A water monitoring system and method

Abstract

The application provides a water conservancy monitoring system and method, and relates to the technical field of water conservancy safety monitoring. The system comprises a decision center and at least one monitoring station. A first monitoring station in the at least one monitoring station identifies the water conservancy risk category of a first monitoring area based on first water conservancy data of the first monitoring area, and transmits the risk category and related second water conservancy data to the decision center. The decision center determines the affected second monitoring area by means of a target risk propagation model in combination with the second water conservancy data of the first monitoring area; the water conservancy risk category of the second monitoring area is determined by analyzing the data of the first and second monitoring areas, and then early warning information containing the risk category is sent. The application cooperates with the decision center through the monitoring station, combines a targeted risk model, avoids redundant operation, improves the water conservancy risk prediction efficiency, enhances the prediction accuracy, and provides effective support for water conservancy risk prevention and control.

Description

Technical Field

[0001] This invention relates to the field of water conservancy safety monitoring technology, and in particular to a water conservancy monitoring system and method. Background Technology

[0002] With the development of sensor networks and IoT technologies, multi-dimensional real-time monitoring of water resources data provides rich information support for the assessment of water resources risks (such as abnormal water levels and excessive structural stress). However, in complex watershed environments, ensuring the accuracy of water resources risk prediction results remains a key technical challenge for the industry.

[0003] Currently, the assessment of water conservancy risks generally adopts a full-data-driven modeling approach. This involves inputting water conservancy data collected from multiple sites into a unified water conservancy risk prediction simulation model without discrimination, and then combining historical data statistics with physical field coupling calculations to complete risk extrapolation. This type of approach attempts to ensure the reliability of prediction results through comprehensive data coverage, forming a standardized processing framework in traditional monitoring scenarios.

[0004] However, this method of performing a complete calculation process on all data results in a significant waste of computing resources on redundant calculations for risk-free areas, leading to low prediction efficiency and difficulty in meeting the rapid decision-making needs in emergency scenarios. Summary of the Invention

[0005] This invention provides a water conservancy monitoring system and method that can avoid the problem of low efficiency in predicting water conservancy risks caused by redundant calculation of all data.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] In a first aspect, the present invention provides a water conservancy monitoring system, comprising a decision-making center and at least one monitoring station. For a first monitoring station among the at least one monitoring station, the first monitoring station is used for:

[0008] Based on the first water resources data corresponding to the first monitoring area, and assuming that water resources risks exist in the first monitoring area, the type of water resources risk in the first monitoring area is determined. The first monitoring area is the monitoring coverage area of ​​the first monitoring station. The type of water resources risk in the first monitoring area and the corresponding second water resources data are sent to the decision-making center. The first water resources data corresponding to the first monitoring area includes the corresponding second water resources data. The second water resources data corresponding to the first monitoring area is determined based on the type of water resources risk in the first monitoring area.

[0009] The decision-making center is used for:

[0010] Based on the second water resources data corresponding to the first monitoring area and the target risk propagation model, a second monitoring area is determined. The second monitoring area is the monitoring coverage area of ​​the second monitoring station among at least one monitoring station. The target risk propagation model is determined based on the types of water resources risks in the first monitoring area. Based on the second water resources data corresponding to the first monitoring area and the first water resources data corresponding to the second monitoring area, if water resources risks are determined to exist in the second monitoring area, the types of water resources risks in the second monitoring area are determined. Early warning information, including the types of water resources risks in the second monitoring area, is sent to the second monitoring station.

[0011] Based on the above technical solution, the present invention can be further improved as follows.

[0012] Furthermore, the first monitoring station is used to determine at least one risk estimation model based on the first water conservancy data corresponding to the first monitoring area. The first water conservancy data corresponding to the first monitoring area includes multiple sub-data; different risk estimation models correspond to different types of water conservancy risks; for any one of the at least one risk estimation models, each risk estimation model is associated with at least one sub-data. Based on inputting the sub-data associated with any risk estimation model into that risk estimation model, the risk value of the water conservancy risk type corresponding to that risk estimation model in the first monitoring area is determined. If the risk value of the water conservancy risk type corresponding to that risk estimation model in the first monitoring area is greater than the risk threshold of that water conservancy risk type, it is determined that a water conservancy risk exists in the first monitoring area. The water conservancy risk type corresponding to any risk estimation model is then determined as the water conservancy risk type of the first monitoring area.

[0013] Furthermore, the sub-data includes water level data, flow rate data, flow velocity data, sediment concentration data, water quality data, hydrological time series data, meteorological data, topographic data, soil data, geological data, engineering stress data, and engineering strain data. Water conservancy risk types include flood risk, drought risk, riverbed siltation risk, water pollution risk, debris flow risk, soil displacement risk, dam instability risk, and secondary geological disaster risk. Specifically, water level data, flow rate data, and hydrological time series data are correlated with the risk estimation model corresponding to flood risk. Flow rate data and water level data are correlated with the risk estimation model corresponding to drought risk. Sediment concentration data and flow velocity data are correlated with the risk estimation model corresponding to riverbed siltation risk; water quality data is correlated with the risk estimation model corresponding to water pollution risk. Meteorological data and topographic data are correlated with the risk estimation model corresponding to debris flow risk. Soil data is correlated with the risk estimation model corresponding to soil displacement risk. Engineering stress data and engineering strain data are correlated with the risk estimation model corresponding to dam instability risk; geological data is correlated with the risk estimation model corresponding to secondary geological disaster risk.

[0014] Furthermore, the first monitoring station is used to determine the sub-data associated with any risk estimation model as the second water conservancy data corresponding to the first monitoring area.

