Reservoir group flood control risk identification early warning system and method

Abstract

The invention provides a flood control risk identification and early warning system and method for a reservoir group, and belongs to the technical field of flood prevention and fighting. According to the method, a standard data set is generated by collecting rainfall, reservoir water level, flood discharge flow and landform data and performing data processing. And based on the standardized data set, constructing a hydrological model of the reservoir group in combination with GIS data, simulating flood discharge capacities under different rainfall scenes, calculating a flood discharge capacity index, a flood control risk index, a downstream vulnerability index, an upstream flood formation index and a midstream flood propagation index, and carrying out weighted fusion to generate a comprehensive risk index. And according to comparison between the comprehensive risk index and preset low, medium and high early warning threshold values, determining a current risk level and issuing early warning information. Meanwhile, user feedback is received, and model parameters and an early warning strategy are dynamically adjusted. According to the invention, through a multi-source data fusion technology, a deep learning algorithm and a dynamic feedback mechanism, accurate identification and timely early warning of the flood control risk of the reservoir group are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of flood prevention and control, and in particular to a reservoir group flood prevention risk identification and early warning system and method. Background Art

[0002] As global climate change intensifies, the frequency and intensity of extreme rainfall events have increased significantly, posing a significant challenge to flood control efforts across reservoirs. Traditional flood early warning methods, which rely primarily on single-source water level monitoring and empirical assessments, are unable to meet the complex and ever-changing needs of flood risk identification. Existing technologies have significant shortcomings in the following areas:

[0003] (1) The response speed and accuracy of the early warning mechanism are insufficient: Traditional early warning systems are mostly based on fixed thresholds and empirical judgments, and lack the ability to dynamically analyze real-time data, resulting in delayed release of early warning information and inability to respond to sudden flood events in a timely manner.

[0004] (2) Lack of dynamic simulation of flood propagation paths: Existing methods are unable to dynamically simulate the propagation paths and speeds of floods under different terrain conditions, resulting in inaccurate vulnerability assessments in downstream areas and difficulty in effectively guiding flood control decisions.

[0005] (3) Lack of user feedback mechanism: The existing system is unable to dynamically adjust model parameters and warning strategies based on user feedback in actual flood events, resulting in difficulty in ensuring the timeliness and accuracy of warning information.

[0006] Therefore, there is an urgent need for a reservoir group flood control risk identification and early warning system and method based on multi-source data fusion, deep learning algorithm and dynamic feedback mechanism to overcome the limitations of existing technologies and achieve accurate identification, dynamic assessment and timely early warning of reservoir group flood control risks. Summary of the Invention

[0007] The purpose of this invention is to provide a flood risk identification and early warning system and method for a reservoir cluster. By collecting real-time rainfall, reservoir water levels, flood discharge, and topographic data, and integrating IoT, satellite remote sensing, and GIS technologies, this system uses deep learning algorithms to fuse and process multi-source data. This system then constructs a hydrological model to simulate flood discharge capacity, calculates a comprehensive risk index, and distributes risk information through a three-level early warning mechanism. The system also receives user feedback and dynamically adjusts model parameters and early warning strategies to improve the accuracy and real-time nature of early warnings, ultimately enabling intelligent management of flood prevention and disaster reduction.

[0008] To achieve the above objectives, the present invention provides a method for identifying and warning flood control risks in a reservoir group, comprising the following steps:

[0009] Step S1: real-time collection of rainfall, reservoir water level, flood discharge, and topographic data, and cleaning, calibration, and fusion of the collected data to generate a standardized data set;

[0010] Step S2: Calculate the flood risk index of the reservoir group based on the standardized data set generated in step S1;

[0011] Combine GIS data to build a hydrological model of the reservoir group, simulate flood discharge capacity under different rainfall scenarios, and calculate the flood discharge capacity index;

[0012] Calculate the vulnerability index of the downstream area based on the population density, infrastructure distribution and terrain characteristics of the downstream area;

[0013] Calculate the flood formation index for the upstream region based on rainfall intensity, soil moisture, and terrain slope;

[0014] Calculate the flood propagation index for the midstream region based on the river channel slope, riverbed roughness, and river channel width;

[0015] The flood control risk index, flood discharge capacity index, downstream vulnerability index, upstream flood formation index and midstream flood propagation index are weighted and integrated to generate a comprehensive risk index;

[0016] Step S3: Set three warning thresholds: low, medium, and high. Compare the comprehensive risk index with the three warning thresholds to obtain the current risk level and issue warning information.

