Irrigation area flood risk prediction system and method

Abstract

The invention provides an irrigation area flood risk prediction system and method, and relates to the technical field of natural disaster early warning, and the system comprises a data acquisition module, a data preprocessing module, a risk prediction module, a decision support module, an interactive visualization module and corresponding functions. According to the invention, high-precision prediction of the irrigation area flood risk is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of natural disaster early warning, and more particularly to a system and method for predicting flood risk in an irrigation area. Background Art

[0002] Currently, flood risk monitoring and prediction for irrigation channel networks is a core component of water conservancy project safety management. Risk evolution is influenced by multiple factors, including water level, flow velocity, meteorological conditions, equipment status, and spatial topology. Traditional prediction methods face significant technical bottlenecks when addressing complex spatiotemporal dynamic risks:

[0003] 1. Spatiotemporal feature modeling is fragmented, and dynamic correlation analysis is insufficient. Existing models often independently process time series data (such as LSTM) or spatial data (such as GNN), lacking the ability to jointly model upstream and downstream channel topological relationships and spatiotemporal coupling characteristics. For example, traditional methods do not incorporate the Transformer self-attention mechanism to capture time-step dependencies, nor do they use Euclidean distance to calculate node spatial weights, resulting in an inability to accurately simulate the spatiotemporal transmission path of risks.

[0004] 2. Lack of dynamic weight adjustment mechanism for multiple factors. Traditional models do not establish a dynamic adjustment system for risk factor weights and are unable to adaptively optimize the contribution of each factor based on real-time business scenarios (such as meteorological conditions and equipment status changes). For example, the lack of a differentiated weight allocation mechanism for different risk factors (such as hydrological anomalies and equipment failures) makes it difficult to reflect the nonlinear changes in risk levels under the coupling of multiple factors, resulting in limited model generalization capabilities.

[0005] 3. Insufficient risk attribution and interpretability, and lack of support for multi-dimensional predictions. Existing technologies often only output a single risk level, lacking the ability to trace risk causes and conduct multi-dimensional predictions. For example, the lack of a coordinated prediction of risk level and risk-causing factors prevents management decisions from providing a complete explanation chain of "risk level-influencing factors-evolution path," resulting in insufficient business guidance for risk warnings.

[0006] Therefore, how to provide an irrigation area flood risk prediction system that can solve the above problems is an issue that those skilled in the art urgently need to solve. Summary of the Invention

[0007] In view of this, the present invention provides a system and method for predicting flood risks in irrigation areas, and provides a risk evolution prediction method that integrates spatiotemporal attention mechanism, dynamic weight adjustment and multi-dimensional prediction, and constructs an interactive visualization platform integrated with the geographic information system (GIS) to achieve high-precision prediction, explainable analysis and real-time monitoring of irrigation area flood risks.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] An irrigation area flood risk prediction system, comprising:

[0010] a data acquisition module for acquiring multi-source risk data and hydrological equipment status data of the irrigation area, wherein the multi-source risk data includes meteorological environment data, and the hydrological equipment status data includes spatial topology data, and constructing a topological relationship model of upstream and downstream equipment nodes of the irrigation area based on the multi-source risk data and the hydrological equipment status data;

[0011] a data preprocessing module, configured to preprocess the multi-source risk data and the hydrological equipment status data respectively, and extract corresponding time feature vectors from the preprocessed multi-source risk data, and extract discrete coding features from the preprocessed hydrological equipment status data;

[0012] A risk prediction module is used to construct a risk prediction model and input the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result;

[0013] The decision support module is used to analyze the risk prediction results obtained by the risk prediction module to achieve auxiliary decision-making.

[0014] Preferably, it also includes:

[0015] An interactive visualization module is connected to the data acquisition module and the risk prediction module, and is used to display the risk prediction results and changes in the data acquired by the data acquisition module.

[0016] Preferably, the specific implementation process of the data preprocessing module includes:

[0017] Numerically encoding the discrete data included in the hydrological equipment status data, extracting corresponding discrete coding features and then normalizing them;

[0018] Performing sliding window feature extraction on the rainfall data in the meteorological environment data to obtain a corresponding time feature vector;

[0019] Gaussian noise is added to key feature data in the meteorological environment data and oversampling is performed to achieve category balance, wherein the key feature data includes rainfall data and water level data.

