Reservoir flood control water level early warning system and method

Abstract

The present application relates to the field of flood control water level warning, and specifically discloses a reservoir flood control water level warning system and method. The system first obtains rainfall data along the upstream river in real time and monitors the water level changes at the reservoir entrance and key river sections. Then, the system uses artificial intelligence technology based on deep learning to analyze the collected rainfall data and water level data to capture the spatiotemporal correlation characteristics between rainfall data at various monitoring points along the river, as well as the spatiotemporal correlation characteristics between water level data at various monitoring points in the water level monitoring area. The system uses rainfall data as the main variable and water level data as the covariate, and conducts a time series interactive analysis of the two to reveal the mechanism of the influence of rainfall changes on water level changes, thereby realizing the prediction of reservoir water level change trends. In this way, the response speed and accuracy of reservoir flood control water level warnings can be effectively improved, the dependence on human resources can be reduced, and the maintenance cost of traditional hydrological models can be reduced.

Description

Technical Field

[0001] The present application relates to the field of flood control water level warning, and more specifically, to a reservoir flood control water level warning system and method. Background Art

[0002] As global climate change intensifies and extreme weather events become more frequent, reservoirs are facing an increasing risk of flood disasters. As crucial infrastructure for flood control, disaster reduction, and water resources management, reservoir water level fluctuations directly impact the safety of life and property of downstream residents and the stability of the ecological environment. Traditional reservoir monitoring and early warning systems often rely on manual observation and empirical judgment, which presents numerous limitations, including slow response times, low early warning accuracy, and high human resource consumption. Therefore, accurately predicting reservoir water level trends and issuing timely early warning information have become key issues in flood prevention efforts.

[0003] In this regard, the invention patent with publication number CN118430190A discloses a reservoir flood prevention water level warning system and method. It monitors the rainfall data along the upstream river and the water level changes at the reservoir entrance and key river sections in real time, and uses the hydrological model to predict and analyze the changing trend of the reservoir water level in a certain period of time in the future. Based on the water level prediction results, it assesses the risk level of the reservoir encountering flood disasters in the future period and issues corresponding early warning prompts.

[0004] However, the above methods mainly predict reservoir water level changes by constructing distributed hydrological models. Although they have achieved certain results in predicting reservoir water level changes, the construction and parameter setting of hydrological models are highly dependent on a large amount of historical data and empirical knowledge, including various geographical information and historical observation data of the upstream basin. As a result, the model construction and adjustment process is complicated and time-consuming. In addition, when faced with a complex and changeable natural environment, the model is difficult to adapt and update quickly, resulting in limited adaptability and generalization capabilities.

[0005] Therefore, an optimized reservoir flood control water level early warning system and method are expected. Summary of the Invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a reservoir flood control water level warning system and method, which first obtains rainfall data along the upstream river in real time, and monitors the water level changes at the reservoir entrance and key river sections, and then uses artificial intelligence technology based on deep learning to perform data analysis on the collected rainfall data and water level data to capture the spatiotemporal correlation characteristics between the rainfall data at each monitoring point along the river, as well as the spatiotemporal correlation characteristics between the water level data at each monitoring point in the water level monitoring area, and uses rainfall data as the main variable and water level data as the covariate, and performs a time series interactive analysis of the two to reveal the impact mechanism of rainfall changes on water level changes, thereby realizing the prediction of reservoir water level change trends. In this way, the response speed and accuracy of reservoir flood control water level warnings can be effectively improved, the dependence on human resources can be reduced, and the maintenance cost of traditional hydrological models can be reduced.

[0007] According to one aspect of the present application, a reservoir flood control water level early warning method is provided, comprising:

[0008] Obtain real-time rainfall data along upstream rivers and monitor water level changes at reservoir entrances and key river sections;

[0009] Based on the acquired rainfall data and water level data, the reservoir water level change trend in the future is predicted and analyzed to obtain the water level prediction results;

[0010] Based on the water level forecast results, the risk level of reservoir flood disasters in the coming period is assessed and corresponding warnings are issued;

[0011] Based on water level forecast results and risk assessment, a flood discharge scheduling plan is formulated and implemented.

[0012] According to another aspect of the present application, a reservoir flood control water level early warning system is provided, comprising:

[0013] The data structured processing module is used to perform structured processing on rainfall data and water level data to obtain the rainfall time series aggregation matrix along the upstream river and the time series aggregation matrix of the entire water level monitoring domain;

[0014] A main variable association coding module is used to use the rainfall time series aggregation matrix along the upstream river as the main variable, and perform association coding on the rainfall time series aggregation matrix along the upstream river to obtain an implicit feature vector of rainfall time series association along the river;

[0015] a covariate association coding module, configured to use the full water level monitoring domain time series aggregation matrix as a covariate, and perform association coding on the full water level monitoring domain time series aggregation matrix to obtain a full water level monitoring domain water level time series implicit coding aggregation matrix;

[0016] A dynamic interaction analysis module is used to perform a main-covariate dynamic interaction based on a fast scanning mechanism on the implicit characteristic vector of the rainfall time series along the river and the implicit coding aggregation matrix of the water level time series in the entire water level monitoring domain to obtain a main-covariate time series interaction response coding vector of the reservoir water level;

[0017] The water level prediction generation module is used to generate the water level prediction result based on the reservoir water level main-covariate time series interaction response coding vector.

[0018] Compared with the existing technology, the reservoir flood control water level warning system and method provided by this application first obtains rainfall data along the upstream river in real time and monitors the water level changes at the reservoir entrance and key river sections. Then, it uses artificial intelligence technology based on deep learning to analyze the collected rainfall data and water level data to capture the spatiotemporal correlation characteristics between rainfall data at each monitoring point along the river, as well as the spatiotemporal correlation characteristics between water level data at each monitoring point in the water level monitoring area. It uses rainfall data as the main variable and water level data as the covariate, and conducts a time series interactive analysis of the two to reveal the impact mechanism of rainfall changes on water level changes, thereby realizing the prediction of reservoir water level change trends. In this way, the response speed and accuracy of reservoir flood control water level warnings can be effectively improved, the dependence on human resources can be reduced, and the maintenance cost of traditional hydrological models can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 Flowchart of a reservoir flood prevention water level warning method according to an embodiment of the present application.

[0021] Figure 2 Schematic diagram of data flow of a reservoir flood control water level warning method according to an embodiment of the present application.

[0022] Figure 3 This is a flowchart of sub-step S1 of the reservoir flood control water level early warning method according to an embodiment of the present application.

