A saltwater tide upstream prediction method and system based on artificial intelligence

Abstract

The present invention relates to the technical field of saltwater tide upstream prediction, and in particular to a saltwater tide upstream prediction method and system based on artificial intelligence. The method comprises the following steps: performing upstream and downstream drop correlation analysis on the estuary to obtain width change drop correlation data; extracting precipitation in non-flood season and flood season on the historical environmental data of the estuary, and performing downstream runoff change analysis on the width change drop correlation data to obtain non-flood season / flood season downstream runoff change data; performing tidal superposition effect analysis and non-flood season / flood season saltwater tide upstream action analysis based on the non-flood season / flood season downstream runoff change data to obtain non-flood season / flood season saltwater tide upstream action data; constructing a saltwater tide upstream prediction model for the non-flood season / flood season saltwater tide upstream action data based on a recurrent neural network algorithm and sending it to a data cloud platform to perform saltwater tide upstream prediction. The present invention makes the saltwater tide upstream prediction technology more accurate by optimizing the saltwater tide upstream prediction technology.

Description

Technical Field

[0001] The present invention relates to the technical field of saltwater tide upstream prediction, and in particular to a saltwater tide upstream prediction method and system based on artificial intelligence. Background Art

[0002] The phenomenon of saltwater tide has a profound impact on the natural environment, ecosystem, economic development and social culture of the Pearl River Delta. This paper aims to provide decision support for reducing the impact of saltwater tide on the environment and enhancing water security in coastal areas by analyzing the relationship between hydrological processes and the evolution of saltwater tide. At present, the research methods of saltwater tide prediction mainly include prediction methods based on mathematical models, prediction methods based on statistical methods, and prediction methods based on artificial neural networks and genetic algorithms.

[0003] The numerical simulation method is mainly based on the Saint-Venant equations to establish a hydrodynamic model and a mathematical model of estuarine salinity mixing, diffusion, and migration and transport to predict the spatiotemporal changes in estuarine salinity. Among them, numerical models such as FVCOM, Delft3D, and MIKE are widely used. This method has high requirements for data such as river networks, topography, and boundary conditions in the basin, and the amount of calculation for numerical simulation of the entire basin is large. There are certain difficulties in large-scale application in specific practice.

[0004] Mathematical statistics methods use historical data to construct statistical correlation models or empirical formulas to analyze estuary salinity changes and predict the spatiotemporal distribution characteristics and invasion range of saltwater tides. These methods can effectively determine the main controlling factors of saltwater tides, and the amount of calculation is relatively low. However, such methods have high requirements for the amount of data, usually requiring a large amount of sample data, and assuming that the data obeys a typical probability distribution. This limits the applicability of the model under complex conditions. However, a traditional saltwater tide prediction method based on artificial intelligence has the problems of inaccurate prediction results and poor adaptability. Summary of the invention

[0005] Based on this, it is necessary to provide an artificial intelligence-based saltwater tide upstream prediction method and system to solve at least one of the above technical problems.

[0006] To achieve the above purpose, a saltwater tide upstream prediction method based on artificial intelligence is provided, the method comprising the following steps:

[0007] Step S1: Acquire historical environmental data of the estuary; perform correlation analysis of the upstream and downstream drop of the river channel at the estuary to obtain width change drop correlation data;

[0008] Step S2: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively; analyzing downstream runoff changes on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and obtaining downstream runoff change data in the non-flood season / flood season;

[0009] Step S3: Analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season; analyze the saltwater tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season to obtain the saltwater tide upstream effect data in the non-flood season / flood season;

[0010] Step S4: construct a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

[0011] The present invention obtains historical environmental data of the estuary, which helps to establish a basic understanding of the region, including hydrology, geomorphology and ecological environment. The key to this step is to analyze the drop between the upstream and downstream of the river, which can reveal the law and reason of the change in the width of the river. By analyzing the relationship between the width change and the drop in the historical data, the influence of water flow on the river morphology can be identified, and how potential hydrological and geological factors affect the stability of the river can be revealed. This analysis helps to predict future trends and provide a scientific basis for formulating corresponding management strategies. In general, this step can provide comprehensive data support for the long-term environmental evolution of the estuary area. In this step, it is first necessary to extract precipitation data in the non-flood season and the flood season, so that the precipitation characteristics of these two periods can be understood respectively. By comparing and analyzing the precipitation data with the width change drop correlation data, the changes in downstream runoff under different precipitation conditions can be revealed. The precipitation in the non-flood season is less, while the precipitation in the flood season is more. This difference will significantly affect the water flow and river width in the downstream. In-depth analysis of these data can reveal the influence of precipitation on downstream runoff, help predict the impact of future precipitation patterns on the estuary environment, and provide valuable reference for environmental protection and water resources management. At this stage, it is first necessary to analyze the tidal superposition effect based on the downstream runoff change data in the non-flood season and flood season. This analysis helps to understand how tides are superimposed on downstream water flow during different precipitation periods, thereby affecting the salinity distribution of the river. By further analyzing these data, data on the upstream action of saltwater tides in the non-flood season and flood season can be obtained. The upstream action of saltwater tides refers to how the salinity of seawater advances upstream of the river with changes in tides and flow. The beneficial effect of this step is to reveal the impact of salinity changes on the ecological environment of the river under different seasons and precipitation conditions, provide a scientific basis for managing saltwater intrusion, and help formulate corresponding prevention and control measures, thereby effectively protecting inland water resources and ecosystems. In this step, a saltwater upstream prediction model will be constructed based on the recurrent neural network (RNN) algorithm. RNN can process time series data and is suitable for predicting the upstream action of saltwater tides with time dependence. By training the model to predict the changing trend of saltwater tide, accurate prediction can be provided for future salinity management. These predicted data will then be sent to the data cloud platform, which can be monitored and updated in real time, thereby improving the timeliness and accuracy of the prediction. Ultimately, this prediction model helps to more effectively deal with the impact of saltwater tide on rivers, optimize the allocation and management of water resources, and provide decision-making support for relevant departments. Therefore, the present invention is an optimization process made to a traditional saltwater tide upstream prediction method based on artificial intelligence, which solves the problems of inaccurate prediction results and poor fitness of a traditional saltwater tide upstream prediction method based on artificial intelligence, improves the accuracy of the prediction results, and improves the fitness.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: Acquire historical environmental data of the estuary;

[0014] Step S12: Analyze the width change of the river basin at the estuary to obtain the river basin width change data;

[0015] Step S13: Based on the river basin width change data, the upstream and downstream drop correlation analysis of the river is performed to obtain width change drop correlation data.

[0016] The present invention obtains the historical environmental data of the estuary as the basis for subsequent analysis. These data usually include information such as river flow, precipitation, river width, sediment type, topography and landform. By collecting these long-term observation data, the natural environment and change trend of the estuary can be fully understood. The beneficial effect of this step is to provide detailed background information for subsequent analysis, so that the study of the change of river width and its causes can be based on real and accurate data. Only by mastering comprehensive historical environmental data can a reliable basis be provided for analyzing the evolution of the river and predicting future changes. The analysis of the change of river basin width at the estuary aims to identify and quantify the change of river width under different time periods or conditions. This usually involves comparative analysis of remote sensing images of the river, ground measurement data, etc., to determine the change trend of river width and its spatial distribution. This analysis helps to reveal the long-term change characteristics of river width and its influencing factors, such as changes in water flow, accumulation or erosion of sediments, etc. By understanding these changes, the stability of the river and its adaptability to the environment can be evaluated, thereby providing a scientific basis for the formulation of river management and protection measures. In this step, the correlation analysis of the upstream and downstream drop of the river channel is carried out based on the data of the change of river basin width. This analysis aims to explore the relationship between the change of river width and the upstream and downstream drop, so as to reveal the mechanism of the influence of drop on the change of river width. Through statistical and modeling methods, the specific effect of drop on width change can be quantified, and the main factors affecting river morphology can be identified. The beneficial effect of this step is that it can reveal the dynamic change law of river basin, provide important information about river evolution, and provide scientific support for river management, flood management and ecological protection. This analysis helps to understand the natural evolution process of the river channel and guide future engineering design and management strategies.

[0017] Preferably, step S2 comprises the following steps:

[0018] Step S21: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively;

[0019] Step S22: performing historical precipitation attenuation calculation on the flood season precipitation data to obtain flood season precipitation attenuation data;

[0020] Step S23: Calculate the upstream runoff reciprocity for the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the non-flood season / flood season upstream runoff reciprocity data;

[0021] Step S24: Based on the non-flood season / flood season upstream runoff reciprocity data, the width change drop correlation data is analyzed for downstream runoff changes to obtain non-flood season / flood season downstream runoff change data.

[0022] The present invention can understand the precipitation characteristics of the two periods by extracting precipitation data in the non-flood season and the flood season from the historical environmental data of the estuary. The precipitation in the non-flood season is usually less, while the precipitation in the flood season increases significantly. This data extraction helps to identify the seasonal changes in precipitation and provides a basis for further analysis. By comparing the precipitation in the non-flood season and the flood season, the different effects of precipitation on the width of the river channel and the basin can be evaluated, thereby providing the necessary input data for the river flow and hydrological model, and then optimizing water resources management and flood control measures. The purpose is to understand the trend of precipitation changes over time by calculating the historical precipitation attenuation of the same period on the flood season precipitation data. The attenuation calculation helps to identify whether there is a trend or pattern of precipitation reduction, thereby reflecting the impact of potential climate change or other environmental factors. This analysis helps to reveal the long-term change law of precipitation in the flood season and evaluate its impact on the river channel and basin. Understanding the attenuation trend of precipitation is of great significance for formulating strategies to deal with extreme weather events and regulate water resources management. The non-flood season precipitation data and the flood season precipitation attenuation data are combined to calculate the upstream runoff reciprocity. Runoff reciprocity refers to the volatility of runoff and the degree of its periodic changes. This calculation can reveal the fluctuations of runoff in the upstream area under different precipitation conditions and evaluate how these fluctuations affect the drop of width changes. By analyzing the reciprocity, we can gain a deeper understanding of the impact of precipitation changes on upstream runoff, provide basic data for further analysis of downstream runoff changes, and thus support river management and water resource scheduling decisions. Based on the upstream runoff reciprocity data in the non-flood season and flood season, the width change drop correlation data is analyzed for downstream runoff changes. This analysis aims to reveal how changes in upstream runoff affect the flow and hydrological conditions downstream. By understanding the impact of runoff reciprocity on downstream runoff, the overall changes to the river in different precipitation periods can be evaluated. This helps predict the flow dynamics in downstream areas, optimize water resource allocation and flood risk management, and provide a scientific basis for future river management and environmental protection.