[0020] Preferably, the risk prediction module specifically includes:

[0021] A risk prediction model construction unit, configured to construct a risk prediction model, wherein the risk prediction model includes a temporal attention model, a spatial attention model, and a multi-task learning spatiotemporal fusion model;

[0022] A risk prediction unit is used to input the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result.

[0023] Preferably, the specific implementation process of the risk prediction model construction unit includes:

[0024] The temporal attention model is implemented based on the Transformer self-attention mechanism model and is used to obtain the dependencies between the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding features.

[0025] The spatial attention model uses the topological relationship included in the upstream and downstream device node topological relationship model to generate node connection weights, with neighboring nodes having higher weights. It also calculates spatial similarity through the Euclidean distance of longitude and latitude, constructs a learnable projection matrix, maps spatial distance to attention weights, and integrates temporal and spatial attention scores to form an enhanced spatiotemporal attention model, thereby obtaining the spatiotemporal transmission characteristics of risks.

[0026] The multi-task learning spatiotemporal fusion model includes a multi-task learning framework with multi-branch outputs, which combines spatiotemporal conduction characteristics for synchronous prediction to obtain corresponding risk prediction results, wherein the risk prediction results include risk levels and risk causes.

[0027] Preferably, the specific implementation process of the spatial attention model includes:

[0028] The node connection weights in the spatial attention model are implemented using a dynamic adjustment mechanism and are dynamically adjusted according to the rainfall data included in the multi-source risk data.

[0029] Preferably, the specific implementation process of the multi-task learning spatiotemporal fusion model includes:

[0030] The multi-task learning spatiotemporal fusion model also includes a dynamic weight adjustment module, which is used to dynamically adjust the contribution weight of each risk factor according to the rainfall data and risk causes upstream of the irrigation area to achieve adaptive feature weight allocation.

[0031] Preferably, the specific implementation process of the interactive visualization module includes:

[0032] The interactive visualization module is based on a GIS interactive platform and visualizes risk levels using color and size.

[0033] Preferably, the specific implementation process of the decision support module includes:

[0034] The decision support module pushes the governance plan corresponding to the risk cause in real time through the expert knowledge graph, where the expert knowledge graph includes: risk level node, risk cause node and governance measure node, and establishes a mapping relationship between risk level-risk cause-governance plan through rule boundaries, and generates corresponding governance plans according to risk level and risk cause.

[0035] The present invention also provides a method for predicting flood risk in an irrigation area, comprising the following steps:

[0036] Acquiring multi-source risk data and hydrological equipment status data of the irrigation district, wherein the multi-source risk data includes meteorological environment data, and the hydrological equipment status data includes spatial topology data, and constructing a topological relationship model of upstream and downstream equipment nodes of the irrigation district based on the multi-source risk data and the hydrological equipment status data;

[0037] Preprocessing the multi-source risk data and the hydrological equipment status data respectively, extracting corresponding time feature vectors from the preprocessed multi-source risk data, and extracting discrete coding features from the preprocessed hydrological equipment status data;

[0038] Constructing a risk prediction model, and inputting the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result;

[0039] The risk prediction results obtained according to the risk prediction model are analyzed to achieve auxiliary decision-making.

[0040] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a system and method for predicting flood risk in an irrigation area, which has the following beneficial effects:

[0041] 1. By integrating temporal attention (Transformer mechanism) and spatial attention (channel topology and spatiotemporal distance modeling), the collaborative representation of risk factors in time series evolution and spatial node transmission is achieved. The accuracy of risk level prediction reaches over 90%, and the accuracy of high-risk scenario identification exceeds 92%, which is significantly better than the traditional single model's ability to characterize spatiotemporal coupled risks.

[0042] 2. Dynamically adjust risk factor weights based on real-time business scenarios such as upstream rainstorm conditions and gate failures, and accurately simulate the risk evolution path under the coupling of multiple factors through hydraulic physics causal chain modeling such as "sedimentation degree → roughness coefficient → flow velocity → water level", thereby improving the robustness of the model under complex working conditions.

