Intelligent short- and long-term fusion flood forecasting method based on rain measurement radar

Abstract

The present application relates to the technical field of hydrological resources, and specifically includes an intelligent short-impending fusion flood forecasting method based on a rain measurement radar, which comprises: receiving rain measurement detection information, performing prediction effect feedback in combination with a short-impending flood monitoring task at an edge adaptive correction node, configuring flood identification features based on a hydrological fusion prediction model, determining a flood peak time combination and a flooding risk, and generating a graded short-impending flood warning signal. The technical problem of the initial configuration of the hydrological fusion prediction model not matching the rain measurement data characteristics and the prediction target, the prediction deviation being unable to be corrected in time, and the deviation accumulation affecting the prediction reliability is solved by using a general hydrological model deployment mode, the technical effect of adaptively deploying the hydrological fusion prediction model by binding the core detection information of the rain measurement radar and the flood prediction accuracy requirement, the cooperative mechanism of the edge adaptive correction node and the RS485 bus interface, effectively avoiding deviation accumulation, and improving the stability of runoff simulation is achieved.

Description

Technical Field

[0001] This invention relates to the field of hydrological resources technology, specifically to an intelligent short-term fusion flood forecasting method based on rainfall radar. Background Technology

[0002] With the frequent occurrence of extreme precipitation events such as short-duration heavy rainfall and typhoon rainstorms, the formation mechanism of river basin floods is becoming increasingly complex, placing higher demands on the timeliness and accuracy of short-term flood forecasting. Rainfall radar, with its advantages of wide detection range, fast data updates, and real-time capture of precipitation spatial distribution, has become the core technical support for short-term flood forecasting. However, current short-term flood forecasting based on rainfall radar adopts a generalized hydrological model architecture, which makes it difficult to quickly adjust forecast parameters when faced with sudden changes in hydrological conditions caused by extreme precipitation. This results in a large deviation in the flood forecast period, inaccurate classification of short-term flood warning signals, and difficulty in effectively supporting emergency response decisions, thus failing to meet the practical needs of river basin flood control and disaster reduction.

[0003] In summary, existing technologies suffer from technical problems such as the mismatch between the initial configuration of the hydrological fusion prediction model and the characteristics of rainfall data and prediction targets, the inability to correct prediction deviations in a timely manner, and the cumulative impact of deviations on prediction reliability. Summary of the Invention

[0004] This application provides an intelligent short-term fusion flood forecasting method based on rainfall radar, aiming to solve the technical problems in the existing technology where the deployment mode of the generalized hydrological model is adopted, the initial configuration of the hydrological fusion prediction model does not match the characteristics of rainfall data and the prediction target, the prediction deviation cannot be corrected in time, and the cumulative deviation affects the reliability of the prediction.

[0005] In view of the above problems, the technical solution to achieve the present application is as follows:

[0006] This application provides an intelligent short-term fusion flood forecasting method based on rainfall radar. The method includes: receiving rainfall detection information, including the detection frequency range, data sampling rate, and covered watershed area, based on the transceiver chip of the rainfall radar; uploading flood forecast accuracy requirements; deploying a hydrological fusion prediction model based on the rainfall detection information; driving an RS485 bus interface; providing prediction effect feedback at an edge adaptive correction node in conjunction with short-term flood monitoring tasks; and simultaneously collecting precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters; configuring flood identification features based on the hydrological fusion prediction model and combining the precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters; connecting the RS485 bus interface and the edge adaptive correction node; parsing the flood identification features; determining the peak flood time combination and inundation risk; and generating graded short-term flood warning signals.

[0007] Preferably, the flood prediction accuracy requirements include flood peak error rate, flood process fit degree, and forecast period achievement rate; based on the hydrological fusion prediction model, Class I and Class II monitoring networks under the watershed hydrological monitoring network are marked.

[0008] Preferably, the monitoring task of short-term flood is read, and the precipitation analysis network in the model architecture of the hydrological fusion prediction model is combined to determine the first prediction feature corresponding to the runoff simulation path of the main precipitation area under the marker of a type of monitoring node; wherein, the type of monitoring node includes at least the rain gauge station in the core area of ​​the main precipitation, the hydrological station at the control section of the main stream of the basin, and the water level station in the key area of ​​high slope confluence.

[0009] Preferably, based on historical flood prediction examples constrained by the flood prediction accuracy requirements, and combined with the hydrological response network in the model architecture of the hydrological fusion prediction model, the second prediction feature corresponding to the next precipitation area confluence simulation path is determined by marking the second type of monitoring nodes; wherein, the second type of monitoring nodes includes at least the rain gauge station in the precipitation impact area, the hydrological station at the confluence section of the tributary of the basin, the water level station in the plain flood detention area, and the underlying surface feature monitoring point.

[0010] Preferably, based on precipitation characteristics and combined with the physical characteristics of the flood peak error rate in the flood prediction accuracy requirements, a minimum peak deviation and an allowable error threshold range are set. The precipitation characteristics include precipitation intensity, precipitation duration, and precipitation spatial distribution. The minimum peak deviation and the allowable error threshold range are added as constraint information to the hydrological fusion prediction model.

[0011] Preferably, based on the required flood prediction accuracy, a watershed hydrological characteristic analysis is performed to obtain precipitation characteristics; based on the precipitation characteristics, combined with the first prediction characteristics and the second prediction characteristics, an adaptation mapping relationship is set; through the adaptation mapping relationship, edge adaptive correction nodes corresponding to the hydrological fusion prediction model are set.