[0023] Figure 4 This is a flowchart of sub-step S4 of the reservoir flood control water level warning method according to an embodiment of the present application.

[0024] Figure 5 This is a flowchart of sub-step S41 of the reservoir flood control water level warning method according to an embodiment of the present application.

[0025] Figure 6 4 is a block diagram of a reservoir flood control water level warning system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0027] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0028] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0029] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0030] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0031] As mentioned in the background technology above, patent CN118430190A proposes a reservoir flood control water level warning method, which includes: obtaining real-time rainfall data along the upstream river and monitoring the water level changes at the reservoir entrance and key river sections; based on the acquired rainfall data and water level data, predicting and analyzing the reservoir water level change trend in a certain period of time in the future to obtain water level prediction results; based on the water level prediction results, evaluating the risk level of the reservoir encountering flood disasters in the future period of time and issuing corresponding warnings; and formulating and implementing a flood discharge scheduling plan based on the water level prediction results and risk assessment.

[0032] The above-mentioned reservoir flood control water level warning method mainly predicts reservoir water level changes by constructing a distributed hydrological model. Although it has achieved certain results in predicting reservoir water level changes, the construction and parameter setting of the hydrological model are highly dependent on a large amount of historical data and empirical knowledge, including various geographical information and historical observation data of the upstream basin. As a result, the model construction and adjustment process is complicated and time-consuming. In addition, when faced with a complex and changeable natural environment, the model is difficult to adapt and update quickly, resulting in its limited adaptability and generalization ability. In response to this technical problem, this application proposes an optimized reservoir flood control water level warning method, which first obtains real-time rainfall data along the upstream river and monitors the water level changes at the reservoir entrance and key river sections. Then, it uses deep learning-based artificial intelligence technology to analyze the collected rainfall data and water level data to capture the spatiotemporal correlation characteristics between rainfall data at each monitoring point along the river, as well as the spatiotemporal correlation characteristics between water level data at each monitoring point in the water level monitoring area. It uses rainfall data as the main variable and water level data as the covariate, and conducts a time series interactive analysis of the two to reveal the impact mechanism of rainfall changes on water level changes, thereby realizing the prediction of reservoir water level change trends. In this way, the response speed and accuracy of reservoir flood control water level warnings can be effectively improved, the dependence on human resources can be reduced, and the maintenance cost of traditional hydrological models can be reduced.

[0033] Supported by modern information technology, obtaining real-time rainfall data along upstream rivers relies primarily on a widely distributed network of rain gauges. These stations are typically equipped with high-precision rain gauges, such as tipping bucket rain gauges, ultrasonic rain gauges, or laser rain gauges, which accurately measure precipitation and record time series information. The selection and placement of rain gauges are crucial, as they directly impact the representativeness and accuracy of the data. Ideally, the placement should cover key areas throughout the basin, particularly those with significant impacts on reservoir water levels, such as the confluence of major tributaries, low-lying terrain, and areas historically prone to flooding. Furthermore, given that different types of rain gauges may exhibit varying performance characteristics under different climatic conditions, in practice, a combination of multiple types of rain gauges may be used to average the data to ensure comprehensive and reliable data collection.

[0034] In addition to fixed ground monitoring points, mobile monitoring equipment also plays a vital role in obtaining real-time rainfall data. For example, unmanned aerial vehicles (UAVs) equipped with miniaturized, portable rainfall sensors can quickly survey large areas in a short period of time. Vehicle-mounted weather stations can quickly reach designated locations for temporary monitoring in emergency situations. These flexible and mobile monitoring methods not only fill in the gaps between fixed stations but also enable timely response to sudden heavy rainfall events, significantly enhancing the flexibility and emergency response capabilities of the overall monitoring system.

[0035] To monitor water level changes at reservoir entrances and key river sections, various types of automated water level monitoring equipment can be deployed there. These include, but are not limited to, float-type water level gauges, pressure-type water level gauges, and radar water level gauges. Float-type water level gauges utilize the principle of a float floating on the water surface, moving with the rise and fall of the water level, converting mechanical motion into an electrical signal output. Pressure-type water level gauges, based on the principles of hydrostatics, measure the pressure at the bottom of the water to determine the water level. Radar water level gauges use non-contact microwave radar technology to calculate the water level based on the time difference between the transmitted and received reflected waves. Each of these different types of water level gauges has its own advantages. For example, float-type water level gauges are suitable for shallow waters, pressure-type water level gauges are suitable for deep water environments, and radar-type water level gauges offer high accuracy and are maintenance-free. Selecting the appropriate instrument combination based on the specific application scenario and environmental conditions ensures the accuracy and reliability of monitoring data.

[0036] In addition to traditional water level gauges, high-definition cameras can be installed near key control points such as dams, spillways, and other important structures. These cameras are not only used to visually monitor water surface conditions and water flow rates, but can also assist in verifying the rationality of data provided by other sensors. For example, when a water level gauge indicates an abnormally high water level, the video image can be viewed to confirm whether there is an actual flood threat or data errors caused by a sensor failure. In addition, the cameras are equipped with infrared night vision capabilities, allowing for effective monitoring even at night or in low-light conditions. All cameras are connected to a central processing center, and the images they capture can be transmitted in real time and analyzed and processed to enable a rapid response to any emergencies.

[0037] In summary, through the above steps, rainfall data along the upstream river and water level changes at the reservoir entrance and key river sections can be systematically obtained, thus laying a solid foundation for subsequent data analysis and processing.

[0038] Figure 1 Flowchart of a reservoir flood prevention water level warning method according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the reservoir flood control water level warning method according to the embodiment of the present application. Figure 1 and Figure 2As shown, the reservoir flood prevention water level warning method includes the following steps: S1, performing structured processing on the rainfall data and the water level data to obtain a rainfall time series aggregation matrix along the upstream river and a time series aggregation matrix of the entire water level monitoring domain; S2, using the rainfall time series aggregation matrix along the upstream river as the main variable, performing association coding on the rainfall time series aggregation matrix along the upstream river to obtain a rainfall time series association implicit feature vector along the river; S3, using the time series aggregation matrix of the entire water level monitoring domain as the covariate, performing association coding on the time series aggregation matrix of the entire water level monitoring domain to obtain a water level time series implicit coding aggregation matrix of the entire water level monitoring domain; S4, performing main-covariate dynamic interaction based on a fast scanning mechanism on the rainfall time series association implicit feature vector along the river and the water level time series implicit coding aggregation matrix of the entire water level monitoring domain to obtain a reservoir water level main-covariate time series interaction response coding vector; S5, generating the water level prediction result based on the reservoir water level main-covariate time series interaction response coding vector.