[0043] 3. Introducing a multi-task learning framework, the risk level and risk cause are output simultaneously, and a traceable chain of "risk result-risk factor-transmission path" is established. The accuracy of risk cause prediction is greatly improved, providing a clear risk cause positioning for management decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0045] Figure 1 This is a structural principle block diagram of an irrigation area flood risk prediction system provided by the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] See also Figure 1 As shown, an embodiment of the present invention discloses an irrigation area flood risk prediction system, comprising:

[0048] The data acquisition module is used to obtain multi-source risk data and hydrological equipment status data of the irrigation area, and construct the topological relationship model of the upstream and downstream equipment nodes of the irrigation area based on the multi-source risk data and hydrological equipment status data. The multi-source risk data includes: hydrological monitoring data (such as water level, flow velocity, flow), meteorological environment data (such as rainfall data and temperature upstream of the irrigation area), risk assessment data (such as historical risk level, historical risk reasons), etc. The hydrological equipment status data includes equipment status data (such as gate status, siltation degree), spatial topological data (such as channel segment number, latitude and longitude coordinates, upstream and downstream relationship of the irrigation area), etc. In the process of constructing the topological relationship model of the upstream and downstream equipment nodes, the timestamp can be parsed into datetime type through pandas to construct the upstream and downstream topological relationship diagram of the channel, and realize the mapping of node ID with channel segment and latitude and longitude;

[0049] A data preprocessing module is used to preprocess the multi-source risk data and the hydrological equipment status data respectively, and extract the corresponding time feature vector from the preprocessed multi-source risk data, and extract the discrete coding feature from the preprocessed hydrological equipment status data;

[0050] The risk prediction module is used to build a risk prediction model and input the pre-processed multi-source risk data, hydrological equipment status data, time feature vectors, and discrete coding features into the risk prediction model for processing to obtain the corresponding risk prediction results;

[0051] The decision support module is used to analyze the risk prediction results obtained by the risk prediction module to assist in decision making.

[0052] Specifically, the implementation process of the data preprocessing module includes:

[0053] (1) The embodiment of the present invention numerically encodes the discrete data included in the hydrological equipment status data (e.g., gate status: open = 1.0, half-open = 0.5, closed = 0.0; siltation degree: low = 1, medium = 2, high = 3; risk reason: no abnormality = 0, excessive rainfall = 1, siltation = 2, gate failure = 3, water level too high = 4), extracts the corresponding time features, and then normalizes them;

[0054] (2) Calculate the sliding window mean of the rainfall in the last three steps by grouping the nodes in the irrigation area to capture the rainfall trend. The specific calculation process is: first, group the original data by node ID and sort by timestamp within each node to ensure the continuity of the time series; then use the rolling function of pandas to calculate the sliding window mean, with a window size of 3 time steps (i.e. 45 minutes) and a minimum number of cycles of 1 to ensure that a valid mean can be calculated even if the data is incomplete. The calculation formula is: the mean of the rainfall in the last three steps = (the rainfall at the current moment + the rainfall in the previous 15 minutes + the rainfall in the previous 30 minutes) / 3. Convert the time series into a sliding window sample of fixed length (the window length is 24 time steps, corresponding to 6 hours of historical data, and predicting the risk level in the next three steps) to ensure the time continuity of a single node, thereby enhancing the time series features;

[0055] (3) In normal scenarios, there are fewer data samples with medium and high risks, so feature perturbation and data enhancement are used. The data samples with medium and high risks include: medium risk samples (risk level is "medium") and high risk samples (risk level is "high"). These samples contain complete 14-dimensional feature data, including hydrological monitoring data (water level, flow velocity), meteorological environment data (rainfall, temperature), equipment status data (gate status, siltation degree), spatial topology data (latitude and longitude, node ID), risk assessment data (risk level, risk cause) and derived feature data (average rainfall of the last three steps, time normalized features). These samples fully participate in the above modeling process and are the core data source for training and prediction of the spatiotemporal attention enhancement model. Gaussian noise (σ = 0.1 to 0.4) is added to the key features of medium and high risk samples (such as rainfall and water level) and oversampled to alleviate class imbalance.

