A multi-dimensional hydrological risk assessment method, device and computer equipment

Abstract

The application relates to a multi-dimensional hydrological risk assessment method, device and computer equipment. The method comprises the following steps: acquiring hydrological real-time data and hydrological historical data of a target water area, and inputting the data into a hydrological risk machine model and a hydrological risk physical model respectively to obtain prediction data of the two models. If the difference between the prediction data of the two models exceeds a preset value, the model parameters are adjusted, the data is input again and prediction is performed until the prediction difference is less than a set threshold. Then, the model result with smaller difference is selected as the water area risk prediction data, and the prediction data is subjected to availability checking. If the data is reliable, the data is used as target risk prediction data for subsequent application. The method can effectively improve the accuracy and timeliness of the prediction result of the hydrological risk, and provides more accurate and real-time data support for water area management, disaster warning and decision making.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a multi-dimensional hydrological risk assessment method, apparatus, and computer equipment. Background Technology

[0002] Traditional hydrological risk assessment techniques typically utilize remote sensing to acquire surface change data, combined with weather forecasts and historical precipitation data. Hydrological models can simulate processes such as precipitation, evaporation, and soil infiltration to predict water flow dynamics within a watershed. Furthermore, by comparing hydrological data under different scenarios, potential flood risks, drought risks, and other hydrological hazards are assessed, providing hydrological risk assessment data. However, traditional techniques often rely on limited observational data and model assumptions, making it difficult to cope with complex hydrological changes and extreme weather events, resulting in poor accuracy and timeliness of hydrological risk predictions. Summary of the Invention

[0003] Therefore, it is necessary to provide a multi-dimensional hydrological risk assessment method, device, and computer equipment that can effectively improve the accuracy and timeliness of hydrological risk prediction results, addressing the aforementioned technical problems.

[0004] Firstly, this application provides a multi-dimensional hydrological risk assessment method, including:

[0005] Acquire real-time and historical hydrological data for the target water area;

[0006] The real-time hydrological data and the historical hydrological data are respectively input into the hydrological risk machine model and the hydrological risk physical model to obtain the risk machine model prediction data and the risk physical model prediction data.

[0007] If the difference between the predicted data from the risk machine model and the predicted data from the risk physical model is greater than a preset difference, the model parameters of the hydrological risk machine model and the hydrological risk physical model are adjusted according to the difference.

[0008] The parameter-tuned hydrological risk machine model is used as the hydrological risk machine model, and the parameter-tuned hydrological risk physical model is used as the hydrological risk physical model. Then, the process returns to the step of inputting the real-time hydrological data and the historical hydrological data into the hydrological risk machine model and the hydrological risk physical model respectively to obtain the risk machine model prediction data and the risk physical model prediction data.

[0009] Until the difference between the risk machine model prediction data and the risk physical model prediction data is less than the preset difference, the risk machine model prediction data or the risk physical model prediction data shall be selected as the water area risk prediction data.

[0010] The water area risk prediction data is subjected to an availability check. If the availability check result indicates that the water area risk prediction data is reliable, the water area risk prediction data is used as the target risk prediction data.

[0011] Secondly, this application also provides a multi-dimensional hydrological risk assessment device, including:

[0012] The hydrological data acquisition module is used to acquire real-time and historical hydrological data of the target water area;

[0013] The hydrological data analysis module is used to input the real-time hydrological data and the historical hydrological data into the hydrological risk machine model and the hydrological risk physical model, respectively, to obtain the risk machine model prediction data and the risk physical model prediction data.

[0014] The hydrological model adjustment module is used to adjust the model parameters of the hydrological risk machine model and the hydrological risk physical model according to the difference when the difference between the predicted data of the risk machine model and the predicted data of the risk physical model is greater than a preset difference.

[0015] The hydrological data analysis module is also used to use the parameter-adjusted hydrological risk machine model as the hydrological risk machine model and the parameter-adjusted hydrological risk physical model as the hydrological risk physical model, and return to execute the step of inputting the real-time hydrological data and the historical hydrological data into the hydrological risk machine model and the hydrological risk physical model respectively to obtain the risk machine model prediction data and the risk physical model prediction data.

[0016] The prediction data selection module is used to select either the risk machine model prediction data or the risk physical model prediction data as water area risk prediction data until the difference between the risk machine model prediction data and the risk physical model prediction data is less than the preset difference.

[0017] The risk data determination module is used to perform an availability check on the water area risk prediction data. If the availability check result indicates that the water area risk prediction data is reliable, the water area risk prediction data is used as the target risk prediction data.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of a multi-dimensional hydrological risk assessment.

[0019] The aforementioned multi-dimensional hydrological risk assessment method, apparatus, and computer equipment input real-time and historical hydrological data of the target water area into a hydrological risk machine model and a hydrological risk physical model. By comparing the prediction results of the two models, when the difference exceeds a preset threshold, the physical parameters in the model are automatically adjusted to reduce model error. After multiple adjustments and iterations, the most reliable risk prediction data is ultimately selected, ensuring the effectiveness of the prediction results in practical applications. Further usability checks on the water area risk prediction data enhance the reliability of the model output, avoiding decision-making errors caused by model errors or data inconsistencies. By integrating the advantages of machine learning and physical modeling, the scientific rigor and practicality of water area risk prediction are improved, thus effectively enhancing the accuracy and timeliness of hydrological risk prediction results and providing more accurate and real-time data support for water area management, disaster early warning, and decision-making. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a diagram illustrating the application environment of a multi-dimensional hydrological risk assessment method in one embodiment.

[0022] Figure 2 This is a flowchart illustrating a multi-dimensional hydrological risk assessment method in one embodiment;

[0023] Figure 3 This is a flowchart illustrating a method for obtaining risk physical model prediction data in one embodiment;

[0024] Figure 4 This is a flowchart illustrating a method for obtaining target risk prediction data in one embodiment;

[0025] Figure 5 This is a flowchart illustrating a method for obtaining the credibility analysis results of risk data in one embodiment.

