Intelligent drought risk management method and system based on disaster attribution analysis

Abstract

The present invention discloses an intelligent drought risk management method and system based on disaster attribution analysis, and the method includes data collection, data enhancement, disaster attribution analysis, disaster dynamic risk assessment and intelligent drought risk management. The present invention relates to the field of drought risk management, and adopts an integrated method of intelligent drought risk management combining disaster attribution analysis and disaster dynamic risk assessment. By performing dynamic analysis of drought disasters from the perspective of attribution, and combining multiple feature optimization models to perform multi-factor feature mining, the complexity processing capability and real-time accuracy of intelligent drought risk management are improved, and the fusion performance of multi-domain data is also improved; an improved support vector machine model combining expert model and graph neural network feature optimization is used to perform disaster attribution analysis; and a gated recurrent unit combining generative adversarial network and graph convolution network feature deepening is used to perform dynamic disaster risk assessment.

Description

Technical Field

[0001] The present invention relates to the field of drought risk management, and in particular to an intelligent drought risk management method and system based on disaster attribution analysis. Background Art

[0002] The intelligent drought risk management method and system based on disaster attribution analysis is to conduct attribution analysis on the multi-dimensional factors of drought occurrence by comprehensively utilizing multi-source data and advanced machine learning and deep learning technologies, so as to identify the main causes and influencing factors of drought. Through disaster attribution analysis, this method can quantitatively reveal the contribution of different factors to drought risk, and provide a scientific basis for real-time assessment of the dynamic changes of drought. Its core role is to improve the accuracy and timeliness of drought risk prediction, help decision makers accurately identify the potential threats of drought, and provide intelligent support for drought emergency response. In addition, the system can also dynamically adjust the risk assessment results according to real-time data, realize continuous monitoring and effective management of drought risks, thereby optimizing resource allocation, reducing losses caused by drought and enhancing disaster response capabilities.

[0003] However, in the existing intelligent drought risk management, there is a problem that the traditional drought risk analysis process relies on statistical analysis and a single machine learning model, which cannot effectively deal with the spatiotemporal variability and complexity of drought risk. At the same time, the existing methods also lack the fusion analysis of multi-dimensional factors, which makes it impossible to fully explore the potential correlation between various factors, and then leads to the technical problem of insufficient accuracy and reliability of drought risk assessment; in the existing disaster attribution analysis methods, there is a problem that traditional disaster attribution often relies on manual analysis and screening and refinement in the feature extraction stage. However, manual methods inevitably have empirical errors, which further leads to the problem that the intelligence of disaster attribution analysis needs to be improved. At the same time, traditional intelligent methods are also difficult to accept complex multi-factor features to form effective attribution outputs, resulting in the technical problem that the intelligence, usability and practicality of existing technologies for drought disaster attribution need to be improved; in the existing disaster dynamic risk assessment, there is a problem that the influencing factors of drought disasters are complex and complicated, and the similarity between the technical characteristics that actually directly affect drought is low, making it difficult to effectively form regular learning, thereby greatly increasing the difficulty of intelligent disaster dynamic risk assessment. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent drought risk management method and system based on disaster attribution analysis. In the existing intelligent drought risk management, the traditional drought risk analysis process relies on statistical analysis and a single machine learning model, which cannot effectively deal with the spatiotemporal variability and complexity of drought risk. At the same time, the existing method also lacks fusion analysis of multi-dimensional factors, resulting in the inability to fully explore the potential correlation between various factors, which in turn leads to the technical problem of insufficient accuracy and reliability of drought risk assessment. This solution creatively adopts an integrated method of intelligent drought risk management that combines disaster attribution analysis and disaster dynamic risk assessment. By The paper conducts drought disaster dynamic analysis from the perspective of attribution, and combines multiple feature optimization models to mine multi-factor features, which improves the complexity processing capability and real-time accuracy of intelligent drought risk management, and also improves the fusion performance of multi-domain data. In the existing disaster attribution analysis methods, traditional disaster attribution often relies on manual analysis and screening and refinement in the feature extraction stage. However, manual methods inevitably have empirical errors, which further leads to the problem that the intelligence of disaster attribution analysis needs to be improved. At the same time, traditional intelligent methods are also difficult to accept complex multi-factor features to form effective attribution outputs, resulting in the existing technology for the intelligence, usability and practicality of drought disaster attribution. In order to solve the technical problems that need to be improved, this solution creatively adopts an improved support vector machine model that combines expert models and graph neural network feature optimization to conduct disaster attribution analysis. By introducing expert models and graph neural networks to optimize the spatiotemporal characteristics of multiple factors, and improving the final attribution effect of the support vector machine model optimization solution, the depth and analysis dimension of disaster attribution are improved, providing effective data support and functional support for subsequent drought management. In view of the fact that the influencing factors of drought disasters are complex and numerous in the existing disaster dynamic risk assessment, and the similarity between the technical characteristics that actually directly affect drought is low, it is difficult to effectively form regular learning, thereby greatly improving the intelligent disaster dynamic risk assessment. To address the technical difficulties of risk assessment, this solution creatively adopts a gated recurrent unit that combines generative adversarial networks and graph convolutional network feature deepening. Through the process of generative adversarial networks and graph convolutional network feature deepening, it learns the relationship between features and complements and generates them. Specifically, through the optimization of the spatiotemporal adjacency matrix of the graph convolutional network, the timing optimization of the generator of the generative adversarial network, and the adaptive function design of the gated recurrent unit, the feature processing and evaluation prediction of the timing network are further performed through the simulated generated drought disaster samples, which improves the accuracy of dynamic disaster risk assessment, optimizes the intelligent method's understanding of the regularity of drought risk, and provides practical experience for the intelligent management of drought disasters.