[0056] In a specific embodiment, it also includes:

[0057] The interactive visualization module is connected with the data acquisition module and the risk prediction module to display the risk prediction results and the changes in the data obtained by the data acquisition module.

[0058] In a specific embodiment, the risk prediction module specifically includes:

[0059] A risk prediction model construction unit is used to construct a risk prediction model, where the risk prediction model includes a temporal attention model, a spatial attention model, and a multi-task learning spatiotemporal fusion model;

[0060] The risk prediction unit is used to input the pre-processed multi-source risk data, hydrological equipment status data, time feature vectors, and discrete coding features into the risk prediction model for processing to obtain the corresponding risk prediction results.

[0061] In a specific embodiment, the specific implementation process of the risk prediction model construction unit includes:

[0062] The temporal attention model is implemented based on the Transformer self-attention mechanism model. It is used to obtain the dependencies between each time step in the preprocessed multi-source risk data, hydrological equipment status data, time feature vectors, and discrete coding features. It can capture the dependencies between each time step in the time series and strengthen the feature aggregation at the critical risk fluctuation moments.

[0063] The spatial attention model uses the topological relationships included in the upstream and downstream device node topology relationship model to generate node connection weights, with neighboring nodes having higher weights. It also calculates spatial similarity through the Euclidean distance of longitude and latitude, constructs a learnable projection matrix, maps spatial distance to attention weights, and integrates temporal and spatial attention scores to form an enhanced spatiotemporal attention model, thereby obtaining the spatiotemporal transmission characteristics of risks.

[0064] The multi-task learning spatiotemporal fusion model includes a multi-task learning framework with multi-branch outputs, which combines the spatiotemporal conduction characteristics for synchronous prediction to obtain the corresponding risk prediction results. The risk prediction results include risk levels and risk causes. The risk causes can output five types of risk causes (too high water level, excessive rainfall, siltation and blockage, gate failure, and no abnormality), and are jointly optimized through the cross-entropy loss function and the risk level prediction branch.

[0065] In a specific embodiment, the specific implementation process of the spatial attention model includes:

[0066] The node connection weights in the spatial attention model are implemented using a dynamic adjustment mechanism, which is dynamically adjusted based on the rainfall data included in the multi-source risk data. The specific implementation process includes:

[0067] The specific expression of node connection weight is:

[0068]

[0069] Where, d ij is the Euclidean distance between nodes i and j in terms of longitude and latitude, ∈ is the smoothing factor, the Euclidean distance weight between nodes i and j is used to measure the spatial proximity of adjacent nodes in the channel network; the topological connection weight reflects the direct influence of the upstream and downstream relationships in the channel; and the comprehensive spatial weight, a learnable projection matrix weight that combines geographic distance and topological relationship, is used for feature aggregation in the enhanced spatial attention module.

[0070] The influence weight of upstream nodes on downstream nodes is 0.1 to 0.3, and the weight coefficient in high-risk scenarios is ×1.2.

[0071] The specific process of dynamically adjusting weights also includes:

[0072] Dynamically adjust feature weights based on upstream heavy rainfall conditions (rainfall > 2.0 mm in the 15th minute after rainfall begins). During heavy rainfall periods, the weight of rainfall is increased to 0.7, the weight of siltation is adjusted to 0.2, the weight of gate status is reduced to 0.1, and the weight of temperature is adjusted to 0.0. During non-heavy rainfall periods, dynamic weights are assigned based on the contribution of siltation and gate status (formula: contribution = weight × standardized score): rainfall is weighted 0.4, siltation 0.3, gate status 0.2, and temperature 0.1. The sum of these weights is normalized to [0, 1]. These weight adjustments address the contribution of each risk factor (rainfall, siltation, gate status, and temperature) to the risk level prediction and are unrelated to node weights. The weight adjustment mechanism is closely related to the aforementioned upstream and downstream topological relationships. The rainfall status of upstream nodes influences the risk assessment weight allocation of downstream nodes, reflecting the spatial transmission characteristics of risk within the channel network.

