A method, system, electronic device, and storage medium for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure.

By using a multi-dimensional time-series alarm hierarchical aggregation method, dynamically adjusting thresholds, clustering noise reduction, and differential push, the problems of high false alarm rate and redundant alarms in traditional alarm monitoring are solved, thereby improving fault location efficiency and notification accuracy.

CN123123121APending Publication Date: 2026-11-17SHENZHEN PUHUI ZHIWEI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611619955.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional alarm monitoring methods in digital infrastructure suffer from problems such as high false alarm rate, numerous redundant alarms, difficulty in fault location, and congestion of notification channels. In particular, fixed thresholds are difficult to adapt to dynamic operating conditions such as equipment aging, seasonal changes, and load fluctuations, and they ignore the spatiotemporal correlation between multiple measurement points.

Method used

A multi-dimensional temporal alarm hierarchical aggregation method is adopted. By accessing remote sensing measurement point data in real time, multi-dimensional temporal semantic outlier features are extracted, thresholds are dynamically adjusted, clustering noise reduction and alarm cluster merging are performed, and a level score is calculated based on static criticality and representative outlier score. Event fingerprints are generated for differential push.

Benefits of technology

It significantly reduced false alarms and redundant alarms, optimized fault location efficiency and notification accuracy, and reduced the workload and resource waste of operation and maintenance personnel.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123121A_ABST
    Figure CN123123121A_ABST
Patent Text Reader

Abstract

This invention discloses a method, system, electronic device, and storage medium for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure, relating to the field of intelligent operation and maintenance technology. The method includes: S10, real-time access to time-series data from various remote signaling points in the digital infrastructure; S20, extraction of multi-dimensional time-series semantic outlier features; S30, dynamic adjustment of thresholds and calculation of multi-dimensional time-series outlier scores; when the outlier score reaches the threshold, the remote signaling point is marked as a time-series alarm point; S40, clustering and denoising the time-series alarm points, merging them into alarm clusters; S50, aggregating multi-dimensional time-series alarm events and calculating the event's hierarchical score; S60, routing the multi-dimensional time-series alarm events to a notification channel; S70, generating a fingerprint of the current event and comparing it with the fingerprint of the last successfully pushed event; only if a difference exists is the event pushed through the notification channel. The beneficial effects of this invention are: significantly reduced false alarm rate and redundant alarms, and optimized fault location efficiency and notification accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology, and more specifically, to a method, system, electronic device, and storage medium for hierarchical aggregation of multi-dimensional time-series alarms for digital infrastructure. Background Technology

[0002] In the operation and maintenance management of digital infrastructure (such as data centers, power distribution rooms, and communication base stations), real-time monitoring of time-series data from various remote sensing measurement points (such as temperature, humidity, voltage, current, and switch status) is crucial for ensuring stable system operation. Traditional alarm monitoring methods typically employ fixed threshold judgments: an alarm is immediately generated when the value of a measurement point exceeds a preset upper or lower limit. However, this fixed threshold approach has many shortcomings in practical applications.

[0003] First, fixed thresholds are ill-suited to dynamic operating conditions such as equipment aging, seasonal changes, and load fluctuations. They are prone to generating numerous false alarms due to momentary fluctuations or normal fluctuations, leading to "alarm fatigue" among maintenance personnel and causing them to overlook truly critical faults. Second, traditional methods often process alarms from each measurement point in isolation, ignoring the spatiotemporal correlations between multiple measurement points within the same physical unit or logical loop. This results in multiple alarms triggered by the same fault source being presented in a scattered manner, increasing the difficulty of fault localization. Furthermore, for the same alarm event, traditional systems repeatedly push notifications with each state change, causing congestion in the notification channel and wasting resources. Maintenance personnel also find it difficult to quickly identify the evolution trend of alarms.