[0023] Preferably, step S23 includes the following steps:

[0024] Step S231: Calculate the upstream average flow rate in the non-flood season / flood season based on the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the upstream average flow rate data in the non-flood season / flood season;

[0025] Step S232: Calculate the upstream flow amplitude of the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data to obtain the non-flood season / flood season upstream flow amplitude data;

[0026] Step S233: Calculate the upstream positive and negative flow averages according to the non-flood season / flood season upstream flow variation data and the non-flood season / flood season upstream average flow data to obtain the upstream positive and negative flow average data;

[0027] Step S234: Calculate the upstream runoff reciprocity based on the non-flood season / flood season upstream flow variation data and the upstream positive and negative flow mean data to obtain the non-flood season / flood season upstream runoff reciprocity data.

[0028] The present invention calculates the average flow in the non-flood season and flood season of the upstream according to the precipitation data in the non-flood season and flood season and the attenuation data of the precipitation in the flood season. This calculation can provide a basic understanding of the upstream flow under different precipitation conditions and reveal how precipitation affects the flow level of the river channel. By comparing the average flow in the non-flood season and the flood season, the specific impact of precipitation changes on the upstream flow can be identified. This analysis is crucial for water resource management and helps to evaluate the demand and supply of water flow in different seasons, thereby optimizing the scheduling and management strategies of water resources. By combining the precipitation data in the non-flood season and the precipitation attenuation data in the flood season, the upstream flow amplitude is calculated for the width change drop correlation data. This process aims to determine the fluctuation range and amplitude of the upstream flow under different precipitation conditions. The flow amplitude calculation can reveal the specific impact of precipitation changes on flow fluctuations and help understand the impact of extreme precipitation events on flow. This analysis is of great significance for evaluating the hydrological dynamics in the basin and its impact on the change of river width, and helps to formulate management and response measures for different flow conditions. The upstream flow variation data in the non-flood season and flood season are combined with the average flow data to calculate the upstream positive and negative flow mean. The calculation of the positive and negative flow mean is to quantify the fluctuation range and extreme values ​​of the flow and reflect the characteristics of flow variation. This analysis helps to reveal the direction and intensity of flow fluctuations and identify the key factors affecting the flow. Understanding these positive and negative flow means is crucial for further analyzing the reciprocating characteristics of runoff and formulating response strategies. It can provide targeted flow regulation measures for river management and improve the predictability of extreme flow events. The upstream runoff reciprocity is calculated using the upstream flow variation data in the non-flood season and flood season and the positive and negative flow mean data. The runoff reciprocity calculation is designed to analyze the periodicity and fluctuation characteristics of flow and reveal the stability and instability of flow. This calculation helps to understand the changing laws of flow and evaluate the runoff fluctuations under different seasons and precipitation conditions. By mastering the runoff reciprocity, we can better predict the future flow change trend, provide a scientific basis for water resources management and flood control, and optimize the hydrological management and ecological protection measures in the basin.

[0029] Preferably, step S3 comprises the following steps:

[0030] Step S31: analyzing the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season;

[0031] Step S32: simulating the wind mixing effect according to the tidal superposition effect data in the non-flood season / flood season and the historical environmental data of the estuary to obtain the wind mixing effect data;

[0032] Step S33: analyzing the upstream action of saltwater tides in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season and the wind mixing effect data, and obtaining the upstream action data of saltwater tides in the non-flood season / flood season.

[0033] The present invention analyzes the tidal superposition effect by analyzing the downstream runoff change data in the non-flood season and the flood season, and can reveal how the tide interacts with the runoff change under different seasons or flow conditions. This analysis helps to understand the impact of the tide on the downstream water body, especially the complex interaction of water level fluctuations and flow changes when the tide and runoff are superimposed. Through the quantitative evaluation of the tidal superposition effect, the potential impact of the tide on the downstream hydrological environment can be identified, and then data support can be provided for tidal management, coastal protection and water quality monitoring. In addition, this analysis can also help understand the impact of tides on sediment transport and ecosystems in rivers or estuaries, and provide a scientific basis for environmental protection and planning in coastal areas. The tidal superposition effect data in the non-flood season and the flood season and the historical environmental data of the estuary are used to simulate the wind mixing effect. Wind mixing refers to the impact of wind on the distribution and mixing of pollutants or sediments in the water body. This simulation helps to understand the role of wind on the stratification and mixing degree of water bodies under the influence of tides. By accurately simulating wind mixing, the impact of wind on water quality, sediment migration and pollutant diffusion can be evaluated, especially under different seasons and meteorological conditions. This helps to develop more effective environmental protection measures, improve the ability to prevent and control water pollution, and optimize water quality management strategies in estuary areas. The tidal superposition effect data of the non-flood season and the flood season and the wind mixing data are used together to analyze the upstream effect of saltwater tide. Saltwater upstream refers to the expansion of saltwater in the upstream direction inland due to factors such as tides and wind. By analyzing the upstream effect of saltwater tide, it can be revealed how saltwater affects the upstream area under different seasons and environmental conditions. This analysis can evaluate the potential impact of saltwater tide on the river channel and its surrounding areas, especially on agricultural water use, water resources management and ecosystems. Understanding the laws and scope of saltwater upstream can help to develop effective protection measures to reduce the negative impact of saltwater intrusion on the ecological environment and water quality, thereby better protecting coastal and inland water resources.

[0034] Preferably, step S31 includes the following steps:

[0035] Step S311: Acquire tidal seasonal regularity data;

[0036] Step S312: Analyze the tidal phase migration effect on the tidal seasonal regularity data according to the downstream runoff variation data in the non-flood season / flood season to obtain the tidal phase migration data;

[0037] Step S313: analyzing the tidal level-runoff intensity coupling effect according to the downstream runoff variation data in the non-flood season / flood season and the tidal phase migration data to obtain the tidal level-runoff coupling intensity data;

[0038] Step S314: performing a superimposed regional diffusion numerical analysis on the downstream runoff change data in the non-flood season / flood season and the tidal phase migration data based on the tide-runoff coupling intensity data to obtain superimposed regional diffusion numerical data;

[0039] Step S315: extracting tide-runoff interaction dynamic features from the superimposed area diffusion numerical data to obtain tide-runoff interaction dynamic data;

[0040] Step S316: Analyze the tidal superposition effect based on the superposition area diffusion numerical data and the tide-runoff interaction dynamic data to obtain the non-flood season / flood season tidal superposition effect data.

[0041] The data of tidal seasonality obtained by the present invention is the basis for studying tidal behavior. These data provide the changing patterns of tides in different seasons, including the high and low tide cycles, frequency and intensity of tides. This information is crucial for understanding the long-term impact of tides on the downstream hydrological environment. By analyzing the seasonal law, the influence of tides on water level changes and the pattern of their seasonal fluctuations can be revealed, laying the foundation for further analysis of tidal effects. This data is of great significance for formulating tidal management strategies, predicting tidal events and evaluating the impact of tides on the environment. The downstream runoff change data in the non-flood season and flood season are combined with the tidal seasonal law data to analyze the migration effect of tidal phase. Tidal phase migration refers to the time offset or position change of tidal waves, which is affected by runoff changes. Through this analysis, we can understand how runoff changes affect the adjustment of tidal phases and the specific impact of this migration on the height of tidal levels. This study helps to reveal the complexity of the interaction between tides and runoff, provide a basis for more accurate prediction of tidal fluctuations, and improve tide-related hydrological forecasts and management strategies. The coupling effect between tide level and runoff intensity is analyzed by combining the downstream runoff change data and tidal phase migration data in the non-flood season and flood season. The analysis of the tidal level-runoff intensity coupling effect can reveal the interaction intensity between tide level and runoff and evaluate how the two jointly affect the downstream hydrological conditions. Through this analysis, the synergistic effects of tide and runoff in different seasons and conditions can be determined, and their combined impacts on water level fluctuations, flow changes and water quality can be further understood. This provides valuable information for watershed management and environmental protection, and helps to optimize water resources management under the interaction of tidal and runoff changes. The tidal level-runoff coupling intensity data are used to numerically analyze the superimposed regional diffusion of the downstream runoff change data and tidal phase migration data in the non-flood season and flood season. The numerical analysis of superimposed regional diffusion aims to evaluate the diffusion characteristics of the superimposed effect of tide and runoff in different regions. This analysis can reveal the spatial interaction between tide and runoff and its impact on the diffusion behavior of water bodies, and provides quantitative data for understanding how tide and runoff jointly affect the distribution and diffusion of downstream water bodies. This is of great value for environmental monitoring, pollutant diffusion assessment and water body management. By analyzing the superimposed regional diffusion numerical data, the tide-runoff interaction dynamic characteristics are extracted. This step focuses on identifying the dynamic characteristics of the interaction between tide and runoff, including how the interaction between tide and runoff affects the movement and diffusion behavior of water bodies. The extracted dynamic characteristic data helps to understand the dynamic interaction process of tide and runoff and its impact on the movement of water bodies, thus providing a basis for further analysis and management. Mastering these interactive dynamic characteristics is of great significance for optimizing water quality management, predicting tidal changes, and responding to watershed management challenges. Based on the superimposed regional diffusion numerical data and tide-runoff interaction dynamic data, a comprehensive analysis of the tidal superposition effect is carried out. The tidal superposition effect analysis reveals how the interaction between tide and runoff jointly affects the overall behavior and environmental conditions of downstream water bodies.Through this comprehensive analysis, the combined impact of the superposition effect of tides and runoff on water level, flow and water quality can be evaluated, revealing how tides and runoff mutually enhance or weaken their environmental effects. This will help to formulate more effective strategies for water resources management, flood control and environmental protection, thereby achieving scientific hydrological prediction and management.

[0042] Preferably, step S4 comprises the following steps:

[0043] Step S41: marking the time series data of the saltwater tide tracing back in the non-flooding season / flooding season to obtain the time series data of the saltwater tide tracing back in the non-flooding season / flooding season;

[0044] Step S42: performing neural network structure matching adjustment on the non-flooding season / flooding season saltwater tide tracing time series data to obtain non-flooding season / flooding season saltwater tide tracing network structure matching data;

[0045] Step S43: constructing a saltwater tide upstream prediction model based on the non-flood season / flood season saltwater tide upstream network structure matching data based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model.

