Typhoon-rainstorm flood composite disaster monitoring and evaluating method and system

Abstract

The invention discloses a typhoon-rainstorm flood composite disaster monitoring and evaluating method and system, and belongs to the field of intelligent disaster monitoring, and the method comprises the following steps: building a disaster recognition model through a machine learning algorithm based on multi-factor data, and obtaining a composite disaster judgment result; based on the composite disaster judgment result, constructing a disaster evaluation model for the area forming the disaster; dynamically updating parameters of the disaster assessment model based on real-time monitoring data to obtain a real-time composite disaster assessment result; based on the real-time composite disaster assessment result, dynamically adjusting loss probability distribution to obtain a comprehensive disaster damage assessment result; and carrying out disaster situation space-time evolution based on the comprehensive disaster damage assessment result to obtain a disaster situation development trend.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent disaster monitoring, and in particular relates to a typhoon-rainstorm flood composite disaster monitoring and assessment method and system. Background Art

[0002] In the monitoring and assessment of typhoon-rainstorm-flood compound disasters, there are the following technical problems: First, how to accurately identify the occurrence of typhoons and rainstorms, and determine whether they constitute a compound event. This requires comprehensive consideration of multiple factors such as the typhoon path, rainstorm area, and occurrence time, and setting a reasonable threshold. Secondly, the loss assessment of compound disasters is difficult. Typhoons and rainstorms have different impact areas and loss mechanisms. How to consider the coupling effect of the two and establish a reasonable loss assessment model requires in-depth research. Furthermore, the occurrence of typhoon-rainstorm compound events has strong randomness and uncertainty, and the evolution process of the disaster is complex and changeable. How to achieve real-time dynamic monitoring of the disaster process and give corresponding assessment results in combination with the disaster loss assessment model puts forward high requirements on technology. Finally, the temporal and spatial scales of typhoon-rainstorm disaster events vary greatly, and the influence of regional topography, landform, land use and other factors is complex. The disaster loss response mechanism in different regions is different. Therefore, the assessment model should have regional applicability and be able to carry out disaster monitoring and assessment according to local conditions. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a typhoon-rainstorm flood combined disaster monitoring and assessment method and system to solve the problems existing in the above-mentioned prior art.

[0004] To achieve the above object, the present invention provides a typhoon-rainstorm flood combined disaster monitoring and assessment method, comprising:

[0005] Based on multi-factor data, a disaster identification model is established through machine learning algorithms to obtain compound disaster judgment results;

[0006] Constructing a disaster assessment model for the area constituting the disaster based on the composite disaster judgment result;

[0007] Dynamically updating the parameters of the disaster assessment model based on real-time monitoring data to obtain real-time composite disaster assessment results;

[0008] Based on the real-time composite disaster assessment result, dynamically adjusting the loss probability distribution to obtain a comprehensive disaster loss assessment result;

[0009] Based on the comprehensive disaster loss assessment results, the spatiotemporal evolution of the disaster situation is carried out to obtain the development trend of the disaster situation.

[0010] Optionally, the disaster identification model obtains a composite disaster judgment result by analyzing rainfall monitoring data of heavy rain occurring on the same day within a 300-kilometer buffer zone of the typhoon path.

[0011] Optionally, when it is determined to be a complex disaster, a regional modeling strategy is adopted to divide the study area into several assessment units, and a complex disaster assessment model is constructed based on the natural geographical attributes of the several assessment units.

[0012] Optionally, the process of obtaining real-time composite disaster assessment results includes:

[0013] Input the real-time monitoring data of the disaster area into the disaster assessment model to obtain the loss distribution of each assessment unit;

[0014] Based on the loss distribution, when the loss of the assessment unit exceeds a preset threshold, the current assessment unit is judged to be a high-risk area, and an identification result of the high-risk area is obtained;

[0015] Based on the location distribution of high-risk areas and road network data, the optimal dispatch route of the rescue team is planned to obtain the rescue route plan;

[0016] Based on the continuous acquisition of real-time monitoring data and the re-evaluation of the disaster damage assessment model, the rescue plan is dynamically adjusted to obtain an adjusted rescue plan.

[0017] Optionally, the process of constructing the disaster assessment model also includes describing the interaction of the multi-factor data through a nonlinear function, introducing the interaction terms of the strong winds and heavy rains brought by the typhoon; and performing storm surge warning based on the disaster assessment model introducing the interaction terms.

[0018] Optionally, the process of obtaining a comprehensive disaster damage assessment result includes:

[0019] Based on historical typhoon-rainstorm flood data, a probability distribution model of compound disaster events is established, and the probability of occurrence of different disaster scenarios is obtained;

[0020] Based on the probability of occurrence of different disaster scenarios, the Monte Carlo simulation method is used to estimate the loss distribution under different disaster scenarios;

[0021] Collect wind speed and water level data in real time, and update the probability distribution model of disaster events based on the wind speed and water level data to obtain an adjusted probability of disaster occurrence;

[0022] Based on the adjusted disaster occurrence probability, the disaster loss distribution of different scenarios is weighted averaged to obtain a comprehensive disaster loss assessment result.

