Space weather disturbance feature recognition and early warning system based on artificial intelligence

Abstract

The invention discloses a space weather disturbance feature recognition and early warning system based on artificial intelligence, and relates to the technical field of space weather disturbance monitoring and early warning, and the system comprises a multi-source data integration and collection module, a double-source data fusion module, a multi-scale disturbance recognition module, a cross-domain cascade risk assessment module and a full-link early warning defense module. By collecting and processing multi-source data, fusing double-source data, accurately identifying disturbance, scientifically evaluating cross-domain risks and building a three-level linkage network, rapid early warning and defense are achieved, system parameters are optimized, and stable operation of key infrastructures is guaranteed. According to the invention, through multi-source data integration and double-source fusion, high-quality data support disturbance identification is provided, and real disturbance and false alarm are accurately distinguished; meanwhile, a cross-domain risk assessment system is constructed, the risk level is scientifically calculated, the propagation process is simulated, rapid early warning and defense are realized based on a three-level linkage network, the prevention and control capability is improved, and stable operation of key infrastructures is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of space weather disturbance monitoring and early warning technology, specifically to an artificial intelligence-based space weather disturbance feature recognition and early warning system. Background Technology

[0002] This invention relates to the field of space weather disturbance monitoring and early warning technology. With the expansion of space activities and the advancement of information society construction, the dependence of key infrastructures such as power grids, communication networks, satellite navigation systems, and financial systems on the space environment has significantly increased. The potential threats posed by space weather disturbances are becoming increasingly prominent. Space weather disturbances can trigger phenomena such as ionospheric fluctuations, geomagnetic anomalies, and solar wind changes. These phenomena may interfere with satellite signal transmission, affect the operation of power equipment, disrupt the stability of communication links, and even cause the interruption of financial transaction systems, impacting social and economic order and public safety. Currently, space weather-related data sources cover various types, including satellite measurements, numerical model simulations, and data from the infrastructure's own operations. However, different data differ in format standards and spatiotemporal scales. How to effectively integrate these scattered data and transform them into effective information that can be used for disturbance identification has become a core requirement for improving space weather early warning capabilities, and it is also driving the development of related monitoring and early warning technologies towards multi-source collaboration, precision, and efficiency.

[0003] Traditional space weather disturbance monitoring and early warning technologies have several limitations in practical applications. In the data processing stage, traditional methods struggle to achieve deep fusion of multi-source data. Data quality often fails to meet the requirements for accurate analysis due to spatiotemporal mismatch, missing data in sparse areas, or noise interference, affecting the reliability of subsequent disturbance identification. In terms of disturbance identification, there is a lack of effective means to address the imbalance between a few extreme event samples and the majority of normal samples in the dataset. Furthermore, fixed feature extraction and clustering strategies are often used, which cannot be dynamically adjusted according to changes in space weather, easily generating false alarm signals and making it difficult to accurately determine the type, intensity, and scope of impact of disturbances. In the field of risk prevention and control, traditional technologies often focus on risk assessment of single infrastructures, neglecting the interdependence between different fields. They cannot simulate the cross-field cascading propagation process of risks, and the early warning and defense measures lack coordination, making it difficult to quickly respond to systemic risks caused by space weather disturbances. Overall, the efficiency and effectiveness of prevention and control need to be improved. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an artificial intelligence-based space weather disturbance feature identification and early warning system. It integrates multi-source data collection, dual-source data fusion, multi-scale disturbance identification, cross-domain cascaded risk assessment, and full-link early warning and defense modules. The system uses high-precision data processing and fusion technology, dynamic PCA and DBSCAN to identify anomalies, combines probability models and node failure propagation algorithms to assess risks, and finally achieves rapid early warning and defense through a three-level linkage network. The full-link response time is controlled within 90 seconds, effectively improving the ability to respond to space weather events.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a space weather disturbance feature recognition and early warning system based on artificial intelligence, the system comprising:

[0006] Multi-source data integration and acquisition module: collects satellite measured data, space weather numerical model data, critical infrastructure basic data and cross-domain dependency data, and performs data standardization processing;

[0007] Dual-source data fusion module: It realizes the spatiotemporal registration of satellite measured data and numerical model data through linear interpolation, corrects data errors by Kalman filtering, completes dual-source data fusion by combining adaptive weighting rules, completes sparse region data by adaptive Kriging interpolation, eliminates data noise by sliding window smoothing technology, and normalizes infrastructure and dependency data.

