Forecasting and early warning method and system for researching flood in hill and plain mixed region

Abstract

The invention discloses a forecasting and early warning method and system for researching flood in a hill and plain mixed region, and belongs to the technical field of flood early warning. Comprising the steps of S1, multi-source data fusion and adaptive preprocessing, S2, lightweight data preprocessing, S3, low-dimensional feature engineering, S4, lightweight dynamic model construction, S5, dynamic early warning threshold generation, S6, rapid early warning and feedback optimization and S7, verification and applicability expansion. On the basis of forecasting and early warning of flood, low parameter requirements can be input, calculation time is short, and both precision and efficiency are considered.

Description

Technical Field

[0001] The present invention relates to the field of flood warning technology, and more particularly to a method and system for studying flood forecasting and warning in mixed hilly and plain areas. Background Art

[0002] Floods are natural disasters caused by heavy rain, torrential downpours, or sustained rainfall that inundate and waterlogge low-lying areas, posing a serious threat to human life, property, and the social economy. Mixed hilly and plain areas are a typical landform in my country. For example, the Qinhuai River Basin in Nanjing, Jiangsu Province, located in the lower reaches of the Yangtze River, is 74% hilly and mountainous, while low-lying plains account for 26%. Surrounded by mountains on all sides and flat in the middle, it forms a complete intermountain basin, a typical mixed hilly and plain area. These areas have short upstream and downstream distances and large elevation differences, resulting in a short source of floodwaters and rapid flow. This leads to low storage capacity in the upper and middle reaches, rapid flood rises, and frequent flood peaks. The lower reaches merge into the Yangtze River, where flood levels are affected by the impact of floodwaters from the Yangtze River downstream, resulting in a high risk of flood disasters.

[0003] In order to effectively prevent and mitigate the impact of flood disasters in such areas, it is necessary to establish scientific and reasonable flood disaster forecasting and early warning methods and systems, provide accurate and reliable flood disaster information in a timely manner, and guide flood prevention and disaster relief work.

[0004] Traditional flood disaster forecasting and early warning methods are primarily based on hydrological and hydrodynamic models. By simulating regional hydrological and hydrodynamic processes, these methods predict hydrological and hydrodynamic elements such as water level, flow rate, and flow velocity at various river control nodes, thereby determining the likelihood and severity of flood disasters. However, this approach has several limitations, including difficulty determining model parameters, high input data requirements, large computational complexity, long computational time, and low efficiency.

[0005] Therefore, a flood forecasting and warning method and system that has low requirements on input parameters, short calculation time, and takes both accuracy and efficiency into consideration and can be applied to mixed areas of hills and plains is desired. Summary of the Invention

[0006] 1. Technical problems to be solved

[0007] In response to the problems existing in the prior art, the purpose of the present invention is to provide a forecasting and early warning method and system for studying floods in mixed areas of hills and plains. On the basis of realizing the forecasting and early warning of floods, the present invention can also input low parameter requirements, shorten the calculation time, and take into account both accuracy and efficiency.

[0008] 2. Technical solution

[0009] In order to solve the above problems, the present invention adopts the following technical solutions:

[0010] A flood forecasting and early warning method for mixed hilly and plain areas, including:

[0011] S1, multi-source data fusion and adaptive preprocessing;

[0012] S2, lightweight data preprocessing;

[0013] S3, low-dimensional feature engineering;

[0014] S4, lightweight dynamic model construction;

[0015] S5, dynamic warning threshold generation;

[0016] S6. Rapid warning and feedback optimization;

[0017] S7. Verification and applicability extension.

[0018] As a preferred solution of the present invention, step S1 includes:

[0019] S101. Low-parameter data collection: Integrate rainfall data from weather stations, river water level monitoring, soil moisture sensors, and satellite remote sensing data. Standardize data to eliminate dimensional differences and use Kriging GIS spatial interpolation technology to supplement missing values, reducing reliance on a single parameter.

[0020] S102, Terrain Adaptive Modeling: Dynamically divide the grid based on the DEM digital elevation model, identify the river network and sub-basin boundaries, process mixed terrain through adaptive resolution technology, and optimize computing efficiency.

