Public broadband early warning method and system based on disaster prevention, electronic equipment, medium and product

Abstract

The invention provides a public broadband early warning method and system based on disaster prevention, electronic equipment, a storage medium and a program product, and aims to solve the problems that disaster prevention early warning is static and lagging, lacks prediction capability, is low in efficiency and is insufficient in precision. The method comprises the following steps: constructing a whole-course routing logic chain from a user broadband account to a core network based on multi-source network data; establishing a network resource geographic information layer containing position information of the user terminal and the network equipment; acquiring predicted disaster data, and performing spatial overlay analysis on the predicted disaster data and the network resource geographic information layer to identify a potential affected area; based on the whole-course routing logic chain and the potential affected area, mapping out all possibly affected user broadband accounts; and based on historical disaster data and network performance data, utilizing a prediction model to evaluate the network vulnerability and the affected user range of the potential affected area. According to the invention, accurate pre-disaster early warning can be realized, and specific affected users can be quickly positioned.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a public broadband early warning method and system, electronic device, computer-readable storage medium, and computer program product based on disaster prevention. Background Technology

[0002] In the field of disaster prevention, mitigation, and emergency management, especially in responding to major natural disasters such as typhoons and earthquakes, emergency management departments and public utility service providers (such as telecommunications operators) face severe challenges. Currently, the assessment of the potential impact of natural disasters on public communication networks mainly relies on manual statistics and reporting after the disaster.

[0003] Specifically, existing technical solutions have the following significant drawbacks:

[0004] Static and Lagging Nature: Network equipment information (such as latitude and longitude) and network topology relationships of operators largely rely on manual input and updates. This static and lagging management method makes it impossible to quickly and accurately establish a correlation model between disaster prediction paths and network resources before a disaster occurs.

[0005] Lack of predictive capabilities: Existing technical solutions mainly focus on post-disaster fault handling and statistics. Maintenance personnel need to log into their respective professional network management systems to manually verify and summarize the affected scope and user information after a disaster damages facilities such as data centers and transmission lines. This approach completely fails to meet the urgent needs of emergency departments to predict impacts before a disaster and thus scientifically deploy rescue forces. For example, while meteorological bureaus can predict typhoon paths, existing technology cannot answer the crucial question: "Under the predicted typhoon path, to what extent will the operator's network and users be affected?"

[0006] Inefficiency and lack of accuracy: During and after a disaster, determining the scope of the impact of a fault relies entirely on the experience of maintenance personnel. Manual cross-system correlation analysis is cumbersome, time-consuming, and prone to omissions, making it difficult to generate an accurate list of affected users. This results in a lack of timely and accurate data support for emergency response decisions.

[0007] Therefore, there is an urgent need in this field for a public broadband early warning method for disaster prevention that can achieve pre-disaster prediction, rapid and accurate assessment during disasters, and fully automated operation, in order to overcome the above-mentioned deficiencies of existing technologies. Summary of the Invention

[0008] To address at least some of the problems existing in current disaster prevention and early warning technologies, such as static and delayed nature, lack of predictive ability, low efficiency, and insufficient accuracy, this disclosure provides a public broadband early warning method and system based on disaster prevention, as well as electronic devices, computer-readable storage media, and computer program products. It achieves accurate pre-disaster early warning by deeply integrating dynamic network topology, device geographic information, and meteorological forecast paths, automatically and quickly locating specific affected users from the disaster map. This solves the pain points of delays and inefficiencies in traditional manual methods, greatly improving emergency response capabilities and network resilience.

[0009] In a first aspect, this disclosure provides a public broadband early warning method based on disaster prevention, the method comprising:

[0010] Construct a complete routing logic chain from user broadband account to core network based on multi-source network data;

[0011] Establish a network resource geographic information layer that includes the location information of user terminals and network devices;

[0012] Acquire predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas;

[0013] Based on the entire routing logic chain and the potentially affected areas, all potentially affected user broadband accounts are mapped out;

[0014] Based on historical disaster data and network performance data, a predictive model is used to assess the network vulnerability of the potentially affected areas and the scope of affected users.

[0015] Furthermore, the construction of the end-to-end routing logical chain from the user's broadband account to the core network based on multi-source network data includes:

[0016] Collect network data from authentication systems, network element management systems, address allocation systems, and link layer discovery protocols;

[0017] Based on the network data, a dynamic end-to-end network topology is constructed;

[0018] Through correlation analysis, a complete routing logical chain is generated from the user's broadband account through the access device, aggregation device to the core router;

[0019] The authentication system is the Radius (Remote Authentication Dial In User Service) system, which is used to collect the association information between the BRAS (Broadband Remote Access Server) and the broadband account.

[0020] The network element management system is used to collect the association information between OLT (Optical Line Terminal) and broadband account;

[0021] The address allocation system is a DHCP (Dynamic Host Configuration Protocol) system, used to collect the correspondence information between BRAS and OLT;

[0022] The link layer discovery protocol is used to collect end-to-end interconnection information between network devices.

[0023] Furthermore, the establishment of a network resource geographic information layer containing location information of user terminals and network devices includes:

[0024] The latitude and longitude information of the user-side ONT (Optical Network Terminal / Optical Modem) is collected using the TR069 protocol;

[0025] Based on the collected latitude and longitude information of the optical modem, a clustering algorithm is used to form a coverage area layer where users gather;

[0026] Associate the coverage area layer and the latitude and longitude of the network device with an online or offline map that uses a tile data model;

[0027] The tile data model is a pyramid structure, with different levels of tiles corresponding to different map resolutions with varying scaling levels.

[0028] Furthermore, the acquisition of predicted disaster data and its spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas includes:

[0029] The predicted path, actual path, eye location, wind circle range, and wind force level data of the typhoon are obtained from the meteorological bureau system through the API (Application Programming Interface).

[0030] The acquired typhoon data is overlaid with the network resource geographic information layer;

[0031] Based on the overlay results, the network devices and user coverage areas within the wind circle's coverage area are automatically identified.

[0032] Furthermore, the assessment of network vulnerability and the scope of affected users in the potentially affected areas using predictive models includes:

[0033] Time series analysis models are used to analyze the long-term trends and periodic patterns between disasters and network performance; and / or,

[0034] The probabilistic variance assessment method is used to evaluate network stability by calculating the degree of network performance fluctuation caused by disaster events.

[0035] Furthermore, the method also includes:

[0036] Collect multi-source auxiliary parameters, which include at least one parameter collected from the following systems:

[0037] The computer room temperature, humidity, battery storage capacity, and input voltage and current parameters are collected from the power monitoring system.

[0038] The number of repeated complaints from users collected from the customer service system is used to identify sensitive users and is weighted in the predictive model.

[0039] The latitude and longitude coordinates, routing information, and laying method parameters of optical cable lines are collected from the pipeline resource system;

[0040] Data on the changes in radio performance of base stations under different rainfall and wind conditions, collected from the wireless network system;

[0041] When assessing network vulnerability, the multi-source auxiliary parameters are input as weighting parameters into the prediction model to optimize its prediction results.

[0042] Furthermore, the method also includes:

[0043] The prediction model is trained offline using historical disaster data, and then analyzed and optimized online using real-time data.

[0044] Furthermore, the method also includes:

[0045] Real-time monitoring of alarm information and port performance data of network devices;

[0046] When a device failure or port congestion is detected, the entire route is used to reverse-engineer and generate a real-time list of affected users.

[0047] Secondly, this disclosure provides a public broadband early warning system based on disaster prevention, the system comprising:

[0048] The routing logic construction module is configured to build a complete routing logic chain from the user's broadband account to the core network based on multi-source network data.

[0049] The geographic information mapping module is configured to create a network resource geographic information layer that includes the location information of user terminals and network devices.

[0050] The spatial analysis module is configured to acquire predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas;

[0051] The user mapping module is configured to map all potentially affected user broadband accounts based on the end-to-end routing logical chain and the potentially affected areas.

[0052] The intelligent prediction module is configured to use a prediction model to assess the network vulnerability of the potentially affected areas and the scope of affected users based on historical disaster data and network performance data.

[0053] Thirdly, this disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores one or more computer programs executable by the at least one processor, the one or more computer programs being executed by the at least one processor to enable the at least one processor to perform the above-described disaster prevention-based public broadband early warning method.

[0054] Fourthly, this disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described disaster prevention-based public broadband early warning method.

[0055] Fifthly, this disclosure provides a computer program product comprising computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is executed in a processor of an electronic device, the processor in the electronic device performs the aforementioned public broadband early warning method based on disaster prevention.