[0073] In a specific embodiment, the specific implementation process of the multi-task learning spatiotemporal fusion model includes:

[0074] The multi-task learning spatiotemporal fusion model also includes a dynamic weight adjustment module, which is used to dynamically adjust the contribution weight of each risk factor according to the rainfall data and risk causes upstream of the irrigation area, thereby realizing adaptive feature weight allocation.

[0075] In a specific embodiment, the specific implementation process of the interactive visualization module includes:

[0076] The interactive visualization module is based on the GIS interactive platform, which visualizes the risk level in color and size. It also has the function of switching between multiple map layers and supports multiple display methods such as standard maps, satellite maps, topographic maps and dark maps. The system can intuitively display the risk level status of each node in the channel network in a spatially distributed manner. By clicking on any node, users can view the monitoring data of the node (such as water level, flow rate, gate status, siltation level, rainfall, etc.), the predicted results of the risk level and risk cause in the next multiple steps (such as the next 3 steps), and the risk transmission path between the node and the upstream and downstream nodes. The system automatically analyzes the spatial transmission relationship of risks, dynamically displays the evolution link of risks from upstream to downstream, and visually annotates the risk transmission path.

[0077] In a specific embodiment, the specific implementation process of the decision support module includes:

[0078] The decision support module pushes the governance plan corresponding to the risk cause in real time through the knowledge graph. The expert knowledge graph includes: risk level node, risk cause node and governance measure node. The mapping relationship between risk level-risk cause-governance plan is established through the rule boundary, and the corresponding governance plan is generated according to the risk level and risk cause.

[0079] For each risk node, the system automatically pushes management recommendations tailored to the current risk level and cause based on the knowledge graph. For example, when a node presents the risk of "excessive water levels," the system pushes targeted measures such as "activating emergency plans" and "diverting floodwaters" in real time, assisting managers in making informed decisions. This entire visual decision-making process integrates risk monitoring, prediction, tracing, transmission analysis, and the push of management recommendations, significantly enhancing the intelligence and responsiveness of irrigation district risk management.

[0080] An embodiment of the present invention further provides a prediction method using the irrigation area flood risk prediction system of any of the above embodiments, comprising the following steps:

[0081] Acquire multi-source risk data and hydrological equipment status data of the irrigation area, wherein the multi-source risk data includes meteorological environment data, and the hydrological equipment status data includes spatial topology data, and construct a topological relationship model of upstream and downstream equipment nodes of the irrigation area based on the multi-source risk data and hydrological equipment status data;

[0082] Preprocess the multi-source risk data and hydrological equipment status data respectively, extract the corresponding time feature vector from the preprocessed multi-source risk data, and extract the discrete coding feature from the preprocessed hydrological equipment status data;

[0083] Construct a risk prediction model and input the pre-processed multi-source risk data, hydrological equipment status data, time feature vectors, and discrete coding features into the risk prediction model for processing to obtain the corresponding risk prediction results;

[0084] Analyze the risk prediction results obtained according to the risk prediction model to assist decision making.

[0085] Subsequently, simulation verification was carried out on the system provided by the embodiment of the present invention, and the risk level prediction accuracy was ≥90%, the high-risk scenario identification accuracy was ≥92%, and the risk cause prediction accuracy was ≥99%. It supports real-time prediction of large-scale irrigation district networks that process 500+ nodes per second.

[0086] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0087] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A flood risk prediction system for irrigation areas, characterized in that: include: a data acquisition module for acquiring multi-source risk data and hydrological equipment status data of the irrigation area, wherein the multi-source risk data includes meteorological environment data, and the hydrological equipment status data includes spatial topology data, and constructing a topological relationship model of upstream and downstream equipment nodes of the irrigation area based on the multi-source risk data and the hydrological equipment status data; a data preprocessing module, configured to preprocess the multi-source risk data and the hydrological equipment status data respectively, and extract corresponding time feature vectors from the preprocessed multi-source risk data, and extract discrete coding features from the preprocessed hydrological equipment status data; A risk prediction module is used to construct a risk prediction model and input the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result; The decision support module is used to analyze the risk prediction results obtained by the risk prediction module to achieve auxiliary decision-making.