[0023] Optionally, based on the comprehensive disaster loss assessment results, a spatiotemporal data analysis method is used to perform a spatiotemporal evolution analysis of the disaster to obtain the distribution characteristics of the disaster at different times and spaces. Based on the distribution characteristics of the disaster at different times and spaces, high-risk areas of the disaster are identified, key time nodes of the disaster are determined, and the severity and scope of the disaster are judged.

[0024] The present invention also provides a typhoon-rainstorm flood compound disaster monitoring and evaluation system, comprising:

[0025] A composite disaster identification module, which is used to obtain typhoon and rainstorm data and determine whether it constitutes a composite disaster through machine learning;

[0026] A sub-regional modeling module, the sub-regional modeling module is used to divide the assessment unit and construct a regional disaster loss assessment model;

[0027] A real-time assessment module, which is used to obtain real-time monitoring data and dynamically update model parameters to perform disaster assessment;

[0028] A coupling effect evaluation module, wherein the coupling effect evaluation module is used to improve the evaluation accuracy by describing the multi-factor interaction through nonlinear functions;

[0029] A probability statistics module, which is used to estimate the distribution of disaster losses under different scenarios and dynamically adjust the probability;

[0030] The disaster situation analysis module is used to analyze the temporal and spatial evolution of the disaster situation and predict the development trend to provide a basis for decision-making.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] The present invention discloses a monitoring and assessment method for typhoon-rainstorm compound disasters. In view of the coupling effect of typhoon and rainstorm disasters, a compound disaster identification model is trained by a machine learning algorithm to determine whether a compound disaster is constituted. For situations where a compound disaster is constituted, a regional modeling strategy is adopted, and the research area is divided into assessment units according to natural attributes such as topography and landforms, and a disaster loss assessment model applicable to the region is constructed in combination with historical disaster data. The model introduces the interaction terms of strong winds and rainstorms brought by typhoons, characterizes the coupling effect of multiple factors through nonlinear functions, and uses probabilistic statistical methods to deal with the uncertainty of disasters. When a disaster occurs, the model parameters are dynamically updated based on real-time monitoring data to achieve real-time assessment of the losses of each assessment unit, and analyze the temporal and spatial evolution trend of the disaster. The present invention improves the accuracy of compound disaster loss assessment by considering the coupling effect and regional differences of disasters, and provides a scientific basis for disaster emergency decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0034] Figure 1 This is a flow chart of a method for monitoring and assessing a typhoon-rainstorm-flood compound disaster according to an embodiment of the present invention;

[0035] Figure 2 A flow chart of compound disaster identification according to an embodiment of the present invention;

[0036] Figure 3 This is a structural diagram of a typhoon-rainstorm-flood complex disaster monitoring and assessment system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0038] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] Embodiment 1

[0040] like Figure 1 As shown, this embodiment provides a typhoon-rainstorm flood combined disaster monitoring and assessment method, including the following steps:

[0041] Step S101, as Figure 2 As shown, based on multi-factor data, a disaster identification model is established through machine learning algorithm to obtain the composite disaster judgment result; the multi-factor data includes typhoon path, rainstorm intensity, rainstorm range, and rainstorm duration.

[0042] Furthermore, multiple factors such as typhoon path, rainstorm intensity, rainstorm range, and rainstorm duration are obtained, and a composite disaster identification model is trained through a machine learning algorithm. The composite disaster identification model is used to determine whether the current typhoon and rainstorm events constitute a composite disaster.

