A trend analysis system and method based on multidimensional disaster warning data

Abstract

The present invention belongs to the field of disaster warning analysis technology, and specifically relates to a trend analysis system and method based on multi-dimensional disaster warning data. This invention significantly improves the accuracy and timeliness of disaster prediction by optimizing matching rules and thresholds in real time, realizes the deep integration and efficient utilization of multi-source data, and provides a scientific basis and technical support for disaster warning and emergency response. At the same time, the system has adaptive learning capabilities and can continuously optimize the model in the constantly updated data to ensure the dynamic accuracy of the prediction results and effectively improve disaster prevention and mitigation capabilities. In addition, the present invention enhances the comprehensiveness and reliability of disaster warning through comprehensive analysis of multi-dimensional data, provides decision makers with an intuitive disaster evolution situation map, helps to quickly formulate response strategies, minimizes disaster losses and improves the efficiency of emergency response.

Description

Technical Field

[0001] The present invention belongs to the technical field of disaster warning analysis, and in particular relates to a trend analysis system and method based on multi-dimensional disaster warning data. Background Art

[0002] Traditional disaster warning methods based on a single data source are no longer able to meet the needs of modern society. Traditional methods often rely solely on a single type of data, such as meteorological data or geological data, for disaster warning and analysis, but ignore the multifactorial and complex nature of disasters. For example, earthquakes may not only be related to geological activities, but may also be affected by multiple factors such as climate change and human activities. Therefore, relying solely on a single data source may lead to inaccurate warnings and even miss important disaster precursor information.

[0003] In the existing technology, there have been some attempts to integrate multi-source data into disaster warning systems, but most of them have problems such as insufficient data fusion and a single warning model, making it difficult to achieve accurate warnings. For example, data fusion only stays at the simple superposition level and lacks in-depth correlation analysis, resulting in insufficient adaptability of the warning model to complex disaster scenarios and the inability to effectively capture the evolution of disasters under the interweaving of multiple factors, thus affecting the timeliness and accuracy of the warning. Based on this, this solution provides a trend analysis method based on multi-dimensional disaster warning data to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a trend analysis system and method based on multi-dimensional disaster warning data, which can effectively integrate multi-source data such as meteorology, geology, and human activities, reveal the evolution law of disasters through deep correlation analysis, and improve the accuracy and timeliness of early warnings.

[0005] The technical solutions adopted by the present invention are as follows:

[0006] A trend analysis method based on multi-dimensional disaster warning data, comprising:

[0007] Collect multi-dimensional heterogeneous data through distributed data collection interfaces, including meteorological data, geological data, and social media data;

[0008] Establish a multi-source event correlation network, where:

[0009] Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data;

[0010] Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events;

[0011] Backtracking the event propagation chain based on the multi-source event association network to identify the disaster source area, and matching the diffusion path template of the current event with the historical similar event library;

[0012] Generate a dynamic disaster simulation sand table, use pulse diffusion ripples to represent the real-time spread of disasters, and use multi-color warning layers to overlay and display the warning levels of different disaster types;

[0013] Based on the actual disaster verification results, incremental learning optimization is performed on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network.

[0014] In a preferred solution, after the multi-dimensional heterogeneous data collection is completed, pre-processing is performed simultaneously, including:

[0015] Meteorological data were filtered using a sliding window method to remove outliers that exceeded the normal fluctuation range;

[0016] The geological data are used to extract deformation characteristics using wavelet transform method and uniformly converted into standard time series format;

[0017] Social media data is calibrated for timestamp deviations across different platforms using the Network Time Protocol to ensure data temporal consistency.

[0018] In a preferred embodiment, the step of performing spatiotemporal matching of abnormal air pressure change events in meteorological data with deformation acceleration events in geological data includes:

[0019] Extract meteorological anomaly events whose air pressure change rate exceeds a preset change threshold from meteorological data, and establish a space-time cylindrical matching window with the center point of the meteorological anomaly event as the origin;

[0020] Conduct time-frequency analysis on geological deformation data in geological data to screen out deformation acceleration events;

[0021] The deformation acceleration events are spatially located and temporally aligned within the space-time cylinder matching window, and the correlation between the deformation acceleration events and the meteorological anomaly events within the space-time cylinder is calculated.

[0022] When the correlation is greater than or equal to a preset correlation threshold, a preliminary correlation is established between the meteorological anomaly event and the deformation acceleration event;

[0023] Generate a heat map of the spatial density of meteorological and geological anomalies, calculate the probability of simultaneous occurrence of meteorological and geological anomalies in overlapping areas, and upgrade the preliminary association between meteorological and geological anomalies to a valid association when the probability of simultaneous occurrence exceeds a preset probability threshold;

[0024] Abnormal meteorological events and abnormal geological events that meet effective correlation requirements are included in the multi-source event correlation network.

