Dynamic environment monitoring data visualization platform and abnormality diagnosis method

Abstract

The invention relates to the technical field of dynamic environment data monitoring, and discloses a dynamic environment monitoring data visualization platform and an abnormality diagnosis method, and the method comprises the steps: recognizing abnormal data in dynamic environment data; constructing a causal transaction graph of the dynamic environment data; exploring a fault path for each item of abnormal data based on the causal transaction graph; calculating the fault confidence of each node in each fault path; screening and trimming each fault path based on the fault confidence of the node; obtaining a weight value of a node based on the causal transaction graph, and calculating the comprehensive causal strength of each fault path based on the fault confidence and the weight value of the node; and determining an actual fault path and diagnosing a fault source based on the comprehensive causal strength. According to the invention, the accuracy of the abnormal diagnosis result of the power environment monitoring system and the reliability of fault source positioning are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power environment data monitoring, and in particular to a power environment monitoring data visualization platform and an abnormality diagnosis method. BACKGROUND

[0002] In order to realize real-time state monitoring and risk early warning of the power environment system, the power environment monitoring system is widely used in the industry, which uses sensors, collection devices and data analysis platforms to continuously monitor various power supply parameters and environmental indicators, which can reflect the state changes of power supply equipment and its operating environment to a certain extent, and assist operation and maintenance personnel to timely discover potential risks. However, the existing power environment monitoring system mostly stays at the level of threshold-based judgment or single-point index abnormality detection, and has some deficiencies.

[0003] The existing monitoring system lacks in-depth modeling of the complex linkage relationship between equipment and environment. The operating state of power equipment and environmental factors have a high coupling relationship, and the existing technology often monitors each parameter in isolation, lacks systematic causal logic reasoning, and cannot accurately judge the relevance and linkage between different parameter changes, making it difficult to effectively identify the root cause of abnormal events.

[0004] Traditional fault diagnosis methods are mostly based on a single reasoning path or fixed diagnosis rules, and lack multi-path and multi-result parallel reasoning capability. In actual systems, the same abnormal phenomenon may be caused by multiple reasons, and there is a complex causal transmission chain between different fault sources. A single reasoning conclusion is easy to misjudge, and cannot fully and accurately reflect the real fault state of the system, limiting the accuracy and comprehensiveness of fault source positioning.

[0005] A data center power environment monitoring system and method are disclosed in Chinese Patent No. CN110046074B, which performs power environment monitoring operations through a set of collection modules and monitoring modules. The collection module uses sensors deployed on the collection module to collect data from the data center power system and sends the collected data to the monitoring module. The monitoring module receives the collection data sent by the collection module, establishes a data analysis model for the collection data, analyzes the collection data using the data analysis model, identifies whether the data center power environment system is abnormal, and simultaneously performs safety warning on the data center power environment system. The purpose of real-time monitoring of the equipment operation and environmental state of the data center power system is achieved, and the safety of data storage is improved. At the same time, the monitoring system can be viewed through an APP, and the desired information can be called at any time and anywhere through voice, improving the convenience of the monitoring system. However, this technical solution still has the problem mentioned in the background of the present application: lack of multi-path and multi-result parallel reasoning capability.

[0006] The information disclosed in this Background section is only for the purpose of increasing an understanding of the general context of the present application and does not necessarily constitute an acknowledgement or any form of suggestion that this information forms part of the prior art already known to a person of ordinary skill in the art. SUMMARY

[0007] The technical problem to be solved by the present application is to overcome the defects of the prior art, and to provide a power environment monitoring data visualization platform and an abnormality diagnosis method, so as to improve the accuracy of abnormality diagnosis results and the reliability of fault source positioning of the power environment monitoring system.

[0008] To solve the above technical problems, the present application provides the following technical solutions:

[0009] In one aspect, the present application provides a power environment monitoring data abnormality diagnosis method, comprising the following steps:

[0010] Identifying abnormal data in power environment data; constructing a causal abnormality graph of power environment data;

[0011] Exploring a fault path for each item of abnormal data based on the causal abnormality graph;

[0012] Calculating the fault confidence of each node in each fault path, and performing linkage consistency verification and compensation correction on the fault confidence of the node;

[0013] Filtering and pruning each fault path based on the fault confidence of the node;

[0014] Obtaining a weight value of the node based on the causal abnormality graph, and calculating a comprehensive causal strength of each fault path based on the fault confidence and the weight value of the node;

[0015] Determining an actual fault path and diagnosing a fault source based on the comprehensive causal strength.

