is assumed that the initial salinity prediction model is a neural network model. The hyperparameters are optimized using a grid search method. For example, when the learning rate is 0.01, the regularization coefficient is 0.1, and the number of hidden layer nodes is 50, the root mean square error of the model on the validation set is minimized to 0.2‰. Therefore, this set of hyperparameters is selected as the current optimal combination.

[0079] Step S53: Repeat step S52 until the root mean square error converges or the preset number of iterations is reached; select the model with the smallest root mean square error from all hyperparameter combinations as the final salinity prediction model.

[0080] In this embodiment of the invention, the grid search method is repeatedly executed until the root mean square error converges or a preset number of iterations is reached. Specifically, the preset number of iterations is set to 100. In each iteration, the root mean square error of the model on the validation set is recorded. If the root mean square error does not change significantly after 10 consecutive iterations, the error is considered to have converged.

[0081] In one implementation of this invention, it is assumed that after 50 iterations, the root mean square error of the model on the validation set stabilizes at 0.15‰, and the error does not change significantly in subsequent iterations. Therefore, the error is considered to have converged. The model with the smallest root mean square error among all hyperparameter combinations is selected as the final salinity prediction model. Finally, a model with a learning rate of 0.01, a regularization coefficient of 0.1, and 50 hidden layer nodes is selected as the final model.

[0082] Of particular importance, the accuracy evaluation of the final salinity forecasting model includes: The predicted salinity sequence and the measured salinity sequence are obtained, aligned by time, and the prediction error of the salinity peak time is calculated by the time-weighted average method. The weight allocation rule of the time-weighted average method is as follows: for each increase of the salinity change rate threshold, the error weight of the corresponding time period is increased by the preset weight increase, and the maximum weight does not exceed the preset upper limit. In this embodiment of the invention, the predicted salinity sequence and the measured salinity sequence of the final salinity forecast model are obtained and aligned by time. A time-weighted average method is used to calculate the prediction error at the peak salinity time to evaluate the model's prediction accuracy during key periods. The weighting rule for the time-weighted average method is as follows: for every increase in the salinity change rate by a preset change rate threshold (e.g., 0.1‰ / h), the corresponding time period error weight is increased by a preset weight increment (e.g., 0.1), with the maximum weight not exceeding a preset upper limit (e.g., 1.0).

[0083] In one implementation of this invention, the predicted salinity sequence is assumed to be 1.2‰, 1.5‰, 1.8‰, 2.0‰, 2.2‰, 2.5‰, and the measured salinity sequence is assumed to be 1.3‰, 1.6‰, 1.9‰, 2.1‰, 2.3‰, 2.6‰. The salinity change rate is assumed to be 0.3‰ / h, 0.4‰ / h, 0.5‰ / h, 0.6‰ / h, 0.7‰ / h. The preset change rate threshold is 0.1‰ / h, the weight increment is 0.1, and the weight cap is 1.0. The calculated weights are 1.0, 1.1, 1.2, 1.3, and 1.4, respectively. Based on these weights, the prediction error at the salinity peak time is calculated, resulting in a weighted error sequence.

[0084] Calculate the arithmetic mean of the weighted errors for all time periods to obtain the weighted average of the time-domain errors; In this embodiment of the invention, the arithmetic mean of the weighted errors for all time periods is calculated to obtain the weighted average of the time-domain errors. Through these steps, the average prediction accuracy of the model over the entire prediction period can be comprehensively evaluated. The smaller the weighted average of the time-domain errors, the higher the prediction accuracy of the model.

[0085] In one implementation of this invention, the weighted error sequence is assumed to be 0.1‰, 0.2‰, 0.3‰, 0.4‰, and 0.5‰. The calculated weighted average of the time-domain error is 0.3‰. The preset error threshold is 0.5‰. Since 0.3‰ is less than the preset error threshold, it indicates that the model has high prediction accuracy in the time domain.

[0086] Fast Fourier transform was performed on the predicted salinity sequence and the measured salinity sequence to extract the dominant frequency component and its harmonic components. In one implementation of this invention, the predicted salinity sequence is assumed to be 1.2‰, 1.5‰, 1.8‰, 2.0‰, 2.2‰, 2.5‰, and the measured salinity sequence is assumed to be 1.3‰, 1.6‰, 1.9‰, 2.1‰, 2.3‰, 2.6‰. Through FFT analysis, the dominant frequency component of the predicted salinity sequence is extracted to be 0.1Hz with an amplitude of 2.0‰; the dominant frequency component of the measured salinity sequence is also extracted to be 0.1Hz with an amplitude of 2.1‰. Simultaneously, the first harmonic component of the predicted salinity sequence is extracted to be 0.2Hz with an amplitude of 0.5‰; the first harmonic component of the measured salinity sequence is also extracted to be 0.2Hz with an amplitude of 0.6‰.