[0056] Beneficial effects:

[0057] This disclosure provides a disaster prevention-based public broadband early warning method and system, electronic equipment, computer-readable storage medium, and computer program product. It automates and refines disaster impact assessment. Through the collaboration of a "full-process routing logic chain" and a "network resource geographic information layer," it can instantly transform abstract disaster paths (such as typhoon wind circles) into a concrete list of user accounts to be affected. Before a disaster occurs, it accurately maps the "disaster impact map," elevating vague regional early warnings to precise household-level potential fault location, providing unprecedented data support for resource pre-positioning and emergency decision-making. It also significantly improves the speed and efficiency of emergency response. Furthermore, it enables fully automated operation, allowing for near real-time monitoring of the threat level and impact scale facing the network, thereby enabling rapid formulation and execution of response measures to minimize the impact of disasters on communication services and user experience. By building a quantitative assessment capability for network resilience, it can assess network "vulnerability" and predict the impact range under different disaster intensities. This allows operators to identify weak points in the network, providing a scientific basis for future network planning and reinforcement investments, thus systematically improving the network's ability to withstand disasters in the long term.

[0058] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0059] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the embodiments of the present disclosure to explain the disclosure and do not constitute a limitation thereof. The above and other features and advantages will become more apparent to those skilled in the art from the detailed description of exemplary embodiments with reference to the accompanying drawings, in which:

[0060] Figure 1 A flowchart illustrating a public broadband early warning method based on disaster prevention provided in Embodiment 1 of this disclosure;

[0061] Figure 2 A schematic diagram of an exemplary end-to-end ledger network topology provided for embodiments of this disclosure;

[0062] Figure 3 A topology diagram for an example of fault or relay blockage determination provided in this disclosure embodiment;

[0063] Figure 4 A block diagram of a public broadband early warning system based on disaster prevention, provided in Embodiment 2 of this disclosure;

[0064] Figure 5 This is a block diagram of an electronic device provided in Embodiment 3 of this disclosure. Detailed Implementation

[0065] To enable those skilled in the art to better understand the technical solutions of this disclosure, exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments of this disclosure to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0066] Where there is no conflict, the various embodiments of this disclosure and the features thereof in the embodiments may be combined with each other.

[0067] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0068] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Words such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0069] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein. Those skilled in the art will understand that the specific order of execution of the steps in the methods described above in the specific embodiments should be determined by their function and possible internal logic.

[0070] To enhance the readability of the embodiments of the present invention, the following is a brief explanation of the English abbreviations that may appear in the embodiments of the present invention:

[0071] Radius is short for Remote Authentication Dial In User Service.

[0072] PPPoE (Point to Point Protocol over Ethernet) carries the PPP protocol (point-to-point connection protocol) on Ethernet. It uses Ethernet to form a network of a large number of hosts, which are connected to the Internet through a remote access device, and each host connected is controlled.

[0073] IPoE (Internet Protocol over Ethernet) is a DHCP+authentication technology. Authentication and billing are based on the physical location of the user (identified by a unique VLAN ID / PVC ID). Users do not need to enter a username and password when accessing the Internet.

[0074] DHCP (Dynamic Host Configuration Protocol) is a local area network protocol that uses the UDP protocol. It has two main uses: automatically assigning IP addresses to internal networks or network service providers, and providing users or internal network administrators with a means of centrally managing all computers. It is described in detail in RFC 2131.

[0075] BRAS (Broadband Remote Access Server) is a new type of access gateway for broadband network applications. It is located at the edge layer of the backbone network and can complete data access for user bandwidth, enabling broadband Internet access for commercial buildings and residential residents, building enterprise intranets, and other applications.

[0076] IP (Internet Protocol) is a protocol designed for communication between interconnected computer networks.

[0077] A VLAN (Virtual LAN) is a logical group of devices and users that are not limited by physical location. They can be organized based on factors such as function, department, and application, and communicate with each other as if they were on the same network segment, hence the name Virtual LAN. A VLAN is a broadcast domain, and communication between VLANs is accomplished through Layer 3 routers.

[0078] QinQ is a descriptive name for a tunneling protocol based on IEEE 802.1Q encapsulation. QinQ technology adds an outer VLAN tag on top of the original VLAN tag (inner tag). The outer tag can mask the inner tag. The outer tag is SVLAN, and the inner tag is CVLAN.

[0079] SVLAN (Service Provider VLAN), outer VLAN, also known as VLAN stacking.

[0080] CVLAN (Customer VLAN), an inner VLAN, also known as a user VLAN.

[0081] LLDP (Link Layer Discovery Protocol) is a proximity discovery protocol. It defines a standard method for Ethernet network devices, such as switches, routers, and wireless LAN access points, to announce their presence to other nodes in the network and to store discovery information for each neighboring device.

[0082] CR (Core Router), also known as "backbone router", is a router located at the center of the network. It connects to the China Mobile Internet backbone network, performs high-speed data forwarding, and acts as an IP metropolitan area network exit device.

[0083] A switch (SW) is a device in a communication system that performs information exchange functions and can provide a dedicated electrical signal path for any two network nodes connected to the switch.

[0084] OLT (optical line terminal) is a terminal device used to connect optical fiber trunk lines.

[0085] ONU (Optical Network Unit) is divided into active optical network units and passive optical network units.

[0086] EMS (network element management system) is a system that manages one or more telecommunications network elements of a specific type.

[0087] LOID stands for Line Identifier. It is a marker configured on the ONT to identify user information. Upper-layer devices will send corresponding broadband service data based on the LOID configured on the ONT.

[0088] SN (Serial Number, also called SerialNo, product serial number) is an identification code for a terminal device. The product serial number is a concept introduced to verify the "legal identity of a product," protecting users' legitimate rights and ensuring they enjoy legal services; each genuine product corresponds to only one product serial number. Other names include: machine code, authentication code, registration application code, etc.

[0089] MAC (Media Access Control) addresses are burned into the Network Interface Card (NIC). A MAC address, also called a hardware address, consists of 48 bits of hexadecimal numbers.

[0090] MAC address authentication is a method of controlling a user's network access permissions based on port and MAC address. It does not require the user to install any client software. The device initiates authentication for that user upon first detecting their MAC address. During authentication, the user does not need to manually enter a username or password.

[0091] PON (Passive Optical Network) refers to an optical distribution network that contains no electronic components or power supplies. The entire ODN is composed of passive components such as optical splitters, eliminating the need for expensive active electronic devices.

[0092] SNMP (Simple Network Management Protocol).

[0093] PDU (Protocol Data Unit) refers to the data unit transmitted between peer layers.

[0094] Trap is one of the core SNMP PDU messages. SNMP agents use Trap to send unsolicited messages to the SNMP management station, typically describing the occurrence of an event. This event can be an alarm, alarm recovery, notification, etc., such as interface UP / DOWN, IP address change, etc.

[0095] syslog is the default logging daemon in Linux systems.

[0096] The public broadband early warning method based on disaster prevention according to embodiments of this disclosure can be executed by electronic devices such as terminal devices or servers. Terminal devices can be in-vehicle devices, user equipment (UE), mobile devices, user terminals, terminals, cellular phones, cordless phones, personal digital assistants (PDAs), handheld devices, computing devices, in-vehicle devices, wearable devices, etc. The method can be implemented by a processor calling computer-readable program instructions stored in memory. Alternatively, the method can be executed by a server.

[0097] Example 1

[0098] Figure 1 This is a flowchart illustrating a public broadband early warning method based on disaster prevention, provided in Embodiment 1 of this disclosure. (Refer to...) Figure 1 The method includes:

[0099] Step S101: Construct a complete routing logical chain from the user's broadband account to the core network based on multi-source network data;

[0100] Step S102: Establish a network resource geographic information layer containing the location information of user terminals and network devices;

[0101] Step S103: Obtain the predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas;

[0102] Step S104: Based on the end-to-end routing logical chain and the potentially affected areas, map out all potentially affected user broadband accounts;

[0103] Step S105: Based on historical disaster data and network performance data, use a predictive model to assess the network vulnerability of the potentially affected areas and the scope of affected users.

[0104] The purpose of this disclosure is to enable operators to determine changes in network (broadband capacity and bandwidth) and bandwidth during the three periods before, during, and after a disaster in disaster prevention, mitigation, and emergency rescue scenarios. Through big data analysis and historical typhoon level projections, it calculates the network's capacity to withstand typhoons and earthquakes, providing a comprehensive assessment of network operations at the city, county, town, and village levels, as well as the network's resilience and user situation. It also implements a fully automated reporting mechanism, eliminating the need for manual data collection and reporting. Simultaneously, it establishes a layer with the meteorological bureau's API map, predicting and assessing disaster situations and resilience by studying overall changes in the network topology, resolving potential network failures, ensuring normal operation of operator services, and preventing failures from impacting user experience.