[0017] Step S4: The system receives user feedback and makes dynamic adjustments based on the user feedback.

[0018] Preferably, in step S1, data collection is performed through Internet of Things devices, satellite remote sensing and geographic information systems, and historical databases.

[0019] Preferably, in step S1, data fusion uses a deep learning algorithm to extract features, including using a convolutional neural network (CNN) to extract features from rainfall, reservoir water level, flood discharge and topography data to obtain multi-source data features, and using a recursive neural network (RNN) to extract features from the time series of rainfall and water level reservoirs to obtain time series features, and then fusing the multi-source data features and the time series features to obtain fused features.

[0020] Preferably, in step S2, the specific operation of calculating the flood control risk index of the reservoir group is as follows:

[0021] FRI t =α×LSTM'(F 融合 )+β×HIST 数据 ;

[0022] Among them, FRI t represents the flood risk index at time step t, α and β represent weight coefficients, and F 融合 Represents the fused features, LSTM' represents the LSTM network that introduces the multi-source data time correlation modeling module, HIST 数据 Indicates the impact of historical data;

[0023] LSTM'(F 融合 ) is calculated as follows:

[0024]

[0025] Among them, A t,i represents the attention weight of time step t on data source i; n represents the number of data sources;

[0026] The multi-source data time correlation modeling module is as follows:

[0027] A t =Attention(F 融合 ,W 时间序列 );

[0028] Among them, A t represents the attention weight vector at time step t; W 时间序列 Represents time series characteristics;

[0029] The calculation formula for the impact of historical data is as follows:

[0030]

[0031] Among them, m represents the number of samples of historical data, FRI ls,j Denotes the flood risk index of the jth historical sample, DCI ls,j represents the flood discharge capacity index of the jth historical sample;

[0032] DCI ls,j =γ×f(W t ,S t )+δ×HD t ;

[0033] Among them, γ and δ represent weight coefficients, f(W t ,S t ) indicates that the current water level W t and facility status S t Function, HD t represents the historical flood discharge data of time step t;

[0034] f(W t ,S t )=aW t +bSt ;

[0035] Among them, a and b represent weight coefficients.

[0036] Preferably, in step S2, the vulnerability index calculation formula of the downstream area is as follows:

[0037] VUL t =c×P t +d×I t +e×TF t ;

[0038] Among them, VUL t represents the vulnerability index, c, d, e represent the weight coefficients, P t represents the population density of the downstream area, I t represents the infrastructure distribution in the downstream area, TF t Indicates the topographic features of the downstream area;

[0039] The calculation formula for the flood formation index in the upstream area is as follows:

[0040] UFI t =g×R t +h×SM t +k×SG t ;

[0041] Among them, UFI t represents the flood formation index, g, h, k represent weight coefficients, R t represents the rainfall in the upstream area, SM t represents the soil moisture in the upstream area, SG t represents the topographic slope of the upstream area;

[0042] The calculation formula for the flood propagation index in the middle reaches is as follows:

[0043] MCI t =l×RG t +o×RC t +p×RW t ;

[0044] Among them, MCI t represents the flood propagation index, l, o, p represent weight coefficients, RG t Indicates the river slope in the middle reaches, RC t Indicates the river channel roughness in the middle reaches, RW t Indicates the width of the river in the midstream area.

[0045] Preferably, in step S2, the calculation formula of the comprehensive risk index is as follows:

[0046] CRI综合 =ι×FRI t +κ×DCI t +λ×VUI t +μ×UFI t +ν×MCI t ;

[0047] Among them, CRI 综合 represents the comprehensive risk index, ι, κ, λ, μ, ν represent weight coefficients, DCI t Represents the flood discharge capacity index.