[0039] In the above reservoir flood control water level warning method, the step S1 is to perform structured processing on the rainfall data and the water level data to obtain a rainfall time series aggregation matrix along the upstream river and a time series aggregation matrix of the entire water level monitoring domain. Figure 3 FIG. 1 is a flow chart of sub-step S1 of the reservoir flood prevention water level warning method according to an embodiment of the present application. Figure 3 As shown, the step S1 includes the following steps: S11, arranging the rainfall data into a rainfall time series aggregation matrix along the upstream river according to the monitoring point sample dimension and the time dimension; S12, arranging the water level data into a time series aggregation matrix of the entire water level monitoring domain according to the time dimension and the position sample dimension.

[0040] Specifically, the step S11 arranges the rainfall data according to the monitoring point sample dimension and the time dimension into a rainfall time series aggregation matrix along the upstream river. It should be understood that, considering that the rainfall data has significant spatiotemporal correlation, that is, there is a correlation between rainfall at different times and in different geographical locations. Therefore, in order to understand the rainfall conditions along the entire upstream river in a broader context, the present application further arranges the collected rainfall data according to the monitoring point sample dimension and the time dimension, so as to construct a matrix structure that can reflect the spatiotemporal correlation of rainfall between the various monitoring points along the river, and obtains the rainfall time series aggregation matrix along the upstream river, which is helpful to subsequently capture the temporal and spatial variation pattern of rainfall along the river.

[0041] Specifically, the step S12 arranges the water level data into the full water level monitoring domain time series aggregation matrix according to the time dimension and the position sample dimension. It should be understood that since the water level monitoring data also has similar spatiotemporal correlation, the present application also arranges the water level data into the full water level monitoring domain time series aggregation matrix according to the time dimension and the position sample dimension, so as to capture the changing patterns of the water level at different time points and different geographical locations. Among them, the position sample dimension represents the upstream and downstream position sequence relationship of each water level monitoring location, while the time dimension reflects the changing trend of the water level data at each location over time. Through this matrix structure, the water level changing trend of each location and the interaction between different locations can be analyzed more accurately.

[0042] Next, since the rainfall along the upstream river is directly related to the water inflow into the reservoir, this application uses the rainfall time series aggregation matrix along the upstream river as the main variable and the full water level monitoring domain time series aggregation matrix as the covariate, and conducts a time series interaction analysis on the two to reveal the impact of rainfall on water level changes, thereby predicting the changing trend of the reservoir water level.

[0043] Specifically, in the above-mentioned reservoir flood control water level warning method, step S2 uses the upstream river rainfall time series aggregation matrix as the main variable, and performs association coding on the upstream river rainfall time series aggregation matrix to obtain the river rainfall time series correlation implicit feature vector. In a specific example of the present application, step S2 includes: inputting the upstream river rainfall time series aggregation matrix into a river rainfall time series correlation feature extractor based on a converter structure to obtain the river rainfall time series correlation implicit feature vector. It should be understood that in order to capture the temporal and spatial correlation change pattern of rainfall along the upstream river, so as to better understand the impact of rainfall on reservoir water level changes, the present application uses a river rainfall time series correlation feature extractor based on a converter structure to perform global association coding processing on the upstream river rainfall time series aggregation matrix to obtain the river rainfall time series correlation implicit feature vector. Those skilled in the art should be aware that the converter structure is based on the self-attention mechanism, which can effectively handle long-distance dependency problems in sequence data, allowing the model to focus on the feature information of all input positions at the same time, thereby capturing the temporal correlation characteristics between the rainfall data of each monitoring point in the rainfall time series aggregation matrix along the upstream river on a global scale, effectively modeling the spatiotemporal correlation characteristics in the rainfall data, and then obtaining a feature representation that can reflect the spatiotemporal variation pattern of rainfall along the river, providing richer feature information for subsequent reservoir water level prediction tasks.

[0044] In the above-mentioned reservoir flood control water level warning method, step S3 uses the full water level monitoring domain time series aggregation matrix as a covariate and performs association coding on the full water level monitoring domain time series aggregation matrix to obtain the full water level monitoring domain water level time series implicit coding aggregation matrix. In a specific example of the present application, step S3 includes: inputting the full water level monitoring domain time series aggregation matrix into a sequence encoder based on an LSTM model to obtain the full water level monitoring domain water level time series implicit coding aggregation matrix. It should be understood that considering that the change in reservoir water level is not only affected by direct rainfall, but also by the hydrological conditions in the entire basin, in order to capture the changing patterns of water level monitoring data at different time points and different geographical locations, the present application uses a sequence encoder based on an LSTM model to perform sequence encoding processing on the full water level monitoring domain time series aggregation matrix. Specifically, the LSTM model, i.e., a long short-term memory network, can effectively process and memorize long-term dependency information in sequence data through its internal gating mechanism (such as input gate, output gate, and forget gate). In the technical solution of the present application, an LSTM model is used to perform forward propagation sequence encoding on the water level data at each water level monitoring location in the full water level monitoring domain time series aggregation matrix. The long-term memory capacity of the LSTM model can be used to effectively capture the water level variation patterns between different geographical locations, fully understand the potential impact of upstream water level changes on reservoir water levels, and thus provide a more accurate basis for predicting reservoir water level variation trends. In a specific implementation, each row vector in the full water level monitoring domain time series aggregation matrix (i.e., the water level data at each water level monitoring location) is used as an input sequence, and is processed layer by layer through multiple hidden layers of the LSTM model to ultimately obtain a hidden state sequence containing the cumulative effect of upstream water level changes, and combine it into a full water level monitoring domain water level time series implicit coding aggregation matrix, thereby achieving an in-depth analysis of the water level variation patterns.