[0021] As a preferred solution of the present invention, step S2 includes:

[0022] S201. Simplified filling of missing values: Use global constants or attribute means to fill missing values, avoiding complex regression or decision tree methods;

[0023] S202. Noise smoothing simplification: using binning method local smoothing technology to replace regression or clustering denoising with large computational complexity, wherein the binning method is to smooth the stored data values by examining the data range.

[0024] As a preferred solution of the present invention, step S3 includes:

[0025] S301, Redundant Feature Screening: Quickly identify redundant attributes using the Pearson product-distance function correlation coefficient, rather than the chi-square test or covariance analysis;

[0026] S302. Attribute subset selection: retain key influencing factors and use data reduction methods to actively eliminate weakly correlated parameters.

[0027] As a preferred solution of the present invention, the data reduction method in step S302 includes: data cube aggregation, data attribute subset selection, dimension reduction, value reduction and discretization;

[0028] Data cube aggregation is used for data in a data cube structure. Data cubes store multidimensional aggregation information. Each cell stores an aggregation value, corresponding to a point in the multidimensional space. Each attribute may have a conceptual hierarchy, allowing data analysis at multiple levels of abstraction.

[0029] Attribute subset selection When the data set to be analyzed contains a large number of attributes, most of which are irrelevant or redundant to the mining task, attribute subset selection can detect and delete irrelevant, redundant or weakly relevant attributes or dimensions. Its goal is to find the minimum attribute set so that the probability distribution of the data class is as close as possible to the original distribution obtained using all attributes. For attribute subset selection;

[0030] Dimensionality reduction uses data encoding or transformation to obtain a reduced or compressed representation of the original data, reducing the number of random variables or attributes considered. If the reduced data can only reconstruct an approximate representation of the original data, then the data reduction is lossy. If the original data can be reconstructed without losing any information, then it is lossless.

[0031] Numerical reduction reduces the amount of data by choosing alternative data representations, replacing or estimating data with smaller data representations;

[0032] Data discretization divides the attribute value domain into intervals to reduce the number of given continuous attribute values. The interval labels can replace the actual data values, and the numerical values of continuous attributes are replaced by a few interval labels.

[0033] As a preferred solution of the present invention, step S4 includes:

[0034] S401, Hybrid hydrological model selection: Combine the distributed hydrological model with the simplified physical mechanism and adopt the improved Model, where f and B are the inverse of soil infiltration capacity, and f and B are used to dynamically adjust runoff calculation;

[0035] S402. Machine Learning Model Optimization: Use a lightweight model such as a random forest or LSTM network to train a time series prediction model with rainfall, water level, and soil moisture as input, and reduce redundant parameters through LASSO regression feature selection.

[0036] As a preferred solution of the present invention, step S5 includes:

[0037] S501. Multi-scenario simulation and threshold delineation: Generate random rainfall scenarios based on historical rainfall data, simulate runoff generation and convergence processes using a distributed hydrological model, and dynamically calculate critical rainfall and water level thresholds for different regions. Simultaneously, a Seq2Seq model, using rainfall, water level, and rise rate as input, predicts disaster processes and generates dynamic thresholds.

[0038] S502. Soil-topography coupling analysis: In plain areas, soil storage and runoff generation are considered, while in mountainous areas, a direct runoff generation mechanism is introduced when the rainfall intensity exceeds the infiltration rate. The runoff inflow time is corrected by combining vegetation coverage and surface evaporation.

[0039] As a preferred solution of the present invention, step S6 includes:

[0040] S601, Real-time dynamic warning release: Integrate warning results through the GIS platform, display high-, medium-, and low-risk areas in a risk heat map, and release warning information through multiple channels such as SMS, broadcast, and App;

[0041] S602, model iterative optimization: Utilize cross-validation and online learning mechanisms, combine real-time monitoring data to dynamically update model parameters, and adjust the hyperparameters of the Seq2Seq model through K-fold cross-validation to ensure generalization ability.