[0048] Preferably, in step S3, when the comprehensive risk index is lower than the low-level warning threshold, it indicates that the current risk is low; when the comprehensive risk index is greater than or equal to the low-level warning threshold and less than or equal to the intermediate warning threshold, it indicates that the current risk is medium; when the comprehensive risk index is greater than the intermediate warning threshold, it indicates that the current risk is high.

[0049] Preferably, in step S3, the warning information is sent via SMS, APP push, broadcast and email channels.

[0050] Preferably, in step S4, user feedback is received, and the model parameters in step S2 and the early warning strategy in step S3 are dynamically adjusted according to the user feedback. The specific steps are as follows:

[0051] Update the LSTM's weight coefficients:

[0052]

[0053] Among them, θ τ+1 Represents the updated weight coefficient, θ τ represents the current weight coefficient, ψ represents the learning rate, represents the error gradient;

[0054] Adjust the warning threshold based on actual water level and rainfall intensity information fed back by users:

[0055] T 低 =ξ-ζ×σ;

[0056] T 中 =ξ+ζ×σ;

[0057] Among them, T 低 Indicates the low-level warning threshold, T 中 represents the intermediate warning threshold, ξ represents the average value of historical data, σ represents the standard deviation of historical data, and ζ represents the adjustment factor.

[0058] The present invention also provides a reservoir group flood prevention risk identification and early warning system, comprising:

[0059] Data acquisition module, used to collect real-time rainfall, reservoir water level, flood discharge and topographic data;

[0060] The data processing module is used to clean, calibrate and fuse the collected data to generate a standardized data set;

[0061] The risk calculation module is used to calculate the flood control risk index, flood discharge capacity index, downstream vulnerability index, upstream flood formation index and midstream flood propagation index of the reservoir group based on the standardized data set, and generate a comprehensive risk index through weighted fusion;

[0062] The early warning module is used to set low, medium and high warning thresholds, compare the comprehensive risk index with the thresholds, and issue corresponding warning information;

[0063] Feedback adjustment module: used to receive user feedback and dynamically adjust the parameters and early warning strategies of the risk calculation model, including updating weight coefficients and adjusting early warning thresholds;

[0064] GIS modeling module: used to build hydrological models of reservoir groups based on geographic information system data and simulate flood discharge capacity under different rainfall scenarios;

[0065] Historical data analysis module: used to analyze the impact of historical data on current flood control risks and optimize model parameters and early warning strategies.

[0066] Therefore, the present invention adopts the above-mentioned reservoir group flood prevention risk identification and early warning system and method, and the beneficial technical effects are as follows:

[0067] (1) Application of multi-source data fusion and deep learning algorithms.

[0068] The present invention uses IoT devices, satellite remote sensing, and geographic information systems (GIS) to collect real-time data on rainfall, reservoir water levels, flood discharge, and topography, and employs deep learning algorithms (such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs)) for feature extraction and data fusion. This multi-source data fusion technology overcomes the limitations of traditional methods that rely solely on a single data source, and can more comprehensively and accurately reflect the real-time status and potential risks of a reservoir cluster. Through the modeling capabilities of deep learning algorithms for time series data, the present invention can dynamically capture the implicit patterns in the data, significantly improving the accuracy and efficiency of data processing.

[0069] (2) Construction of dynamic risk assessment model and analysis of spatiotemporal correlation.

[0070] The present invention achieves a dynamic assessment of flood control risks of reservoir groups in both temporal and spatial dimensions, overcoming the limitations of existing technologies in single-dimensional analysis. This is specifically reflected in the following aspects:

[0071] Dynamic simulation of time dimension:

[0072] This paper introduces a long short-term memory (LSTM) network and an attention mechanism to dynamically model the time series characteristics of flood formation, propagation, and flood discharge capacity. Through the LSTM network, the system can capture the long-term dependencies between time series data such as rainfall intensity and reservoir water level changes, accurately predicting flood formation and development trends. For example, the system can identify the real-time impact of sudden changes in rainfall intensity on reservoir flood discharge capacity, thereby providing early warning of potential flood risks.