[0008] Multi-scale disturbance identification module: Based on standardized data, features are extracted and disturbances are identified. An improved SMOTE-ENN algorithm is used to handle data imbalance. Dynamic PCA and DBSCAN density clustering technology are combined to identify abnormal disturbances and eliminate false alarms. The disturbance type, intensity and range of influence are determined by comparing with a preset disturbance feature library.

[0009] Cross-domain cascading risk assessment module: It uses a probability model to calculate the direct risk level of a single domain, simulates the cross-domain risk cascading propagation process through the cross-domain cascading risk propagation index, combines density clustering to verify and screen effective propagation paths, calculates the macro-systemic risk index, and outputs the risk level of a single domain, the cascading failure chain, and the systemic risk results.

[0010] The end-to-end early warning and defense module establishes a three-level linkage network, realizes hierarchical early warning command transmission and linkage triggering based on TCP / IP and MQTT protocols, executes single-domain defense measures and cross-domain collaborative defense operations, pushes early warning information through SMS and Web platforms, and collects response data to optimize system parameters.

[0011] Furthermore, the multi-source data integration and acquisition module acquires GPS-TEC ionospheric data, solar wind plasma density data, geomagnetic component observation data, global ionospheric model GIM output data, power grid topology, communication base station distribution information, satellite navigation receiver parameters, and financial transaction system configuration data through the IGS data interface, satellite ground receiving station, space weather monitoring and early warning center server, and infrastructure operation unit management platform, respectively. All data is converted into JSON format and field naming and data type are standardized.

[0012] Furthermore, the dual-source data fusion module uses a 1-degree × 1-degree latitude and longitude grid and UTC hourly intervals as spatiotemporal references. It interpolates non-hourly data to the hourly time using linear interpolation with an interpolation step of 1 minute. It employs Kalman filtering technology to correct errors by using satellite measured data as observations and numerical model data as predictions. It determines the reliability weight of measured data based on the degree of agreement between measured data and historical true values ​​over the past 72 hours, with a value range of 0.3 to 0.8. It uses adaptive Kriging interpolation to complete sparse region data and eliminates data noise using a 30-minute sliding window smoothing technique.

[0013] Furthermore, when the multi-scale perturbation identification module uses the improved SMOTE-ENN algorithm to handle the data imbalance problem, it defines extreme event samples as the minority class, selects 5 most similar samples in the feature space for each minority class sample, generates 3 to 5 new samples through linear interpolation, and then uses the edit nearest neighbor algorithm to remove overlapping noise samples for the majority class samples.

[0014] Furthermore, when the multi-scale disturbance identification module combines dynamic PCA and DBSCAN density clustering technology to identify abnormal disturbances and eliminate false alarms, it dynamically adjusts the size of the PCA sliding window according to the intensity of solar activity, performs principal component analysis on the magnetic field plasma density data within the window, extracts the top three principal components with a cumulative variance contribution rate of not less than 90%, identifies potential abnormal data by calculating residuals and setting a threshold of 3 times the standard deviation, and then uses DBSCAN density clustering technology to confirm the real disturbance.

[0015] Furthermore, the multi-scale disturbance identification module determines the disturbance type, intensity, and impact range by comparing it with a preset disturbance feature library. The feature library contains feature parameter ranges for four types of disturbances: ionospheric traveling wave disturbances, strong ionospheric scintillation, high-speed solar wind flow, and geomagnetic storms. The disturbance type is determined by calculating the matching degree, the intensity level is determined by combining the comprehensive disturbance feature index S, and the impact range is determined according to the 1-degree × 1-degree latitude and longitude grid distribution.