[0012] Preferably, an initial set of prediction parameters is generated based on the minimized peak deviation and the allowable error threshold range; based on the initial set of prediction parameters, the fitness is evaluated according to the objective function corresponding to the hydrological fusion prediction model, and a fitness sequence is determined in descending order of the fitness values; the initial set of prediction parameters is iteratively optimized according to the fitness sequence.

[0013] Preferably, based on historical flood prediction examples constrained by the flood prediction accuracy requirements, a hydrological state space, a prediction action space, and a reward signal are defined; and the hydrological state space and the prediction action space are updated based on the observation feedback data from the short-term flood monitoring task.

[0014] Preferably, based on the reward signal, the prediction deviation evolution path of the short-term flood monitoring task is analyzed, and positive incentives are extracted; at the same time, a dynamic optimization loop is established through the updated hydrological state space and prediction action space.

[0015] Preferably, under extreme precipitation conditions including short-term heavy rainfall and typhoon rainstorms, weak links and parameter-sensitive intervals in flood prediction are identified; based on the weak links and parameter-sensitive intervals in flood prediction, key enhancement features are determined; the edge adaptive correction node is invoked for online reconfiguration, and the model parameters of the hydrological fusion prediction model are dynamically recalibrated in combination with the key enhancement features.

[0016] In summary, one or more technical solutions provided in this application achieve the technical effect of binding the core detection information of rainfall radar with the accuracy requirements of flood prediction, enabling the adaptive deployment of hydrological fusion prediction models, and the collaborative mechanism of edge adaptive correction nodes and RS485 bus interfaces, effectively avoiding the accumulation of deviations and improving the stability of runoff simulation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the intelligent short-term fusion flood forecasting method based on rainfall radar. Detailed Implementation

[0019] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0020] Example 1

[0021] The present application will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, this application provides an intelligent short-term fusion flood forecasting method based on rainfall radar, wherein the method includes:

[0022] S1: Receive rainfall detection information, including the detection frequency range, data sampling rate, and covered watershed area, based on the transceiver chip of the rainfall radar; S2: Upload the flood prediction accuracy requirements, deploy a hydrological fusion prediction model in conjunction with the rainfall detection information, and drive the RS485 bus interface. At the edge adaptive correction node, combine the short-term flood monitoring task to provide prediction effect feedback, and simultaneously collect precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast period deviation parameters.

[0023] Specifically, the rain-measuring radar transceiver chip is used to transmit and receive radar signals, and obtain precipitation-related detection information by analyzing the echo signals; the detection frequency band range refers to the frequency range of the radar signal, which determines the intensity and type of precipitation that the radar can detect; the data sampling rate refers to the frequency at which the radar collects data per unit time, affecting the spatiotemporal resolution of the data; the covered watershed area refers to the geographical range that the radar can detect, which determines its ability to monitor watershed precipitation; the hydrological fusion prediction model is a model that comprehensively considers rain-measuring radar data and hydrological characteristics, used to predict the occurrence and development of floods; the RS485 bus interface is a serial communication interface used to connect the edge adaptive correction node and the main control system to realize data transmission and interaction.

[0024] Edge adaptive correction nodes refer to computing nodes deployed at the edge of the monitoring network, used to process and correct the output of the prediction model in real time, and to provide feedback on the prediction effect in conjunction with short-term flood monitoring tasks. Precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters are used to evaluate the performance and reliability of the prediction model. Among them, precipitation estimation accuracy parameters reflect the accuracy of precipitation prediction; runoff simulation stability parameters measure the model's adaptability to runoff changes; and flood forecast deviation parameters represent the deviation between the predicted flood occurrence time and the actual time.

[0025] Execution steps: The rain-measuring radar receives rain-measuring detection information through its transceiver chip, including the detection frequency range, data sampling rate, and covered watershed area. This information is the basic data for flood prediction and can reflect the spatial distribution and intensity changes of precipitation in real time. Specifically, the detection frequency range of the rain-measuring radar can cover different precipitation intensities from light rain to heavy rain, the data sampling rate can reach once per minute, and the covered watershed area can reach thousands of square kilometers, thus providing high spatiotemporal resolution data support for flood prediction.

[0026] The system uploads flood forecast accuracy requirements and deploys a hydrological fusion prediction model based on rainfall and meteorological data. This model connects to edge adaptive correction nodes via an RS485 bus interface. It incorporates short-term flood monitoring tasks to provide forecast feedback and simultaneously collects parameters such as precipitation estimation accuracy, runoff simulation stability, and flood lead time deviation. Furthermore, a real-time feedback mechanism promptly identifies and corrects forecast deviations, preventing their accumulation. In a short-duration heavy rainfall event, the model's initial peak flood time may deviate from the actual time. Real-time feedback from the edge adaptive correction nodes allows for timely adjustment of model parameters, significantly improving the timeliness and reliability of the forecast. Simultaneously, the collected parameter data further optimizes the runoff simulation stability of the hydrological fusion prediction model, ensuring high forecast accuracy even under extreme precipitation conditions.

[0027] S3: Based on the hydrological fusion prediction model, and combined with the precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters, configure flood identification features; S4: Connect the RS485 bus interface and the edge adaptive correction node, analyze the flood identification features, determine the peak flood time combination and inundation risk, and generate graded short-term flood warning signals.

[0028] Specifically, flood identification features refer to the key features extracted by analyzing multi-dimensional information such as rainfall radar data, precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters, which can characterize the occurrence, development, and risk of floods. These features are used to identify the intensity, scope, and potential impact of floods and form the basis for generating flood warning signals. Flood peak time combinations refer to the time series of flood peak occurrences, including the arrival time of the flood peak, the duration of the flood peak, and the intervals between multiple flood peaks. By analyzing flood peak time combinations, the dynamic changes of floods can be predicted more accurately, providing a temporal reference for flood control decisions.