[0046] The present invention obtains the time series data of the salt tide tracing back effect by time series labeling the salt tide tracing back effect data in the non-flood season and the flood season. Time series labeling is to classify and mark the salt tide tracing back effect data in chronological order to reflect its changing characteristics at different time points. This process can record in detail the changing trend of the salt tide under different seasons and meteorological conditions, including its intensity, duration and spatial distribution. This time series data is of great significance for identifying the periodic changes, abnormal events and long-term trends of the salt tide tracing back, and provides an accurate data basis for subsequent analysis. This not only helps to understand the timeliness characteristics of the salt tide tracing back, but also provides an important basis for formulating effective prevention and control measures and making scientific predictions. The neural network structure matching adjustment is performed on the salt tide tracing back effect time series data in the non-flood season and the flood season to obtain the network structure matching data of the salt tide tracing back. The neural network structure matching adjustment is to optimize the data representation method by using deep learning technology, and to perform structured analysis on the time series data through the neural network model to reveal the potential patterns and characteristics of the salt tide tracing back. Through this adjustment, the modeling ability of complex time series data can be improved, and the key influencing factors of saltwater tide tracing characteristics can be extracted, so that the prediction model can more accurately capture the dynamic changes and trends of saltwater tide. This process helps to optimize data processing and analysis, making subsequent predictions more accurate and reliable. The recurrent neural network (RNN) algorithm is used to construct a prediction model for the aforementioned saltwater tide tracing network structure matching data. Recurrent neural networks are particularly suitable for processing time series data because they can remember and utilize time dependencies in sequences. By constructing a saltwater tide tracing prediction model, the trend and impact of future saltwater tide changes can be predicted based on historical data. This prediction model can provide early warnings on the intensity, scope and time of occurrence of saltwater tide tracing, providing a scientific basis for preventing and managing saltwater tides. Accurate predictions can help relevant departments formulate effective response strategies, reduce the negative impact of saltwater tides on coastal areas, and protect the environment and economic interests.

[0047] Preferably, step S42 includes the following steps:

[0048] Step S421: using the hidden layer in the bidirectional RNN to perform overfitting optimization processing on the time series data of the saltwater tide tracing back in the non-flooding season / flooding season, and obtaining the fitting data of the saltwater tide tracing back in the non-flooding season / flooding season;

[0049] Step S422: gradient optimization is performed on the non-flood season / flood season saltwater tide upward effect fitting data through a multi-layer dense connection model to obtain non-flood season / flood season saltwater tide upward effect gradient optimization data;

[0050] Step S423: adjusting the error loss of the non-flooding season / flooding season salty tide upward action time series data according to the non-flooding season / flooding season salty tide upward action fitting data and the non-flooding season / flooding season salty tide upward action gradient optimization data to obtain the non-flooding season / flooding season salty tide upward action error loss adjustment data;

[0051] Step S424: performing neural network structure matching adjustment according to the non-flooding season / flooding season salty tide tracing back fitting data, the non-flooding season / flooding season salty tide tracing back gradient optimization data and the non-flooding season / flooding season salty tide tracing back error loss adjustment data to obtain the non-flooding season / flooding season salty tide tracing back network structure matching data.

[0052] The present invention uses a hidden layer in a bidirectional recurrent neural network (RNN) to perform overfitting optimization on the saltwater tide tracing time series data in non-flood season and flood season. Bidirectional RNN can process forward and reverse information flows at the same time, thereby capturing the complex time series dependencies in the time series data. The purpose of overfitting optimization is to improve the fitting accuracy of the model to the saltwater tide tracing data, and obtain more accurate fitting data by reducing model errors and enhancing the expression ability of data features. This optimization process helps to improve the understanding of time series data, so that subsequent models can more effectively learn and predict saltwater tide behavior, and improve the prediction ability and reliability of the overall model. The saltwater tide tracing fitting data is gradient optimized by a multi-layer dense connection model (such as a multi-layer perceptron). The multi-layer dense connection model establishes rich connections between different network layers to enhance the expression ability and computing power of the model. The gradient optimization technology adjusts the model parameters by minimizing the loss function, thereby improving the accuracy of the model for the saltwater tide data. This optimization process can reduce the prediction error, improve the accuracy of the fitting data, and generate optimized saltwater tide tracing gradient data. This provides an optimization basis for subsequent error adjustment and improves the performance and accuracy of the model in practical applications. The fitting data of the saltwater upwelling effect and the gradient optimization data are combined to adjust the error loss. This process involves comparing the error between the actual time series data and the model prediction results, and adjusting them according to the fitting and optimization results. By analyzing and adjusting the error loss, the model's ability to accurately predict the saltwater upwelling effect data can be improved and the model's error can be reduced. This error loss adjustment data helps to optimize the model's prediction performance, making the final prediction results closer to the actual situation, thereby improving the reliability and accuracy of the model in practical applications. Combined with the fitting data of the saltwater upwelling effect, the gradient optimization data and the error loss adjustment data, the neural network structure is matched and adjusted. By optimizing the network structure, the adaptability and prediction ability of the model for the saltwater upwelling effect data can be improved. The matching adjustment of the neural network structure involves adjusting the network's hierarchy, nodes, and connection methods to better reflect the characteristics and complexity of the data. This process ensures that the network can effectively process and predict the saltwater upwelling effect data, so that the final network structure matching data can more accurately capture the behavioral characteristics of the saltwater tide, thereby providing more reliable prediction results.

[0053] Preferably, step S43 includes the following steps:

[0054] Step S431: dividing the non-flooding season / flooding season saltwater tide tracing network structure matching data into a training set and a test set, respectively obtaining a non-flooding season / flooding season saltwater tide tracing network structure training set and a non-flooding season / flooding season saltwater tide tracing network structure test set;

[0055] Step S432: Based on the hidden layer of the recurrent neural network algorithm, neuron training matching is performed on the non-flood season / flood season saltwater tide traceability network structure training set to obtain training set neuron matching data;

[0056] Step S433: setting the sequence length of the training set neuron matching data to obtain neuron matching sequence length data;

[0057] Step S434: constructing an initial saltwater tide tracing prediction model for the non-flooding season / flooding season saltwater tide tracing network structure training set based on the training set neuron matching data and neuron matching sequence length data to obtain an initial saltwater tide tracing prediction model;

[0058] Step S435: iteratively optimize the initial saltwater tide upstream prediction model through the non-flood season / flood season saltwater tide upstream network structure test set to obtain the saltwater tide upstream prediction model.

[0059] The present invention divides the saltwater tide tracing network structure matching data into a training set and a test set. The training set is used for model training, and the test set is used to evaluate the performance of the model. This division process is a key step to ensure that the model can make effective predictions on unseen data. By dividing the data into two independent subsets, the model can be prevented from overfitting during the training process, thereby improving its generalization ability in practical applications. The training set contains a sufficient amount of data for model learning and optimization, while the test set is used to verify the performance of the model on actual data to ensure the reliability and accuracy of the prediction results. The training set data is matched with neuron training using a recurrent neural network (RNN) algorithm. The hidden layer of the RNN can capture the temporal features in the data, and the weights and biases of the neurons are optimized through the training process to improve the matching accuracy of the saltwater tide tracing data. The results of the neuron training matching generate the training set neuron matching data, which reflects the performance of the model when processing temporal data. This training process enables the model to better understand the dynamic changes of the saltwater tide tracing, and provides an optimized parameter basis for the subsequent prediction model construction. The sequence length is set for the neuron matching data in the training set. This process aims to determine the appropriate length of each time series so that the neural network can effectively learn and remember the information in the sequence. The sequence length setting affects the input dimension and learning ability of the model. By setting the sequence length reasonably, the model can optimize the capture effect of time series data and improve the accurate prediction of salt tide traceability characteristics. This step ensures that the model can handle time series data of different lengths, thereby improving the stability and accuracy of the prediction. The initial salt tide traceability prediction model is constructed using the training set neuron matching data and sequence length data. By integrating these data, the model can make preliminary salt tide traceability predictions based on the information in the training set. This model building process is a key stage in model development. It converts the features in the training set into predictive capabilities and provides a basis for subsequent optimization. The initial model can capture the basic patterns and trends of salt tide traceability, laying a solid foundation for iterative optimization. The initial salt tide traceability prediction model is iteratively optimized using the test set. The test set is used to evaluate the actual performance of the model and to reduce the prediction error by continuously adjusting the model parameters and structure. The iterative optimization process includes continuously adjusting the model's training process, learning rate, and other hyperparameters to improve the model's prediction accuracy and generalization ability. This optimization process can make the model better adapt to the actual data, and ultimately obtain a more accurate and reliable saltwater tide upstream prediction model, providing effective support for practical applications.

[0060] Preferably, the present invention further provides a saltwater tide tracing prediction system based on artificial intelligence, which is used to execute the saltwater tide tracing prediction method based on artificial intelligence as described above, and the saltwater tide tracing prediction system based on artificial intelligence comprises:

[0061] The upstream and downstream drop analysis module is used to obtain historical environmental data of the estuary; the upstream and downstream drop correlation analysis of the estuary is carried out to obtain width change drop correlation data.