[0042] As a preferred solution of the present invention, step S7 includes:

[0043] S701. Multi-case verification: Select a typical hilly transition zone and compare historical disaster data with model prediction results. The evaluation accuracy F1 score is greater than 0.85 and the computational time for a single simulation is less than 10 minutes.

[0044] S702, modular expansion design: Encapsulate the model into independent modules to support fast switching of different regional parameters such as permeability coefficient library in plain area and slope in mountain area.

[0045] A flood forecasting and warning system for mixed hilly and plain areas includes: a lightweight data acquisition and fusion module, an adaptive cleaning module, a hybrid model inference core, a spatiotemporal compression module, a database module, a hydrological forecast module, a warning application module, a hierarchical warning decision module, a three-dimensional spatial warning mapping module, and a crowdsourced data feedback channel;

[0046] The lightweight data acquisition and fusion module integrates data from multiple types of sensors such as water level stations, rain gauges, and weather radars, supports hourly or minute-level data stream access, uses a sliding window mean method to replace complex parameter input, and automatically calibrates key thresholds through historical data to reduce manual intervention. The adaptive cleaning module has a built-in dynamic filling strategy for missing values, and automatically identifies outliers in combination with covariance analysis. The spatiotemporal dimension compression module uses the DBA average sequence algorithm DTW dynamic time warping technology to align multi-site data in time and space to reduce data redundancy. The database module is used to provide sufficient training samples for hydrological forecasts and provide discriminant data support for risk zoning warnings. The database module includes a historical hydrological monitoring database module, current real-time water conservancy project information The hydrological forecast module selects input parameters based on the database module and establishes a regional hydrological forecast through LSTM deep learning. The early warning application module generates graded warning labels based on the water level values of the control nodes predicted by the hydrological forecast module and the corresponding warning water levels for reference in decision-making and command. The graded warning decision module evaluates the model reliability in real time based on the Nash coefficient and triggers model self-checking when NSE < 0.95. The three-dimensional spatial warning mapping module is used to map the water level prediction value to the transition zone from hills to plains in combination with the DEM digital elevation model to generate a flood depth heat map. The crowdsourcing data feedback channel is used to access flood images on social media and use CV technology to assist in verifying the accuracy of the warning area.

[0047] The historical hydrological monitoring database module includes historical rainfall and evaporation information of each rain gauge in the region, historical water level and flow information of the control station, and tide level information along the river at the regional exit. The current real-time water conservancy project information database module includes the length, direction, bottom elevation, bottom width, and slope ratio information of the main flood-carrying river channels in the region, as well as the scheduling and application rules of reservoirs, flood storage areas, embankment areas, and sluice stations in the region. The future hypothetical scenario database module is based on the historical rainfall of each rain gauge in the region, and obtains rainfall and rainfall distribution at different frequencies through frequency calculation. Through the historical Yangtze River tide data, frequency calculation is performed to obtain tide types at different frequencies. Based on the hydrological and hydrodynamic model simulation calculation, the water level, flow, flow velocity and other information of the main control stations under different future flood and tide combination scenarios are obtained, thereby forming a future hypothetical scenario database.

[0048] 3. Beneficial effects

[0049] Compared with the prior art, the advantages of the present invention are:

[0050] (1) The present invention only requires historical water level and rainfall station time series data, avoiding complex parameter calibration. It is suitable for mixed terrain areas with scarce data. Through the pattern library and DTW distance matching, it can adaptively identify sudden floods in hilly areas and slow floods in plain areas. The ASCS algorithm reduces the time consumption of LSTM parameter optimization by 40%, and combines the attention mechanism to filter redundant information. The single prediction time is less than 3 seconds.