[0026] Figure 6 This is a flowchart illustrating the method for obtaining the credibility analysis results of risk data in another embodiment;

[0027] Figure 7 This is a structural block diagram of a multi-dimensional hydrological risk assessment device in one embodiment.

[0028] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0030] This application provides a multi-dimensional hydrological risk assessment method that can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0031] In one exemplary embodiment, such as Figure 2 As shown, a multi-dimensional hydrological risk assessment method is provided, which is then applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 212. Wherein:

[0032] Step 202: Obtain real-time hydrological data and historical hydrological data for the target water area.

[0033] The target water area can be a specific body of water or area for hydrological risk prediction. This could be a river, lake, watershed, or reservoir, etc., and it is usually necessary to monitor the hydrological conditions of the area in order to predict and manage potential floods, droughts, or other hydrological risks.

[0034] Among them, real-time hydrological data can be data on various hydrological parameters (such as water level, flow velocity, precipitation, temperature, etc.) of the target water area collected in real time at a specific point in time or within a time period.

[0035] Among them, historical hydrological data can be hydrological data related to the target water area that has been recorded and stored over a period of time, such as historical water level, flow rate, and precipitation.

[0036] Specifically, real-time hydrological data for the target water area is monitored in real time using sensors, remote sensing equipment, weather stations, and other devices installed around the water area to track variables such as water level, flow velocity, precipitation, temperature, and wind speed. Historical hydrological data for the target water area comes from meteorological bureaus, hydrological monitoring stations, and other institutions, and typically includes records of hydrological events over many years, such as historical precipitation, basin flood events, drought conditions, historical water levels, historical flow rates, and historical precipitation.

[0037] Step 204: Input the real-time hydrological data and historical hydrological data into the hydrological risk machine model and the hydrological risk physical model, respectively, to obtain the risk machine model prediction data and the risk physical model prediction data.

[0038] Among them, the hydrological risk machine model can be a model based on machine learning or data mining techniques, used to analyze and predict the hydrological risks of water areas, obtained by training the model to identify patterns and regularities in historical data.

[0039] Among them, the physical model of hydrological risk can be a model that simulates the hydrological processes of a body of water using physical equations and laws (such as flow equations, precipitation-runoff models, etc.) based on hydrological and meteorological principles. It simulates physical processes such as water flow, precipitation, and evaporation in a body of water through mathematical modeling and numerical calculations, thereby assessing hydrological risk. Unlike machine models, physical models focus on the physical mechanisms and laws of the real world.

[0040] The risk machine model prediction data can be the prediction results obtained through a hydrological risk machine learning model. This data typically includes estimates of hydrological variables (such as water level, flow, etc.) over a future period, as well as risk assessment results (such as flood risk, drought risk, etc.) calculated based on these prediction data.

[0041] Among them, the risk physics model prediction data can be the prediction results obtained through hydrological risk physics models. These data calculate the hydrological state of the water body (such as water level, flow velocity, etc.) and its changing trends by simulating hydrological processes and physical laws, and conduct risk assessments based on these prediction results.

[0042] Specifically, real-time and historical hydrological data need to be input into two different types of models. The hydrological risk machine model (such as support vector machines, decision trees, random forests, and neural networks) learns patterns from historical data and predicts hydrological risk based on both real-time and historical data, particularly for complex nonlinear relationships. The hydrological risk physical model, on the other hand, is based on fundamental principles of hydrology and meteorology. It uses hydrodynamic equations (such as precipitation-runoff models and surface flow models) to simulate the physical changes in water bodies, based on both real-time and historical hydrological data. These two types of models predict hydrological risk from different perspectives, resulting in prediction data from the risk machine model and the risk physical model.

[0043] Step 206: If the difference between the predicted data from the risk machine model and the predicted data from the risk physics model is greater than the preset difference, adjust the model parameters of the hydrological risk machine model and the hydrological risk physics model according to the difference.

[0044] The difference can be the deviation or difference between the prediction data of the hydrological risk machine model and the prediction data of the hydrological risk physical model. The difference is used to measure the similarity or consistency of the outputs of the two models.

[0045] The preset difference can be a tolerance threshold set in advance during the model adjustment process. It is used to measure the magnitude of the difference between the prediction results of the machine learning model and the physical model. If the difference between the prediction results of the two models exceeds this threshold, it means that the model predictions are inconsistent and further adjustment of the model parameters is required; if the difference is within the threshold range, the predictions of the two models are considered to be sufficiently consistent.

[0046] Specifically, if the difference between the predictions from the risk machine model and the risk physics model exceeds a preset tolerance threshold, it indicates that the prediction accuracy of one or both models needs improvement. In this case, the model parameters need to be adjusted based on the difference in prediction results to narrow the prediction discrepancy between the two models. For the hydrological risk machine model, the model's fitting ability is optimized by adjusting its hyperparameters (such as learning rate, number of hidden layer nodes, regularization terms, etc.) to reduce overfitting or underfitting. For the hydrological risk physics model, the focus of adjustment is on the parameters involved in the physical equations, such as the friction coefficient of water flow and the precipitation-runoff conversion coefficient, to more accurately simulate actual hydrological processes. Since there may be conflicting adjustments during the adjustment of the two models, optimization algorithms such as particle swarm optimization (PSO), genetic algorithm (GA), or Bayesian optimization are used to automatically explore the parameter space and find parameter combinations that can improve the prediction results of both models simultaneously. The optimization process is usually iterative. By repeatedly adjusting parameters and verifying outputs, the difference between the machine learning model and the physical model is gradually reduced, resulting in the adjusted hydrological risk machine model and the adjusted hydrological risk physical model.

[0047] Step 208: Using the adjusted hydrological risk machine model as the hydrological risk machine model and the adjusted hydrological risk physical model as the hydrological risk physical model, return to the execution step of inputting real-time hydrological data and historical hydrological data into the hydrological risk machine model and the hydrological risk physical model respectively to obtain the risk machine model prediction data and the risk physical model prediction data.