[0015] Furthermore, the decision-making center is used to determine the scope of water conservancy risks in the first monitoring area based on the second water conservancy data corresponding to the first monitoring area and the target risk propagation model. Based on the scope of water conservancy risks in the first monitoring area, the second monitoring area is determined.

[0016] Furthermore, the decision-making center uses the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area to determine the risk value of each type of water conservancy risk occurring in the second monitoring area. If the risk value of any type of water conservancy risk in the second monitoring area is greater than the risk threshold for that type of water conservancy risk, then the water conservancy risk for that type of water conservancy risk in the second monitoring area is determined. This type of water conservancy risk is then identified as the water conservancy risk type for the second monitoring area.

[0017] Furthermore, any risk estimation model includes:

[0018] The first feature extraction layer includes a classification network, a convolutional neural network, and a graph neural network. The classification network categorizes the sub-data associated with any risk estimation model into temporal and spatial data, and inputs the temporal data into the convolutional neural network and the spatial data into the graph neural network. The convolutional neural network extracts features from the temporal data to obtain the corresponding temporal features. The graph neural network extracts features from the spatial data to obtain the spatial features.

[0019] The first fusion layer is used to fuse temporal and spatial features to obtain fused features.

[0020] The first risk assessment layer is used to map the fusion features to the corresponding risk values ​​of water conservancy risks.

[0021] The beneficial effects of adopting the above-mentioned further scheme are: by constructing a risk estimation model containing multiple network structures, and by extracting and fusing classification features from the data, it is possible to fully explore the temporal and spatial information in the data, more accurately assess the water conservancy risk value, improve the accuracy and reliability of risk estimation, and provide technical support for the accurate identification of water conservancy risks.

[0022] Furthermore, the target risk propagation model includes:

[0023] The recurrent neural network layer is used to simulate the process of the range of water conservancy risk in the first monitoring area changing over time based on the second data corresponding to the first monitoring area, and outputs the range of risk range in the first monitoring area at different time steps; the spatiotemporal aggregation layer is used to aggregate the range of risk range in the first monitoring area at different time steps, and output the range of water conservancy risk in the first monitoring area through weighted average and extreme value analysis.

[0024] Furthermore, the decision-making center uses the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area as inputs into the cross-regional joint risk assessment model to determine the risk value of each type of water conservancy risk occurring in the second monitoring area. The cross-regional joint risk assessment model includes:

[0025] The second feature extraction layer employs a dual-channel attention mechanism to extract features from the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, respectively. It outputs the temporal and spatial features of the second water conservancy data corresponding to the first monitoring area, and vice versa. The dual-channel attention mechanism includes a temporal attention submodule and a spatial attention submodule. The temporal attention submodule extracts and enhances the temporal features of the input data, while the spatial attention submodule extracts and enhances the spatial features of the input data. The risk association layer outputs a risk association matrix between the first and second monitoring areas based on the scope of water conservancy risks in the first monitoring area.

[0026] The second fusion layer is used to fuse the temporal and spatial features of the second water conservancy data corresponding to the first monitoring area, the temporal and spatial features of the first water conservancy data corresponding to the second monitoring area, and the risk correlation matrix between the first and second monitoring areas to obtain at least one fusion matrix; one fusion matrix corresponds to one type of water conservancy risk; the fully connected neural network layer is used to output the risk value of water conservancy risk corresponding to any type of water conservancy risk in the second monitoring area for any fusion matrix in the at least one fusion matrix.

[0027] The beneficial effects of this invention are:

[0028] This invention effectively overcomes the technical challenges of full-data modeling by constructing a water conservancy monitoring system that coordinates a decision-making center with at least one monitoring station. The first monitoring station, based on primary water conservancy data from a primary monitoring area, accurately identifies the water conservancy risks and their types within the area. This changes the traditional approach of indiscriminately processing all data, avoids redundant calculations in risk-free areas, and significantly reduces unnecessary computational power consumption.

[0029] Once the first monitoring station identifies a risk, only the relevant second-level water conservancy data is transmitted to the decision-making center. The decision-making center then uses the corresponding target risk propagation model to determine the second monitoring area based on the type of risk. This on-demand model retrieval and targeted analysis approach, compared to the traditional method of processing all data using a unified model, further focuses on key data and core risk areas, significantly improving computational efficiency.

[0030] When determining the types of water conservancy risks in the second monitoring area, the decision-making center conducts precise analysis based on the screened data from the first and second monitoring areas and sends early warning information to the second monitoring stations. The entire process abandons the traditional "one-size-fits-all" approach to basin-wide data, instead adopting a highly efficient model of "local preliminary assessment - precise transmission - targeted analysis." This ensures the accuracy of water conservancy risk prediction while significantly reducing computation time, effectively meeting the needs of rapid decision-making in emergency scenarios, improving the timeliness and accuracy of water conservancy risk prevention and control, and reducing disaster losses.

[0031] Secondly, this invention provides a water conservancy monitoring method. In this method, based on first water conservancy data corresponding to a first monitoring area, and upon determining that a water conservancy risk exists in the first monitoring area, the type of water conservancy risk in the first monitoring area is determined. Based on second water conservancy data corresponding to the first monitoring area and a target risk propagation model, a second monitoring area is determined. The first water conservancy data corresponding to the first monitoring area includes the second water conservancy data corresponding to the first monitoring area; the second water conservancy data corresponding to the first monitoring area is determined based on the type of water conservancy risk in the first monitoring area; the target risk propagation model is determined based on the type of water conservancy risk in the first monitoring area. Based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, upon determining that a water conservancy risk exists in the second monitoring area, the type of water conservancy risk in the second monitoring area is determined. An early warning message is issued, which includes the type of water conservancy risk in the second monitoring area.