[0005] The technical solution adopted by the present invention is as follows: The intelligent drought risk management method based on disaster attribution analysis provided by the present invention comprises the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data enhancement;

[0008] Step S3: Disaster attribution analysis;

[0009] Step S4: Disaster dynamic risk assessment;

[0010] Step S5: Smart drought risk management.

[0011] Furthermore, in step S1, the data collection is used to collect the original data required for intelligent drought risk management, specifically to collect multi-dimensional data related to drought, and obtain the original data for intelligent drought risk management through data collection;

[0012] The intelligent drought risk management raw data specifically includes meteorological data, soil data, remote sensing feature data, agricultural water use data, geographic information support data and climate change data.

[0013] Further, in step S2, the data enhancement is used to optimize and enhance the original data, specifically to enhance the original data of the intelligent drought risk management to obtain enhanced intelligent drought risk management data, and specifically includes the following steps:

[0014] Step S21: data cleaning;

[0015] Step S22: missing value optimization;

[0016] Step S23: Specification enhancement;

[0017] Step S24: timing processing enhancement;

[0018] Step S25: initial feature optimization;

[0019] Step S26: Data integration and enhancement.

[0020] Further, in step S3, the disaster attribution analysis is used to identify and analyze the multi-dimensional factors that cause drought. Specifically, based on the enhanced intelligent drought risk management data, an improved support vector machine model combining an expert model and graph neural network feature optimization is used to perform disaster attribution analysis to obtain drought disaster attribution analysis reference data, which specifically includes the following steps:

[0021] Step S31: expert model attribution feature optimization, specifically using a lightweight expert model to perform disaster attribution analysis feature screening optimization to obtain expert model optimization feature data;

[0022] Step S32: constructing a graph neural feature optimization subnet, specifically, constructing a graph neural feature optimization subnet through node representation and graph convolution operation, combining the expert model optimization feature data and the enhanced intelligent drought risk management data, performing graph neural network feature optimization, and obtaining spatiotemporal optimization feature data;

[0023] The calculation formula for the improved objective function is:

[0024] ;

[0025] In the formula, F eS is the objective function of the improved support vector machine model, min is the minimum value function, w is the weight of the disaster influencing factor, which is used to indicate the influence of each disaster attribution factor in the spatiotemporal optimization feature data on the occurrence of drought disasters, o is the error variable, which is used to indicate the error in the disaster attribution analysis process, C is the soft interval parameter, which is used to indicate the control degree of the error in the training process of the disaster attribution analysis model, n is the total number of spatiotemporal optimization feature data, i is the spatiotemporal optimization feature data index, is the spatial difference balance factor, x i is the feature data optimized by the i-th expert model, is the i-th spatiotemporal optimization feature data;

[0026] Step S33: construct an improved support vector machine model, specifically construct a support vector machine model with improved objective function, and perform support vector machine classification and regression analysis based on the spatiotemporal optimization feature data to obtain support vector disaster attribution output data.

[0027] Step S34: training the disaster attribution analysis model, specifically, training the disaster attribution analysis model by optimizing the expert model attribution features, constructing the graph neural feature optimization subnet, and constructing the improved support vector machine model to obtain the disaster attribution analysis model Model RC ;

[0028] Step S35: Disaster attribution analysis, specifically, using the disaster attribution analysis model Model based on the enhanced intelligent drought risk management data RC , conduct disaster attribution analysis, and obtain reference data for drought disaster attribution analysis;

[0029] The drought disaster attribution analysis reference data specifically includes meteorological factor analysis, soil moisture factor analysis, water resource utilization factor analysis, land use analysis and climate change analysis.

[0030] Further, in step S4, the dynamic disaster risk assessment is used to dynamically assess the risk level of drought in real time, specifically, based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data, a gated recurrent unit combining generative adversarial networks and graph convolutional network feature deepening is used to perform dynamic disaster risk assessment to obtain dynamic disaster risk assessment data, specifically including the following steps:

[0031] Step S41: graph convolution feature extraction, specifically constructing a graph convolution neural network to perform time-space structured feature extraction, and extracting disaster risk features based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data;

[0032] The spatiotemporal structured feature extraction is specifically performed by introducing a spatiotemporal weighted adjacency matrix to optimize graph convolution feature extraction. The calculation formula of the spatiotemporal weighted adjacency matrix is:

[0033] ;

[0034] In the formula, is the spatiotemporal weighted adjacency matrix, which is used to replace the normalized adjacency matrix in the original graph convolutional neural network. a is the spatial feature weight, A is the spatial adjacency matrix, which is used to represent the normalized adjacency matrix in the original graph convolutional neural network. b is the temporal feature weight, W(t) is the temporal correlation matrix, and t is the time index.