[0004] To this end, the present invention provides a method, system, electronic device and storage medium for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure, which can significantly reduce false alarm rate and redundant alarms, and optimize fault location efficiency and notification accuracy. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a method, system, electronic device, and storage medium for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure, which can significantly reduce false alarm rate and redundant alarms, and optimize fault location efficiency and notification accuracy.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure, wherein the improvement is that the method for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure includes the following steps: S10, real-time access to time-series data from various remote sensing measurement points in the digital infrastructure; S20, extract multidimensional temporal semantic outlier features for each remote sensing measurement point at each sampling time; wherein, the multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency, and anomaly duration; S30, dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and calculate its multidimensional time-series outlier based on displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the gate opening threshold, mark the remote signaling point as a time-series alarm point. S40, perform clustering and noise reduction on all marked time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster; S50 aggregates alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculates the event's level score based on the static criticality and representative outlier of the alarm cluster. S60, based on the grade score, routes multi-dimensional time-series alarm events to the corresponding notification channels; S70, generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only if there is a difference will the event be pushed through the corresponding notification channel.

[0007] Furthermore, in step S30, the specific method for dynamically adjusting the threshold based on the historical smoothed values ​​of each remote sensing point within its sliding window and calculating its multidimensional time-series outliers is as follows: For each time-series semantic outlier feature of each remote sensing measurement point, a sliding window is maintained, and the mean of the smoothed values ​​within the window is calculated as the local normal baseline. Calculate the absolute deviation of the current smoothed value from the baseline, and combine it with a preset deviation scale to obtain the deviation ratio of each feature dimension; The multidimensional temporal outlier is obtained by weighted summation of the deviation ratios across multiple feature dimensions, and its calculation expression is as follows: In the formula, Represents the multidimensional temporal outlier of measurement point i; Indicates the deviation ratio of the displacement strength dimension; This indicates the deviation ratio of the displacement frequency dimension; Indicates the deviation ratio of the abnormal persistence dimension; This represents the weight of the displacement intensity dimension in group separation; This indicates the weight of the displacement frequency dimension in the group splitting; This represents the weight of the anomaly persistence dimension in the group dissociation; and .

[0008] Furthermore, the deviation from the scale needs to be slowly corrected based on feedback from false alarms or missed alarms in manual annotation. The correction formula is as follows: In the formula, This indicates the deviation of dimension d from the scale. Indicates the deviation from the scale correction step size; This represents the scale residual of dimension d.

[0009] Furthermore, in step S40, the specific method for clustering and denoising all marked time-series alarm points is as follows: Calculate the feature vector distance between any two alarm points The feature vector distance is determined by the normalized time difference, whether the loops are the same, whether the measurement point types are of the same type, and the outlier difference. The feature vector distance is converted into similarity using the following formula: and similarity Reaching the preset lower limit of clustering The alarm points are merged into one alarm cluster.

[0010] Furthermore, in step S50, the formula for calculating the grade score is: In the formula, This indicates the level score of alarm cluster c; This indicates the weight of static keyness in the ranking score; This indicates the static criticality of alarm cluster c; The alarm cluster c represents the outlier, and the data within the cluster is taken. The maximum value.

[0011] An improved multi-dimensional time-series alarm hierarchical aggregation system for digital infrastructure includes: The data access module is used to access time-series data from various remote sensing measurement points in the digital infrastructure in real time; The feature extraction module is used to extract multidimensional temporal semantic outlier features of each remote sensing measurement point at each sampling time; the multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency, and anomaly duration; The adaptive threshold module is used to dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and to calculate its multidimensional time-series outlier based on displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the gate opening threshold, the remote signaling point is marked as a time-series alarm point. The clustering and noise reduction module is used to cluster and reduce noise for all labeled time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster. The hierarchical aggregation module is used to aggregate alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculate the level score of the event based on the static criticality and representative outlier of the alarm cluster. The hierarchical routing module is used to route multi-dimensional time-series alarm events to the corresponding notification channels based on hierarchical scores. The difference push module is used to generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only when there is a difference will the event be pushed through the corresponding notification channel.

[0012] An electronic device, improved in that it includes at least one processor and at least one memory, wherein, The memory stores computer-readable instructions; The computer-readable instructions are executed by one or more processors, enabling the electronic device to implement the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation method as described above.

[0013] An improvement of a storage medium having computer-readable instructions stored thereon, wherein the computer-readable instructions are executed by one or more processors to implement the multi-dimensional time-series alarm hierarchical aggregation method for digital infrastructure as described above.