[0025] In a preferred embodiment, the step of confidence matching the keywords in the social media data with the abnormal air pressure change events and the deformation acceleration events includes:

[0026] Perform natural language processing on social media data to extract keywords containing disaster semantic features and compile them into a disaster keyword library;

[0027] Cross-match the disaster keyword database with meteorological and geological anomalies in time and space, and calculate the confidence scores of keywords and events;

[0028] Filter out keyword and abnormal event pairs with confidence scores higher than the preset confidence threshold, establish the association relationship between keywords and abnormal events, and incorporate them into the multi-source event association network.

[0029] In a preferred embodiment, the step of backtracking and identifying the disaster source area based on the event propagation chain of the multi-source event correlation network includes:

[0030] In the multi-source event association network, the current disaster event is used as the propagation endpoint, and the associated event nodes are traced back step by step. A set of candidate source regions is generated based on the time difference between the associated events and the degree of spatial distance attenuation.

[0031] Double verification of the candidate source region set;

[0032] The first level of verification involves searching for previous disaster events with the same disaster characteristics through the historical case database, and matching the historical source coordinates whose spatiotemporal propagation trajectory matching degree is greater than or equal to the preset matching degree threshold;

[0033] The second level of verification uses real-time monitoring data to review the deformation of the candidate source area to verify whether the surface displacement rate exceeds the geological stability threshold. When the surface displacement rate exceeds the geological stability threshold, the corresponding candidate source area is locked.

[0034] The candidate source areas that meet both the first and second verifications are extracted, and the comprehensive confidence is calculated. The candidate source area with the highest comprehensive confidence is confirmed as the disaster source area.

[0035] In a preferred solution, the step of matching the diffusion path template of the current event based on the historical similar event library includes:

[0036] Retrieve historical events with the same disaster type identifier as the current event from the historical similar event database and extract the spatiotemporal diffusion path feature vector of the historical event. The spatiotemporal diffusion path feature vector includes the disaster propagation direction angle sequence, the disaster intensity attenuation coefficient matrix, and the cross-medium propagation node coordinate set.

[0037] Calculate the temporal similarity between the feature vector of the current event and the temporal diffusion path feature vector of the historical event, and filter out similar historical events with a temporal similarity higher than a preset threshold, as well as the temporal diffusion path templates corresponding to the similar historical events;

[0038] Keywords in social media data and confidence scores of similar historical events are mapped into path correction coefficients, and the spatiotemporal diffusion path template is adjusted by the path correction coefficient to generate a predicted diffusion path network containing multiple branch paths, which is used as the diffusion path template of the current disaster event.

[0039] In a preferred embodiment, the steps of generating a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time propagation range of the disaster, and using multi-color warning layers to overlay and display warning levels of different disaster types include:

[0040] Construct a basic framework for disaster simulation sandbox on a 3D geographic information platform, and dynamically draw disaster propagation ripples using real-time monitoring data;

[0041] According to the disaster propagation direction in the diffusion path template, the expansion speed and intensity attenuation coefficient of the pulse ripple are adjusted through the ripple advancement mechanism;

[0042] Set different color layers according to the warning level and display them in real time;

[0043] Among them, when abnormal meteorological events and abnormal geological events occur at the same time, they are displayed by superimposing two-color spiral ripples, and the ripple interval is used to represent the interaction intensity between disasters. The smaller the interval, the stronger the correlation.

[0044] In a preferred embodiment, the step of performing incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the actual disaster verification results includes:

[0045] Obtain real-time disaster spread boundary information through satellite remote sensing images and ground sensor networks, and compare it pixel by pixel with the predicted spread path network to generate a prediction error distribution map;

[0046] According to the temporal and spatial overlap in the prediction error distribution graph, the time window and spatial matching range parameters in the spatiotemporal matching rule are dynamically adjusted;

[0047] Establish a comparison mechanism between the actual disaster development speed and the predicted speed, and adjust the diffusion path template matching threshold based on the continuous comparison results;

[0048] Set the automatic optimization termination condition. When the prediction error rate after multiple consecutive optimizations is lower than the preset error threshold, stop the optimization and maintain the current optimal matching threshold and spatiotemporal matching rules.

[0049] The present invention also provides a trend analysis system based on multi-dimensional disaster warning data, using the above-mentioned trend analysis method based on multi-dimensional disaster warning data, comprising:

[0050] Data acquisition module, used to collect multi-dimensional heterogeneous data through distributed data acquisition interfaces, including meteorological data, geological data, and social media data;

[0051] The correlation network building module is used to build a multi-source event correlation network, where:

[0052] Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data;

[0053] Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events;

[0054] The diffusion prediction module is used to identify the disaster source area by tracing back the event propagation chain based on the multi-source event association network, and to match the diffusion path template of the current event with the historical similar event library;

[0055] The simulation and warning module is used to generate a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time spread of disasters, and using multi-color warning layers to overlay and display the warning levels of different disaster types;

[0056] The incremental learning module is used to perform incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the actual disaster verification results.

