Disaster grade prediction method and system combined with regional iron tower monitoring data

Abstract

The application provides a disaster grade prediction method and system combining tower monitoring data of a target area, relates to the technical field of disaster grade prediction, and comprises the following steps: collecting tower data of a target area in real time, extracting key features of the tower data to obtain tower features; obtaining expert knowledge, combining historical disaster data and the expert knowledge to construct a disaster grade analysis model; inputting the tower features into the disaster grade analysis model to perform disaster grade evaluation and output a disaster grade evaluation result; formulating an emergency response plan based on the disaster grade evaluation result and executing the emergency response plan. The application can solve the technical problem that the accuracy and effectiveness of disaster grade evaluation are affected due to the failure to accurately analyze the diversity of disaster types and influence mechanisms in the prior art, achieve the technical goal of accurate disaster grade evaluation and dynamic emergency response capability in a multi-disaster scenario, and improve the accuracy of disaster evaluation.

Description

Technical Field

[0001] This application relates to the field of disaster prediction technology, and in particular to a method and system for predicting disaster levels by combining regional tower monitoring data. Background Technology

[0002] With the acceleration of urbanization and the increasing frequency of natural disasters, higher demands are being placed on the monitoring, evaluation, and management of steel towers. However, current technical methods for steel tower monitoring and management still have certain shortcomings.

[0003] Currently, existing technologies lack the comprehensive assessment capabilities for multiple disaster types. Different disasters (such as typhoons, earthquakes, and floods) have different manifestations and impact mechanisms. Single disaster assessment models struggle to cope with complex multi-hazard scenarios, leading to reduced accuracy in disaster severity assessments and consequently affecting the effectiveness of emergency response. When disaster types or severity differ, traditional assessment methods often fail to adjust emergency plans in a timely manner, resulting in inadequate allocation of relief resources and an inability to minimize losses. For example, in some regions, disaster severity assessments fail to accurately reflect the impact of different disasters on infrastructure, potentially leading to severely affected areas not receiving sufficient relief resources in a timely manner.

[0004] In summary, existing technologies suffer from technical problems due to the failure to accurately analyze the diversity of disaster types and impact mechanisms, making it difficult to address complex multi-hazard scenarios based on a single disaster assessment model, which further affects the accuracy of disaster level assessment and the effectiveness of emergency response. Summary of the Invention

[0005] The purpose of this application is to provide a disaster level prediction method and system that combines regional tower monitoring data, in order to solve the technical problem in the existing technology that the lack of accurate analysis of the diversity of disaster types and impact mechanisms makes it difficult to deal with complex multi-hazard scenarios based on a single disaster assessment model, which further affects the accuracy of disaster level assessment and the effectiveness of emergency response.

[0006] In view of the above problems, this application provides a disaster level prediction method and system that combines regional tower monitoring data.

[0007] Firstly, this application provides a disaster level prediction method that combines regional tower monitoring data. This method is implemented through a disaster level prediction system that combines regional tower monitoring data. The method includes: real-time acquisition of tower data from a target area; extraction of key features from the tower data to obtain tower features; acquisition of expert knowledge; construction of a disaster level analysis model by combining historical disaster data and the expert knowledge; inputting the tower features into the disaster level analysis model for disaster level assessment; and outputting the disaster level assessment result. Based on the disaster level assessment result, an emergency response plan is formulated and executed.

[0008] Secondly, this application also provides a disaster level prediction system combining regional tower monitoring data, used to execute the disaster level prediction method combining regional tower monitoring data as described in the first aspect, including: a feature extraction module, which is used to collect tower data in the target area in real time and extract key features from the tower data to obtain tower features; a model building module, which is used to acquire expert knowledge and build a disaster level analysis model by combining historical disaster data and the expert knowledge; a level assessment module, which is used to input the tower features into the disaster level analysis model to assess the disaster level and output the disaster level assessment result; and a plan execution module, which is used to formulate an emergency response plan based on the disaster level assessment result and execute the emergency response plan.

[0009] The technical solution provided in this application has at least the following technical effects or advantages: It obtains tower data by real-time collection of data from the target area, extracts key features from the tower data to obtain tower features; acquires expert knowledge, and constructs a disaster level analysis model by combining historical disaster data and the expert knowledge; inputs the tower features into the disaster level analysis model for disaster level assessment, and outputs the disaster level assessment result; formulates an emergency response plan based on the disaster level assessment result, and executes the emergency response plan. In other words, by achieving the technical goal of accurate disaster level assessment and dynamic emergency response capabilities under multi-hazard scenarios, it achieves the technical effects of improving the accuracy, real-time nature, and flexibility of disaster assessment, optimizing the allocation of rescue resources, and enhancing emergency response efficiency and post-disaster recovery capabilities.

[0010] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1This is a flowchart illustrating the disaster level prediction method based on regional tower monitoring data used in this application.

[0013] Figure 2 This is a schematic diagram of the disaster level prediction system that incorporates regional tower monitoring data, as described in this application.

[0014] Figure labeling: Feature extraction module 11, Model building module 12, Level evaluation module 13, Plan execution module 14. Detailed Implementation

[0015] This application provides a disaster level prediction method and system that combines regional tower monitoring data. It addresses the technical problem in existing technologies where the lack of accurate analysis of the diversity of disaster types and impact mechanisms makes it difficult to address complex multi-hazard scenarios using a single disaster assessment model, further impacting the accuracy of disaster level assessment and the effectiveness of emergency response. The application aims to achieve accurate disaster level assessment and dynamic emergency response capabilities in multi-hazard scenarios, thereby improving the accuracy, real-time nature, and flexibility of disaster assessment, optimizing the allocation of rescue resources, and enhancing emergency response efficiency and post-disaster recovery capabilities.

[0016] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0017] To facilitate understanding, the main implementation concepts of the various embodiments of the present invention will be briefly described first.

[0018] In the current situation, after a disaster occurs, in addition to the disaster site and limited disaster situations, national, provincial, and municipal management departments need to quickly understand and grasp the type and level of the disaster. However, currently, after a disaster occurs, multiple levels of management departments are unaware that a disaster has occurred or its level, and therefore cannot carry out effective dispatching, assessment, and rescue work.