[0035] Step S42: Dynamically generate risk assessment features, specifically, construct a generative adversarial network, and optimize the generator structure in the original generative adversarial network by introducing a temporal generator, perform adversarial training based on the disaster risk features, and generate risk assessment dynamic features through adversarial training;

[0036] The temporal generator specifically introduces a temporal convolution layer and a standard generator model for integration, and optimizes the extraction of temporal features through the temporal convolution layer. The calculation formula is:

[0037] ;

[0038] Where, X dy is the generated risk assessment dynamic feature, TSG(·) is the temporal generator operation function, which is used to represent the integration of the temporal convolutional layer and the standard generator model, X G is the disaster risk characteristic, t is the time index;

[0039] Step S43: constructing a gated cyclic feature analysis subnet, specifically constructing a gated cyclic unit including a reset gate and an update gate, performing disaster dynamic risk feature analysis, and combining the factor analysis data in the drought disaster attribution analysis reference data, introducing an adaptive gating function, performing time series feature analysis, and obtaining time series drought risk change law feature data;

[0040] The adaptive gating function is used to improve the computational adaptability of the reset gate and the update gate, and the calculation formula is:

[0041] ;

[0042] In the formula, The reset gate output after optimization by the adaptive gating function, sig(·) is the S-type activation function, W r is the reset gate weight value, [·,·,·] is the feature connection operator, and h t-1 is the hidden state of the previous moment, t is the time index, x t It is the raw data input of the gated loop feature analysis subnet, which is used to represent the risk assessment dynamic feature data. r is to reset the gate adaptation parameters, b r is the reset gate bias term, is the update gate output after optimization by the adaptive gating function, W z Is to update the gate weight value, Adapt z is the update gate adaptive parameter, b z is the update gate bias term, r t is the original reset gate output without optimization of the adaptive gating function, M is the total number of adaptive factors, which is used to represent the types of factors in the drought disaster attribution analysis reference data, is the attribution influence parameter of the mth adaptive factor, which is used to indicate the sharing degree of drought disaster caused by the attribution parameter, z t is the original update gate output without optimization by the adaptive gating function, is the update gate adjustment factor, the specific value range is [0.5,1], and the parameter is adjusted by cross-validation method;

[0043] Step S44: training a dynamic disaster risk assessment model, specifically, training a dynamic disaster risk assessment model by extracting graph convolution features, dynamically generating risk assessment features, and constructing a gated cyclic feature analysis subnet to obtain a dynamic disaster risk assessment model Model AP ;

[0044] Step S45: Disaster dynamic risk assessment, specifically, using the disaster dynamic risk assessment model Model based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data AP , conduct disaster dynamic risk assessment and obtain disaster dynamic risk assessment data;

[0045] The disaster dynamic risk assessment data specifically includes risk level data, drought risk factor analysis data and risk change characteristic data.

[0046] Further, in step S5, the intelligent drought risk management is used to perform comprehensive drought risk management in combination with disaster attribution and risk assessment, specifically, combining the drought disaster attribution analysis reference data and the disaster dynamic risk assessment data to perform intelligent drought risk management and obtain comprehensive reference data for intelligent drought risk management.

[0047] The intelligent drought risk management system based on disaster attribution analysis provided by the present invention comprises a data acquisition module, a data enhancement module, a disaster attribution analysis module, a disaster dynamic risk assessment module and an intelligent drought risk management module;

[0048] The data acquisition module is used for data acquisition, and obtains the original data of intelligent drought risk management through data acquisition, and sends the original data of intelligent drought risk management to the data enhancement module;

[0049] The data enhancement module is used for data enhancement, and obtains enhanced intelligent drought risk management data through data enhancement, and sends the enhanced intelligent drought risk management data to the disaster attribution analysis module and the disaster dynamic risk assessment module;

[0050] The disaster attribution analysis module is used for disaster attribution analysis, obtains drought disaster attribution analysis reference data through disaster attribution analysis, and sends the drought disaster attribution analysis reference data to the disaster dynamic risk assessment module and the intelligent drought risk management module;

[0051] The disaster dynamic risk assessment module is used for disaster dynamic risk assessment, obtains disaster dynamic risk assessment data through disaster dynamic risk assessment, and sends the disaster dynamic risk assessment data to the intelligent drought risk management module;

[0052] The intelligent drought risk management module is used for intelligent drought risk management, and through intelligent drought risk management, comprehensive reference data for intelligent drought risk management is obtained.

[0053] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0054] (1) In the existing intelligent drought risk management, the traditional drought risk analysis process relies on statistical analysis and a single machine learning model, which cannot effectively deal with the spatiotemporal variability and complexity of drought risk. At the same time, the existing methods also lack the fusion analysis of multi-dimensional factors, which makes it impossible to fully explore the potential correlation between various factors, and thus leads to the insufficient accuracy and reliability of drought risk assessment. This solution creatively adopts an integrated intelligent drought risk management method that combines disaster attribution analysis and disaster dynamic risk assessment. By conducting a dynamic analysis of drought disasters from the perspective of attribution and combining multiple feature optimization models to mine multi-factor features, the complexity processing capability and real-time accuracy of intelligent drought risk management are improved, and the fusion performance of multi-domain data is also improved;

[0055] (2) In the existing disaster attribution analysis methods, traditional disaster attribution often relies on manual analysis and screening and refinement in the feature extraction stage. However, manual methods are inevitably prone to empirical errors, which further leads to the problem that the intelligence of disaster attribution analysis needs to be improved. At the same time, traditional intelligent methods are also difficult to accept complex multi-factor features to form effective attribution outputs, resulting in the technical problem that the intelligence, usability and practicality of existing technologies for drought disaster attribution need to be improved. This scheme creatively adopts an improved support vector machine model that combines expert models and graph neural network feature optimization to conduct disaster attribution analysis. By introducing expert models and graph neural networks to optimize the spatiotemporal characteristics of multiple factors, and by improving the final attribution effect of the support vector machine model optimization scheme, the depth and analysis dimension of disaster attribution are improved, providing effective data support and functional support for subsequent drought management;