[0051] (2) The present invention uses a lightweight model to achieve MAE ≤ 0.017m within a forecast period of 1-3 hours, meeting the real-time warning needs. The Nash coefficient reaches 0.992, which is 15% higher than the traditional model, and the underreporting rate of extreme events is reduced by 22%. The dynamic time-bending average sequence reduces pattern redundancy, making the peak prediction error stable within ±5%. It supports multi-site data fusion and quickly matches similar flood patterns through hierarchical clustering of symbolic distance, realizing basin-level linkage warning, taking into account both short-duration warning in hilly areas and long-duration scheduling needs in plain areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of a method for forecasting and warning floods in mixed hilly and plain areas according to the present invention;

[0053] Figure 2 This is a module diagram of a forecast and early warning system for studying floods in mixed hilly and plain areas according to the present invention. DETAILED DESCRIPTION

[0054] 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 them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

[0055] Example:

[0056] See also Figure 1 , a flood forecasting and early warning method for mixed hilly and plain areas, including:

[0057] S1, multi-source data fusion and adaptive preprocessing;

[0058] S2, lightweight data preprocessing;

[0059] S3, low-dimensional feature engineering;

[0060] S4, lightweight dynamic model construction;

[0061] S5, dynamic warning threshold generation;

[0062] S6. Rapid warning and feedback optimization;

[0063] S7. Verification and applicability extension.

[0064] In a specific embodiment of the present invention, by only requiring historical water level and rainfall station time series data, complex parameter calibration is avoided, and it is suitable for mixed terrain areas with scarce data. Through the pattern library and DTW distance matching, it can adaptively identify sudden floods in hilly areas and slow rise processes in plain areas. The ASCS algorithm reduces the time cost of LSTM parameter optimization by 40%, combines the attention mechanism to filter redundant information, and the single prediction time cost is less than 3 seconds. Through the lightweight model, MAE ≤ 0.017m within the forecast period of 1-3 hours, meeting the real-time warning needs. The Nash coefficient reaches 0.992, which is 15% higher than the traditional model, and the extreme event omission rate is reduced by 22%. The dynamic time warping average sequence reduces pattern redundancy, so that the peak prediction error is stabilized within ±5%. It supports multi-site data fusion and quickly matches similar flood patterns through signed distance hierarchical clustering to achieve basin-level linkage warning, taking into account the short-duration warning in hilly areas and the long-duration scheduling needs in plain areas.

[0065] Specifically, step S1 includes:

[0066] S101. Low-parameter data collection: Integrate rainfall data from weather stations, river water level monitoring, soil moisture sensors, and satellite remote sensing data. Standardize data to eliminate dimensional differences and use Kriging GIS spatial interpolation technology to supplement missing values, reducing reliance on a single parameter.

[0067] S102, Terrain Adaptive Modeling: Dynamically divide the grid based on the DEM digital elevation model, identify the river network and sub-basin boundaries, and process mixed terrain through adaptive resolution technology, such as using coarse-grained grids in plains and fine-grained grids in mountainous areas to optimize computing efficiency.

[0068] Specifically, step S2 includes:

[0069] S201. Simplified filling of missing values: Use global constants or attribute means to fill missing values, avoiding complex regression or decision tree methods;

[0070] S202. Simplified noise smoothing: The binning method is used to replace the computationally intensive regression or clustering denoising. The binning method is to smooth the stored data values by examining the data values around the data. The stored values are divided into some bins. Since only the values of the nearest neighbors are examined, the binning method performs local smoothing.

[0071] Specifically, step S3 includes:

[0072] S301. Redundant feature screening: Redundant attributes are quickly identified through the Pearson product distance function correlation coefficient, rather than the chi-square test or covariance analysis. The Pearson product distance function estimates the correlation r and g of the numerical attributes A and B, which are defined as formula 1. Formula 1 is:

[0073]

[0074] Where N is the number of tuples (samples), a i and b i are the values of tuple (sample) i in attributes A and B, and are the means of attributes A and B, σ A and σ B are the standard deviations of attributes A and B in these N samples, is the inner product of the attributes A and B in these N samples, -1≤r A,B ≤1, if r A，B Greater than 0, indicating that the value of attribute A increases as the value of attribute B increases, r A，B The larger the r, the stronger the correlation, that is, the greater the possibility that each attribute contains another attribute, so the larger the r A，B The value indicates that attribute A or B can be removed as redundant. A，B If it is equal to 0, then A and B are independent and unrelated. If it is less than 0, then A and B are negatively correlated, indicating that each attribute prevents the other from appearing.

