Link channel abnormal state prediction method
By constructing an active early warning model based on multi-dimensional real-time data analysis, the problems of passive response and coarse prediction granularity in abnormal channel states in wireless networks are solved, enabling refined judgment and dynamic adaptation of anomalies, and improving the accuracy and practicality of network operation and maintenance.
Patent Information
- Application Number
- CN202511862153.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-03-13
AI Technical Summary
The prediction of abnormal channel states in existing wireless networks suffers from problems such as passive response, single analysis method, poor adaptability, insufficient utilization of multidimensional data, weak model generalization and adaptive ability, and coarse prediction granularity. These issues result in insufficient prediction accuracy and practicality, making it difficult to meet the high reliability requirements of complex heterogeneous networks.
A two-dimensional curve model is constructed using multiple abnormal state indicators to monitor the changes in the danger values of abnormal state indicators in real time. Similarity is calculated by Pearson correlation coefficient or cosine similarity to generate risk levels and issue alarms. Combined with online learning and parameter tuning, an active early warning model is constructed to achieve refined judgment on the type, severity, duration and scope of impact of anomalies.
It improves the accuracy of predictions, reduces false alarms and false negatives, enables early warning, adapts to dynamic changes in the network environment, eliminates the need for frequent manual intervention, and significantly enhances the foresight and accuracy of network operation and maintenance.
Smart Images
Figure CN121664335A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of link channel fault monitoring technology, and more specifically to a method for predicting abnormal link channel conditions. Background Technology
[0002] With the expansion of wireless network scale and the large-scale application of heterogeneous networking, the frequency of abnormal channel conditions in the network is increasing. These anomalies are often sudden and unpredictable, easily leading to degraded communication quality, data transmission interruptions, and impaired user experience, thus causing varying degrees of economic and efficiency losses for users. Therefore, predicting abnormal channel conditions during network operation has become a key challenge for intelligent operation and maintenance.
[0003] Currently, the prediction of abnormal wireless channel conditions has the following limitations: 1) Reactive response rather than proactive early warning. Most existing systems rely on detection and alarm mechanisms after anomalies occur, lacking the ability to predict potential risks in advance. This "post-event remediation" model cannot meet the stringent continuity requirements of high-reliability business operations and cannot effectively prevent losses from occurring; 2) The analysis methods are simplistic and lack adaptability. Traditional methods often rely on simple threshold judgments or analyze data from a single source. In complex heterogeneous network environments, channel state fluctuations are influenced by multiple coupled factors. Such methods struggle to adapt to the dynamic fluctuations of network traffic in different scenarios and time periods, resulting in high false alarm and false negative rates, and making it difficult to accurately capture true anomalies. 3) Insufficient utilization of multidimensional and massive operational data. As network scale expands, the resulting situational data is characterized by large volume, high dimensionality, and strong dynamism. Existing technologies lack efficient multidimensional feature extraction and fusion analysis capabilities, failing to fully exploit the temporal patterns and correlations hidden in historical and real-time data, resulting in limited accuracy of prediction models; 4) Weak model generalization and adaptation capabilities. Many machine learning-based solutions rely on static historical datasets in specific scenarios for training, requiring high data stability and depending on the quality of feature engineering. When network topology, user behavior, and business types change, the model is prone to failure, requiring extensive manual intervention for retraining. It is difficult to capture sudden anomalies and achieve continuous self-optimization and generalization across scenarios and networks.
[0004] 5) The prediction granularity is coarse and its practicality is low. Some prediction schemes can only determine "whether it is abnormal", but cannot make more refined inferences about the type of abnormality, severity level, potential impact range and possible duration, making it difficult for operation and maintenance personnel to formulate accurate and differentiated handling strategies.
[0005] These shortcomings collectively result in insufficient accuracy, timeliness, and practicality of existing channel anomaly prediction technologies, hindering the improvement of network operation and maintenance intelligence. Summary of the Invention
[0006] In view of the above problems, the present invention is proposed to provide a link channel anomaly state prediction method to overcome or at least partially solve the above problems, aiming to meet the need for high-precision and proactive prediction of channel anomaly states in complex heterogeneous networks.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, embodiments of the present invention provide a method for predicting abnormal states of a link channel, comprising the following steps: An abnormal state model of the target fault is established, which is a two-dimensional curve model constructed based on multiple abnormal state indicators and corresponding hazard values associated with the current target fault. The abnormal state indicators are monitored in real time, and when the danger value changes, the similarity between the current state model and the abnormal state model is calculated. If the similarity reaches a preset threshold, a risk level is generated and an alarm is issued.