[0056] (3) In view of the fact that the influencing factors of drought disasters are complex and numerous in the existing dynamic risk assessment of disasters, and the similarity between the technical features that actually directly affect drought is low, it is difficult to effectively form regularity learning, which greatly increases the difficulty of intelligent dynamic risk assessment of disasters. This scheme creatively adopts a gated recurrent unit that combines generative adversarial networks and graph convolutional network feature deepening. Through the process of generative adversarial networks and graph convolutional network feature deepening, the relationship between features is learned and supplemented and generated. Specifically, through the optimization of the spatiotemporal adjacency matrix of the graph convolutional network, the timing optimization of the generator of the generative adversarial network, and the adaptive function design of the gated recurrent unit, the drought disaster samples generated by simulation are further processed and evaluated and predicted by the timing network, which improves the accuracy of dynamic risk assessment of disasters and optimizes the regularity of drought risk by intelligent methods, providing practical experience for the intelligent management of drought disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1A schematic diagram of a process flow of an intelligent drought risk management method based on disaster attribution analysis provided by the present invention;

[0058] Figure 2 A schematic diagram of an intelligent drought risk management system based on disaster attribution analysis provided by the present invention;

[0059] Figure 3 Schematic diagram of the process of data enhancement in step S2;

[0060] Figure 4 This is a flow chart of disaster attribution analysis in step S3;

[0061] Figure 5 This is a flow chart of step S4 disaster dynamic risk assessment.

[0062] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0064] Example 1, see Figure 1 The present invention provides an intelligent drought risk management method based on disaster attribution analysis, which comprises the following steps:

[0065] Step S1: data collection;

[0066] Step S2: data enhancement;

[0067] Step S3: Disaster attribution analysis;

[0068] Step S4: Disaster dynamic risk assessment;

[0069] Step S5: Smart drought risk management.

[0070] By performing the above operations, in view of the fact that in the existing intelligent drought risk management, the traditional drought risk analysis process relies on statistical analysis and a single machine learning model, which cannot effectively deal with the spatiotemporal variability and complexity of drought risk. At the same time, the existing methods also lack the fusion analysis of multi-dimensional factors, which makes it impossible to fully explore the potential correlation between various factors, and thus leads to the insufficient accuracy and reliability of drought risk assessment. Technical problems, this scheme creatively adopts an integrated method of intelligent drought risk management that combines disaster attribution analysis and disaster dynamic risk assessment. By conducting a dynamic analysis of drought disasters from the perspective of attribution, and combining a variety of feature optimization models for multi-factor feature mining, the complexity processing capability and real-time accuracy of intelligent drought risk management are improved, and the fusion performance of multi-field data is also improved.

[0071] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data required for intelligent drought risk management, specifically to collect multi-dimensional data related to drought, and obtain the original data for intelligent drought risk management through data collection;

[0072] The original data of intelligent drought risk management specifically include meteorological data, soil data, remote sensing characteristic data, agricultural water use data, geographic information support data and climate change data;

[0073] The meteorological data specifically includes temperature, precipitation, humidity, wind speed and air pressure data;

[0074] The soil data specifically includes soil moisture, soil water content and soil temperature data;

[0075] The remote sensing characteristic data specifically include drought detection characteristics, land use characteristics and surface moisture information data;

[0076] The agricultural water use data specifically include irrigation water volume, agricultural water resource allocation and agricultural water consumption data;

[0077] The geographic information support data specifically include land use type, terrain information, watershed division, water body distribution and population density characteristic data;

[0078] The climate change data mentioned above specifically refers to long-term climate change reference data.

[0079] Example 3, see Figure 1 , Figure 2 and Figure 3This embodiment is based on the above embodiment. In step S2, the data enhancement is used to optimize and enhance the original data, specifically to enhance the original data of the intelligent drought risk management to obtain enhanced intelligent drought risk management data, and specifically includes the following steps:

[0080] Step S21: data cleaning, specifically, cleaning the raw data of the intelligent drought risk management by data deduplication and outlier statistical detection to obtain cleaned drought risk management data;

[0081] Step S22: missing value optimization, specifically, clustering and filling the missing data in the cleaned drought risk management data to obtain missing value optimized drought risk management data;

[0082] Step S23: specification enhancement, specifically, performing Z-score standardization on the numerical data in the missing value optimized drought risk management data, and performing normalization encoding on the categorical data to obtain specification enhanced drought risk management data;

[0083] Step S24: time series processing enhancement, specifically, performing time series smoothing and difference processing enhancement on the data in the specification enhanced drought risk management data to obtain time series optimized drought risk management data;

[0084] Step S25: initial feature optimization, specifically, using an artificial feature selection and screening method to perform preliminary feature screening on the time series optimized drought risk management data to obtain initial optimized feature data;

[0085] Step S26: data integration enhancement, specifically, data enhancement and data integration and fusion are performed through the data cleaning, the missing value optimization, the specification enhancement, the time series processing enhancement and the feature initial optimization to obtain enhanced intelligent drought risk management data.

