Efficient electric power operation and maintenance data acquisition system

Abstract

The invention discloses an efficient electric power operation and maintenance data acquisition system, and relates to the technical field of electric power operation and maintenance, and the system comprises a high-frequency sampling module which is used for dynamically adjusting the sampling frequency and obtaining key data in the electric power operation and maintenance process; the high-frequency sampling module comprises a data sampling unit, an anomaly detection unit and a dynamic adjustment unit. By designing the high-frequency sampling module, the dynamic adjustment of the sampling frequency is realized, the accuracy of data acquisition is improved, the subtle change of the running state of the equipment is accurately captured, and the problem that the traditional static sampling cannot cope with diversified scenes is solved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of electric power operation and maintenance, and particularly relates to an efficient electric power operation and maintenance data acquisition system. BACKGROUND

[0002] With the rapid development of the electric power industry and the continuous advancement of the smart grid construction, as a core network supporting the safe and stable operation of the electric power system, the electric power communication network is increasingly expanding in scale, and the types and quantities of equipment are continuously increasing, and the functions are also increasingly powerful, which also brings new challenges to the operation and maintenance management of the electric power communication network.

[0003] The traditional electric power system adopts a fixed sampling frequency, which cannot be dynamically adjusted according to time conditions, resulting in a mismatch between data acquisition and demand, and it is difficult to adapt to complex operation and maintenance scenarios, leading to passive operation and maintenance, resource waste and delayed fault response.

[0004] Patent CN114386632B discloses a power distribution operation and maintenance system based on electric power big data, which realizes the efficiency of electric power distribution operation and maintenance, promotes the stable and safe development of the electric power grid, reduces the equipment maintenance cost, and improves the work efficiency of electric power distribution operation and maintenance.

[0005] The above-mentioned patent solves the problem that the existing operation and maintenance management process of the electric power distribution network has great one-sidedness and inaccuracy in the operation and maintenance management mode, and it is difficult to quickly judge the fault conditions occurring in the operation process of the distribution network, and cannot guarantee the safety and stability of the operation of the distribution network, which hinders the development of the electric power distribution network, but there is still optimization space in the setting of the operation and maintenance data acquisition frequency, and the application dynamically adjusts the operation and maintenance data acquisition frequency to solve the problem that the traditional static sampling cannot cope with diversified scenarios.

[0006] Therefore, the application provides an efficient electric power operation and maintenance data acquisition system for dynamically adjusting the operation and maintenance data acquisition frequency. SUMMARY

[0007] The application aims to provide an efficient electric power operation and maintenance data acquisition system to solve the technical problems in the background.

[0008] To achieve the above-mentioned purpose, the application provides the following technical scheme: an efficient electric power operation and maintenance data acquisition system, comprising a high-frequency sampling module, the high-frequency sampling module is used for dynamically adjusting the sampling frequency and acquiring key data in the electric power operation process.

[0009] The high-frequency sampling module comprises a data sampling unit, an abnormality detection unit and a dynamic adjustment unit, the data sampling unit is connected with the abnormality detection unit and the dynamic adjustment unit through signals, and the abnormality detection unit is connected with the dynamic adjustment unit through signals.

[0010] The data sampling unit sets different sampling requirements according to the characteristics and operation and maintenance requirements of various key devices in the power operation and maintenance environment, performs high-frequency power operation and maintenance data sampling, and collects log data, index data and call chain data of the devices;

[0011] The abnormality detection unit performs multi-source data fusion on the log, index and call chain data, calculates an abnormality score through reconstruction loss and adversarial loss, constructs a fault propagation model, locates a core abnormal node, and outputs a root cause hypothesis list;

[0012] The dynamic adjustment unit identifies the need to adjust the sampling frequency according to the root cause hypothesis list, generates an adjustment strategy, and records the time of each sampling frequency adjustment and the sampling frequency before and after the adjustment.

[0013] Preferably, the high-frequency sampling module is connected with a weight adjustment module through a signal, and the weight adjustment module is used for dynamically adjusting the weight of the log, index and call chain data in the fusion process;

[0014] The weight adjustment module comprises a network construction unit, a feature aggregation unit and a weight adjustment unit, the network construction unit is connected with the feature aggregation unit and the data sampling unit through a signal, the feature aggregation unit is connected with the weight adjustment unit through a signal, and the weight adjustment unit is connected with the abnormality detection unit through a signal;

[0015] The network construction unit inputs the log, index and call chain as node features into a heterogeneous graph neural network, defines different types of meta-paths, captures high-order semantic relationships, and obtains complex dependencies between cross-modal data;

[0016] The feature aggregation unit aggregates node features using an attention mechanism, dynamically identifies the association strength of nodes and neighbor nodes, assigns attention weights to different node features, and preliminarily identifies the relative importance of different node features in the fusion process;

[0017] The weight adjustment unit assigns a confidence weight to the node features based on the historical performance of the node features, adjusts the attention weights of different node features using an adaptive fusion gating method, and generates a weight.