[0019] The disaster level prediction method and system provided by this invention, which combines regional tower monitoring data, achieves the technical goals of accurate disaster level assessment and dynamic emergency response capabilities under multi-hazard scenarios. This results in improved accuracy, real-time performance, and flexibility of disaster assessment, optimized allocation of rescue resources, and enhanced emergency response efficiency and post-disaster recovery capabilities.

[0020] Example 1, please refer to the appendix. Figure 1This application provides a disaster level prediction method that combines regional tower monitoring data, applied to a disaster level prediction system that combines regional tower monitoring data, specifically including:

[0021] Step 1: Collect tower data in the target area in real time, and extract key features from the tower data to obtain tower features.

[0022] Specifically, sensors and monitoring equipment installed within the area continuously collect various status data of the towers, including tower tilt, temperature, wind speed, and power load. This data is transmitted wirelessly to a data processing system for further analysis. After acquiring the raw data from these towers, key features are extracted to select the indicators that best reflect the tower's health and performance. For example, data such as wind speed, vibration amplitude, and changes in the tower's power load can directly affect the tower's stability and failure risk. The extraction process is typically completed using statistical methods or machine learning algorithms to select the data features most relevant to the tower's safety, efficiency, and operational status, ultimately yielding the tower's characteristics. These tower characteristics will be used for subsequent risk assessment and decision analysis, accurately evaluating the tower's health and providing a basis for subsequent maintenance and early warning measures.

[0023] Step 2: Acquire expert knowledge and construct a disaster level analysis model by combining historical disaster data with the expert knowledge.

[0024] Specifically, acquiring expert knowledge refers to collecting and organizing the professional experience and theoretical knowledge of experts in the field regarding disasters and their impacts, which can provide important basis for disaster prediction and assessment. For example, experts may summarize the types of failures that may occur in iron towers under different wind speeds based on long-term research, or determine the degree of impact of a certain type of disaster on iron towers based on historical experience. This expert knowledge includes the classification of disasters, the analysis of influencing factors, and post-disaster recovery plans. When combining historical disaster data, by analyzing records of past disaster events, including information such as the time, location, and intensity of the disasters, and combining this with expert knowledge, a more accurate disaster level analysis model can be established. For example, historical data may indicate that the probability of iron tower collapse is high in a certain area during a typhoon with a specific wind speed (such as 30 meters per second), and experts may provide a formula relating wind speed and collapse probability in that area. By combining this historical data and expert knowledge, a disaster level analysis model that comprehensively considers multiple factors such as wind speed, precipitation, and temperature can be constructed. This model can predict the impact of disasters on iron towers in real time and assess the disaster level, such as "particularly severe" or "severe." This model, which combines expert knowledge with historical data, can not only improve the accuracy of predictions but also provide scientific guidance for disaster response.

[0025] Step 3: Input the tower characteristics into the disaster level analysis model to conduct a disaster level assessment and output the disaster level assessment results.

[0026] Specifically, inputting tower characteristics into the disaster level analysis model refers to inputting tower status information collected through sensors, such as tower tilt angle, offline rate, and power load, into a pre-trained disaster level analysis model for calculation and analysis. The core of this model lies in its ability to identify the disaster type implied by the input characteristics and to assess the level based on the key indicators of that disaster type. Internally, the model learns from historical disaster data and expert knowledge to establish a mapping relationship between key characteristics and disaster levels for different disaster types. For example, in the case of typhoon disasters, the model will focus on characteristics such as high wind speeds, heavy rainfall, increased tower vibration amplitude, abnormally increased tilt angles, severe fluctuations in power load, and the resulting increase in regional power outage rates. Wind speeds reaching specific thresholds (e.g., 25 m / s) or tilt angles exceeding critical values ​​(e.g., 2 degrees) may directly trigger higher disaster levels (e.g., "major"). Flood disasters: The model focuses on characteristics such as high water levels, high immersion rates, saturated soil moisture, tower foundation displacement / settlement (which can be indirectly reflected by tilt angle), and a significant increase in regional power outage rates (due to power facility failures caused by flooding). In this type of disaster, ambient temperature is usually not a major driving factor or assessment criterion. Water levels exceeding warning lines or immersion rates exceeding a specific percentage (e.g., 30%) may be classified as "major" disasters. Snow and ice disasters: The model focuses on characteristics such as low temperatures, high humidity (leading to icing), ice thickness on towers and lines, increased structural loads (which can be reflected by tilt or vibration), and insulator flashover risk (which may affect offline rates). Ice thickness reaching a critical value or abnormal structural loads may lead to an upgrade in the disaster level. Earthquake disasters: The model focuses on characteristics such as ground vibration intensity (e.g., peak ground acceleration), tower structural deformation (tilt angle, abnormal vibration modes), foundation displacement, and potential chain reactions (e.g., equipment damage leading to a sharp increase in offline rates and power outage rates). Vibration intensity exceeding the intensity threshold is crucial for assessment.

[0027] Tower characteristics reflect the tower's health condition and potential damage during a disaster. The disaster severity analysis model assesses the impact of a disaster by analyzing these input characteristics and combining them with disaster severity classification standards. The disaster severity classification is based on the extent of damage to tower infrastructure (communication, power), the scope of functional interruption, and the impact on the overall operation of the region.

[0028] Extremely serious: refers to a situation where a disaster causes severe damage or collapse of a large number of iron towers, and the regional communication and power functions are completely or nearly completely paralyzed (e.g., offline rate > 50%, power outage rate > 40%), posing an extremely serious threat to people's lives and property, social order and the operation of critical infrastructure.

[0029] Major: refers to a situation where a disaster causes significant damage or severe functional interruption to a considerable number of towers within a region (e.g., offline rate of 20%-50%, power outage rate of 20%-40%), resulting in widespread disruption of communication and power services, causing a major impact on production and daily life, and requiring the immediate activation of a large-scale emergency response.

[0030] Significant: This refers to a situation where a disaster causes damage to some towers in a localized area or limits their functionality (e.g., offline rate of 5%-20%, power outage rate of 10%-20%), resulting in partial interruption of communication and power services, and causing a significant impact on production and daily life in the localized area, requiring timely and targeted emergency measures.