[0045] In the above-mentioned reservoir flood control water level warning method, the step S4 performs a dynamic interaction of the main and covariates based on the fast scanning mechanism on the implicit characteristic vector of the rainfall time series association along the river and the implicit coding aggregation matrix of the water level time series of the entire water level monitoring domain to obtain the main and covariate time series interaction response coding vector of the reservoir water level. That is, in order to reveal the interaction between rainfall and water level changes, the present application further performs an interactive response analysis on the implicit characteristic vector of the rainfall time series association along the river and the implicit coding aggregation matrix of the water level time series of the entire water level monitoring domain, so as to predict the changing trend of the reservoir water level. In particular, considering that there may be different correlations between water level changes and rainfall in different geographical locations. Therefore, in order to improve the relevance and efficiency of data interaction response analysis, this application proposes a dynamic interaction method of principal-covariate based on a fast scanning mechanism, which quickly identifies and strengthens the key connection between rainfall and water level changes by learning the semantic correlation between the implicit feature vector of rainfall time series along the river and the water level change characteristics of each water level monitoring location, while filtering out the interference of irrelevant or noisy data, thereby achieving accurate prediction of reservoir water level change trends. Among them, Figure 4 FIG. 4 is a flow chart of sub-step S4 of the reservoir flood control water level warning method according to an embodiment of the present application. Figure 4 As shown, the step S4 includes the following steps: S41, based on the semantic correlation distribution between the implicit feature vector associated with the rainfall time series along the river and the implicit coding aggregation matrix of the water level time series of the entire water level monitoring domain, screening out a subsequence of the water level time series implicit coding vector of the water level monitoring domain from the implicit coding aggregation matrix of the water level time series of the entire water level monitoring domain; S42, performing cross-domain interactive fusion on the subsequence of the implicit feature vector associated with the rainfall time series along the river and the implicit coding vector of the water level time series of the water level monitoring domain to obtain the principal-covariate time series interactive response coding vector of the reservoir water level.

[0046] Specifically, Figure 5 FIG. 4 is a flow chart of sub-step S41 of the reservoir flood prevention water level warning method according to an embodiment of the present application. Figure 5As shown, the step S41 includes the steps of: S411, splitting the water level time series implicit coding aggregation matrix of the entire water level monitoring domain according to the position sample dimension to obtain a sequence of water level monitoring domain water level time series implicit coding vectors; S412, calculating the mutual information between the implicit feature vector associated with the rainfall time series along the river and each water level monitoring domain water level time series implicit coding vector in the sequence of the water level monitoring domain water level time series implicit coding vector to obtain a sequence of reservoir water level main-covariate fast semantic matching factors; S413, identifying the first maximum value and the second maximum value from the sequence of reservoir water level main-covariate fast semantic matching factors; S414, based on the positions of the first maximum value and the second maximum value in the sequence of the reservoir water level main-covariate fast semantic matching factors, determining a subsequence of the water level monitoring domain water level time series implicit coding vector that is quickly matched from the sequence of the water level monitoring domain water level time series implicit coding vector.

[0047] More specifically, the S411 is expressed as follows:

[0048]

[0049] in, represents a sequence of implicit coding vectors of the water level time series in the water level monitoring domain, 、 、 and Respectively represent the first, second, and third in the sequence of water level time series implicit coding vectors of the water level monitoring domain and The implicit coding vector of the water level time series in the water level monitoring domain, is the number of implicit coding vectors of the water level time series in the water level monitoring domain.

[0050] That is, the water level time series implicit coding aggregation matrix of the entire water level monitoring domain is split into a sequence of water level time series implicit coding vectors of the water level monitoring domain according to the position sample dimension, wherein each vector corresponds to a specific water level monitoring position.

[0051] More specifically, the step S412 is expressed as follows:

[0052]

[0053]

[0054]

[0055] in, represents the implicit feature vector of the temporal correlation of rainfall along the river, represents the sigmoid normalization function, represents the implicit eigenvector of the normalized rainfall time series correlation along the river, Indicates the Normalized water level monitoring domain water level time series implicit coding vector, represents the information entropy of the implicit feature vector associated with the normalized rainfall time series along the river, Indicates the The information entropy of the implicit coding vector of the water level time series in the normalized water level monitoring domain, represents the normalized rainfall time series associated implicit feature vector along the river and the The joint entropy between the implicit coding vectors of the water level time series in the normalized water level monitoring domain, Indicates the A fast semantic matching factor for the main and covariates of reservoir water levels.

[0056] That is, by calculating the mutual information between the implicit coding vectors of the water level time series in each water level monitoring domain and the implicit feature vectors associated with the rainfall time series along the river, the feature correlation between the two is evaluated, and a sequence of rapid semantic matching factors of the reservoir water level main and covariates is generated. It should be understood that the larger the mutual information value, the higher the correlation between the two; conversely, the smaller the mutual information value, the weaker the correlation between the two. In this way, not only can the water level change characteristics that are most relevant to rainfall changes be effectively identified, but also an objective standard can be provided for subsequent rapid matching in a quantitative manner.

[0057] More specifically, the step S413 is expressed as follows:

[0058]

[0059]

[0060] in, represents the maximum value function, and represent the first maximum value and the second maximum value respectively.

[0061] More specifically, the step S414 is expressed as follows:

[0062]

[0063]

[0064]

[0065] in, Indicates the index corresponding to the maximum value. and Respectively represent the position indexes of the first maximum value and the second maximum value in the sequence of the reservoir water level main-covariate fast semantic matching factor, and represent the implicit coding vectors of the water level time series in the water level monitoring domain corresponding to the first maximum value and the second maximum value, respectively. A subsequence representing the implicit coding vector of the water level time series in the water level monitoring domain.

[0066] That is, two maximum values ​​are queried from the sequence of rapid semantic matching factors of the reservoir water level main-covariate, and based on the positions of the two maximum values ​​in the sequence, a subsequence of the water level monitoring domain water level time series implicit coding vector that is significantly correlated with rainfall is extracted from the corresponding sequence of the water level monitoring domain water level time series implicit coding vector, to achieve preliminary feature selection and filtering, so as to concentrate resources to process the most relevant interactive information, reduce unnecessary calculations, and reduce noise interference in the water level change information.

[0067] Specifically, in a specific example of the present application, step S42 includes: performing a linear transformation on the implicit feature vector associated with the rainfall time series along the river to obtain a query vector and a value vector, and using a subsequence of the implicit coding vector of the water level time series in the water level monitoring domain as a subsequence of the key vector, inputting the query vector, the value vector, and the subsequence of the key vector into a cross-domain query coding module based on a converter structure to obtain the reservoir water level principal-covariate time series interaction response coding vector, which is expressed as follows:

[0068]

[0069]

[0070]

[0071] in, Indicates the The implicit coding vector of the water level time series in the water level monitoring domain, and denote the query embedding matrix and the value embedding matrix respectively, and Represent the query bias vector and value bias vector respectively, and Represent the query vector and value vector respectively, S represents the The length of the implicit coding vector of the water level time series in each water level monitoring domain, represents the transpose of a vector, represents the normalized exponential function, represents the matrix multiplication operation, Represents the reservoir water level main-covariate time series interaction response encoding vector.