[0105] This disclosure utilizes data fusion and intelligent analysis to achieve accurate prediction of the impact of disasters on public broadband networks. Specifically, it includes:

[0106] Step S101: Construct a complete routing logical chain from the user's broadband account to the core network based on multi-source network data, including:

[0107] Data Acquisition: The following multi-source data is collected periodically (e.g., every minute) or triggered via automated scripts and network management interfaces:

[0108] Parse and extract the following fields from the PPPoE authentication messages of the Radius system: user account, BRAS IP address, BRAS physical slot number, port number, SVLAN, and CVLAN.

[0109] From the EMS network element management system, resource data of the OLT device is obtained directly via SNMP protocol, including: OLT management IP, LOID or SN code of the connected ONU, PON port number of OLT, port number of ONU and its corresponding VLAN information.

[0110] Extract the "BRAS IP" and "OLT alias" fields from the IPTV (IPoE) authentication messages of the DHCP system, and establish the hierarchical association between BRAS and OLT.

[0111] Enable the LLDP protocol on routers, switches, and OLTs in the network, and have the IP network management system collect the interconnection information between each pair of devices, including the local device IP, local port, peer device IP, and peer port, to form the original link dataset.

[0112] Topology Construction and Route Generation: The collected data is pushed to a big data platform (such as one built on Hadoop or Spark). The platform first uses the APRIORI association rule algorithm or Graph Convolutional Network (GCN) to clean, merge, and restore the complex "spider web" links formed by LLDP, automatically generating an accurate, dynamically updated end-to-end network topology map. Subsequently, by associating data from Radius, EMS, and DHCP, user accounts are bound to specific OLT and BRAS ports, thereby calculating the complete end-to-end logical routing chain from "user account -> optical modem (ONT) -> OLT -> aggregation switch -> BRAS -> core router (CR)," and persistently storing it in the database.

[0113] This embodiment of the disclosure automates and enables real-time network resource management through the aforementioned steps, completely changing the traditional static ledger model that relies on manual data entry and is slow to update. It generates precise logical dependencies down to the household level, laying an irreplaceable data foundation for subsequent rapid and accurate fault impact analysis.

[0114] Step S102: Establish a network resource geographic information layer containing the location information of user terminals and network devices, including:

[0115] Location information collection:

[0116] User terminal: Through the operator's unified TR069 protocol management platform, a query command is sent to the optical modem (ONT) in the user's home to obtain the latitude and longitude information obtained from its built-in GPS module or by resolving the IP address.

[0117] Network equipment: Obtain the precise latitude and longitude of the computer room or base station where the core router, BRAS, OLT and other network equipment are located from the resource management system or manually entered ledgers.

[0118] Layer Construction: The massive amounts of user terminal latitude and longitude information collected are analyzed using the DBSCAN clustering algorithm to create user cluster "heat maps" or coverage area polygons based on communities and villages. Simultaneously, network devices are treated as point features. All this geographic information is linked to the operator's offline electronic maps (organized using a pyramid tile model, such as ZoomLevel 15 with street-level detail). The map itself has already loaded electronic fences (i.e., administrative boundary polygons) for provinces, cities, districts / counties, and townships / streets.

[0119] By establishing a geographic information layer for network resources, the virtual, logical network world is accurately mapped to the real physical geographic space. Through the clustering of user locations, an advancement from "individual user points" to "user coverage" is achieved, making regional impact assessments possible and corresponding to administrative levels.

[0120] Step S103: Acquire predicted disaster data and perform spatial overlay analysis, including:

[0121] Data Acquisition: Access the provincial / municipal meteorological bureau's data service platform periodically (e.g., every 10 or 30 minutes) through RESTful API interfaces to obtain structured disaster forecast data, such as the typhoon's predicted path (a line composed of a series of latitude and longitude points), the radius of the level 7 / 10 wind circle (forming a polygon), the storm's center location, movement speed, and central pressure.

[0122] Spatial Analysis: In a Geographic Information System (GIS) engine (such as PostGIS or GeoPandas), a spatial overlay analysis is performed between the wind circle polygons provided by the meteorological bureau and the "network resource geographic information layer" established in step two. Specifically, an "Intersect" query is executed to quickly identify all user coverage areas and network device locations that fall within or intersect with the wind circle polygons.

[0123] Spatial overlay analysis enables deep fusion of cross-domain data, directly applying the predictive capabilities of meteorological science to communication network operation and maintenance. It can automatically and quickly identify micro-level, specific network asset risk points from macro-level disaster forecasts, completing the critical transformation from "natural disasters" to "network disasters."

[0124] Step S104: Based on the end-to-end routing logical chain and the potentially affected areas, map out all potentially affected user broadband accounts.

[0125] After spatial analysis identifies the "potentially affected areas," the system generates a "list of affected network devices" (e.g., OLT-01, OLT-02). Subsequently, the system performs an efficient query in the database, based on the end-to-end routing logical chain generated in step one, to identify all user broadband accounts whose routing paths contain these OLT devices, ultimately generating an accurate list of affected user accounts.

[0126] This allows for precise location of disaster impact down to the household level, transforming a vague early warning for "a region" into a precise early warning for "a specific list of users." This list serves as the most direct basis for subsequent targeted user notifications and resource allocation.

[0127] Step S105: Based on historical disaster data and network performance data, use a predictive model to assess the network vulnerability of the potentially affected areas and the scope of affected users.

[0128] The predictive model used in this step integrates multiple algorithms. For example, the ARIMA time series model is used to analyze the relationship between base station outage rates and wind speed and rainfall during historical typhoons; analysis of variance is used to calculate the probability and stability of fiber optic cable interruptions (i.e., signal transmission interruptions) under specific wind force levels. The model's input data includes historical disaster data and real-time network performance (such as port traffic and CRC error counts). Evaluation and output: The model analyzes the user list obtained in the previous step and outputs two types of results: Network vulnerability: A quantitative rating (e.g., "high risk," "medium risk," "low risk") is given for network facilities in potentially affected areas. Affected user scope: The proportion and number of users whose services will ultimately be interrupted under specific disaster intensities are predicted, and different levels such as "inevitable interruption" and "high probability of interruption" can be distinguished.

[0129] The model-based prediction system represents a leap from qualitative judgment to quantitative forecasting, enabling operators to not only know "there will be an impact," but also predict "how severe the impact will be." This provides operators with scientific decision-making support, allowing them to prioritize responses and allocate limited emergency resources to the most critical and vulnerable areas, thereby maximizing disaster prevention benefits. Through continuous data feedback, the model can continuously self-optimize, forming a virtuous cycle of becoming increasingly intelligent with use.

[0130] This disclosed embodiment, through the collaboration of an "end-to-end routing logic chain" and a "network resource geographic information layer," can instantly transform abstract disaster paths (such as typhoon wind circles) into a concrete list of user accounts to be affected. Before a disaster occurs, a precise "disaster map" is drawn, elevating vague regional warnings to precise household-level potential fault location, providing unprecedented data support for resource pre-positioning and emergency decision-making. It also significantly improves the speed and efficiency of emergency response. Furthermore, it enables fully automated operation, allowing for near real-time monitoring of the threat level and impact scale facing the network, thereby enabling rapid formulation and execution of countermeasures to minimize the impact of disasters on communication services and user experience. By constructing a quantitative assessment capability for network resilience, it is possible to assess network "vulnerability" and predict the impact range under different disaster intensities. This allows operators to identify weak points in the network, providing a scientific basis for future network planning and reinforcement investments, thus systematically improving the network's ability to withstand disasters in the long term.

[0131] Furthermore, the construction of the end-to-end routing logical chain from the user's broadband account to the core network based on multi-source network data includes:

[0132] Collect network data from authentication systems, network element management systems, address allocation systems, and link layer discovery protocols;

[0133] Based on the network data, a dynamic end-to-end network topology is constructed;

[0134] Through correlation analysis, a complete routing logical chain is generated from the user's broadband account through the access device, aggregation device to the core router;

[0135] The authentication system is the Radius system, which is used to collect the association information between the Broadband Remote Access Server (BRAS) and the broadband account.

[0136] The network element management system is used to collect the association information between the optical line terminal (OLT) and the broadband account;

[0137] The address allocation system is a DHCP system, used to collect the correspondence information between BRAS and OLT;

[0138] The Link Layer Discovery Protocol (LLDP) is used to collect end-to-end interconnection information between network devices.