[0008] Furthermore, the multiple abnormal state indicators include one or more of the following: latency, packet loss rate, signal-to-noise ratio, signal strength, bit error rate, handover success rate, jitter, and interference.
[0009] Furthermore, the danger value of the abnormal state indicator is obtained by querying a preset danger value mapping table, which is used to define the mapping relationship between the state indicator parameter values and danger values corresponding to different target faults under different links.
[0010] Furthermore, the maximum danger value corresponding to each state indicator parameter value in the danger value mapping table is set according to the degree of correlation between the current indicator and the target fault.
[0011] Furthermore, the similarity between the current state model and the abnormal state model is calculated using the Pearson correlation coefficient; when all abnormal state indicators have the same risk value, cosine similarity is used for calculation.
[0012] Furthermore, when monitoring the abnormal state indicators in real time, if the parameter value of a certain abnormal state indicator cannot be obtained, the danger value of the current abnormal indicator is filled with zero, and then similarity calculation is performed.
[0013] Furthermore, the risk level is determined based on the similarity and the highest risk value among all abnormal state indicators monitored in real time.
[0014] Furthermore, it also includes correcting the abnormal state model, the steps of which include: After the target fault occurs, the abnormal state model is revised based on the difference between the actual danger value of each abnormal state indicator in this fault and the corresponding danger value in the abnormal state model and the preset model revision table.
[0015] Furthermore, if the target fault contains an abnormal state indicator where the dangerous value is padded to zero, the revision of the current abnormal state indicator is skipped when revising the abnormal state model.
[0016] The link channel abnormality prediction method disclosed in this invention can not only predict whether an anomaly will occur, but also focuses on making refined judgments and graded early warnings on the type, severity, possible duration, and scope of impact of the anomaly. Compared with the prior art, the beneficial effects include at least the following: 1. This paper proposes to comprehensively utilize heterogeneous data such as signal strength, bit error rate, traffic characteristics, and handover success rate to conduct in-depth fusion and correlation analysis, maximize the value of the data, fully reflect the consistent characteristics of the anomaly, thereby significantly improving the accuracy of prediction and reducing false alarms and missed alarms. 2. Construct a proactive early warning model based on multi-dimensional real-time data analysis, transforming post-event remediation into pre-event early warning, significantly improving the foresight of network operation and maintenance, and fundamentally reducing the risk of service interruption and economic losses caused by channel anomalies. 3. The abnormal state model of this application can learn and optimize parameters online in real time based on feedback data and new network deployment scenarios, without the need for frequent manual intervention and retraining. It can adapt to the dynamic changes of heterogeneous network environment and ensure the continuous effectiveness of prediction. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0018] Figure 1 This is a flowchart of the link channel abnormal state prediction method of the present invention; Figure 2 A schematic diagram of the abnormal state model provided by the present invention; Figure 3 This is a trend modeling diagram provided in an embodiment of the present invention; Figure 4 This is a trend analysis chart provided in an embodiment of the present invention; Figure 5 The model correction diagram provided in the embodiments of the present invention; Figure 6This is a schematic diagram illustrating the intelligent analysis of the results based on channel state prediction in this invention. Detailed Implementation
[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] The development of wireless communication technology has always been accompanied by profound changes in network architecture. To meet the explosive growth in capacity and coverage demands, wireless networks have evolved into complex heterogeneous networks containing various elements. At the same time, the introduction of new technologies such as massive MIMO and millimeter-wave communication has not only improved system performance but also greatly increased the complexity and dynamism of wireless channel components.
[0021] Against this backdrop, accurate perception and intelligent management of network channel status have become crucial for ensuring user experience and network performance. Channel status information is no longer merely the basis for link adaptation or resource scheduling, but also key situational data for assessing overall network health and predicting potential risks. Therefore, analyzing massive amounts of network operational status data to achieve intelligent prediction of network anomalies is an inevitable technological path towards highly autonomous networks and zero-fault operation and maintenance.
[0022] This application addresses the problems of passive response, poor adaptability, limited accuracy, and weak real-time performance in existing technologies by providing a link channel anomaly prediction method. First, it proposes to comprehensively utilize heterogeneous data such as signal strength, bit error rate, traffic characteristics, and handover success rate for deep fusion and correlation analysis to uncover consistent features that better reflect the essence of the anomaly, thereby significantly improving prediction accuracy and reducing false alarms and missed alarms. Furthermore, it designs an efficient feature extraction and selection algorithm to automatically and efficiently extract key temporal features and patterns strongly correlated with channel anomalies from massive, high-dimensional, and dynamic network operation status data, maximizing data value. Finally, it constructs a proactive early warning model based on multi-dimensional real-time data analysis, transforming post-event remediation into pre-event warning, significantly improving the foresight of network operation and maintenance, and fundamentally reducing the risk of service interruption and economic losses caused by channel anomalies. Simultaneously, this model can learn and optimize parameters online in real-time based on data and new network deployment scenarios, adapting to dynamic changes in heterogeneous network environments without frequent manual intervention and retraining, ensuring the continuous effectiveness of predictions.