[0031] General: This refers to situations where a disaster causes minor damage to a few or a small number of towers or results in brief alarms (e.g., offline rate <5%, power outage rate <10%), while communication and power services remain largely normal or experience only minor disturbances. The impact is limited and can be handled according to standard procedures. Ultimately, the disaster severity assessment results output by the model will inform decision-makers of the disaster's severity level, such as "particularly serious," "serious," "relatively serious," or "general," helping to quickly identify the disaster's risk level and providing a scientific basis for timely emergency response measures.

[0032] Step 4: Develop an emergency response plan based on the disaster level assessment results, and execute the emergency response plan.

[0033] Specifically, based on the severity and scope of the disaster, specific emergency response measures are formulated according to the current disaster situation through analysis of the disaster level assessment results. Disaster level assessment results typically include a classification of disaster intensity, such as "major," "relatively serious," or "general," reflecting the severity of the disaster. The core objective of the emergency response measures is to rapidly restore the communication / power functions of towers, prevent secondary disasters, and ensure the operation of critical infrastructure. For example, if a region is assessed as "major" due to a typhoon (e.g., regional power outage rate 20%-40%, offline rate 20%-50%), the emergency response plan will prioritize the repair and power restoration of critical tower facilities. Specific measures may include: immediately dispatching repair teams and emergency power generation equipment to core hub sites; urgently reinforcing or isolating towers that are tilted beyond limits or structurally damaged; coordinating with the power grid company to prioritize the restoration of power supply to main lines; activating backup communication links to ensure communication in critical areas; and conducting intensive monitoring of towers in high-risk areas to prevent cascading failures. The response action requires rapid and concentrated resources in the core damaged areas. For "particularly severe" disasters (e.g., offline rate > 50%, power outage rate > 40%), the emergency response plan implies the need to initiate the highest level of resource mobilization and cross-regional coordination. Measures may include: mobilizing large-scale repair forces and heavy equipment across provinces and cities; deploying numerous mobile generators to establish temporary power networks; organizing expert teams to conduct safety assessments and develop repair plans for large-scale damaged areas; implementing network-wide communication traffic scheduling and emergency communication support; and establishing a 24 / 7 uninterrupted monitoring and command system. The response action is characterized by its large scale, wide scope, long duration, and high degree of multi-departmental coordination, with the goal of quickly containing the spread of the disaster and gradually restoring the backbone network's functionality. When developing an emergency response plan, differentiated response strategies will be adopted based on different disaster levels and the extent of damage to the tower infrastructure they reflect, with significant differences in resource investment scale, response speed, task priority, and coordination scope.

[0034] Based on the established emergency response measures, resources are organized and allocated to implement rescue work. This involves actual task allocation, personnel mobilization, and material transportation. For example, in the event of a "major" typhoon, the emergency management department will immediately dispatch rescue teams to the disaster area to provide medical assistance, repair power facilities, and coordinate road traffic through transportation departments. During the execution of the plan, resource utilization is adjusted and optimized in a timely manner to ensure the smooth progress of all rescue operations and achieve the goals of minimizing disaster losses and restoring normal order.

[0035] The disaster level prediction method combining regional tower monitoring data is applied to the disaster level prediction system combining regional tower monitoring data. It can achieve the technical goals of accurate disaster level assessment and dynamic emergency response capability under multi-hazard scenarios, thereby improving the accuracy, real-time performance and flexibility of disaster assessment, optimizing the allocation of rescue resources, and enhancing emergency response efficiency and post-disaster recovery capabilities.

[0036] Furthermore, this application also includes: collecting real-time online and offline status data of target towers in the target area through a wireless sensor network to obtain an offline rate; counting the number of offline towers based on the tower fault records in the target area to obtain offline data; monitoring the water level and water accumulation range in the target area in real time using a water immersion sensor to calculate the area affected by water immersion to obtain a water immersion rate; monitoring the power supply status of the towers in the target area to obtain information on out-of-power towers and calculate the power outage rate; and combining the offline rate, the number of offline towers, the water immersion rate, and the power outage rate to obtain the tower data.

[0037] Specifically, a wireless sensor network (WSN) is a sensor system that transmits wireless signals to monitor and collect various data. Installing wireless sensors within a target area allows for real-time acquisition of the tower's operational status, recording its online status. Online status indicates the tower is functioning normally, while offline status means the tower's signal is lost or it is malfunctioning. The offline rate refers to the proportion of offline towers within the target area out of the total number of towers within a specific time period. This data provides insight into the overall health of the towers within the target area.

[0038] Tower failure records refer to the detailed information recorded by the relevant management system whenever a tower fails, including the time and type of failure. Within a target area, these records can be used to count the number of towers that fail and go offline within a certain period. The statistical results are offline data, providing information on the severity and frequency of failures within the area. For example, if ten towers fail and go offline within a week, it means that ten towers are offline each week. Offline data can reflect the overall reliability of the towers and the potential impact of disasters.

[0039] Water immersion sensors are devices specifically designed to monitor water levels and flooding conditions. They can record the water level height and flooded area in real time during a disaster. The percentage of the area affected by flooding represents the proportion of the total area of ​​the target region. The flooding rate can be used to assess the impact of a disaster on the tower and its surrounding environment, especially after floods or torrential rains, where flooding can damage tower infrastructure.

[0040] Monitoring the power supply status of transmission towers refers to the real-time detection of whether the towers are in a normal power supply state using devices such as current sensors and voltage sensors installed on the towers or power lines. When a power supply interruption is detected, the tower is determined to be in a "power outage" state. Information about outage towers includes the tower's identification and the start time of the outage. The outage rate refers to the proportion of towers in a power outage state within a specific time period within a target area. Power outages are a direct impact of disasters (such as strong winds breaking power lines, floods inundating substations, and earthquakes damaging the power grid) on power infrastructure and are a key indicator for assessing the severity of disasters and the functional integrity of transmission towers.