[0073] Regional correlation analysis in spatial dimension:

[0074] The present invention comprehensively considers the dynamic correlation among the upstream, midstream and downstream areas, calculates the upstream flood formation index, midstream flood propagation index and downstream vulnerability index respectively, and generates a comprehensive risk index through weighted fusion.

[0075] Upstream areas: Dynamically assess flood potential by analyzing rainfall intensity, soil moisture, and terrain slope. For example, when upstream rainfall intensity exceeds the soil's absorptive capacity, the system can quickly identify areas at high risk of flooding.

[0076] Midstream: Simulating flood propagation paths and speeds in the midstream, taking into account river slope, riverbed roughness, and river width. This dynamic simulation helps predict when floods will reach downstream, providing critical support for flood control decisions.

[0077] Downstream areas: Assess the vulnerability of downstream areas based on population density, infrastructure distribution, and terrain characteristics. For example, the system can identify potential damage that flooding may cause to downstream towns and critical infrastructure (such as bridges and hospitals).

[0078] Comprehensive assessment combining time and space:

[0079] By combining dynamic analysis across time and space, this invention can comprehensively quantify flood risk for a reservoir cluster. For example, during an extreme rainfall event, the system can simultaneously identify high-risk areas for flooding upstream, changes in flood propagation paths midstream, and areas of increased vulnerability downstream, thereby enabling precise location and tiered early warning of flood risks. This dynamic assessment model, combining time and space, significantly improves the comprehensiveness and timeliness of risk assessments, overcoming the shortcomings of existing technologies in single-dimensional analysis.

[0080] (3) Adaptive adjustment mechanism driven by user feedback.

[0081] The present invention dynamically adjusts model parameters and early warning strategies by receiving user feedback, including updating weight coefficients and adjusting warning thresholds. This adaptive adjustment mechanism overcomes the shortcomings of existing systems, which are unable to dynamically optimize based on actual flood events. This ensures that early warning information can be continuously improved and adapt to complex and changing flood scenarios. For example, when user feedback indicates a deviation between the actual water level and the model's prediction, the system can dynamically update the weight coefficients using the learning rate and error gradient, optimizing the model parameters and improving the accuracy of early warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 This is a flow chart of a reservoir group flood risk identification and early warning method according to the present invention;

[0083] Figure 2 This is a structural diagram of a reservoir group flood prevention risk identification and early warning system according to the present invention. DETAILED DESCRIPTION

[0084] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0085] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0086] Example 1

[0087] like Figure 1 As shown, a method for identifying and warning flood control risks of a reservoir group includes the following steps:

[0088] Step S1: Real-time collection of rainfall, reservoir water level, flood discharge and topography data, cleaning, calibrating and fusing the collected data to generate a standardized data set.

[0089] Data collection is carried out through IoT devices, satellite remote sensing and geographic information systems, and historical databases.

[0090] IoT devices: Sensors deployed in reservoirs, rivers, and key infrastructure (such as bridges and sluice gates) collect real-time data on rainfall, reservoir water levels, flood discharge, and more.

[0091] Satellite remote sensing: Obtain remote sensing images of topography, soil moisture, and flood coverage.

[0092] Geographic Information System (GIS): Provides static data such as the geographical distribution of reservoir groups, downstream population density, and infrastructure distribution.

[0093] Historical database: Call up data on past flood events, including rainfall intensity, reservoir discharge volume and disaster conditions.

[0094] The data is processed as follows:

[0095] Missing value processing: Missing values in sensor data are filled using time series interpolation.

[0096] Noise filtering: High-frequency noise in rainfall and water level data was removed by low-pass filtering.

[0097] Data calibration: Compare IoT devices with satellite remote sensing data to calibrate sensor deviations.

[0098] Data fusion uses deep learning algorithms for feature extraction, including using convolutional neural networks (CNN) to extract features from rainfall, reservoir water level, flood discharge and topography data to obtain multi-source data features, and using recurrent neural networks (RNN) to extract features from the time series of rainfall and water level reservoirs to obtain time series features. The multi-source data features and time series features are then fused to obtain the fused features.