[0016] Furthermore, when the cross-domain cascaded risk assessment module calculates the direct risk level of a single domain using a probabilistic model, it takes the disturbance intensity output by the multi-scale disturbance identification module as input and establishes a probabilistic mapping relationship by combining the vulnerability parameters of the power grid, communication network, satellite navigation system, and financial system. The vulnerability parameter of the power grid is the transformer's resistance to geomagnetic induced current threshold, the communication network's resistance to ionospheric scintillation threshold, the satellite navigation system's positioning accuracy threshold, and the financial system's core communication link stability threshold. The disturbance intensity is quantitatively compared with the vulnerability thresholds of each domain. When the disturbance intensity is ≤30% of the vulnerability threshold, it is judged as Level 1 risk; when it is 31%-50% of the vulnerability threshold, it is judged as Level 2 risk; when it is 51%-70% of the vulnerability threshold, it is judged as Level 3 risk; when it is 71%-90% of the vulnerability threshold, it is judged as Level 4 risk; and when it is >90% of the vulnerability threshold, it is judged as Level 5 risk. The calculation of the direct risk level of each single domain is completed through this probabilistic mapping relationship.

[0017] Furthermore, in the cross-domain cascading risk assessment module, the calculation formula for the cross-domain cascading risk propagation index is as follows: ,in As a systemic cascading risk index, The initial direct risk value, This is the infrastructure dependency weight matrix. This represents the infrastructure vulnerability attenuation coefficient. For propagation path weighting coefficients, The number of upstream infrastructure types, Number of downstream infrastructure types This represents the number of effective cascading propagation paths.

[0018] Furthermore, when the cross-domain cascaded risk assessment module combines density clustering to verify and screen effective propagation paths, it uses the potential propagation paths constructed from the cross-domain dependency matrix as clustering objects, sets the neighborhood radius to 0.3 path similarity units, and the minimum number of samples to 3. Through density clustering, the paths are classified according to the strength of dependency and the similarity of propagation logic. Paths with no less than 3 samples within a cluster, a dependency weight of no less than 0.7, and a vulnerability decay coefficient of no less than 0.8 are selected as effective propagation paths. The macro-systemic risk index is calculated using the formula in claim 8 and further mapped to three levels of risk: low, medium, and high. When the systemic risk index ≤ 3, it is considered low risk; when 3 < systemic risk index ≤ 7, it is considered medium risk; and when the systemic risk index > 7, it is considered high risk.

[0019] Furthermore, the multi-domain linkage network of the full-link early warning and defense module includes a three-level architecture of spacecraft-satellite-ground, using X-band inter-satellite links and S-band telemetry and control links for communication, and using a dedicated fiber optic network and 4G / 5G network to transmit data within the ground system; the graded early warning is triggered based on the systemic risk index, and the full-link response time is controlled within 90 seconds.

[0020] Compared with existing technologies, this AI-based space weather disturbance feature recognition and early warning system has the following advantages:

[0021] I. This invention achieves standardized processing and efficient fusion of different types of space weather-related data through multi-source data integration and dual-source fusion technology, solving problems such as data spatiotemporal mismatch, sparsity, and noise interference. It provides high-quality data support for disturbance identification. In the disturbance identification stage, by optimizing the data imbalance processing method and dynamically adjusting feature extraction and clustering technology, it accurately distinguishes between real disturbances and false alarm signals, effectively determining the type, intensity, and impact range of disturbances, and significantly improving the accuracy and completeness of disturbance identification. This synergistic optimization of data processing and identification mechanisms breaks the limitations of traditional single data sources or fixed identification patterns, making the monitoring of space weather disturbances more comprehensive and reliable, providing accurate and effective basis for subsequent risk assessment, and helping to detect potential space weather threats in advance.

[0022] Second, this invention constructs a cross-domain cascaded risk assessment system, combining the characteristics and vulnerabilities of each critical infrastructure sector to scientifically calculate the direct risk level of a single sector. Simultaneously, it simulates the cross-domain risk propagation process, filters effective propagation paths, quantifies macro-systemic risks, and clearly presents the risk spread trend and impact. Relying on a three-level linked, full-link early warning and defense network, it achieves rapid transmission of early warning instructions, multi-domain collaborative defense response, and timely push of early warning information. Furthermore, it continuously optimizes system parameters through response data feedback. This end-to-end design, from risk assessment to early warning and defense, opens up cross-domain collaborative response channels, enhances the ability to prevent and control chain risks caused by space weather disturbances, effectively reduces the impact of disturbances on key sectors such as power, communications, satellite navigation, and finance, and ensures the stable operation of related infrastructure and systems.