[0041] Combining offline rate, number of offline towers, flooding rate, and power outage rate yields a more comprehensive data report on tower health. This data comprehensively considers the tower's communication status, fault conditions, the impact of flooding in the surrounding environment, and power supply availability, providing a foundation for subsequent disaster assessment, risk analysis, and emergency response. For example, if a region has a high offline rate, a large number of offline towers, a high flooding rate, and a high power outage rate, it indicates that the disaster impact in that region is extremely severe, with significant damage to infrastructure (communication and power), potentially requiring immediate emergency repairs or relocation.

[0042] By comprehensively analyzing this data, a comprehensive assessment of the tower and its surrounding environment can be formed, potential problems can be identified in a timely manner and preventive measures can be taken, ensuring that the data at each stage can provide an accurate basis for the next step of decision-making, thereby achieving efficient and accurate risk warning and management.

[0043] Furthermore, this application also includes: obtaining denoised data by noise filtering the tower data; obtaining identification data by outlier detection and outlier removal of the denoised data; obtaining standard data by data normalization of the identification data; obtaining imputed data by filling missing values ​​in the standard data; and obtaining tower features by extracting offline rate, offline quantity, water immersion rate, power outage rate, tilt angle, and environmental conditions based on the imputed data.

[0044] Specifically, tower data is often affected by instability in the external environment or the equipment itself, resulting in noise in the data. Noise refers to irrelevant or inaccurate parts of the data, which may be caused by electromagnetic interference, sensor malfunctions, or transmission problems. Noise filtering uses algorithms to remove this irrelevant or inaccurate information, leaving only the useful signal. For example, wind speed data collected by sensors may fluctuate abnormally due to equipment problems. In this case, a noise filtering algorithm will eliminate these invalid fluctuations, thus obtaining more stable wind speed data—this is denoised data. Denoising-processed data is more accurate and helps with subsequent analysis and decision-making.

[0045] Even after noise removal, data may still contain extreme values ​​or data points that do not conform to normal patterns; these are called outliers. Outlier detection uses statistical methods or machine learning algorithms to identify these values ​​that do not fall within the expected range. For example, if a wind speed suddenly shows a value as high as 50 meters per second, while the normal wind speed range is usually between 5 and 20 meters per second, this data is an outlier. Outlier removal removes these unreasonable values ​​from the data to avoid affecting subsequent analysis. After outlier removal, the remaining data is the identified data, representing the processed dataset that conforms to the normal range.

[0046] Data normalization refers to transforming data of different dimensions or orders of magnitude into a standardized range, typically mapping data to the range of 0 and 1. This avoids excessively large numerical differences between different features, which could negatively impact model training. For example, the wind speed data for an iron tower might be ten meters per second, while the temperature data might be twenty degrees Celsius. Direct comparison due to their different dimensions can lead to bias. Normalization converts this data into a unified standard form, making it easier for the model to process. Standardized data, having undergone normalization, tends to perform better in subsequent algorithms.

[0047] Missing values ​​refer to data points that were not recorded or transmitted during the data acquisition process, resulting in incomplete data. Missing value imputation involves methods such as estimation, interpolation, or using averages of surrounding data to fill in the missing information. For example, if wind speed data was not collected for a certain period, imputation methods might estimate a reasonable value based on the trend of data before and after that period. The imputed data is called filled data; the integrity of this data is restored, and it can be used for subsequent analysis.

[0048] The extraction of offline rate, number of offline towers, water immersion rate, power outage rate, tilt angle, and environmental conditions refers to extracting the most representative information about the tower's condition from complete and processed data (impacted data), i.e., key feature extraction. Statistical or machine learning methods are used to select the most valuable features for disaster prediction and tower health assessment. Among these, the power outage rate, as a quantitative indicator directly reflecting the interruption of power supply to towers, is one of the key features for assessing the severity of disasters (especially those affecting power infrastructure). After extracting key features, the resulting tower feature dataset can provide a reliable basis for further model training, analysis, and decision-making.

[0049] Noise filtering removes useless information, outlier detection and removal eliminate erroneous data (i.e., outliers) to avoid affecting subsequent analysis, normalization unifies the scale of different data, missing value filling repairs data gaps, and feature extraction obtains key information that effectively reflects the tower's condition (including communication status, power supply, structural stability, and environmental factors). This ensures that the data can play its maximum role in subsequent analysis, thus providing accurate basis for disaster prediction, risk assessment, and decision support.

[0050] Furthermore, this application also includes: constructing data dimensions based on time, location, disaster type, intensity, tower offline rate, number of offline towers, and impact range; extracting historical disaster data based on the data dimensions; combining the historical disaster data and the expert knowledge to extract offline rate, number of offline towers, flooding rate, power outage rate, tilt angle, and environmental conditions to obtain input features; combining the historical disaster data and the expert knowledge to extract disaster levels of particularly serious, serious, relatively serious, and general to obtain output targets; constructing a deep neural network architecture, training the deep neural network architecture through the input features and the output targets to obtain the disaster level analysis model.

[0051] Specifically, in disaster management and risk assessment, data dimensions are combinations and arrangements of various information used to describe multiple aspects of a particular event. The time dimension refers to the specific time when the disaster occurred, while the spatial dimension refers to the specific area where the disaster occurred. Disaster type and intensity describe the type of disaster (e.g., typhoon, earthquake, flood) and its intensity (e.g., wind speed, magnitude). Tower downtime rate refers to the proportion of towers that lose signal due to the disaster; the number of downtime towers refers to the total number of towers that have lost signal; and the impact range refers to the extent to which the disaster affects the surrounding area. The combination of these dimensions comprehensively reflects the spatiotemporal distribution of the disaster and its impact on towers, providing complete background information for subsequent analysis.

[0052] Historical disaster data refers to detailed records of past disaster events, including information such as the time, location, type, and intensity of the disaster. Historical data is a crucial foundation for disaster severity analysis. By extracting relevant historical disaster data from the aforementioned data dimensions, patterns and trends of past disaster events can be identified, providing a reference for disaster prediction. For example, typhoon data from the past five years can be extracted to analyze typhoon intensity, impact range, and damage to power towers, helping to build risk assessment models for future disasters.