[0029] Inundation risk refers to the areas that may be inundated by floods and their risk levels. By analyzing the topography, water level changes, and flood propagation paths of the watershed, it is determined which areas may be inundated, as well as the depth and extent of inundation, thereby assessing the inundation risk. Graded short-term flood warning signals are warning signals classified according to the severity and risk level of the flood. Specifically, graded short-term flood warning signals can be divided into three levels: low risk, medium risk, and high risk. Each level corresponds to different response measures so that appropriate flood control measures can be taken based on the warning signal.

[0030] Execution steps: Based on the hydrological fusion prediction model, and combining precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters, flood identification features are configured. This transforms the single precipitation data-driven approach into a multi-source data-driven approach. By comprehensively analyzing multi-dimensional parameters, the characteristics and risks of floods are reflected more comprehensively. Furthermore, by analyzing rainfall radar data and precipitation estimation accuracy parameters, the intensity and distribution of precipitation are determined; combined with runoff simulation stability parameters, the propagation path and velocity of floods within the watershed are assessed; and the flood forecast deviation parameter can correct the accuracy of prediction time. Through the comprehensive analysis of these parameters, more accurate flood identification features are generated, thereby improving the reliability of flood prediction.

[0031] By connecting the RS485 bus interface to the edge adaptive correction node, flood identification features are analyzed to determine the peak flood time combination and inundation risk, and graded short-term flood warning signals are generated. Furthermore, through the collaborative work of the RS485 bus interface and the edge adaptive correction node, rapid data transmission and real-time processing are achieved. The flood identification features are transmitted to the edge adaptive correction node through the RS485 bus interface, and the node analyzes these features in real time to determine the peak flood time combination and inundation risk, generating graded short-term flood warning signals, which can provide more refined guidance for flood control decisions.

[0032] Furthermore, the method of this application also includes:

[0033] The flood prediction accuracy requirements include flood peak error rate, flood process fit rate, and forecast period compliance rate; based on the hydrological fusion prediction model, Class I and Class II monitoring networks under the watershed hydrological monitoring network are marked.

[0034] Specifically, flood forecast accuracy requirements refer to the specific requirements for the accuracy of flood forecast results, including peak flood error rate, flood process fit, and forecast period achievement rate, used to evaluate the performance and reliability of the forecast model. Peak flood error rate measures the error ratio between the predicted peak flow and the actual peak flow, reflecting the accuracy of the forecast model's estimation of the peak flow. Flood process fit assesses the similarity between the predicted flood process and the actual flood process, reflecting the model's ability to fit the overall dynamics of the flood. Specifically, the flood process includes the flow rate change curve over time. Forecast period achievement rate measures the degree of matching between the predicted flood occurrence time and the actual occurrence time, reflecting the model's accuracy in predicting the flood's forecast period.

[0035] A watershed hydrological monitoring network refers to various hydrological monitoring stations and equipment distributed within a watershed, used for real-time monitoring of hydrological information such as precipitation, water level, and flow. These stations and equipment are classified into Class I and Class II monitoring networks based on their importance and function. Class I monitoring networks include rain gauge stations in the main precipitation core area, hydrological stations at the control sections of the main stream, and water level stations in key high-slope confluence areas, used to monitor the main precipitation areas and key confluence paths within the watershed. Class II monitoring networks include rain gauge stations in the secondary precipitation impact area, hydrological stations at the confluence sections of tributaries, water level stations in the plain flood detention area, and monitoring points on underlying surface characteristics, used to monitor secondary precipitation areas and tributary confluence. Labeling the watershed hydrological monitoring network involves classifying and labeling various hydrological monitoring stations and equipment within the watershed to better allocate and optimize monitoring resources and improve forecast accuracy.

[0036] Execution steps: The accuracy requirements for flood forecasting are clearly defined as the flood peak error rate, flood process fit, and forecast period achievement rate. These three rates provide specific targets and constraints for optimizing the hydrological fusion forecasting model. Based on the hydrological fusion forecasting model, the watershed hydrological monitoring network is classified into Class I and Class II monitoring networks. Specifically, monitoring resources are rationally allocated according to the importance and function of the monitoring stations to improve the relevance and efficiency of forecasts. Class I monitoring networks are mainly concentrated in the core precipitation area and the control sections of the main stream of the watershed. These stations can monitor the precipitation intensity and water level changes of the main stream in the main precipitation area in real time, providing key data support for flood forecasting. By analyzing the data from these stations, the runoff simulation path and peak flow in the main precipitation area can be accurately predicted, thereby improving the flood peak error rate and flood process fit.

[0037] The second-class monitoring network is distributed in the secondary precipitation impact area and tributary confluence sections. It is used to monitor the precipitation in the secondary precipitation area and the confluence of tributaries. By analyzing the data of the second-class monitoring network, the flood prediction model can be further improved and the forecast period can be increased. The first-class monitoring network can quickly capture the precipitation intensity changes in the main precipitation area, while the second-class monitoring network can monitor the water level changes of tributaries flowing into the main stream. Through classification, labeling and optimized configuration, the accuracy and timeliness of flood prediction can be significantly improved, and the arrival time and inundation range of the flood peak can be predicted more accurately, thereby generating more reliable short-term flood warning signals.