[0062] The runoff change analysis module is used to extract the precipitation in the non-flood season and flood season from the historical environmental data of the estuary, and obtain the precipitation data in the non-flood season and the precipitation data in the flood season respectively; the downstream runoff change analysis is performed on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and the downstream runoff change data in the non-flood season / flood season is obtained;

[0063] The salt tide upstream effect analysis module is used to analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season, and obtain the tidal superposition effect data in the non-flood season / flood season; analyze the salt tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season, and obtain the salt tide upstream effect data in the non-flood season / flood season;

[0064] The model building module is used to build a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on the recurrent neural network algorithm, obtain the saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

[0065] The beneficial effect of the present invention is that obtaining historical environmental data of the estuary helps to establish a basic understanding of the region, including hydrology, geomorphology and ecological environment. The key to this step is to analyze the drop between the upstream and downstream of the river, which can reveal the law and reason of the change in the width of the river. By analyzing the relationship between the width change and the drop in the historical data, the influence of water flow on the river morphology can be identified, and how potential hydrological and geological factors affect the stability of the river can be revealed. This analysis helps to predict future trends and provide a scientific basis for formulating corresponding management strategies. In general, this step can provide comprehensive data support for the long-term environmental evolution of the estuary area. In this step, it is first necessary to extract precipitation data in the non-flood season and the flood season, so that the precipitation characteristics of these two periods can be understood respectively. By comparing and analyzing the precipitation data with the width change drop correlation data, the changes in downstream runoff under different precipitation conditions can be revealed. The precipitation in the non-flood season is less, while the precipitation in the flood season is more. This difference will significantly affect the water flow and river width in the downstream. In-depth analysis of these data can reveal the influence of precipitation on downstream runoff, help predict the impact of future precipitation patterns on the estuary environment, and provide valuable reference for environmental protection and water resources management. At this stage, it is first necessary to analyze the tidal superposition effect based on the downstream runoff change data in the non-flood season and flood season. This analysis helps to understand how tides are superimposed on downstream water flow during different precipitation periods, thereby affecting the salinity distribution of the river. By further analyzing these data, data on the upstream action of saltwater tides in the non-flood season and flood season can be obtained. The upstream action of saltwater tides refers to how the salinity of seawater advances upstream of the river with changes in tides and flow. The beneficial effect of this step is to reveal the impact of salinity changes on the ecological environment of the river under different seasons and precipitation conditions, provide a scientific basis for managing saltwater intrusion, and help formulate corresponding prevention and control measures, thereby effectively protecting inland water resources and ecosystems. In this step, a saltwater upstream prediction model will be constructed based on the recurrent neural network (RNN) algorithm. RNN can process time series data and is suitable for predicting the upstream action of saltwater tides with time dependence. By training the model to predict the changing trend of saltwater tide, accurate prediction can be provided for future salinity management. These predicted data will then be sent to the data cloud platform, which can be monitored and updated in real time, thereby improving the timeliness and accuracy of the prediction. Ultimately, this prediction model helps to more effectively deal with the impact of saltwater tide on rivers, optimize the allocation and management of water resources, and provide decision-making support for relevant departments. Therefore, the present invention is an optimization process made to a traditional saltwater tide upstream prediction method based on artificial intelligence, which solves the problems of inaccurate prediction results and poor fitness of a traditional saltwater tide upstream prediction method based on artificial intelligence, improves the accuracy of the prediction results, and improves the fitness. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1A schematic diagram of the steps of a saltwater tide upstream prediction method based on artificial intelligence;

[0067] Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG.

[0068] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0069] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0070] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0071] 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 only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0072] To achieve this, please refer to Figure 1 to Figure 2 , a saltwater tide upstream prediction method based on artificial intelligence, the method comprising the following steps:

[0073] Step S1: Acquire historical environmental data of the estuary; perform correlation analysis of the upstream and downstream drop of the river channel at the estuary to obtain width change drop correlation data;

[0074] Step S2: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively; analyzing downstream runoff changes on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and obtaining downstream runoff change data in the non-flood season / flood season;

[0075] Step S3: Analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season; analyze the saltwater tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season to obtain the saltwater tide upstream effect data in the non-flood season / flood season;

[0076] Step S4: construct a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

[0077] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of a saltwater tide prediction method based on artificial intelligence of the present invention. In this example, the saltwater tide prediction method based on artificial intelligence includes the following steps:

[0078] Step S1: Acquire historical environmental data of the estuary; perform correlation analysis of the upstream and downstream drop of the river channel at the estuary to obtain width change drop correlation data;

[0079] In the embodiment of the present invention, it is necessary to obtain historical environmental data of the estuary, specifically including historical water level, flow, tide level, sediment content and other data. These data are usually recorded and accumulated by hydrological monitoring stations for a long time, and can be obtained by accessing the local hydrological database and the data sharing platform of the environmental monitoring agency, and the data is required to cover historical records of different periods and years. For the collection and collation of data, data screening rules can be used to ensure the continuity and integrity of the data, focusing on the historical river channel changes and hydrological characteristics of the estuary area. Next, according to the acquired data, the upstream and downstream of the river channel are subjected to drop correlation analysis, that is, by comparing the water level, flow, terrain changes and other factors of the upstream and downstream, the drop change map is calculated and drawn. Through high-precision terrain data and remote sensing imaging technology, the river section is analyzed, the significant area of ​​width change is identified, and the time series changes of the river channel morphology are compared to analyze the drop changes caused by factors such as topography, deposition, erosion, etc. Finally, the above data are integrated and processed to form "width change drop correlation data", which lays the foundation for the subsequent runoff change analysis.

[0080] Step S2: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively; analyzing downstream runoff changes on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and obtaining downstream runoff change data in the non-flood season / flood season;

[0081] In the embodiment of the present invention, it is first necessary to further subdivide and extract the historical environmental data of the estuary, and focus on extracting the precipitation data of the non-flood season and the flood season. The specific operation is to classify the historical precipitation data of many years according to the flood season (usually June to September each year) and the non-flood season through the time series segmentation method, ensure the accurate division of the data interval, and exclude the influence of abnormal precipitation values ​​on the average precipitation. After extraction, the "non-flood season precipitation data" and "flood season precipitation data" are obtained respectively, and statistical analysis is performed on them to generate precipitation characteristics of typical years and abnormal years. Subsequently, the precipitation data of the non-flood season and the flood season are combined with the "width change drop correlation data" to carry out the change analysis of the downstream runoff. The specific method is: according to the known river width change and precipitation change, the physical hydrological relationship formula is used to derive the change trend of river runoff under different precipitation conditions. First, the runoff change equation of the estuary section is established, the width drop is combined with the precipitation, and the distribution of water in the non-flood season and the flood season in the downstream and the influence on the river width are calculated. The numerical integration technology is used to analyze the water flow process, and the velocity and flow change characteristics of the runoff at different widths and heights are obtained, and finally the "non-flood season / flood season downstream runoff change data" are generated. This data will serve as an important input for predicting the impact of saltwater tides, and provide detailed hydrological dynamic information for subsequent environmental change analysis.

[0082] Step S3: Analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season; analyze the saltwater tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season to obtain the saltwater tide upstream effect data in the non-flood season / flood season;

[0083] In the embodiment of the present invention, it is first necessary to analyze the tidal superposition effect of the "non-flood season / flood season downstream runoff change data". The tidal superposition effect analysis is to combine the runoff change data in the non-flood season and flood season with the tidal data of the estuary to accurately evaluate the impact of the tide on the downstream runoff. The specific operation includes: extracting historical tidal data, such as tide level, tidal direction, tidal cycle, etc., and using the long-term series data of the tidal observation station to perform statistical analysis on the tidal characteristics of different periods. The tidal water level and the downstream runoff are superimposed by the hourly comparison method to generate a tidal-runoff combination diagram. By calculating the superposition amplitude and superposition time series of tide and runoff, the degree of influence of tide on runoff is evaluated, focusing on analyzing the tidal top-up effect, tidal backflow effect and the dynamic impact on estuary runoff, and finally obtaining the "non-flood season / flood season tidal superposition effect data". After completing the tidal superposition effect analysis, the above data is used to analyze the saltwater upstream effect in the non-flood season and flood season. The specific operation is to combine the tidal superposition effect data with the saltwater tide observation data at the estuary, conduct salinity front analysis, and use the salinity data of the water quality monitoring station and the analysis results of the measured water samples to dynamically simulate the salinity changes under different tidal conditions. By establishing a recursive formula for salinity along the way, the changing trend of salinity during the upward movement of the tide is compared and calculated. The analysis includes the length of the saltwater tide upward movement, the speed of the upward movement, and the formation conditions of the high-salinity interval, and then evaluates the impact of the saltwater tide on the estuary ecosystem. Finally, the "non-flood season / flood season saltwater tide upward movement data" is generated, which provides accurate boundary conditions for subsequent prediction steps.

[0084] Step S4: construct a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

[0085] In the embodiment of the present invention, a saltwater tide tracing prediction model is constructed based on the "non-flood season / flood season saltwater tide tracing data". First, the recurrent neural network (RNN) algorithm is used to extract features and recognize time series patterns on the saltwater tide tracing data. The specific implementation steps include: extracting time series features from the "non-flood season / flood season saltwater tide tracing data", such as the time when the salinity peak occurs, the rate of change of the tracing speed, the phase relationship between the tide and the salinity, etc. The long short-term memory (LSTM) unit of the RNN is used to construct a time series prediction structure, and the input data is nonlinearly mapped through the step-by-step operation of the input layer, the hidden layer and the output layer. The focus is on parameter optimization during the training process, and the network weights and biases are adjusted by the back propagation algorithm to make the model fit the historical data to the best state. After the model training is completed, the obtained saltwater tide tracing prediction model is verified for many times, and the prediction accuracy of the model is verified by the historical data backtesting method to ensure that the model has a high prediction ability for saltwater tide tracing under different periods and different tidal conditions. Next, the verified models will be packaged and uploaded to the designated cloud platform through the data transmission protocol. The cloud platform serves as the model execution environment and provides the model's real-time prediction function. After receiving the real-time estuary tide and runoff data, the platform will call the prediction model to output the trend and possibility of future saltwater tides, providing decision makers with a basis for saltwater tide prevention and control, thereby realizing the intelligent management of the estuary's ecological environment.

[0086] Preferably, step S1 comprises the following steps:

[0087] Step S11: Acquire historical environmental data of the estuary;

[0088] Step S12: Analyze the width change of the river basin at the estuary to obtain the river basin width change data;

[0089] Step S13: Based on the river basin width change data, the upstream and downstream drop correlation analysis of the river is performed to obtain width change drop correlation data.