[0072] That is, a query vector and a value vector are constructed based on the implicit feature vector associated with the rainfall time series along the river, and a subsequence of the water level time series implicit encoding vector of the selected water level monitoring domain is used as a subsequence of the key vector. Cross-domain query interaction is performed through the converter, and the self-attention mechanism is used to learn the associated dependency relationship between rainfall information and water level change information, and generate a reservoir water level main-covariate time series interaction response encoding vector, thereby achieving an in-depth understanding of the associated response relationship between upstream rainfall changes and reservoir water level changes.

[0073] In the above-mentioned reservoir flood control water level warning method, the step S5 generates the water level prediction result based on the reservoir water level main-covariate time series interaction response coding vector. In a specific example of the present application, the step S5 includes: inputting the reservoir water level main-covariate time series interaction response coding vector into a water level prediction engine based on the RNN model to obtain the water level prediction result. It should be known to those skilled in the art that the RNN model is a recurrent neural network model, which can effectively memorize and utilize the information of the previous moment to affect the output of the current moment through its internal loop structure. Here, the RNN model receives the reservoir water level main-covariate time series interaction response coding vector as input, and uses its internal loop structure to memorize and utilize the rainfall information and the associated impact information of rainfall on water level changes contained in the reservoir water level main-covariate time series interaction response coding vector, thereby accurately predicting the future change trend of the reservoir water level. In addition, since the RNN model can share parameters between each time step and maintain a continuous state update process, it can maintain coherence and consistency when predicting future values, making the output water level prediction results smoother, reducing abrupt fluctuations, and closer to the gradual changes in actual physical processes, thereby further improving the stability and reliability of the prediction.

[0074] Here, when the implicit feature vector of the rainfall time series association along the river and the implicit coding aggregation matrix of the water level time series in the entire water level monitoring domain are used to represent the rainfall time series association characteristics along the river and the water level time series association characteristics of each water level monitoring location respectively, when performing semantic query encoding based on the fast scanning mechanism, the differences in fast scanning semantic queryability caused by the differences in source time series distribution are taken into account, and it is expected to improve the semantic consistency aggregation expression effect of the reservoir water level main-covariate time series interaction response coding vector obtained by response semantic encoding.

[0075] In a preferred example of the present application, before inputting the reservoir water level principal-covariate time series interaction response encoding vector into a water level prediction engine based on an RNN model to obtain a water level prediction result, the reservoir water level principal-covariate time series interaction response encoding vector is subjected to semantic consistency aggregation optimization to obtain an optimized reservoir water level principal-covariate time series interaction response encoding vector, and the process includes the following steps:

[0076] The internal structure of the reservoir water level principal-covariate time series interaction response encoding vector is revealed based on the independent component analysis paradigm to obtain the principal component association activation matrix of the principal-covariate time series interaction response and the principal-covariate time series interaction response characteristic fluctuation representation matrix, which are expressed as:

[0077]

[0078]

[0079]

[0080] in, represents the reservoir water level main-covariate time series interaction response encoding vector, and are the first and second order coding vectors of the reservoir water level main-covariate time series interaction response. and The eigenvalues ​​at the positions, The principal component activation matrix representing the principal-covariate time series interaction response The value of the position, Represents the characteristic fluctuation matrix of the main-covariate time series interaction response The value of the position.

[0081] The principal component association activation matrix of the principal-covariate temporal interaction response and the characteristic fluctuation representation matrix of the principal-covariate temporal interaction response are respectively used as high-dimensional directional embedding coding matrices, and the reservoir water level principal-covariate temporal interaction response encoding vector is subjected to high-dimensional nonlinear embedding coding to obtain the principal-covariate temporal interaction response semantic association excitation encoding vector and the principal-covariate temporal interaction response semantic fluctuation excitation encoding vector, which are expressed as:

[0082]

[0083]

[0084] in, represents the principal component activation matrix of the principal-covariate time series interaction response, represents the characteristic fluctuation matrix of the main-covariate time series interaction response, represents matrix multiplication, represents the linear rectification function, represents the semantic association-stimulated encoding vector of the primary-covariate temporal interaction response, Represents the encoding vector that activates the temporal interaction between the principal and covariate in response to semantic fluctuations.

[0085] The main-covariate temporal interaction response semantic association excitation coding vector and the main-covariate temporal interaction response semantic fluctuation excitation coding vector are auto-correlatedly coded to obtain the main-covariate temporal interaction response semantic structure consistency potential coding matrix, which is expressed as:

[0086]

[0087] in, represents the length of the vector, represents the transpose of a vector, represents the normalized exponential function, The potential encoding matrix representing the semantic structure consistency of the principal-covariate temporal interaction response.

[0088] After feature fusion of the semantic association excitation coding vector of the main-covariate temporal interaction response and the semantic fluctuation excitation coding vector of the main-covariate temporal interaction response, the result of feature fusion is input into the main-covariate temporal interaction response semantic structure consistency potential coding matrix to perform semantic structure consistency explicit modeling to obtain the optimized reservoir water level main-covariate temporal interaction response coding vector, which is expressed as:

[0089]

[0090] in, Represents an activation function, such as the sigmoid function, Indicates feature fusion, such as cascade, Represents the weight matrix, obtained through training, Represents the optimized reservoir water level main-covariate time series interaction response encoding vector.

[0091] Accordingly, based on the analysis of the local regional activation focus of each eigenvalue of the reservoir water level main-covariate time series interaction response encoding vector using the independent component analysis paradigm, a high-dimensional nonlinear embedding network is used to correlate and analyze the potential spatial structural similarity between local structural features and eigenvectors. Thus, the sparsity of the representation vector is optimized with the help of autocorrelation to establish a mapping relationship between the correlation of structural elements and the complexity of the system hierarchy. This promotes the convergence of the reservoir water level main-covariate time series interaction response encoding vector toward the semantic consistency axis to enhance the semantic consistency aggregation expression effect of the reservoir water level main-covariate time series interaction response encoding vector. In this way, the accuracy of the water level prediction result obtained by inputting the reservoir water level main-covariate time series interaction response encoding vector into the water level prediction engine based on the RNN model is improved.