[0016] As a preferred solution of the power environment monitoring data abnormality diagnosis method described in the present application, wherein: the power environment data includes power data and environment data; wherein the power data includes operating parameters of power supply equipment in a power system; the environment data is monitoring data of a working environment of the power supply equipment in the power system;

[0017] Each type of power environment data in each partition is recorded as an item of power environment data separately;

[0018] The identification of abnormal data in power environment data specifically includes:

[0019] respectively, for each item of power environment data, a time series is established; for any item of power environment data, the latest m monitoring values in the time series are extracted; based on the latest m monitoring values, a change rate and a fluctuation degree of the power environment data are calculated; m is a positive integer;

[0020] a reference value interval, a change rate threshold, and a fluctuation degree threshold are set for each item of power environment data respectively;

[0021] If the power environment data satisfies at least one of the following conditions: the change rate is greater than the change rate threshold, the fluctuation degree is greater than the fluctuation degree threshold, and at least n of the m monitoring values are out of the reference value interval, then the power environment data is abnormal data; n is a positive integer less than or equal to m.

[0022] As a preferred scheme of the power environment monitoring data anomaly diagnosis method described in the present application, wherein: the causal abnormality graph includes nodes and directed edges; any node corresponds to an item of power environment data; any directed edge is used to connect two nodes, and points from an upstream node to a downstream node;

[0023] If any two nodes are connected by a directed edge, then the corresponding two nodes have an abnormal linkage relationship, which specifically includes: when the power environment data corresponding to the upstream node of the corresponding two nodes is abnormal, there is a probability that it will cause the power environment data corresponding to the downstream node to be abnormal, and the probability is equal to the edge weight of the corresponding directed edge;

[0024] Collect historical data, and count the probability that any item of power environment data will cause any item of power environment data other than itself to be abnormal when it is abnormal, and assign a weight value to the corresponding directed edge in the causal abnormality graph.

[0025] As a preferred scheme of the power environment monitoring data anomaly diagnosis method described in the present application, wherein: the fault path includes nodes and directed edges in the causal abnormality graph; when exploring a fault path for any abnormal data, the node corresponding to the corresponding abnormal data is marked as a target node; any fault path is acyclic and unbranched, and the target node has no downstream node;

[0026] The method of exploring the fault path is as follows: taking the target node as the starting point, exploring the nodes in the causal abnormality graph in the direction opposite to the directed edges; each time a node is explored, the nodes and directed edges currently contained in the fault path are updated, and the product of the weight values of all directed edges currently contained in the fault path is calculated; if the product of the weight values is less than a preset minimum weight threshold, the exploration is stopped, and a fault path is obtained.

[0027] As a preferred scheme of the power environment monitoring data anomaly diagnosis method, the method for calculating the fault confidence of each node in each fault path comprises the following steps: assigning an initial value to the fault confidence based on the time sequence of the power environment data corresponding to the node, specifically comprising the following steps: if the power environment data corresponding to the node is normal data, the fault confidence is 0; otherwise:

[0028] The mean value of the latest m monitoring values in the time sequence is calculated as the observation value of the node; the change rate and the fluctuation degree calculated based on the latest m monitoring values are obtained; the initial value of the fault confidence is assigned based on any one of the absolute value of the difference between the observation value and the midpoint of the reference value interval, the difference between the change rate and the change rate threshold, and the difference between the fluctuation degree and the fluctuation degree threshold;

[0029] The linkage consistency verification of the fault confidence of the node comprises abnormal direction verification and abnormal lag verification of each non-target node; if any non-target node fails to pass the abnormal direction verification or the abnormal lag verification, the confidence of the corresponding non-target node is set to 0.

[0030] As a preferred scheme of the power environment monitoring data anomaly diagnosis method, the abnormal direction verification specifically comprises the following steps: extracting the difference between the observation value of any non-target node and the observation value of the adjacent downstream node in the fault path and the midpoint of the reference value interval, respectively, to identify the abnormal direction of the non-target node and the adjacent downstream node; the abnormal direction comprises abnormal rise and abnormal fall;

[0031] The abnormal linkage direction of the non-target node and the adjacent downstream node is identified; the abnormal linkage direction comprises positive linkage and negative linkage; wherein, the positive linkage indicates that the abnormal directions of the non-target node and the adjacent downstream node are the same, and the negative linkage indicates that the abnormal directions are opposite;

[0032] If the abnormal directions of the non-target node and the adjacent downstream node do not satisfy the abnormal linkage direction, the non-target node fails to pass the abnormal direction verification

[0033] As a preferred scheme of the power environment monitoring data anomaly diagnosis method, the abnormal lag verification specifically comprises the following steps: the time sequence corresponding to each node in the fault chain is intercepted through a sliding window with a length of m monitoring values, and it is judged whether the power environment data is abnormal data after each sliding interception; the time point at which each power environment data starts to be abnormal is determined as the abnormal starting point of the corresponding node in the fault chain based on the judgment result after each sliding interception.