[0023] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0025] Figure 1 A flowchart illustrating the workflow of an AI-based space weather disturbance feature recognition and early warning system;

[0026] Figure 2 The module architecture and data flow diagram of an AI-based space weather disturbance feature recognition and early warning system;

[0027] Figure 3 This is a flowchart of the core technology of the multi-scale disturbance identification module in an AI-based space weather disturbance feature recognition and early warning system. Detailed Implementation

[0028] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0029] Example 1:

[0030] Strong ionospheric scintillation affects the scene.

[0031] Space weather monitoring satellites have detected strong precursor signals of ionospheric scintillation in the South China region. Communication base stations, satellite navigation terminals, and financial trading systems in this region are sensitive to changes in the ionosphere. Strong scintillation may cause signal interruption, positioning errors, and transaction delays. The space weather disturbance feature identification and early warning system based on this invention is then activated to carry out full-process disturbance identification, risk assessment, and defense response.

[0032] Implementation steps, such as Figure 1 As shown:

[0033] The multi-source data integration and acquisition module acquires GPS-TEC ionospheric data through the IGS data interface, solar wind plasma density data through satellite ground receiving stations, geomagnetic component observation data and global ionospheric model GIM output data through the space weather monitoring and early warning center server, and collects regional power grid topology, communication base station distribution information, satellite navigation receiver parameters, and financial transaction system configuration data through the infrastructure operation unit management platform. This multi-source data acquisition method ensures coverage of core space weather parameters and critical infrastructure information. All data is then converted to JSON format with standardized field naming and data types, ensuring a consistent structure for data from different sources and formats, providing a standardized data foundation for subsequent processing by the dual-source data fusion module.

[0034] The dual-source data fusion module uses a 1°×1° latitude and longitude grid and UTC hourly intervals as spatiotemporal references to ensure the spatiotemporal alignment of satellite measured data and numerical model data. Linear interpolation is used to interpolate non-hourly acquired data to the hourly time, with an interpolation step size of 1 minute, filling data gaps at different sampling intervals and ensuring the continuity of the data in the temporal dimension. Kalman filtering is employed, using satellite measured data as observations and numerical model data as predictions, to correct errors caused by equipment and model biases, making the data closer to real-world space weather conditions. Based on the close match between measured data and historical values ​​over the past 72 hours, [further details are needed for accurate translation]. The reliability weight of the measured data was set at 0.7, and dual-source data fusion was completed according to the adaptive weighting rule. This ensured that the fused data retained the authenticity of the measured data while also incorporating the continuity of the numerical model data. To address the data sparsity problem in some mountainous areas of South China, adaptive Kriging interpolation was used to complete the data in sparse areas, solving the problem of insufficient data coverage. Random noise in the data was eliminated through a 30-minute sliding window smoothing technique, improving data stability. At the same time, infrastructure and dependency data such as power grid topology and communication base station distribution were normalized, allowing infrastructure data of different scales to participate in the calculation of the subsequent cross-domain cascaded risk assessment module.

[0035] The multi-scale disturbance identification module first uses an improved SMOTE-ENN algorithm to address the data imbalance problem, defining extreme event samples such as strong ionospheric scintillation as a minority class. For each minority class sample, the five most similar samples are selected in the feature space, and three new samples are generated through linear interpolation to supplement the minority class sample count and avoid bias towards the majority class data. Then, the edit nearest neighbor algorithm is used to remove overlapping noise samples from the majority class samples, reducing the interference of invalid data on subsequent identification. Subsequently, dynamic PCA and DBSCAN density clustering techniques are combined to achieve abnormal disturbance identification and false alarm rejection. The size of the PCA sliding window is adjusted according to the current solar activity intensity, and principal component analysis is performed on the magnetic field plasma density data within the window. The top three principal components with a cumulative variance contribution rate of 93% are extracted, focusing on key data features to reduce computational complexity. The residuals are then calculated. A threshold of 3 times the standard deviation was set to identify potential anomalies. DBSCAN density clustering technology was then used to cluster these anomalies, confirming the actual disturbance signals and eliminating false alarms caused by equipment malfunctions or data transmission errors. Finally, the disturbance type, intensity, and impact range were determined by comparing the data with a pre-set disturbance feature library. This library contains characteristic parameter ranges for four types of disturbances: ionospheric traveling wave disturbances, strong ionospheric scintillation, high-velocity solar wind flow, and geomagnetic storms. By calculating the matching degree between the current data and the various disturbance parameters in the feature library, the disturbance type was determined to be strong ionospheric scintillation. Combined with the comprehensive disturbance characteristic index S, the intensity level was determined to be level 4. Based on a 1°×1° latitude and longitude grid distribution, the disturbance's impact range was determined to cover the entire grid of 105°-115° E and 20°-30° N in the South China region, providing accurate disturbance information for subsequent risk assessment. Figure 3 As shown.