[0014] The beneficial effects of this invention are as follows: This invention dynamically adjusts the threshold based on the historical smooth value of each remote sensing measurement point within its sliding window and calculates its multidimensional temporal outliers. When the outlier reaches the opening threshold, the remote sensing measurement point is marked as a temporal alarm point, solving the problem of false alarms caused by the inability of fixed thresholds to adapt to dynamic operating conditions. By clustering and denoising all marked temporal alarm points, alarm points meeting preset similarity conditions are merged into an alarm cluster, and alarm clusters belonging to the same physical unit or logical loop are aggregated into a single multidimensional temporal alarm event, solving the problem of scattered alarm presentation and difficulty in fault location caused by isolated measurement points. By generating a fingerprint of the current event and comparing it with the fingerprint of the last successfully pushed event, push is only made through the corresponding notification channel when a difference exists, solving the problem of notification channel congestion and resource waste caused by repeated pushes of the same alarm event. Therefore, this invention can significantly reduce false alarm rate and redundant alarms, and optimize fault location efficiency and notification accuracy. Attached Figure Description

[0015] Figure 1 This is an overall flowchart of a multi-dimensional time-series alarm hierarchical aggregation method for digital infrastructure according to the present invention; Figure 2 This is a block diagram of a multi-dimensional time-series alarm hierarchical aggregation system for digital infrastructure according to the present invention; Figure 3 A hardware structure diagram of an electronic device is shown as an exemplary embodiment; Figure 4 A block diagram illustrating an electronic device as an exemplary embodiment. Detailed Implementation

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] The following will clearly and completely describe the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention. Furthermore, all connections / linkages involved in the patent do not simply refer to direct contact between components, but rather to the ability to form a better connection structure by adding or reducing connecting accessories according to specific implementation conditions. The various technical features in this invention can be combined interactively without contradicting each other.

[0018] Reference Figure 1 As shown, this invention discloses a method for hierarchical aggregation of multi-dimensional time-series alarms for digital infrastructure. The method includes the following steps: S10, real-time access to time-series data from various remote sensing measurement points in the digital infrastructure; It should be noted that in this embodiment, the digital infrastructure includes sites requiring continuous power and network operation, such as the park's power distribution room, data center power modules, and network rooms. The measurement points are sourced from the power Internet of Things (IoT). Remote signaling points are equipment status-related remote control signals, with values ​​primarily indicating on / off status, fault / reset status, and online / offline status. The system acquires real-time time-series data from each remote signaling point through the data access module of the edge node at a fixed sampling period (e.g., one frame per second). Accompanying telemetry data (e.g., current, voltage, communication interruption counts) is also accessed, but only for feature supplementation in subsequent clustering distance calculations; no separate alarm engine is used. Missing data uses the value from the previous frame to avoid all data reporting immediately upon online connection. The system performs a data acquisition operation at a fixed period (e.g., every 1 second) to ensure data timeliness.

[0019] S20, extract multidimensional temporal semantic outlier features of each remote sensing measurement point at each sampling time; the multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency and anomaly duration.

[0020] It should be noted that, in this embodiment, the multidimensional temporal semantic outlier features include at least three dimensions: displacement intensity, displacement frequency, and anomaly persistence. The original feature extraction methods for each dimension are as follows: Displacement intensity dimension: The separation / merging state of remote signaling is encoded as 0 / 1, and the absolute value of the difference between the current beat value and the previous beat value is used as the original displacement intensity.

[0021] Displacement frequency dimension: Count the number of displacements of the measuring point within the sliding window at the current sampling time, divide by the window length W, and obtain the normalized displacement frequency.

[0022] Abnormal persistence dimension: If the deviation of the current frame from the window baseline persists, the persistence frame count is incremented by one; otherwise, it is cleared to zero and then divided by W to be truncated to 1.