[0090] In the embodiment of the present invention, the historical environmental data of the estuary must first be obtained, and the required data include long-term monitoring data of various hydrological and environmental factors such as water level, flow, tidal level, river width change, sediment content, terrain change, salinity, temperature, wind speed, wind direction, etc. The acquisition method is to access the data platform of hydrological monitoring stations and environmental monitoring centers at all levels, query the long-term observation data related to the estuary area, and extract key data information. In the data collection process, ensure that the data covers the estuary conditions in different years, seasons, and climatic conditions, and pay special attention to the time series integrity and spatial distribution consistency of the data. After obtaining the data, the original data is preprocessed, including data denoising, outlier identification and elimination, to ensure the authenticity and validity of the data. Through data sorting and archiving, a standardized estuary historical environmental data file is formed to provide detailed basic information for subsequent analysis steps. It is necessary to conduct a detailed change analysis of the width of the river basin at the estuary. The specific operation is to use remote sensing images, historical aerial photos, geographic information system (GIS) data, etc. to extract the width of the river channel in multiple periods. First, high-resolution remote sensing images of different periods are selected to monitor the width change of the river in time series. The image segmentation algorithm is used to accurately extract the river boundary line, and the width change amplitude of each period is calculated by image registration and difference method. The cross-section of the river at each time node is analyzed, and the river width change curve is drawn in combination with the longitudinal section analysis of the river. The narrowing or expansion area and change rate of the river in different time periods are calculated. In order to improve the accuracy of the analysis, the field surveying and mapping data are used for correction, and the deviation between the interpretation results of the remote sensing images and the actual surveying and mapping data is compared to ensure the reliability of the analysis results. Finally, the width change data of each period are integrated to generate "river basin width change data", which provides an accurate basis for width change for the subsequent upstream and downstream drop correlation analysis. The drop correlation analysis of the upstream and downstream of the river is carried out, and the "river basin width change data" obtained previously is used as the basic data. First, by mapping the elevation data of the upstream and downstream of the estuary, the longitudinal section profile of the river is obtained, and the width change is superimposed and analyzed with the elevation data of the river. Specific operations include: using digital elevation model (DEM) data, combined with actual surveyed elevation points, to construct the longitudinal profile of the upstream and downstream river channels. According to the overlapping area of ​​width change and elevation change, the drop value of the river channel is calculated, and the upstream and downstream water level difference and flow velocity changes are analyzed by differential method. Then, the drop of different river sections is statistically analyzed to identify the areas where the width change affects the kinetic energy of the water flow, especially the bottleneck section of the river channel and the water flow acceleration section. The degree of influence of the river channel drop on the water flow is further calculated by the gravity hydrodynamic formula, and the width change data is correlated with the drop distribution to determine the characteristic values ​​of the key drop correlation points. Finally, the various analysis data are integrated to generate "width change drop correlation data" to provide accurate physical conditions for subsequent hydrological effects and saltwater tide upstream analysis.

[0091] Preferably, step S2 comprises the following steps:

[0092] Step S21: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively;

[0093] Step S22: performing historical precipitation attenuation calculation on the flood season precipitation data to obtain flood season precipitation attenuation data;

[0094] Step S23: Calculate the upstream runoff reciprocity for the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the non-flood season / flood season upstream runoff reciprocity data;

[0095] Step S24: Based on the non-flood season / flood season upstream runoff reciprocity data, the width change drop correlation data is analyzed for downstream runoff changes to obtain non-flood season / flood season downstream runoff change data.

[0096] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0097] Step S21: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively;

[0098] In an embodiment of the present invention, the historical environmental data of the estuary is accurately extracted for non-flooding season and flooding season precipitation. The specific operation steps include: first, according to the historical precipitation data set, the long-term precipitation data of the estuary and its upstream area are extracted, and the time is marked, and the data is classified according to year, month and specific time period. The non-flooding season is usually the dry season, from October to March of the following year, and the flooding season is from June to September. The precipitation data of these periods are summarized and counted respectively. The multidimensional precipitation analysis method is adopted to perform weighted accumulation of daily precipitation data to obtain the total precipitation of each period. Through data cleaning, the missing values ​​and outliers are corrected to ensure the continuity and accuracy of the data. Finally, the extracted precipitation data is archived to generate "non-flooding season precipitation data" and "flooding season precipitation data" for subsequent hydrological changes and runoff analysis.

[0099] Step S22: performing historical precipitation attenuation calculation on the flood season precipitation data to obtain flood season precipitation attenuation data;

[0100] In the embodiment of the present invention, the precipitation attenuation calculation of the historical period is performed on the precipitation data of the flood season, the purpose of which is to evaluate the decreasing or increasing trend of precipitation in different years. The specific method is: first, according to the time series of the precipitation data of the flood season, the precipitation data of adjacent years are compared year by year to calculate the annual precipitation change range. Using the attenuation coefficient formula, the fluctuation range of the historical precipitation in the same period is quantitatively analyzed, and the attenuation factor of the precipitation in each year is calculated to identify the change trend of the precipitation. Through the segmented statistical method, the precipitation in different stages of the flood season (such as the early flood season, the middle flood season, and the late flood season) is refined and processed, and the precipitation attenuation rate of each stage is calculated. For the extreme points and mutation points of precipitation, the linear regression method is used to correct the long-term change trend of precipitation data. Finally, the "flood season precipitation attenuation data" is generated to provide precipitation attenuation characteristic parameters for runoff analysis.

[0101] Step S23: Calculate the upstream runoff reciprocity for the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the non-flood season / flood season upstream runoff reciprocity data;

[0102] In the embodiment of the present invention, the reciprocating degree of upstream runoff is calculated for the "width change drop correlation data" according to the "non-flood season precipitation data" and the "flood season precipitation attenuation data". The calculation of the reciprocating degree of upstream runoff reflects the flow fluctuation of upstream runoff under different seasons and precipitation conditions. The operation steps include: first, combining the precipitation data with the river width change data, and using the hydraulic balance formula to calculate the flow change rate of the upstream runoff. Secondly, using the reciprocating degree calculation formula, the back-and-forth fluctuation of the runoff at each section of the river channel is evaluated, and the reciprocating characteristics of the runoff are obtained by calculating the flow reversal frequency of the runoff under different water levels and flow conditions. The runoff data of the non-flood season and the flood season are calculated separately, with special attention paid to the conversion frequency between high flow and low flow, to ensure that the calculation results can truly reflect the degree of runoff reciprocating. Finally, the "non-flood season / flood season upstream runoff reciprocating degree data" is obtained, which is used to analyze the impact of upstream water volume changes on downstream.

[0103] Step S24: Based on the non-flood season / flood season upstream runoff reciprocity data, the width change drop correlation data is analyzed for downstream runoff changes to obtain non-flood season / flood season downstream runoff change data.

[0104] In the embodiment of the present invention, the downstream runoff change analysis is performed on the "width change drop correlation data" based on the "non-flood season / flood season upstream runoff reciprocity data". The specific implementation operation includes: combining the upstream runoff reciprocity data with the river width change and drop correlation data, and using the hydrodynamic analysis method to establish a relationship model between the downstream flow and the upstream runoff reciprocity. By quantitatively analyzing the impact of the upstream runoff reciprocity on the downstream hydraulic conditions, the flow velocity and flow rate change trends of different downstream river sections are calculated. The cross-section analysis method is used to accurately identify key hydrodynamic change points in the downstream river channel, such as water swirls, vortex generation areas, and areas where the flow rate slows down or increases. The analysis results are further verified and adjusted in combination with actual measurement data to ensure accurate reflection of downstream runoff changes. Finally, the "non-flood season / flood season downstream runoff change data" is output to provide a scientific basis for subsequent tidal effects and saltwater tide upstream predictions.

[0105] Preferably, step S23 includes the following steps:

[0106] Step S231: Calculate the upstream average flow rate in the non-flood season / flood season based on the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the upstream average flow rate data in the non-flood season / flood season;

[0107] Step S232: Calculate the upstream flow amplitude of the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data to obtain the non-flood season / flood season upstream flow amplitude data;

[0108] Step S233: Calculate the upstream positive and negative flow averages according to the non-flood season / flood season upstream flow variation data and the non-flood season / flood season upstream average flow data to obtain the upstream positive and negative flow average data;

[0109] Step S234: Calculate the upstream runoff reciprocity based on the non-flood season / flood season upstream flow variation data and the upstream positive and negative flow mean data to obtain the non-flood season / flood season upstream runoff reciprocity data.

[0110] In the embodiment of the present invention, the average flow rate of the upstream non-flood season and flood season is calculated by combining the "non-flood season precipitation data" and the "flood season precipitation attenuation data" with the "width change drop correlation data". The specific operation includes: first, using the historical precipitation data and the width change and drop information of the upstream river channel, the flow statistical model of the non-flood season and the flood season is established. The data of the non-flood season and the flood season are imported respectively, and based on the continuous flow measurement method, the instantaneous flow of each section of the upstream in each time period is accumulated and calculated to obtain the daily and monthly average flow. The data is standardized to remove the influence of abnormal flow peak on the mean value to ensure the accuracy of the calculation. The multi-stage integration method is used to sum the average flow of each section, and the weighted average processing is performed on each basin point, and finally the "non-flood season / flood season upstream average flow data" is obtained to provide key basic data for the upstream runoff characteristics analysis. The upstream flow amplitude is calculated for the "width change drop correlation data" through the "non-flood season precipitation data" and the "flood season precipitation attenuation data". The specific implementation process is as follows: First, according to the historical flow records of the estuary and upstream areas, the flow fluctuations in each period of the non-flood season and the flood season are analyzed in detail, and the maximum and minimum flow values ​​in each period are extracted. The flow amplitude calculation formula is used to analyze the difference between the maximum flow and the minimum flow to calculate the upstream flow amplitude. Combined with the hydrological division and the characteristics of river channel changes, multi-point sampling analysis of the flow amplitude is carried out, and time series statistics are performed to identify periodic and sudden fluctuations. The difference method is further used to refine the flow amplitude, analyze the amplitude characteristics in different time periods, and finally generate "non-flood season / flood season upstream flow amplitude data", which is used for subsequent flow mean and reciprocity calculations. The "non-flood season / flood season upstream flow amplitude data" and "non-flood season / flood season upstream average flow data" are used to calculate the upstream positive and negative flow mean. The operation steps include: First, the upstream flow data is symbolized, and all data greater than the average flow is classified as positive flow, and data less than the average flow is classified as negative flow. The positive and negative flows are accumulated and calculated separately, and the changing trend of the flow under different conditions is determined by counting the mean of the positive and negative flows point by point. The two-way mean analysis method is adopted to comprehensively consider the frequency and amplitude of each flow point, accurately calculate the frequency of flow change through the frequency curve, and adjust the weight to ensure that the mean of positive and negative flows can fully reflect the flow characteristics of the upstream basin. Finally, the "upstream positive and negative flow mean data" is calculated and output to provide a flow mean basis for the analysis of reciprocity. The calculation of the reciprocity of upstream runoff is based on the "non-flood season / flood season upstream flow amplitude data" and "upstream positive and negative flow mean data", with the aim of quantitatively evaluating the reciprocating characteristics of upstream runoff. The specific operations include: first, using the flow amplitude data, calculate the number of frequent reversals of flow on each section to quantify the reciprocating degree of flow changes.Combined with the average data of positive and negative flows, the intensity and impact range of flow reversal are judged, and the frequency superposition method is used to count the number of reciprocating fluctuations and their distribution in the entire basin. The runoff reciprocity formula is used to perform multiple rounds of integration on the time series data of positive and negative flows to obtain the overall reciprocity index. The reciprocity index is compared with the historical typical year data to evaluate the abnormal reciprocating behavior of upstream runoff and ensure the scientificity and accuracy of the calculation results. Finally, the "non-flood season / flood season upstream runoff reciprocity data" is generated to guide the subsequent analysis and prediction of the upstream effect of saltwater tides.