[0036] The cross-domain cascaded risk assessment module first uses a probabilistic model to calculate the direct risk level of a single domain. Using the level 4 strong ionospheric scintillation intensity output by the multi-scale disturbance identification module as input, a probabilistic mapping relationship is established by combining vulnerability parameters of each domain. Specifically, the power grid references the transformer's resistance to geomagnetic induced current threshold, the communication network references the ionospheric scintillation threshold, the satellite navigation system references the positioning accuracy threshold, and the financial system references the core communication link stability threshold. The disturbance intensity is then quantitatively compared with the vulnerability thresholds of each domain. In this case, the strong ionospheric scintillation intensity reached 78% of the vulnerability threshold for the communication network, 82% of the vulnerability threshold for the satellite navigation system, 25% of the vulnerability threshold for the power grid, and 65% of the vulnerability threshold for the financial system. Based on this, the communication network is classified as level 4 risk, the satellite navigation system as level 4 risk, the power grid as level 1 risk, and the financial system as level 3 risk, clearly defining the degree of direct risk faced by each domain. Subsequently, the cross-domain cascaded risk propagation index is used to simulate the cross-domain risk cascade propagation process. The formula for calculating the cross-domain cascaded risk propagation index is as follows: ,in As a systemic cascading risk index, The initial direct risk value, This is the infrastructure dependency weight matrix. This represents the infrastructure vulnerability attenuation coefficient. For propagation path weighting coefficients, The number of upstream infrastructure types, Number of downstream infrastructure types To determine the effective cascading propagation paths, starting with the high risks in communication networks and satellite navigation systems, the path of risk transmission to the financial system and power grid is reconstructed. Effective propagation paths are verified and screened using density clustering. Potential propagation paths constructed from cross-domain dependency matrices are used as clustering objects, with a neighborhood radius of 0.3 path similarity units and a minimum sample size of 3. Density clustering categorizes paths based on dependency strength and propagation logic similarity, selecting two effective propagation paths with at least 3 samples per cluster, a dependency weight of at least 0.7, and a vulnerability attenuation coefficient of at least 0.8. Paths without practical propagation significance are excluded. Finally, the macro-systemic risk index is calculated using the systemic cascading risk index formula, resulting in a macro-systemic risk index of 5.5. Based on the mapping rule (3 < systemic risk index ≤ 7 indicates medium risk), it is classified as a medium risk level. The risk level of each individual domain, the two cascading failure chains, and the systemic medium risk result are output, providing a clear risk decision-making basis for the full-link early warning and defense module.

[0037] The end-to-end early warning and defense module establishes a three-tiered network linking spacecraft, satellites, and the ground. It utilizes X-band inter-satellite links for high-speed data transmission between spacecraft and satellites, and S-band telemetry and control links to ensure command interaction between satellites and ground systems. Within the ground system, a dedicated fiber optic network transmits core early warning data, and 4G / 5G networks transmit terminal early warning information, ensuring efficient and stable data transmission between all levels of nodes. Based on TCP / IP and MQTT protocols, it implements tiered early warning command transmission and linkage triggering, triggering corresponding early warning mechanisms based on the systemic medium-risk level. Single-domain defense measures are implemented: the communication network activates anti-flicker signal modulation technology to enhance signal anti-interference capabilities; the satellite navigation system switches to dual-frequency positioning mode to reduce positioning deviations caused by flicker; the financial system activates a backup timing source to avoid satellite navigation timing anomalies affecting transactions; and the power grid strengthens monitoring system inspections to ensure smooth operation. The communication links are stable; cross-domain collaborative defense operations are carried out, coordinating the communication network to prioritize the core communication links of the financial system and the power grid, and the satellite navigation system pushes positioning deviation compensation parameters to financial institutions to achieve multi-domain resource linkage; early warning information is pushed to the operation and management personnel of the communication, satellite navigation, financial, and power sectors via SMS, and the characteristics of strong ionospheric scintillation, the risk level of each sector, the cascading failure chain, and defense suggestions are displayed in real time on the web platform to ensure that relevant personnel are promptly informed; the system continuously collects defense response data from each sector, and feeds back data such as communication link stability, positioning deviation correction effect, and transaction system operation status to the core algorithm module to dynamically optimize system parameters such as dual-source data fusion weight and multi-scale disturbance identification threshold, with the entire link response time controlled within 85 seconds, ensuring that defense measures are quickly implemented and effectively reduce the impact of strong ionospheric scintillation. Figure 2 As shown.