[0032] Thirdly, the present invention provides an electronic device, comprising: a memory and one or more processors; the memory and the processors are coupled; wherein the memory stores computer program code, the computer program code including computer instructions, which, when executed by the processor, cause the electronic device to perform the method described in the second aspect above.

[0033] Fourthly, a computer-readable storage medium is provided, including computer instructions that, when executed on an electronic device, cause the electronic device to perform the method described in the second aspect above.

[0034] Fifthly, a computer program product is provided that, when run on a computer, causes the computer to perform the method described in the second aspect above.

[0035] It is understood that the beneficial effects achieved by the method of the second aspect, the electronic device of the third aspect, the computer-readable storage medium of the fourth aspect, and the computer program product of the fifth aspect can be referred to the beneficial effects of the first aspect and any of its possible design embodiments, which will not be repeated here. Attached Figure Description

[0036] Figure 1 This invention provides a schematic diagram of the structure of a water conservancy monitoring system;

[0037] Figure 2 This invention provides a signaling interaction diagram between various monitoring stations and the decision-making center.

[0038] Figure 3 A schematic diagram of the structure of a risk estimation model provided by the present invention;

[0039] Figure 4 A schematic diagram of the structure of a risk propagation model provided by the present invention;

[0040] Figure 5 This is a schematic diagram of the structure of a cross-regional joint risk assessment model provided by the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. In the description of this application, unless otherwise stated, " / " indicates that the objects before and after are in an "or" relationship. For example, A / B can represent A or B. "And / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Furthermore, in the description of this application, unless otherwise stated, "multiple" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple. Furthermore, to facilitate a clear description of the technical solutions in the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish identical or similar items with substantially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" are not necessarily different. Meanwhile, in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate that something is being used as an example, illustration, or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes.

[0042] With the rapid development of sensor networks and the Internet of Things (IoT) technologies, the field of water conservancy engineering has achieved real-time acquisition and transmission of multi-dimensional data such as water level, flow rate, stress, and strain. This massive amount of monitoring data provides a rich information foundation for water conservancy risk assessment. However, in complex and ever-changing watershed environments, traditional water conservancy monitoring solutions still face the dual challenges of insufficient accuracy in risk prediction and low computational efficiency.

[0043] Currently, a full-data-driven modeling approach is commonly used for water conservancy risk assessment. This method indiscriminately inputs all collected hydrological, environmental, and structural data into a unified water conservancy risk prediction simulation model, combining historical data statistics with physical field coupling calculations to extrapolate risks. Although this method constructs a standardized processing framework through comprehensive data coverage, it essentially performs a complete calculation process on all data, resulting in a significant amount of computing resources being consumed in redundant calculations in risk-free areas. For example, performing a full-process simulation of flood risk across the entire basin during dry seasons, or continuously performing complex calculations of dam stress in structurally healthy areas, not only wastes computing resources but also makes it difficult to meet the rapid decision-making needs of emergency scenarios such as sudden floods and dam failures in terms of risk prediction efficiency. Furthermore, the unified model cannot perform refined analysis of the characteristics of different types of water conservancy risks (such as floods, debris flows, and water pollution), further weakening the accuracy of risk prediction.

[0044] Therefore, how to ensure the accuracy of water conservancy risk prediction while achieving high efficiency in data processing and model calculation has become a key technical problem that urgently needs to be solved in the field of water conservancy monitoring.

[0045] To address the aforementioned problems, this invention provides a water conservancy monitoring system and method that enables targeted processing of multi-source water conservancy data and cross-regional risk projection, avoiding the inefficiency caused by redundant calculations of all data. Based on a dynamic matching algorithm model for risk types and combined with a risk estimation and propagation model constructed using digital twin technology, this invention can accurately identify risks and predict their impact range, improving the accuracy of water conservancy risk assessment in complex watershed environments. Furthermore, by screening key data and employing targeted analysis mechanisms, this invention significantly reduces computational power consumption, meeting the rapid decision-making needs in emergency scenarios and providing efficient technical support for water conservancy project safety and watershed disaster prevention.

[0046] This invention provides a water conservancy monitoring system, see [link to relevant documentation]. Figure 1 The water conservancy monitoring system provided by the present invention includes a decision-making center 101 and at least one monitoring station (e.g., monitoring station 102 and monitoring station 103).

[0047] For any one of the at least one monitoring station (also referred to as the first monitoring station in this embodiment), see Figure 2The first monitoring station can determine the type of water conservancy risk in the first monitoring area based on the first water conservancy data corresponding to the first monitoring area, provided that a water conservancy risk exists in the first monitoring area. Furthermore, the first monitoring station can send the type of water conservancy risk and the corresponding second water conservancy data for the first monitoring area to the decision-making center. Here, the first monitoring area is the monitoring coverage area of ​​the first monitoring station. The first water conservancy data corresponding to the first monitoring area includes the corresponding second water conservancy data. The second water conservancy data corresponding to the first monitoring area is determined based on the type of water conservancy risk in the first monitoring area.

[0048] The decision-making center can determine the second monitoring area based on the second water resources data corresponding to the first monitoring area and the target risk propagation model. Then, based on the second water resources data corresponding to the first monitoring area and the first water resources data corresponding to the second monitoring area, if a water resources risk is determined to exist in the second monitoring area, the decision-making center can determine the type of water resources risk in the second monitoring area and send early warning information to the second monitoring stations. The second monitoring area is the monitoring coverage area of ​​the second monitoring station among at least one monitoring station, and the target risk propagation model is determined based on the type of water resources risk in the first monitoring area. The early warning information includes the type of water resources risk in the second monitoring area.