[0139] Building a complete routing logical chain from the user's broadband account to the core network includes the following steps:

[0140] Step 1: Establish a pre-defined network topology and latitude / longitude coordinates. The following data needs to be collected:

[0141] RADIUS system message: This message contains the possible uniqueness of the logical topology for a user's internet access, a five-tuple, the BRAS IP address, the BRAS slot number, the BRAS port number, the BRAS SVLAN, and the BRAS CVLAN. When a broadband user accesses the network via a single BRAS router, this five-tuple information can determine the uniqueness of the customer. However, in live networks, there may be redundant BRAS accesses, requiring the use of other data for verification.

[0142] Data packets from the access network manufacturer's EMS system: The following fields are extracted from the EMS system's resource information: OLTIP, LOID SN authentication (MAC authentication), OLT device name, slot PON port number, ONU slot port number, SVLAN, and CVLAN. This information associates the physical port and logical information of a customer's access information under the OLT. This association with other packets plays a crucial role, as other packets cannot be associated with the specific physical port of the last associated device.

[0143] DHCP system data packets: These packets contain key information such as the BRAS IP address and OLT alias. Their primary function is to determine the mapping between the OLT and the BRAS, identifying the network elements connected to the BRAS.

[0144] LLDP collects end-to-end data packets: Routers, switches, OLTs, and hosts exchange IP addresses and physical port information with each other during interconnection. This allows you to obtain the topology information for the entire physical route from the router to the end OLT.

[0145] The data acquisition and processing process includes:

[0146] 1.1 Acquisition of the association information between BRAS and broadband account based on message data from the Radius system.

[0147] PPPoE authentication is based on the user's logical physical location for authentication and billing. It records the user's online and offline authentication records. By parsing the PPPoE packets of the Radius system, fields such as the IP address, BRAS slot, port number, SVLAN, and CVLAN packets of the Internet access authentication BRAS can be extracted.

[0148] 1.2 Resource information collection based on direct procurement network elements: association information between OLT and broadband account.

[0149] Information such as OLT and ONU uplink and downlink devices, slots, ports, and VLANs is collected directly via SNMP. The following fields are extracted by reading network element information via the SNMP protocol: OLT IP, LOID SN authentication (MAC authentication), OLT device name, slot PON port number, ONU slot port number, SVLAN, and CVLAN.

[0150] 1.3 Acquisition of the correspondence between BRAS and OLT in message data of IPTV-based DHCP system.

[0151] Extract the IP address and OLT alias of the Internet access authentication BRAS from the IPTV IPOE authentication message on the existing network to obtain the relationship between the two.

[0152] 1.4 Collection of end-to-end ledger information based on Link Layer Neighbor Discovery Protocol (LLDP)

[0153] Deploy the LLDP protocol on routers, switches, and host devices within the metropolitan area network to automatically generate end-to-end interconnection ledger information for each device in the IP network management system. The IP network management system then generates a network topology diagram based on network hierarchy or device naming rules.

[0154] All carrier metropolitan area network (MAN) topologies are tree-structured, with four regions planned: Region A is the core router layer (CR routers); Region B is the aggregation router layer, containing BRAS, SR, and MSE routers; Region C is defined as the aggregation switch layer, containing SW switches (C1, C2, C3, etc., as multiple cascaded relationships exist within Region C); and Region D is defined as the access layer, containing OLT switches. With these four regions defined within the network framework, various devices automatically read their manufacturer's MIB information from the IP management system to obtain their model numbers and automatically populate the corresponding regions. However, the interrelationships between these regions are unclear, especially in Region C. Using LLDP protocol messages, which exchange IP addresses and physical ports, a spiderweb-like structure is created. The APRIORI algorithm in the big data platform is then used to automatically reconstruct this complex spiderweb structure.

[0155] 1.5 Pushing data to the big data platform based on the IP network management system

[0156] In metropolitan area network (MAN) systems, IP network management systems collect trap and syslog alarms, device performance data, and network topology information. This information is then pushed to a big data platform.

[0157] Trap and syslog alerts can inform the system of the currently faulty network element and its IP address. Device performance data can be extracted to determine port congestion. The network topology map helps identify whether there are other trunk links protecting the network element in case of failure; or, in multi-truncation scenarios, if one trunk fails, causing congestion in the remaining trunks.

[0158] Step 2: Generate the entire route;

[0159] Based on the device information (BRAS, OLT, ONU, etc.), link information, and VLAN information collected in step 1, the big data platform analyzes and correlates the data to calculate the entire route from the user's broadband account to the CR.

[0160] The above calculates the relationship between CR and OLT. The remaining part is the access part from OLT to optical modem ONT. From points 1.1 and 1.2, the specific ONT access information of OLT can be obtained.

[0161] Step 3: End-to-end dynamic resource tree for big data;

[0162] The same devices in each business path are pushed to the big data platform to form a dynamic resource tree end-to-end ledger.

[0163] like Figure 2 The image shows an exemplary end-to-end ledger. Figure 2 In this scenario, the IP network management system can read information about six network elements via SNMP. The model information of each element can be found in the MIB database. Based on the device model, devices r1-c-gddg-ssh and r2-c-gddg-NCxsk are located in area A; device 120.80.152.31 is in area B; 120.80.236.175 and 10.0.28.7 are in area C; and 10.2.95.2 is in area D. Devices in areas A and B exchange port information via LLDP protocol messages. Areas B and D also exchange interconnection IP addresses and port information. Using the APRIORI algorithm in the big data platform, the relationship between C1 and C2 can be determined. This is because B and C1 interact, C1 and C2 exchange information, and C2 and D exchange IP addresses and physical ports. This is a simple topology diagram. In the live network, the relationship between B, C, and D will be more replicated. In this diagram, there will be many horizontal lines or links that span several network elements.

[0164] This embodiment deploys data acquisition probes to periodically (e.g., every 5 minutes) poll the Radius system logs, using regular expressions to parse PPPoE packets and extract key fields such as Calling-Station-Id (user account), NAS-IP-Address (BRAS IP), and NAS-Port-Id (port). Simultaneously, it securely accesses the EMS of various vendors' OLTs via the SNMPv3 protocol, using predefined OIDs (e.g., 1.3.6.1.4.1.XXXX.1.1.3.0) to read ONU registration information in batches. DHCP packets are captured via mirrored switch ports, and the Option 82 field is parsed to associate BRAS and OLT aliases. LLDP is globally enabled on all Layer 2 / Layer 3 devices in the network, and the central network management system collects lldpLocPortTable and lldpRemTable via SNMP to obtain device adjacency relationships. This data is pushed to a Kafka message queue in real time for stream processing by Spark Streaming jobs. An improved FP-growth algorithm (an association rule algorithm) is employed to perform frequent itemset mining on massive LLDP adjacency relationships, automatically eliminating temporary loops and erroneous links to reconstruct a stable and accurate tree or ring network topology. Subsequently, using (BRAS IP, SVLAN, CVLAN) and (OLT IP, PON port, ONU SN) as composite keys, a JOIN operation is performed on Radius, EMS, and DHCP data. Finally, a complete logical routing chain in the form of key-value pairs ("User Account -> OLTPON Port -> Uplink Switch -> BRAS Port -> CR") is generated and stored in an in-memory computing engine (such as Redis). This step achieves fully automated and real-time construction of network topology and user routing relationships, improving the traditional model of relying on manual ledgers and weekly update cycles to the minute level. This provides a dynamic data foundation with millisecond-level query capabilities for subsequent precise impact analysis, and the automatic algorithmic correction significantly improves the accuracy and reliability of the data.

[0165] Furthermore, the establishment of a network resource geographic information layer containing location information of user terminals and network devices includes:

[0166] The latitude and longitude information of the user-side optical modem (ONT) is collected using the TR069 protocol.

[0167] Based on the collected latitude and longitude information of the optical modem, a clustering algorithm is used to form a coverage area layer where users gather;

[0168] Associate the coverage area layer and the latitude and longitude of the network device with an online or offline map that uses a tile data model;

[0169] The tile data model is a pyramid structure, with different levels of tiles corresponding to different map resolutions with varying scaling levels.

[0170] By collecting latitude and longitude information from user-side optical modems (ONTs) using the TR069 protocol, a TR069 ACS (Auto-Configuration Server) cluster can be deployed within the operator's integrated network management system. Through configured automated tasks, GetParameterValues ​​requests are periodically (e.g., daily at 2 AM) sent to all online ONTs across the network in batches to query the Device.DeviceInfo.GPSLocation parameter. For ONTs that do not support GPS modules, the system will invoke a high-precision IP address positioning service (e.g., combining IP address with base station LAC / CI information) to obtain their latitude and longitude coordinates via a REST API interface. The data format is uniformly WGS-84 coordinate system (longitude, latitude). This enables automated and large-scale collection of geographic location information from massive amounts of user terminals, replacing traditional manual data entry methods, ensuring the timeliness and accuracy of location data, and providing a reliable data source for subsequent spatial analysis.