[0038] In summary, in scenarios where strong ionospheric scintillation affects the South China region, the system acquires and standardizes space weather and infrastructure data through a multi-source data integration and acquisition module, laying the foundation for subsequent processing. A dual-source data fusion module uses a specific spatiotemporal reference, combined with techniques such as linear interpolation and Kalman filtering, to complete data fusion and optimization, improving data quality. A multi-scale disturbance identification module uses improved SMOTE-ENN, dynamic PCA, and DBSCAN technologies to confirm and identify strong ionospheric scintillation and its parameters. A cross-domain cascaded risk assessment module calculates the risk levels and effective propagation paths across multiple domains, determining the risks within the system. A full-link early warning and defense module establishes a three-tiered network, executes defense measures, and provides feedback for optimization, responding within 85 seconds, effectively reducing the impact of strong ionospheric scintillation on communication, navigation, and other fields.

[0039] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A space weather disturbance feature recognition and early warning system based on artificial intelligence, characterized in that, The system includes: Multi-source data integration and acquisition module: collects satellite measured data, space weather numerical model data, critical infrastructure basic data and cross-domain dependency data, and performs data standardization processing; Dual-source data fusion module: It realizes the spatiotemporal registration of satellite measured data and numerical model data through linear interpolation, corrects data errors by Kalman filtering, completes dual-source data fusion by combining adaptive weighting rules, completes sparse region data by adaptive Kriging interpolation, eliminates data noise by sliding window smoothing technology, and normalizes infrastructure and dependency data. Multi-scale disturbance identification module: Based on standardized data, features are extracted and disturbances are identified. An improved SMOTE-ENN algorithm is used to handle data imbalance. Dynamic PCA and DBSCAN density clustering technology are combined to identify abnormal disturbances and eliminate false alarms. The disturbance type, intensity and range of influence are determined by comparing with a preset disturbance feature library. Cross-domain cascading risk assessment module: It uses a probability model to calculate the direct risk level of a single domain, simulates the cross-domain risk cascading propagation process through the cross-domain cascading risk propagation index, combines density clustering to verify and screen effective propagation paths, calculates the macro-systemic risk index, and outputs the risk level of a single domain, the cascading failure chain, and the systemic risk results. The end-to-end early warning and defense module establishes a three-level linkage network, realizes hierarchical early warning command transmission and linkage triggering based on TCP / IP and MQTT protocols, executes single-domain defense measures and cross-domain collaborative defense operations, pushes early warning information through SMS and Web platforms, and collects response data to optimize system parameters.

2. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, The multi-source data integration and acquisition module acquires GPS-TEC ionospheric data, solar wind plasma density data, geomagnetic component observation data, global ionospheric model GIM output data, power grid topology, communication base station distribution information, satellite navigation receiver parameters, and financial transaction system configuration data through the IGS data interface, satellite ground receiving station, space weather monitoring and early warning center server, and infrastructure operation unit management platform, respectively. All data is converted into JSON format and field naming and data type are standardized.

3. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, The dual-source data fusion module uses a 1-degree × 1-degree latitude and longitude grid and UTC hourly time as the spatiotemporal reference. It interpolates non-hourly data to the hourly time using linear interpolation with an interpolation step of 1 minute. It employs Kalman filtering technology to correct errors by using satellite measured data as observations and numerical model data as predictions. The reliability weight of the measured data is determined based on the degree of agreement between the measured data in the past 72 hours and the historical true values, with a value range of 0.3 to 0.