[0057] And, an electronic device, comprising:

[0058] at least one processor;

[0059] and a memory communicatively coupled to the at least one processor;

[0060] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the above-mentioned trend analysis method based on multi-dimensional disaster warning data.

[0061] The technical effects achieved by the present invention are:

[0062] The present invention significantly improves the accuracy and timeliness of disaster prediction by optimizing matching rules and thresholds in real time, realizes the deep integration and efficient utilization of multi-source data, and provides a scientific basis and technical support for disaster warning and emergency response. At the same time, the system has adaptive learning capabilities and can continuously optimize the model in the constantly updated data to ensure the dynamic accuracy of the prediction results and effectively improve the disaster prevention and mitigation capabilities. In addition, the present invention enhances the comprehensiveness and reliability of disaster warning through comprehensive analysis of multi-dimensional data, provides decision makers with an intuitive disaster evolution situation map, helps to quickly formulate response strategies, minimizes disaster losses and improves the efficiency of emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flow chart of the method of the present invention;

[0064] Figure 2 It is a schematic diagram of the system modules of the present invention;

[0065] Figure 3 It is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0067] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0068] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive of other embodiments.

[0069] See also Figure 1 As shown, the present invention provides a trend analysis method based on multi-dimensional disaster warning data, comprising:

[0070] S1. Collect multi-dimensional heterogeneous data through distributed data collection interfaces, including meteorological data, geological data, and social media data;

[0071] In step S1, when performing disaster warning, it is first necessary to deploy a sensor network in the area where disaster prediction is performed to monitor and transmit data in real time, ensure the comprehensiveness and timeliness of information, and provide a solid foundation for subsequent analysis. Specifically, multi-dimensional heterogeneous data is collected through a distributed data acquisition interface, including but not limited to meteorological data, geological data, and social media data. Meteorological data can provide important information about weather changes, such as temperature, humidity, wind speed, air pressure, etc. Geological data involves key indicators such as crustal movement, seismic activity, and soil moisture. Social media data can reflect the public's attention and response to disaster events, and the keywords contained therein can be used as auxiliary information for disaster warning. After the multi-dimensional heterogeneous data is collected, pre-processing is performed simultaneously, including:

[0072] Meteorological data were filtered using a sliding window method to remove outliers that exceeded the normal fluctuation range;

[0073] The geological data are used to extract deformation characteristics using wavelet transform method and uniformly converted into standard time series format;

[0074] Social media data uses the Network Time Protocol to calibrate timestamp deviations across different platforms to ensure data temporal consistency;

[0075] Specifically, after completing the collection of multi-dimensional heterogeneous data, synchronous preprocessing steps are required to ensure the quality and consistency of the data. For meteorological data, the sliding window method is used for filtering, which can effectively identify and eliminate outliers that exceed the normal fluctuation range, thereby ensuring the accuracy and reliability of the data. For geological data, the wavelet transform method is used to extract the deformation characteristics, which can more accurately capture the subtle changes in geological activities. In addition, in order to facilitate subsequent data processing and analysis, the geological deformation characteristics will be uniformly converted into a standard time series format. Finally, for social media data, the timestamps on different platforms are calibrated through the network time protocol, which can effectively eliminate the possible timing deviations between different platforms, ensure the consistency of the data in the time dimension, and provide a corresponding basis for subsequent data analysis.

[0076] S2. Establish a multi-source event correlation network, where:

[0077] Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data;

[0078] Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events;

[0079] In step S2, after the multi-dimensional heterogeneous data is collected and preprocessed, a multi-source event association network is established. The purpose of the multi-source event association network is to integrate and analyze event information from different data sources. In this process, it is necessary to perform spatiotemporal matching on abnormal air pressure change events in meteorological data and deformation acceleration events in geological data to determine the causal relationship between the two. In addition, the keywords in social media data are matched with abnormal air pressure change events and deformation acceleration events for confidence. By analyzing the information on social media, the severity of the disaster event and the public's reaction are assisted in judging, thereby providing a more comprehensive perspective for disaster warning. Among them, the step of performing spatiotemporal matching on abnormal air pressure change events in meteorological data and deformation acceleration events in geological data includes:

[0080] Extract meteorological anomaly events whose air pressure change rate exceeds a preset change threshold from meteorological data, and establish a space-time cylindrical matching window with the center point of the meteorological anomaly event as the origin;

[0081] Conduct time-frequency analysis on geological deformation data in geological data to screen out deformation acceleration events;

[0082] The deformation acceleration events are spatially located and temporally aligned within the space-time cylinder matching window, and the correlation between the deformation acceleration events and the meteorological anomaly events within the space-time cylinder is calculated.