[0111] Preferably, step S3 comprises the following steps:

[0112] Step S31: analyzing the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season;

[0113] Step S32: simulating the wind mixing effect according to the tidal superposition effect data in the non-flood season / flood season and the historical environmental data of the estuary to obtain the wind mixing effect data;

[0114] Step S33: analyzing the upstream action of saltwater tides in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season and the wind mixing effect data, and obtaining the upstream action data of saltwater tides in the non-flood season / flood season.

[0115] In the embodiment of the present invention, the tidal superposition effect analysis is performed through the "non-flood season / flood season downstream runoff change data" to accurately evaluate the impact of tides on runoff. The specific implementation process is: first, the downstream runoff change data is divided into time periods according to the non-flood season and the flood season, and the corresponding tidal periodicity data are extracted respectively. Using the tidal water level monitoring data and the downstream flow velocity change data, the superposition principle of hydraulics is applied to calculate the degree of influence of the tide on the downstream runoff. Combined with the phase relationship between the instantaneous flow velocity and the tidal water level, the tidal superposition effect is analyzed time by time period to identify the impact of the tide on the downstream flow velocity and flow direction. The Fourier transform method is used to convert the tidal superposition effect data from the time domain to the frequency domain, and the contribution rate of the tidal frequency to the runoff change is further analyzed. Finally, through the tidal-runoff interaction calculation, the analysis results are output as "non-flood season / flood season tidal superposition effect data" for subsequent wind mixing simulation. The wind mixing effect simulation is carried out according to the "non-flood season / flood season tidal superposition effect data" and "estuary historical environmental data" to quantify the mixing effect of wind on tide and runoff. The specific operation steps include: first, obtain the wind data in the non-flood season and flood season, including characteristic parameters such as wind speed, wind direction, and wind frequency. The wind data is segmented and statistically analyzed to distinguish the mixing intensity under different wind speed ranges. Based on the tidal superposition effect data, a dynamic model of wind-tidal interaction is constructed to simulate the process of wind pushing, rotating and turbulent formation on the water body. The wind speed, tidal flow velocity, and flow direction are coupled and calculated, and the mixing effect of wind on the tide is quantified by comparing the degree of water mixing under different wind speed conditions. The particle tracking method is used to track and analyze the water path during the mixing process to further verify the influence of wind mixing on the upstream of the tide. Finally, the "wind mixing effect data" is output to provide dynamic input parameters for the subsequent analysis of the upstream effect of salt tide. Based on the "non-flood season / flood season tidal superposition effect data" and "wind mixing effect data", the non-flood season / flood season salt tide upstream action analysis is carried out to clarify the upstream behavior of salt tide under different hydrological conditions. The specific implementation process includes: first, extracting the key parameters of tidal frequency and tidal height in the non-flood season and flood season tidal superposition effect data, and coupling them with the wind speed and water mixing degree in the wind mixing effect data. By calculating the salt tide diffusion range during the peak tidal period and the period with high wind speed, a salt tide upstream prediction model is established. Using the salinity gradient measurement method, the salinity changes under different tidal and wind conditions are analyzed, and the model parameters of the upstream action are calibrated in combination with the salinity profile data measured on site. The salt tide diffusion distance calculation method is used to simulate the salt tide upstream limit under different conditions, and analyze its impact on the upstream freshwater environment. Finally, the "non-flood season / flood season salt tide upstream action data" is formed to provide a scientific basis for salt tide risk assessment and prevention and control strategy formulation.

[0116] Preferably, step S31 includes the following steps:

[0117] Step S311: Acquire tidal seasonal regularity data;

[0118] Step S312: Analyze the tidal phase migration effect on the tidal seasonal regularity data according to the downstream runoff variation data in the non-flood season / flood season to obtain the tidal phase migration data;

[0119] Step S313: analyzing the tidal level-runoff intensity coupling effect according to the downstream runoff variation data in the non-flood season / flood season and the tidal phase migration data to obtain the tidal level-runoff coupling intensity data;

[0120] Step S314: performing a superimposed regional diffusion numerical analysis on the downstream runoff change data in the non-flood season / flood season and the tidal phase migration data based on the tide-runoff coupling intensity data to obtain superimposed regional diffusion numerical data;

[0121] Step S315: extracting tide-runoff interaction dynamic features from the superimposed area diffusion numerical data to obtain tide-runoff interaction dynamic data;

[0122] Step S316: Analyze the tidal superposition effect based on the superposition area diffusion numerical data and the tide-runoff interaction dynamic data to obtain the non-flood season / flood season tidal superposition effect data.

[0123] In the embodiment of the present invention, it is necessary to extract seasonal regularity data of tides from the data of long-term hydrological monitoring stations. The specific operation includes: sorting the tidal monitoring data, and screening out the time series data such as tidal water level, tidal range, and tidal time within at least 10 years. The data is statistically analyzed by season, the changing characteristics of tidal water level in spring, summer, autumn and winter are analyzed, and the average value, maximum value, minimum value and corresponding occurrence time of tidal water level in different seasons are extracted. Combined with key factors such as moon phase and tidal cycle, the high and low peak periods of seasonal tides are identified. The tidal data is extracted for periodicity by time series analysis, the main cycle and phase characteristics of seasonal tides are obtained, and the tidal intensity and frequency of different seasons are comprehensively evaluated, and finally the "tidal seasonal regularity data" is obtained to provide basic data support for the subsequent tidal phase migration effect analysis. The tidal phase migration effect analysis of the "tidal seasonal regularity data" is carried out according to the "non-flood season / flood season downstream runoff change data", the purpose of which is to evaluate the impact of downstream runoff changes on the tidal phase. The operation steps include: first, compare and analyze the downstream runoff change data with the tidal seasonal law data, focusing on the time overlap between the runoff peak period and the tidal high and low tide phases. Use the phase difference calculation method to compare the tidal phase offset between the non-flood season and the flood season, and calculate the degree of runoff shift on the tidal phase. Use the time difference correction method to fine-tune the tidal phase day by day and hour by hour, and analyze the impact of runoff changes on the time displacement of tidal peaks and troughs. By summarizing the analysis results into a time series phase migration diagram, extracting key phase change features, and obtaining "tidal phase migration data", it provides accurate tidal phase information for subsequent coupling effect analysis. According to the "non-flood / flood season downstream runoff change data" and "tidal phase migration data", the tidal level-runoff intensity coupling effect analysis is carried out to quantitatively evaluate the interaction intensity between tidal level and runoff. The specific operation is: first, extract the peak and low value data of the tidal level in the tidal phase migration data, and match them with the high and low flow data of runoff changes. The coupling intensity calculation method is used to quantify the correlation between tide level and runoff. The coupling intensity values ​​under different phases are obtained by calculating the amplitude and direction of the influence of the peak tide level on the runoff velocity. Further cross-analysis is performed on different tide level and runoff combinations to construct the tide level-runoff intensity matrix and analyze its dynamic changes in the non-flood season and flood season. Finally, the "tide level-runoff coupling intensity data" is output as the input data for the subsequent numerical analysis of superimposed regional diffusion. Based on the "tide level-runoff coupling intensity data", the "non-flood season / flood season downstream runoff change data" and "tidal phase migration data" are numerically analyzed for superimposed regional diffusion to evaluate the diffusion characteristics of tides and runoff under different phases. The implementation steps are as follows: First, the tide level-runoff coupling intensity data is paired with the tide level time series in the tidal phase migration data to calculate the diffusion effect when the tide level and runoff are superimposed on each time period. The numerical integration method is used to calculate the diffusion speed and range of the water body in the superimposed area, and the changes in the water quality gradient during the diffusion process are tracked in real time.Combined with the actual terrain data, the diffusion values ​​are spatially calibrated and the regional boundary conditions in the diffusion model are adjusted. The numerical analysis results are presented in graphical form to accurately identify the key diffusion areas where tides and runoff are superimposed. Finally, the "superimposed area diffusion numerical data" is generated to provide a basis for the subsequent extraction of tide-runoff interaction dynamic characteristics. Tide-runoff interaction dynamic characteristics are extracted from the "superimposed area diffusion numerical data" to identify the interaction characteristics of tides and runoff in space and time. The operation process includes: by segmenting the hydrodynamic data of the superimposed area, the velocity distribution characteristics under different tide levels and runoff conditions are identified. Using the hydrodynamic feature extraction algorithm, the hydrodynamic characteristics of the key areas are parameterized, such as vortex intensity, flow direction change rate, etc. Combined with the velocity field information in the diffusion numerical data, the driving force and resistance of the tide on the runoff are calculated, and the tide-runoff interaction coefficient is extracted. The vorticity analysis method is further used to identify the rotation and mixing characteristics of tides and runoff in different time periods. Finally, the "tide-runoff interaction dynamic data" is formed to provide a dynamic basis for the analysis of tidal superposition effects. The tidal superposition effect analysis is carried out by combining the "superimposed regional diffusion numerical data" and the "tidal-runoff interaction dynamic data" to quantify the superimposed influence of tides in the non-flood season and the flood season. The specific operation steps are: Integrate the tide-runoff interaction dynamic data with the diffusion numerical data, and calculate the superimposed effect of tides on runoff in different seasons by comparing the flow velocity, diffusion range and dynamic characteristics in different time periods. Apply the dynamic superposition model to cumulatively calculate the interaction force of tides and runoff, and identify the superimposed peak effect of tides in different phases and seasons. Compare the analysis results with the measured tidal level and runoff changes to verify the accuracy of the superimposed effect. Finally, the "non-flood season / flood season tidal superposition effect data" is generated to provide accurate input for the saltwater tide upstream prediction method.

[0124] Preferably, step S4 comprises the following steps:

[0125] Step S41: Time series marking of the saltwater tide upstream action data in the non-flood season / flood season to obtain the saltwater tide upstream action time series data in the non-flood season / flood season;

[0126] Step S42: performing neural network structure matching adjustment on the non-flooding season / flooding season saltwater tide tracing time series data to obtain non-flooding season / flooding season saltwater tide tracing network structure matching data;

[0127] Step S43: constructing a saltwater tide upstream prediction model based on the non-flood season / flood season saltwater tide upstream network structure matching data based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model.