2. The irrigation area flood risk prediction system according to claim 1, characterized in that: Also includes: An interactive visualization module is connected to the data acquisition module and the risk prediction module, and is used to display the risk prediction results and changes in the data acquired by the data acquisition module.

3. The irrigation area flood risk prediction system according to claim 1, characterized in that: The specific implementation process of the data preprocessing module includes: Numerically encoding the discrete data included in the hydrological equipment status data, extracting corresponding discrete coding features and then normalizing them; Performing sliding window feature extraction on the rainfall data in the meteorological environment data to obtain a corresponding time feature vector; Gaussian noise is added to key feature data in the meteorological environment data and oversampling is performed to achieve category balance, wherein the key feature data includes rainfall data and water level data.

4. The irrigation area flood risk prediction system according to claim 3, characterized in that: The risk prediction module specifically includes: A risk prediction model construction unit, configured to construct a risk prediction model, wherein the risk prediction model includes a temporal attention model, a spatial attention model, and a multi-task learning spatiotemporal fusion model; A risk prediction unit is used to input the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result.

5. The irrigation area flood risk prediction system according to claim 4, characterized in that: The specific implementation process of the risk prediction model construction unit includes: The temporal attention model is implemented based on the Transformer self-attention mechanism model and is used to obtain the dependencies between the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding features. The spatial attention model uses the topological relationship included in the upstream and downstream device node topological relationship model to generate node connection weights, calculates spatial similarity through the Euclidean distance of longitude and latitude, constructs a learnable projection matrix, maps spatial distance to attention weight, and integrates time series and spatial attention scores to form an enhanced spatiotemporal attention model, thereby obtaining the spatiotemporal transmission characteristics of risks. The multi-task learning spatiotemporal fusion model includes a multi-task learning framework with multi-branch outputs, which combines spatiotemporal conduction characteristics for synchronous prediction to obtain corresponding risk prediction results, wherein the risk prediction results include risk levels and risk causes.

6. The irrigation area flood risk prediction system according to claim 5, characterized in that: The specific implementation process of the spatial attention model includes: The node connection weights in the spatial attention model are implemented using a dynamic adjustment mechanism and are dynamically adjusted according to the rainfall data included in the multi-source risk data.

7. The irrigation area flood risk prediction system according to claim 5, characterized in that: The specific implementation process of the multi-task learning spatiotemporal fusion model includes: The multi-task learning spatiotemporal fusion model also includes a dynamic weight adjustment module, which is used to dynamically adjust the contribution weight of each risk factor according to the rainfall data and risk causes upstream of the irrigation area to achieve adaptive feature weight allocation.

8. The irrigation area flood risk prediction system according to claim 1, characterized in that: The specific implementation process of the interactive visualization module includes: The interactive visualization module is based on a GIS interactive platform and visualizes risk levels using color and size.

9. The irrigation area flood risk prediction system according to claim 5, characterized in that: The specific implementation process of the decision support module includes: The decision support module pushes the governance plan corresponding to the risk cause in real time through the expert knowledge graph, where the expert knowledge graph includes: risk level node, risk cause node and governance measure node, and establishes a mapping relationship between risk level-risk cause-governance plan through rule boundaries, and generates corresponding governance plans according to risk level and risk cause.

10. A prediction method using the irrigation area flood risk prediction system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Acquiring multi-source risk data and hydrological equipment status data of the irrigation district, wherein the multi-source risk data includes meteorological environment data, and the hydrological equipment status data includes spatial topology data, and constructing a topological relationship model of upstream and downstream equipment nodes of the irrigation district based on the multi-source risk data and the hydrological equipment status data; Preprocessing the multi-source risk data and the hydrological equipment status data respectively, extracting corresponding time feature vectors from the preprocessed multi-source risk data, and extracting discrete coding features from the preprocessed hydrological equipment status data; Constructing a risk prediction model, and inputting the pre-processed multi-source risk data, the hydrological equipment status data, the time feature vector, and the discrete coding feature into the risk prediction model for processing to obtain a corresponding risk prediction result; The risk prediction results obtained according to the risk prediction model are analyzed to achieve auxiliary decision-making.

Images