[0099] Step S2: Calculate the flood risk index of the reservoir group based on the standardized data set generated in step S1;

[0100] Combining GIS data to build a hydrological model of the reservoir group, simulate the flood discharge capacity under different rainfall scenarios, and calculate the flood discharge capacity index. This is the existing technology and is briefly described below.

[0101] To construct a hydrological model for a reservoir cluster, GIS data are first collected, including terrain elevation, land use type, soil type, river network, and watershed boundaries. The collected GIS data are then preprocessed, including data cleaning, missing value filling, and noise filtering, to ensure data accuracy and usability. GIS tools are then used to delineate the watershed, determine the catchment area of each reservoir, and extract key watershed characteristics, such as drainage area, river channel length, slope, and river network density. Based on this GIS data, a distributed hydrological model (such as the SWAT model or the HEC-HMS model) is constructed. The model inputs include rainfall, evaporation, soil moisture, and terrain slope, and the outputs are reservoir discharge, water level changes, and flood propagation paths. Finally, the model is calibrated using historical flood event data, and model parameters are adjusted to improve simulation accuracy. The model's accuracy and reliability are verified by comparing the simulation results with actual observations.

[0102] Calculate the vulnerability index of the downstream area based on the population density, infrastructure distribution and terrain characteristics of the downstream area;

[0103] Calculate the flood formation index for the upstream region based on rainfall intensity, soil moisture, and terrain slope;

[0104] Calculate the flood propagation index for the midstream region based on the river channel slope, riverbed roughness, and river channel width;

[0105] The flood control risk index, flood discharge capacity index, downstream vulnerability index, upstream flood formation index and midstream flood propagation index are weighted and integrated to generate a comprehensive risk index.

[0106] The specific operations for calculating the flood control risk index of a reservoir group are as follows:

[0107] FRI t =α×LSTM'(F 融合 )+β×HIST 数据 (1);

[0108] Among them, FRI t represents the flood risk index at time step t, α and β represent weight coefficients, and F 融合 Represents the fused features, LSTM' represents the LSTM network that introduces the multi-source data time correlation modeling module, HIST 数据 Indicates the impact of historical data;

[0109] LSTM'(F 融合 ) is calculated as follows:

[0110]

[0111] Among them, A t,i represents the attention weight of time step t on data source i;

[0112] The multi-source data time correlation modeling module is as follows:

[0113] A t =Attention(F 融合 ,W 时间序列 )(3);

[0114] Among them, A t represents the attention weight vector at time step t; W 时间序列 Represents time series characteristics;

[0115] The calculation formula for the impact of historical data is as follows:

[0116]

[0117] Among them, m represents the number of samples of historical data, FRI ls,j Denotes the flood risk index of the jth historical sample, DCI ls,j represents the flood discharge capacity index of the jth historical sample;

[0118] DCI ls,j=γ×f(W t ,S t )+δ×HD t (5);

[0119] Among them, γ and δ represent weight coefficients, f(W t ,S t ) indicates that the current water level W t and facility status S t Function, HD t represents the historical flood discharge data of time step t;

[0120] f(W t ,S t )=aW t +bS t (6);

[0121] Among them, a and b represent weight coefficients.

[0122] The vulnerability index calculation formula for the downstream area is as follows:

[0123] VUL t =c×P t +d×I t +e×TF t (7);

[0124] Among them, VUL t represents the vulnerability index, c, d, e represent the weight coefficients, P t represents the population density of the downstream area, I t represents the infrastructure distribution in the downstream area, TF t Indicates the topographic features of the downstream area;

[0125] The calculation formula for the flood formation index in the upstream area is as follows:

[0126] UFI t =g×R t +h×SM t +k×SG t (8);

[0127] Among them, UFI t represents the flood formation index, g, h, k represent weight coefficients, R t represents the rainfall in the upstream area, SM t represents the soil moisture in the upstream area, SG t represents the topographic slope of the upstream area;

[0128] The calculation formula for the flood propagation index in the middle reaches is as follows:

[0129] MCI t=l×RG t +o×RC t +p×RW t (9);

[0130] Among them, MCI t represents the flood propagation index, l, o, p represent weight coefficients, RG t Indicates the river slope in the middle reaches, RC t Indicates the river channel roughness in the middle reaches, RW t Indicates the width of the river in the midstream area.