[0083] When the correlation is greater than or equal to a preset correlation threshold, a preliminary correlation is established between the meteorological anomaly event and the deformation acceleration event;

[0084] Generate a heat map of the spatial density of meteorological and geological anomalies, calculate the probability of simultaneous occurrence of meteorological and geological anomalies in overlapping areas, and upgrade the preliminary association between meteorological and geological anomalies to a valid association when the probability of simultaneous occurrence exceeds a preset probability threshold;

[0085] Incorporate meteorological anomaly events and geological anomaly events that meet effective correlation into the multi-source event correlation network;

[0086] Specifically, in order to perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data, it is first necessary to extract meteorological anomaly events whose air pressure change rate exceeds a preset change threshold from the meteorological data set. Once the meteorological anomaly event is successfully identified, a spatiotemporal cylindrical matching window with the center point of the meteorological anomaly event as the origin will be established for subsequent matching work. At the same time, time-frequency analysis will be performed on the geological deformation data in the geological data in order to screen out deformation acceleration events. Time-frequency analysis can reveal the acceleration trend of geological deformation within a specific time period, ensuring that the screened deformation acceleration events are highly timely and accurate. After identifying the deformation acceleration event, the deformation acceleration event will be spatially located and time-aligned with the meteorological anomaly event within the established spatiotemporal cylindrical matching window, so as to ensure that the deformation acceleration event and the meteorological anomaly event are accurately matched in space and time. Subsequently, the correlation between the deformation acceleration event and the meteorological anomaly event within the spatiotemporal cylinder is calculated to evaluate the correlation between the two. When the calculated correlation is greater than or equal to the preset correlation threshold, a preliminary association between meteorological anomaly events and deformation acceleration events can be established. In order to further verify the effectiveness of the preliminary association, it is necessary to generate a spatial density heat map of meteorological anomaly events and geological anomaly events. Through the spatial density heat map of meteorological anomaly events and geological anomaly events, we can intuitively understand the spatial density of meteorological anomaly events and geological anomaly events. Then it is necessary to calculate the probability of simultaneous occurrence of meteorological anomaly events and geological anomaly events in the overlapping area (simultaneous occurrence probability = the number of simultaneous occurrences of meteorological anomaly events and geological anomaly events in the overlapping area / total monitoring times). If the simultaneous occurrence probability exceeds the preset probability threshold, the preliminary association between meteorological anomaly events and geological anomaly events will be upgraded to a valid association. Otherwise, the correlation between the two will be considered weak and will not be included in the multi-source event association network. Finally, all meteorological anomaly events and geological anomaly events that meet the valid association conditions will be included in the multi-source event association network, integrating data from anomaly events from different sources.

[0087] In addition, the steps of matching the keywords in the social media data with the abnormal air pressure change events and deformation acceleration events with confidence include:

[0088] Perform natural language processing on social media data to extract keywords containing disaster semantic features and compile them into a disaster keyword library;

[0089] Cross-match the disaster keyword database with meteorological and geological anomalies in time and space, and calculate the confidence scores of keywords and events;

[0090] Filter out keyword and abnormal event pairs with confidence scores higher than the preset confidence threshold, establish correlations between keywords and abnormal events, and incorporate them into the multi-source event correlation network;

[0091] Specifically, when matching the keywords in social media data with abnormal air pressure change events and deformation acceleration events, natural language processing is first performed on the social media data to extract keywords containing disaster semantic features. After screening and sorting, the keywords are summarized into a disaster keyword library to facilitate subsequent analysis and matching. The constructed disaster keyword library is then cross-matched with meteorological and geological abnormal events in time and space to calculate the confidence score between each keyword and a specific event ( , where represents the confidence score, 、 and Respectively represent the weight coefficients of time overlap, space overlap and keyword occurrence frequency, Indicates the overlap duration between the keyword appearance time window and the duration of the abnormal event. Indicates the total duration of the abnormal event, Indicates the overlapping area between the keyword and the area affected by the abnormal event. Indicates the total area of the region affected by the abnormal event, represents the frequency of keyword occurrence within the overlapping spatiotemporal range), thereby evaluating the correlation between keywords and abnormal events. Finally, keyword and abnormal event pairs with confidence scores higher than the preset threshold are screened out, and the screened keyword and event pairs are used to establish corresponding association relationships and incorporated into the multi-source event association network.

[0092] S3, based on the event propagation chain of the multi-source event association network, backtracking to identify the disaster source area, and matching the diffusion path template of the current event with the historical similar event library;

[0093] In step S3, the event propagation chain based on the multi-source event association network is traced back to identify the source area of the disaster. By analyzing the correlation between events, the starting point of the disaster can be traced back. This method helps to formulate effective response measures. At the same time, the diffusion path template of the current event is matched with the historical similar event library. By comparing historical data, the development trend of the disaster and the possible scope of impact can be predicted. The step of tracing back the event propagation chain based on the multi-source event association network to identify the source area of the disaster includes:

[0094] In the multi-source event association network, the current disaster event is used as the propagation endpoint, and the associated event nodes are traced back step by step. A set of candidate source regions is generated based on the time difference between the associated events and the degree of spatial distance attenuation.