8. It uses adaptive Kriging interpolation to complete the sparse region data and eliminates data noise using a 30-minute sliding window smoothing technique.

4. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, When the multi-scale perturbation identification module uses the improved SMOTE-ENN algorithm to handle the data imbalance problem, it defines extreme event samples as the minority class. For each minority class sample, it selects 5 most similar samples in the feature space, generates 3 to 5 new samples through linear interpolation, and then uses the edit nearest neighbor algorithm to remove overlapping noise samples from the majority class samples.

5. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, The multi-scale disturbance identification module combines dynamic PCA and DBSCAN density clustering technology to identify abnormal disturbances and eliminate false alarms. It dynamically adjusts the size of the PCA sliding window according to the intensity of solar activity, performs principal component analysis on the magnetic field plasma density data within the window, extracts the top three principal components with a cumulative variance contribution rate of not less than 90%, identifies potential abnormal data by calculating residuals and setting a threshold of 3 times the standard deviation, and then uses DBSCAN density clustering technology to confirm the real disturbance.

6. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, The multi-scale disturbance identification module determines the disturbance type, intensity, and impact range by comparing it with a preset disturbance feature library. The feature library contains feature parameter ranges for four types of disturbances: ionospheric traveling wave disturbances, strong ionospheric scintillation, high-speed solar wind flow, and geomagnetic storms. The disturbance type is determined by calculating the matching degree, the intensity level is determined by combining the comprehensive disturbance feature index S, and the impact range is determined according to the 1-degree × 1-degree latitude and longitude grid distribution.

7. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, When the cross-domain cascaded risk assessment module calculates the direct risk level of a single domain using a probabilistic model, it takes the disturbance intensity output by the multi-scale disturbance identification module as input and establishes a probabilistic mapping relationship with the vulnerability parameters of the power grid, communication network, satellite navigation system, and financial system. The vulnerability parameter of the power grid is the transformer's resistance to geomagnetic induced current threshold, the communication network's resistance to ionospheric scintillation threshold, the satellite navigation system's positioning accuracy threshold, and the financial system's core communication link stability threshold. The disturbance intensity is quantitatively compared with the vulnerability thresholds of each domain. When the disturbance intensity is ≤30% of the vulnerability threshold, it is judged as Level 1 risk; 31%-50% is judged as Level 2 risk; 51%-70% is judged as Level 3 risk; 71%-90% is judged as Level 4 risk; and >90% is judged as Level 5 risk. The direct risk level of each single domain is calculated through this probabilistic mapping relationship.

8. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, In the cross-domain cascading risk assessment module, the formula for calculating the cross-domain cascading risk propagation index is as follows: ,in As a systemic cascading risk index, The initial direct risk value, This is the infrastructure dependency weight matrix. This represents the infrastructure vulnerability attenuation coefficient. For propagation path weighting coefficients, The number of upstream infrastructure types, Number of downstream infrastructure types This represents the number of effective cascading propagation paths.

9. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, When the cross-domain cascaded risk assessment module combines density clustering to verify and screen effective propagation paths, it uses the potential propagation paths constructed from the cross-domain dependency matrix as clustering objects, sets the neighborhood radius to 0.3 path similarity units, and the minimum number of samples to 3. Through density clustering, the paths are classified according to the strength of dependency and the similarity of propagation logic. Paths with no less than 3 samples in the cluster, a dependency weight of no less than 0.7, and a vulnerability decay coefficient of no less than 0.8 are selected as effective propagation paths. The macro-systemic risk index is calculated using the formula in claim 8 and further mapped to three levels of risk: low, medium, and high. When the systemic risk index ≤ 3, it is low risk; when 3 < systemic risk index ≤ 7, it is medium risk; and when the systemic risk index > 7, it is high risk.

10. The space weather disturbance feature identification and early warning system based on artificial intelligence according to claim 1, characterized in that, The multi-domain linkage network of the full-link early warning and defense module includes a three-level architecture of spacecraft-satellite-ground, using X-band inter-satellite links and S-band telemetry and control links for communication, and using fiber optic private networks and 4G / 5G networks to transmit data within the ground system; the graded early warning is triggered based on the systemic risk index, and the full-link response time is controlled within 90 seconds.

Images