[0023] The original features of the above three dimensions are subjected to exponential smoothing, and the smoothing formula is as follows: In the formula, This represents the smoothed value of measurement point i at time t in dimension d; This represents the smoothed value of measurement point i at the previous time (t-1) in dimension d, i.e., the smoothed result of the previous time step. This represents the original feature of measurement point i at time t in dimension d; This represents the exponential smoothing coefficient, with a value ranging from 0.6 to 0.8. The larger the value, the stronger the smoothing effect, and the less likely it is to increase the intensity dimension of a single shot with burrs. When there is no historical data for a new measurement point. Use the current one directly Initially, the first W frames only accumulate windows and do not judge outliers to avoid all reports immediately upon going online. The purpose of smoothing is to remove single-frame spikes from the intensity dimension, but the frequency dimension can still reflect the jitter situation. In the specific implementation of this invention, this embodiment extracts semantic features from three dimensions: displacement intensity, displacement frequency, and anomaly duration. This upgrades the simple judgment of single-point exceeding the limit to a multi-dimensional temporal semantic description, providing rich discrimination criteria for subsequent adaptive thresholding and clustering aggregation. It fundamentally solves the problem of weak correlation between points in traditional methods, where points are only hard-bound by names or rules.

[0024] S30, dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and calculate its multidimensional time-series outlier based on displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the gate opening threshold, mark the remote signaling point as a time-series alarm point. It should be noted that in this embodiment, a sliding window is maintained for each temporal semantic outlier feature of each remote sensing point. The window length W is 20-60 frames; for example, 30 frames, with a frame interval of 1 second or 2 seconds. After the window has been open for W frames, the mean of the smoothed values ​​within the window is calculated as the local normal baseline, using the following formula: In the formula, Let d be the window mean of measurement point i in dimension d; This represents the smoothing value for the k-th frame within the window.

[0025] Calculate the absolute deviation of the current smoothed value from the baseline: ; Combined with preset dimensions Deviation from scale The deviation ratio for each feature dimension is obtained, and its calculation formula is as follows: In the formula, To prevent division by small amounts; The multidimensional temporal outlier is obtained by weighted summation of the deviation ratios of the three feature dimensions, and its calculation expression is as follows: In the formula, Represents the multidimensional temporal outlier of measurement point i; Indicates the deviation ratio of the displacement strength dimension; This indicates the deviation ratio of the displacement frequency dimension; Indicates the deviation ratio of the abnormal persistence dimension; This represents the weight of the displacement intensity dimension in group separation; This indicates the weight of the displacement frequency dimension in the group splitting; This represents the weight of the anomaly persistence dimension in the group dissociation; and The weights can be adjusted according to the actual scenario: if you want to focus more on intensity changes, you can increase the weight. To suppress shaking, you can increase... If you want the abnormality to be more prominent over a longer period of time, you can increase the size. .

[0026] A hysteresis threshold mechanism is used to determine whether to open or close the gate. When the measuring point is currently in a non-alarm state, and When this occurs, mark the measuring point as a timing alarm point (gate opening); When the measuring point is currently in an alarm state, and When the alarm status of the measuring point is cancelled (gate closed); when When the value is between the two thresholds, maintain the original state and avoid back-and-forth fluctuations near the threshold.

[0027] in, The threshold value for opening the gate ranges from 0.50 to 0.65. This is the gate closing threshold, with a value range of 0.30-0.40. Must be higher than When the window is not full, If the value is 0, no alarm points will be generated.

[0028] Furthermore, the deviation from the scale needs to be slowly corrected based on feedback from false alarms or missed alarms in manual annotation. The correction formula is as follows: In the formula, This indicates the deviation of dimension d from the scale. Indicates the deviation from the scale correction step size; This represents the scale residual of dimension d.

[0029] When the on-duty personnel mark an alarm as a false alarm >0 indicates a scale increase, making it more difficult to open the gate with the same burrs; this is marked as a missed report. <0, scale decreases, more sensitive. A single correction cuts off within a predetermined range, prohibiting... Change to the following; In implementing this invention, this embodiment dynamically adjusts the threshold using historical smooth values ​​within a sliding window and controls the opening and closing of the gate using a hysteresis mechanism, completely solving the problem of fixed thresholds in existing technologies. Specifically, the threshold of each measuring point adaptively changes according to its recent behavior. The average value and scale of points prone to jitter will automatically increase, and the same spikes will not exceed the opening threshold. When a true displacement occurs, the intensity dimension deviates significantly, and outliers quickly exceed the threshold, preventing missed alarms. At the same time, the deviation scale can be slowly corrected based on manual feedback, further suppressing false alarms. Noise reduction is no longer done after the alarm is generated, but is intercepted in advance at the threshold stage. That is, the threshold itself moves with the window, and most of the jitter noise can be blocked at the edge.