[0095] Double verification of the candidate source region set;

[0096] The first level of verification involves searching for previous disaster events with the same disaster characteristics through the historical case database, and matching the historical source coordinates whose spatiotemporal propagation trajectory matching degree is greater than or equal to the preset matching degree threshold;

[0097] The second level of verification uses real-time monitoring data to review the deformation of the candidate source area to verify whether the surface displacement rate exceeds the geological stability threshold. When the surface displacement rate exceeds the geological stability threshold, the corresponding candidate source area is locked.

[0098] Extract candidate source areas that meet both the first and second verifications, calculate the comprehensive confidence, and confirm the candidate source area with the highest comprehensive confidence as the disaster source area;

[0099] Specifically, when identifying the disaster source area, first, in the multi-source event association network, the current disaster event is used as the end point of propagation, and then the associated event nodes are traced back step by step. In this process, it is necessary to generate a set of candidate source areas based on the time difference between the occurrence of the associated events and the attenuation degree of spatial distance. Then, the set of candidate source areas is double-verified. The first step of the double verification is to search through the historical case library to find the preceding disaster events with the same disaster characteristics. By matching the spatiotemporal propagation trajectory, the historical source coordinates with a matching degree greater than or equal to the preset matching degree threshold are found. The second step of the double verification is to use real-time monitoring data to conduct deformation review of the candidate source area, mainly to verify whether the surface displacement rate exceeds the geological stability threshold. If the surface displacement rate does exceed the geological stability threshold, the corresponding candidate source area will be locked. Finally, the candidate source area that meets both the first and second verifications is extracted, and the comprehensive confidence of the two is calculated. The comprehensive confidence can be calculated by weighting multiple dimensional indicators such as the frequency of disaster events, the scope of impact, and the similarity of historical cases. Finally, the candidate source area with the highest comprehensive confidence is confirmed as the disaster source area.

[0100] Secondly, the steps of matching the diffusion path template of the current event with the historical similar event library include:

[0101] Retrieve historical events with the same disaster type identifier as the current event from the historical similar event database and extract the spatiotemporal diffusion path feature vector of the historical event. The spatiotemporal diffusion path feature vector includes the disaster propagation direction angle sequence, the disaster intensity attenuation coefficient matrix, and the cross-medium propagation node coordinate set.

[0102] Calculate the temporal similarity between the feature vector of the current event and the temporal diffusion path feature vector of the historical event, and filter out similar historical events with a temporal similarity higher than a preset threshold, as well as the temporal diffusion path templates corresponding to the similar historical events;

[0103] Keywords in social media data and confidence scores of similar historical events are mapped into path correction coefficients. The spatiotemporal diffusion path template is adjusted by the path correction coefficients to generate a predicted diffusion path network containing multiple branch paths, which is used as the diffusion path template for the current disaster event.

[0104] In this embodiment, when matching the diffusion path template of the current event according to the historical similar event library, firstly, the historical events with the same disaster type identifier as the current event are retrieved from the historical similar event library, and then the spatiotemporal diffusion path feature vector of the historical event is extracted. Here, the spatiotemporal diffusion path feature vector includes the disaster propagation direction angle sequence, the disaster intensity attenuation coefficient matrix and the cross-medium propagation node coordinate set, which provides basic data for subsequent analysis. Then, the time domain similarity between the feature vector of the current event and the spatiotemporal diffusion path feature vector of the historical event is calculated (specifically, it can be achieved through dynamic time warping algorithm or cosine similarity algorithm), so as to be able to screen Similar historical events with temporal similarity higher than a preset threshold are selected. For each screened similar historical event, its corresponding spatiotemporal diffusion path template is further extracted and used for subsequent path prediction. Finally, the keywords in the social media data and the confidence scores of similar historical events are mapped into path correction coefficients. The mapping method can be a linear regression or neural network model. The spatiotemporal diffusion path template is fine-tuned in combination with the path correction coefficient to generate a predicted diffusion path network containing multiple branch paths. The predicted diffusion path network includes not only the main path but also possible branch paths, thereby providing a more comprehensive and flexible diffusion path template for current disaster events.

[0105] S4. Generate a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time spread of the disaster, and use multi-color warning layers to overlay and display the warning levels of different disaster types;

[0106] In step S4, in order to intuitively display the real-time propagation range and warning level of the disaster, this embodiment generates a dynamic disaster simulation sand table. In the dynamic disaster simulation sand table, the pulse diffusion ripples can represent the real-time propagation range of the disaster, and the multi-color warning layer is superimposed to display the warning levels of different disaster types, so that managers can intuitively understand the severity of the disaster and the affected area. The steps of generating a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time propagation range of the disaster, and using multi-color warning layers to superimpose and display the warning levels of different disaster types include:

[0107] Construct a basic framework for disaster simulation sandbox on a 3D geographic information platform, and dynamically draw disaster propagation ripples using real-time monitoring data;

[0108] According to the disaster propagation direction in the diffusion path template, the expansion speed and intensity attenuation coefficient of the pulse ripple are adjusted through the ripple advancement mechanism;

[0109] Set different color layers according to the warning level and display them in real time;

[0110] When meteorological and geological anomalies occur simultaneously, they are displayed using a two-color spiral ripple overlay. The ripple interval represents the intensity of the interaction between the disasters. The smaller the interval, the stronger the correlation.