[0048] Specifically, after adjusting the parameters of both the hydrological risk machine model and the hydrological risk physical model, the adjusted hydrological risk machine model is used as the hydrological risk machine model, and the adjusted hydrological risk physical model is used as the hydrological risk physical model. Real-time and historical hydrological data are then input into both models for new predictions. At this point, the deviation between the output data of the two new models should be reduced. If the difference between the predicted data from the risk machine model and the risk physical model is still greater than a preset threshold, it indicates that one or both models need further optimization, and therefore, the model parameters need to be adjusted continuously. Through this iterative process of continuously optimizing the model parameters, until the difference between the predicted data from the risk machine model and the risk physical model is less than a preset difference, it means that the two prediction models in different directions can output the same prediction results, reducing the possibility of model error.

[0049] Step 210: Select either the risk machine model prediction data or the risk physical model prediction data as the water area risk prediction data until the difference between the risk machine model prediction data and the risk physical model prediction data is less than the preset difference.

[0050] Among them, the water area risk prediction data can be selected from the water area risk assessment results obtained from the hydrological risk machine model and the hydrological risk physical model.

[0051] Specifically, when the difference between the prediction data from the risk machine model and the risk physics model is less than a preset difference—that is, when the difference between the prediction results of the hydrological risk machine model and the hydrological risk physics model is within a tolerable range—it is necessary to select either the prediction data from the risk machine model or the risk physics model as the water area risk prediction data. At this point, the choice between the prediction results from the hydrological risk machine model and the hydrological risk physics model can be made based on the actual situation. For example, if a certain model performs better in a specific water environment, its prediction result can be prioritized; or, based on the interpretability and reliability of the model, the more reliable prediction data can be selected.

[0052] Step 212: Perform an availability check on the water area risk prediction data. If the availability check result indicates that the water area risk prediction data is reliable, then use the water area risk prediction data as the target risk prediction data.

[0053] Among them, availability checking can be a process of verifying the quality of water risk prediction data, with the aim of ensuring the reliability and applicability of the prediction data.

[0054] Among them, target risk prediction data can refer to water risk prediction data that has been verified as reliable after availability checks, and this data is ultimately used as the basis for decision support.

[0055] Specifically, usability checks involve verification across multiple dimensions: First, the completeness of the water risk prediction data is checked to ensure no data is missing and all indicators can be calculated correctly or meet objective conditions. Second, the accuracy of the water risk prediction data is assessed to determine whether it is constructed from known historical or reference data and follows reasonable objective laws, ensuring the model output has no significant bias or errors. Third, the timeliness of the water risk prediction data is checked to confirm that it is based on the latest hydrological conditions and meteorological forecasts and is applicable to the current risk prediction scenario. Finally, the reliability of the water risk prediction data is further confirmed by comparing it with historical observation data or using retrospective analysis. If the water risk prediction data passes all these checks, it indicates that the data is reliable and usable and can be used as target risk prediction data. If the check results indicate that the water risk prediction data is flawed or does not conform to reality, the accuracy of the model needs to be reassessed or data needs to be recollected to improve the reliability and practicality of the prediction results.

[0056] In the aforementioned multi-dimensional hydrological risk assessment method, real-time and historical hydrological data of the target water area are input into a hydrological risk machine model and a hydrological risk physical model. By comparing the prediction results of the two models, when the difference exceeds a preset threshold, the physical parameters in the model are automatically adjusted to reduce model error. After multiple adjustments and iterations, the most reliable risk prediction data can be selected, ensuring the effectiveness of the prediction results in practical applications. Further usability checks on the water area risk prediction data enhance the credibility of the model output, avoiding decision-making errors caused by model errors or data inconsistencies. By integrating the advantages of machine learning and physical modeling, the scientific rigor and practicality of water area risk prediction are improved, thus effectively enhancing the accuracy and timeliness of hydrological risk prediction results and providing more accurate and real-time data support for water area management, disaster early warning, and decision-making.

[0057] In one exemplary embodiment, such as Figure 3 As shown, the hydrological risk physical model includes a water quality-discharge coupled sub-model and a topographic-flood coupled sub-model; real-time hydrological data and historical hydrological data are input into the hydrological risk physical model to obtain the risk physical model prediction data, including steps 302 to 306.

[0058] in:

[0059] Step 302: Input real-time hydrological data and historical hydrological data into the water quality-flow coupling sub-model to obtain water quality-flow coupling data.

[0060] The water quality-flow coupling sub-model can be a mathematical model describing the interaction between water flow and water quality. Its core principle is that the dynamic changes in water flow affect the diffusion, propagation, and sedimentation of pollutants in the water, while changes in pollutant concentration affect the characteristics of water flow.

[0061] Among them, water quality-flow coupling data can be data calculated through a water quality-flow coupling sub-model, reflecting the relationship between water flow and water quality. These data include the velocity and flow rate of water under specific hydrological conditions and their impact on the diffusion of pollutants in the water.

[0062] Specifically, real-time and historical hydrological data are input into the water quality-flow coupling sub-model. This sub-model primarily focuses on the interaction between water flow and water quality. It estimates water quality changes by simulating the processes of water velocity, flow rate, and the diffusion, sedimentation, and degradation of pollutants in the water. For example, it can calculate the concentration changes of dissolved oxygen, nitrogen and phosphorus pollutants, and heavy metals in rivers. Therefore, the water quality-flow coupling sub-model first calculates the flow velocity, flow rate, and their distribution, and uses flow equations (such as the Navier-Stokes equation or shallow water equations) to describe the spatial and temporal changes in the flow. Next, it simulates the diffusion, deposition, and degradation of pollutants in the water flow using water quality transport equations (such as the Advancement-Diffusion equation). During the calculation, the water quality-flow coupling sub-model also considers changes in water quality, such as the concentration changes of dissolved oxygen, nitrogen and phosphorus, and heavy metals in the water. Finally, the water quality-flow coupling sub-model outputs the coupling results of water quality and water flow, thus obtaining water quality-flow coupling data.