[0038] Furthermore, for a type of monitoring network under the marker watershed hydrological monitoring network, the method of this application includes:

[0039] By reading short-term flood monitoring tasks and combining the precipitation analysis network in the model architecture of the hydrological fusion prediction model, the first prediction feature corresponding to the runoff simulation path of the main precipitation area under a type of monitoring node is determined; wherein, the type of monitoring node includes at least the rain gauge station in the core area of ​​the main precipitation area, the hydrological station at the control section of the main stream of the basin, and the water level station in the key area of ​​high slope confluence.

[0040] Specifically, short-term flood monitoring refers to real-time monitoring of short-term, imminent flood events, including high-frequency monitoring of hydrological parameters such as precipitation, water level, and flow rate to obtain real-time data on flood occurrence and development. Precipitation analysis networks are used to analyze precipitation data and extract precipitation-related features, such as precipitation intensity, precipitation duration, and spatial distribution, to provide basic input for flood forecasting. The first predictive feature refers to the key parameters that characterize the flood characteristics of the main precipitation area, extracted by analyzing the runoff simulation path of the main precipitation area, and used for subsequent flood forecasting and early warning.

[0041] One type of monitoring node refers to important monitoring stations in the watershed hydrological monitoring network used to monitor the main precipitation area and key confluence paths. These include: rain gauge stations in the core precipitation area, which monitor precipitation in the main precipitation area and are an important data source for flood forecasting; hydrological stations at the control sections of the main stream, which monitor water level and flow changes in the main stream and are key stations for assessing flood propagation paths and intensity; and water level stations in key high-slope confluence areas, which monitor confluence in high-slope areas, where the confluence velocity is fast and has a significant impact on flood formation.

[0042] Execution steps: The system reads short-term flood monitoring data and, combined with the precipitation analysis network in the hydrological fusion prediction model, determines the first predictive feature corresponding to the simulated runoff path of the main precipitation area under a class of monitoring node markers. By analyzing data from key monitoring stations in the main precipitation area, it extracts core features closely related to flood formation and development, thus providing high-precision input for flood prediction. Specifically, precipitation intensity and duration are analyzed from precipitation data obtained from rain gauges in the core precipitation area; water level and flow data from hydrological stations at the main stream control sections are used to assess the flood propagation path and velocity; and the dynamic changes in confluence are analyzed from data from water level stations in key high-slope confluence areas. This data is input into the precipitation analysis network, processed, and analyzed to extract the first predictive feature, such as peak runoff time, peak flow, and confluence velocity in the main precipitation area. This enables the hydrological fusion prediction model to more accurately simulate the formation and development of floods, significantly improving the accuracy and timeliness of flood prediction and providing more reliable support for flood control decisions.

[0043] Furthermore, for the Class II monitoring network under the watershed hydrological monitoring network, the method of this application includes:

[0044] Based on historical flood prediction examples constrained by the aforementioned flood prediction accuracy requirements, and combined with the hydrological response network in the model architecture of the hydrological fusion prediction model, the second prediction feature corresponding to the next precipitation area confluence simulation path is determined by marking two types of monitoring nodes; wherein, the two types of monitoring nodes include at least the rain gauge station in the affected area of ​​the next precipitation, the hydrological station at the confluence section of the tributary of the basin, the water level station in the plain flood detention area, and the underlying surface feature monitoring point.

[0045] Specifically, historical flood prediction examples under the requirement of flood prediction accuracy refer to examples selected from past flood prediction cases that meet the requirements, including the flood peak error rate, flood process fit, and forecast period achievement rate, and are used to analyze and optimize the current flood prediction model; the hydrological response network is used to simulate the hydrological response process of the watershed, and predicts the dynamic changes of floods by analyzing hydrological processes such as precipitation, runoff, and flow; the second prediction feature refers to the key parameters that can characterize the flood characteristics of the region by analyzing the runoff simulation path of the sub-precipitation area, which are used to supplement and improve the flood prediction model and improve the prediction accuracy.

[0046] Category II monitoring nodes refer to monitoring stations in the watershed hydrological monitoring network used to monitor secondary precipitation areas and tributary confluence paths. These include: rain gauge stations in secondary precipitation impact areas, used to monitor precipitation in these areas, where precipitation contributes relatively little to floods; hydrological stations at tributary confluence sections, used to monitor water level and flow changes when tributaries flow into the main stream, which helps assess the impact of tributaries on main stream floods; water level stations in plain flood detention areas, used to monitor water level changes in flood detention areas in plains, where water level changes significantly affect flood propagation and inundation range; and underlying surface characteristic monitoring points, used to monitor changes in soil, vegetation, and other characteristics of the watershed's underlying surface, which affect precipitation infiltration and confluence processes.

[0047] Execution steps: Based on historical flood prediction examples constrained by flood prediction accuracy requirements, and combined with the hydrological response network in the hydrological fusion prediction model, the second prediction feature corresponding to the sub-precipitation area confluence simulation path marked by the second-class monitoring nodes is determined. By analyzing the data of the sub-precipitation area and tributary confluence paths, the flood prediction model is supplemented and improved, enhancing the comprehensiveness and accuracy of the prediction. The second prediction feature provides the flood prediction model with key information on the sub-precipitation area and tributary confluence paths, enabling the model to more comprehensively simulate the formation and development process of floods. Preferably, the data from the second-class monitoring nodes supplements the information of the main precipitation area, allowing the flood prediction model to more comprehensively consider the hydrological dynamics within the basin, thereby significantly improving the accuracy and reliability of flood prediction.