[0053] In disaster assessment, input features refer to the dataset used to train the model, reflecting the performance and health status of the towers after a disaster. By combining historical disaster data and expert knowledge, indicators such as offline rate, number of offline towers, flooding rate, power outage rate, tilt angle, and environmental conditions are extracted. The offline rate and number of offline towers reflect the extent to which towers lose communication signals due to disasters, while the flooding rate measures the degree of impact of floods on the region. The power outage rate directly reflects the power supply interruption caused by disasters and is a key indicator for assessing damage to power infrastructure. The tilt angle represents the structural deformation of the towers, and environmental conditions may include factors such as wind speed and precipitation. These feature data will serve as model inputs to help predict the specific impact of disasters on the towers.

[0054] The output target is the model's prediction result, representing the disaster's severity level. In disaster management, disasters are classified into four levels based on their impact: extremely severe, severe, relatively severe, and general. Extremely severe: Causes widespread communication disruptions in multiple provinces (autonomous regions, municipalities) or serious damage to important communication hubs, or devastating blows to national-level important communication infrastructure, seriously affecting national security, economic operation, social stability, or public interests, requiring the activation of a national-level emergency response. Severe: Causes widespread communication disruptions in multiple local networks within a province (autonomous region, municipality) or damage to important provincial communication hubs, seriously affecting the region's economic operation, social stability, or public interests, requiring the activation of a provincial-level emergency response. Relatively severe: Causes widespread communication disruptions in a single local network within a province (autonomous region, municipality) or damage to important communication facilities, significantly impacting the region's economic operation, social stability, or public interests, requiring the activation of an emergency response by relevant municipal or provincial departments. General: Causes communication disruptions in a localized area or damage to some communication facilities, with a limited impact, mainly affecting the communication needs of users in a localized area, and is handled by the municipal or county-level communications authorities. By combining historical disaster data and expert knowledge, each type of disaster event can be assigned a corresponding level as the model's target output, helping decision-makers quickly identify the severity of the disaster. Deep neural networks are machine learning models commonly used to process complex data, simulating the brain's learning process through connections between multiple layers of neurons. In disaster level analysis, the architecture of a deep neural network is trained by input features (such as offline rate, number of offline events, flooding rate, power outage rate, tilt angle, environmental conditions, etc.) and the output target (such as disaster level), continuously adjusting its internal parameters to optimize the model's predictive ability. Through training, the neural network can learn the complex relationship between disasters and tower damage and functional interruptions (communication, power) from historical data, thereby establishing an accurate disaster level analysis model. This model can assess the impact of current disasters in real time and provide decision support for emergency response.

[0055] By combining expert knowledge and historical data, features that contribute to comprehensive disaster level prediction (covering communication disruptions, power outages, structural deformation, and environmental factors) are extracted. An accurate disaster analysis model is then trained using a deep neural network architecture, providing necessary data support for the accuracy of the final disaster assessment model.

[0056] Furthermore, this application also includes: combining the historical disaster data and the expert knowledge to analyze the cascading risk effects between target towers and obtain multi-factor dynamic variables; based on the cascading risk effects of the multi-factor dynamic variables, obtaining associated input features and associated output targets; and inputting the associated input features and associated output targets into the input features and the output targets, respectively.

[0057] Specifically, historical disaster data records detailed information about various past disaster events. Expert knowledge refers to the professional understanding and summary of disaster impacts based on the experience and theories of domain experts. When data are combined, it is possible to identify and analyze potential cascading risk effects between target towers. A cascading risk effect refers to the direct or indirect impact that damages one tower on other nearby towers, creating a chain reaction. For example, during a typhoon, if a tower is damaged by wind or loses power, it may not only cause additional stress on the physical structure of nearby towers but also affect adjacent facilities that rely on its functions (such as communication relay stations and regional power junctions) due to power or communication outages, ultimately causing wider physical damage and functional paralysis. By analyzing these factors, multiple dynamic variables related to the cascading effect can be identified, such as wind speed, precipitation, soil moisture, and tower power outage status / outage rate, to obtain a comprehensive understanding of the disaster's impact on the towers and their support network.

[0058] In the analysis of cascading risk effects, the complex interactions between multiple factors (such as wind speed, precipitation, tower damage, and power outages) are considered to determine whether these factors collectively determine the overall performance of the tower network and the risk propagation path during a disaster. Correlated input features refer to useful data extracted from these multi-factor dynamic variables that characterize cascading risk relationships, serving as input to the (enhanced) model to more accurately predict disaster impacts. Correlated output targets refer to the predicted results based on these input features that consider cascading effects, such as upgraded disaster levels, wider-ranging tower damage, regional functional disruption procedures, or the complexity and time required for post-disaster recovery. Output targets are generated based on input features (such as wind speed, power outage rate, and tilt) and their interrelationships. For example, if a region experiences wind speeds of 30 meters per second, a power outage rate exceeding a threshold, and water accumulation reaching 200 millimeters, it can be predicted that the tower network in that region may face extremely high risks of systemic damage or functional collapse, and a corresponding upgraded disaster level (e.g., from "relatively high" to "major") can be given.

[0059] By supplementing the input features with related input features and the output targets with related output targets, and by inputting these input features and output target relationships that take into account the cascading effects into the model, the model can learn from them and more deeply identify the complex nonlinear relationships between the input features and output targets, especially under the influence of cascading risks. This allows the model to more accurately predict the potential spread and ultimate impact range of disasters when they occur, enabling prediction and decision-making.

[0060] By analyzing the cascading risk effects between target towers, we can obtain various dynamic factors (including physical damage factors and functional interruption factors) that affect the health of the tower network. This further helps to extract related input features and related output targets that better reflect the system's vulnerability and risk propagation, providing richer and more insightful training data for machine learning models and significantly improving the model's evaluation and prediction capabilities in complex multi-hazard cascading scenarios.

[0061] Furthermore, this application also includes: identifying time-dependent dynamic variables of the multi-factor dynamic variables to obtain time-series variables; obtaining time-series associated input features and time-series associated output targets based on the cascading risk effects of the multi-factor dynamic variables; and inputting the time-series associated input features and the time-series associated output targets into the associated input features and the associated output targets, respectively.