[0128] In the embodiment of the present invention, the "non-flood season / flood season salt tide tracing action data" is time-series labeled for the purpose of obtaining accurate salt tide tracing time-series change rules. First, the start time, duration, peak time and corresponding water level data of the salt tide tracing are extracted from the obtained non-flood season and flood season salt tide tracing action data. Time series analysis is performed on these data, and the change process of the salt tide tracing is gradually labeled in combination with the time nodes of tidal water level and runoff changes. Using interpolation and trend analysis technology, the key period of salt tide tracing is subdivided into multiple time nodes, and the traceability intensity and speed characteristics of each node are analyzed. The time series labeling results are summarized to form a data set, and the salt tide tracing behavior of each period is recorded and marked on the corresponding time axis to obtain the "non-flood season / flood season salt tide tracing action time series data", which provides a time series input basis for the matching adjustment of the subsequent neural network structure. The neural network structure matching adjustment is performed on the "non-flood season / flood season salt tide tracing action time series data" to optimize the input structure of the prediction model. First, the feature extraction algorithm is used to quantize the time series data, and the key feature parameters in the time series (such as the amplitude, frequency, and rate of change of the salt tide) are extracted and converted into a feature matrix that can be recognized by the neural network. Next, the existing neural network structure is matched and adjusted by the network structure optimization method to ensure that the network can fully capture the temporal change characteristics of the salt tide. During the adjustment process, the number of nodes and connection weights of the input layer, hidden layer, and output layer are optimized in turn, so that the network structure maintains a high degree of matching with the temporal data characteristics of the salt tide. After multiple iterative adjustments and precision tests, the "non-flooding / flooding season salt tide tracing network structure matching data" is finally output to provide a structured data basis for the next step of building a prediction model. The prediction model of the "non-flooding / flooding season salt tide tracing network structure matching data" is constructed based on the recurrent neural network algorithm, with the goal of forming an accurate salt tide tracing prediction model. First, the network structure matching data is imported into the recurrent neural network (RNN) framework, and the salt tide tracing features in the data are analyzed layer by layer using the RNN's temporal learning ability. Through the iterative process of feedforward and feedback, the network can capture the long-term dependencies and short-term volatility characteristics in the time series. Next, the model is trained, and the connection weights of the network are adjusted through multiple rounds of learning and error back propagation to improve the model's prediction accuracy for saltwater tide upstream behavior. During the training process, the convergence speed and prediction accuracy of the model are optimized through continuous verification and parameter adjustment, so that the model has a high adaptability to the saltwater tide upstream characteristics in both the non-flood season and the flood season. Finally, the "saltwater tide upstream prediction model" is obtained, which can predict the saltwater tide upstream process under different conditions in real time, and provide scientific support for saltwater tide risk management and prevention and control.

[0129] Preferably, step S42 includes the following steps:

[0130] Step S421: using the hidden layer in the bidirectional RNN to perform overfitting optimization processing on the time series data of the saltwater tide tracing back in the non-flooding season / flooding season, and obtaining the fitting data of the saltwater tide tracing back in the non-flooding season / flooding season;

[0131] Step S422: gradient optimization is performed on the non-flood season / flood season saltwater tide upward effect fitting data through a multi-layer dense connection model to obtain non-flood season / flood season saltwater tide upward effect gradient optimization data;

[0132] Step S423: adjusting the error loss of the non-flooding season / flooding season salty tide upward action time series data according to the non-flooding season / flooding season salty tide upward action fitting data and the non-flooding season / flooding season salty tide upward action gradient optimization data to obtain the non-flooding season / flooding season salty tide upward action error loss adjustment data;

[0133] Step S424: performing neural network structure matching adjustment according to the non-flooding season / flooding season salty tide tracing back fitting data, the non-flooding season / flooding season salty tide tracing back gradient optimization data and the non-flooding season / flooding season salty tide tracing back error loss adjustment data to obtain the non-flooding season / flooding season salty tide tracing back network structure matching data.

[0134] In the embodiment of the present invention, the hidden layer in the bidirectional RNN (bidirectional recurrent neural network) is used to perform overfitting optimization on the "non-flooding season / flooding season saltwater tide tracing time series data". First, the bidirectional RNN is configured to read the forward and reverse information of the input time series data at the same time to ensure that the time-dependent characteristics of the data are fully learned. Through the back propagation mechanism of the hidden layer, the subtle changes and features of the input data are learned multiple times to capture the detailed fluctuation characteristics of the saltwater tide tracing. In this process, the focus of overfitting optimization is to adjust the connection weights of the hidden layer so that it suppresses non-significant features, and reduce excessive dependence on training data through regularization means to improve the generalization ability of the model. Finally, through multiple optimization iterations, the "non-flooding season / flooding season saltwater tide tracing effect fitting data" is output, which provides a high-precision fitting result for subsequent gradient optimization. Gradient optimization is performed on the "non-flooding season / flooding season saltwater tide tracing effect fitting data" to improve the training accuracy of the fitting data. In the implementation process of gradient optimization, a multi-layer dense connection model is first used to construct an optimization framework, which passes the fitting data layer by layer through the fully connected layer and calculates the current gradient value in each layer. Through multiple forward propagation and back propagation, the gradient direction and amplitude of each layer are gradually adjusted to reduce the error of the model to the input data. The gradient descent method is used to dynamically adjust the learning rate of each layer to avoid the problem of gradient explosion or disappearance, and optimize the convergence speed of the gradient. After multiple iterations and error correction, the "non-flood season / flood season salt tide up-gradient optimization data" is obtained. This data can reflect the response degree of the model to the time series characteristics under different gradients, laying the foundation for the subsequent error loss adjustment. Combined with the "non-flood season / flood season salt tide up-gradient optimization data" and "non-flood season / flood season salt tide up-gradient optimization data", the "non-flood season / flood season salt tide up-gradient time series data" is adjusted for error loss. First, the fitting data is compared with the actual time series data, and the error value is calculated through the loss function. The error loss adjustment process uses error evaluation methods such as mean square error (MSE) to gradually analyze the deviation between the fitting data and the real data. Combined with the gradient optimization data, the error curve is fine-tuned, and the learning process of the neural network is accurately corrected through the optimization solution of the loss function. The adjusted error value is used to reversely adjust the connection weights and bias values ​​of the neural network to minimize the prediction error, and finally output the "non-flood season / flood season saltwater tide traceback error loss adjustment data" to provide a corrected time series basis for the network structure matching adjustment. According to the "non-flood season / flood season saltwater tide traceback effect fitting data", "non-flood season / flood season saltwater tide traceback gradient optimization data" and "non-flood season / flood season saltwater tide traceback error loss adjustment data", the neural network structure is matched and adjusted. First, combined with the fitting data and gradient optimization results, the weak links of the neural network in time series data learning are identified, and the network's feature capture ability is adjusted by systematically optimizing the network hierarchy structure and connection weights.This process improves the response sensitivity of the network when processing the time series data of saltwater upwelling through parameter tuning and inter-layer weight reconstruction. The adjusted network relearns the error loss adjustment data through the feedback mechanism, thereby further optimizing the network's matching of time series characteristics. Finally, through repeated adjustments, the "non-flood season / flood season saltwater upwelling network structure matching data" is obtained, providing accurate data support for the subsequent construction of saltwater upwelling prediction models.

[0135] Preferably, step S43 includes the following steps:

[0136] Step S431: dividing the non-flooding season / flooding season saltwater tide tracing network structure matching data into a training set and a test set, respectively obtaining a non-flooding season / flooding season saltwater tide tracing network structure training set and a non-flooding season / flooding season saltwater tide tracing network structure test set;

[0137] Step S432: Based on the hidden layer of the recurrent neural network algorithm, neuron training matching is performed on the non-flood season / flood season saltwater tide traceability network structure training set to obtain training set neuron matching data;

[0138] Step S433: setting the sequence length of the training set neuron matching data to obtain neuron matching sequence length data;

[0139] Step S434: constructing an initial saltwater tide tracing prediction model for the non-flooding season / flooding season saltwater tide tracing network structure training set based on the training set neuron matching data and neuron matching sequence length data to obtain an initial saltwater tide tracing prediction model;

[0140] Step S435: iteratively optimize the initial saltwater tide upstream prediction model through the non-flood season / flood season saltwater tide upstream network structure test set to obtain the saltwater tide upstream prediction model.

[0141] In the embodiment of the present invention, the "non-flooding season / flooding season saltwater tide tracing network structure matching data" is divided into a training set and a test set. In the specific implementation process, the random segmentation method is first used to divide the data according to a fixed ratio (such as 7:3 or 8:2) to ensure that the training set and the test set can represent the overall characteristics of the data. After the division, the training set is mainly used for parameter training of the neural network model, and the test set is used for model verification and evaluation. In the division process, the temporal characteristics of the data are strictly maintained to ensure that the data distribution of the training and test sets is consistent. Finally, the "non-flooding season / flooding season saltwater tide tracing network structure training set" and the "non-flooding season / flooding season saltwater tide tracing network structure test set" are output. These two data sets provide basic data support for subsequent neuron matching and model construction. Based on the hidden layer of the recurrent neural network (RNN) algorithm, the "non-flooding season / flooding season saltwater tide tracing network structure training set" is matched with neuron training. In the specific implementation process, by constructing the hidden layer structure of the RNN, the input and output relationship of the neuron is defined, and the connection weights of the hidden layer neurons are gradually optimized. First, each time series sample in the training set is input into the hidden layer, and the activation value of each neuron is calculated by forward propagation, and the interaction relationship between neurons is recorded. Then, the connection weight of each neuron is adjusted by the back propagation algorithm to achieve the best match with the characteristics of the training set data. This process will continue for multiple rounds of iterations until the output of the neuron is highly consistent with the expected results of the training set samples, and the "training set neuron matching data" is output, which provides a basis for the subsequent sequence length setting. The sequence length of neuron matching is set according to the "training set neuron matching data". During the implementation process, the optimal sequence length is determined by analyzing the matching situation at different time steps in the training set neuron matching data. First, the output error and matching accuracy of the neuron are calculated at each time step, and the corresponding sequence length is recorded. During the setting process, by adjusting the range of sequence length (such as 1 to 10 time steps), the sequence length value that can minimize the error and improve the matching accuracy is selected. Finally, the "neuron matching sequence length data" is output, which is used to guide the construction of the initial saltwater tide upward prediction model, so that the model can accurately capture the time series characteristics of the saltwater tide upward. Based on the "training set neuron matching data" and "neuron matching sequence length data", the initial saltwater tide upstream prediction model is constructed for the "non-flood season / flood season saltwater tide upstream network structure training set". In the specific implementation, by inputting the training set data into the set RNN model, the model performs forward calculation according to the set sequence length and neuron matching relationship, and outputs the saltwater tide upstream prediction value for each time step. In this process, the structure of the model will be continuously adjusted to best fit the training set characteristics. The constructed initial model will perform multiple forward and reverse iterations on the training data, and gradually converge to a state with the minimum error. Finally, the "initial saltwater tide upstream prediction model" is output. The model has preliminary saltwater tide upstream prediction capabilities and provides a basic framework for subsequent iterative optimization.The "initial saltwater up-trace prediction model" is iteratively optimized using the "non-flood / flood season saltwater up-trace network structure test set". In the specific operation, the test set data is input into the initial prediction model, the difference between the model output value and the actual test set value is calculated, and the error size of the model is evaluated through the loss function. Combined with the error value, the parameters and structure of the model are adjusted through multiple iterations, and the weights and biases of the hidden layer neurons are optimized, so that the prediction error of the model for the test data is gradually reduced. After each round of iterative optimization, the prediction accuracy and generalization ability of the model are evaluated in real time until the prediction results of the model for the test set meet the expected standards. Finally, the "saltwater up-trace prediction model" is output. After multiple rounds of optimization, the model can accurately predict the saltwater up-trace behavior in the non-flood and flood seasons, and provide data support for relevant prevention and early warning measures.