[0034] The time difference between the abnormal starting point of any non-target node and the adjacent downstream node in the fault path is calculated as the abnormal lag time of the corresponding non-target node; a lag time interval is set for each directed edge in the causal anomaly graph; if the abnormal lag time of the non-target node is not located in the lag time interval of the directed edge between the non-target node and the adjacent downstream node, the non-target node fails to pass the abnormal lag verification.

[0035] As a preferred scheme of the power environment monitoring data anomaly diagnosis method described in the application, the screening and pruning of each fault path specifically includes: after assigning an initial value to the fault confidence of each node in each fault path, screening and pruning the fault path once; after verifying the linkage consistency of the fault confidence of each node, screening and pruning the fault path twice.

[0036] The method of screening or pruning the fault path once or twice is as follows:

[0037] If the fault confidence of all non-target nodes except the target node in any fault path is 0, the corresponding fault path is deleted.

[0038] If any fault path contains at least one non-target node with a fault confidence of 0, and at least one non-target node with a confidence of 0 exists between the target node and the non-target node with a fault confidence of 0 in the fault path, the corresponding fault path is deleted.

[0039] The method of pruning the fault path once or twice is as follows:

[0040] If any fault path contains at least one non-target node with a fault confidence of 0, and no non-target node with a confidence of 0 exists between any non-target node with a fault confidence of 0 and the target node in the fault path, the corresponding fault path is pruned, specifically including: deleting all non-target nodes with a fault confidence of 0 in the fault path.

[0041] As a preferred scheme of the power environment monitoring data anomaly diagnosis method described in the application, the weight value of any node is the weight value of the directed edge between the corresponding node and the adjacent downstream node in the corresponding fault path; the comprehensive causal strength of any fault path is calculated based on the fault confidence and the weight value of the node, specifically including: multiplying the fault confidence of each node by the corresponding weight value to obtain the weighted confidence of the node; summing up the weighted confidences of all nodes except the target node in the fault path as the comprehensive causal strength.

[0042] The method for determining the actual fault path is as follows: the fault paths are sorted according to the size of the comprehensive causal strength; and the fault path with the largest comprehensive causal strength is selected as the actual fault path.

[0043] The method for diagnosing the fault source is as follows: the first node in the actual fault path is marked as the fault source; and the first node is a node without an upstream node in the actual fault path.

[0044] In a second aspect, the application provides a power environment monitoring data visualization platform, comprising an anomaly identification module, a causal modeling module, a causal reasoning module, a path evaluation module, and a visualization module; wherein:

[0045] The anomaly identification module is configured to identify abnormal data in the power environment data.

[0046] The causal modeling module is configured to construct a causal anomaly graph of the power environment data.

[0047] The causal reasoning module is configured to explore fault paths for each abnormal data based on the causal anomaly graph, and to screen and prune each fault path.

[0048] The path evaluation module is configured to calculate the comprehensive causal strength of each fault path, and to determine an actual fault path and diagnose a fault source based on the comprehensive causal strength.

[0049] The visualization module is configured to visually display the power environment data, the causal anomaly graph, the actual fault path, and the fault source.

[0050] Compared with the prior art, the application has the following beneficial effects:

[0051] The application realizes multi-dimensional data fusion of power supply equipment and environmental factors by constructing a causal anomaly graph, designing a reasoning strategy with abnormal splitting capability, and combining power environment linkage consistency verification and node confidence dynamic adjustment mechanism, which can systematically reflect the complex linkage relationship between the power system and the environment system, and improve the comprehensiveness and accuracy of abnormal analysis.

[0052] Through the structured modeling of the causal anomaly graph, the causal transmission logic between the monitoring data is clearly expressed; the linkage consistency verification mechanism is introduced, the directionality and hysteresis of the node anomaly in the reasoning path are jointly verified, and the physical logic rationality of the diagnosis result is ensured, which significantly reduces the risk of misdiagnosis and missed diagnosis.

[0053] The dynamic supplementary adjustment mode of the node confidence of the application optimizes the confidence evaluation based on factors such as multi-path co-occurrence, regional focus, environmental anomaly assistance, and historical fault tendency, which further improves the stability and reliability of abnormal positioning, helps to locate the key investigation area, and improves the practicality of fault diagnosis. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0055] Figure 1 A flowchart of the method for diagnosing abnormal power environment monitoring data provided in this application;

[0056] Figure 2 This is a schematic diagram of the structure of the power environment monitoring data visualization platform provided in this application. Detailed Implementation

[0057] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0058] Example 1

[0059] This embodiment describes a method for diagnosing anomalies in power environment monitoring data, referring to... Figure 1 The method includes the following steps: identifying anomalous data in the dynamic environment data;

[0060] The power environment data includes power data and environmental data; wherein, the power data includes the operating parameters of the power supply equipment in the power system; and the environmental data is the monitoring data of the working environment of the power supply equipment in the power system.