[0075] Specifically, the data reduction method in step S302 includes: data cube aggregation, data attribute subset selection, dimension reduction, value reduction and discretization;

[0076] Data cube aggregation is used for data in a data cube structure. Data cubes store multidimensional aggregation information. Each cell stores an aggregation value, corresponding to a point in the multidimensional space. Each attribute may have a conceptual hierarchy, allowing data analysis at multiple levels of abstraction.

[0077] Attribute subset selection: When the dataset to be analyzed contains a large number of attributes, most of which are irrelevant or redundant to the mining task, attribute subset selection can detect and remove irrelevant, redundant, or weakly relevant attributes or dimensions. Its goal is to find the minimum attribute set so that the probability distribution of the data class is as close as possible to the original distribution obtained by using all attributes. Its advantage is that it reduces the number of attributes that appear in the discovered pattern, making the pattern easier to understand. For attribute subset selection, exhaustive search to find the best attribute subset may be unrealistic. Therefore, heuristic algorithms that compress the search space are often used.

[0078] Dimensionality reduction uses data encoding or transformation to obtain a reduced or compressed representation of the original data, reducing the number of random variables or attributes considered. If the reduced data can only reconstruct an approximate representation of the original data, then the data reduction is lossy. If the original data can be reconstructed without losing any information, then it is lossless.

[0079] Numerical reduction reduces the amount of data by choosing alternative data representations, replacing or estimating data with smaller data representations;

[0080] Data discretization divides the attribute value domain into intervals to reduce the number of given continuous attribute values. The interval labels can replace the actual data values. Replacing the numerical values of continuous attributes with a few interval labels can reduce and simplify the original data, making the data analysis results of unsupervised learning concise, easy to use, and have a knowledge-level representation.

[0081] Specifically, step S4 includes:

[0082] S401, Hybrid hydrological model selection: Combine the distributed hydrological model with the simplified physical mechanism and adopt the improved Model, where f and B are the inverse of soil infiltration capacity, and f and B are used to dynamically adjust runoff calculations;

[0083] S402. Machine Learning Model Optimization: Use a lightweight model such as a random forest or LSTM network to train a time series prediction model with rainfall, water level, and soil moisture as input. LASSO regression feature selection is used to reduce redundant parameters and improve computational speed.

[0084] Specifically, step S5 includes:

[0085] S501. Multi-scenario simulation and threshold delineation: Generate random rainfall scenarios based on historical rainfall data, simulate runoff generation and convergence processes using a distributed hydrological model, and dynamically calculate critical rainfall and water level thresholds for different regions. Simultaneously, a Seq2Seq model, using rainfall, water level, and rise rate as input, predicts disaster processes and generates dynamic thresholds.

[0086] S502. Soil-topography coupling analysis: In plain areas, soil storage and runoff generation are considered, while in mountainous areas, a direct runoff generation mechanism is introduced when the rainfall intensity exceeds the infiltration rate. The runoff inflow time is corrected by combining vegetation coverage and surface evaporation.

[0087] Specifically, step S6 includes:

[0088] S601. Real-time dynamic warning release: Integrate warning results through the GIS platform, display high-, medium-, and low-risk areas in a risk heat map, and release warning information through multiple channels such as SMS, broadcast, and apps to shorten response time.

[0089] S602, model iterative optimization: Utilize cross-validation and online learning mechanisms, combine real-time monitoring data to dynamically update model parameters, and adjust the hyperparameters of the Seq2Seq model through K-fold cross-validation to ensure generalization ability.

[0090] Specifically, step S7 includes:

[0091] S701. Multi-case verification: Select a typical hilly transition zone and compare historical disaster data with model prediction results. The evaluation accuracy F1 score is greater than 0.85 and the computational time for a single simulation is less than 10 minutes.

[0092] S702. Modular expansion design: Encapsulate the model into independent modules to support fast switching of different regional parameters such as permeability coefficient library in plain area and slope in mountain area, thus improving the universality of the method.