[0030] S40, perform clustering and noise reduction on all marked time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster; It should be noted that, in this embodiment, for all timing alarm points currently in the open state, the feature vector distance is calculated pairwise. The feature vector distance is determined by the normalized time difference, whether the loops are the same, whether the measurement point types are of the same type, and the outlier difference. Each component is first compressed to [0,1] before the Euclidean distance is calculated. When a component is missing, it is counted as 1 (partially dissimilar) to avoid hard merging. For example, time difference normalization: the time difference between two alarm points is divided by the maximum allowable time difference (e.g., 30 seconds), and truncated to [0,1]; the same loop is counted as 0, and different loops as 1; the measurement point types are of the same type, and different loops as 1; the outlier difference is the difference between the two points. The absolute value of the difference.

[0031] The feature vector distance is converted into similarity using the following formula: In the formula, The value range is (0,1]. The smaller the value, the higher the similarity. (The similarity is then considered.) Reaching the preset lower limit of clustering The alarm points are merged into one alarm cluster, in the example. Set the value to 0.7. The merging operation uses a disjoint-set data structure algorithm; single-point clustering is also allowed. Outliers are selected within each cluster. The largest point is used as the representative point. If points are tied, the point with higher preset criticality is selected, followed by the point with an earlier displacement time. The remaining points within the cluster are added to the member list. Through clustering noise reduction, large-scale displacements caused by communication interruptions are grouped into a single cluster, eliminating the need for dozens of independent alarms. Points that do not meet the similarity lower limit are not forcibly merged to avoid bundling unrelated loops together.

[0032] In the specific implementation of this invention, multiple alarms triggered by the same fault source (such as a single incoming line displacement) are automatically merged into a cluster due to their small time difference, identical circuits, similar types, and small feature vector distance, resulting in high similarity and preventing them from reporting independently. Similarly, large-scale displacements caused by intermittent communication interruptions are also clustered together because their time difference is close to zero and they share the same gateway, preventing confusion with genuine power outages. The association between points no longer relies on hard binding by names or rules but is dynamically determined by the vector distance composed of multi-dimensional temporal semantic features. This directly solves the two major problems of high redundancy and weak association in existing technologies.

[0033] S50 aggregates alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculates the event's level score based on the static criticality and representative outlier of the alarm cluster. It should be noted that in this embodiment, alarm clusters within the same circuit or cabinet are further aggregated into a single multi-dimensional time-series alarm event. Alarms that cross circuits but are already merged into the same cluster in S40 follow the cluster and are not separated. The event record information includes: the set of member measurement points, the primary remote signaling status (representing the status of the point), and the static criticality. , representing outliers And rating .

[0034] Among them, static keyness Take the maximum value of the preset criticality score for all measurement points within the cluster. The preset criticality score is set in the system configuration: high criticality scores (e.g., 0.90~0.95) are assigned to critical equipment such as incoming lines, UPS inputs, and ATS; medium criticality scores (e.g., 0.50~0.60) are assigned to ordinary feeders; and low criticality scores (e.g., 0.30) are assigned to socket-level equipment. This represents outlier segments. Take outgroups from all measurement points within the cluster The maximum value.

[0035] The formula for calculating the rating is: In the formula, This indicates the level score of alarm cluster c; This indicates the weight of static keyness in the ranking score, with a value ranging from 0.3 to 0.5. This indicates the static criticality of alarm cluster c; The alarm cluster c represents the outlier, and the data within the cluster is taken. The maximum value. When When the trend is biased towards criticality, a change in the entry line makes it easier to enter a higher-level channel; when the trend is biased towards outliers, a sharp fluctuation in a regular point may also improve the level, but it is still limited by... The difference. Empty clusters are not included in the ranking score.