[0018] Preferably, the high-frequency sampling module is connected with an ad hoc network communication module through a signal, and the ad hoc network communication module is used for stable transmission of power operation and maintenance data;

[0019] The ad hoc network communication module comprises a communication evaluation unit, a network management unit and a push decision unit, the communication evaluation unit is connected with the network management unit and the push decision unit through a signal, and the network management unit is connected with the push decision unit through a signal;

[0020] The communication evaluation unit periodically evaluates the performance of the power line carrier and the micro-power wireless communication channel, uses a weighted grey correlation degree analysis method to sort the channel performance, and selects the channel with the best performance for communication.

[0021] The network management unit monitors the status of each node in the network, constructs a network topology graph, detects node faults and communication interruptions, automatically reconfigures the network topology, and generates a channel switching strategy.

[0022] The push decision unit performs rule matching on the collected and fused multi-source data, calculates the relevance of the multi-source data to the current operation and maintenance task, sets a push threshold according to a threshold update strategy, and compares the relevance and the push threshold.

[0023] Preferably, the high-frequency sampling module is connected with an anomaly evaluation module through a signal, and the anomaly evaluation module is used for dynamically allocating an anomaly level.

[0024] The anomaly evaluation module includes an intelligent word segmentation unit, an entity recognition unit, and a level evaluation unit, the intelligent word segmentation unit is connected with the entity recognition unit through a signal, the entity recognition unit is connected with the level evaluation unit through a signal, and the level evaluation unit is connected with an anomaly detection unit through a signal.

[0025] The intelligent word segmentation unit identifies the text information of the multi-source fusion data, uses a deep learning model to intelligently segment the text describing the device anomaly in the power operation and maintenance environment, and divides the word boundaries.

[0026] The entity recognition unit uses a named entity recognition technology, combines an entity type library, automatically identifies and labels key entities in the anomaly text, and constructs the association relationship between the key entities.

[0027] The level evaluation unit extracts features for level evaluation in the anomaly text, sets initial weights of the features, identifies the influence of time, device type, and operating environment on the feature weights, dynamically adjusts the feature weights, and uses a deep neural network to perform level evaluation on the anomaly.

[0028] Preferably, the high-frequency sampling module is connected with a data warehouse module through a signal, and the data warehouse module is used for storage and retrieval of operation and maintenance data.

[0029] The data warehouse module includes a storage strategy unit, an index optimization unit, and a period management unit, the storage strategy unit is connected with the index optimization unit through a signal, and the index optimization unit is connected with the period management unit through a signal.

[0030] The storage strategy unit combines the sampling frequency information of the power operation and maintenance data, analyzes the data volume growth trend, predicts the future storage demand of the power operation and maintenance data, and adjusts the storage proportion of the distributed file system and the columnar database.

[0031] The index optimization unit constructs a real-time index for the newly added data stream of the sampling frequency promotion according to the sampling frequency information, generates a query log according to the result of the user accessing and querying the power operation and maintenance data, and identifies the query frequency;

[0032] The period management unit sets the storage level of the power operation and maintenance data, and generates a data migration strategy according to the query frequency information and the sampling frequency information of the power operation and maintenance data.

[0033] Preferably, the abnormality evaluation module is connected with a data enhancement module through a signal, and the data enhancement module is used to generate high-quality enhanced samples from abnormal texts;

[0034] The data enhancement module comprises a text enhancement unit and a quality control unit, the text enhancement unit is connected with the quality control unit through a signal, and the quality control unit is connected with an entity recognition unit through a signal;

[0035] The text enhancement unit sorts the importance descriptions in the abnormal texts, adopts a masking strategy, reconstructs the masked abnormal texts, and generates enhanced samples;

[0036] The quality control unit calculates the difference between the enhanced samples and the abnormal texts, selects enhanced samples with low repetition, integrates the enhanced samples and the abnormal texts, and generates an expanded data set.

[0037] Preferably, the sampling requirements in the data sampling unit include:

[0038] Sampling requirements of device type characteristics: key devices prone to failure in power operation and maintenance are sampled in priority, and non-key devices are sampled regularly;

[0039] Sampling requirements of operation and maintenance requirements: high sampling frequency is set for key data that needs to be monitored in real time;

[0040] Sampling requirements of environmental factor characteristics: high sampling frequency is set under harsh environmental conditions;

[0041] Sampling requirements of historical data: high sampling frequency is set for devices that have frequently appeared abnormally in history.

[0042] Preferably, the adjustment strategy in the dynamic adjustment unit is to increase the sampling frequency of the device detected to have an abnormality.

[0043] Preferably, the key entity in the entity recognition unit is an entity related to the abnormal condition of the device, including: device name, abnormal type, running environment, and abnormal position.

[0044] Preferably, the data migration strategy in the cycle management unit is: according to the sampling and query frequency of the power operation and maintenance data, the power operation and maintenance data is divided into high-frequency, medium-frequency and low-frequency data, and the high-frequency, medium-frequency and low-frequency data are respectively migrated to the hot data layer, the warm data layer and the cold data layer.