[0063] The expression for the water quality-flow coupling sub-model is as follows:

[0064]

[0065] Where Q1(t) is the first water flow rate at time t, P(t) is the real-time precipitation at time t, and Q h λ(t) represents the historical flow rate at time t, A is the catchment area, λ is the coupling regulation coefficient between precipitation and catchment area, δ is the influence coefficient of historical flow rate on current flow rate, α is the influence coefficient of historical flow rate on current flow rate prediction, κ is the decay rate of the influence of historical flow rate on flow rate, C(t) is the concentration of water quality factor at time t, D is the diffusion rate of substances within the water body, V is the water volume, γ is the coefficient of precipitation on water quality factor at time t, f(·) is the nonlinear influence function of the first flow rate at time t, and τC(t) is the feedback effect of water quality factor concentration on flow rate at time t. The cumulative effect of the coupling effect between flow rate and water quality concentration over time.

[0066] Step 304: Input real-time hydrological data and historical hydrological data into the topographic flood coupling sub-model to obtain topographic flood coupling data.

[0067] The topographic-flood coupling sub-model can be a mathematical model that describes the interaction between hydrological conditions and topographic features. This model helps predict the occurrence, spread, and extent of floods by simulating the effects of precipitation, watershed topography (such as slope and riverbed morphology), and other factors on flood formation and propagation.

[0068] Among them, topographic flood coupling data can be generated through a topographic flood coupling sub-model, reflecting information such as flood propagation, flood accumulation areas, and water level changes under specific topographic and hydrological conditions. This data includes the flood flow path, inundated area, and water flow velocity and water level changes.

[0069] Specifically, real-time and historical hydrological data are input into the topographic flood coupling sub-model. This sub-model primarily focuses on simulating the interaction between hydrological conditions and topographic features, particularly in precipitation-induced flood scenarios, examining how topographic features influence water propagation paths, flood extent, and flow velocity. The topography of a watershed (e.g., mountains, plains, riverbeds) determines how precipitation converges into rivers and ultimately leads to flooding. For example, steep mountains may cause rapid water accumulation after heavy rain, resulting in dramatic floods, while flat areas may experience waterlogging or stagnant water. Therefore, the topographic flood coupling sub-model uses topographic features and hydrodynamic equations to simulate the water propagation process. Shallow water equations (Saint-Venant equations) or hydrodynamic models are used to describe changes in water level and velocity. Based on precipitation and topographic conditions, the model calculates the water convergence process (e.g., water converging from upstream to downstream) and the water distribution path. By considering different topographic conditions (e.g., mountains, plains), the model further calculates the spread, headwaters, and inundation areas of the water flow. Based on this, the model calculates the spatiotemporal evolution of floods, simulates the rise in water levels and the potential impact of floods in different regions, and couples the above data to obtain topographic flood coupled data.

[0070] The expression for the topographic-flood coupled sub-model is as follows:

[0071]

[0072] Where Q2(t) is the second flow rate at time t, h(x,y,t) is the spatial distribution of water level, v(x,y,t) is the flow velocity vector, P(t) is the real-time precipitation at time t, S is the water area, F(t) is the flood risk index of any area of ​​the water body at time t, σ(x,y) is the watershed water accumulation coefficient, θ(x,y,t) is the flow resistance coefficient, and DEM(t) is the real-time topographic data. A flood source tracing function that combines the effects of precipitation, flow rate, and topography.

[0073] Step 306: Perform dynamic feedback coupling processing on the water quality flow coupled data and the topographic flood coupled data to obtain the risk physical model prediction data.

[0074] In this dynamic feedback coupling process, water quality changes and flood propagation are mutually influential. The outputs of the two models need to be adjusted based on real-time data to ensure that the models accurately reflect their interaction. During the dynamic feedback process, changes in water flow may affect the propagation of pollutants, while changes in pollutant concentration will affect the dynamic characteristics of water flow, and vice versa.

[0075] Specifically, since water quality-flow coupled data provides the relationship between pollutant diffusion and water flow, while topographic flood coupled data provides the flood propagation path and impact range, these two prediction results are combined and adjusted. For example, the water flow path and water level changes predicted by the topographic flood model affect the transport velocity and concentration of pollutants, while water quality changes (such as changes in pollutant concentration) may alter the dynamic characteristics of water flow, affecting flood spread and water level changes. By introducing nonlinear adjustment factors, especially when the interaction between water quality and flood is strong, it is ensured that the outputs of the two models can reflect the complex relationship between water quality changes and flood dynamics. The models self-optimize, adjusting parameters such as water flow, pollutant diffusion, and flood propagation, making the final risk physics model prediction data more accurately reflect actual hydrological risks.

[0076] The expression for dynamic feedback coupling processing is as follows:

[0077]

[0078] Among them, R final(t) The data used in the risk physics model prediction are as follows: ρ1 is the hydrological risk feedback coefficient, ρ2 is the flood risk feedback coefficient, ξ1 is the hydrological risk nonlinear adjustment coefficient, and ξ2 is the flood risk nonlinear adjustment coefficient. ρ1 and ρ2 are time-adaptive and can be adjusted based on historical data and different scenarios; ξ1 and ξ2 can be set through data fitting or experiments, depending on factors such as watershed, climate, and historical floods, and are dynamically adjusted based on real-time data.

[0079] In this embodiment, by inputting real-time and historical hydrological data into the water quality-discharge coupling sub-model and the topographic flood coupling sub-model, detailed prediction data related to water quality and floods can be obtained respectively. The water quality-discharge coupling sub-model helps simulate the interaction between water flow and water quality, assessing the diffusion, deposition, and degradation processes of pollutants in water, while the topographic flood coupling sub-model reflects the interaction between hydrological conditions and topography, predicting the occurrence and propagation of floods. By dynamically feeding back and coupling these two types of data, the mutual influence between water quality and floods can be simulated more accurately, resulting in comprehensive risk physics model prediction data. Overall, this provides more comprehensive and accurate data support for water area risk prediction, helping decision-makers better address water pollution prevention and flood management, and improving the reliability of risk assessment and the scientific basis of decision-making.