[0093] See also Figure 2 , a flood forecasting and early warning system for mixed hilly and plain areas, including:

[0094] Lightweight data acquisition and fusion module, adaptive cleaning module, hybrid model inference core, spatiotemporal dimension compression module, database module, hydrological forecast module, warning application module, hierarchical warning decision module, three-dimensional space warning mapping module and crowdsourcing data feedback channel;

[0095] The lightweight data acquisition and fusion module integrates data from multiple types of sensors, including water level stations, rainfall stations, and weather radars. It supports hourly or minute-level data stream access, uses a sliding window averaging method to replace complex parameter inputs, and automatically calibrates key thresholds using historical data, reducing manual intervention.

[0096] The adaptive cleaning module has a built-in dynamic filling strategy for missing values and automatically identifies outliers in combination with covariance analysis;

[0097] The spatiotemporal dimension compression module uses the DBA average sequence algorithm and DTW dynamic time warping technology to align the multi-site data in time and space to reduce data redundancy;

[0098] The database module is used to provide sufficient training samples for hydrological forecasts and provide discriminant data support for risk zoning warnings. The database module includes a historical hydrological monitoring database module, a current water conservancy project information database module, and a future hypothetical scenario database module.

[0099] The hydrological forecast module selects input parameters based on the database module and establishes regional hydrological forecasts through LSTM deep learning;

[0100] The early warning application module generates graded early warning labels based on the water level values of the control nodes predicted by the hydrological forecast module and the corresponding warning water levels for reference in decision-making and command.

[0101] For example, the warning water level at Dongshan Station in the Qinhuai River Basin is 8.8m. When the water level is forecast to be between 8.8 and 9.5m, a blue warning will be issued; when the water level is forecast to be between 9.5 and 10.5m, a yellow warning will be issued; when the water level is forecast to be between 10.5 and 11.5m, an orange warning will be issued; and when the water level is forecast to be above 1.5m, a red warning will be issued.

[0102] The hierarchical warning decision module evaluates the model reliability in real time based on the Nash coefficient, and triggers model self-checking when NSE < 0.95;

[0103] The three-dimensional spatial warning mapping module is used to combine the DEM digital elevation model to map the water level prediction value to the transition zone from hills to plains and generate a flood depth heat map.

[0104] The crowdsourcing data feedback channel is used to access flood images on social media and use CV technology to assist in verifying the accuracy of the warning area.

[0105] The historical hydrological monitoring database module includes historical rainfall and evaporation information of each rainfall station in the region, historical water level and flow information of the control station, and tide level information along the river at the regional exit.

[0106] The current real-time water conservancy project information database module includes information on the length, direction, bottom elevation, bottom width, and slope ratio of the main flood channels in the region, as well as information on the scheduling and operation rules of water conservancy projects in the region, including reservoirs, flood storage areas, polder areas, and sluice stations;

[0107] The future hypothetical scenario database module is based on the historical rainfall of each rain gauge in the region. Through frequency calculation, it obtains the rainfall and rainfall distribution at different frequencies, such as 5-year, 10-year, 20-year, 50-year, and 100-year. Through the historical Yangtze River tide data, frequency calculation is performed to obtain the tide types at different frequencies, such as 2-year, 5-year, 10-year, 20-year, and 50-year. Based on the simulation calculation of the hydrological and hydrodynamic model, the water level, flow rate, flow velocity and other information of the main control stations under different future flood and tide combination scenarios are obtained, thereby forming a future hypothetical scenario database.

[0108] The usage of the hydrological forecast module includes:

[0109] The input parameters are the current water level and flow in the area, as well as the rainfall in the previous n hours and the next n hours, where n is the forecast period;

[0110] After preparing the input data, normalize the data and divide it into training set and test set, selecting 80% of the data as the training set and 20% of the data as the test set;

[0111] Use the training data to input the LSTM model, adjust the parameters according to the model characteristics to achieve the optimal model, and use ReLU as the activation function;

[0112] After the model training is completed, the model is used to predict the test samples, and then denormalized and compared with the actual results to complete the model verification.