[0023] Furthermore, this application not only predicts whether an anomaly will occur, but also strives to make refined judgments and graded early warnings on the type, severity, possible duration and scope of impact of the anomaly, providing hierarchical and operable decision-making basis for network automated operation and maintenance systems, thereby assisting operation and maintenance personnel in achieving precise intervention.
[0024] In one specific embodiment, the link channel abnormal state prediction method provided by the present invention, such as... Figure 1 The steps include: S1. Establish an abnormal state model of the target fault, wherein the abnormal state model is a two-dimensional curve model constructed based on multiple abnormal state indicators and corresponding hazard values associated with the current target fault. S2. Monitor the abnormal state indicators in real time, and calculate the similarity between the current state model and the abnormal state model when the danger value changes. S3. If the similarity reaches a preset threshold, a risk level is generated and an alarm is issued.
[0025] Because the target faults belong to different objects, the abnormal state models also differ. Therefore, this application constructs a corresponding abnormal state model for each link. Considering that even devices of the same type may have individual differences, to improve the accuracy of prediction, even two links of the same type will have their own exclusive abnormal state models.
[0026] The following is a detailed explanation of the execution process for each step.
[0027] In an optional embodiment, in step S1, the abnormal state model is a two-dimensional curve model constructed based on multiple abnormal state indicators and corresponding hazard values associated with the current target fault; such as Figure 2 As shown; The abnormal status indicators include one or more of the following: latency, packet loss rate, signal-to-noise ratio, signal strength, bit error rate, handover success rate, jitter, and interference. The more abnormal status indicators involved in the calculation, the more accurate the prediction result. The danger value of each abnormal status indicator is obtained by querying a preset danger value mapping table. This danger value mapping table defines the mapping relationship between the status indicator parameter values and danger values corresponding to different target faults under different links, and needs to be preset according to the different characteristics of different links. For example, when the latency of an optical fiber link is ≥100ms, the danger value is 95; when the latency of a satellite link is ≥3s, the danger value is 95.
[0028] Preferably, the hazard value mapping table is preset inside the server. For faults where no target fault exists, a sample model is built based on the current fault as the basis for alarms of subsequent faults with the same characteristics.
[0029] Furthermore, the maximum danger value corresponding to each status indicator parameter value in the danger value mapping table is set according to the degree of correlation between the current indicator and the target fault. For example, the signal-to-noise ratio and signal strength are more closely related to the channel link disconnection fault, and their maximum danger value can reach 95, while the handover success rate is less closely related to the channel link disconnection fault, and its maximum danger value is only 80.
[0030] In an optional embodiment, in step S2, when the danger value changes, the similarity between the current state model and the abnormal state model is calculated. In this embodiment, the Pearson correlation coefficient is first used to calculate the similarity between the current state model and the abnormal state model. When all abnormal state indicators have the same danger value and the Pearson correlation coefficient cannot be used for calculation, the cosine similarity is used for calculation.
[0031] Furthermore, when monitoring the abnormal state indicators in real time, if the parameter value of a certain abnormal state indicator cannot be obtained, the danger value of the current abnormal indicator is filled with zero, and then similarity calculation is performed.
[0032] In an optional embodiment, in step S3, if the similarity reaches a preset threshold, a risk level is generated and an alarm is issued. In this embodiment, the risk level is determined based on a combination of the similarity and the highest danger value among all abnormal state indicators monitored in real time.
[0033] In some implementation schemes, refer to Table 1; Table 1
[0034] If the analysis result is ≥0.6, a risk is considered to exist. The risk level is further determined by combining the highest hazard value among all current abnormal state indicators; that is, if the highest hazard value is ≥60, the risk is classified as level 4. In this embodiment, the risk is divided into four levels, with level 1 being the most severe. To further optimize the above technical solution, this application also includes correcting the abnormal state model, the steps of which include: After the target failure occurs, the abnormal state model is revised based on the difference between the actual danger value of each abnormal state indicator in this failure and the corresponding danger value in the abnormal state model, and a preset model revision table. In some implementation schemes, the revision table is shown in Table 2; Table 2
[0035] In this embodiment, if there is an abnormal state indicator in the target fault where the dangerous value is padded to zero, the revision of the current abnormal state indicator is skipped when revising the abnormal state model.