[0086] Example 4, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S3, the disaster attribution analysis is used to identify and analyze the multi-dimensional factors that cause drought. Specifically, based on the enhanced intelligent drought risk management data, an improved support vector machine model combining an expert model and graph neural network feature optimization is used to perform disaster attribution analysis to obtain drought disaster attribution analysis reference data. Specifically, the following steps are included:

[0087] Step S31: expert model attribution feature optimization, specifically using a lightweight expert model to perform disaster attribution analysis feature screening optimization to obtain expert model optimization feature data;

[0088] The lightweight expert model specifically adopts the DeepSeekV2Lite model;

[0089] The expert model optimizes the feature data, specifically by calculating through mutual information feature selection, and the calculation formula is:

[0090] ;

[0091] In the formula, MI(X j ,Y) is the output of the mutual information feature selection calculation function, which is used to represent the value of mutual information. By screening the disaster attribution analysis features with large mutual information values, feature screening optimization is performed to obtain the expert model optimization feature data, where X j is the jth disaster attribution analysis feature, used to represent the initial optimized feature data in the jth enhanced intelligent drought risk management data, j is the disaster attribution analysis feature index, Y is the disaster parameter, used as the target variable for mutual information feature selection calculation and represents the probability value of disaster occurrence, F expert (X) is the expert model optimization feature output, wherein X is the original feature data, which is used to represent the initial optimized feature data in the enhanced intelligent drought risk management data, and x j is the disaster attribution analysis feature after optimization by the j-th expert model, which is used to represent the optimized feature output of the j-th expert model. P(·) is the probability distribution calculation function. P(x j ,Y) specifically represents the disaster attribution analysis feature x after the jth expert model is optimized j and the joint probability distribution of the disaster parameter Y, P(x j ) specifically represents the disaster attribution analysis feature x after the jth expert model is optimized j The marginal probability distribution of , P(Y) specifically represents the marginal probability distribution of the disaster parameter Y, and DeepSeekV2Lite(·) is a lightweight expert model optimization function;

[0092] Step S32: constructing a graph neural feature optimization subnet, specifically, constructing a graph neural feature optimization subnet through node representation and graph convolution operation, combining the expert model optimization feature data and the enhanced intelligent drought risk management data, performing graph neural network feature optimization, and obtaining spatiotemporal optimization feature data;

[0093] Step S33: constructing an improved support vector machine model, specifically constructing a support vector machine model with improved objective function, and performing support vector machine classification and regression analysis based on the spatiotemporal optimization feature data to obtain support vector disaster attribution output data;

[0094] The calculation formula for the improved objective function is:

[0095] ;

[0096] In the formula, F eS is the objective function of the improved support vector machine model, min is the minimum value function, w is the weight of the disaster influencing factor, which is used to indicate the influence of each disaster attribution factor in the spatiotemporal optimization feature data on the occurrence of drought disasters, o is the error variable, which is used to indicate the error in the disaster attribution analysis process, C is the soft interval parameter, which is used to indicate the control degree of the error in the training process of the disaster attribution analysis model, n is the total number of spatiotemporal optimization feature data, i is the spatiotemporal optimization feature data index, is the spatial difference balance factor, x i is the feature data optimized by the i-th expert model, is the i-th spatiotemporal optimization feature data;

[0097] Step S34: training the disaster attribution analysis model, specifically, training the disaster attribution analysis model by optimizing the expert model attribution features, constructing the graph neural feature optimization subnet, and constructing the improved support vector machine model to obtain the disaster attribution analysis model Model RC ;

[0098] Step S35: Disaster attribution analysis, specifically, using the disaster attribution analysis model Model based on the enhanced intelligent drought risk management data RC , conduct disaster attribution analysis, and obtain reference data for drought disaster attribution analysis;

[0099] The reference data for drought disaster attribution analysis specifically include meteorological factor analysis, soil moisture factor analysis, water resource utilization factor analysis, land utilization analysis and climate change analysis;

[0100] The data form of the drought disaster attribution analysis reference data is specifically quantitative percentage data of factor contribution, and the value range is [0,100%].

[0101] By performing the above operations, in the existing disaster attribution analysis methods, the traditional disaster attribution often relies on manual analysis and screening and refinement in the feature extraction stage. However, manual methods inevitably have empirical errors, which further leads to the problem that the intelligence of disaster attribution analysis needs to be improved. At the same time, traditional intelligent methods are also difficult to accept complex multi-factor characteristics to form effective attribution outputs, resulting in the technical problem that the intelligence, usability and practicality of existing technologies for drought disaster attribution need to be improved. This scheme creatively adopts an improved support vector machine model that combines expert models and graph neural network feature optimization to conduct disaster attribution analysis. By introducing expert models and graph neural networks to optimize the spatiotemporal characteristics of multiple factors, and by improving the final attribution effect of the support vector machine model optimization scheme, the depth and analysis dimension of disaster attribution are improved, providing effective data support and functional support for subsequent drought management.

[0102] Example 5, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment. In step S4, the dynamic disaster risk assessment is used to dynamically assess the risk level of drought in real time. Specifically, based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data, a gated recurrent unit combining generative adversarial networks and graph convolutional network feature deepening is used to perform dynamic disaster risk assessment to obtain dynamic disaster risk assessment data. Specifically, the following steps are included:

[0103] Step S41: graph convolution feature extraction, specifically constructing a graph convolution neural network to perform time-space structured feature extraction, and extracting disaster risk features based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data;

[0104] The spatiotemporal structured feature extraction is specifically performed by introducing a spatiotemporal weighted adjacency matrix to optimize graph convolution feature extraction. The calculation formula of the spatiotemporal weighted adjacency matrix is:

[0105] ;

[0106] In the formula, is the spatiotemporal weighted adjacency matrix, which is used to replace the normalized adjacency matrix in the original graph convolutional neural network. a is the spatial feature weight, A is the spatial adjacency matrix, which is used to represent the normalized adjacency matrix in the original graph convolutional neural network. b is the temporal feature weight, W(t) is the temporal correlation matrix, and t is the time index.