[0043] As a specific implementation of this embodiment, in the process of building a composite disaster identification model, it is first necessary to collect and organize data on typhoon paths, rainstorm intensity, rainstorm range and duration. For example, data on all recorded typhoon and rainstorm events in the past ten years are obtained from the database of the Meteorological Bureau, including the path coordinates, maximum wind speed, rainfall, and rainfall duration of the typhoon. In the process of data cleaning, it is necessary to eliminate those data entries with incomplete or obvious errors, such as records of sudden interruption of the typhoon path or abnormally low rainstorm intensity. Next, feature extraction is one of the key steps. Specifically, features such as wind speed at typhoon path points, path buffer, rainfall change rate intersecting with the buffer, and maximum duration of rainstorm are extracted from the original meteorological data. These features can help the model better understand the behavior patterns and interactions of typhoons and rainstorms. Data normalization ensures the effectiveness of model training. Due to the different dimensions and magnitudes of different data indicators such as wind speed and rainfall, directly using these data may cause deviations in the model training process. By scaling all feature values ​​to the same range, such as between 0 and 1, model training can be made more stable and efficient. The support vector machine (SVM) algorithm is used for model training because SVM has shown good performance in dealing with small samples, nonlinear and high-dimensional pattern recognition. Through grid search and cross-validation methods, different hyperparameter combinations (such as kernel function type, penalty parameter C and kernel parameter γ) can be systematically searched and evaluated to find the optimal model parameter settings. After the model training is completed, the real-time monitored typhoon and rainstorm data are input into the model for prediction. For example, if the currently monitored typhoon path and rainstorm intensity pattern are similar to the pattern marked as a composite disaster in the training data, the model may predict that this event will constitute a composite disaster. At this time, the early warning mechanism will be triggered, and the relevant departments will receive the early warning information and initiate the corresponding emergency plan according to the early warning level. Continuous monitoring and dynamic updating are necessary because the actual path and intensity of typhoons and rainstorms may be different from the initial prediction. Real-time updating of the model input data can adjust the prediction results and ensure the timeliness and accuracy of emergency response measures. Finally, by collecting the actual impact data after the disaster, such as the affected area and economic losses, the prediction effect of the model can be evaluated, and the model can be further optimized and improved accordingly. Regularly retraining the model to adapt to changes in disaster characteristics is a key measure to ensure the long-term effectiveness and adaptability of the model.

[0044] Furthermore, based on real-time monitoring data of multi-dimensional factors such as typhoon path, rainstorm intensity, and duration, and using the trained composite disaster identification model for prediction, it is determined that if there is rainstorm on the same day within the 300-kilometer buffer zone of the typhoon path, it is a typhoon-rainstorm composite event.

[0045] As a specific implementation method of this embodiment, multi-dimensional real-time monitoring data such as typhoon path, rainstorm intensity, duration, etc. are obtained to construct a training data set for the composite disaster identification model. Machine learning algorithms such as support vector machines or random forests are used to train and optimize the composite disaster identification model based on the training data set. When typhoon path data is monitored, the 300-kilometer buffer area range is determined with the typhoon path as the center. The rainstorm monitoring data in the buffer area is obtained, and characteristic parameters such as rainstorm intensity and duration are extracted. The typhoon gale and rainstorm characteristic parameters are input into the trained composite disaster identification model to perform composite disaster assessment and analysis. If the evaluation result of the composite disaster identification model shows that there is a rainstorm event in the buffer area at the same time, it is determined to be a typhoon-rainstorm composite event. According to the evaluation results of the typhoon-rainstorm composite event, early warning information is automatically generated and released through multiple channels to guide relevant departments to take disaster prevention and mitigation measures.

[0046] Step S102, constructing a disaster assessment model for the area causing the disaster based on the composite disaster judgment result;

[0047] Furthermore, if the current typhoon and rainstorm events constitute a composite disaster, then the study area is divided into several assessment units using a regional modeling strategy in view of the differences in the impact areas and loss mechanisms of the composite disasters. For each assessment unit, its topography, landform, land use and other natural geographical attributes are obtained, and the disaster response characteristics of each assessment unit are extracted from historical disaster data to construct a regionally applicable disaster loss assessment model.