[0061] The preferred power data in this embodiment includes, but is not limited to, voltage data, current data, power factor, load level, battery pack internal resistance, and battery temperature; the preferred environmental data in this embodiment includes, but is not limited to, ambient temperature, ambient humidity, air conditioning operating status, water immersion monitoring information, and smoke alarm information.

[0062] The power supply equipment and its working environment are divided into zones according to functional areas or physical locations, and each type of power environment data in each zone is recorded as a separate power environment data item.

[0063] The embodiment preferably partitions the power supply equipment according to functional areas, for example, divides the power supply equipment into a power cabinet area, a UPS equipment area, and a battery pack area; the embodiment also preferably partitions the working environment of the power supply equipment according to physical locations, for example, a first equipment room partition and a second equipment room partition correspond to an independent physical equipment room where the power supply equipment is located; through the above partitioning, the physical area to which each item of power environment data belongs is classified and categorized, facilitating subsequent modeling of the causal relationship of power environment data in different areas and between areas.

[0064] The method comprises the following steps:

[0065] A monitoring period is set; at the beginning of each monitoring period, each item of power environment data is synchronously collected;

[0066] A time series of each item of power environment data is established, and the time series of each item of power environment data is timestamp-aligned;

[0067] For any item of power environment data, the latest m monitoring values in the time series are extracted; the change rate and the fluctuation degree of the power environment data are calculated based on the latest m monitoring values; m is a positive integer; the embodiment preferably takes the change rate as the average slope of the power environment data, and takes the fluctuation degree as the standard deviation of the power environment data.

[0068] A reference value interval, a change rate threshold, and a fluctuation degree threshold are set for each item of power environment data;

[0069] If the power environment data meets at least one of the following conditions: the change rate is greater than the change rate threshold, the fluctuation degree is greater than the fluctuation degree threshold, and at least n of the m monitoring values are outside the reference value interval, then the power environment data is abnormal data; n is a positive integer less than or equal to m.

[0070] In the embodiment, the reference value interval, the change rate threshold, and the fluctuation degree threshold can be set for each item of power environment data based on the statistical distribution of the device technical specification parameters and historical data;

[0071] A causal anomaly graph of the power environment data is constructed;

[0072] The causal anomaly graph comprises nodes and directed edges; any node corresponds to an item of power environment data; any directed edge is used to connect two nodes, and points from an upstream node to a downstream node;

[0073] If any two nodes are connected by a directed edge, then the corresponding two nodes have an abnormal linkage relationship, which specifically includes: when the dynamic environment data corresponding to the upstream node of the corresponding two nodes is abnormal, there is a probability that the dynamic environment data corresponding to the downstream node is abnormal, and the probability is equal to the edge weight of the corresponding directed edge;

[0074] Collect historical data, count the probability of causing any item of dynamic environment data abnormality when any item of dynamic environment data is abnormal, and assign a value to the weight value of the corresponding directed edge in the causal abnormality graph.

[0075] Explore a fault path for each abnormal data based on the causal abnormality graph;

[0076] The fault path includes nodes and directed edges in the causal abnormality graph; when exploring a fault path for any abnormal data, the node corresponding to the corresponding abnormal data is marked as a target node; any fault path is acyclic and has no branches, and the target node has no downstream nodes;

[0077] The method for exploring the fault path is as follows: taking the target node as the starting point, exploring nodes in the causal abnormality graph in the reverse direction of the directed edge; each time a node is explored, the nodes and directed edges currently included in the fault path are updated, and the product of the weight values of all directed edges currently included in the fault path is calculated; if the product of the weight values is less than a preset minimum weight threshold, the exploration is stopped, and a fault path is obtained.

[0078] In this embodiment, the process of exploring the fault path is to perform multi-path backtracking in the reverse direction of the directed edge in the graph to find a set of upstream nodes that may cause the current abnormality; each fault path represents a possible causal chain of abnormal data. The minimum weight threshold is used to limit the lower limit of the credibility of the fault path, control the effectiveness of the fault path in reasoning, and avoid the existence of a large number of low-confidence and weakly-associated nodes in the fault path.