[0036] In this specific implementation of the invention, multiple alarm clusters from the same circuit or cabinet are grouped into a single event. Instead of a string of red dots on the screen, the screen displays an aggregated event with a member list, allowing maintenance personnel to easily identify the root cause. Simultaneously, the calculation of the severity score integrates device criticality and outlier status, laying a quantitative foundation for subsequent hierarchical routing and avoiding the drawback of traditional methods that, while hierarchical, do not segment notifications. S60, based on the grade score, routes multi-dimensional time-series alarm events to the corresponding notification channels; It should be noted that in this embodiment, two levels of routing thresholds are preset: the first level is the lower routing threshold. and the lower limit of secondary routing Specifically, when ≥ The event is routed to the duty channel, and an instant message is pushed to the duty personnel; when ≤ < When the event is routed to the work group channel, it is pushed to the corresponding work group; when < At this time, events are only written to the edge daily report and not pushed as instant messages. Channels are configured on edge nodes and are not hardcoded to any particular SMS gateway. When communication interruption characteristics are obvious (same gateway, simultaneous widespread changes, accompanied by an increase in the interruption count), events are still aggregated according to the above process, but the static criticality is... Use communication-related criticality, which is lower than the criticality of incoming line tripping, to avoid intermittent outages directly hitting the duty channel.

[0037] S70, generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only if there is a difference will the event be pushed through the corresponding notification channel.

[0038] It should be noted that, in this embodiment, an event fingerprint is maintained for each multi-dimensional time-series alarm event. The event fingerprint includes at least the set of member measurement points, the level, and the main remote signaling status. The event identifier uses a loop plus representation of the measurement point, which is stable within the same field.

[0039] The fingerprint of the currently captured event is compared item by item with the fingerprint of the last successfully pushed event: If any discrepancies exist (including member additions or removals, level upgrades or downgrades, primary remote signaling status changes from split to merge or from merge to split, reverting, etc.), the discrepancy will be pushed through the routing channel determined by S60, with the push message carrying the member list and level information, and updating the pushed fingerprints. If there is no difference, suppress this push, only update the current event snapshot on the edge node, and do not send any notification.

[0040] When the event disappears (all alarm points in the cluster have been shut down) and an alarm was previously pushed, a reset notification is sent to the original channel, and then the fingerprint is cleared.

[0041] In implementing this invention, this embodiment only pushes notifications when there is a substantial change in the event state (addition or removal of members, level increase or decrease, or main state reversal); otherwise, notifications are suppressed. The notification volume perfectly matches the actual changes on-site, and the duty phone is no longer bombarded with duplicate messages. This directly solves the problem of channel congestion and alarm fatigue caused by repeated pushes of the same alarm event in the prior art.

[0042] Reference Figure 2 As shown, this invention also discloses a multi-dimensional time-series alarm hierarchical aggregation system for digital infrastructure, comprising: The data access module 601 is used to access time-series data from various remote sensing measurement points in the digital infrastructure in real time. The feature extraction module 602 is used to extract multidimensional temporal semantic outlier features of each remote sensing measurement point at each sampling time; the multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency and anomaly duration; The adaptive threshold module 603 is used to dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and to calculate its multidimensional time-series outlier based on the displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the opening threshold, the remote signaling point is marked as a time-series alarm point. The clustering and noise reduction module 604 is used to perform clustering and noise reduction on all the marked time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster. The hierarchical aggregation module 605 is used to aggregate alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculate the level score of the event based on the static criticality and representative outlier of the alarm cluster. The graded routing module 606 is used to route multi-dimensional time-series alarm events to the corresponding notification channels according to grade scores. The difference push module 607 is used to generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only when there is a difference will the push be made through the corresponding notification channel.

[0043] It should be noted that the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation strategy provided in the above embodiments is only an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation system 600 will be divided into different functional modules to complete all or part of the functions described above. Furthermore, the embodiments of the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation system 600 and the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation method belong to the same concept. The specific way each module performs its operation has been described in detail in the method embodiments, and will not be repeated here.

[0044] Figure 3 A schematic diagram of the structure of an electronic device according to an exemplary embodiment is shown.

[0045] It should be noted that this electronic device is merely an example adapted to the present invention and should not be construed as providing any limitation on the scope of use of the present invention. Furthermore, this electronic device should not be interpreted as requiring or depending on having... Figure 3 One or more components of the exemplary electronic device 2000 shown.

[0046] The hardware structure of electronic devices 2000 can vary significantly due to differences in configuration or performance, such as... Figure 3 As shown, the electronic device 2000 includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) 270.

[0047] Specifically, power supply 210 is used to provide operating voltage for various hardware devices on electronic device 2000.

[0048] Interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. Of course, in other examples adapted to this invention, interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input / output interface 235, and at least one USB interface 237, etc. Figure 3 As shown, this does not constitute a specific limitation.