[0131] The calculation formula for the comprehensive risk index is as follows:

[0132] CRI 综合 =ι×FRI t +κ×DCI t +λ×VUI t +μ×UFI t +ν×MCI t (10);

[0133] Among them, CRI 综合 represents the comprehensive risk index, ι, κ, λ, μ, ν represent weight coefficients, DCI t It represents the flood discharge capacity index, and its calculation formula is the same as formula (5).

[0134] Step S3: Set low, medium, and high warning thresholds, compare the comprehensive risk index with the three-level warning thresholds, obtain the current risk level, and issue warning information (warning information is sent via SMS, APP push, broadcast, and email channels).

[0135] When the comprehensive risk index is lower than the low-level warning threshold, it indicates that the current risk is low. When the comprehensive risk index is greater than or equal to the low-level warning threshold and less than or equal to the medium-level warning threshold, it indicates that the current risk is medium. When the comprehensive risk index is greater than the medium-level warning threshold, it indicates that the current risk is high.

[0136] Step S4: Receive user feedback and dynamically adjust the model parameters in step S2 and the early warning strategy in step S3 according to the user feedback.

[0137] Receive user feedback and dynamically adjust the model parameters in step S2 and the warning strategy in step S3 based on the user feedback. The specific steps are as follows:

[0138] Update the LSTM's weight coefficients:

[0139]

[0140] Among them, θ τ+1 Represents the updated weight coefficient, θτ represents the current weight coefficient, ψ represents the learning rate, represents the error gradient;

[0141] Adjust the warning threshold based on actual water level and rainfall intensity information fed back by users:

[0142] T 低 =ξ-ζ×σ(12);

[0143] T 中 =ξ+ζ×σ(13);

[0144] Among them, T 低 Indicates the low-level warning threshold, T 中 represents the intermediate warning threshold, ξ represents the average value of historical data, σ represents the standard deviation of historical data, and ζ represents the adjustment factor.

[0145] The present invention will be further described below through simulation experiments.

[0146] Data source:

[0147] Historical data: data on flood events in a certain place over the past 10 years.

[0148] Simulation data: simulates moderate rainfall (100 mm / 24 h), extreme rainfall (200 mm / 24 h) and continuous rainfall events (rainfall of 120 mm, 150 mm and 180 mm over three days).

[0149] Data generation method:

[0150] Generate rainfall data randomly using a normal distribution.

[0151] Simulate reservoir water level and flood discharge flow based on rainfall and reservoir flood discharge capacity.

[0152] Extract static data such as terrain slope and river width from the GIS database.

[0153] In order to illustrate the effect of the present invention, the following comparison model is also set up:

[0154] Comparative model 1: Reducing the upstream flood formation index.

[0155] Comparative model 2: reducing the flood propagation index in the middle reaches.

[0156] Comparative Model 3: Reducing the downstream vulnerability index.

[0157] Comparative model 4: Simultaneously reducing the flood formation index in the upper and middle reaches.

[0158] The simulation results are shown in Table 1. By comparing the performance of the complete model with several comparison models under different rainfall scenarios, the simulation experiments verified the impact of various factors on the performance of the reservoir flood risk identification and early warning system. The complete model demonstrated high accuracy in calculating the comprehensive risk index and high early warning accuracy under all rainfall scenarios.