[0079] Calculate the fault confidence of each node in each fault path, and perform linkage consistency verification and compensation correction on the fault confidence of the node;

[0080] The method for calculating the fault confidence of each node in each fault path is to assign an initial value to the fault confidence based on the time series of the dynamic environment data corresponding to the node, which specifically includes: if the dynamic environment data corresponding to the node is non-abnormal data, the fault confidence is 0; otherwise:

[0081] The mean value of the latest m monitoring values in the time sequence is calculated as an observation value of the node; a change rate and a fluctuation degree calculated based on the latest m monitoring values are obtained; an initial value of the fault confidence is assigned based on any one of the absolute value of the difference between the observation value and the midpoint of the reference value interval, the difference between the change rate and the change rate threshold, and the difference between the fluctuation degree and the fluctuation degree threshold;

[0082] In the embodiment, the greater any one of the above differences (or absolute values thereof) is, the greater the initial value of the fault confidence is.

[0083] The linkage consistency verification of the fault confidence of the node includes abnormal direction verification and abnormal lag verification on each non-target node; if any non-target node fails to pass the abnormal direction verification or the abnormal lag verification, the confidence of the corresponding non-target node is set to 0;

[0084] Simple data anomaly is not equal to the causality logic of the fault path. Only when the abnormal linkage direction and the abnormal lag time of adjacent nodes in the fault path match, the fault confidence of the node in the fault path is valid.

[0085] The abnormal direction verification specifically includes: extracting the difference between the observation value and the midpoint of the reference value interval of any non-target node and the adjacent downstream node in the fault path, respectively, to identify the abnormal direction of the non-target node and the adjacent downstream node; the abnormal direction includes abnormal rise and abnormal fall;

[0086] In the embodiment, if the difference between the observation value and the midpoint of the reference value interval of the non-target node or the adjacent downstream node is greater than 0, the abnormal direction of the corresponding node is abnormal rise, otherwise, the abnormal direction is abnormal fall.

[0087] The abnormal linkage direction of the non-target node and the adjacent downstream node is identified; the abnormal linkage direction includes positive linkage and negative linkage; wherein the positive linkage indicates that the abnormal directions of the non-target node and the adjacent downstream node are the same, and the negative linkage indicates that the abnormal directions are opposite; for example, the positive linkage is that the UPS load rise leads to the rise of the current, and the negative linkage is that the increase of the current leakage leads to the fall of the terminal voltage.

[0088] If the abnormal directions of the non-target node and the adjacent downstream node do not satisfy the abnormal linkage direction, the non-target node fails to pass the abnormal direction verification.

[0089] The abnormal direction verification can exclude the fault path nodes with obvious logical errors and ensure the overall physical logic reliability of the fault path.

[0090] The abnormal lag verification specifically comprises: intercepting a time sequence corresponding to each node in the fault chain through a sliding window with a length of m monitoring values, judging whether the power environment data is abnormal data after each sliding interception; determining a time point at which each power environment data starts to appear abnormal as an abnormal starting point of the corresponding node in the fault chain based on the judgment result after each sliding interception;

[0091] For example, sliding interception is performed in the direction from new to old along the time sequence, and the power environment data is judged as abnormal data after the first several sliding interceptions. If the power environment data is judged as non-abnormal from a certain sliding interception, the time point corresponding to the midpoint of the window of the sliding interception is the time point at which the power environment data starts to appear abnormal.

[0092] The time difference between the abnormal starting points of any non-target node and the adjacent downstream node in the fault path is calculated as the abnormal lag time of the corresponding non-target node; a lag time interval is set for each directed edge in the causal anomaly graph; if the abnormal lag time of the non-target node is not located in the lag time interval of the directed edge between the non-target node and the adjacent downstream node, the non-target node fails to pass the abnormal lag verification.

[0093] The lag time interval of each directed edge is set based on experience and represents the time difference between the start of the abnormality of the upstream node and the downstream node. For example, after UPS overload, the battery temperature usually has a lag of several minutes; after the humidity increases, the insulation performance needs to accumulate for a certain period of time; if the abnormality of the downstream node is earlier or far lags behind the abnormality of the upstream node, it is considered that the abnormality of the upstream node is not transmitted to the downstream node. Through the abnormal lag verification, the causal logic error of the fault path caused by the monitoring system error and noise interference can be prevented, and the fault path of the false exploration can be excluded.

[0094] Through the above linkage consistency verification mechanism, simple independence parameter abnormality and real linkage abnormality consistent with the causal logic of the system can be effectively distinguished; thereby ensuring that the final output diagnosis result is not only based on the data abnormality itself, but also based on the logical causal linkage between the data, and the accuracy and interpretability of the diagnosis are improved.