[0111] Specifically, in order to effectively generate a dynamic disaster simulation sand table, the basic framework of the disaster simulation sand table is first constructed on the three-dimensional geographic information platform. Then, the ripples of disaster propagation are dynamically drawn using real-time monitoring data to ensure that the sand table can reflect the development of the disaster in real time. Then, according to the disaster propagation direction defined in the diffusion path template, the expansion speed and intensity attenuation coefficient of the pulse ripple are adjusted through the ripple advancement mechanism to ensure that the dynamic changes of the ripple can accurately simulate the propagation characteristics of the disaster. In order to intuitively display different warning levels, different color layers are set according to the warning levels, and the different color layers are superimposed on the sand table in real time, so that users can clearly identify the disaster warning status of each area. It should be noted that when meteorological anomalies and geological anomalies occur at the same time, a two-color spiral ripple superposition display is adopted, and the ripple interval is used to represent the interaction intensity between the disasters. The smaller the ripple interval, the stronger the correlation between the two disasters, thereby providing an accurate basis for emergency decision-making.

[0112] S5. Based on the actual disaster verification results, incremental learning optimization is performed on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network;

[0113] In step S5, to ensure the accuracy of the multi-source event correlation network, incremental learning optimization is performed on the spatiotemporal matching rules and the diffusion path template matching threshold in the multi-source event correlation network based on the actual disaster verification results. As more actual data accumulates, it can continuously self-optimize and improve the accuracy of disaster analysis and early warning, thereby better serving disaster prevention and mitigation work. The step of incremental learning optimization of the spatiotemporal matching rules and the diffusion path template matching threshold in the multi-source event correlation network based on the actual disaster verification results includes:

[0114] Obtain real-time disaster spread boundary information through satellite remote sensing images and ground sensor networks, and compare it pixel by pixel with the predicted spread path network to generate a prediction error distribution map;

[0115] According to the temporal and spatial overlap in the prediction error distribution graph, the time window and spatial matching range parameters in the spatiotemporal matching rule are dynamically adjusted;

[0116] Establish a comparison mechanism between the actual disaster development speed and the predicted speed, and adjust the diffusion path template matching threshold based on the continuous comparison results;

[0117] Set the automatic optimization termination condition. When the prediction error rate after multiple consecutive optimizations is lower than the preset error threshold, stop the optimization and maintain the current optimal matching threshold and spatiotemporal matching rules.

[0118] Specifically, in order to improve the accuracy of spatiotemporal matching rules and the efficiency of diffusion path template matching in multi-source event association networks, satellite remote sensing images and ground sensor networks are first used to obtain the boundary information of disaster spread in real time. Then, the boundary information of disaster spread obtained in real time is compared and analyzed with the predicted diffusion path network pixel by pixel to generate a prediction error distribution map. The prediction error distribution map can intuitively show the difference between the predicted path and the actual diffusion situation. Then, according to the overlap of time and space in the prediction error distribution map, the time window and space matching range parameters in the spatiotemporal matching rules are dynamically adjusted to ensure that the spatiotemporal matching rules are more in line with the actual development of the disaster. For example, if the predicted path deviates far from the actual diffusion path, the time window is shortened and the spatial matching range is expanded, and vice versa. By continuously comparing the actual disaster development speed with the predicted speed, the diffusion path template matching threshold is fine-tuned until the prediction error rate is stably lower than the preset threshold, so as to ensure that the disaster diffusion trend can be accurately reflected. In order to ensure the efficiency and effectiveness of the optimization process, an automatic optimization termination condition is also set. When the prediction error rate after multiple consecutive optimizations is lower than the preset error threshold, the optimization process will be automatically stopped and the current optimal matching threshold and spatiotemporal matching rules will be maintained. Based on this method, not only the accuracy of the prediction is guaranteed, but also unnecessary waste of computing resources is avoided.

[0119] See also Figure 2 As shown, a trend analysis system based on multi-dimensional disaster warning data uses the above-mentioned trend analysis method based on multi-dimensional disaster warning data, including:

[0120] Data acquisition module, used to collect multi-dimensional heterogeneous data through distributed data acquisition interfaces, including meteorological data, geological data, and social media data;

[0121] The correlation network building module is used to build a multi-source event correlation network, where:

[0122] Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data;

[0123] Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events;

[0124] The diffusion prediction module is used to identify the disaster source area by tracing back the event propagation chain based on the multi-source event association network, and to match the diffusion path template of the current event with the historical similar event library;

[0125] The simulation and warning module is used to generate a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time spread of disasters, and using multi-color warning layers to overlay and display the warning levels of different disaster types;

[0126] The incremental learning module is used to perform incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the actual disaster verification results.