[0159] Specifically, the comprehensive risk index of the complete model was 0.45 under moderate rainfall scenarios, 0.85 under extreme rainfall scenarios, and gradually increased from 0.5 to 0.75 under continuous rainfall scenarios, with warning accuracy rates reaching 95%, 90%, and 88%, respectively. In contrast, after reducing the upstream flood formation index and midstream flood propagation index in the comparison model, the calculated comprehensive risk index and warning accuracy decreased, with the impact being particularly significant under extreme rainfall and continuous rainfall scenarios. For example, the comprehensive risk index of comparison model 4 under extreme rainfall scenarios was only 0.78, and the warning accuracy dropped to 80%. This demonstrates that considering all factors in the complete model is crucial to ensuring the system's high performance and reliability.

[0160] Table 1 Simulation results

[0161]

[0162]

[0163] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0164] Therefore, the present invention adopts the aforementioned reservoir flood risk identification and early warning system and method. By collecting real-time rainfall, reservoir water levels, flood discharge, and topographic data, and integrating IoT, satellite remote sensing, and GIS technologies, the system uses deep learning algorithms to fuse and process multi-source data. This system then constructs a hydrological model to simulate flood discharge capacity, calculates a comprehensive risk index, and distributes risk information through a three-level early warning mechanism. Furthermore, the system receives user feedback and dynamically adjusts model parameters and early warning strategies to improve the accuracy and real-time nature of early warnings, ultimately achieving intelligent management of flood prevention and disaster reduction.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying and warning flood control risks in a reservoir group, characterized in that: The following steps are involved: Step S1: real-time collection of rainfall, reservoir water level, flood discharge, and topographic data, and cleaning, calibration, and fusion of the collected data to generate a standardized data set; Step S2: Calculate the flood risk index of the reservoir group based on the standardized data set generated in step S1; Combine GIS data to build a hydrological model of the reservoir group, simulate flood discharge capacity under different rainfall scenarios, and calculate the flood discharge capacity index; Calculate the vulnerability index of the downstream area based on the population density, infrastructure distribution and terrain characteristics of the downstream area; Calculate the flood formation index for the upstream region based on rainfall intensity, soil moisture, and terrain slope; Calculate the flood propagation index for the midstream region based on the river channel slope, riverbed roughness, and river channel width; The flood control risk index, flood discharge capacity index, downstream vulnerability index, upstream flood formation index and midstream flood propagation index are weighted and integrated to generate a comprehensive risk index; Step S3: Set three warning thresholds: low, medium, and high. Compare the comprehensive risk index with the three warning thresholds to obtain the current risk level and issue warning information. Step S4: The system receives user feedback and makes dynamic adjustments based on the user feedback.

2. A reservoir group flood risk identification and early warning method according to claim 1, characterized in that: In step S1, data collection is performed through IoT devices, satellite remote sensing and geographic information systems, and historical databases.

3. A reservoir group flood risk identification and early warning method according to claim 1, characterized in that: In step S1, data fusion uses a deep learning algorithm to extract features, including using a convolutional neural network (CNN) to extract features from rainfall, reservoir water level, flood discharge, and topography data to obtain multi-source data features, and using a recurrent neural network (RNN) to extract features from the time series of rainfall and water level reservoirs to obtain time series features. Then, the multi-source data features and the time series features are fused to obtain the fused features.

4. A reservoir group flood risk identification and early warning method according to claim 1, characterized in that: In step S2, the specific operations for calculating the flood risk index of the reservoir group are as follows: FREE t =α×LSTM'(F 融合 )+β×HIST 数据 ; Among them, FRI t represents the flood risk index at time step t, α and β represent weight coefficients, and F 融合 Represents the fused features, LSTM' represents the LSTM network that introduces the multi-source data time correlation modeling module, HIST 数据 Indicates the impact of historical data; LSTM'(F 融合 ) is calculated as follows: Among them, A t,i represents the attention weight of time step t on data source i; n represents the number of data sources; The multi-source data time correlation modeling module is as follows: A t =Attention(F 融合 ,W 时间序列 ); Among them, A t represents the attention weight vector at time step t; W 时间序列 Represents time series characteristics; The calculation formula for the impact of historical data is as follows: Among them, m represents the number of samples of historical data, FRI ls,j Denotes the flood risk index of the jth historical sample, DCI ls,j represents the flood discharge capacity index of the jth historical sample; DCI ls,j =γ×f(W t ,S t )+δ×HD t ; Among them, γ and δ represent weight coefficients, f(W t ,S t ) indicates that the current water level W t and facility status S t Function, HD t Represents historical flood discharge data at time step t.