[0048] As a specific implementation of this embodiment, obtaining the natural geographical attribute data such as the topography, landform, and land use of the study area is the basis for disaster damage assessment. For example, a certain study area is located in the southeast coast, with complex terrain, including mountains, plains and hills, and various land use types, including farmland, urban built-up areas, forests, etc. These data are divided into several assessment units, and divided according to the topographic and geomorphic characteristics and land use types, such as dividing mountains and forests into one unit, plains and farmland into another unit, and urban built-up areas as a single unit. For each assessment unit, the damage data of typhoon and rainstorm disaster events are extracted from the historical disaster database. Assume that there are records of typhoon and rainstorm events that occurred in the region in the past ten years in the database, including data such as the affected area, economic losses, casualties, etc. of each event. When extracting these data, it is necessary to ensure the integrity and accuracy of the data. For example, for mountain and forest units, the extracted data may show that a typhoon caused a large area of ​​forest to fall, with economic losses of tens of millions of yuan; while for urban built-up area units, the data may show that a rainstorm caused urban waterlogging, with more serious economic losses and casualties. The extracted disaster loss data is preprocessed to remove outliers and missing values. For example, if the economic loss data in a certain record is abnormally high, it is verified that it is due to data entry errors and should be removed; for missing values, they can be filled by interpolation or using average values. After preprocessing, a standardized disaster loss data set is obtained to ensure the reliability of subsequent analysis. The disaster loss data of each assessment unit are analyzed to extract key indicators that reflect the disaster loss response characteristics of the region. For example, the intensity of disaster loss can be expressed by the ratio of economic loss to the affected area, and the scope of disaster loss can be measured by the proportion of the affected area to the total area. For example, the disaster loss intensity of mountain and forest units may be low, but the scope of disaster loss is wide; while the disaster loss intensity of urban built-up area units is high, but the scope of disaster loss is relatively concentrated. According to the natural geographical attributes and disaster loss response characteristics of each assessment unit, machine learning algorithms such as support vector machines or random forests are used to train the disaster loss assessment model for the region. Assuming that the support vector machine algorithm is selected, it is necessary to first select a suitable kernel function, such as the radial basis function (RBF), and then optimize the model hyperparameters, such as the penalty parameter C and the kernel function parameter γ, through grid search and cross-validation methods. Through training, a disaster loss assessment model for each assessment unit is obtained. Using the trained disaster loss assessment model, a new typhoon and rainstorm compound disaster event is quickly assessed to predict its disaster loss impact in each assessment unit. For example, when a new typhoon and rainstorm event occurs, based on real-time monitoring data, it is input into the model, and the model will output the estimated economic losses and affected areas of each assessment unit. Assume that the model predicts that the economic losses of urban built-up area units will reach hundreds of millions of yuan, and the affected areas of mountain and forest units will exceed thousands of hectares. The disaster loss assessment results of each assessment unit are summarized to obtain a comprehensive disaster loss assessment result for the entire study area.For example, by summarizing the estimated economic losses and affected areas of each unit, the overall economic losses and affected areas of the entire study area can be obtained, providing decision support for disaster emergency management and post-disaster reconstruction. Assuming that the comprehensive assessment results show that the overall economic losses of the entire study area will reach tens of billions of yuan and the affected area will exceed tens of thousands of hectares, the relevant departments can formulate corresponding emergency response measures and post-disaster reconstruction plans based on this assessment result. The reason for this is that through the detailed division of assessment units and analysis of disaster loss data, the disaster loss characteristics of different regions can be more accurately grasped, thereby improving the accuracy and pertinence of disaster loss assessment. The use of machine learning algorithms for model training can make full use of the information in historical data and improve the reliability of predictions. The final comprehensive disaster loss assessment results provide decision makers with a comprehensive and scientific basis, which helps to improve the efficiency and effectiveness of disaster response. Through continuous model training and updating, it can also adapt to changes in disaster characteristics and further improve the robustness and practicality of disaster loss assessment.

[0049] Step S103, dynamically updating the parameters of the disaster assessment model based on the real-time monitoring data to obtain real-time flood disaster assessment results; the process of obtaining real-time flood disaster assessment results includes: inputting the real-time monitoring data of the disaster area into the disaster assessment model to obtain the loss distribution of each assessment unit; based on the loss distribution, when the loss of the assessment unit exceeds a preset threshold, the current assessment unit is judged to be a severely affected area, and the identification result of the severely affected area is obtained; based on the location distribution and road network data of the severely affected area, the optimal dispatch route of the rescue team is planned to obtain a rescue route plan; based on the continuous acquisition of real-time monitoring data and the re-evaluation of the disaster loss assessment model, the rescue plan is dynamically adjusted to obtain an adjusted rescue plan.

[0050] Furthermore, when a typhoon-rainstorm combined disaster occurs, real-time monitoring data such as rainfall, wind speed, and water level in the disaster area are obtained, and the real-time monitoring data are input into a pre-constructed disaster loss assessment model. The parameters of the disaster loss assessment model are dynamically updated through an online learning algorithm, and a real-time assessment of the disaster situation at the current moment is performed to obtain the loss distribution of each assessment unit.