[0087] The accuracy of the final salinity forecast model is assessed based on the weighted average, dominant frequency component, and harmonic components.

[0088] Of particular importance is the assessment of the accuracy of the final salinity forecast model based on the weighted average, dominant frequency component, and its harmonic components, including: The dominant frequency deviation between the predicted and measured salinity sequences is calculated based on their respective dominant frequency components, and the harmonic amplitude ratio deviation between the predicted and measured salinity sequences is calculated based on their respective harmonic components. In this embodiment of the invention, the dominant frequency deviation refers to the difference between the dominant frequency of the predicted salinity sequence and the dominant frequency of the measured salinity sequence. The harmonic amplitude ratio deviation refers to the deviation of the ratio of the harmonic amplitude of the predicted salinity sequence to the harmonic amplitude of the measured salinity sequence.

[0089] In one implementation of this invention, it is assumed that the dominant frequency of the predicted salinity sequence is 0.1 Hz with an amplitude of 2.0‰; the dominant frequency of the measured salinity sequence is 0.1 Hz with an amplitude of 2.1‰. The first harmonic component of the predicted salinity sequence is 0.2 Hz with an amplitude of 0.5‰; the first harmonic component of the measured salinity sequence is 0.2 Hz with an amplitude of 0.6‰. The calculated dominant frequency deviation is 0.0 Hz, and the harmonic amplitude ratio deviation is (0.5‰ / 0.6‰)-1=-0.167.

[0090] If the main frequency deviation between the predicted salinity sequence and the measured salinity sequence is less than or equal to the preset frequency deviation threshold and the harmonic amplitude ratio deviation is less than or equal to the preset harmonic threshold, then the frequency domain consistency of the final salinity forecast model is deemed to have passed. In one implementation of this invention, it is assumed that the preset frequency offset threshold is 0.01Hz and the preset harmonic threshold is 0.1Hz. The calculated dominant frequency deviation is 0.0Hz, which is less than the preset frequency offset threshold; the harmonic amplitude ratio deviation is -0.167, which is less than the preset harmonic threshold. Therefore, the frequency domain consistency of the final salinity prediction model is determined to be satisfactory.

[0091] When the weighted average of the time domain error is less than or equal to the preset error threshold and the frequency domain consistency is passed, the accuracy of the final salinity forecast model is determined to be up to standard; otherwise, the model parameter optimization process is triggered.

[0092] In one implementation of this invention, a preset error threshold is assumed to be 0.5‰. The calculated weighted average of the time-domain error is 0.3‰, which is less than the preset error threshold. Meanwhile, frequency domain consistency has been achieved. Therefore, the accuracy of the final salinity forecast model is determined to be satisfactory. If the weighted average of the time-domain error is 0.6‰, exceeding the preset error threshold, or if frequency domain consistency fails, a model parameter optimization process is triggered to readjust the model parameters and improve prediction accuracy.

[0093] The data acquisition module is used to determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section. The event determination module is used to determine the current period as a high salinity event when the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration; based on the high salinity event, it extracts the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations to form high salinity event data; A standard construction module is used to calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combined with the flow change rate of the upstream freshwater flow monitoring station, a salinity upstream intrusion discrimination standard is constructed, which includes the salinity abrupt change threshold and the flow salinity response delay threshold. The factor screening module is used to calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section. It screens the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. Among them, the key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. The model building module is used to construct an initial salinity forecast model based on the temporal variation patterns of key influencing factors and the criteria for identifying upstream saltwater intrusion; the initial salinity forecast model is then optimized using high salinity event data to obtain the final salinity forecast model.

[0094] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application be incorporated into the invention.

[0095] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for constructing a saltwater intrusion salinity forecast model, characterized in that, Includes the following steps: Step S1: Determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section; Step S2: When the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration, the current period is determined as a high salinity event; based on the high salinity event, the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations are extracted to form high salinity event data; Step S3: Calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combine the flow change rate of the upstream freshwater flow monitoring station to construct the salinity upstream intrusion discrimination criteria, which include the salinity abrupt change threshold and the flow salinity response delay threshold. Step S4: Calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section, and screen the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. The key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. Step S5: Based on the temporal variation patterns of key influencing factors and combined with the criteria for identifying upstream saltwater intrusion, construct an initial salinity forecast model; optimize the initial salinity forecast model using high salinity event data to obtain the final salinity forecast model.