[0092] Next, to assess the risk level of reservoir flooding in the coming period based on water level forecasts and issue corresponding warnings, the entire process requires integration from multiple perspectives, including the establishment of a risk assessment indicator system, multi-source data analysis and integration, risk classification and warning level setting, and a warning information release mechanism. The following is a comprehensive and detailed description of this process:

[0093] First, in order to accurately assess the risk level of reservoirs encountering flood disasters, a scientific and reasonable risk assessment indicator system must be established. This system should include but is not limited to the following aspects: Water level thresholds: setting different levels of warning water levels (such as warning water levels, dangerous water levels, etc.). These thresholds are determined based on the experience summary of historical flood events and professional analysis of the reservoir's carrying capacity; Flow changes: paying attention to the changing trend of inflow flow, especially the sudden increase in flow in a short period of time, which is often a precursor to the arrival of a flood peak; Rainfall intensity and duration: The rainfall situation in the upstream basin directly determines the amount of water flowing into the reservoir. Heavy rainfall or long-term continuous rainfall will significantly increase the risk of flooding; Topographic and geomorphic characteristics: analyzing the topographic undulation of the upstream basin , slope, and soil type, all of which affect the rate of rainwater infiltration and the process of runoff formation; vegetation coverage: good vegetation can slow down the rate of surface runoff and reduce the possibility of sediment entering rivers; changes in climate patterns: considering the impact of global climate change, long-term climate trend forecasts are also instructive for short-term flood forecasts; impact of human activities: the degree of urbanization, agricultural irrigation needs, etc. may change the characteristics of local water cycles and increase or reduce the risk of floods; emergency response capabilities: assess the ability of local governments and relevant departments to respond to emergencies, such as the professional quality of rescue teams and the status of material reserves.

[0094] After obtaining the data for the above indicators, multi-source data analysis and fusion are carried out. Here, relatively simple and intuitive methods such as linear regression models and statistical methods can be used. Based on the results obtained from the multi-source data analysis, the risks faced by the reservoir are quantitatively scored according to pre-defined rules, and different risk levels are divided accordingly. Generally speaking, it can be divided into four levels: low risk, medium risk, high risk and extremely high risk. Each level corresponds to a different warning level (such as blue warning, yellow warning, orange warning and red warning), and is accompanied by specific response measures. The setting of the warning level should take into account the actual situation and allow dynamic adjustment to adapt to changing environmental conditions.

[0095] To ensure that warning information reaches its target audience quickly and effectively, it is necessary to build a multi-layered, comprehensive information dissemination network. This can be achieved through the following methods: SMS platforms, which send brief and clear warning messages to residents in the affected area via text messages; social media platforms, which leverage social channels such as Weibo and WeChat official accounts to expand information coverage and promote community interaction and communication; traditional media channels, which continue to leverage the role of traditional media such as television and radio to ensure that information reaches all people; community bulletin boards, which set up bulletin boards in community centers or public places to post the latest warning information; and emergency broadcast systems, which install wireless transmitters in specific areas that can continue to operate during power outages, ensuring that important notifications are delivered to all corners of the population.

[0096] Once the alert level is determined, it's crucial to prepare appropriate follow-up measures and contingency plans. This includes an evacuation plan with detailed information such as evacuation routes, assembly points, and transportation arrangements; stockpiling sufficient daily necessities and emergency supplies; organizing a medical team on standby, with ambulances, emergency equipment, and medicines; and planning post-disaster reconstruction plans, prioritizing ecological protection and restoration to prevent secondary disasters.

[0097] In summary, through the aforementioned series of measures, a closed-loop management system, from risk assessment to early warning issuance and emergency response, can be established to effectively prevent and mitigate losses caused by flood disasters. This process relies not only on advanced technology and scientific methods, but also on multi-party collaboration and efficient communication to ensure smooth execution of each link and to be fully prepared for potential natural disasters.

[0098] Then, based on the risk assessment results, a flood discharge scheduling plan is formulated and implemented. This process relies not only on accurate water level forecasts and risk assessment results, but also requires full consideration of multiple factors, including reservoir operating regulations, the safety of downstream residents, ecological and environmental protection, and the rational use of water resources. Throughout this process, it is crucial to ensure the timeliness and accuracy of information and maintain effective communication with all stakeholders.

[0099] First, after receiving the water level forecast based on rainfall data along the upstream river and a time-series aggregate matrix for the entire water level monitoring domain, the latest weather forecast, real-time water level trends, and other natural or human factors that may affect water levels in the coming days should be quickly analyzed. Based on this information, a decision should be made as to whether flood discharge operations are necessary and the initial target flood discharge volume should be set.

[0100] Next, to optimize flood discharge strategies, mathematical programming methods can be applied to find the optimal discharge path. For example, linear programming or dynamic programming techniques can be used to find solutions that maximize flood control effectiveness while ensuring downstream safety, while complying with legal regulations for reservoir operation (such as minimum ecological flow requirements). Furthermore, seasonal factors should be taken into account, such as maintaining high water levels during the dry season for subsequent use and lowering them appropriately before the rainy season to reserve capacity for possible heavy rains. Simulation software can also be used to test different flood discharge scenarios, evaluate their effectiveness, and select the most appropriate solution.

[0101] Once a specific flood discharge plan has been decided, preparations must begin immediately. This requires notifying relevant departments and individuals according to established procedures to ensure all necessary preparations are in place. If a risk assessment indicates a high likelihood of flooding, the pre-established emergency plan should be activated immediately. In accordance with the plan, personnel should be organized to inspect and maintain flood discharge facilities to ensure they are in good working order. Emergency rescue teams should also be on standby to ensure a rapid response in the event of an emergency.

[0102] Before officially commencing flood discharge, it should be reconfirmed that all conditions have been met, including but not limited to: the spillway gates have been fully overhauled and can be opened and closed normally; downstream communities have been fully informed of the impending flood discharge and have taken necessary precautions; and the communication system is unobstructed and can receive feedback from the site at any time. The spillway gates or other flood discharge facilities should then be gradually opened according to the pre-calculated flood discharge plan. During this process, special attention should be paid to monitoring water flow rate, water quality, and other conditions to prevent secondary pollution or other problems caused by the flood discharge. If any abnormal conditions are discovered during the flood discharge, such as the actual water level rising faster than expected, the flood discharge strategy will need to be flexibly adjusted to ensure the safety and controllability of the entire process.