[0095] The fault confidence of each node is compensated and corrected, specifically comprising:

[0096] If any node exists in more than one fault path, the fault confidence of the corresponding node is increased, and the more times the node appears in different fault paths, the greater the increase in the fault confidence;

[0097] If the same node appears in multiple independent reasoning failure paths, and these paths are different but all take the node as the key relay, it indicates that the node has a greater abnormal influence on the power environment system. Accordingly, the confidence of the node can be appropriately increased to enhance the consistency and credibility of the overall abnormal explanation.

[0098] If the power environment data corresponding to at least M nodes correspond to the same partition of the power supply equipment and its working environment, the failure confidence of all nodes corresponding to the partition is increased.

[0099] If the abnormal nodes in multiple failure paths are concentrated in a certain functional area or physical location, such as most paths involving the UPS system, a certain battery pack, etc., it indicates that the overall risk level of the area is higher. Appropriately increasing the failure confidence of all related nodes in the area helps to locate the key investigation area and improve the practicality of fault diagnosis.

[0100] Based on the failure confidence of the node, each failure path is screened and pruned;

[0101] The screening and pruning of each failure path specifically includes: after assigning an initial value to the failure confidence of each node in each failure path, screening and pruning the failure path once; after verifying the linkage consistency of the failure confidence of each node, screening and pruning the failure path twice;

[0102] The method for screening or screening the failure path twice is as follows:

[0103] If the failure confidence of all non-target nodes except the target node in any failure path is 0, the corresponding failure path is deleted;

[0104] If any failure path contains at least one non-target node with a failure confidence of 0, and at least one of the non-target nodes with a failure confidence of 0 in the failure path exists between the target node and the non-target node with a confidence of 0, the corresponding failure path is deleted.

[0105] The method for pruning or pruning the failure path twice is as follows:

[0106] If any failure path contains at least one non-target node with a failure confidence of 0, and any non-target node with a failure confidence of 0 in the failure path does not exist between the target node and the non-target node with a confidence of 0, the corresponding failure path is pruned, specifically including: deleting all non-target nodes with a failure confidence of 0 in the failure path.

[0107] Based on the weight value of the node obtained based on the causal anomaly graph, and based on the failure confidence and weight value of the node, the comprehensive causal strength of each failure path is calculated;

[0108] The weight value of any node is the weight value of the directed edge between the corresponding node and the adjacent downstream node in the corresponding fault path; the comprehensive causal strength of any fault path is calculated based on the node-based fault confidence and the weight value, specifically comprising: multiplying the fault confidence of each node by the corresponding weight value to obtain the weighted confidence of the node; summing up the weighted confidence of all nodes in the fault path except the target node as the comprehensive causal strength.

[0109] An actual fault path is determined based on the comprehensive causal strength, and a fault source is diagnosed.

[0110] The method for determining the actual fault path is as follows: the fault paths are sorted according to the size of the comprehensive causal strength; the fault path with the largest comprehensive causal strength is selected as the actual fault path.

[0111] The method for diagnosing the fault source is as follows: the first node in the actual fault path is marked as the fault source; the first node is a node in the actual fault path that does not have an upstream node.

[0112] Embodiment 2

[0113] This embodiment is the second embodiment of the present application; based on the same inventive concept as Embodiment 1, refer to Figure 2 This embodiment introduces a power environment monitoring data visualization platform, which includes an anomaly identification module, a causal modeling module, a causal reasoning module, a path evaluation module, and a visualization module; wherein:

[0114] The anomaly identification module is used to identify abnormal data in power environment data; this module accurately marks the abnormal state of various parameters by real-time monitoring of power data of power supply equipment and environmental data of working environment, combined with dynamic abnormal intervals, to ensure the effectiveness and timeliness of the data source.

[0115] The causal modeling module is used to construct a causal anomaly graph of power environment data; this module establishes a directed graph structure reflecting the causal logic of data based on device operation logic, environmental linkage mechanism, and historical data relationship, providing a complete data correlation basis for subsequent fault reasoning and path analysis.

[0116] The causal reasoning module explores fault paths for each abnormal data based on the causal anomaly graph, and screens and prunes each fault path; this module excavates multiple potential fault causal chains through reverse search, and dynamically screens out low-confidence or unreasonable redundant paths, improving reasoning efficiency and accuracy.

[0117] The path evaluation module is used to calculate the comprehensive causal strength of each fault path, and determine the actual fault path and diagnose the fault source based on the comprehensive causal strength.

[0118] The visualization module is used for visualizing the power environment data, the causal abnormality graph, the actual fault path and the fault source, so as to facilitate the operation and maintenance personnel to quickly understand the system state, trace the abnormal transmission path, and improve the fault troubleshooting and disposal efficiency.