[0051] As a specific implementation of this embodiment, obtaining real-time monitoring data of rainfall, wind speed and water level in the disaster area is the basis for disaster damage assessment. For example, during the passage of Typhoon Haiyan, a meteorological monitoring station in a coastal city recorded an hourly rainfall of 100 mm, a wind speed of 30 m / s, and a water level rise of 2 m above the warning line. These data are transmitted to the data processing center in real time through the sensor network to provide basic data for subsequent disaster damage assessment. The monitoring data is input into a pre-built disaster damage assessment model, which is usually trained based on historical disaster data and machine learning algorithms. Assume that the city has built a disaster damage assessment model based on a random forest algorithm, and the model input includes parameters such as rainfall, wind speed, and water level, and the output is the disaster damage level of each assessment unit. After inputting real-time monitoring data, the model preliminarily assesses the disaster damage in different areas. The use of an online learning algorithm to dynamically update the parameters of the disaster damage assessment model is to enable the model to adapt to changes in the current disaster situation. For example, over time, the path and intensity of a typhoon may change, resulting in a significant increase in rainfall and wind speed in certain areas. Through the online learning algorithm, the model can adjust parameters in real time, such as increasing the weight of the impact of rainfall on disaster losses, so as to make the assessment results more accurate. Through the updated disaster loss assessment model, the typhoon-rainstorm compound disaster at the current moment is evaluated in real time to obtain the loss distribution of each assessment unit. Assume that the model assessment results show that the disaster loss level in the eastern part of the city is "serious", the western part is "moderate", and the southern part is "mild". This distribution can help decision makers quickly identify the hardest hit areas. If the loss of an assessment unit exceeds the preset threshold, the assessment unit is judged to be a hard hit area and rescue needs to be arranged first. For example, the disaster loss level is set as "serious" as the threshold, and the eastern part is judged to be a hard hit area. At this time, the rescue department needs to mobilize resources immediately and give priority to rescue in this area. According to the location distribution of the hard hit areas and combined with the road network data, the optimal dispatch route of the rescue team is planned to ensure that it reaches the hard hit areas in the shortest time. Assuming that there are three optional rescue routes in the eastern part, through the analysis of the geographic information system (GIS), an optimal route that avoids congestion and damaged sections is selected, which is expected to shorten the rescue time by 20 minutes. During the rescue process, real-time monitoring data of the disaster area is continuously obtained and input into the disaster damage assessment model for re-evaluation, and the rescue plan is adjusted dynamically. For example, during the rescue process, it was found that the rainfall in a certain area of ​​the East District suddenly increased. The model re-evaluation results showed that the disaster damage level in the area rose to "extremely serious". The rescue department immediately adjusted the plan and dispatched additional rescue forces. The loss distribution obtained by the assessment is correlated with the population distribution data of the disaster area to predict possible casualties and property losses, providing a decision-making basis for post-disaster recovery and reconstruction. Assuming that the East District is densely populated, the model predicts that there may be a large number of casualties and property losses in the area. Based on this, government departments formulate detailed evacuation and resettlement plans and give priority to the reconstruction of the area.Through this series of steps, the disaster damage assessment model can not only reflect the changes in the disaster situation in real time, but also provide a scientific basis for rescue and reconstruction. The acquisition of real-time monitoring data and the dynamic update of the model ensure the accuracy of the assessment results; the planning of the optimal dispatch route improves the rescue efficiency; the dynamic adjustment of the rescue plan reduces the risk of secondary disasters; and the correlation analysis of population distribution data provides strong support for post-disaster recovery and reconstruction. This comprehensive disaster damage assessment and rescue strategy has significantly improved the disaster response capabilities and protected the lives and property of the people.

[0052] Step S104, taking into account the coupling effect of typhoons and rainstorms, introduces the interaction term of typhoons and rainstorms into the disaster loss assessment model, and uses nonlinear functions to characterize the interaction of factors such as typhoon intensity, rainstorm intensity, and duration, so as to improve the assessment accuracy of the comprehensive impact of compound disasters and conduct storm surge warnings.

[0053] Furthermore, historical data of typhoons and rainstorms are obtained, including data on factors such as typhoon intensity, rainstorm intensity, and duration; the obtained typhoon and rainstorm data are preprocessed to remove outliers and missing values, and the data are standardized; based on the preprocessed typhoon and rainstorm data, a nonlinear disaster loss assessment model containing interaction terms is constructed, wherein the interaction terms are used to characterize the interaction of factors such as typhoon intensity, rainstorm intensity, and duration; machine learning algorithms such as support vector machines or random forests are used to train the constructed disaster loss assessment model to obtain the optimal parameters of the model; new typhoon and rainstorm data are predicted using the trained disaster loss assessment model to obtain the comprehensive impact assessment results of compound disasters; the assessment results are analyzed to determine whether the assessment accuracy meets the preset threshold, and if not, return to step 3 to adjust the model structure or parameters until the accuracy requirements are met; the disaster loss assessment model that meets the accuracy requirements is applied to actual typhoon and rainstorm disaster warnings and loss assessments to provide decision support for disaster prevention and mitigation.