[0107] Step S42: Dynamically generate risk assessment features, specifically, construct a generative adversarial network, and optimize the generator structure in the original generative adversarial network by introducing a temporal generator, perform adversarial training based on the disaster risk features, and generate risk assessment dynamic features through adversarial training;

[0108] The temporal generator specifically introduces a temporal convolution layer and a standard generator model for integration, and optimizes the extraction of temporal features through the temporal convolution layer. The calculation formula is:

[0109] ;

[0110] In the formula, X dy is the generated risk assessment dynamic feature, TSG(·) is the temporal generator operation function, which is used to represent the integration of the temporal convolutional layer and the standard generator model, X G is the disaster risk characteristic, t is the time index;

[0111] Step S43: constructing a gated cyclic feature analysis subnet, specifically constructing a gated cyclic unit including a reset gate and an update gate, performing disaster dynamic risk feature analysis, and combining the factor analysis data in the drought disaster attribution analysis reference data, introducing an adaptive gating function, performing time series feature analysis, and obtaining time series drought risk change law feature data;

[0112] The adaptive gating function is used to improve the computational adaptability of the reset gate and the update gate, and the calculation formula is:

[0113] ;

[0114] In the formula, The reset gate output after optimization by the adaptive gating function, sig(·) is the S-type activation function, W r is the reset gate weight value, [·,·,·] is the feature connection operator, and h t-1 is the hidden state of the previous moment, t is the time index, x t It is the raw data input of the gated loop feature analysis subnet, which is used to represent the risk assessment dynamic feature data. r is to reset the gate adaptation parameters, b r is the reset gate bias term, is the update gate output after optimization by the adaptive gating function, W z Is to update the gate weight value, Adapt z is the update gate adaptive parameter, b z is the update gate bias term, r tis the original reset gate output without optimization of the adaptive gating function, M is the total number of adaptive factors, which is used to represent the factor types in the drought disaster attribution analysis reference data, and its specific value is 5, m is the adaptive factor index, and according to the value of the adaptive factor index m, it is used to represent meteorological factor analysis, soil moisture factor analysis, water resource utilization factor analysis, land use analysis and climate change analysis in turn, is the attribution influence parameter of the mth adaptive factor, which is used to indicate the sharing degree of drought disaster caused by the attribution parameter, z t is the original update gate output without optimization by the adaptive gating function, is the update gate adjustment factor, the specific value range is [0.5,1], and the parameter is adjusted by cross-validation method;

[0115] Step S44: training a dynamic disaster risk assessment model, specifically, training a dynamic disaster risk assessment model by extracting graph convolution features, dynamically generating risk assessment features, and constructing a gated cyclic feature analysis subnet to obtain a dynamic disaster risk assessment model Model AP ;

[0116] Step S45: Disaster dynamic risk assessment, specifically, using the disaster dynamic risk assessment model Model based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data AP , conduct disaster dynamic risk assessment and obtain disaster dynamic risk assessment data;

[0117] The disaster dynamic risk assessment data specifically includes risk level data, drought risk factor analysis data and risk change characteristic data.

[0118] By executing the above operations, in view of the fact that in the existing dynamic risk assessment of disasters, the influencing factors of drought disasters are complex and numerous, and the similarity between the technical features that actually directly affect droughts is low, making it difficult to effectively form regular learning, thereby greatly increasing the difficulty of intelligent dynamic risk assessment of disasters. This solution creatively adopts a gated recurrent unit that combines generative adversarial networks and graph convolutional network feature deepening. Through the process of generative adversarial networks and graph convolutional network feature deepening, the relationship between features is learned and supplemented and generated. Specifically, through the optimization of the spatiotemporal adjacency matrix of the graph convolutional network, the timing optimization of the generator of the generative adversarial network, and the adaptive function design of the gated recurrent unit, the feature processing and evaluation prediction of the timing network are further performed through the simulated generated drought disaster samples, which improves the accuracy of dynamic risk assessment of disasters, optimizes the regular understanding of drought risks by intelligent methods, and provides practical experience for the intelligent management of drought disasters.

[0119] Example 6, see Figure 1 , Figure 2 This embodiment is based on the above embodiment. In step S5, the intelligent drought risk management is used to perform comprehensive drought risk management in combination with disaster attribution and risk assessment. Specifically, the drought disaster attribution analysis reference data and the disaster dynamic risk assessment data are combined to perform intelligent drought risk management and obtain comprehensive reference data for intelligent drought risk management.

[0120] Embodiment 7, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the intelligent drought risk management system based on disaster attribution analysis provided by the present invention includes a data acquisition module, a data enhancement module, a disaster attribution analysis module, a disaster dynamic risk assessment module and an intelligent drought risk management module;

[0121] The data acquisition module is used for data acquisition, and obtains the original data of intelligent drought risk management through data acquisition, and sends the original data of intelligent drought risk management to the data enhancement module;

[0122] The data enhancement module is used for data enhancement, and obtains enhanced intelligent drought risk management data through data enhancement, and sends the enhanced intelligent drought risk management data to the disaster attribution analysis module and the disaster dynamic risk assessment module;

[0123] The disaster attribution analysis module is used for disaster attribution analysis, obtains drought disaster attribution analysis reference data through disaster attribution analysis, and sends the drought disaster attribution analysis reference data to the disaster dynamic risk assessment module and the intelligent drought risk management module;

[0124] The disaster dynamic risk assessment module is used for disaster dynamic risk assessment, obtains disaster dynamic risk assessment data through disaster dynamic risk assessment, and sends the disaster dynamic risk assessment data to the intelligent drought risk management module;

[0125] The intelligent drought risk management module is used for intelligent drought risk management, and through intelligent drought risk management, comprehensive reference data for intelligent drought risk management is obtained.