[0049] In some embodiments, a first monitoring station is used to determine at least one risk estimation model based on first water resources data corresponding to a first monitoring area. The first water resources data corresponding to the first monitoring area includes multiple sub-data sets. Different risk estimation models correspond to different types of water resources risks. For any one of the at least one risk estimation models, each risk estimation model is associated with at least one sub-data set. Based on inputting the sub-data associated with each risk estimation model into that model, a risk value for the water resources risk type corresponding to that model in the first monitoring area is determined. If the risk value for the water resources risk type corresponding to that model in the first monitoring area is greater than the risk threshold for that model, it is determined that a water resources risk exists in the first monitoring area. The water resources risk type corresponding to that model is then determined as the water resources risk type for the first monitoring area.

[0050] As can be seen, this invention employs multiple risk estimation models and combines different sub-data to quantitatively assess water conservancy risks, making risk assessment more scientific and accurate. It can more accurately identify the specific types of water conservancy risks existing in the first monitoring area, providing a reliable basis for subsequent prevention and control measures.

[0051] In some embodiments, sub-data includes water level data, flow rate data, flow velocity data, sediment concentration data, water quality data, hydrological time series data, meteorological data, topographic data, soil data, geological data, engineering stress data, and engineering strain data. Types of water conservancy risks include flood risk, drought risk, riverbed siltation risk, water pollution risk, debris flow risk, soil displacement risk, dam instability risk, and secondary geological disaster risk.

[0052] Among them, water level data, flow data, hydrological time series data are linked to the risk estimation model corresponding to flood risk.

[0053] Specifically, the risk estimation model corresponding to flood risk can be expressed as:

[0054]

[0055] Among them, R flood H represents the risk value corresponding to the flood risk in the corresponding monitoring area; H represents the real-time water level in the corresponding monitoring area; H thr Q represents the warning water level threshold for the corresponding monitoring area; Q represents the real-time flow rate for the corresponding monitoring area; Q thr This indicates the safe flow threshold for the corresponding monitoring area (determined based on the river's water carrying capacity); This represents the rate of change of water level per unit time in the corresponding monitoring area. α, β, and γ are all weighting coefficients for the corresponding terms (those skilled in the art can set them according to actual scenarios and needs).

[0056] The flow data, water level data and the corresponding risk estimation model for drought risk are correlated.

[0057] Specifically, the risk estimation model corresponding to drought risk can be expressed as:

[0058]

[0059] Among them, R drought This represents the risk value corresponding to the drought risk in the monitored area; Q std H represents the historical average flow rate for the corresponding monitoring area during the same period; std This indicates the historical average water level for the corresponding monitoring area during the same period.

[0060] The data on sediment concentration and flow velocity are correlated with the risk estimation model corresponding to the risk of river siltation.

[0061] Specifically, the risk estimation model corresponding to the risk of river siltation can be expressed as:

[0062]

[0063] Among them, R siltS represents the risk value corresponding to the riverbed siltation risk in the corresponding monitoring area; S represents the real-time sediment concentration in the corresponding monitoring area; S std This represents the real-time sediment concentration threshold for the corresponding monitoring area; v represents the real-time flow velocity for the corresponding monitoring area; v crit This indicates the critical scouring velocity for the corresponding monitoring area.

[0064] The water quality data is correlated with the risk estimation model corresponding to the water pollution risk.

[0065] Specifically, the risk estimation model corresponding to water pollution risk can be expressed as:

[0066]

[0067] Among them, R pollution This represents the risk value corresponding to the water pollution risk in the monitored area; n represents the number of pollutant types; w i C represents the weighting coefficient of the i-th pollutant (based on pollutant toxicity settings); i C represents the real-time concentration of the i-th pollutant; i-thr This represents the safe concentration threshold for the i-th pollutant.

[0068] The meteorological data, topographic data and the risk estimation model corresponding to debris flow risk are correlated.

[0069] Specifically, the risk estimation model corresponding to debris flow risk can be expressed as:

[0070]

[0071] Among them, R landslide This represents the risk value corresponding to the debris flow risk in the monitored area; k represents the preset correction coefficient; p crit θ represents the critical rainfall threshold for the corresponding monitoring area; p represents the rainfall in the corresponding monitoring area per unit time; θ represents the topographic slope of the corresponding monitoring area.

[0072] The soil data is correlated with the risk estimation model corresponding to soil displacement risk.

[0073] Specifically, the risk estimation model corresponding to soil displacement risk can be expressed as:

[0074]

[0075] Among them, R soil This represents the risk value corresponding to the soil displacement risk in the corresponding monitoring area; w represents the real-time soil moisture content in the corresponding monitoring area; w sat represents the soil saturation moisture content of the corresponding monitoring area; c represents the soil cohesion of the corresponding monitoring area.

[0076] The engineering stress data, engineering strain data and the risk estimation model corresponding to the dam instability risk are correlated.

[0077] Specifically, the risk estimation model corresponding to the dam instability risk can be expressed as:

[0078]

[0079] Among them, R dam σ represents the risk value corresponding to the dam instability risk in the corresponding monitoring area; σ represents the real-time stress on the dam body in the corresponding monitoring area; ∈ represents the real-time strain on the dam body in the corresponding monitoring area; σ allow Indicates the allowable stress threshold of the dam material in the corresponding monitoring area; ∈ allow This indicates the allowable strain threshold of the dam material in the corresponding monitoring area.

[0080] The geological data is linked to the risk estimation model corresponding to the risk of secondary geological disasters.

[0081] Specifically, the risk estimation model corresponding to the risk of secondary geological disasters can be expressed as:

[0082]

[0083] Among them, R geohazard D represents the risk value corresponding to the geological secondary disaster risk in the corresponding monitoring area; D represents the geological disaster monitoring index value in the corresponding monitoring area; D max F represents the historical maximum geological hazard monitoring index value for the corresponding monitoring area; F represents the geological structural vulnerability score for the corresponding monitoring area; F thr This indicates the threshold for geological structural vulnerability in the corresponding monitoring area.