[0054] As a specific implementation of this embodiment, obtaining historical data of typhoons and rainstorms is the basic work of disaster loss assessment. For example, typhoon and rainstorm data in a coastal area in the past ten years are collected, including key factors such as typhoon intensity (such as maximum central wind speed), rainstorm intensity (such as 24-hour rainfall), and duration. These data can be obtained from historical records of the meteorological department, or they can be comprehensively sorted out through satellite remote sensing data and ground monitoring station data. Preprocessing the acquired data is an important step to ensure data quality. First, remove outliers, such as abnormally high wind speed or rainfall recorded by a monitoring station due to equipment failure. Secondly, fill in missing values, such as rainfall data missing on a certain day due to equipment failure, which can be interpolated and filled by data from neighboring monitoring stations. Finally, the data is standardized so that data of different dimensions can be compared on the same scale, such as converting wind speed and rainfall into standard scores (Z scores) respectively. The purpose of constructing a nonlinear disaster loss assessment model containing interaction terms is to more accurately reflect the combined impact of typhoons and rainstorms. For example, the model not only contains the two main effect terms of typhoon intensity and rainstorm intensity, but also their interaction terms (such as typhoon intensity × rainstorm intensity). This interaction term can capture the nonlinear effects of typhoons and rainstorms. For example, when strong typhoons are accompanied by heavy rainfall, disaster losses will increase significantly. The purpose of training the model with machine learning algorithms such as support vector machines (SVM) or random forests (RF) is to find the optimal model parameters. For example, when using the random forest algorithm, the number of trees, the depth of the trees and other parameters are adjusted to find the parameter combination that makes the model perform best on historical data. During the training process, the cross-validation method is used to divide the data into training sets and validation sets to ensure the generalization ability of the model. The trained model is used to predict new typhoon and rainstorm data to obtain the comprehensive impact assessment results of the compound disaster. For example, when the new typhoon "Maria" is about to land, the meteorological department predicts that its maximum wind speed will be 40 meters per second and the 24-hour rainfall will be 200 mm. These data are input into the model, and the model outputs the comprehensive impact assessment results of the typhoon-rainstorm compound disaster, such as the expected economic losses and casualties. The evaluation results are analyzed to determine whether their accuracy meets the preset threshold in order to ensure the reliability of the model. For example, if the accuracy threshold of the model is set to 90%, and the accuracy of the model on the validation set is only 85%, it does not meet the requirements. At this time, it is necessary to return to step 3 and adjust the model structure or parameters, such as adding more interaction items or replacing the machine learning algorithm, until the accuracy of the model reaches the preset threshold. Applying the disaster damage assessment model that meets the accuracy requirements to the actual typhoon and rainstorm disaster warning and loss assessment and storm surge warning can provide decision support for disaster prevention and mitigation. For example, before Typhoon Maria landed, the model predicted that the disaster damage level in the eastern district of a coastal city was "serious". Based on this, government departments deployed rescue forces in advance to evacuate residents in high-risk areas and reduce casualties and property losses.Through this series of steps, the disaster loss assessment model can not only more accurately reflect the combined impact of typhoons and heavy rains, but also provide a scientific basis for disaster prevention and mitigation. Data preprocessing ensures the quality of model input, the introduction of interaction terms improves the model's fitting ability, the application of machine learning algorithms optimizes the model's parameters, and the accuracy verification of the assessment results ensures the reliability of the model. The final practical application converts the value of the model into actual disaster prevention and mitigation effects. This comprehensive disaster loss assessment method significantly improves disaster response capabilities through various technical means and rigorous implementation steps. Data preprocessing and standardization ensure the availability and consistency of data, the introduction of interaction terms and the application of machine learning algorithms improve the model's prediction accuracy, and accuracy verification ensures the reliability of the model. The final practical application maximizes the value of the model and provides strong decision-making support for disaster prevention and mitigation.

[0055] Step S105, dynamically adjusting the loss probability distribution based on the real-time flood disaster assessment result to obtain a comprehensive disaster loss assessment result;

[0056] Among them, in view of the uncertainty of complex disasters, the disaster loss assessment model adopts a probabilistic statistical method to estimate the distribution of disaster losses under different scenarios based on the probability distribution of disaster events; obtains real-time monitoring data of the disaster area, dynamically adjusts the probability of disaster scenarios, and obtains a dynamically changing loss probability distribution.

[0057] Furthermore, based on the historical disaster data of the disaster area, a probability distribution model of compound disaster events is established to obtain the probability of occurrence of different disaster scenarios. For each disaster scenario, the Monte Carlo simulation method is used to estimate the loss distribution under the scenario based on the disaster damage function. Real-time monitoring data of the disaster area, including multi-source heterogeneous data such as meteorology, geology, and hydrology, is obtained and pre-processed through data fusion technology. According to the real-time monitoring data, the probability distribution model of disaster events is dynamically updated to obtain the real-time changes in the probability of occurrence of various disaster scenarios. For disaster scenarios with changed probabilities, Monte Carlo simulation is re-performed to obtain dynamically changing loss distributions. The loss distributions of different scenarios are weighted averaged, with the weights being the dynamic probability of occurrence of each scenario, to obtain a comprehensive disaster loss assessment result. Machine learning algorithms such as support vector machines and random forests are used to train prediction models based on the results of disaster loss assessment to achieve rapid loss assessment of new disaster scenarios.