[0048] Furthermore, to meet the accuracy requirements of flood forecasting, the method in this application also includes:

[0049] Based on precipitation characteristics and the physical properties of the flood peak error rate required for flood prediction accuracy, a minimum peak deviation and an allowable error threshold range are set. The precipitation characteristics include precipitation intensity, precipitation duration, and precipitation spatial distribution. The minimum peak deviation and the allowable error threshold range are added as constraint information to the hydrological fusion prediction model.

[0050] Specifically, precipitation characteristics refer to the physical properties of precipitation events, including precipitation intensity, precipitation duration, and spatial distribution. These are important input parameters for flood forecasting and directly affect the formation and development of floods. Precipitation intensity refers to the amount of precipitation per unit time, usually expressed in millimeters per hour. Precipitation duration refers to the length of time precipitation lasts, usually expressed in hours. Spatial distribution of precipitation refers to the distribution of precipitation in geographical space, reflecting the intensity differences of precipitation in different regions. The physical characteristics of the flood peak error rate refer to the error range of the flood peak flow in flood forecasting. The flood peak error rate reflects the degree of deviation between the predicted flood peak and the actual flood peak and is one of the key indicators for measuring the accuracy of flood forecasting.

[0051] Minimizing peak deviation refers to optimizing model parameters to minimize the deviation between the predicted and actual flood peaks. The allowable error threshold range refers to the maximum range of flood peak error that is permitted in flood prediction, usually set according to actual flood control needs. For example, the allowable error threshold range can be set to ±10%. Constraint information refers to the conditions used to limit the range of model parameter values ​​during model optimization. Minimizing peak deviation and the allowable error threshold range are added as constraint information to the hydrological fusion prediction model to ensure that the accuracy of the model prediction meets the requirements.

[0052] Execution steps: Based on precipitation characteristics including precipitation intensity, duration, and spatial distribution, and combined with the physical characteristics of the flood peak error rate required for flood prediction accuracy, a minimum peak deviation and an allowable error threshold range are set. By quantifying precipitation characteristics and prediction accuracy requirements, clear objectives and constraints are provided for the optimization of the hydrological fusion prediction model. The minimum peak deviation and the allowable error threshold range are added as constraints to the hydrological fusion prediction model. The model parameters are optimized through these constraints to ensure that the predicted flood peak flow is within the allowable error range. Specifically, by adjusting parameters in the model, including the runoff coefficient and soil permeability, the deviation between the predicted flood peak flow and the actual flood peak flow is minimized. At the same time, by setting the allowable error threshold range, the prediction accuracy and reliability of the hydrological fusion prediction model under extreme precipitation conditions are ensured.

[0053] Furthermore, based on precipitation characteristics and considering the physical properties of the flood peak error rate in the required flood forecast accuracy, the method of this application also includes:

[0054] Based on the required flood prediction accuracy, watershed hydrological characteristics are analyzed to obtain precipitation characteristics; based on the precipitation characteristics, combined with the first prediction characteristics and the second prediction characteristics, an adaptation mapping relationship is set; through the adaptation mapping relationship, edge adaptive correction nodes corresponding to the hydrological fusion prediction model are set.

[0055] Specifically, watershed hydrological characteristic analysis refers to the systematic analysis of hydrological characteristics within a watershed, including factors such as precipitation, runoff, runoff, soil moisture content, and topography, to understand the watershed's hydrological response mechanism and provide basic data for flood forecasting. Adaptive mapping relationships refer to the correspondence established between precipitation characteristics and first and second prediction characteristics, used to adjust and optimize the parameters of the hydrological fusion prediction model so that it can better adapt to different precipitation conditions and watershed characteristics. Edge adaptive correction nodes refer to computing nodes deployed at the network edge, used to correct and optimize the output of the prediction model in real time, dynamically adjusting model parameters through adaptive mapping relationships to improve prediction accuracy.

[0056] Execution steps: Based on the accuracy requirements of flood prediction, watershed hydrological characteristics are analyzed to obtain precipitation characteristics. This involves analyzing hydrological data within the watershed to extract precipitation-related features, such as precipitation intensity, duration, and spatial distribution. Based on the obtained precipitation characteristics, combined with first and second prediction features, an adaptation mapping relationship is established. By establishing the correspondence between precipitation characteristics and prediction features, model parameters are optimized to better adapt to different precipitation conditions and watershed characteristics. Edge adaptive correction nodes are set corresponding to the hydrological fusion prediction model through the adaptation mapping relationship. The model output is corrected in real time through edge computing nodes to ensure the accuracy and timeliness of the prediction results. Specifically, the edge adaptive correction nodes can dynamically adjust model parameters, such as runoff coefficient and soil permeability, based on real-time monitoring data to adapt to actual hydrological conditions. Through this adaptation mapping relationship and edge adaptive correction mechanism, model parameters can be dynamically adjusted, and prediction results can be corrected in real time, significantly improving the accuracy and reliability of flood prediction.

[0057] Furthermore, by incorporating the minimized peak deviation and the allowable error threshold range as constraint information into the hydrological fusion prediction model, the method of this application also includes:

[0058] Based on the minimized peak deviation and the allowable error threshold range, an initial prediction parameter solution set is generated; based on the initial prediction parameter solution set, the fitness is evaluated according to the objective function corresponding to the hydrological fusion prediction model, and the fitness sequence is determined in descending order of the fitness values; the initial prediction parameter solution set is iteratively optimized according to the fitness sequence.