[0062] Specifically, multi-factor dynamic variables refer to multiple factors considered when analyzing the impact of disasters on power towers, such as wind speed, precipitation, tower offline rate, and tower power outage rate. These variables change over time. Time dependence means that the changes in these variables depend not only on the immediate state of the disaster but also on past trends. For example, wind speed may increase sharply in a short period, while precipitation or regional power outage rate may show a cumulative or spreading trend. Therefore, identifying time-dependent dynamic variables involves analyzing the patterns of these variables over time (such as rate of change, trend, and periodicity) to identify time-series variables. These time-series variables reflect the patterns of disaster factors and their impacts (including functional interruptions) over time, and can more accurately predict the evolution trajectory of tower risks during disaster development. For example, suppose that the power outage rate in a certain area rapidly increases from 5% to 40% within six hours after a typhoon makes landfall, while the wind speed remains at 25 meters per second. Such time-series data can identify that the power system is suffering severe damage, and the tower network faces the risk of large-scale functional failure.

[0063] The cascading risk effect refers to the influence of one factor (such as wind speed, flooding, or power outage) on other factors and their interactions, and this interaction often evolves over time. For example, strong winds may cause a power tower to tilt, which in turn may increase the load on the tower, thus affecting the stability of other nearby towers; similarly, a power outage on a critical tower may, over time, cause nearby communication towers that rely on its power supply to go offline due to the depletion of their backup power. These cascading effects are usually time-dependent, so combining these multi-factor dynamic variables and their interactions with time factors yields time-series correlated input features. Input features may include wind speed change rate, precipitation accumulation, power outage rate growth trend, and the lag changes in the correlation between offline rate and power outage rate, which can reflect the mutual influence and risk transmission path between various factors during the disaster process. The time-series correlated output target refers to the predicted results based on these time-evolving time-series input features, such as the disaster level at a future time point, the degree of tower damage or functional interruption, and its potential impact range. For example, if the wind speed in a certain area is expected to continue to increase in the next two hours, the current power outage rate in the area has exceeded 30% and is still rising, and is accompanied by continuous heavy rainfall, the model can predict that the tower network in the area may be severely damaged and its functions may be interrupted in the next few hours. The output target is updated to the "particularly serious" disaster level and a warning is issued for the impact on a wider area.

[0064] By supplementing the input features with time-series correlated data and supplementing the output targets with time-series correlated data, the model can learn and identify the dynamic relationships between these input features and output targets across time steps. This allows the model to dynamically predict the evolution of disasters and their cascading impacts (especially communication and power outages) based on real-time time-series data during actual disasters, enabling more accurate predictions and decisions. For example, input data might include wind speed trends and power outage rate growth curves over the past few hours, while the output target might be a prediction of changes in the overall functional status level of the tower network or the regional disaster level over the next few hours.

[0065] By identifying the time dependence of multi-factor dynamic variables (especially indicators of functional disruption, such as power outage rates), time-series variables are obtained, deeply reflecting the dynamic trends of different factors and their interactions during disasters. By analyzing the complex correlations between time-series data, time-series correlated input features and time-series correlated output targets are obtained, providing training data for the model to characterize dynamic processes and risk propagation. By learning the relationships between these inputs and outputs over time, the model can proactively predict the risk evolution of the tower network and formulate more predictive emergency response strategies based on real-time changing factors and their historical trends when actual disasters occur.

[0066] Furthermore, this application also includes: converting the disaster level assessment results and the emergency response plan into digital decisions; generating a rescue plan through the digital decisions; pushing the rescue plan through a real-time information push system via multiple channels; scheduling work based on the push results; and completing the execution of the emergency response plan.

[0067] Specifically, the disaster impact assessment results obtained through disaster level analysis models (including comprehensive assessments of communication outages, power outages, structural damage, etc.) are combined with pre-defined emergency response plans for different disaster types and impact dimensions (such as power outages), transforming them into a digital form that can be processed by a computer system. Disaster level assessment results typically include the severity of the disaster (e.g., "major" or "extremely major") and its key causative factors (e.g., "accompanying large-scale power outages"), while emergency response plans are emergency measures developed based on different disaster levels and their specific characteristics. This transformation process converts unstructured data into a structured digital form that can be used for decision support. For example, if a region's disaster level is "major" and the assessment results show a regional power outage rate exceeding 40%, based on the pre-defined emergency response plan, digital decisions might include "activating emergency power repair teams," "dispatching mobile power generation equipment to critical sites," "prioritizing power supply to core communication nodes," and conventional actions such as "activating emergency teams" and "dispatching resources." These digital decisions can then be further used for precise scheduling and allocation.

[0068] The automated system generates specific, targeted rescue plans based on the disaster severity assessment results (especially power outages) and emergency response plans. These plans include details such as the necessary rescue actions, resource allocation (e.g., the ratio of power repair resources, communication support resources, and conventional rescue resources), and personnel deployment. For example, when the disaster severity level is "extremely severe" and accompanied by an extremely high power outage rate, the rescue plan includes large-scale mobilization of power repair forces, deployment of numerous emergency power generation facilities, priority repair of the main power grid, as well as additional rescue personnel, mobilization of more materials and equipment, and the development of an emergency evacuation plan. The generation of rescue plans relies on the system's understanding and application of digital decision-making (especially information reflecting the severity of power outages) to ensure that all emergency measures accurately match the actual damage caused by the disaster, particularly the functional disruption of critical infrastructure.

[0069] The generated rescue plans, which include specific tasks such as power restoration, will be promptly sent to all relevant personnel, such as emergency management departments, rescue teams, power repair departments, telecommunications operators, medical rescue teams, and resource allocation centers, through various means (such as SMS, email, mobile application push notifications, and dedicated emergency communication channels). This ensures that all decision-makers and implementers can receive the latest rescue instructions simultaneously and in a timely manner when a disaster occurs, achieving rapid information transmission and sharing. For example, if a region experiences a "major" disaster accompanied by severe power outages (outage rate > 30%), the system will simultaneously push plans containing specific instructions such as power repair priorities, emergency power supply point settings, and communication support requirements to local governments, emergency management bureaus, regional command centers of power grid companies, and emergency centers of telecommunications operators through preset channels. This ensures that power restoration, communication support, and life-saving forces can respond quickly and collaboratively.