[0103] To ensure transparency and public trust during the flood release process, it is essential to establish information disclosure channels to promptly inform residents of affected areas of the latest water conditions and preventative measures. Official announcements can be released through various media platforms, such as social media accounts, local television and radio stations, to keep the public informed of the latest developments. Furthermore, community meetings should be actively organized to gather feedback and enhance public understanding and cooperation with flood prevention efforts. Good social communication not only enhances public security but also helps reduce unnecessary panic and social instability.

[0104] Throughout the flood discharge process, it is also necessary to closely monitor environmental changes both inside and outside the reservoir, such as changes in rainfall intensity, river water levels, and reservoir water levels. If actual conditions deviate from expectations, a swift response is required, requiring flexible adjustments to flood discharge strategies to ensure the ultimate goal of protecting the lives and property of downstream residents and maintaining a stable natural environment.

[0105] Finally, a comprehensive review after the flood discharge is essential. Evaluating the effectiveness and existing problems of the flood discharge operation based on actual events and collected data, updating and iterating relevant models and algorithms, and improving prediction accuracy and response speed are also important means of continuously improving and perfecting the flood discharge scheduling system. In summary, formulating and implementing flood discharge scheduling plans is a comprehensive task involving multidisciplinary knowledge and technical application. Through scientific and rational planning, effective technical support, and good social coordination, we can minimize the damage caused by floods and waterlogging disasters, safeguard the safety of people's lives and property, and ensure the sustainable development of the natural environment.

[0106] In summary, the reservoir flood control water level warning method based on the embodiment of the present application is explained. It first obtains rainfall data along the upstream river in real time and monitors the water level changes at the reservoir entrance and key river sections. Then, artificial intelligence technology based on deep learning is used to analyze the collected rainfall data and water level data to capture the spatiotemporal correlation characteristics between rainfall data at each monitoring point along the river, as well as the spatiotemporal correlation characteristics between water level data at each monitoring point in the water level monitoring area. Rainfall data is used as the main variable and water level data as the covariate. By performing a time series interactive analysis on the two, the mechanism of the influence of rainfall changes on water level changes is revealed, thereby realizing the prediction of the reservoir water level change trend. In this way, the response speed and accuracy of reservoir flood control water level warnings can be effectively improved, the dependence on human resources can be reduced, and the maintenance cost of traditional hydrological models can be reduced.

[0107] Furthermore, a reservoir flood prevention water level early warning system is also provided.

[0108] Figure 6 FIG is a block diagram of a reservoir flood control water level warning system according to an embodiment of the present application. Figure 6 As shown, the reservoir flood control water level warning system 100 according to the embodiment of the present application includes: a data structured processing module 110, which is used to perform structured processing on rainfall data and water level data to obtain a rainfall time series aggregation matrix along the upstream river and a time series aggregation matrix of the entire water level monitoring domain; a main variable association coding module 120, which is used to use the rainfall time series aggregation matrix along the upstream river as the main variable, and perform association coding on the rainfall time series aggregation matrix along the upstream river to obtain a rainfall time series correlation implicit feature vector along the river; a covariate association coding module 130, which is used to use the rainfall time series aggregation matrix along the entire water level monitoring domain as the main variable, and perform association coding on the rainfall time series aggregation matrix along the upstream river to obtain a rainfall time series correlation implicit feature vector along the river; The time series aggregation matrix is ​​used as a covariate, and the time series aggregation matrix of the entire water level monitoring domain is associated and coded to obtain the time series implicit coding aggregation matrix of the water level in the entire water level monitoring domain; the dynamic interaction analysis module 140 is used to perform a main-covariate dynamic interaction based on a fast scanning mechanism on the implicit characteristic vector of the rainfall time series association along the river and the time series implicit coding aggregation matrix of the water level in the entire water level monitoring domain to obtain the main-covariate time series interaction response coding vector of the reservoir water level; the water level prediction generation module 150 is used to generate the water level prediction result based on the time series interaction response coding vector of the main-covariate of the reservoir water level.

[0109] Here, those skilled in the art will appreciate that the specific operations of each module in the above reservoir flood control water level warning system have been described in detail above. Figures 1 to 5 The invention has been described in detail in the description of the reservoir flood control water level early warning method, and therefore, its repeated description will be omitted.

[0110] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details of the above embodiments are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0111] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0112] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be encompassed therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0113] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0114] Finally, it should be noted that the above description has been provided for purposes of illustration and description. Furthermore, the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A reservoir flood control water level early warning method, comprising: Real-time rainfall data along the upstream river is obtained, and water level changes at the reservoir entrance and key river sections are monitored; based on the obtained rainfall data and water level data, the reservoir water level change trend in the future is predicted and analyzed to obtain water level prediction results; based on the water level prediction results, the risk level of the reservoir encountering flood disasters in the future is assessed and corresponding warnings are issued; based on the water level prediction results and risk assessment, a flood discharge scheduling plan is formulated and implemented. The characteristics of the system are that based on the obtained rainfall data and water level data, the reservoir water level change trend in the future is predicted and analyzed to obtain water level prediction results, including: Performing structured processing on the rainfall data and the water level data to obtain a rainfall time series aggregation matrix along the upstream river and a time series aggregation matrix of the entire water level monitoring domain; Taking the rainfall time series aggregation matrix along the upstream river as the main variable, performing association coding on the rainfall time series aggregation matrix along the upstream river to obtain the rainfall time series association implicit feature vector along the river; Taking the full water level monitoring domain time series aggregation matrix as a covariate, performing association coding on the full water level monitoring domain time series aggregation matrix to obtain the full water level monitoring domain water level time series implicit coding aggregation matrix; Performing a main-covariate dynamic interaction based on a fast scanning mechanism on the rainfall time series associated implicit feature vector along the river and the water level time series implicit coding aggregation matrix of the entire water level monitoring domain to obtain a reservoir water level main-covariate time series interaction response coding vector; Inputting the reservoir water level main-covariate time series interaction response encoding vector into the water level prediction engine based on the RNN model to obtain the water level prediction result; Before inputting the reservoir water level principal-covariate time series interaction response encoding vector into a water level prediction engine based on an RNN model to obtain a water level prediction result, the reservoir water level principal-covariate time series interaction response encoding vector is subjected to semantic consistency aggregation optimization to obtain an optimized reservoir water level principal-covariate time series interaction response encoding vector, the process of which includes the following steps: The internal structure of the reservoir water level principal-covariate time series interaction response encoding vector is revealed based on the independent component analysis paradigm to obtain the principal component correlation activation matrix of the principal-covariate time series interaction response and the principal-covariate time series interaction response characteristic fluctuation representation matrix; Using the principal component association activation matrix of the principal-covariate temporal interaction response and the characteristic fluctuation representation matrix of the principal-covariate temporal interaction response as high-dimensional directional embedding coding matrices, high-dimensional nonlinear embedding coding is performed on the principal-covariate temporal interaction response coding vector of the reservoir water level to obtain the principal-covariate temporal interaction response semantic association excitation coding vector and the principal-covariate temporal interaction response semantic fluctuation excitation coding vector; Performing autocorrelation coding on the semantic association excitation coding vector of the main-covariate temporal interaction response and the semantic fluctuation excitation coding vector of the main-covariate temporal interaction response to obtain a main-covariate temporal interaction response semantic structure consistency potential coding matrix; After feature fusion of the main-covariate temporal interaction response semantic association excitation coding vector and the main-covariate temporal interaction response semantic fluctuation excitation coding vector, the result of feature fusion is input into the main-covariate temporal interaction response semantic structure consistency potential coding matrix for semantic structure consistency explicit modeling to obtain the optimized reservoir water level main-covariate temporal interaction response coding vector.