2. The method according to claim 1, wherein, The construction of the criteria for determining upstream saltwater intrusion in step S3 includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; The peak salinity times of the saltwater intrusion monitoring section and the downstream tide level monitoring station in the synchronous salinity time series are located, and the peak salinity time difference is calculated based on the peak salinity times. When the peak salinity time difference is within the preset time difference range, the current event is judged as a typical propagation time lag characteristic of saltwater intrusion. Based on the synchronous salinity time series, the salinity variation amplitude of the saltwater intrusion monitoring section and the downstream tide level monitoring station during the same ebb tide event was calculated respectively. The salinity variation amplitude is the difference between the peak salinity and the valley salinity. Calculate the ratio of the salinity change amplitude at the saltwater intrusion monitoring section to that at the downstream tide level monitoring station. When the ratio is within the preset amplitude ratio range, the current event is judged as a typical amplitude response characteristic of saltwater intrusion. The flow rate change rate and the time lag of salinity change at the saltwater intrusion monitoring section were calculated based on the synchronous salinity time series. When the salinity change lag is less than or equal to the preset lag threshold, and the decrease in the flow rate is greater than or equal to the preset decrease threshold, the current event is judged as a typical linkage feature of freshwater flow inhibiting saltwater intrusion.

3. The method according to claim 1, wherein, In step S4, key influencing factors that dominate the upstream intrusion of saltwater intrusion are screened through correlation analysis, including: Obtain the synchronous flow-salinity ratio sequence of the saltwater intrusion monitoring section and the upstream freshwater flow monitoring station, as well as the salinity gradient sequence of the saltwater intrusion monitoring section and the downstream tide level monitoring station; The flow-salinity ratio sequence and the salinity gradient sequence are smoothed by a sliding window according to a preset smoothing window, resulting in a smoothed flow-salinity ratio sequence and a smoothed salinity gradient sequence. The correlation between the smoothed flow-salinity ratio sequence and the smoothed salinity gradient sequence was calculated separately. The significance test threshold was set to be less than the preset significance level threshold, and the absolute value of the correlation coefficient was greater than the preset correlation coefficient threshold. Factors that met the conditions were selected as initial candidate key influencing factors. The first-order absolute difference between adjacent time periods is calculated based on the preset time interval for the synchronous flow salinity ratio sequence. If the absolute difference of N consecutive time periods is greater than or equal to the preset mutation threshold, the current time period is determined to be a valid mutation. The value of N ranges from 3 to 5. Candidate key impact factors are determined based on the spatiotemporal consistency between the effective mutation period and the initial candidate key impact factors.

4. The method according to claim 1, wherein, Step S4, the calculation of the salinity gradient, includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; For each sampling time, the difference between the salinity value at the saltwater intrusion monitoring section and the salinity value at the downstream tide level monitoring station is calculated based on the synchronous salinity time series. Obtain the distance between the river centerline of the saltwater intrusion monitoring section and the downstream tide level monitoring station; The salinity gradient is calculated based on the difference and the distance from the river centerline.

5. The method according to claim 1, wherein, The calculation of the peak salinity time difference in step S3 includes: Obtain synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station; Based on the preset analysis window, the sliding window method is used to scan the synchronous salinity time series of the saltwater intrusion monitoring section and the downstream tide level monitoring station to locate the local salinity peak time. Calculate the time lag between the synchronous salinity time series of the saltwater intrusion monitoring section and the downstream tide level monitoring station, and calibrate the timestamp deviation between the saltwater intrusion monitoring section and the downstream tide level monitoring station based on the time lag to obtain the calibrated synchronous salinity time series between the saltwater intrusion monitoring section and the downstream tide level monitoring station. The peak salinity time difference between the salinity monitoring section and the downstream tide level monitoring station was calculated based on the synchronous salinity time series of the calibrated salinity monitoring section and the downstream tide level monitoring station. If the peak salinity time difference exceeds the preset time difference threshold or is negative, then the data synchronization and peak detection parameters should be rechecked.