[0058] As a specific implementation of this embodiment, in disaster risk management, establishing a probability distribution model for compound disaster events is a key step. For example, by analyzing historical flood data, a model is established to predict the probability of such a disaster. It is an effective technical means to use the Monte Carlo simulation method to estimate the loss distribution under a specific disaster scenario. Through a large number of random sampling, the Monte Carlo method can generate a loss distribution close to the actual situation. The acquisition and preprocessing of real-time monitoring data is another important link in disaster management. For example, through water level sensors and wind speed anemometers, water level and wind speed data can be collected in real time. After these data are processed by data fusion technology, they can be used to update the probability distribution model of disaster events, making the model closer to the current environment and conditions. Dynamically updating the probability distribution model of disaster events is a continuous process. With the continuous input of new data, the model will adjust the probability of disaster occurrence in real time. For example, if the river water rises sharply due to continuous rainfall for several days, the probability of flood occurrence will increase. This dynamic adjustment makes disaster warning more accurate and timely. When the probability of a disaster scenario changes, it is necessary to re-perform the Monte Carlo simulation. In this way, an updated loss distribution can be obtained based on the latest probability distribution, providing data support for the adjustment of disaster response measures. The key step to obtain the comprehensive disaster loss assessment result is to take the weighted average of the loss distribution of different scenarios, where the weight is the dynamic probability of each scenario. This step ensures that the assessment results can reflect the losses corresponding to the most likely disaster scenario. Finally, using machine learning algorithms such as support vector machines and random forests to train prediction models based on the disaster loss assessment results can achieve rapid loss assessment of new disaster scenarios. This method uses models trained with historical data to quickly give loss predictions when new disasters occur, helping decision makers respond quickly.

[0059] Step S106, based on the comprehensive disaster damage assessment result, the spatiotemporal evolution of the disaster situation is performed to obtain the development trend of the disaster situation;

[0060] Among them, on the basis of loss assessment, the temporal and spatial evolution process of the disaster is further analyzed, the high-risk areas and key time nodes of the disaster are identified, and the development trend of the disaster is predicted to generate a decision-making basis for disaster emergency response.

[0061] Furthermore, according to the results of disaster loss assessment, the spatiotemporal data analysis method is used to analyze the spatiotemporal evolution process of the disaster, and the distribution characteristics of the disaster in different time and space are obtained. By analyzing the spatiotemporal distribution characteristics of the disaster, the high-risk areas of the disaster are identified, the key time nodes of the disaster are determined, and the severity and impact range of the disaster are judged. Machine learning algorithms, such as time series prediction and spatial interpolation, are used to predict the development trend of the disaster and estimate the changes in the disaster in the future. According to the results of high-risk area identification and key time nodes, the key areas and key time periods of disaster emergency response are determined, and targeted emergency measures are formulated. Through the results of disaster trend prediction, the development direction and change trend of the disaster are judged, the emergency response strategy is adjusted, and rescue resources are reasonably allocated. The analysis results of disaster loss assessment, high-risk area identification, key node determination, trend prediction, etc. are integrated to form a decision-making basis for disaster emergency response and support emergency command decision-making. Visualization technology is used to present the disaster analysis and prediction results in the form of maps, curves, etc., to intuitively display the spatiotemporal evolution and development trend of the disaster, so that emergency management personnel can quickly understand the dynamics of the disaster.

[0062] As a specific implementation of this embodiment, when conducting disaster loss assessment, a spatiotemporal data model is established by analyzing historical disaster data. For example, in a coastal area, by analyzing the typhoon paths and impact ranges in the past ten years, it can be identified that certain low-lying areas are often flooded. This analysis can be performed using a geographic information system (GIS). By superimposing the data of historical disaster events on a map, it is possible to intuitively see which areas are most severely affected by disasters within a specific time. After identifying high-risk areas, time series prediction methods can be used to predict the time of occurrence and possible impact of future disasters. For example, if a certain area will experience floods every rainy season, the flood trend in the future rainy season can be predicted by analyzing the rainfall and flood occurrence time points in the past few years. This prediction can help local governments and rescue organizations prepare in advance, such as stockpiling disaster relief supplies in advance and formulating evacuation plans. After determining the key time nodes and high-risk areas of the disaster, the key areas and key time periods of the emergency response are also clarified. For example, when it is predicted that an upcoming typhoon may land in a coastal city, the city and its surrounding areas become the key areas of emergency response. The critical time period is usually a few days before and after the typhoon makes landfall, when rescue resources, such as rescue teams and disaster relief supplies, need to be deployed intensively. Emergency response strategies can be adjusted based on the disaster trend forecast results. For example, if the forecast results show that a flood will not only affect low-lying areas, but may also affect several towns upstream, then the emergency response strategy needs to be adjusted to include these newly identified high-risk areas. This adjustment ensures that rescue resources can be allocated more effectively, thereby reducing the impact of the disaster. Finally, integrating all this information and presenting it through visualization technology is the key to ensuring the efficiency and accuracy of information communication. For example, a dynamic disaster monitoring dashboard can be created to display real-time meteorological data, high-risk areas, predicted disaster development trends, and other information. Such a dashboard can not only help emergency managers quickly obtain the latest disaster information, but also help the public understand the upcoming risks and take appropriate preventive measures.