2. The reservoir flood control water level early warning method according to claim 1, characterized in that: The rainfall data and the water level data are subjected to structured processing to obtain a rainfall time series aggregation matrix along the upstream river and a time series aggregation matrix of the entire water level monitoring domain, including: Arranging the rainfall data into a rainfall time series aggregation matrix along the upstream river according to the monitoring point sample dimension and the time dimension; The water level data are arranged into the full water level monitoring domain time series aggregation matrix according to the time dimension and the location sample dimension.

3. The reservoir flood control water level early warning method according to claim 2, characterized in that: Performing association coding on the rainfall time series aggregation matrix along the upstream river to obtain the rainfall time series association implicit feature vector along the river, including: The rainfall time series aggregation matrix along the upstream river is input into the rainfall time series correlation feature extractor along the river based on the converter structure to obtain the rainfall time series correlation implicit feature vector along the river.

4. The reservoir flood control water level early warning method according to claim 3, characterized in that: Performing association coding on the full water level monitoring domain time series aggregation matrix to obtain the full water level monitoring domain water level time series implicit coding aggregation matrix, including: The full water level monitoring domain time series aggregation matrix is ​​input into a sequence encoder based on the LSTM model to obtain the full water level monitoring domain water level time series implicit coding aggregation matrix.

5. The reservoir flood control water level early warning method according to claim 4, characterized in that: The main-covariate dynamic interaction based on the fast scanning mechanism is performed on the rainfall time series associated implicit feature vector along the river and the water level time series implicit coding aggregation matrix of the entire water level monitoring area to obtain the reservoir water level main-covariate time series interaction response coding vector, including: Based on the semantic correlation distribution between the implicit feature vector associated with the rainfall time series along the river and the water level time series implicit coding aggregation matrix of the entire water level monitoring domain, a subsequence of the water level time series implicit coding vector of the water level monitoring domain is filtered out from the water level time series implicit coding aggregation matrix of the entire water level monitoring domain; A cross-domain interactive fusion is performed on the subsequences of the rainfall time series associated implicit feature vector along the river and the water level time series implicit coding vector of the water level monitoring domain to obtain the reservoir water level main-covariate time series interactive response coding vector.

6. The reservoir flood control water level early warning method according to claim 5, characterized in that: Based on the semantic correlation distribution between the implicit feature vector associated with the rainfall time series along the river and the water level time series implicit coding aggregation matrix of the entire water level monitoring domain, a subsequence of the water level time series implicit coding vector of the water level monitoring domain is screened out from the water level time series implicit coding aggregation matrix of the entire water level monitoring domain, including: Splitting the water level time series implicit coding aggregation matrix of the entire water level monitoring domain according to the position sample dimension to obtain a sequence of water level time series implicit coding vectors of the water level monitoring domain; Calculating the mutual information between the implicit feature vector associated with the rainfall time series along the river and each water level monitoring domain water level time series implicit coding vector in the sequence of the water level monitoring domain water level time series implicit coding vectors to obtain a sequence of reservoir water level main-covariate fast semantic matching factors; identifying a first maximum value and a second maximum value from a sequence of the reservoir water level primary-covariate fast semantic matching factors; Based on the positions of the first maximum value and the second maximum value in the sequence of the reservoir water level main-covariate fast semantic matching factors, a subsequence of the water level monitoring domain water level time series implicit coding vector that is quickly matched is determined from the sequence of the water level monitoring domain water level time series implicit coding vector.

7. The reservoir flood control water level early warning method according to claim 6, characterized in that: Cross-domain interactive fusion is performed on the subsequences of the rainfall time series associated implicit feature vector along the river and the water level time series implicit coding vector of the water level monitoring domain to obtain the reservoir water level main-covariate time series interactive response coding vector, including: A linear transformation is performed on the implicit feature vector associated with the rainfall time series along the river to obtain a query vector and a value vector, and a subsequence of the implicit coding vector of the water level time series in the water level monitoring domain is used as a subsequence of the key vector. The query vector, the value vector and the subsequence of the key vector are input into a cross-domain query coding module based on a converter structure to obtain the reservoir water level principal-covariate time series interaction response coding vector.

8. A reservoir flood control water level early warning system, used to execute the method according to any one of claims 1 to 7, characterized in that: include: The data structured processing module is used to perform structured processing on rainfall data and water level data to obtain the rainfall time series aggregation matrix along the upstream river and the time series aggregation matrix of the entire water level monitoring domain; A main variable association coding module is used to use the rainfall time series aggregation matrix along the upstream river as the main variable, and perform association coding on the rainfall time series aggregation matrix along the upstream river to obtain an implicit feature vector of rainfall time series association along the river; a covariate association coding module, configured to use the full water level monitoring domain time series aggregation matrix as a covariate, and perform association coding on the full water level monitoring domain time series aggregation matrix to obtain a full water level monitoring domain water level time series implicit coding aggregation matrix; A dynamic interaction analysis module is used to perform a main-covariate dynamic interaction based on a fast scanning mechanism on the implicit characteristic vector of the rainfall time series along the river and the implicit coding aggregation matrix of the water level time series in the entire water level monitoring domain to obtain a main-covariate time series interaction response coding vector of the reservoir water level; The water level prediction generation module is used to generate the water level prediction result based on the reservoir water level main-covariate time series interaction response coding vector.

Images