[0126] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0127] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0128] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An intelligent drought risk management method based on disaster attribution analysis, characterized by: The method comprises the following steps: Step S1: data collection, obtaining raw data for intelligent drought risk management; Step S2: data enhancement to obtain enhanced intelligent drought risk management data; Step S3: disaster attribution analysis, used to identify and analyze the multi-dimensional factors that cause drought, specifically based on the enhanced intelligent drought risk management data, using an improved support vector machine model that combines expert model and graph neural network feature optimization to perform disaster attribution analysis, and obtain drought disaster attribution analysis reference data, specifically including the following steps: step S31: expert model attribution feature optimization; step S32: constructing a graph neural feature optimization subnet; step S33: constructing an improved support vector machine model; step S34: disaster attribution analysis model training; step S35: disaster attribution analysis; Step S33: constructing an improved support vector machine model, specifically constructing a support vector machine model with improved objective function, and performing support vector machine classification and regression analysis based on spatiotemporal optimization feature data to obtain support vector disaster attribution output data; The calculation formula for the improved objective function is: In the formula, F eS is the objective function of the improved support vector machine model, min is the minimum value function, w is the weight of the disaster influencing factor, which is used to indicate the influence of each disaster attribution factor in the spatiotemporal optimization feature data on the occurrence of drought disasters, o is the error variable, which is used to indicate the error in the disaster attribution analysis process, C is the soft interval parameter, which is used to indicate the control degree of the error in the training process of the disaster attribution analysis model, n is the total number of spatiotemporal optimization feature data, i is the spatiotemporal optimization feature data index, λ is the spatial difference balance factor, x i is the feature data optimized by the i-th expert model, is the i-th spatiotemporal optimization feature data; Step S4: dynamic disaster risk assessment, used for real-time dynamic assessment of drought risk level, specifically based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data, using a gated cyclic unit that combines generative adversarial networks and graph convolutional network feature deepening to perform dynamic disaster risk assessment, and obtain dynamic disaster risk assessment data, specifically including the following steps: step S41: graph convolution feature extraction; step S42: dynamic generation of risk assessment features; step S43: construction of a gated cyclic feature analysis subnet; step S44: training of a dynamic disaster risk assessment model; step S45: dynamic disaster risk assessment; Step S5: Intelligent drought risk management, obtaining comprehensive reference data for intelligent drought risk management.

2. The intelligent drought risk management method based on disaster attribution analysis according to claim 1 is characterized in that: In step S1, the intelligent drought risk management raw data specifically includes meteorological data, soil data, remote sensing feature data, agricultural water use data, geographic information support data and climate change data.

3. The intelligent drought risk management method based on disaster attribution analysis according to claim 2 is characterized in that: In step S2, the data enhancement is used to optimize and enhance the original data, specifically to enhance the original data of the intelligent drought risk management to obtain enhanced intelligent drought risk management data, and specifically includes the following steps: step S21: data cleaning; step S22: missing value optimization; step S23: specification enhancement; step S24: time series processing enhancement; step S25: feature initial optimization; step S26: data integration enhancement.

4. The intelligent drought risk management method based on disaster attribution analysis according to claim 3 is characterized in that: In step S3, the disaster attribution analysis is used to identify and analyze the multi-dimensional factors that cause drought. Specifically, based on the enhanced intelligent drought risk management data, an improved support vector machine model combining an expert model and graph neural network feature optimization is used to perform disaster attribution analysis to obtain drought disaster attribution analysis reference data, which specifically includes the following steps: Step S31: expert model attribution feature optimization, specifically using a lightweight expert model to perform disaster attribution analysis feature screening optimization to obtain expert model optimization feature data; Step S32: constructing a graph neural feature optimization subnet, specifically, constructing a graph neural feature optimization subnet through node representation and graph convolution operation, combining the expert model optimization feature data and the enhanced intelligent drought risk management data, performing graph neural network feature optimization, and obtaining spatiotemporal optimization feature data; Step S34: training the disaster attribution analysis model, specifically, training the disaster attribution analysis model by optimizing the expert model attribution features, constructing the graph neural feature optimization subnet, and constructing the improved support vector machine model to obtain the disaster attribution analysis model Model RC ; Step S35: Disaster attribution analysis, specifically, using the disaster attribution analysis model Model based on the enhanced intelligent drought risk management data RC , conduct disaster attribution analysis, and obtain reference data for drought disaster attribution analysis.

5. The intelligent drought risk management method based on disaster attribution analysis according to claim 4 is characterized in that: The drought disaster attribution analysis reference data specifically includes meteorological factor analysis, soil moisture factor analysis, water resource utilization factor analysis, land use analysis and climate change analysis.