[0171] Based on the collected latitude and longitude information of the optical modems, a clustering algorithm is used to form a coverage area layer representing user clusters. A big data platform (such as Spark) can be used to perform distributed processing on the collected latitude and longitude points. First, data cleaning is performed to remove outlier points (such as latitude and longitude values ​​of 0 or points outside administrative boundaries). Then, DBSCAN (a density-based spatial clustering algorithm) is used, with a neighborhood radius ε = 50 meters and a minimum number of points MinPts = 10 as clustering parameters. The algorithm aggregates densely adjacent user points into multiple polygonal regions, each polygon representing a user cluster (such as a residential community or village). The clustering results are stored in GeoJSON format, containing attributes such as polygon boundary coordinates, region ID, and the number of users included.

[0172] By intelligently aggregating massive numbers of discrete user points into coverage areas with clearly defined geographical boundaries, the complexity of data processing is significantly reduced, while user distribution becomes clearer. This regionalized representation facilitates spatial relationship analysis with administrative boundaries (electronic fences), laying the foundation for regional-level disaster impact assessments.

[0173] The process involves associating the covered area layer and network device latitude and longitude with online or offline maps using a tile data model. This includes building a web-based map service system and organizing map data using a pyramid tile data model. Map base tiles are organized by zoom level; for example, Zoom Level 15 corresponds to street-level detail (1:144,447 scale). On the server side, a PostGIS spatial database stores the user's covered area polygons and network device locations (latitude and longitude). When a client requests a map, the map engine (such as OpenLayers) quickly retrieves and returns the user's covered area and network device data within the current view bounding box and zoom level using spatial SQL queries (such as ST_Intersects), overlaying it onto the corresponding tile base map for rendering. When using an offline map, users can dynamically load the corresponding tile data based on the current zoom level. This is achieved by linking the city, district, county, town, and village electronic fences on the online map with the latitude and longitude of the covered area layer.

[0174] The tile model supports fast map loading and smooth zooming, and combined with the efficient retrieval capabilities of spatial databases, enables real-time visualization of millions of geographic features. This association method allows network resources (users and devices) to be precisely "pinned" onto the map, providing efficient technical support for subsequent spatial overlay analysis with meteorological data, while ensuring the system's response performance under large-scale data volumes. Offline map data models need to support efficient data storage and fast retrieval. A common model is to divide map data into multiple tiles, each tile representing a region on the map, and each tile has a unique identifier. Through the hierarchical structure of the tiles, map zooming operations can be implemented.

[0175] Through the three steps described above, this embodiment successfully constructs an accurate and dynamic geographic information layer for network resources, mapping abstract broadband users and devices to real geographic space, providing a crucial foundation for spatial analysis of disaster impacts. It addresses the pain point of missing or inaccurate geographic location information in traditional network management, achieving "spatialization" and "visualization" of network resource management, and significantly improving the accuracy and efficiency of disaster early warning.

[0176] Furthermore, the acquisition of predicted disaster data and its spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas includes:

[0177] The predicted path, actual path, eye location, wind circle range, and wind force level data of the typhoon can be obtained from the meteorological bureau system through the application programming interface (API).

[0178] The acquired typhoon data is overlaid with the network resource geographic information layer;

[0179] Based on the overlay results, the network devices and user coverage areas within the wind circle's coverage area are automatically identified.

[0180] Identifying potentially affected areas requires first obtaining typhoon forecast paths, actual paths, eye locations, wind circle extent, and wind speed data from the meteorological bureau system via an Application Programming Interface (API). This is achieved by deploying a data acquisition service in the early warning system's backend. This service calls standardized RESTful APIs provided by the meteorological department (e.g., the China Meteorological Administration (CMA) or the local meteorological observatory's early warning data interface), uses an authorized API key for authentication, and obtains structured typhoon data in JSON or XML format. The acquisition task is managed by a scheduling framework (such as Apache Airflow) and set to a high-frequency triggering mode (e.g., every 10 minutes) to ensure data timeliness. Key data fields acquired include: `forecast_path` (a string of predicted path lines consisting of a series of latitude and longitude points), `current_center` (the current location of the eye), `radius_7` and `radius_10` (representing the radii of wind circles at levels 7 and 10, respectively), and `max_wind_speed` (maximum wind speed).

[0181] This step enables automated and standardized integration with authoritative meteorological data, transforming unstructured weather forecasts into machine-readable and processable structured spatiotemporal data. This provides accurate and timely input for subsequent automated spatial analysis, completely changing the inefficient model that relies on manual interpretation of meteorological graphic announcements.

[0182] The acquired typhoon data is then overlaid with the network resource geographic information layer. The system backend uses a professional Geographic Information System (GIS) engine (such as PostGIS, GeoServer, or ArcGIS Server). First, the acquired typhoon data is processed in real time: using the eye coordinates as the center and the wind circle radius as the buffer distance, the ST_Buffer function of GIS dynamically generates circular or polygonal geometric objects to represent the coverage areas of different wind levels. Simultaneously, the predicted path point strings are constructed as LineString objects. Subsequently, in the GIS engine, these dynamically generated typhoon geometric layers are spatially registered with the network resource geographic information layer (containing "user coverage area polygons" and "network device locations"), which is persistently stored in the spatial database, under a unified coordinate system (e.g., WGS84). This transforms the abstract wind circle and path data from the meteorological field into precise geospatial graphics in real time, aligning them with the communication network resource layer in a unified digital map space. This step is a prerequisite for accurate spatial calculations, ensuring that data from different sources can be correlated under the same reference.

[0183] Based on the overlay results, the system automatically identifies network devices and user coverage areas within the wind circle's coverage area. It performs automated spatial relationship queries, returning the IDs of all network devices and user areas whose geometric locations intersect with or are located within the wind circle polygon, thus achieving accurate identification. This enables instantaneous and automatic mapping from "geospatial" to a "network asset list." The system can automatically output a detailed list of communication facilities and user groups located within the disaster threat area without requiring manual map comparison. This is not merely a simple visualization overlay, but rather generates a structured risk list with clear operational guidance value. It serves as direct input for subsequent impact mapping and prediction, significantly improving the accuracy and speed of emergency response.

[0184] Furthermore, the assessment of network vulnerability and the scope of affected users in the potentially affected areas using predictive models includes:

[0185] Time series analysis models are used to analyze the long-term trends and periodic patterns between disasters and network performance; and / or,

[0186] The probabilistic variance assessment method is used to evaluate network stability by calculating the degree of network performance fluctuation caused by disaster events.

[0187] The transformation of qualitative risk identification into quantitative scientific assessment is achieved through the collaborative work of two types of mathematical models:

[0188] 1. Using time series analysis models to analyze the long-term trends and periodic patterns between disasters and network performance: This method is used to uncover the inherent patterns between historical disaster events and network failure indicators.

[0189] Data Preparation: Collect time-series data of key performance indicators (KPIs) of the local network during each typhoon passage over the past 5-10 years. For example, collect hourly data on "OLT uplink packet loss rate", "number of base station outages", or "number of fiber optic cable interruption alarms" as dependent variables; simultaneously, collect corresponding meteorological data as independent variables, such as "maximum sustained wind speed" and "cumulative rainfall".

[0190] Model Building and Training: A seasonal autoregressive integrated moving average (SARIMA) model is built using statistical analysis tools (such as Python's statsmodels library). The model fitting process includes: testing the stationarity of the series and performing differencing (I), determining the autoregressive order (AR) and the moving average order (MA), and introducing a seasonal period (S, e.g., a 24-hour or yearly period). The model parameters are trained using methods such as maximum likelihood estimation.

[0191] Applications and Predictions: When new typhoon forecast data is input, the predicted wind speed-time curve and other parameters are used as exogenous variables and fed into a pre-trained SARIMA model. The model will output predicted values ​​and confidence intervals for network performance indicators (such as the number of out-of-service base stations) over a future period (e.g., hourly during the typhoon's impact). This reveals the trends and periodic patterns of network performance changes with disaster intensity.

[0192] This transforms experience-based, fuzzy judgments into quantitative trend predictions based on historical data patterns. It enables operations and maintenance personnel to anticipate the timing and severity of network performance degradation, providing a scientific basis for implementing phased emergency measures (such as activating specific plans before peak wind speeds), and achieving a clear understanding of the disaster's impact.

[0193] 2. A probabilistic variance assessment method is adopted to evaluate the stability of the network by calculating the degree of network performance fluctuation caused by disaster events: This method is used to quantitatively assess the reliability and uncertainty of the network when facing disasters of similar intensity.

[0194] Define the event and probability: For a specific evaluation object (such as a backbone optical cable), define a random variable X to represent its state under a "specific wind force level (such as level 12 wind)". Its possible values ​​(xi) include: "complete interruption", "performance degradation but usable", and "normal". Based on historical data statistics, calculate the prior probability P(xi) of each state.