[0142] Preferably, the present invention further provides a saltwater tide tracing prediction system based on artificial intelligence, which is used to execute the saltwater tide tracing prediction method based on artificial intelligence as described above, and the saltwater tide tracing prediction system based on artificial intelligence comprises:

[0143] The upstream and downstream drop analysis module is used to obtain historical environmental data of the estuary; the upstream and downstream drop correlation analysis of the estuary is carried out to obtain width change drop correlation data.

[0144] The runoff change analysis module is used to extract the precipitation in the non-flood season and flood season from the historical environmental data of the estuary, and obtain the precipitation data in the non-flood season and the precipitation data in the flood season respectively; the downstream runoff change analysis is performed on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and the downstream runoff change data in the non-flood season / flood season is obtained;

[0145] The salt tide upstream effect analysis module is used to analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season, and obtain the tidal superposition effect data in the non-flood season / flood season; analyze the salt tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season, and obtain the salt tide upstream effect data in the non-flood season / flood season;

[0146] The model building module is used to build a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on the recurrent neural network algorithm, obtain the saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

[0147] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0148] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A saltwater tide upstream prediction method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Acquire historical environmental data of the estuary; perform correlation analysis of the upstream and downstream drop of the river channel at the estuary to obtain width change drop correlation data; Step S2: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively; analyzing downstream runoff changes on the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation data, and obtaining downstream runoff change data in the non-flood season / flood season. Step S2 includes the following steps: Step S21: extracting precipitation in the non-flood season and the flood season from the historical environmental data of the estuary, and obtaining precipitation data in the non-flood season and precipitation data in the flood season respectively; Step S22: performing historical precipitation attenuation calculation on the flood season precipitation data to obtain flood season precipitation attenuation data; Step S23: Calculate the upstream runoff reciprocity of the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data to obtain the non-flood season / flood season upstream runoff reciprocity data. Step S23 includes the following steps: Step S231: Calculate the upstream average flow rate in the non-flood season / flood season based on the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data, and obtain the upstream average flow rate data in the non-flood season / flood season; Step S232: Calculate the upstream flow amplitude of the width change drop correlation data according to the non-flood season precipitation data and the flood season precipitation attenuation data to obtain the non-flood season / flood season upstream flow amplitude data; Step S233: Calculate the upstream positive and negative flow averages according to the non-flood season / flood season upstream flow variation data and the non-flood season / flood season upstream average flow data to obtain the upstream positive and negative flow average data; Step S234: Calculate the upstream runoff reciprocity based on the non-flood season / flood season upstream flow amplitude data and the upstream positive and negative flow mean data to obtain the non-flood season / flood season upstream runoff reciprocity data; Step S24: analyzing downstream runoff changes based on the width change drop correlation data of upstream runoff reciprocity in the non-flood season / flood season, and obtaining downstream runoff change data in the non-flood season / flood season; Step S3: Analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season; analyze the saltwater tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season to obtain the saltwater tide upstream effect data in the non-flood season / flood season; Step S4: construct a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

2. The artificial intelligence-based saltwater tide prediction method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire historical environmental data of the estuary; Step S12: Analyze the width change of the river basin at the estuary to obtain the river basin width change data; Step S13: Based on the river basin width change data, the upstream and downstream drop correlation analysis of the river is performed to obtain width change drop correlation data.

3. The artificial intelligence-based saltwater tide prediction method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: analyzing the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season to obtain the tidal superposition effect data in the non-flood season / flood season; Step S32: simulating the wind mixing effect according to the tidal superposition effect data in the non-flood season / flood season and the historical environmental data of the estuary to obtain the wind mixing effect data; Step S33: analyzing the upstream action of saltwater tides in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season and the wind mixing effect data, and obtaining the upstream action data of saltwater tides in the non-flood season / flood season.

4. The artificial intelligence-based saltwater tide upstream prediction method according to claim 3 is characterized in that: Step S31 The following steps are involved: Step S311: Acquire tidal seasonal regularity data; Step S312: Analyze the tidal phase migration effect on the tidal seasonal regularity data according to the downstream runoff variation data in the non-flood season / flood season to obtain the tidal phase migration data; Step S313: analyzing the tidal level-runoff intensity coupling effect according to the downstream runoff variation data in the non-flood season / flood season and the tidal phase migration data to obtain the tidal level-runoff coupling intensity data; Step S314: performing a superimposed regional diffusion numerical analysis on the downstream runoff change data in the non-flood season / flood season and the tidal phase migration data based on the tide-runoff coupling intensity data to obtain superimposed regional diffusion numerical data; Step S315: extracting tide-runoff interaction dynamic features from the superimposed area diffusion numerical data to obtain tide-runoff interaction dynamic data; Step S316: Analyze the tidal superposition effect based on the superposition area diffusion numerical data and the tide-runoff interaction dynamic data to obtain the non-flood season / flood season tidal superposition effect data.

5. The artificial intelligence-based saltwater tide upstream prediction method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: marking the time series data of the saltwater tide tracing back in the non-flooding season / flooding season to obtain the time series data of the saltwater tide tracing back in the non-flooding season / flooding season; Step S42: performing neural network structure matching adjustment on the non-flooding season / flooding season saltwater tide tracing time series data to obtain non-flooding season / flooding season saltwater tide tracing network structure matching data; Step S43: constructing a saltwater tide upstream prediction model based on the non-flood season / flood season saltwater tide upstream network structure matching data based on a recurrent neural network algorithm to obtain a saltwater tide upstream prediction model.

6. The artificial intelligence-based saltwater tide upstream prediction method according to claim 5, characterized in that: Step S42 includes the following steps: Step S421: using the hidden layer in the bidirectional RNN to perform overfitting optimization processing on the time series data of the saltwater tide tracing back in the non-flooding season / flooding season, and obtaining the fitting data of the saltwater tide tracing back in the non-flooding season / flooding season; Step S422: gradient optimization is performed on the non-flood season / flood season saltwater tide upward effect fitting data through a multi-layer dense connection model to obtain non-flood season / flood season saltwater tide upward effect gradient optimization data; Step S423: adjusting the error loss of the non-flooding season / flooding season salty tide upward action time series data according to the non-flooding season / flooding season salty tide upward action fitting data and the non-flooding season / flooding season salty tide upward action gradient optimization data to obtain the non-flooding season / flooding season salty tide upward action error loss adjustment data; Step S424: performing neural network structure matching adjustment according to the non-flooding season / flooding season salty tide tracing back fitting data, the non-flooding season / flooding season salty tide tracing back gradient optimization data and the non-flooding season / flooding season salty tide tracing back error loss adjustment data to obtain the non-flooding season / flooding season salty tide tracing back network structure matching data.

7. The artificial intelligence-based saltwater tide upstream prediction method according to claim 5, characterized in that: Step S43 includes the following steps: Step S431: dividing the non-flooding season / flooding season saltwater tide tracing network structure matching data into a training set and a test set, respectively obtaining a non-flooding season / flooding season saltwater tide tracing network structure training set and a non-flooding season / flooding season saltwater tide tracing network structure test set; Step S432: Based on the hidden layer of the recurrent neural network algorithm, neuron training matching is performed on the non-flood season / flood season saltwater tide traceability network structure training set to obtain training set neuron matching data; Step S433: setting the sequence length of the training set neuron matching data to obtain neuron matching sequence length data; Step S434: constructing an initial saltwater tide tracing prediction model for the non-flooding season / flooding season saltwater tide tracing network structure training set based on the training set neuron matching data and neuron matching sequence length data to obtain an initial saltwater tide tracing prediction model; Step S435: iteratively optimize the initial saltwater tide upstream prediction model through the non-flood season / flood season saltwater tide upstream network structure test set to obtain the saltwater tide upstream prediction model.

8. A saltwater tide upstream prediction system based on artificial intelligence, characterized in that: Used to execute the saltwater tide upstream prediction method based on artificial intelligence as claimed in claim 1, the saltwater tide upstream prediction system based on artificial intelligence comprises: The upstream and downstream drop analysis module is used to obtain historical environmental data of the estuary; the upstream and downstream drop correlation analysis of the estuary is carried out to obtain width change drop correlation data; The runoff change analysis module is used to extract the precipitation in the non-flood season and flood season from the historical environmental data of the estuary, and obtain the precipitation data in the non-flood season and the precipitation data in the flood season respectively; the downstream runoff change analysis is performed on the width change drop correlation data according to the precipitation data in the non-flood season and the precipitation data in the flood season, and the downstream runoff change data in the non-flood season / flood season is obtained; The salt tide upstream effect analysis module is used to analyze the tidal superposition effect according to the downstream runoff change data in the non-flood season / flood season, and obtain the tidal superposition effect data in the non-flood season / flood season; analyze the salt tide upstream effect in the non-flood season / flood season according to the tidal superposition effect data in the non-flood season / flood season, and obtain the salt tide upstream effect data in the non-flood season / flood season; The model building module is used to build a saltwater tide upstream prediction model for the saltwater tide upstream action data in the non-flood season / flood season based on the recurrent neural network algorithm, obtain the saltwater tide upstream prediction model, and send the saltwater tide upstream prediction model to the data cloud platform to perform saltwater tide upstream prediction.

Images