6. The intelligent drought risk management method based on disaster attribution analysis according to claim 5 is characterized in that: In step S4, the dynamic disaster risk assessment is used to dynamically assess the risk level of drought in real time. Specifically, based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data, a gated recurrent unit combining generative adversarial networks and graph convolutional network feature deepening is used to perform dynamic disaster risk assessment to obtain dynamic disaster risk assessment data, which specifically includes the following steps: Step S41: graph convolution feature extraction, specifically constructing a graph convolution neural network to perform time-space structured feature extraction, and extracting disaster risk features based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data; The spatiotemporal structured feature extraction is specifically performed by introducing a spatiotemporal weighted adjacency matrix to optimize graph convolution feature extraction. The calculation formula of the spatiotemporal weighted adjacency matrix is: A′=aA+bW(t); Where A′ is the spatiotemporal weighted adjacency matrix, which is used to replace the normalized adjacency matrix in the original graph convolutional neural network, a is the spatial feature weight, A is the spatial adjacency matrix, which is used to represent the normalized adjacency matrix in the original graph convolutional neural network, b is the temporal feature weight, W(t) is the temporal correlation matrix, and t is the time index; Step S42: Dynamically generate risk assessment features, specifically, construct a generative adversarial network, and optimize the generator structure in the original generative adversarial network by introducing a temporal generator, perform adversarial training based on the disaster risk features, and generate risk assessment dynamic features through adversarial training; The temporal generator specifically introduces a temporal convolution layer and a standard generator model for integration, and optimizes the extraction of temporal features through the temporal convolution layer. The calculation formula is: X dy =TSG(X G ,t); Where, X dy is the generated risk assessment dynamic feature, TSG(·) is the temporal generator operation function, which is used to represent the integration of the temporal convolutional layer and the standard generator model, X G is the disaster risk characteristic, t is the time index; Step S43: constructing a gated cyclic feature analysis subnet, specifically constructing a gated cyclic unit including a reset gate and an update gate, performing disaster dynamic risk feature analysis, and combining the factor analysis data in the drought disaster attribution analysis reference data, introducing an adaptive gating function, performing time series feature analysis, and obtaining time series drought risk change law feature data; The adaptive gating function is used to improve the computational adaptability of the reset gate and the update gate, and the calculation formula is: In the formula, r′ t The reset gate output after the adaptive gating function optimization, sig(·) is the S-type activation function, W r is the reset gate weight value, [·,·,·] is the feature connection operator, and h t-1 is the hidden state of the previous moment, t is the time index, x t It is the raw data input of the gated loop feature analysis subnet, which is used to represent the risk assessment dynamic feature data. r is to reset the gate adaptation parameters, b r is the reset gate bias term, z′ t is the update gate output after optimization by the adaptive gating function, W z Is to update the gate weight value, Adapt z is the update gate adaptive parameter, b z is the update gate bias term, r t is the original reset gate output without optimization of the adaptive gating function, M is the total number of adaptive factors, which is used to represent the types of factors in the drought disaster attribution analysis reference data, m is the adaptive factor index, α m is the attribution influence parameter of the mth adaptive factor, which is used to indicate the sharing degree of drought disaster caused by the attribution parameter. t is the original update gate output without adaptive gating function optimization, β is the update gate adjustment factor, the specific value range is [0.5, 1], and the parameter is adjusted by cross-validation method; Step S44: training a dynamic disaster risk assessment model, specifically, training a dynamic disaster risk assessment model by extracting graph convolution features, dynamically generating risk assessment features, and constructing a gated cyclic feature analysis subnet to obtain a dynamic disaster risk assessment model Model AP ; Step S45: Disaster dynamic risk assessment, specifically, using the disaster dynamic risk assessment model Model based on the enhanced intelligent drought risk management data and the drought disaster attribution analysis reference data AP , conduct disaster dynamic risk assessment and obtain disaster dynamic risk assessment data; The disaster dynamic risk assessment data specifically includes risk level data, drought risk factor analysis data and risk change characteristic data.

7. The intelligent drought risk management method based on disaster attribution analysis according to claim 6 is characterized in that: In step S5, the intelligent drought risk management is used to perform comprehensive drought risk management in combination with disaster attribution and risk assessment, specifically, the intelligent drought risk management is performed in combination with the drought disaster attribution analysis reference data and the disaster dynamic risk assessment data to obtain comprehensive reference data for intelligent drought risk management.

8. An intelligent drought risk management system based on disaster attribution analysis, used to implement the intelligent drought risk management method based on disaster attribution analysis as described in any one of claims 1 to 7, characterized in that: It includes data collection module, data enhancement module, disaster attribution analysis module, disaster dynamic risk assessment module and intelligent drought risk management module.

9. The intelligent drought risk management system based on disaster attribution analysis according to claim 8 is characterized in that: The data acquisition module is used for data acquisition, and obtains the original data of intelligent drought risk management through data acquisition, and sends the original data of intelligent drought risk management to the data enhancement module; The data enhancement module is used for data enhancement, and obtains enhanced intelligent drought risk management data through data enhancement, and sends the enhanced intelligent drought risk management data to the disaster attribution analysis module and the disaster dynamic risk assessment module; The disaster attribution analysis module is used for disaster attribution analysis, obtains drought disaster attribution analysis reference data through disaster attribution analysis, and sends the drought disaster attribution analysis reference data to the disaster dynamic risk assessment module and the intelligent drought risk management module; The disaster dynamic risk assessment module is used for disaster dynamic risk assessment, obtains disaster dynamic risk assessment data through disaster dynamic risk assessment, and sends the disaster dynamic risk assessment data to the intelligent drought risk management module; The intelligent drought risk management module is used for intelligent drought risk management, and through intelligent drought risk management, comprehensive reference data for intelligent drought risk management is obtained.

Images