[0070] After the rescue plan is sent to relevant departments through a real-time information push system, these departments make specific work arrangements and dispatch based on this information. Work dispatch includes assigning specific tasks, determining the priority of resource use, and coordinating various forces for emergency rescue. For example, if a region is notified that additional medical teams and supplies are needed, the command center will dispatch the appropriate teams and supplies to the site based on the current disaster situation and resource availability. Ultimately, all work dispatch and command execution form a unified action chain to ensure the smooth implementation of the emergency response plan.

[0071] By transforming disaster severity assessments and emergency response plans into digital decisions, and then using these decisions to generate specific rescue plans, these plans are promptly delivered to relevant personnel through multiple channels. Ultimately, work scheduling ensures the effective execution of rescue operations, thereby ensuring that emergency response can be carried out efficiently when a disaster occurs.

[0072] Furthermore, this application also includes: the disaster level analysis model is a recurrent neural network model, used to learn the changing trends of disaster factors over time and predict future disaster levels.

[0073] Specifically, recurrent neural network (RNN) models are deep learning models specifically designed for processing and learning time-series data. Compared to traditional neural network models, RNN models have memory capabilities, allowing them to store past information (such as historical wind speeds, precipitation, and power outage rates) and use this information in current calculations. This enables RNN models to better understand and predict dynamic processes that change over time, such as the evolution of natural disasters and their cascading effects on infrastructure (such as communication towers and power supply).

[0074] Learning the changing trends of disaster factors over time means that models analyze historical disaster data to identify key disaster factors, such as wind speed, rainfall, tower tilt angle, tower offline rate, and tower power outage rate, and to understand the patterns of these factors' changes over time and their temporal correlations. For example, a model might find that before a typhoon arrives, wind speed gradually increases, while rainfall and potential power outage risk increase rapidly in a short period; or during a prolonged flood, there is a clear positive correlation and lag effect between flooding rate and power outage rate. By learning these trends and correlation patterns, models can more accurately predict future disaster levels and the key issues they may involve (such as widespread power outages).

[0075] Predicting future disaster levels refers to a model's forecast of the severity of a disaster over a future period based on learned trends and interactions of disaster factors. Disaster levels are typically categorized into four levels: extremely severe, severe, relatively severe, and moderate, representing the extent and scope of the disaster's impact. For example, a model might predict that within the next 6-12 hours, the average wind speed in a region will consistently exceed 25 meters per second, and the regional power outage rate is expected to rise from the current 20% to over 50%. A comprehensive assessment would indicate an "extremely severe" disaster level, meaning the highest level of emergency response must be activated immediately, including large-scale evacuation and intensive repairs of power facilities.

[0076] In summary, recurrent neural network models, by learning the changing trends and complex relationships of multi-dimensional disaster factors, including power outage rates, over time, can more accurately predict future disaster levels and their key characteristics (such as the severity and spread risk of power outages). This provides a scientific basis for more precise and forward-looking disaster warnings and targeted emergency responses. For example, during typhoon season, the model can predict the intensity and path of typhoons, as well as the potential scope and evolution of power outages they may cause. This helps relevant departments not only to evacuate residents and allocate rescue resources in a timely manner, but also to deploy power repair forces and allocate emergency power generation equipment in advance, optimizing resource allocation and thus minimizing disaster losses and accelerating the recovery of critical infrastructure functions.

[0077] In summary, the disaster level prediction method combining regional tower monitoring data provided in this application has the following technical effects: Tower data is obtained by real-time collection of data from the target area; key features are extracted from the tower data to obtain tower features; expert knowledge is acquired, and a disaster level analysis model is constructed by combining historical disaster data and the expert knowledge; the tower features are input into the disaster level analysis model for disaster level assessment, and the disaster level assessment result is output; an emergency response plan is formulated based on the disaster level assessment result, and the emergency response plan is executed. In other words, by achieving the technical goal of accurate disaster level assessment and dynamic emergency response capability under multi-hazard scenarios, the method improves the accuracy, real-time performance, and flexibility of disaster assessment, optimizes the allocation of rescue resources, and enhances emergency response efficiency and post-disaster recovery capabilities.

[0078] Example 2: Based on the disaster level prediction method combining regional tower monitoring data in the aforementioned examples, and using the same inventive concept, this application also provides a disaster level prediction system combining regional tower monitoring data. Please refer to the appendix. Figure 2 The system includes: a feature extraction module 11, which collects tower data from the target area in real time and extracts key features from the tower data to obtain tower features; a model building module 12, which acquires expert knowledge and combines historical disaster data with the expert knowledge to build a disaster level analysis model; a level assessment module 13, which inputs the tower features into the disaster level analysis model to assess the disaster level and outputs the disaster level assessment result; and a plan execution module 14, which formulates an emergency response plan based on the disaster level assessment result and executes the emergency response plan.

[0079] Furthermore, the disaster level prediction system combining regional tower monitoring data is also used for: collecting real-time online and offline status data of target towers in the target area through a wireless sensor network to obtain the offline rate; counting the number of offline towers based on the tower fault records in the target area to obtain offline data; monitoring the water level and water accumulation range of the target area in real time using water immersion sensors to calculate the area affected by water immersion to obtain the water immersion rate; monitoring the power supply status of towers in the target area to obtain information on outage towers and calculate the power outage rate; and combining the offline rate, the number of offline towers, and the water immersion rate to obtain the tower data.

[0080] Furthermore, the disaster level prediction system combining regional tower monitoring data is also used for: obtaining denoised data by filtering the tower data for noise; obtaining identification data by detecting and removing outliers from the denoised data; obtaining standard data by normalizing the identification data; obtaining imputed data by filling in missing values ​​in the standard data; and obtaining tower characteristics by extracting offline rate, offline quantity, water immersion rate, power outage rate, tilt angle, and environmental conditions based on the imputed data.