[0063] Embodiment 2

[0064] like Figure 3 As shown, this embodiment also provides a typhoon-rainstorm flood compound disaster monitoring and assessment system, which mainly includes:

[0065] The composite disaster identification module is used to obtain typhoon and rainstorm data and determine whether it constitutes a composite disaster through machine learning;

[0066] The regional modeling module is used to divide the assessment units and build a disaster loss assessment model with regional applicability;

[0067] Real-time assessment module, used to obtain real-time monitoring data and dynamically update model parameters for disaster assessment;

[0068] The coupling effect evaluation module is used to improve the evaluation accuracy by describing the multi-factor interaction through nonlinear functions;

[0069] Probabilistic statistics module, used to estimate the distribution of disaster losses under different scenarios and dynamically adjust the probability;

[0070] The disaster analysis module is used to analyze the temporal and spatial evolution of the disaster and predict development trends to provide a basis for decision-making.

[0071] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for monitoring and assessing typhoon-rainstorm-flood compound disasters, characterized in that: The following steps are involved: Based on multi-factor data, a disaster identification model is established through machine learning algorithms to obtain compound disaster judgment results; Constructing a disaster assessment model for the area constituting the disaster based on the composite disaster judgment result; Dynamically updating the parameters of the disaster assessment model based on real-time monitoring data to obtain real-time composite disaster assessment results; Based on the real-time composite disaster assessment result, dynamically adjusting the loss probability distribution to obtain a comprehensive disaster loss assessment result; Based on the comprehensive disaster loss assessment results, the spatiotemporal evolution of the disaster situation is carried out to obtain the development trend of the disaster situation.

2. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 1 is characterized in that: The disaster identification model is based on the rainfall monitoring data of heavy rain occurring on the same day within the 300-kilometer buffer zone of the typhoon path to obtain the composite disaster judgment result.

3. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 1 is characterized in that: When a complex disaster is determined, the study area is divided into several assessment units using a regional modeling strategy, and a complex disaster assessment model is constructed based on the natural geographical attributes of the assessment units.

4. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 3 is characterized in that: The process of obtaining real-time composite hazard assessment results includes: Input the real-time monitoring data of the disaster area into the disaster assessment model to obtain the loss distribution of each assessment unit; Based on the loss distribution, when the loss of the assessment unit exceeds a preset threshold, the current assessment unit is judged to be a high-risk area, and an identification result of the high-risk area is obtained; Based on the location distribution of high-risk areas and road network data, the optimal dispatch route of the rescue team is planned to obtain the rescue route plan; Based on the continuous acquisition of real-time monitoring data and the re-evaluation of the disaster damage assessment model, the rescue plan is dynamically adjusted to obtain an adjusted rescue plan.

5. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 4 is characterized in that: The process of constructing the disaster assessment model also includes describing the interaction of the multi-factor data through a nonlinear function, introducing the interaction terms of the strong winds and heavy rains brought by the typhoon; and performing storm surge warning based on the disaster assessment model with the interaction terms introduced.

6. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 1 is characterized in that: The process of obtaining comprehensive damage assessment results includes: Based on historical typhoon-rainstorm flood data, a probability distribution model of compound disaster events is established, and the probability of occurrence of different disaster scenarios is obtained; Based on the probability of occurrence of different disaster scenarios, the Monte Carlo simulation method is used to estimate the loss distribution under different disaster scenarios; Collect wind speed and water level data in real time, and update the probability distribution model of disaster events based on the wind speed and water level data to obtain an adjusted probability of disaster occurrence; Based on the adjusted disaster occurrence probability, the disaster loss distribution of different scenarios is weighted averaged to obtain a comprehensive disaster loss assessment result.

7. The typhoon-rainstorm flood combined disaster monitoring and assessment method according to claim 6 is characterized in that: Based on the comprehensive disaster loss assessment results, a spatiotemporal data analysis method is used to analyze the spatiotemporal evolution of the disaster to obtain the distribution characteristics of the disaster at different times and spaces. Based on the distribution characteristics of the disaster at different times and spaces, high-risk areas of the disaster are identified, key time nodes of the disaster are determined, and the severity and scope of the disaster are judged.

8. A typhoon-rainstorm flood combined disaster monitoring and assessment system, characterized in that: include: A composite disaster identification module, which is used to obtain typhoon and rainstorm data and determine whether it constitutes a composite disaster through machine learning; A sub-regional modeling module, the sub-regional modeling module is used to divide the assessment unit and construct a regional disaster loss assessment model; A real-time assessment module, which is used to obtain real-time monitoring data and dynamically update model parameters to perform disaster assessment; A coupling effect evaluation module, wherein the coupling effect evaluation module is used to improve the evaluation accuracy by describing the multi-factor interaction through nonlinear functions; A probability statistics module, which is used to estimate the distribution of disaster losses under different scenarios and dynamically adjust the probability; The disaster situation analysis module is used to analyze the temporal and spatial evolution of the disaster situation and predict the development trend to provide a basis for decision-making.

Images