6. The method according to claim 1, wherein, Step S4, the calculation of the flow-to-salinity ratio, includes: The daily average salinity and daily average flow rate of the saltwater intrusion monitoring section and the upstream freshwater flow monitoring station were obtained synchronously. Data from days with missing or excessive salinity in the daily average salinity are removed to obtain the effective daily average salinity sequence. Remove date data with a flow rate of 0 or exceeding the historical maximum flow rate from the daily average flow rate to obtain the effective daily average flow rate sequence; The ratio of the daily average flow rate of the upstream freshwater flow monitoring station to the daily average salinity of the saltwater intrusion monitoring section is calculated based on the effective daily average salinity sequence and the effective daily average flow rate sequence, and a flow-salinity ratio sequence is generated. If the absolute value of the flow-salinity ratio in the flow-salinity ratio sequence is greater than a preset absolute value threshold, then the meteorological data for the corresponding date is obtained; Determine whether the current anomaly is a natural anomaly based on the meteorological data for the corresponding date; otherwise, discard the corresponding data to obtain the flow-salinity ratio.

7. The method according to claim 1, wherein, The setting of the preset salinity threshold in step S2 includes: The historical salinity of the saltwater intrusion monitoring section during a preset historical period is obtained, including the highest daily salinity value and the corresponding collection time. Data that shows daily salinity fluctuations exceeding a preset threshold and for which there are no associated meteorological or hydrological events during that period are removed from historical salinity data to generate cleaned historical salinity data. The historical salinity of the cleaning process is divided into years, and the Kth percentile of salinity during the flood season and non-flood season of each year is calculated. The average value of the Kth percentile during the flood season of each year is taken as the preset salinity threshold, where the value of K ranges from 70% to 90%.

8. The method according to claim 1, wherein, The construction of high salinity event data in step S2 also includes: Obtain daily salinity data from saltwater intrusion monitoring sections, and calculate salinity fluctuation values ​​between adjacent sampling times based on the daily salinity data; If the daily salinity fluctuation value exceeds the preset fluctuation threshold, and there is no related meteorological or hydrological event to support the current event, the sampled value at the corresponding time in the daily salinity data will be removed to obtain the salinity data after removing outliers. After removing outliers, check for consecutive missing periods in the salinity data. If missing periods exist, obtain daily salinity data from adjacent hydrological monitoring stations for the same period and use linear interpolation to fill in the missing values ​​to obtain a complete salinity sequence. The complete salinity sequence is integrated with the corresponding freshwater flow and tidal level data in a time series manner to generate high salinity event data.

9. The method according to claim 1, wherein, The optimization of the initial salinity prediction model in step S5 includes: Step S51: Divide the high salinity event data into a training set and a validation set in chronological order, with the training set accounting for more than or equal to 70% and the validation set accounting for less than or equal to 30%. Step S52: Optimize the learning rate, regularization coefficient and number of hidden layer nodes of the initial salinity prediction model, with the goal of minimizing the root mean square error of the salinity prediction value. The root mean square error is calculated based on the validation set data. Step S53: Repeat step S52 until the root mean square error converges or the preset number of iterations is reached; select the model with the smallest root mean square error from all hyperparameter combinations as the final salinity prediction model. 10.A system for constructing a saltwater intrusion salinity forecast model, characterized in that, For executing the method for constructing a salinity forecasting model for upstream intrusion as described in claim 1, the system for constructing the salinity forecasting model for upstream intrusion includes: The data acquisition module is used to determine the saltwater intrusion monitoring section of the target estuary and obtain the real-time salinity of the saltwater intrusion monitoring section. The event determination module is used to determine the current period as a high salinity event when the real-time salinity exceeds the preset salinity threshold and continues to rise for a preset duration; based on the high salinity event, it extracts the daily average salinity, freshwater flow and tide level data of the associated hydrological monitoring stations to form high salinity event data; A standard construction module is used to calculate the ratio of the peak salinity time difference and the salinity change amplitude between the salinity monitoring section and the downstream tide level monitoring station; combined with the flow change rate of the upstream freshwater flow monitoring station, a salinity upstream intrusion discrimination standard is constructed, which includes the salinity abrupt change threshold and the flow salinity response delay threshold. The factor screening module is used to calculate the flow-salinity ratio and salinity gradient of the saltwater intrusion monitoring section. It screens the key influencing factors that dominate the upstream intrusion of saltwater intrusion through correlation analysis. Among them, the key influencing factors include the abrupt change magnitude of the flow-salinity ratio and the rate of change of the salinity gradient. The model building module is used to construct an initial salinity forecast model based on the temporal variation patterns of key influencing factors and the criteria for identifying upstream saltwater intrusion; the initial salinity forecast model is then optimized using high salinity event data to obtain the final salinity forecast model.

Images