[0084] As can be seen, this invention clarifies the correspondence between various types of sub-data and different types of water conservancy risks, making the construction of risk estimation models more targeted, comprehensively covering multiple types of water conservancy risks, improving the water conservancy monitoring system's ability to identify various risks, and ensuring the safety of water conservancy projects and the surrounding environment.

[0085] In some embodiments, the first monitoring station is used to determine the sub-data associated with any risk estimation model as the second water conservancy data corresponding to the first monitoring area.

[0086] In other words, by selecting risk-related sub-data as the second type of water conservancy data for reporting, this invention reduces the amount of data transmitted and improves data transmission efficiency, while ensuring that the decision-making center obtains key data closely related to the risks, which facilitates subsequent analysis and processing.

[0087] In some embodiments, the decision center is used to determine the scope of water conservancy risk in the first monitoring area based on the second water conservancy data corresponding to the first monitoring area and the target risk propagation model. Based on the scope of water conservancy risk in the first monitoring area, the second monitoring area is determined. This invention utilizes the target risk propagation model to determine the scope of risk impact, thereby accurately locating the potentially affected second monitoring area. This makes the early warning scope more scientific and reasonable, avoids unnecessary early warning spread, improves the utilization efficiency of early warning resources, and ensures that potentially affected areas can receive timely early warnings.

[0088] In some embodiments, the decision center is used to determine the risk value of each type of water conservancy risk occurring in the second monitoring area based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area. The water conservancy risk of any type of water conservancy risk occurring in the second monitoring area is determined based on the fact that the risk value of any type of water conservancy risk in the second monitoring area is greater than the risk threshold of any type of water conservancy risk. This type of water conservancy risk is then identified as the water conservancy risk type of the second monitoring area.

[0089] In other words, by comprehensively analyzing data from two regions, this invention quantifies and assesses various water conservancy risks in the second monitoring region, thereby improving the accuracy and reliability of water conservancy risk identification in the second monitoring region. This allows for a more comprehensive understanding of the water conservancy risk situation in the region and provides strong support for formulating effective prevention and control strategies.

[0090] In some embodiments, see Figure 3 Any risk estimation model includes a first feature extraction layer, a first fusion layer, and a first risk assessment layer.

[0091] The first feature extraction layer further includes a classification network, a convolutional neural network, and a graph neural network. Considering the heterogeneous nature of multi-source data in water conservancy monitoring, the classification network first accurately classifies the sub-data associated with any risk estimation model, dividing parameters such as water level data and flow data that change over time into time-series data, while classifying data with spatial distribution characteristics such as topographic data and geological data into spatial data. Based on the differences in data types, the time-series data is directed to the convolutional neural network, and the spatial data is fed into the graph neural network.

[0092] Convolutional neural networks, with their local perception and weight sharing mechanisms, can efficiently extract features such as trends and periodic changes in time-series data. For example, they can identify the fluctuation trend of water level over time and the abnormal change points of flow, thereby obtaining the time-series features corresponding to the time-series data. Graph neural networks, on the other hand, construct topological relationship graphs based on spatial data. Through the information transmission between nodes and edges, they can deeply explore the spatial relationships between regions, such as the influence of terrain slope on water flow path and the transmission effect of geological structure on soil displacement risk, and thus output the spatial features of spatial data.

[0093] After processing by the first feature extraction layer, the first fusion layer uses weighted summation and channel concatenation to fuse temporal and spatial features, integrating the dynamic evolution of the temporal dimension with the structural correlation of the spatial dimension into a unified fused feature to comprehensively reflect the spatiotemporal coupling characteristics of water conservancy risks. Finally, the first risk assessment layer, based on the mapping relationship trained from historical risk data, inputs the fused features into a logistic regression or deep learning classifier, calculates and outputs the corresponding water conservancy risk value, achieving a complete transformation from raw data to quantitative risk assessment.

[0094] As can be seen, this invention, by constructing a risk estimation model incorporating multiple network structures, performs classification feature extraction and fusion on data, fully mining the temporal and spatial information within the data. Compared to traditional single-model processing methods, it avoids the problem of insufficient feature extraction caused by mixed data types. This model, by specifically processing data from different dimensions, effectively improves the analytical capabilities for complex water conservancy risks such as floods and debris flows, more accurately assesses water conservancy risk values, reduces misjudgments and omissions, and improves the accuracy and reliability of risk estimation. This provides technical support for the accurate identification of water conservancy risks, ensuring that decision-making centers can conduct timely and accurate cross-regional risk early warning and prevention deployments based on reliable risk assessment results.

[0095] In some embodiments, see Figure 4 The target risk propagation model includes a recurrent neural network layer and a spatiotemporal aggregation layer.

[0096] The recurrent neural network layer, targeting the second set of data corresponding to the first monitoring area—key sub-data closely related to different types of water conservancy risks (such as water level and flow data under flood risk, and meteorological and topographic data under debris flow risk)—establishes a time-dimensional information transmission mechanism using gated recurrent units (GRUs) or long short-term memory (LSTM) networks. This layer can capture the temporal dependencies in the risk evolution process, such as the rate of rise of flood levels at different times and the expansion trend of debris flow impact range with rainfall duration, thereby dynamically simulating the process of water conservancy risk range changes over time in the first monitoring area. By setting multiple time steps (such as in hours or days), the recurrent neural network layer outputs risk impact range data at different stages, providing time-seriesd risk evolution information for subsequent analysis.