[0195] Calculating Expectation and Variance: Assign a cost or severity score to each state (e.g., Interruption = 10, Deterioration = 3, Normal = 0), and then calculate the expected value E(X) of the event, which reflects the average severity. The key step is to calculate the variance Var(X) = Σ[Score(xi) - E(X)]²×P(xi). The magnitude of the variance Var(X) directly measures stability: a large variance indicates that the facility performs inconsistently in similar disasters, with significant fluctuations, high vulnerability, and poor stability; a small variance indicates that its performance consistently approaches expectations, demonstrating good resilience and high stability.

[0196] Vulnerability assessment output: The above variance calculation is performed on all critical facilities (such as computer rooms and optical distribution boxes) in the entire "potentially affected area", and the scores are normalized to generate a quantitative heat map of network vulnerability or a vulnerability level list (such as high, medium and low) for the area.

[0197] The probabilistic variance assessment method provides an innovative and inherently stable metric for evaluating network vulnerability. It focuses not only on "how bad it will be on average" (expected value) but also on "how uncertain the outcome is" (variance). This helps operators accurately identify the most unreliable and inconsistent "weak links" in their infrastructure, allowing them to allocate limited hardening resources to the critical nodes that best improve overall network stability, thus maximizing the return on disaster recovery investments.

[0198] Time series models provide longitudinal (time dimension) prediction curves, answering the question of "how the impact evolves during a disaster"; variance assessment methods provide horizontal (spatial dimension) stability scores, answering the question of "which nodes are most unreliable and at the highest risk". The combination of these two methods constitutes a comprehensive, multi-dimensional quantitative assessment system for the network resilience of "potentially affected areas," elevating prediction results from simple user number estimates to scientific decision support reports that include temporal trends and spatial vulnerability distributions, significantly enhancing the depth and value of early warning information.

[0199] Furthermore, the method also includes:

[0200] Collect multi-source auxiliary parameters, which include at least one parameter collected from the following systems:

[0201] The computer room temperature, humidity, battery storage capacity, and input voltage and current parameters are collected from the power monitoring system.

[0202] The number of repeated complaints from users collected from the customer service system is used to identify sensitive users and is weighted in the predictive model.

[0203] The latitude and longitude coordinates, routing information, and laying method parameters of optical cable lines are collected from the pipeline resource system;

[0204] Data on the changes in radio performance of base stations under different rainfall and wind conditions, collected from the wireless network system;

[0205] When assessing network vulnerability, the multi-source auxiliary parameters are input as weighting parameters into the prediction model to optimize its prediction results.

[0206] By constructing an enhanced predictive model that comprehensively reflects both the network's inherent "constitution" and its external "environment," deep fusion and intelligent weighting of multi-source heterogeneous data are achieved. This upgrades the early warning system from a "disaster impact warning device" to a "network resilience diagnosis and optimization platform," improving prediction accuracy. The model construction process includes:

[0207] Automated acquisition and standardized processing of multi-source auxiliary parameters:

[0208] Power environment parameter acquisition: The power environment monitoring system is polled in real time (e.g., every 2 minutes) via SNMP / Modbus protocol. Key indicators collected include: computer room temperature / humidity (to assess condensation and heat dissipation risks), current individual cell voltage and estimated remaining capacity of the battery pack (used to calculate the theoretical battery life based on the current load after a mains power outage, in minutes), and three-phase input voltage imbalance and current harmonic distortion rate (to assess power quality and potential power outage risks). After data cleaning, the data is stored in a time-series database (e.g., InfluxDB).

[0209] User perception parameter collection: Data is incrementally extracted daily from the customer service system's work order database using ETL tools (such as Kettle) or APIs. Business rules are defined: The number of complaints received by the same broadband account in the past 30 days due to "network interruption" or "slow internet speed" that have not been fully resolved is counted. Users with ≥3 complaints are marked as "highly sensitive users," and their respective OLTs or coverage areas are linked for further counting.

[0210] Physical route parameter acquisition: By calling the spatial query interface of the pipeline resource management system (GIS) (such as WFS service), based on the optical cable ID in the network topology, obtain its precise linear geographic path, laying method (enumerated values: overhead, pipeline, direct burial, underwater), and construction year / last maintenance year. These attributes will be quantified into "inherent risk coefficients", for example: overhead = 0.85, pipeline = 0.40, direct burial = 0.15; if the service life exceeds 15 years, the coefficient is multiplied by an aging factor of 1.3.

[0211] Wireless Environment Parameter Acquisition: Historical data was extracted from the wireless network performance management system. Regression analysis was used to calculate the "Received Signal Reference Power (RSRP) attenuation slope (dB / mm)" for each base station under different rainfall intensities (mm / h), and the "Block Error Rate (BLER) growth rate" within a specific wind direction sector. These slope values ​​characterize the base station's electromagnetic sensitivity to wind and rain.

[0212] Feature engineering and model weighted input:

[0213] Feature Construction: The original parameters described above are then constructed into machine learning features. For example:

[0214] For each network node (such as an aggregation room), construct the following characteristics: [battery life level, percentage of highly sensitive users, average risk coefficient of uplink optical cable].

[0215] Construct features for each user service path: [the maximum proportion of overhead optical cables traversed by the path, and the rain attenuation sensitivity of the terminal base station].

[0216] Model Integration and Weighting: During the training phase of the prediction model (such as using XGBoost or LightGBM gradient boosting tree models), these constructed features are input along with core disaster intensity features and network performance baseline features. The model automatically learns the importance (gain) of each feature through the training process. For example, the model may reveal that, in a typhoon warning scenario, the feature "the proportion of the maximum overhead optical cable along the path" contributes as much as 30% to the predicted probability of service interruption (gain value), far exceeding other features.

[0217] Weighted prediction: In actual prediction, the model performs a comprehensive calculation based on the input feature values ​​and their importance weights obtained during training. For example, even if two regions are predicted to have the same wind force, but one region has a high proportion of overhead optical cables and short battery life, the model will output a significantly higher "network vulnerability index" and a larger "proportion of expected affected users" for it.

[0218] By constructing an enhanced predictive model using multi-source auxiliary parameters, a qualitative leap in prediction accuracy can be achieved. By introducing multi-dimensional parameters such as power, customer service, pipeline, and wireless, the model evolves from merely describing "external pressure exerted by disasters" to simultaneously assessing "the network's own resilience." This is analogous to assessing human health by considering not only viral intensity (wind force) but also immunity, medical history, and lifestyle habits, transforming predictions from rough estimates to precise diagnoses and significantly reducing false alarms and missed alarms. It drives precise investment in network hardening. The model's automatically learned feature importance (such as the extremely high weight of "overhead optical cable ratio") provides data-driven decision-making for network planning and investment. Operators can clearly know which areas' overhead optical cables should be prioritized for ducting upgrades or which data centers should have their batteries upgraded to most effectively improve overall network resilience and maximize the benefits of disaster prevention investments. It also enhances the intelligence level of emergency response scheduling: when an early warning is generated, the system can not only list the number of affected users but also identify how many "highly sensitive users" (those prone to complaints) are included, and how long the backup power supply in the affected data centers can sustain operation. This enables emergency command centers to develop differentiated support strategies, such as prioritizing the deployment of generators to data centers with short battery life, or sending personalized reassurance notifications to highly sensitive users in advance, thereby improving customer perception and emergency response efficiency.

[0219] Furthermore, the method also includes:

[0220] The prediction model is trained offline using historical disaster data, and then analyzed and optimized online using real-time data.

[0221] The core of this "offline-online" linked machine learning operations (MLOps) pipeline is used for online analysis and model optimization. The specific process is as follows:

[0222] 1. Offline training phase:

[0223] Data Preparation and Feature Engineering: On a big data platform (such as Apache Spark), a historical training dataset covering many years is compiled. This dataset includes not only network performance data during historical typhoons, rainstorms, and other disasters (such as alarm logs and device performance metrics), but also multi-source auxiliary parameters aligned with it, as well as data on the actual impact after the disaster (i.e., "labeled data," such as lists of users whose actual services were interrupted and lists of faulty devices in each region). All raw data undergoes rigorous cleaning, alignment, and feature construction to form a standardized training feature table.

[0224] Model Training and Validation: Using mainstream machine learning frameworks (such as TensorFlow or PyTorch), with the aforementioned feature table as input and "whether business interruption occurred" or "the proportion of affected users" as prediction targets, the selected prediction model (such as gradient boosting trees or deep neural networks) is trained. The training process employs time-series cross-validation; for example, it always uses data from the past N-1 years to predict the situation in year N, simulating real-world forward-looking prediction scenarios and avoiding data skipping. Through hyperparameter tuning, the optimal model parameters are determined, generating a baseline model version (e.g., model_v1.0).