[0036] As the service runs, the danger values of each abnormal state indicator in the abnormal state model will be continuously adjusted, thereby improving the accuracy of prediction.
[0037] In one example, the analysis and prediction steps for a channel link disconnection state include: Trend modeling, reference Figure 3 Based on the danger values of all abnormal state indicators before the fault occurs, a two-dimensional curve model is constructed according to the target fault object, the associated network elements of each event, and the danger values of the indicators after the event information is converted. Trend analysis, when the danger value of a certain abnormal state indicator in the abnormal state model changes, such as... Figure 4 The danger range for each abnormal state indicator is [0, 100]. If a specific abnormal state indicator cannot be obtained, it is automatically filled with zeros. Further analysis is performed using the Pearson correlation coefficient (if all related items have the same danger value, cosine similarity is used for analysis); and risk warnings are issued based on the Fenix results. Model correction involves revising various abnormal state indicators in the abnormal state model after a fault occurs. If any abnormal state indicator for the new fault contains zero-padding, then those zero-padding indicators are not revised. In this example, the revision result is as follows: Figure 5 .
[0038] The abnormal state prediction method provided in this application has advantages such as high accuracy, high real-time performance, and low resource consumption. Furthermore, it can serve as a reference for various intelligent analyses, including fault correlation analysis, link resource allocation rationality analysis, and service transmission trend prediction analysis. It provides analytical basis for multiple intelligent analysis functions in operation and maintenance management services. For details, please refer to... Figure 6 ,include: 1. After channel state prediction analysis, hidden operational status data related to channel link failures can be found.
[0039] 2. After channel state prediction analysis, based on the network resource allocation among nodes and the service data transmission requirements, it can be determined whether the network resource allocation meets the environmental needs, and reasonable suggestions can be given.
[0040] 3. After channel state prediction analysis, the transmission path information between each node can be output. Combined with the service transmission information, link topology correlation, etc., the transmission trend of the service can be analyzed.
[0041] It should be noted that the implementation of the method of the present invention may rely on a computer program, which may run on computer devices such as smartphones, tablets, and personal computers or on servers. This embodiment does not limit the subject on which the method is run.
[0042] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0043] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting abnormal states of a link channel, characterized in that, include: An abnormal state model of the target fault is established, which is a two-dimensional curve model constructed based on multiple abnormal state indicators and corresponding hazard values associated with the current target fault. The abnormal state indicators are monitored in real time, and when the danger value changes, the similarity between the current state model and the abnormal state model is calculated. If the similarity reaches a preset threshold, a risk level is generated and an alarm is issued.
2. The link channel abnormal state prediction method according to claim 1, characterized in that, The multiple abnormal status indicators include one or more of the following: latency, packet loss rate, signal-to-noise ratio, signal strength, bit error rate, handover success rate, jitter, and interference.
3. The link channel abnormal state prediction method according to claim 1, characterized in that, The danger value of the abnormal state indicator is obtained by querying a preset danger value mapping table. The danger value mapping table is used to define the mapping relationship between the state indicator parameter values and danger values corresponding to different target faults under different links.
4. The link channel abnormal state prediction method according to claim 3, characterized in that, The maximum danger value corresponding to each state indicator parameter value in the danger value mapping table is set according to the degree of correlation between the current indicator and the target fault.
5. The link channel abnormal state prediction method according to claim 1, characterized in that, The similarity between the current state model and the abnormal state model is calculated using the Pearson correlation coefficient; when all abnormal state indicators have the same risk value, the cosine similarity is used for calculation.
6. The link channel abnormal state prediction method according to claim 1, characterized in that, When monitoring the abnormal status indicators in real time, if the parameter value of a certain abnormal status indicator cannot be obtained, the danger value of the current abnormal indicator is filled with zero, and then similarity calculation is performed.
7. The link channel abnormal state prediction method according to claim 1, characterized in that, The risk level is determined based on the similarity and the highest risk value among all abnormal state indicators monitored in real time.
8. The link channel abnormal state prediction method according to claim 1, characterized in that, It also includes correcting the abnormal state model, the steps of which include: After the target fault occurs, the abnormal state model is revised based on the difference between the actual danger value of each abnormal state indicator in this fault and the corresponding danger value in the abnormal state model and the preset model revision table.
9. The link channel abnormal state prediction method according to claim 6, characterized in that, It also includes correcting the abnormal state model, the steps of which include: After the target fault occurs, the abnormal state model is revised based on the difference between the actual danger value of each abnormal state indicator in this fault and the corresponding danger value in the abnormal state model and the preset model revision table. If there are abnormal state indicators in the target fault whose danger values are filled to zero, the revision of the current abnormal state indicator is skipped when revising the abnormal state model.