[0113] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solutions and improved concepts of the present invention within the technical scope disclosed by the present invention, and these changes should be covered by the scope of protection of the present invention.

Claims

1. A flood forecasting and early warning method for mixed hilly and plain areas, characterized by: include: S1, multi-source data fusion and adaptive preprocessing; S2, lightweight data preprocessing; S3, low-dimensional feature engineering; S4, lightweight dynamic model construction; S5, dynamic warning threshold generation; S6. Rapid warning and feedback optimization; S7. Verification and applicability extension.

2. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 1 is characterized in that: The step S1 comprises: S101. Low-parameter data collection: Integrate rainfall data from weather stations, river water level monitoring, soil moisture sensors, and satellite remote sensing data. Standardize data to eliminate dimensional differences and use Kriging GIS spatial interpolation technology to supplement missing values, reducing reliance on a single parameter. S102, Terrain Adaptive Modeling: Dynamically divide the grid based on the DEM digital elevation model, identify the river network and sub-basin boundaries, process mixed terrain through adaptive resolution technology, and optimize computing efficiency.

3. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 2 is characterized in that: The step S2 comprises: S201. Simplified filling of missing values: Use global constants or attribute means to fill missing values, avoiding complex regression or decision tree methods; S202. Noise smoothing simplification: using binning method local smoothing technology to replace regression or clustering denoising with large computational complexity, wherein the binning method is to smooth the stored data values by examining the data range.

4. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 3 is characterized in that: The step S3 comprises: S301, Redundant Feature Screening: Quickly identify redundant attributes using the Pearson product-distance function correlation coefficient, rather than the chi-square test or covariance analysis; S302. Attribute subset selection: retain key influencing factors and use data reduction methods to actively eliminate weakly correlated parameters.

5. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 4 is characterized in that: The data reduction method in step S302 includes: data cube aggregation, data attribute subset selection, dimension reduction, value reduction and discretization; Data cube aggregation is used for data in a data cube structure. Data cubes store multidimensional aggregation information. Each cell stores an aggregation value, corresponding to a point in the multidimensional space. Each attribute may have a conceptual hierarchy, allowing data analysis at multiple levels of abstraction. Attribute subset selection When the data set to be analyzed contains a large number of attributes, most of which are irrelevant or redundant to the mining task, attribute subset selection can detect and delete irrelevant, redundant or weakly relevant attributes or dimensions. Its goal is to find the minimum attribute set so that the probability distribution of the data class is as close as possible to the original distribution obtained using all attributes. For attribute subset selection; Dimensionality reduction uses data encoding or transformation to obtain a reduced or compressed representation of the original data, reducing the number of random variables or attributes considered. If the reduced data can only reconstruct an approximate representation of the original data, then the data reduction is lossy. If the original data can be reconstructed without losing any information, then it is lossless. Numerical reduction reduces the amount of data by choosing alternative data representations, replacing or estimating data with smaller data representations; Data discretization divides the attribute value domain into intervals to reduce the number of given continuous attribute values. The interval labels can replace the actual data values, and the numerical values of continuous attributes are replaced by a few interval labels.

6. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 5 is characterized in that: The step S4 comprises: S401, Hybrid hydrological model selection: Combine the distributed hydrological model with the simplified physical mechanism and adopt the improved Model, where f and B are the inverse of soil infiltration capacity, and f and B are used to dynamically adjust runoff calculation; S402. Machine Learning Model Optimization: Use a lightweight model such as a random forest or LSTM network to train a time series prediction model with rainfall, water level, and soil moisture as input, and reduce redundant parameters through LASSO regression feature selection.

7. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 6 is characterized in that: The step S5 comprises: S501. Multi-scenario simulation and threshold delineation: Generate random rainfall scenarios based on historical rainfall data, simulate runoff generation and convergence processes using a distributed hydrological model, and dynamically calculate critical rainfall and water level thresholds for different regions. Simultaneously, a Seq2Seq model, using rainfall, water level, and rise rate as input, predicts disaster processes and generates dynamic thresholds. S502. Soil-topography coupling analysis: In plain areas, soil storage and runoff generation are considered, while in mountainous areas, a direct runoff generation mechanism is introduced when the rainfall intensity exceeds the infiltration rate. The runoff inflow time is corrected by combining vegetation coverage and surface evaporation.

8. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 7 is characterized in that: The step S6 comprises: S601, Real-time dynamic warning release: Integrate warning results through the GIS platform, display high-, medium-, and low-risk areas in a risk heat map, and release warning information through multiple channels such as SMS, broadcast, and App; S602, model iterative optimization: Utilize cross-validation and online learning mechanisms, combine real-time monitoring data to dynamically update model parameters, and adjust the hyperparameters of the Seq2Seq model through K-fold cross-validation to ensure generalization ability.

9. The method for forecasting and warning floods in mixed hilly and plain areas according to claim 8 is characterized in that: The step S7 comprises: S701. Multi-case verification: Select a typical hilly transition zone and compare historical disaster data with model prediction results. The evaluation accuracy F1 score is greater than 0.85 and the computational time for a single simulation is less than 10 minutes. S702, modular expansion design: Encapsulate the model into independent modules to support fast switching of different regional parameters such as permeability coefficient library in plain area and slope in mountain area.

10. A forecast and early warning system for studying floods in mixed hilly and plain areas, applied to the forecast and early warning method for studying floods in mixed hilly and plain areas as described in claims 1-9, characterized in that: include: Lightweight data acquisition and fusion module, adaptive cleaning module, hybrid model inference core, spatiotemporal dimension compression module, database module, hydrological forecast module, warning application module, hierarchical warning decision module, three-dimensional space warning mapping module and crowdsourcing data feedback channel; The lightweight data acquisition and fusion module integrates data from multiple types of sensors such as water level stations, rain gauges, and weather radars, supports hourly or minute-level data stream access, uses a sliding window mean method to replace complex parameter input, and automatically calibrates key thresholds through historical data to reduce manual intervention. The adaptive cleaning module has a built-in dynamic filling strategy for missing values, and automatically identifies outliers in combination with covariance analysis. The spatiotemporal dimension compression module uses the DBA average sequence algorithm DTW dynamic time warping technology to align multi-site data in time and space to reduce data redundancy. The database module is used to provide sufficient training samples for hydrological forecasts and provide discriminant data support for risk zoning warnings. The database module includes a historical hydrological monitoring database module, current real-time water conservancy project information The hydrological forecast module selects input parameters based on the database module and establishes a regional hydrological forecast through LSTM deep learning. The early warning application module generates graded warning labels based on the water level values of the control nodes predicted by the hydrological forecast module and the corresponding warning water levels for reference in decision-making and command. The graded warning decision module evaluates the model reliability in real time based on the Nash coefficient and triggers model self-checking when NSE < 0.

95. The three-dimensional spatial warning mapping module is used to map the water level prediction value to the transition zone from hills to plains in combination with the DEM digital elevation model to generate a flood depth heat map. The crowdsourcing data feedback channel is used to access flood images on social media and use CV technology to assist in verifying the accuracy of the warning area. The historical hydrological monitoring database module includes historical rainfall and evaporation information of each rain gauge in the region, historical water level and flow information of the control station, and tide level information along the river at the regional exit. The current real-time water conservancy project information database module includes the length, direction, bottom elevation, bottom width, and slope ratio information of the main flood-carrying river channels in the region, as well as the scheduling and application rules of reservoirs, flood storage areas, embankment areas, and sluice stations in the region. The future hypothetical scenario database module is based on the historical rainfall of each rain gauge in the region, and obtains rainfall and rainfall distribution at different frequencies through frequency calculation. Through the historical Yangtze River tide data, frequency calculation is performed to obtain tide types at different frequencies. Based on the hydrological and hydrodynamic model simulation calculation, the water level, flow, flow velocity and other information of the main control stations under different future flood and tide combination scenarios are obtained, thereby forming a future hypothetical scenario database.

Images