[0059] Specifically, the initial prediction parameter set refers to a set of initial parameter values ​​generated during model optimization based on minimizing the peak deviation and the allowable error threshold range. These parameter values ​​are the starting point for model optimization and are used for subsequent fitness evaluation and iterative optimization. The objective function is used to evaluate the fitness of the initial prediction parameter set, that is, the performance of these parameters in flood prediction, usually based on indicators such as prediction accuracy and error rate. Fitness refers to the degree of performance of the parameter set under the objective function. The higher the fitness, the higher the accuracy and the smaller the error of the parameter set in flood prediction. The fitness sequence is the sequence of all initial prediction parameter sets sorted from high to low fitness, used to guide the iterative optimization process, prioritizing the optimization of parameter sets with higher fitness. Iterative optimization refers to gradually improving the prediction accuracy of the model by adjusting and optimizing parameters multiple times. In each iteration, the parameter set with better performance is selected for adjustment based on the fitness sequence to find the optimal parameter combination.

[0060] Execution steps: First, generate an initial set of predicted parameters based on minimizing the peak deviation and the allowable error threshold range. This provides a reasonable set of starting parameter values ​​for the optimization of the hydrological fusion prediction model, including key parameters such as runoff coefficient and soil permeability, ensuring that the optimization process starts from a point close to the optimal solution. Second, based on the initial set of predicted parameters, evaluate the fitness according to the objective function corresponding to the hydrological fusion prediction model. The objective function is usually based on indicators such as prediction accuracy and error rate. Calculate the fitness value of each initial set of predicted parameters using the objective function, and determine the fitness sequence according to the fitness values ​​from largest to smallest. Furthermore, prioritize the parameter set with higher fitness for optimization by sorting.

[0061] The initial prediction parameter set is iteratively optimized according to the fitness sequence. In each iteration, the parameter set with higher fitness is selected for adjustment, gradually improving the model's prediction accuracy. Specifically, in the first iteration, the parameter set with the highest fitness is selected for adjustment, and parameters such as the runoff coefficient and soil permeability are fine-tuned to reduce the flood peak error rate and increase fitness. In subsequent iterations, other parameter sets are further optimized to gradually improve the overall prediction accuracy. This systematic optimization of the hydrological fusion prediction model significantly improves the model's prediction accuracy and reliability, especially in extreme precipitation events, effectively addressing rapid changes in hydrological conditions and ensuring the model's prediction performance under different conditions.

[0062] Furthermore, the method of this application also includes:

[0063] Based on historical flood prediction examples constrained by the required accuracy of flood prediction, a hydrological state space, a prediction action space, and a reward signal are defined; the hydrological state space and the prediction action space are updated based on the observation feedback data from the short-term flood monitoring task.

[0064] Specifically, the hydrological state space refers to the set of all possible variables describing the state of the watershed hydrological system, including precipitation intensity, water level, flow rate, soil moisture content, and runoff path, used to characterize the hydrological features of the watershed at a certain moment; the prediction action space refers to the set of all possible prediction actions that the model can take, including adjusting model parameters, selecting different prediction models, and changing the prediction time step, used to optimize prediction results; the reward signal refers to the signal used in reinforcement learning to evaluate the effect of the model's prediction actions. The reward signal is usually a numerical value, representing the positive or negative impact of the prediction action on model performance. In flood prediction, the reward signal can be an improvement in prediction accuracy, a reduction in error, etc.; historical flood prediction instances refer to records of past flood predictions, including actual observation data, model prediction results, and prediction accuracy evaluation indicators, such as flood peak error rate, flood process fit, and forecast period achievement rate, used to train and optimize the model; the observation feedback data of the short-term flood monitoring task refers to the data acquired in real time during the short-term near-term flood monitoring process, including monitoring information such as precipitation, water level, and flow rate, used to dynamically update the model's state space and action space to adapt to real-time changing hydrological conditions.

[0065] Execution steps: Based on historical flood prediction examples constrained by flood prediction accuracy requirements, define the hydrological state space, prediction action space, and reward signal to provide a basic framework for the dynamic optimization of the hydrological fusion prediction model. This ensures that the hydrological fusion prediction model can learn the optimal prediction strategy based on historical data. Specifically, the prediction action space includes adjusting parameters such as the confluence coefficient and soil permeability; the reward signal is a positive incentive and can be defined as an improvement in prediction accuracy.

[0066] Based on the observation feedback data from the short-term flood monitoring mission, the hydrological state space and prediction action space are updated. The model's state and action are dynamically adjusted through real-time monitoring data to ensure that the model can adapt to real-time changes in hydrological conditions and improve prediction accuracy. Specifically, based on this real-time data, the hydrological state space is updated by incorporating the current precipitation intensity and water level change rate into the state variables. At the same time, the prediction action space is adjusted based on the model's real-time prediction error to further reduce the flood peak error rate. Ideally, this model effectively responds to rapid changes in hydrological conditions during extreme precipitation events, dynamically optimizes the hydrological fusion prediction model, significantly improves the model's prediction accuracy, and ensures that it maintains high reliability throughout the real-time monitoring and prediction process.

[0067] Furthermore, the method of this application also includes:

[0068] Based on the reward signal, the evolution path of the prediction deviation of the short-term flood monitoring task is analyzed, and positive incentives are extracted; at the same time, a dynamic optimization loop is established through the updated hydrological state space and prediction action space.

[0069] Specifically, the prediction bias evolution path refers to the trajectory of the deviation between the prediction results and the actual observation data over time in short-term flood monitoring tasks. By analyzing the prediction bias evolution path, we can understand how the prediction bias is generated and changes, thus providing a basis for model optimization. Positive incentives refer to the factors or events in the prediction bias evolution path that lead to improved prediction accuracy or reduced bias. Positive incentives can serve as reward signals to reinforce the learning process and guide model optimization. Dynamic optimization loop refers to the process of establishing a continuously optimized model based on the updated hydrological state space and prediction action space. In the optimization loop, the hydrological fusion prediction model continuously adjusts its parameters according to the new state and actions to achieve a gradual improvement in prediction accuracy.