[0080] In one exemplary embodiment, such as Figure 4 As shown, an availability check is performed on the water area risk prediction data. If the availability check result indicates that the water area risk prediction data is reliable, then the water area risk prediction data is used as the target risk prediction data, including steps 402 to 408. Wherein:

[0081] Step 402: Construct risk data constraints based on real-time hydrological data, historical hydrological data, and real-time topographic data of the target water area.

[0082] Real-time topographic data can be detailed information about the target water body and its surrounding topography obtained through remote sensing technology, geographic information systems (GIS), or other monitoring tools within a specific time period. This data includes, but is not limited to, the topographic relief, slope, riverbed morphology, watershed distribution, soil type, and other geographical factors that may affect water flow and water level.

[0083] Risk data constraints can be limitations or rules set using actual observation data, historical hydrological data, real-time topographic data, etc., when conducting water area risk prediction. These conditions are used to ensure that the predicted data output by the hydrological model conforms to reality and avoids unreasonable results. For example, water level data may be limited by certain physical or geographical features, flow velocity may not exceed a certain critical value, or under specific topographic conditions, water flow in certain areas may not reach a certain preset range.

[0084] Specifically, when constructing risk data constraints for water area risk prediction, it is necessary to fully utilize current real-time hydrological data, historical hydrological data, and real-time topographic data to determine reasonable boundary conditions. This data includes, but is not limited to, current water level, flow velocity, precipitation, historical basin flood events, and topographic information of the water area (such as slope, riverbed morphology, and basin area). By integrating these factors, constraints on the water area risk prediction data can be established to ensure that all prediction results are within physically and hydrologically feasible limits. For example, the water level should not exceed a certain height, and the flow velocity should meet certain physical constraints. These constraints effectively limit the model's output range and avoid unrealistic prediction results.

[0085] Step 404: Based on the risk data constraints, perform a hydrological logic check on the water area risk prediction data to obtain the risk data logic check data.

[0086] Among them, hydrological logic checking can be a verification of water area risk prediction data, mainly checking whether the hydrological data output by the model conforms to actual hydrological laws and physical logic.

[0087] Among them, the risk data logic check data can be the check results after hydrological logic check, which is used to further ensure that these prediction data are logically sound.

[0088] Specifically, based on the constraints of the risk data, the system checks hydrological parameters such as water level, flow velocity, and precipitation in the water risk prediction data to ensure that the relationships between them are reasonable, thus obtaining risk data logic check data. For example, water level changes over time should be stable, without sudden extreme fluctuations; there should be a certain physical relationship between water flow and water level, and there should be no abnormal situations such as excessively high flow or excessively low water level. Moreover, the hydrological logic check process also compares with historical actual observation results to verify whether the model output is consistent with actual hydrological patterns. If there are illogical situations in the prediction data, such as unreasonable fluctuations in water level or abnormal changes in flow velocity, the system will identify them and take further action.

[0089] Step 406: If there are no hydrological logical anomalies in the risk data characterization water area risk prediction data after the risk data logic check, perform a credibility analysis on the water area risk prediction data to obtain the credibility analysis results of the risk data.

[0090] Credibility analysis can be a process of evaluating water risk prediction data to verify the reliability and accuracy of the prediction results. By retrospectively analyzing the model's performance in similar scenarios, credibility analysis can assess the accuracy of the model's output. This analysis may also include error analysis to detect discrepancies between the model's predicted data and actual data. If the prediction results are consistent with historical data and the error is within an acceptable range, then the prediction data can be considered reliable.

[0091] The credibility analysis results of risk data can be conclusions drawn after conducting a credibility analysis on the water risk prediction data. This result indicates the reliability of the prediction data, specifically manifested in the error between the prediction results and actual observation data, the error range of the model, and the degree of consistency between the prediction data and the actual data. If the credibility analysis results show that the prediction data error is small and consistent with historical data and physical constraints, then the data is considered credible. If the credibility analysis results show that the data has significant bias or inconsistency, then the model needs to be readjusted or the input data further corrected.

[0092] Specifically, after verifying that the risk data logic check shows no hydrological logical anomalies in the water area risk prediction data, the next step is to conduct a credibility analysis of the water area risk prediction data. Credibility analysis primarily verifies the accuracy of the model's output by comparing the predicted data with historical observation data. This includes backtesting analysis, where the model is used to predict historical events and compared with the actual results to check whether the model can accurately predict risks in similar historical contexts. Furthermore, credibility analysis includes error assessment, examining the error range between the model output and actual observations. For the credibility of the prediction data, an error tolerance threshold is typically set. Only when the error of the prediction result is within the tolerance range is the data considered credible, ensuring that the final prediction data has sufficient reliability, thus obtaining the credibility analysis results for the risk data.

[0093] Step 408: If the risk data credibility analysis results indicate that the water area risk prediction data is credible, then the water area risk prediction data shall be used as the target risk prediction data.

[0094] Specifically, if the credibility analysis results indicate that the water area risk prediction data is credible, meaning the credibility analysis shows that the water area risk prediction data meets the accuracy requirements and there are no obvious deviations or discrepancies with reality, the final prediction data will be deemed credible and returned as the target risk prediction data. If the credibility analysis fails, the model needs to be readjusted or the input data further corrected to ensure that the prediction data conforms to the actual situation and the actual risk needs of the water area, thus ensuring the effectiveness and operability of the decision.

[0095] In this embodiment, by constructing risk data constraints based on real-time hydrological data, historical data, and real-time topographic data of the target water area, it is ensured that the water area risk prediction data is within the range of physical and hydrological laws, avoiding unreasonable prediction results. Hydrological logic checks further verify whether the risk prediction data conforms to hydrological laws, ensuring that there are no logical problems such as abnormal water levels or unreasonable flow velocities. If the risk data logic check passes, a credibility analysis is performed. Through historical comparison of the data and model verification, the reliability of the prediction results is ensured. Finally, when the credibility analysis results indicate that the prediction data is credible, it can be used as target risk prediction data for actual decision-making, ensuring the scientific nature, accuracy, and operability of the risk prediction data. This provides reliable data support for water area risk management and emergency response, effectively improving the quality and efficiency of decision-making.