[0081] Furthermore, the disaster level prediction system combining regional tower monitoring data is also used for: constructing data dimensions based on time, location, disaster type, intensity, tower offline rate, number of offline towers, and impact range; extracting historical disaster data based on the data dimensions; combining the historical disaster data and expert knowledge to extract offline rate, number of offline towers, flooding rate, power outage rate, tilt angle, and environmental conditions to obtain input features; combining the historical disaster data and expert knowledge to extract disaster levels of particularly severe, severe, relatively severe, and general to obtain output targets; constructing a deep neural network architecture, training the deep neural network architecture through the input features and the output targets to obtain the disaster level analysis model.

[0082] Furthermore, the disaster level prediction system combining regional tower monitoring data is also used to: combine the historical disaster data and the expert knowledge to analyze the chain risk effect between target towers and obtain multi-factor dynamic variables; based on the chain risk effect of the multi-factor dynamic variables, obtain associated input features and associated output targets; and input the associated input features and associated output targets to the input features and the output targets, respectively.

[0083] Furthermore, the disaster level prediction system combining regional tower monitoring data is also used for: identifying time-dependent dynamic variables of the multi-factor dynamic variables to obtain time-series variables; obtaining time-series associated input features and time-series associated output targets based on the cascading risk effects of the multi-factor dynamic variables; and inputting the time-series associated input features and the time-series associated output targets into the associated input features and the associated output targets, respectively.

[0084] Furthermore, the disaster level prediction system that combines regional tower monitoring data is also used to: transform the disaster level assessment results and the emergency response plan to obtain digital decisions; generate rescue plans through the digital decisions; push the rescue plans through a real-time information push system through multiple channels; and schedule work according to the push results to complete the execution of the emergency response plan.

[0085] Furthermore, the disaster level prediction system that combines regional tower monitoring data is also used to: use a recurrent neural network model as the disaster level analysis model to learn the changing trends of disaster factors over time and predict future disaster levels.

[0086] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The disaster level prediction method and specific examples combining regional tower monitoring data in the aforementioned embodiment one are also applicable to the disaster level prediction system combining regional tower monitoring data in this embodiment. Through the foregoing detailed description of the disaster level prediction method combining regional tower monitoring data, those skilled in the art can clearly understand the disaster level prediction system combining regional tower monitoring data in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0087] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0088] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A disaster level prediction method that binds regional tower monitoring data, characterized by, The method comprises the following steps: real-time acquisition of target area to obtain tower data, key feature extraction of the tower data to obtain tower features; acquiring expert knowledge, combining historical disaster data and the expert knowledge to construct a disaster grade analysis model; inputting the tower features into the disaster grade analysis model for disaster grade evaluation, and outputting disaster grade evaluation results; formulating an emergency response plan based on the disaster grade evaluation results, and executing the emergency response plan; the combination of historical disaster data and expert knowledge to construct a disaster grade analysis model comprises: constructing data dimensions with time, location, disaster type, intensity, tower offline rate, offline quantity and influence range; extracting historical disaster data according to the data dimensions; combining the historical disaster data and the expert knowledge, extracting offline rate, offline quantity, water immersion rate, tilt angle and environmental conditions to obtain input features; combining the historical disaster data and the expert knowledge, extracting disaster grade to obtain output target; constructing a deep neural network architecture, training the deep neural network architecture through the input features and the output target, and obtaining the disaster grade analysis model; before training the deep neural network architecture through the input features and the output target, it comprises: combining the historical disaster data and the expert knowledge, analyzing the chain risk effect between target towers, and acquiring multi-factor dynamic variables; based on the chain risk effect of the multi-factor dynamic variables, acquiring associated input features and associated output targets; inputting the associated input features and the associated output targets into the input features and the output targets respectively.

2. The disaster level prediction method of claim 1, wherein, Real-time acquisition of target area to obtain tower data, comprising: acquiring real-time online and real-time offline state data of target towers in the target area through a wireless sensor network to obtain an offline rate; statistically analyzing the number of offline towers according to the tower failure records of the target area to obtain offline data; using a water immersion sensor to monitor the water level and water accumulation range of the target area in real time, calculating the area ratio affected by water immersion to obtain a water immersion rate; combining the offline rate, the offline quantity and the water immersion rate to obtain the tower data.

3. The disaster level prediction method of claim 1, wherein, Key feature extraction of the tower data to obtain tower features, comprising: noise filtering of the tower data to obtain denoised data; abnormal value detection and abnormal value elimination of the denoised data to obtain identified data; data normalization processing of the identified data to obtain standard data; missing value filling of the standard data to obtain filled data; offline rate, offline quantity, water immersion rate, tilt angle and environmental condition extraction according to the filled data to obtain the tower features.

4. The disaster level prediction method of claim 1, wherein, Before inputting the associated input features and the associated output targets into the input features and the output targets respectively, it comprises: time-dependent dynamic variable identification of the multi-factor dynamic variables to obtain time series variables; based on the chain risk effect of the multi-factor dynamic variables, acquiring time series associated input features and time series associated output targets; The time-correlated input features and the time-correlated output targets are input into the correlated input features and the correlated output targets, respectively.

5. The disaster level prediction method of claim 1, wherein, The emergency response plan is executed, including: The disaster level assessment result and the emergency response plan are converted to obtain a digital decision; A rescue plan is generated through the digital decision, the rescue plan is pushed through a real-time information pushing system in multiple channels, work scheduling is performed according to a pushing result, and execution of the emergency response plan is completed.

6. The disaster scale prediction method of claim 1, wherein, The disaster level analysis model is a recurrent neural network model, which is used to learn a change trend of disaster factors over time and predict a future disaster level.

7. A disaster level prediction system that combines regional tower monitoring data, characterized by, Steps for implementing the disaster level prediction method of the combined regional tower monitoring data according to any one of claims 1 to 6, including: A feature extraction module is configured to collect tower data in a target region in real time, extract key features from the tower data, and obtain tower features; A model construction module is configured to obtain expert knowledge, combine historical disaster data and the expert knowledge, and construct a disaster level analysis model; A level assessment module is configured to input the tower features into the disaster level analysis model for disaster level assessment, and output a disaster level assessment result; A plan execution module is configured to develop an emergency response plan based on the disaster level assessment result, and execute the emergency response plan.

Images