[0127] In the above, the data acquisition module is responsible for collecting multi-dimensional heterogeneous data through the distributed data acquisition interface. The multi-dimensional heterogeneous data includes but is not limited to meteorological data, geological data and social media data, which provides a rich source of information for subsequent analysis. The core function of the association network construction module is to establish a multi-source event association network. In the multi-source event association network, it will perform spatiotemporal matching of abnormal air pressure change events in meteorological data and deformation acceleration events in geological data to discover the potential connection between the two, and match the keywords in social media data with abnormal air pressure change events and deformation acceleration events with confidence, thereby enhancing the accuracy of event association. The diffusion prediction module uses multi-source The event association network retrospectively identifies the source area of the disaster and matches the diffusion path template of the current event based on the historical similar event library to provide a scientific basis for disaster prediction. The deduction and warning module is responsible for generating a dynamic disaster deduction sand table, characterizing the real-time propagation range of the disaster through pulse diffusion ripples. At the same time, it also uses multi-color warning layers to overlay and display the warning levels of different disaster types, making the warning information more intuitive and easy to understand. The role of the incremental learning module is to perform incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the verification results of the actual disaster situation. In this way, the system can continuously improve itself and improve the accuracy of disaster analysis and warning.

[0128] See also Figure 3 As shown, an electronic device includes:

[0129] at least one processor;

[0130] and a memory communicatively coupled to the at least one processor;

[0131] The memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the above-mentioned trend analysis method based on multi-dimensional disaster warning data.

[0132] The processors of the above-mentioned electronic devices can be of various types, such as CPU, GPU or TPU, to adapt to different computing needs and ensure efficient processing of multi-dimensional disaster data. The memory can be of various types, such as RAM, ROM or SSD, to store large amounts of data, support fast reading and writing, and ensure stable operation of the system. In addition, the electronic devices are also equipped with high-precision sensors to monitor environmental changes in real time to ensure the accuracy and timeliness of data collection, as well as arithmetic units, input devices and output devices. The arithmetic unit can be FPGA or ASIC, which is responsible for high-speed parallel computing. Input devices such as keyboards and touch screens are easy to operate, and output devices such as display screens and printers intuitively display the results. They work together as a whole to improve the response speed and decision-making support capabilities of the disaster warning system.

[0133] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0134] The foregoing is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained herein shall, unless otherwise specified or limited, be implemented in accordance with conventional means in the art.

Claims

1. A trend analysis method based on multi-dimensional disaster warning data, characterized by: include: Collect multi-dimensional heterogeneous data through distributed data collection interfaces, including meteorological data, geological data, and social media data; Establish a multi-source event correlation network, where: Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data; Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events; Backtracking the event propagation chain based on the multi-source event association network to identify the disaster source area, and matching the diffusion path template of the current event with the historical similar event library; Generate a dynamic disaster simulation sand table, use pulse diffusion ripples to represent the real-time spread of disasters, and use multi-color warning layers to overlay and display the warning levels of different disaster types; Based on the actual disaster verification results, incremental learning optimization is performed on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network; The step of backtracking and identifying the disaster source area based on the event propagation chain of the multi-source event correlation network includes: In the multi-source event association network, the current disaster event is used as the propagation endpoint, and the associated event nodes are traced back step by step. A set of candidate source regions is generated based on the time difference between the associated events and the degree of spatial distance attenuation. Double verification of the candidate source region set; The first level of verification involves searching for previous disaster events with the same disaster characteristics through the historical case database, and matching the historical source coordinates whose spatiotemporal propagation trajectory matching degree is greater than or equal to the preset matching degree threshold; The second level of verification uses real-time monitoring data to review the deformation of the candidate source area and verify whether the surface displacement rate exceeds the geological stability threshold. When the surface displacement rate exceeds the geological stability threshold, the corresponding candidate source area is locked; Extract candidate source areas that meet both the first and second verifications, calculate the comprehensive confidence, and confirm the candidate source area with the highest comprehensive confidence as the disaster source area; The step of matching the diffusion path template of the current event based on the historical similar event library includes: Retrieve historical events with the same disaster type identifier as the current event from the historical similar event database and extract the spatiotemporal diffusion path feature vector of the historical event. The spatiotemporal diffusion path feature vector includes the disaster propagation direction angle sequence, the disaster intensity attenuation coefficient matrix, and the cross-medium propagation node coordinate set. Calculate the temporal similarity between the feature vector of the current event and the temporal diffusion path feature vector of the historical event, and filter out similar historical events with a temporal similarity higher than a preset threshold, as well as the temporal diffusion path templates corresponding to the similar historical events; Keywords in social media data and confidence scores of similar historical events are mapped into path correction coefficients, and the spatiotemporal diffusion path template is adjusted by the path correction coefficient to generate a predicted diffusion path network containing multiple branch paths, which is used as the diffusion path template of the current disaster event.