[0096] In one exemplary embodiment, such as Figure 5 As shown, a credibility analysis is performed on the water area risk prediction data to obtain the credibility analysis results, including steps 502 to 506. Wherein:

[0097] Step 502: Use Latin hypercube sampling to quantify the uncertainty of water area risk prediction data and obtain risk data uncertainty information.

[0098] Among these, the uncertainty information in risk data can be obtained by analyzing the range and uncertainty of the input parameters of the model, which determines the level of uncertainty in the risk prediction data for water areas. This uncertainty usually stems from factors such as variations in the input data, measurement errors, and model assumptions. For example, input data such as precipitation, flow velocity, and topography may have certain prediction errors or fluctuations.

[0099] Specifically, Latin Hypercube Sampling (LHS) is a statistical sampling method primarily used to quantify uncertainty and improve the representativeness of simulation results. Therefore, the input variables of the risk prediction model (such as water level, flow velocity, and precipitation) are first determined. Then, the LHS method is used to uniformly sample from the possible range of water area risk prediction data, ensuring that the sampling results cover all possible scenarios. This generates multiple sets of different input combinations, which are then simulated separately to obtain a series of water area risk prediction data. These prediction data are used to quantify the uncertainty of the prediction results, revealing how the prediction results change under different input conditions. Finally, using the multiple sets of results generated by LHS sampling, the uncertainty information of the water area risk prediction data can be calculated.

[0100] Step 504: Perform sensitivity analysis on the water area risk prediction data to obtain risk data sensitivity information.

[0101] Sensitivity analysis can be used to assess the degree of influence of input variables on model output. In water risk prediction, sensitivity analysis identifies which factors play an important role in model output by systematically changing input variables (such as water level, flow velocity, precipitation, etc.) and observing their impact on results (such as flood risk, water level changes, etc.).

[0102] The risk data sensitivity information can be derived through sensitivity analysis, reflecting the specific degree of influence of different input variables on the risk prediction results for water areas. This information reveals which factors have a significant impact on the model's output data. For example, if a change in an input variable (such as precipitation or flow velocity) leads to a significant change in the model's prediction results, it indicates that the variable has high sensitivity to the prediction results.

[0103] Specifically, the first step is to select key input variables that influence water area risk prediction, such as precipitation, flow velocity, and topographic features of the watershed. Then, by systematically testing changes in these input variables, their impact on water area risk prediction data (such as water level and flow rate) is analyzed. During sensitivity analysis, the values ​​of the input variables are typically changed gradually, and the changes in the water area risk prediction data output are observed to identify which factors have a significant impact on the data. Through sensitivity analysis, it is possible to determine which input variables are most sensitive and which factors' changes significantly affect the risk prediction results, thus obtaining risk data sensitivity information.

[0104] Step 506: Determine the credibility analysis results of the risk data based on the uncertainty information and sensitivity information of the risk data.

[0105] Specifically, the uncertainty information in risk data reveals the potential fluctuation range of different input variables during the simulation process, reflecting the model's sensitivity to various uncertainties. Meanwhile, risk data sensitivity information helps identify which variables play a decisive role in changes to risk prediction results by assessing the impact of input variables on the prediction outcomes. Combining these two approaches, the uncertainty information in risk data determines which key input variables have large fluctuation ranges, thus affecting the stability of the results; simultaneously, the sensitivity information indicates which variables' changes significantly affect the model output. Finally, by comprehensively considering this information, the model's performance and reliability under different conditions are evaluated, resulting in a comprehensive risk data credibility analysis.

[0106] In this embodiment, by using Latin hypercube sampling to quantify the uncertainty of water area risk prediction data, the potential fluctuations in input data and model predictions can be comprehensively assessed, thereby revealing the range of variation and sources of uncertainty in the prediction results. Next, sensitivity analysis of the risk prediction data can identify which input variables have the greatest impact on the model results, helping to determine the key factors requiring focus in the model. Combining uncertainty and sensitivity information, risk data credibility analysis not only quantifies the contribution of different factors to the results but also comprehensively assesses the reliability of the prediction data. Overall, this improves the transparency and credibility of risk prediction, providing more accurate and reliable data support for water area risk management, and helping decision-makers make more effective responses.

[0107] In one exemplary embodiment, such as Figure 6 As shown, the credibility analysis results of risk data are determined based on the uncertainty information and sensitivity information of risk data, including steps 602 to 606.

[0108] in:

[0109] Step 602: Based on the anomaly detection algorithm, perform anomaly inference on the uncertainty information and sensitivity information of the risk data to obtain anomaly inference data.

[0110] Anomaly detection algorithms are techniques used to identify data points in a dataset that do not conform to expected patterns or deviate from normal behavior. In water risk prediction, anomaly detection algorithms are used to discover data anomalies that may be caused by measurement errors, inaccurate model assumptions, or extreme events. Commonly used anomaly detection methods include statistical methods (such as the standard deviation method and box plot method) and machine learning methods (such as isolation forests and support vector machines).

[0111] Anomaly inference can be a further analysis process of data anomalies discovered using anomaly detection algorithms. Anomaly inference deduces potential causes or patterns of anomalies by deeply mining the data.

[0112] Anomaly inference data can be the result of the anomaly inference process, representing abnormal patterns or data points identified in the dataset. This data may be deviations from the norm that occur during the prediction process, or changes caused by external factors (such as extreme weather events, sudden disasters, etc.).

[0113] Specifically, anomaly detection algorithms are used to analyze the uncertainty and sensitivity information of risk data to find parts of the data that do not conform to the usual pattern. For example, by performing statistical analysis on input variables (such as water level, flow velocity, precipitation, etc.), data points that deviate from the normal range can be identified. These data points may be caused by abnormal inputs, incorrect measurements, or model assumptions, and outliers in the data can be identified to obtain anomaly inference data.