[0070] Execution steps: Based on the reward signal analysis, the evolution path of the prediction bias in the short-term flood monitoring task is analyzed to extract positive incentives. By analyzing the changing trend of the prediction bias, factors that can effectively improve prediction accuracy are identified, thus providing direction for model optimization. Through the updated hydrological state space and prediction action space, a dynamic optimization loop is established. Using real-time monitoring data and historical experience, the state and action of the model are dynamically adjusted to form a continuously optimized closed-loop system. The updated hydrological state space may include the latest information such as precipitation intensity and water level change rate; the prediction action space may include adjusting parameters such as runoff coefficient and soil permeability.

[0071] Based on this updated information, the effectiveness of the current action is evaluated according to the reward signal after each prediction, and parameters are adjusted according to positive incentives. Specifically, if adjusting the confluence coefficient can reduce prediction bias, this parameter will continue to be optimized in subsequent predictions. At the same time, other relevant parameters such as soil permeability will be adjusted to further improve prediction accuracy. Preferably, the mechanism based on reward signals and dynamic optimization cycles can effectively cope with rapid changes in hydrological conditions and provide more accurate decision support for flood control and disaster reduction.

[0072] Furthermore, the method of this application also includes:

[0073] Under extreme precipitation conditions, including short-duration heavy rainfall and typhoon rainstorms, weak links and parameter-sensitive intervals in flood prediction are identified; based on the weak links and parameter-sensitive intervals, key enhancement features are determined; the edge adaptive correction node is invoked for online reconfiguration, and the model parameters of the hydrological fusion prediction model are dynamically recalibrated in combination with the key enhancement features.

[0074] Specifically, extreme precipitation conditions refer to extreme weather conditions such as short-duration heavy rainfall and typhoon rainstorms, which typically result in high precipitation intensity, short duration, and uneven spatial distribution, placing higher demands on the accuracy and reliability of flood prediction models. Weak links in flood prediction refer to the parts of the flood prediction model that perform poorly under extreme precipitation conditions, such as insufficient prediction accuracy, slow response speed, and poor parameter adaptability, which may lead to increased prediction bias and affect the accuracy of flood control decisions. Parameter-sensitive intervals refer to the ranges where changes in model parameters significantly affect the prediction results. Under extreme precipitation conditions, parameters such as runoff coefficient and soil permeability are more sensitive to the prediction results and require special attention and adjustment. Enhancing key features refers to identifying features or parameters that can significantly improve prediction accuracy after identifying weak links and sensitive intervals, such as the relationship between precipitation intensity and peak flow, and changes in runoff paths. Online reconfiguration refers to the process of dynamically adjusting model parameters or structure based on real-time data during model operation. Through online reconfiguration, the model can quickly adapt to hydrological changes under extreme precipitation conditions.

[0075] Execution steps: Under extreme precipitation conditions, including short-duration heavy rainfall and typhoon rainstorms, identify weak links and parameter sensitive ranges in flood prediction. Further, by analyzing the prediction results under extreme precipitation conditions, identify the shortcomings of the hydrological fusion prediction model and the sensitive ranges of key parameters, providing a basis for subsequent optimization. Specifically, by analyzing the evolution path of prediction bias, if it is found that the hydrological fusion prediction model performs poorly in handling high precipitation intensity and rapid runoff, the runoff coefficient and soil permeability are identified as sensitive parameters. Changes in these parameters under extreme precipitation conditions significantly affect the prediction results and are the weak links in the model.

[0076] Based on the identified weak links and parameter-sensitive intervals in flood forecasting, key enhancement features are determined. By analyzing these weak links and sensitive intervals, key features that can significantly improve forecast accuracy are extracted. Analysis reveals that key enhancement features include the nonlinear relationship between precipitation intensity and peak flow, and changes in confluence paths. These features better reflect the hydrological response mechanism under extreme precipitation conditions. Online reconfiguration of edge adaptive correction nodes is performed, and the model parameters of the hydrological fusion forecasting model are dynamically recalibrated in conjunction with the enhanced key features. Furthermore, the model parameters are adjusted in real time through edge computing nodes to ensure the forecast accuracy and reliability of the model under extreme precipitation conditions. The optimal mechanism, based on weak link identification, key feature extraction, and online reconfiguration, effectively addresses hydrological changes under extreme precipitation conditions, dynamically adjusts the model parameters of the hydrological fusion forecasting model, significantly improves the accuracy and timeliness of short-term flood forecasting, provides more reliable decision support for flood control and disaster reduction, and effectively reduces the risk of flood disasters.

[0077] In summary, the beneficial effects of the embodiments of this application are:

[0078] This application provides an intelligent short-term fusion flood forecasting method based on a rainfall radar transceiver chip. It achieves the technical effect of binding rainfall radar core detection information with flood forecast accuracy requirements, adaptively deploying a hydrological fusion forecasting model, and driving the RS485 bus interface. At the edge adaptive correction node, it combines short-term flood monitoring tasks to provide prediction effect feedback, simultaneously collecting precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters. Based on the hydrological fusion forecasting model, and combining these parameters, flood identification features are configured. The RS485 bus interface and the edge adaptive correction node are connected to analyze the flood identification features, determine the peak flood time combination and inundation risk, and generate graded short-term flood warning signals.