[0119] The specific function implementation of each module above refers to the related content in the power environment monitoring data abnormality diagnosis method described in Embodiment 1, and is not described herein.

[0120] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media containing computer usable program codes (including but not limited to disk memory, CD-ROM, optical memory, etc.).

[0121] The embodiments of the present application are described above in combination with the drawings, but the present application is not limited to the specific embodiments described above, and the specific embodiments described above are only illustrative but not restrictive. Those skilled in the art can make many forms under the inspiration of the present application without departing from the purpose and the protected scope of the present application, and these all belong to the protection of the present application.

Claims

1. A method for diagnosing anomalies in power environment monitoring data, characterized in that: Includes the following steps: Identify anomalous data in dynamic environmental data; construct a causal anomaly diagram of dynamic environmental data; Based on the aforementioned causal anomaly diagram, explore fault paths for each abnormal data item; Calculate the fault confidence of each node in each fault path, and perform linkage consistency verification and compensation correction on the fault confidence of the nodes; Each faulty path is filtered and pruned based on the fault confidence of the node. The weight values ​​of nodes are obtained based on the causal anomaly graph, and the comprehensive causal strength of each fault path is calculated based on the fault confidence and weight values ​​of the nodes. The actual fault path is determined and the fault source is diagnosed based on the comprehensive causal strength.

2. The method for diagnosing abnormal power environment monitoring data as described in claim 1, characterized in that: The power environment data includes power data and environmental data; wherein, the power data includes the operating parameters of the power supply equipment in the power system; and the environmental data is the monitoring data of the working environment of the power supply equipment in the power system. The power supply equipment and its working environment are divided into zones according to functional areas or physical locations, and each type of power environment data in each zone is recorded as a separate power environment data item. The identification of abnormal data in the dynamic environment data specifically includes: Establish time series for each piece of dynamic environmental data; for any piece of dynamic environmental data, extract the latest m monitoring values ​​from the time series; calculate the rate of change and fluctuation of the dynamic environmental data based on the latest m monitoring values; m is a positive integer; For each dynamic environment data point, set a reference value range, a change rate threshold, and a fluctuation level threshold. If the dynamic environment data meets the following conditions: the rate of change is greater than the rate of change threshold, the degree of fluctuation is greater than the degree of fluctuation threshold, and at least n of the m monitored values ​​exceed at least one of the reference value ranges, then the dynamic environment data is abnormal data; n is a positive integer less than or equal to m.

3. The method for diagnosing abnormal power environment monitoring data as described in claim 2, characterized in that: The causal anomaly graph includes nodes and directed edges; each node corresponds to a dynamic environment data point; each directed edge connects two nodes and points from the upstream node to the downstream node. If any two nodes are connected by a directed edge, then the two nodes have an abnormal linkage relationship. Specifically, when the power environment data of the upstream node is abnormal, there is a probability that the power environment data of the downstream node will be abnormal, and the probability is equal to the edge weight of the directed edge. Collect historical data, calculate the probability that any abnormality in a dynamic environmental data point will cause an abnormality in any other dynamic environmental data point besides itself, and assign weight values ​​to the corresponding directed edges in the causal anomaly graph.

4. The method for diagnosing abnormal power environment monitoring data as described in claim 3, characterized in that: The fault path includes nodes and directed edges in the causal anomaly graph; when exploring a fault path for any abnormal data, the node corresponding to the abnormal data is marked as the target node; any fault path is acyclic and has no branches, and the target node has no downstream nodes; The method for exploring fault paths is as follows: Starting from the target node, explore nodes in the direction of the directed edges in the causal anomaly graph; for each node explored, update the nodes and directed edges currently included in the fault path, and calculate the product of the weight values ​​of all directed edges currently included in the fault path; if the product of the weight values ​​is less than a preset minimum weight threshold, stop exploring and obtain a fault path.

5. The method for diagnosing abnormal power environment monitoring data as described in claim 4, characterized in that: The method for calculating the fault confidence of each node in each fault path involves assigning an initial value to the fault confidence based on the time series of the power environment data corresponding to the node. Specifically, this includes: if the power environment data corresponding to the node is not abnormal data, then the fault confidence is 0; otherwise: Calculate the mean of the latest m monitoring values ​​in the time series as the observed value of the node; obtain the rate of change and the degree of fluctuation calculated based on the latest m monitoring values; assign an initial value to the fault confidence based on any one of the following: the absolute value of the difference between the observed value and the midpoint of the reference value interval, the difference between the rate of change and the rate of change threshold, and the difference between the degree of fluctuation and the degree of fluctuation threshold. The linked consistency verification of the fault confidence of the nodes includes abnormal direction verification and abnormal lag verification for each non-target node; if any non-target node fails the abnormal direction verification or abnormal lag verification, the confidence of the corresponding non-target node is set to 0.