[0225] Model deployment preparation: Package the trained benchmark model and its required feature processing pipeline (such as normalizer and encoder), containerize it, and push it to the model repository to prepare for online service.

[0226] Online analysis and model optimization phase:

[0227] Online Inference and Service: Production models deployed in the model repository are published as high-performance microservice APIs through model service frameworks (such as TensorFlow Serving or KServe). When a real-time alert task is triggered, the preprocessed real-time feature vectors are sent to this API, and the model performs millisecond-level inference to generate prediction results.

[0228] Online learning and feedback loop: The system establishes a continuous data feedback loop. After each disaster event, the system automatically collects the "actual impact results" of the disaster and compares them with the model's previous "predictions," forming new training samples with "prediction-actual" labeled pairs. These new samples are fed into a stream processing pipeline in real time.

[0229] Short-term optimization (incremental learning / online learning): For models that support online learning, new samples can be used directly for incremental updates of the model in mini-batch form, enabling the model to quickly absorb the latest experience and adapt to minor changes in the network or climate.

[0230] Long-term optimization (regular full retraining): The system sets a scheduled task (e.g., weekly or monthly, or after accumulating a certain amount of new data) to automatically trigger a new full offline training. This training will use the entire dataset containing all historical data and the latest feedback data to retrain a completely new model version (e.g., model_v1.1). After the new model has undergone rigorous A / B testing or shadow mode verification that its performance is superior to the online version, it will be automatically or, after approval, rolled out to the production environment, replacing the old model.

[0231] Offline training enables the predictive model to continuously learn and evolve, dynamically adapting to changes in network structure, the deployment of new equipment, and shifts in climate patterns. This ensures the long-term effectiveness and accuracy of the early warning system, preventing the model from becoming obsolete due to environmental changes, creating a virtuous cycle of "becoming more accurate with use," and significantly enhancing the system's long-term investment value.

[0232] Furthermore, the method also includes:

[0233] Real-time monitoring of alarm information and port performance data of network devices;

[0234] When a device failure or port congestion is detected, the entire route is used to reverse-engineer and generate a real-time list of affected users.

[0235] By linking equipment performance with resources, and combining the dynamic resource tree end-to-end ledger with scenarios of equipment failure or relay congestion, real-time estimation of the customer list affected by failures or relay congestion is achieved.

[0236] like Figure 3 The diagram illustrates how to determine faults or trunk congestion (in both non-redundant and redundant scenarios). With two BRAS1 and BRAS2 devices, if BRAS2 fails, the original service traffic from the OLT connected to the switch (SW) would flow through both devices; now, all traffic flows through BRAS1. The network element topology (LLDP) changes, and due to the loss of BRAS2, only one uplink connection (BRAS1) remains. The performance (traffic) of the interaction port between BRAS1 and the switch (SW) is obtained from the IP network management system. Congestion is then assessed based on this traffic, and correlated with the number of users connected to the OLT. Simultaneously, an API query is initiated to the Radius database to retrieve the accounts of the users connected to the OLT. Radius then provides the number of online users and a list. Furthermore, big data analysis is used to estimate the time of the fault over the past few days, identifying frequently online users at different time intervals, and recording the number of users affected by the fault handling at different time intervals, along with a possible list. In this topology, if the switch (SW) fails, the real-time list of affected OLT network elements can be found in the Radius logs, showing the users who have just gone offline. It can accurately determine the number of users affected at the time and the list of affected users; and predict the time period during which repairs are not yet restored and the number of affected users for each time period based on big data (automatically learning and judging the potential impact based on the number of users collected daily).

[0237] The embodiments disclosed herein construct an independent real-time event analysis and impact localization pipeline, which runs in parallel with the prediction and early warning process, forming a dual-drive system of "pre-event early warning" and "in-event handling".

[0238] Real-time monitoring of network device alarm information and port performance data is achieved by deploying a high-performance streaming data processing engine (such as Apache Flink or Apache Storm). This engine consumes two main types of data streams from the IP network management system in real time: Alarm information stream: It receives and parses SNMP Trap and Syslog messages actively reported by network devices. Using a predefined regular expression rule base, it identifies key fault events such as linkDown (link interruption), coldStart (device cold start / reboot), and bgpBackwardTransition (BGP session interruption). Performance data stream: Through SNMP polling or NetFlow / sFlow telemetry technology, it continuously collects port performance data from key network devices (such as BRAS, aggregation switches, and OLT uplink ports). Core metrics include ingress / egress bandwidth utilization, packet error rate, and packet loss rate. Dynamic thresholds are set (e.g., utilization consistently >90% for 30 seconds or a sudden increase in packet error rate) to generate performance degradation events. It achieves millisecond-level awareness of network operating status, transforming traditional discrete and passive alarm processing into continuous and proactive network health monitoring, and providing event triggering sources for real-time impact analysis.

[0239] When device failure or port congestion is detected, the system, in conjunction with the end-to-end routing, reverse-engineers and generates a real-time list of affected users. When the stream processing engine identifies a high-priority failure event (such as a linkdown on an aggregation switch SW1) or a persistent congestion event, it immediately triggers an impact analysis job. Fault location: Extract the faulty device identifier (e.g., IP address) from the event. Topology query: Starting from the faulty device's IP address, query the generated and maintained end-to-end routing logical chain in memory (which can be viewed as a huge "user-device" dependency graph). Perform a fast reverse graph traversal algorithm. User list generation: The algorithm searches downstream (on the user side) along the dependencies, quickly finding all OLTs with the faulty device as a necessary node. Then, by real-time association with the online user table (e.g., real-time query from the Radius server), it obtains a list of all currently online user broadband accounts under these OLTs.

[0240] Output: A structured report is generated within seconds (usually <5 seconds), including: the faulty device, the time of failure, the number of affected OLTs, the estimated number of real-time online users affected, and a list of specific accounts (which can be anonymized). This report is automatically pushed to the operations and maintenance command dashboard and customer service system via API.

[0241] This disclosed embodiment achieves precise, second-level localization and quantification of the impact from "equipment failure" to "user impact." It automates and instantly completes tasks that traditionally required multi-departmental collaboration and hours of manual investigation across multiple systems. It also improves the closed-loop technical solution, enhancing the system's practical value: the system not only possesses pre-disaster warning capabilities but also in-disaster diagnostic capabilities. This constitutes a complete technical closed loop from "prediction" to "verification" to "response," greatly enhancing the practicality and indispensability of the entire solution in real-world operational scenarios.

[0242] This publicly disclosed implementation achieves intelligent management of the entire process of public broadband network impact before, during, and after disasters by deeply integrating multi-source network data from operators with meteorological disaster forecasts. By automatically constructing a precise end-to-end routing logical chain for "users-devices" and a geographic information layer for network resources, and performing spatial overlay analysis with real-time meteorological data, the system can automatically and accurately predict the potential affected user range and network vulnerability before a disaster, achieving a qualitative leap from "passive response" to "proactive early warning." During a disaster, the system can locate the real-time list of users affected by the fault within seconds based on real-time alarms, greatly improving emergency response efficiency. Simultaneously, through continuous multi-source data learning and model optimization, the system ensures the long-term stability and self-evolution of early warning accuracy. Ultimately, it builds an automated, precise, and intelligent network disaster resilience system for operators, significantly improving business assurance capabilities, user perception, and scientific decision-making levels.

[0243] Example 2

[0244] Embodiment 2 of this disclosure provides a public broadband early warning system based on disaster prevention, such as... Figure 4 As shown, the system includes:

[0245] The routing logic construction module 11 is configured to construct a complete routing logic chain from the user's broadband account to the core network based on multi-source network data.

[0246] Geographic information mapping module 12 is configured to create a network resource geographic information layer containing location information of user terminals and network devices.

[0247] Spatial analysis module 13 is configured to acquire predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas;

[0248] User mapping module 14 is configured to map all potentially affected user broadband accounts based on the end-to-end routing logical chain and the potentially affected areas;

[0249] The intelligent prediction module 15 is configured to use a prediction model to assess the network vulnerability of the potentially affected areas and the scope of affected users based on historical disaster data and network performance data.

[0250] Furthermore, the routing logic construction module 11 is specifically configured as follows:

[0251] Collect network data from authentication systems, network element management systems, address allocation systems, and link layer discovery protocols;

[0252] Based on the network data, a dynamic end-to-end network topology is constructed;

[0253] Through correlation analysis, a complete routing logical chain is generated from the user's broadband account through the access device, aggregation device to the core router;

[0254] The authentication system is the Radius system, which is used to collect the association information between the Broadband Remote Access Server (BRAS) and the broadband account.