[0097] After processing by the recurrent neural network layer, the spatiotemporal aggregation layer receives risk impact range data at each time step. Combining the spatial distribution characteristics of water conservancy risks, it employs a dual strategy of weighted averaging and extreme value analysis for data integration. During weighted averaging, weights are assigned to each data point based on the degree of risk impact at different time points (e.g., data from peak flood periods have higher weights than those from stable periods), while also considering the impact of spatial location on risk propagation (e.g., downstream areas have higher weights than upstream areas). Extreme value analysis focuses on extracting boundary information of the risk impact range (e.g., maximum inundation area, farthest impact distance) to avoid the loss of key risk information due to averaging. Through this aggregation process, the spatiotemporal aggregation layer ultimately outputs a precise water conservancy risk impact range for the first monitoring area, covering the entire time and spatial domain.

[0098] As can be seen, this invention utilizes a recurrent neural network layer and a spatiotemporal aggregation layer to construct a target risk propagation model. Compared to traditional static risk assessment methods, this model can dynamically simulate the changes in the risk impact range over time, effectively solving the problem of the separation between temporal and spatial characteristics in the risk propagation process. Through an aggregation strategy of weighted averaging and extreme value analysis, the overall accuracy of the risk range calculation is ensured while highlighting the impact of key risk nodes, making the determination of the risk impact range more consistent with reality. Based on the precise risk range output by this model, the decision-making center can quickly locate the affected secondary monitoring area and send early warning information to the corresponding stations in advance, gaining valuable response time for water conservancy risk prevention and control. This significantly enhances the foresight and effectiveness of water conservancy risk prevention and control, and improves the scientific rigor and efficiency of basin disaster emergency management.

[0099] In some embodiments, the decision center is used to determine the risk value of each type of water conservancy risk occurring in the second monitoring area by inputting the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area into a cross-regional joint risk assessment model.

[0100] Among them, see Figure 5 The cross-regional risk joint assessment model includes a second feature extraction layer, a risk association layer, a second fusion layer, and a fully connected neural network layer.

[0101] The second feature extraction layer utilizes a dual-channel attention mechanism to refine the data from the two regions separately. The temporal attention submodule employs a gated recurrent structure to dynamically assign weights to time-series data (such as water level fluctuations over time and rainfall temporal distribution), focusing on capturing key time nodes in the risk evolution process (such as the moment of flood peak and periods of continuous heavy rainfall). The spatial attention submodule, on the other hand, constructs regional topological relationships based on a graph convolutional network, enhancing feature extraction from high-risk spatial areas (such as river bends and geologically vulnerable zones) through information transfer between nodes. After this processing, this layer outputs the temporal and spatial features of the data from the two regions, achieving in-depth mining of spatiotemporal information within the data.

[0102] The risk association layer constructs a risk association matrix based on the risk impact range of the first monitoring area output by the target risk propagation model, combined with the spatial relationship between the two monitoring areas and the connectivity of water conservancy facilities. Each element in the matrix represents the propagation weight of different risk types between the two areas. For example, the transmission coefficient of flood risk can be determined by analyzing the upstream and downstream relationship of the river channel, or the probability of the impact of secondary geological disasters can be assessed based on the distribution of geological faults, thereby quantifying the correlation strength between risks between regions.

[0103] The second fusion layer, acting as the information integration hub, performs three-dimensional tensor fusion of the features extracted by the dual-channel attention mechanism with the risk correlation matrix. For each type of water conservancy risk, this layer uses channel splicing and weighted summation to integrate the risk characteristics of the first monitoring area, the local characteristics of the second monitoring area, and the correlation between the two areas into a single fusion matrix, ensuring that each matrix contains complete information on cross-regional risk propagation.

[0104] Finally, the fully connected neural network layer receives the fusion matrix and performs nonlinear mapping through a multilayer perceptron structure, transforming the matrix features into specific risk values. The network is trained and optimized using historical cross-regional risk event data, enabling it to accurately output the probability values ​​of various water conservancy risk types occurring in the second monitoring area, providing a quantitative basis for risk level determination.

[0105] As can be seen, this invention overcomes the limitations of traditional independent regional assessments by constructing a cross-regional joint risk assessment model. The dual-channel attention mechanism effectively separates and enhances the spatiotemporal characteristics of the data, avoiding feature ambiguity caused by information mixing; the introduction of the risk correlation matrix establishes a quantitative link for risk propagation between regions, enabling the model to capture the transmission patterns of cross-regional risks. Through multi-layer fusion and intelligent computing, this model analyzes the water conservancy risk relationship between two regions more comprehensively and deeply. Compared with single-region assessments or simple data overlay methods, it significantly improves the accuracy and completeness of risk assessment, accurately identifying the risk values ​​of various water conservancy risks in the second monitoring region. This provides scientific and reliable decision support for the collaborative prevention and control of cross-regional water conservancy risks, helping to achieve global and refined management of watershed risks.

[0106] In some embodiments, the warning information may also include risk level information, probability of risk occurrence, and scope of risk impact.

[0107] Among them, the risk level information is based on the risk value output by the risk estimation model. It is divided according to established standards (such as general, relatively large, major, and particularly major) and presented with color coding (blue, yellow, orange, and red) or text description. Its source is the quantitative mapping of the integrated characteristics by the first risk assessment layer. After standardization processing by the decision center, it forms an intuitive risk level identifier.

[0108] The probability of risk occurrence is a specific value output by the fully connected neural network layer after performing a nonlinear mapping on the fusion matrix. For example, "the probability of drought occurring within the next 24 hours is 65%". This data comes from the probabilistic calculation of the risk characteristics of the second monitoring area by the fully connected neural network layer in the cross-regional risk joint assessment model.