[0049] The memory 250 serves as a carrier for resource storage and can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored on it include the operating system 251, application programs 253, and data 255, etc., and the storage method can be temporary storage or permanent storage.

[0050] The operating system 251 is used to manage and control the various hardware devices and application programs 253 on the electronic device 2000, so as to enable the central processing unit 270 to perform calculations and processing on the massive data 255 in the memory 250. It can be Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0051] Application 253 is a computer-readable instruction based on operating system 251 that performs at least one specific task, and may include at least one module ( Figure 3 (Not shown), each module may contain computer-readable instructions for electronic device 2000. For example, the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation device can be regarded as application 253 deployed on electronic device 2000.

[0052] Data 255 may be signal information, etc., and is stored in memory 250.

[0053] The central processing unit 270 may include one or more processors and is configured to communicate with the memory 250 via at least one communication bus to read computer-readable instructions stored in the memory 250, thereby enabling the computation and processing of massive amounts of data 255 in the memory 250. For example, a multi-dimensional time-series alarm hierarchical aggregation method for digital infrastructure can be implemented by the central processing unit 270 reading a series of computer-readable instructions stored in the memory 250.

[0054] Furthermore, the present invention can also be implemented through hardware circuits or a combination of hardware circuits and software. Therefore, the implementation of the present invention is not limited to any specific hardware circuit, software, or combination thereof.

[0055] Please see Figure 4 This invention provides an electronic device 4000, which may include: a desktop computer, a laptop computer, a server, etc., with sensor recognition capabilities.

[0056] exist Figure 4 In this context, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0057] The data interaction between the processor 4001 and the memory 4003 can be achieved through at least one communication bus 4002. This communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0058] Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0059] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0060] The memory 4003 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program instructions or code in the form of instructions or data structures and accessible by the electronic device 4000, but not limited thereto.

[0061] The memory 4003 stores computer-readable instructions, and the processor 4001 can read the computer-readable instructions stored in the memory 4003 through the communication bus 4002.

[0062] The computer-readable instructions are executed by one or more processors 4001 to implement the digital infrastructure multi-dimensional timing alarm hierarchical aggregation method in the above embodiments.

[0063] Furthermore, this embodiment of the invention provides a storage medium storing computer-readable instructions, which are executed by one or more processors to implement the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation method as described above.

[0064] Compared with existing technologies, the present invention offers the following advantages: By dynamically adjusting the threshold based on the historical smooth values ​​of each remote sensing measurement point within its sliding window and calculating its multidimensional temporal outliers, the remote sensing measurement point is marked as a temporal alarm point when the outlier reaches the opening threshold, thus solving the false alarm problem caused by the inability of fixed thresholds to adapt to dynamic operating conditions; by clustering and denoising all marked temporal alarm points, alarm points that meet preset similarity conditions are merged into an alarm cluster, and alarm clusters belonging to the same physical unit or logical loop are aggregated into a multidimensional temporal alarm event, thus solving the problem of scattered alarm presentation and difficulty in fault location caused by isolated processing of measurement points; by generating a fingerprint of the current event and comparing it with the fingerprint of the last successfully pushed event, the notification is pushed only when there is a difference, thus solving the problem of notification channel congestion and resource waste caused by repeated pushes of the same alarm event. Therefore, the present invention can significantly reduce the false alarm rate and redundant alarms, and optimize fault location efficiency and notification accuracy.

[0065] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for multi-dimensional time-series alarm hierarchical aggregation of digital infrastructure, characterized in that, The method for hierarchical aggregation of multi-dimensional time-series alarms for digital infrastructure includes the following steps: S10, real-time access to time-series data from various remote sensing measurement points in the digital infrastructure; S20, extract multidimensional temporal semantic outlier features for each remote sensing measurement point at each sampling time; wherein, the multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency, and anomaly duration; S30, dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and calculate its multidimensional time-series outlier based on displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the gate opening threshold, mark the remote signaling point as a time-series alarm point. S40, perform clustering and noise reduction on all marked time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster; S50 aggregates alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculates the event's level score based on the static criticality and representative outlier of the alarm cluster. S60, based on the grade score, routes multi-dimensional time-series alarm events to the corresponding notification channels; S70, generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only if there is a difference will the event be pushed through the corresponding notification channel.