5. A reservoir group flood risk identification and early warning method according to claim 4, characterized in that: In step S2, the vulnerability index of the downstream area is calculated as follows: FILL t =c×P t +d×I t +e×TF t ; Among them, VUL t represents the vulnerability index, c, d, e represent the weight coefficients, P t represents the population density of the downstream area, I t represents the infrastructure distribution in the downstream area, TF t Indicates the topographic features of the downstream area; The calculation formula for the flood formation index in the upstream area is as follows: UFI t =g×R t +h×SM t +k×SG t ; Among them, UFI t represents the flood formation index, g, h, k represent weight coefficients, R t represents the rainfall in the upstream area, SM t represents the soil moisture in the upstream area, SG t represents the topographic slope of the upstream area; The calculation formula for the flood propagation index in the middle reaches is as follows: MCI t =l×RG t +o×RC t +p×RW t ; Among them, MCI t represents the flood propagation index, l, o, p represent weight coefficients, RG t Indicates the river slope in the middle reaches, RC t Indicates the river channel roughness in the middle reaches, RW t Indicates the width of the river channel in the midstream area.

6. A reservoir group flood risk identification and early warning method according to claim 5, characterized in that: In step S2, the calculation formula of the comprehensive risk index is as follows: CRI 综合 =ι×FRI t +κ×DCI t +λ×VUI t +μ×UFI t +ν×MCI t ; Among them, CRI 综合 Denotes the comprehensive risk index, DCI t represents the flood discharge capacity index, and ι, κ, λ, μ, and ν represent weight coefficients.

7. A reservoir group flood risk identification and early warning method according to claim 1, characterized in that: In step S3, when the comprehensive risk index is lower than the low-level warning threshold, it indicates that the current risk is low; when the comprehensive risk index is greater than or equal to the low-level warning threshold and less than or equal to the intermediate warning threshold, it indicates that the current risk is medium; when the comprehensive risk index is greater than the intermediate warning threshold, it indicates that the current risk is high.

8. The method for identifying and warning flood risk of a reservoir group according to claim 1, characterized in that: In step S3, the warning information is sent through SMS, APP push, broadcast and email channels.

9. A reservoir group flood risk identification and early warning method according to claim 1, characterized in that: In step S4, user feedback is received and the model parameters in step S2 and the early warning strategy in step S3 are dynamically adjusted according to the user feedback. The specific steps are as follows: Update the LSTM's weight coefficients: i τ+1 =θ τ -ψ×▽E τ ; Among them, θ τ+1 Represents the updated weight coefficient, θ τ represents the current weight coefficient, ψ represents the learning rate, ▽E τ represents the error gradient; Adjust the warning threshold based on actual water level and rainfall intensity information fed back by users: T 低 =ξ-ζ×σ; T 中 =ξ+ζ×σ; Among them, T 低 Indicates the low-level warning threshold, T 中 represents the intermediate warning threshold, ξ represents the average value of historical data, σ represents the standard deviation of historical data, and ζ represents the adjustment factor.

10. A reservoir group flood risk identification and early warning system, characterized in that: include: Data acquisition module, used to collect real-time rainfall, reservoir water level, flood discharge and topographic data; The data processing module is used to clean, calibrate and fuse the collected data to generate a standardized data set; The risk calculation module is used to calculate the flood control risk index, flood discharge capacity index, downstream vulnerability index, upstream flood formation index and midstream flood propagation index of the reservoir group based on the standardized data set, and generate a comprehensive risk index through weighted fusion; The early warning module is used to set low, medium and high warning thresholds, compare the comprehensive risk index with the thresholds, and issue corresponding warning information; Feedback adjustment module: used to receive user feedback and dynamically adjust the parameters and early warning strategies of the risk calculation model, including updating weight coefficients and adjusting early warning thresholds.

Images