[0109] The scope of risk impact is determined by simulating the evolution of risk over time through the recurrent neural network layer of the target risk propagation model, and then by weighted averaging and extreme value analysis through the spatiotemporal aggregation layer. It is presented in the form of coordinates, map labels or text descriptions, such as "This debris flow risk may affect the area within 3 kilometers northeast of the XX mountain range". Its essence is the spatial expansion of the risk impact range of the first monitoring area and the precise positioning of the affected area of ​​the second monitoring area.

[0110] The present invention also provides a water conservancy monitoring method, which can be applied to the water conservancy monitoring system provided by the present invention to perform some or all of the steps that the water conservancy monitoring system provided by the present invention can achieve.

[0111] Specifically, in the water conservancy monitoring method provided by this invention, based on the first water conservancy data corresponding to the first monitoring area, if it is determined that there is a water conservancy risk in the first monitoring area, the type of water conservancy risk in the first monitoring area is determined. Based on the second water conservancy data corresponding to the first monitoring area and the target risk propagation model, a second monitoring area is determined. The first water conservancy data corresponding to the first monitoring area includes the second water conservancy data corresponding to the first monitoring area; the second water conservancy data corresponding to the first monitoring area is determined based on the type of water conservancy risk in the first monitoring area; the target risk propagation model is determined based on the type of water conservancy risk in the first monitoring area. Based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, if it is determined that there is a water conservancy risk in the second monitoring area, the type of water conservancy risk in the second monitoring area is determined. An early warning message is issued, which includes the type of water conservancy risk in the second monitoring area.

[0112] In some solutions, multiple embodiments of this application can be combined, and the combined solution can be implemented. Optionally, some operations in the processes of each method embodiment may be combined, and / or the order of some operations may be changed. Furthermore, the execution order between the steps of each process is merely exemplary and does not constitute a limitation on the execution order between steps; other execution orders are also possible. It is not intended to indicate that the execution order is the only possible order in which these operations can be performed. Those skilled in the art will conceive of various ways to reorder the operations described herein. In addition, it should be noted that the process details involved in one embodiment of this document are similarly applicable to other embodiments, or different embodiments may be combined.

[0113] Furthermore, some steps in the method embodiments can be equivalently replaced with other possible steps. Alternatively, some steps in the method embodiments may be optional and can be deleted in certain use cases. Or, other possible steps may be added to the method embodiments. Moreover, the various method embodiments can be implemented individually or in combination.

[0114] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above.

[0115] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0116] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0117] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, in essence, or the part that contributes, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A water conservancy monitoring system, characterized in that, Includes a decision-making center and at least one monitoring site; For the first monitoring station among the at least one monitoring station, the first monitoring station is used for: Based on the first water conservancy data corresponding to the first monitoring area, if it is determined that there is water conservancy risk in the first monitoring area, the type of water conservancy risk in the first monitoring area is determined. The first monitoring area is the monitoring coverage area of ​​the first monitoring station; The water conservancy risk types of the first monitoring area and the corresponding second water conservancy data of the first monitoring area are sent to the decision-making center; the first water conservancy data corresponding to the first monitoring area includes the corresponding second water conservancy data of the first monitoring area; the second water conservancy data corresponding to the first monitoring area is determined based on the water conservancy risk types of the first monitoring area. The decision-making center is used for: Based on the second water conservancy data corresponding to the first monitoring area and the target risk propagation model, the scope of water conservancy risk in the first monitoring area is determined; the target risk propagation model is determined based on the types of water conservancy risks in the first monitoring area. Based on the scope of the water conservancy risks in the first monitoring area, a second monitoring area is determined; The second monitoring area is the monitoring coverage area of ​​the second monitoring station among the at least one monitoring stations; Based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, if it is determined that there is water conservancy risk in the second monitoring area, the type of water conservancy risk in the second monitoring area is determined. Send early warning information to the second monitoring station; The early warning information includes the types of water conservancy risks in the second monitoring area; The target risk propagation model includes: A recurrent neural network layer is used to simulate the process of the range of water conservancy risk in the first monitoring area changing over time based on the second water conservancy data corresponding to the first monitoring area, and output the risk range data of the first monitoring area at different time steps. The spatiotemporal aggregation layer is used to aggregate the risk impact range data of the first monitoring area at different time steps, and output the impact range of water conservancy risks in the first monitoring area through weighted average and extreme value analysis.

2. The system according to claim 1, characterized in that, In the context of determining the type of water conservancy risk in the first monitoring station based on the first water conservancy data corresponding to the first monitoring area, the first monitoring station is specifically used for: Based on the first water conservancy data corresponding to the first monitoring area, at least one risk estimation model is determined; the first water conservancy data corresponding to the first monitoring area includes multiple sub-data; different risk estimation models correspond to different types of water conservancy risks. For any one of the at least one risk estimation models, each risk estimation model is associated with at least one sub-data; Based on inputting the sub-data associated with any of the risk estimation models into any of the risk estimation models, the risk value of the water conservancy risk type corresponding to any of the risk estimation models in the first monitoring area is determined; Based on the fact that the risk value of any water conservancy risk type corresponding to any risk estimation model in the first monitoring area is greater than the risk threshold of any water conservancy risk type corresponding to any risk estimation model, it is determined that there is water conservancy risk in the first monitoring area. The water conservancy risk type corresponding to any of the risk estimation models is determined as the water conservancy risk type of the first monitoring area.