[0079] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0080] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart short-term fusion flood forecasting method based on rainfall radar, characterized in that, The method includes: Based on the transceiver chip of the rain-measuring radar, it receives rain-measuring detection information including the detection frequency band range, data sampling rate, and coverage area of ​​the watershed; Upload the flood forecast accuracy requirements, deploy the hydrological fusion forecast model in conjunction with the rainfall detection information, and drive the RS485 bus interface. At the edge adaptive correction node, combine the short-term flood monitoring task to provide forecast effect feedback, and simultaneously collect precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast period deviation parameters. Based on the hydrological fusion prediction model, and combined with the precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters, flood identification features are configured. Connect the RS485 bus interface and the edge adaptive correction node, analyze the flood identification features, determine the peak flood time combination and inundation risk, and generate graded short-term flood warning signals. The required accuracy of flood forecasting includes the flood peak error rate, the fit of the flood process, and the rate of achievement of the forecast period. Based on the aforementioned hydrological fusion prediction model, Class I and Class II monitoring networks are marked under the watershed hydrological monitoring network. The method further includes uploading flood forecast accuracy requirements, and also includes: Based on precipitation characteristics and the physical properties of the flood peak error rate required for flood prediction accuracy, a minimum peak deviation and an allowable error threshold range are set. The precipitation characteristics include precipitation intensity, precipitation duration, and precipitation spatial distribution. The minimum peak deviation and the allowable error threshold range are used as constraint information and added to the hydrological fusion prediction model; The method further includes: Identify weak links and parameter-sensitive ranges in flood forecasting under extreme precipitation conditions, including short-term heavy rainfall and typhoon rainstorms; Based on the aforementioned weak links and parameter sensitive ranges in flood prediction, key enhancement features were identified. The edge adaptive correction node is invoked for online reconfiguration, and the model parameters of the hydrological fusion prediction model are dynamically recalibrated in combination with the enhanced key features. Edge adaptive correction nodes refer to computing nodes deployed at the edge of the monitoring network, used to process and correct the output of the prediction model in real time; Flood identification features refer to the key features extracted by analyzing multi-dimensional information such as rainfall radar data, precipitation estimation accuracy parameters, runoff simulation stability parameters, and flood forecast deviation parameters. These features can characterize the occurrence, development, and risk of floods and are used to identify the intensity, extent, and potential impact of floods. They are the basis for generating flood warning signals. By analyzing rainfall radar data and precipitation estimation accuracy parameters, the intensity and distribution of precipitation are determined; combined with runoff simulation stability parameters, the propagation path and velocity of floods within the watershed are assessed; flood forecast deviation parameters can correct the accuracy of prediction time; through comprehensive analysis of these parameters, more accurate flood identification features are generated, thereby improving the reliability of flood prediction.

2. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 1, characterized in that, A method for marking a type of monitoring network within a watershed hydrological monitoring network includes: By reading short-term flood monitoring tasks and combining the precipitation analysis network in the model architecture of the hydrological fusion prediction model, the first prediction feature corresponding to the simulated runoff path of the main precipitation area under a certain type of monitoring node marking is determined. Among them, the first type of monitoring nodes includes at least rain gauge stations in the core precipitation area, hydrological stations at the control section of the main stream of the basin, and water level stations in the key area of ​​high-slope confluence.

3. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 2, characterized in that, The method for marking Class II monitoring groups within a watershed hydrological monitoring network includes: Based on historical flood prediction examples under the aforementioned flood prediction accuracy requirements, and combined with the hydrological response network in the model architecture of the hydrological fusion prediction model, the second prediction feature corresponding to the next precipitation area confluence simulation path marked by the second type of monitoring node is determined. Among them, the second type of monitoring nodes includes at least rain gauge stations in the secondary precipitation impact area, hydrological stations at the confluence sections of tributaries in the basin, water level stations in the plain flood detention area, and monitoring points for underlying surface characteristics.

4. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 3, characterized in that, Based on precipitation characteristics and considering the physical properties of the flood peak error rate required for flood forecast accuracy, the method further includes: Based on the aforementioned flood forecasting accuracy requirements, watershed hydrological characteristics are analyzed to obtain precipitation features; Based on the precipitation characteristics, and in combination with the first prediction characteristics and the second prediction characteristics, an adaptation mapping relationship is set. The edge adaptive correction node corresponding to the hydrological fusion prediction model is set through the adaptation mapping relationship.

5. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 3, characterized in that, The method further includes adding the minimized peak deviation and the allowable error threshold range as constraint information to the hydrological fusion prediction model: Based on the minimized peak deviation and the allowable error threshold range, an initial set of prediction parameters is generated; Based on the initial set of prediction parameters, the fitness is evaluated according to the objective function corresponding to the hydrological fusion prediction model, and the fitness sequence is determined in descending order of the fitness values. The initial set of prediction parameters is iteratively optimized according to the fitness sequence.

6. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 3, characterized in that, The method further includes: Based on historical flood prediction examples constrained by the aforementioned flood prediction accuracy requirements, the hydrological state space, prediction action space, and reward signal are defined. Based on the observation feedback data from the short-term flood monitoring mission, the hydrological state space and prediction action space are updated.

7. The intelligent short-term fusion flood forecasting method based on rainfall radar as described in claim 6, characterized in that, The method further includes: Based on the reward signal, the evolution path of the prediction deviation of the short-term flood monitoring task is analyzed, and positive incentives are extracted. At the same time, a dynamic optimization loop is established through the updated hydrological state space and prediction action space.

Images