2. A trend analysis method based on multi-dimensional disaster warning data according to claim 1, characterized in that: After the multi-dimensional heterogeneous data collection is completed, pre-processing is performed simultaneously, including: Meteorological data were filtered using a sliding window method to remove outliers that exceeded the normal fluctuation range; The geological data are used to extract deformation characteristics using wavelet transform method and uniformly converted into standard time series format; Social media data is calibrated for timestamp deviations across different platforms using the Network Time Protocol to ensure data temporal consistency.

3. The trend analysis method based on multi-dimensional disaster warning data according to claim 1, characterized in that: The step of performing spatiotemporal matching of the abnormal air pressure change events in the meteorological data and the deformation acceleration events in the geological data includes: Extract meteorological anomaly events whose air pressure change rate exceeds a preset change threshold from meteorological data, and establish a space-time cylindrical matching window with the center point of the meteorological anomaly event as the origin; Conduct time-frequency analysis on geological deformation data in geological data to screen out deformation acceleration events; The deformation acceleration events are spatially located and temporally aligned within the space-time cylinder matching window, and the correlation between the deformation acceleration events and the meteorological anomaly events within the space-time cylinder is calculated. When the correlation is greater than or equal to a preset correlation threshold, a preliminary correlation is established between the meteorological anomaly event and the deformation acceleration event; Generate a heat map of the spatial density of meteorological and geological anomalies, calculate the probability of simultaneous occurrence of meteorological and geological anomalies in overlapping areas, and upgrade the preliminary association between meteorological and geological anomalies to a valid association when the probability of simultaneous occurrence exceeds a preset probability threshold; Abnormal meteorological events and abnormal geological events that meet effective correlation requirements are included in the multi-source event correlation network.

4. The trend analysis method based on multi-dimensional disaster warning data according to claim 1, characterized in that: The step of matching the keywords in the social media data with the abnormal air pressure change events and the deformation acceleration events with confidence includes: Perform natural language processing on social media data to extract keywords containing disaster semantic features and compile them into a disaster keyword library; Cross-match the disaster keyword database with meteorological and geological anomalies in time and space, and calculate the confidence scores of keywords and events; Filter out keyword and abnormal event pairs with confidence scores higher than the preset confidence threshold, establish the association relationship between keywords and abnormal events, and incorporate them into the multi-source event association network.

5. The trend analysis method based on multi-dimensional disaster warning data according to claim 1, characterized in that: The steps of generating a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time propagation range of the disaster, and using multi-color warning layers to overlay and display warning levels of different disaster types include: Construct a basic framework for disaster simulation sandbox on a 3D geographic information platform, and dynamically draw disaster propagation ripples using real-time monitoring data; According to the disaster propagation direction in the diffusion path template, the expansion speed and intensity attenuation coefficient of the pulse ripple are adjusted through the ripple advancement mechanism; Set different color layers according to the warning level and display them in real time; Among them, when abnormal meteorological events and abnormal geological events occur at the same time, they are displayed by superimposing two-color spiral ripples, and the ripple interval is used to represent the interaction intensity between disasters. The smaller the interval, the stronger the correlation.

6. The trend analysis method based on multi-dimensional disaster warning data according to claim 1, characterized in that: The step of performing incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the actual disaster verification results includes: Obtain real-time disaster spread boundary information through satellite remote sensing images and ground sensor networks, and compare it pixel by pixel with the predicted spread path network to generate a prediction error distribution map; According to the temporal and spatial overlap in the prediction error distribution graph, the time window and spatial matching range parameters in the spatiotemporal matching rule are dynamically adjusted; Establish a comparison mechanism between the actual disaster development speed and the predicted speed, and adjust the diffusion path template matching threshold based on the continuous comparison results; Set the automatic optimization termination condition. When the prediction error rate after multiple consecutive optimizations is lower than the preset error threshold, stop the optimization and maintain the current optimal matching threshold and spatiotemporal matching rules.

7. A trend analysis system based on multi-dimensional disaster warning data, characterized by: The trend analysis method based on multidimensional disaster warning data according to any one of claims 1 to 6 comprises: Data acquisition module, used to collect multi-dimensional heterogeneous data through distributed data acquisition interfaces, including meteorological data, geological data, and social media data; The correlation network building module is used to build a multi-source event correlation network, where: Perform spatiotemporal matching between abnormal air pressure change events in meteorological data and deformation acceleration events in geological data; Confidence matching of keywords in social media data with abnormal air pressure change events and deformation acceleration events; The diffusion prediction module is used to identify the disaster source area by tracing back the event propagation chain based on the multi-source event association network, and to match the diffusion path template of the current event with the historical similar event library; The simulation and warning module is used to generate a dynamic disaster simulation sand table, using pulse diffusion ripples to represent the real-time spread of disasters, and using multi-color warning layers to overlay and display the warning levels of different disaster types; The incremental learning module is used to perform incremental learning optimization on the spatiotemporal matching rules and diffusion path template matching thresholds in the multi-source event association network based on the actual disaster verification results.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the trend analysis method based on multidimensional disaster warning data according to any one of claims 1 to 6.

Images