[0255] The network element management system is used to collect the association information between the optical line terminal (OLT) and the broadband account;

[0256] The address allocation system is a DHCP system, used to collect the correspondence information between BRAS and OLT;

[0257] The link layer discovery protocol is used to collect end-to-end interconnection information between network devices.

[0258] Furthermore, the geographic information mapping module 12 is specifically configured as follows:

[0259] The latitude and longitude information of the user-side optical modem (ONT) is collected using the TR069 protocol.

[0260] Based on the collected latitude and longitude information of the optical modem, a clustering algorithm is used to form a coverage area layer where users gather;

[0261] Associate the coverage area layer and the latitude and longitude of the network device with an online or offline map that uses a tile data model;

[0262] The tile data model is a pyramid structure, with different levels of tiles corresponding to different map resolutions with varying scaling levels.

[0263] Furthermore, the spatial analysis module 13 is specifically configured as follows:

[0264] The predicted path, actual path, eye location, wind circle range, and wind force level data of the typhoon can be obtained from the meteorological bureau system through the application programming interface (API).

[0265] The acquired typhoon data is overlaid with the network resource geographic information layer;

[0266] Based on the overlay results, the network devices and user coverage areas within the wind circle's coverage area are automatically identified.

[0267] Furthermore, the user mapping module 14 is specifically configured as follows:

[0268] Time series analysis models are used to analyze the long-term trends and periodic patterns between disasters and network performance; and / or,

[0269] The probabilistic variance assessment method is used to evaluate network stability by calculating the degree of network performance fluctuation caused by disaster events.

[0270] Furthermore, the system also includes an optimization module 16:

[0271] The optimization module 16 is configured to collect multi-source auxiliary parameters, which include at least one parameter collected from the following systems:

[0272] The computer room temperature, humidity, battery storage capacity, and input voltage and current parameters are collected from the power monitoring system.

[0273] The number of repeated complaints from users collected from the customer service system is used to identify sensitive users and is weighted in the predictive model.

[0274] The latitude and longitude coordinates, routing information, and laying method parameters of optical cable lines are collected from the pipeline resource system;

[0275] Data on the changes in radio performance of base stations under different rainfall and wind conditions, collected from the wireless network system;

[0276] Furthermore, when assessing network vulnerability, the multi-source auxiliary parameters are input as weighting parameters into the prediction model to optimize its prediction results.

[0277] Furthermore, the optimization module 16 is also configured to use historical disaster data to train the prediction model offline, and use real-time data for online analysis and model optimization.

[0278] Furthermore, the system also includes a monitoring module 17;

[0279] The monitoring module 17 is configured to monitor the alarm information and port performance data of network devices in real time.

[0280] When a device failure or port congestion is detected, the entire route is used to reverse-engineer and generate a real-time list of affected users.

[0281] The public broadband early warning system based on disaster prevention in this embodiment is used to implement the public broadband early warning method based on disaster prevention in the first embodiment of the method, so the description is relatively simple. For details, please refer to the relevant descriptions in the previous method embodiments, which will not be repeated here.

[0282] Figure 5 This is a block diagram of an electronic device provided in Embodiment 3 of this disclosure.

[0283] Reference Figure 5 This disclosure provides an electronic device comprising: at least one processor 701; at least one memory 702; and one or more I / O interfaces 703 connected between the processor 701 and the memory 702; wherein the memory 702 stores one or more computer programs executable by the at least one processor 701, the one or more computer programs being executed by the at least one processor 701 to enable the at least one processor 701 to execute the aforementioned disaster prevention-based public broadband early warning method.

[0284] This disclosure also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the aforementioned disaster prevention-based public broadband early warning method. The computer-readable storage medium may be volatile or non-volatile.

[0285] This disclosure also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device executes the above-described public broadband early warning method based on disaster prevention.

[0286] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0287] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0288] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0289] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0290] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0291] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0292] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0293] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0294] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0295] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for general illustrative purposes only and should not be construed as limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this disclosure as set forth by the appended claims.

Claims

1. A public broadband early warning method based on disaster prevention, characterized in that, The method includes: Construct a complete routing logic chain from user broadband account to core network based on multi-source network data; Establish a network resource geographic information layer that includes the location information of user terminals and network devices; Acquire predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas; Based on the entire routing logic chain and the potentially affected areas, all potentially affected user broadband accounts are mapped out; Based on historical disaster data and network performance data, a predictive model is used to assess the network vulnerability of the potentially affected areas and the scope of affected users.

2. The method according to claim 1, characterized in that, The construction of the end-to-end routing logical chain from user broadband account to core network based on multi-source network data includes: Collect network data from authentication systems, network element management systems, address allocation systems, and link layer discovery protocols; Based on the network data, a dynamic end-to-end network topology is constructed; Through correlation analysis, a complete routing logical chain is generated from the user's broadband account through the access device, aggregation device to the core router; The authentication system is a remote authentication dial-up user service Radius system, used to collect the association information between the broadband remote access server BRAS and the broadband account. The network element management system is used to collect the association information between the optical line terminal (OLT) and the broadband account; The address allocation system is a Dynamic Host Configuration Protocol (DHCP) system, used to collect the correspondence information between BRAS and OLT; The link layer discovery protocol is used to collect end-to-end interconnection information between network devices.

3. The method according to claim 1, characterized in that, The process of establishing a network resource geographic information layer that includes the location information of user terminals and network devices includes: The latitude and longitude information of the user-side optical modem (ONT) is collected using the TR069 protocol. Based on the collected latitude and longitude information of the optical modem, a clustering algorithm is used to form a coverage area layer where users gather; Associate the coverage area layer and the latitude and longitude of the network device with an online or offline map that uses a tile data model; The tile data model is a pyramid structure, with different levels of tiles corresponding to different map resolutions with varying scaling levels.

4. The method according to claim 1, characterized in that, The process of acquiring predicted disaster data and performing spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas includes: The predicted path, actual path, eye location, wind circle range, and wind force level data of the typhoon can be obtained from the meteorological bureau system through the application programming interface (API). The acquired typhoon data is overlaid with the network resource geographic information layer; Based on the overlay results, the network devices and user coverage areas within the wind circle's coverage area are automatically identified.

5. The method according to claim 1, characterized in that, The assessment of network vulnerability and the range of affected users in the potentially affected areas using predictive models includes: Time series analysis models are used to analyze the long-term trends and periodic patterns between disasters and network performance; and / or, The probabilistic variance assessment method is used to evaluate network stability by calculating the degree of network performance fluctuation caused by disaster events.

6. The method according to claim 1, characterized in that, The method further includes: Collect multi-source auxiliary parameters, which include at least one parameter collected from the following systems: The computer room temperature, humidity, battery storage capacity, and input voltage and current parameters are collected from the power monitoring system. The number of repeated complaints from users collected from the customer service system is used to identify sensitive users and is weighted in the predictive model. The latitude and longitude coordinates, routing information, and laying method parameters of optical cable lines are collected from the pipeline resource system; Data on the changes in radio performance of base stations under different rainfall and wind conditions, collected from the wireless network system; When assessing network vulnerability, the multi-source auxiliary parameters are input as weighting parameters into the prediction model to optimize its prediction results.

7. The method according to claim 6, characterized in that, The method further includes: The prediction model is trained offline using historical disaster data, and then analyzed and optimized online using real-time data.

8. The method according to claim 1, characterized in that, The method further includes: Real-time monitoring of alarm information and port performance data of network devices; When a device failure or port congestion is detected, the entire route is used to reverse-engineer and generate a real-time list of affected users.

9. A public broadband early warning system based on disaster prevention, characterized in that, The system includes: The routing logic construction module is configured to build a complete routing logic chain from the user's broadband account to the core network based on multi-source network data. The geographic information mapping module is configured to create a network resource geographic information layer that includes the location information of user terminals and network devices. The spatial analysis module is configured to acquire predicted disaster data and perform spatial overlay analysis with the network resource geographic information layer to identify potentially affected areas; The user mapping module is configured to map all potentially affected user broadband accounts based on the end-to-end routing logical chain and the potentially affected areas. The intelligent prediction module is configured to use a prediction model to assess the network vulnerability of the potentially affected areas and the scope of affected users based on historical disaster data and network performance data.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, the one or more of the computer programs being executed by the at least one processor to enable the at least one processor to perform the disaster prevention-based public broadband early warning method as described in any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the public broadband early warning method based on disaster prevention as described in any one of claims 1-8.

12. A computer program product, characterized in that, Includes computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device performs the public broadband early warning method based on disaster prevention as described in any one of claims 1-8.

Images