[0114] Step 604: Based on the deviation propagation model, perform deviation propagation analysis on the uncertainty information and sensitivity information of the risk data to obtain deviation propagation analysis data.

[0115] Among them, the bias propagation model can be a mathematical model that assesses how the uncertainty of input variables propagates during model calculation and ultimately affects the model output. In water risk prediction, the bias propagation model propagates the uncertainty of input data (such as measurement errors, estimation biases, etc.) to all levels of the model, assessing the impact of these uncertainties on the final prediction results. This model calculates how the bias of the input data propagates during the calculation process using mathematical formulas (such as Taylor expansion, Monte Carlo methods, etc.) and quantifies the impact of this propagation on the prediction results, thereby helping to understand the sources and range of uncertainty in the prediction.

[0116] Among them, bias propagation analysis can be the analysis process performed by the bias propagation model. It is used to quantify how the bias of the input variables affects the final prediction result. By systematically analyzing the uncertainty of each input variable, its contribution to the model output is evaluated.

[0117] Among them, deviation propagation analysis data can be the results obtained from the deviation propagation analysis process, indicating which factors have a significant impact on the final prediction result during the deviation propagation process of input variables. This data reveals which input variables in the model significantly contribute to changes in risk prediction results, helping to identify which factors require focused attention and control.

[0118] Specifically, risk data uncertainty information provides the range of uncertainty for input variables (such as precipitation, flow velocity, and water level), which typically originate from measurement errors, model assumptions, or natural variations. Risk data sensitivity information provides the degree of influence of each input variable on the prediction result, identifying which variables are highly sensitive to the final outcome. For example, small changes in some variables may lead to large fluctuations in the prediction result, while changes in other variables have a smaller impact. Based on this uncertainty and sensitivity information, the bias propagation model calculates the specific contribution of the biases to the final result by propagating the biases of the input variables to the model's output layer. This quantifies how the uncertainty of each input parameter accumulates during the prediction process and identifies which factors play a dominant role in the model. Ultimately, bias propagation analysis data is obtained.

[0119] Step 606: Multiply the anomaly inference data with the deviation propagation analysis data to obtain the risk data credibility analysis results.

[0120] Specifically, anomaly inference data is multiplied by deviation propagation analysis data to comprehensively consider the impact of anomaly information and deviation propagation on the reliability of risk data. Anomaly inference data provides clues about which data points may be anomalies, while deviation propagation analysis data reveals the degree of influence of different uncertainty sources on the results. By multiplying anomaly inference data with deviation propagation analysis data, a comprehensive risk data reliability analysis result can be obtained, reflecting the degree of uncertainty and anomaly impact on the data, and providing a quantitative evaluation of the accuracy of water area risk prediction.

[0121] In this embodiment, by using anomaly detection algorithms to infer anomalies in the uncertainty and sensitivity information of risk data, it is possible to effectively identify and eliminate anomalous data that deviates from the conventional pattern, ensuring more accurate and reliable analysis results. Further utilizing a deviation propagation model to analyze the deviation propagation of uncertainty and sensitivity information quantifies how the uncertainty of the input data affects the final prediction results, thereby helping to identify and reduce the propagation of errors. By combining anomaly inference data with deviation propagation analysis data, the impact of anomalous factors and deviations in the data can be comprehensively considered, resulting in a more comprehensive and accurate risk data credibility analysis. Overall, this improves the scientific rigor and credibility of risk prediction, provides stronger support for water area risk assessment and decision-making, and helps to formulate more precise management and response measures.

[0122] Based on the same inventive concept, this application also provides a multi-dimensional hydrological risk assessment device for implementing the aforementioned multi-dimensional hydrological risk assessment method, such as... Figure 7As shown, the device includes: a hydrological data acquisition module 702, a hydrological data analysis module 704, a hydrological model adjustment module 706, a prediction data selection module 708, and a risk data determination module 710. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of the one or more multi-dimensional hydrological risk assessment device embodiments can be found in the limitations of a multi-dimensional hydrological risk assessment method described above, and will not be repeated here.

[0123] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface; wherein, the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0124] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0125] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0126] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0127] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0128] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A multi-dimensional hydrologic risk assessment method, characterized by, The method comprises: acquiring hydrological real-time data and hydrological historical data of a target water area; inputting the hydrological real-time data and the hydrological historical data into a hydrological risk machine model and a hydrological risk physical model respectively to obtain risk machine model prediction data and risk physical model prediction data; in a case where a difference between the risk machine model prediction data and the risk physical model prediction data is greater than a preset difference, adjusting model parameters of the hydrological risk machine model and the hydrological risk physical model according to the difference, comprising: optimizing hyperparameters of the hydrological risk machine model, the hyperparameters comprising a learning rate, a number of hidden layer nodes and a regularization term; and optimizing physical equation parameters of the hydrological risk physical model, the parameters comprising a water flow friction coefficient and a precipitation-runoff conversion coefficient; the optimization adopts at least one of a particle swarm optimization, a genetic algorithm or a Bayesian optimization to reduce the prediction difference between the hydrological risk machine model and the hydrological risk physical model and iteratively converge; taking the adjusted hydrological risk machine model as the hydrological risk machine model and taking the adjusted hydrological risk physical model as the hydrological risk physical model, and returning to execute the step of inputting the hydrological real-time data and the hydrological historical data into the hydrological risk machine model and the hydrological risk physical model respectively to obtain the risk machine model prediction data and the risk physical model prediction data; in a case where the difference between the risk machine model prediction data and the risk physical model prediction data is less than the preset difference, selecting the risk machine model prediction data or the risk physical model prediction data as water area risk prediction data; performing availability checking on the water area risk prediction data, and taking the water area risk prediction data as target risk prediction data in a case where the availability checking result indicates that the water area risk prediction data is reliable.