6. The method for diagnosing abnormal power environment monitoring data as described in claim 5, characterized in that: The abnormal direction verification specifically includes: extracting the difference between the observed value and the midpoint of the reference value interval for any non-target node and the adjacent downstream node in the fault path, and identifying the abnormal direction of the non-target node and the adjacent downstream node; the abnormal direction includes abnormal rise and abnormal fall. Identify the abnormal linkage direction between a non-target node and its adjacent downstream node; the abnormal linkage direction includes positive linkage and negative linkage; wherein, positive linkage means that the abnormal direction between the non-target node and its adjacent downstream node is the same, and negative linkage means that the abnormal direction is opposite. If the abnormal direction between the non-target node and its adjacent downstream node does not satisfy the abnormal linkage direction, then the non-target node fails the abnormal direction verification.

7. The method for diagnosing abnormal power environment monitoring data as described in claim 6, characterized in that: The abnormal lag verification specifically includes: extracting the time series corresponding to each node in the fault chain through a sliding window of length m monitoring values, and determining whether the power environment data is abnormal after each sliding extraction; determining the time point at which each power environment data begins to show abnormality based on the judgment result after each sliding extraction, as the abnormal starting point of the corresponding node in the fault chain; Calculate the time difference between the abnormal starting point of any non-target node and its adjacent downstream node in the fault path, and use it as the abnormal lag time of the corresponding non-target node; set a lag time interval for each directed edge in the causal anomaly graph; if the abnormal lag time of the non-target node is not located in the lag time interval of the directed edge between the non-target node and its adjacent downstream node, then the non-target node fails the abnormal lag verification.

8. The method for diagnosing abnormal power environment monitoring data as described in claim 7, characterized in that: The process of filtering and pruning each fault path specifically includes: after assigning an initial value to the fault confidence of each node in each fault path, performing a first filtering and pruning of the fault path; and after verifying the consistency of the fault confidence of each node, performing a second filtering and pruning of the fault path. The methods for performing one or two filtering of faulty paths are as follows: If, in any fault path, the fault confidence of all non-target nodes except the target node is 0, then the corresponding fault path is deleted. If any fault path contains at least one non-target node with a fault confidence of not 0, and at least one of the non-target nodes with a fault confidence of not 0 in the fault path has a non-target node with a confidence of 0 between it and the target node, then the corresponding fault path is deleted. The methods for performing one or two trims on faulty paths are as follows: If any fault path contains at least one non-target node with a fault confidence of not 0, and there are no non-target nodes with a confidence of 0 between any non-target node with a fault confidence of not 0 and the target node, then the corresponding fault path is pruned, specifically including: deleting all non-target nodes with a fault confidence of 0 in the fault path.

9. The method for diagnosing abnormal power environment monitoring data as described in claim 8, characterized in that: The weight of any node is the weight of the directed edge between the corresponding node and the adjacent downstream node in the corresponding fault path. The calculation of the comprehensive causal strength of any fault path based on the fault confidence and weight value of nodes specifically includes: multiplying the fault confidence of each node by the corresponding weight value to obtain the weighted confidence of the node; and calculating the sum of the weighted confidence of all nodes in the fault path except the target node as the comprehensive causal strength. The method for determining the actual fault path is as follows: sort the fault paths according to the magnitude of the comprehensive causal strength; select the fault path with the largest comprehensive causal strength as the actual fault path; The method for diagnosing the fault source is as follows: the first node in the actual fault path is marked as the fault source; the first node is a node in the actual fault path that does not have an upstream node.

10. A power environment monitoring data visualization platform, used to implement the power environment monitoring data anomaly diagnosis method as described in any one of claims 1-9, characterized in that: It includes an anomaly detection module, a causal modeling module, a causal reasoning module, a path evaluation module, and a visualization module; among which: The anomaly identification module is used to identify abnormal data in the power environment data; The causal modeling module is used to construct a causal anomaly diagram of dynamic environmental data; The causal reasoning module explores fault paths for each abnormal data item based on the causal anomaly graph, and filters and prunes each fault path; The path evaluation module is used to calculate the comprehensive causal strength of each fault path, and to determine the actual fault path and diagnose the fault source based on the comprehensive causal strength. The visualization module is used to visualize and display power environment data, cause-effect anomaly diagrams, actual fault paths, and fault sources.

Images