2. The method of claim 1, wherein, In step S30, the specific method for dynamically adjusting the threshold based on the historical smoothing value of each remote sensing point within its sliding window, and calculating its multidimensional time-series outliers based on displacement intensity, displacement frequency, and anomaly persistence, is as follows: For each time-series semantic outlier feature of each remote sensing measurement point, a sliding window is maintained, and the mean of the smoothed values ​​within the window is calculated as the local normal baseline. Calculate the absolute deviation of the current smoothed value from the baseline, and combine it with a preset deviation scale to obtain the deviation ratio of each feature dimension; The multi-dimensional time series outlier scores are obtained by weighted sum of the deviation ratios of the multiple characteristic dimensions, and the calculation expression is as follows: ; wherein, represents the multi-dimensional time series outlier score of the measuring point i; represents the deviation ratio of the displacement strength dimension; represents the deviation ratio of the displacement frequency dimension; represents the deviation ratio of the abnormal duration dimension; represents the weight of the displacement strength dimension in the in-group deviation score; represents the weight of the displacement frequency dimension in the in-group deviation score; represents the weight of the abnormal duration dimension in the in-group deviation score; and .

3. The method for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure according to claim 2, characterized in that, The deviation scale needs to be slowly corrected based on feedback from false alarms or missed alarms in manual annotation. The correction formula is as follows: In the formula, This indicates the deviation of dimension d from the scale. Indicates the deviation from the scale correction step size; This represents the scale residual of dimension d.

4. The method for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure according to claim 1, characterized in that, In step S40, the specific method for clustering and denoising all marked time-series alarm points is as follows: Calculate the feature vector distance between any two alarm points The feature vector distance is determined by the normalized time difference, whether the loops are the same, whether the measurement point types are of the same type, and the outlier difference. The feature vector distance is converted into similarity using the following formula: and similarity Reaching the preset lower limit of clustering The alarm points are merged into one alarm cluster.

5. The method for hierarchical aggregation of multi-dimensional time-series alarms in digital infrastructure according to claim 1, characterized in that, In step S50, the formula for calculating the grade score is: In the formula, This indicates the level score of alarm cluster c; This indicates the weight of static keyness in the ranking score; This indicates the static criticality of alarm cluster c; The alarm cluster c represents the outlier, and the data within the cluster is taken. The maximum value.

6. A multi-dimensional time-series alarm hierarchical aggregation system for digital infrastructure, characterized in that, include: The data access module is used to access time-series data from various remote sensing measurement points in the digital infrastructure in real time; The feature extraction module is used to extract multidimensional temporal semantic outlier features of each remote sensing measurement point at each sampling time. The multidimensional temporal semantic outlier features include at least displacement intensity, displacement frequency, and anomaly persistence; The adaptive threshold module is used to dynamically adjust the threshold according to the historical smooth value of each remote signaling point within its sliding window, and to calculate its multidimensional time-series outlier based on displacement intensity, displacement frequency and anomaly persistence. When the outlier reaches the gate opening threshold, the remote signaling point is marked as a time-series alarm point. The clustering and noise reduction module is used to cluster and reduce noise for all labeled time-series alarm points, and merge alarm points that meet the preset similarity conditions into an alarm cluster. The hierarchical aggregation module is used to aggregate alarm clusters belonging to the same physical unit or logical loop into a single multi-dimensional time-series alarm event, and calculate the level score of the event based on the static criticality and representative outlier of the alarm cluster. The hierarchical routing module is used to route multi-dimensional time-series alarm events to the corresponding notification channels based on hierarchical scores. The difference push module is used to generate the fingerprint of the current event and compare it with the fingerprint of the last successfully pushed event. Only when there is a difference will the event be pushed through the corresponding notification channel.

7. An electronic device, characterized in that, Includes at least one processor and at least one memory, wherein, The memory stores computer-readable instructions; The computer-readable instructions are executed by one or more processors, causing the electronic device to implement the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation method as described in any one of claims 1-5.

8. A storage medium having computer-readable instructions stored thereon, characterized in that, The computer-readable instructions are executed by one or more processors to implement the digital infrastructure multi-dimensional time-series alarm hierarchical aggregation method according to any one of claims 1-5.