3. The system according to claim 2, characterized in that, The sub-data includes water level data, flow rate data, flow velocity data, sediment concentration data, water quality data, hydrological time series data, meteorological data, topographic data, soil data, geological data, engineering stress data, and engineering strain data. The types of water conservancy risks include flood risk, drought risk, river siltation risk, water pollution risk, debris flow risk, soil displacement risk, dam instability risk, and secondary geological disaster risk. Specifically, the water level data, the flow rate data, and the hydrological time series data are associated with the risk estimation model corresponding to the flood risk; the flow rate data and the water level data are associated with the risk estimation model corresponding to the drought risk; the sediment content data and the flow velocity data are associated with the risk estimation model corresponding to the riverbed siltation risk; the water quality data are associated with the risk estimation model corresponding to the water pollution risk; the meteorological data and the topographic data are associated with the risk estimation model corresponding to the debris flow risk; the soil data are associated with the risk estimation model corresponding to the soil displacement risk; the engineering stress data and the engineering strain data are associated with the risk estimation model corresponding to the dam instability risk; and the geological data are associated with the risk estimation model corresponding to the geological secondary disaster risk.

4. The system according to claim 3, characterized in that, After the first monitoring station determines that there is a water conservancy risk in the first monitoring area, the first monitoring station is also used for: The sub-data associated with any of the risk estimation models is determined as the second water conservancy data corresponding to the first monitoring area.

5. The system according to claim 4, characterized in that, In the decision-making center's function of determining the type of water conservancy risk in the second monitoring area based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, when it is determined that there is a water conservancy risk in the second monitoring area, the decision-making center is specifically used for: Based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, the risk value of water conservancy risk for each of the aforementioned water conservancy risk types is determined in the second monitoring area. Based on the fact that the risk value of any type of water conservancy risk in the second monitoring area is greater than the risk threshold of the water conservancy risk of any type of water conservancy risk, the water conservancy risk of any type of water conservancy risk in the second monitoring area is determined. Any of the aforementioned water conservancy risk types is identified as the water conservancy risk types of the second monitoring area.

6. The system according to claim 5, characterized in that, The risk estimation model includes: The first feature extraction layer includes a classification network, a convolutional neural network, and a graph neural network. The classification network is used to classify the sub-data associated with any risk estimation model into temporal data and spatial data, and to input the temporal data into the convolutional neural network and the spatial data into the graph neural network. The convolutional neural network is used to extract features from the temporal data to obtain the temporal features corresponding to the temporal data. The graph neural network is used to extract features from the spatial data to obtain the spatial features of the spatial data. The first fusion layer is used to fuse the temporal features and the spatial features to obtain fused features; The first risk assessment layer is used to map the fused features to the corresponding risk values ​​of water conservancy risks.

7. The system according to claim 6, characterized in that, In the decision-making center's function of determining the risk value of each type of water conservancy risk occurring in the second monitoring area based on the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, the decision-making center is specifically used for: Based on inputting the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area into a cross-regional joint risk assessment model, the risk value of each type of water conservancy risk occurring in the second monitoring area is determined; wherein, the cross-regional joint risk assessment model includes: The second feature extraction layer employs a dual-channel attention mechanism to extract features from the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area, respectively. It outputs the temporal and spatial features of the second water conservancy data corresponding to the first monitoring area, and the temporal and spatial features of the first water conservancy data corresponding to the second monitoring area. The dual-channel attention mechanism includes a temporal attention submodule and a spatial attention submodule. The temporal attention submodule is used to extract and enhance the temporal features of the input data, and the spatial attention submodule is used to extract and enhance the spatial features of the input data. The risk correlation layer is used to output a risk correlation matrix between the first monitoring area and the second monitoring area based on the scope of the impact of water conservancy risks in the first monitoring area. The second fusion layer is used to fuse the temporal and spatial characteristics of the second water conservancy data corresponding to the first monitoring area, the temporal and spatial characteristics of the first water conservancy data corresponding to the second monitoring area, and the risk correlation matrix between the first monitoring area and the second monitoring area to obtain at least one fusion matrix; one fusion matrix corresponds to one type of water conservancy risk. A fully connected neural network layer is used to output, based on any one of the at least one fusion matrix, the risk value of the water conservancy risk corresponding to any one of the fusion matrices occurring in the second monitoring area.

8. A water conservancy monitoring method, characterized in that, The method, applied to the hydraulic monitoring system according to any one of claims 1-7, comprises: The first monitoring station is controlled to determine the type of water conservancy risk in the first monitoring area based on the first water conservancy data corresponding to the first monitoring area; the first monitoring area is the monitoring coverage area of ​​the first monitoring station. The first monitoring station is controlled to send the water conservancy risk type of the first monitoring area and the second water conservancy data corresponding to the first monitoring area to the decision-making center; the first water conservancy data corresponding to the first monitoring area includes the second water conservancy data corresponding to the first monitoring area; the second water conservancy data corresponding to the first monitoring area is determined based on the water conservancy risk type of the first monitoring area. The decision center controls the determination of the scope of water conservancy risks in the first monitoring area based on the second water conservancy data corresponding to the first monitoring area and the target risk propagation model; the target risk propagation model is determined based on the types of water conservancy risks in the first monitoring area. The decision-making center controls the determination of a second monitoring area based on the impact range of water conservancy risks in the first monitoring area; the second monitoring area is the monitoring coverage area of ​​the second monitoring station among the at least one monitoring station. The decision center controls the second water conservancy data corresponding to the first monitoring area and the first water conservancy data corresponding to the second monitoring area to determine the type of water conservancy risk in the second monitoring area if it is determined that there is a water conservancy risk in the second monitoring area. The decision-making center controls the sending of early warning information to the second monitoring station; the early warning information includes the types of water conservancy risks in the second monitoring area. The target risk propagation model includes: A recurrent neural network layer is used to simulate the process of the range of water conservancy risk in the first monitoring area changing over time based on the second water conservancy data corresponding to the first monitoring area, and output the risk range data of the first monitoring area at different time steps. The spatiotemporal aggregation layer is used to aggregate the risk impact range data of the first monitoring area at different time steps, and output the impact range of water conservancy risks in the first monitoring area through weighted average and extreme value analysis.

Images