2. The method of claim 1, wherein, The hydrological risk physical model comprises a water quality flow coupling sub-model and a terrain flood coupling sub-model; the inputting the hydrological real-time data and the hydrological historical data into the hydrological risk physical model to obtain the risk physical model prediction data comprises: inputting the hydrological real-time data and the hydrological historical data into the water quality flow coupling sub-model to obtain water quality flow coupling data; inputting the hydrological real-time data and the hydrological historical data into the terrain flood coupling sub-model to obtain terrain flood coupling data; performing dynamic feedback coupling processing on the water quality flow coupling data and the terrain flood coupling data to obtain the risk physical model prediction data; wherein the dynamic feedback coupling processing is weighting and nonlinear adjustment of the water quality flow coupling data and the terrain flood coupling data according to a hydrological risk feedback coefficient, a flood risk feedback coefficient, a hydrological risk nonlinear adjustment coefficient and a flood risk nonlinear adjustment coefficient; the risk physical model prediction data is used for training and supervision of the hydrological risk machine model.

3. The method of claim 2, wherein, an expression of the water quality flow coupling sub-model is in, for t The first water flow rate at any given moment P ( t )for t Real-time precipitation at any given moment. for t Historical flow at any given moment A For the drainage area, This is the coupling regulation coefficient between precipitation and watershed area. The influence coefficient of historical traffic on current traffic. The influence coefficient of historical flow on current flow forecast. The decay rate of the historical flow's impact on the flow rate. C ( t )for t Concentration of water quality factors at any given time D The diffusion rate of substances within the water body. V For water volume, for t The coefficient of the impact of precipitation on water quality factors at any given time. For the first water flow rate at t Nonlinear influence function at time step In order to be in t The feedback effect of water quality factor concentration on flow rate at any given time. The cumulative effect of the coupling effect between flow rate and water quality concentration over time.

4. The method of claim 2, wherein, an expression of the terrain flood coupling sub-model is wherein, is t the second water flow rate at time t, is the spatial distribution of water level, is the velocity vector, P ( t ) is t the real-time precipitation at time t, S is the water area, F ( t ) is the flood risk index of any region of the water area at time t, t is the water area, is the flood risk index of any region of the water area at time t, is the water area, DEM is the water area, t is the real-time terrain data, is the flood tracing function incorporating the effects of precipitation, flow rate and terrain.

5. The method of claim 2, wherein, An expression of the dynamic feedback coupling process is wherein, is a risk physical model prediction data, is a hydrological risk feedback coefficient, is a flood risk feedback coefficient, is a hydrological risk non-linear adjustment coefficient, is a flood risk non-linear adjustment coefficient; is a hydrological risk feedback term, is a flood risk feedback term.

6. The method of claim 1, wherein, The availability check on the water area risk prediction data, in the case that the check result of the availability check represents that the water area risk prediction data is credible, taking the water area risk prediction data as target risk prediction data, comprises: According to the hydrological real-time data, the hydrological historical data and the real-time terrain data of the target water area, constructing a risk data constraint condition; According to the risk data constraint condition, performing hydrological logic check on the water area risk prediction data to obtain risk data logic check data; In the case that the risk data logic check data represents that the water area risk prediction data does not exist hydrological logic abnormality, performing credibility analysis on the water area risk prediction data to obtain risk data credibility analysis result; In the case that the risk data credibility analysis result represents that the water area risk prediction data is credible, taking the water area risk prediction data as the target risk prediction data.

7. The method of claim 6, wherein, The credibility analysis on the water area risk prediction data comprises: Using Latin hypercube sampling, quantifying the uncertainty of the water area risk prediction data to obtain risk data uncertainty information; Performing sensitivity analysis on the water area risk prediction data to obtain risk data sensitivity information; According to the risk data uncertainty information and the risk data sensitivity information, determining the risk data credibility analysis result.

8. The method of claim 7, wherein, The determination of the risk data credibility analysis result according to the risk data uncertainty information and the risk data sensitivity information comprises: According to an anomaly detection algorithm, performing anomaly inference on the risk data uncertainty information and the risk data sensitivity information to obtain anomaly inference data; According to a bias propagation model, performing bias propagation analysis on the risk data uncertainty information and the risk data sensitivity information to obtain bias propagation analysis data; Multiplying the anomaly inference data and the bias propagation analysis data to obtain the risk data credibility analysis result.

9. A multi-dimensional hydrologic risk assessment apparatus, characterized by, The device comprises: A hydrological data acquisition module for acquiring hydrological real-time data and hydrological historical data of a target water area; A hydrological data analysis module for inputting the hydrological real-time data and the hydrological historical data into a hydrological risk machine model and a hydrological risk physical model respectively to obtain risk machine model prediction data and risk physical model prediction data; The hydrological model adjustment module is configured to adjust model parameters of the hydrological risk machine model and the hydrological risk physical model according to a difference between the risk machine model prediction data and the risk physical model prediction data when the difference is greater than a preset difference, including: optimizing hyperparameters of the hydrological risk machine model, the hyperparameters including a learning rate, a number of hidden layer nodes, and a regularization term; and optimizing physical equation parameters of the hydrological risk physical model, the parameters including a flow friction coefficient and a precipitation-runoff conversion coefficient; the optimization employs at least one of a particle swarm optimization, a genetic algorithm, or a Bayesian optimization to reduce the prediction difference between the hydrological risk machine model and the hydrological risk physical model and to iteratively converge; The hydrological data analysis module is further configured to return to execute the step of inputting the hydrological real-time data and the hydrological historical data into the hydrological risk machine model and the hydrological risk physical model to obtain the risk machine model prediction data and the risk physical model prediction data, with the hydrological risk machine model after parameter adjustment as the hydrological risk machine model and the hydrological risk physical model after parameter adjustment as the hydrological risk physical model; The prediction data selection module is configured to select the risk machine model prediction data or the risk physical model prediction data as the water area risk prediction data until the difference between the risk machine model prediction data and the risk physical model prediction data is less than the preset difference. The risk data determination module is configured to perform an availability check on the water area risk prediction data, and take the water area risk prediction data as target risk prediction data when the availability check result indicates that the water area risk prediction data is reliable. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The processor executes the computer program to implement the steps of the method in any one of claims 1 to 8.

Images