[0045] Compared with the prior art, the present application has the following advantages:

[0046] 1. The present application realizes dynamic adjustment of sampling frequency by designing a high-frequency sampling module, improves the accuracy of data acquisition, accurately captures the subtle changes of the device running state, and solves the problem that traditional static sampling cannot cope with diversified scenes.

[0047] 2. The present application solves the problem of fixed data fusion weight in traditional data acquisition system, which cannot adapt to different scenes and requirements, improves the accuracy and efficiency of data fusion by designing a weight adjustment module.

[0048] 3. The present application realizes stable transmission of power operation and maintenance data by designing a self-organizing network communication module, reduces the operation and maintenance problems caused by communication failure, enhances the reliability of the power system, and realizes intelligent push of operation and maintenance data.

[0049] 4. The present application realizes the evaluation of power operation and maintenance exception by designing an exception evaluation module, improves the efficiency of data processing and exception evaluation, and solves the problems of inaccurate exception information processing and strong subjectivity of exception level evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 It is a schematic diagram of the high-frequency sampling module of the present application.

[0051] Figure 2 It is a schematic diagram of the weight adjustment module of the present application.

[0052] Figure 3 It is a schematic diagram of the self-organizing network communication module of the present application.

[0053] Figure 4 It is a schematic diagram of the exception evaluation module of the present application.

[0054] Figure 5 It is a schematic diagram of the data warehouse module of the present application.

[0055] Figure 6 It is a schematic diagram of the data enhancement module of the present application.

[0056] Figure 7 It is a schematic diagram of the working process of the exception evaluation module of the present application.

[0057] Figure 8 It is a schematic diagram of the system working process of the present application. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0059] Embodiment 1, please refer to Figure 1 and Figure 8 A high-efficiency power operation and maintenance data acquisition system includes a high-frequency sampling module for dynamically adjusting the sampling frequency to obtain key data in the power operation and maintenance process. The high-frequency sampling module includes a data sampling unit, an anomaly detection unit, and a dynamic adjustment unit. The data sampling unit sets different sampling requirements according to the characteristics and operation and maintenance requirements of various key devices in the power operation and maintenance environment, performs high-frequency power operation and maintenance data sampling, and collects log data, index data, and call chain data of the device. The anomaly detection unit performs multi-source data fusion on the log, index, and call chain data, calculates an anomaly score through reconstruction loss and adversarial loss, constructs a fault propagation model, locates the core anomaly node, and outputs a root cause hypothesis list. The dynamic adjustment unit identifies the need to adjust the sampling frequency based on the root cause hypothesis list, generates an adjustment strategy, and records the time of each sampling frequency adjustment and the sampling frequency before and after the adjustment.

[0060] Further, the data sampling unit sets corresponding sampling requirements according to device type characteristics, operation and maintenance stage characteristics, environmental factor characteristics, and historical data characteristics, etc. For example, for devices that are critical and prone to failure in the power system, devices that have failed during operation and maintenance, or devices that have experienced anomalies in the past, the sampling frequency is increased to capture subtle changes in the device's operating state and quickly locate the fault. For normally operating devices or non-critical devices, a lower sampling frequency is set to balance data acquisition volume and system resource consumption, and to reduce unnecessary data redundancy. In addition, the data sampling unit sets corresponding sampling requirements in combination with data types. For example, for device operation logs, a higher sampling frequency is set to record detailed operation records and state changes of the device. For performance indicators of the device, such as voltage, current, temperature, etc., the corresponding sampling period is set according to the indicator change rate to ensure the timeliness and accuracy of the data. For call chain data between systems, the data sampling unit sets the corresponding sampling frequency according to the frequency of system interaction, and the system collects power operation and maintenance data according to the sampling frequency set by the data sampling unit.

[0061] The anomaly detection unit receives power operation and maintenance data such as logs, indicators and call chain data from the data sampling unit, weights the log, indicator and call chain features according to the weight of each data source, obtains the fused multi-source data, reconstructs the multi-source data in the normal state through the self-encoder, calculates the difference between the multi-source data and the reconstructed data, takes the reconstruction error as part of the anomaly score, introduces an adversarial training mechanism, calculates the adversarial loss through the generative adversarial network, combines the reconstruction loss and the adversarial loss to calculate the comprehensive anomaly score, judges whether the device is in an abnormal state, after detecting the anomaly, starts the root cause positioning process, uses the heterogeneous graph neural network and the multi-source data to construct a fault propagation model, locates the fault source, and generates a detailed root cause analysis report including the fault propagation path, the recommended repair measures and the like; the dynamic adjustment unit receives the root cause hypothesis list transmitted by the anomaly detection unit, identifies the related information of the abnormal device, generates the corresponding adjustment strategy, and increases the sampling frequency of the abnormal device. In addition, if the anomaly in the root cause hypothesis list involves a key device, the dynamic adjustment unit increases the sampling frequency of the key device and synchronously increases the sampling frequency of the associated devices; the adjusted sampling frequency is transmitted to the data sampling unit in real time through the dynamic adjustment unit, and at the same time, the dynamic adjustment unit records the related information of the sampling frequency for subsequent analysis and tracing.

[0062] The high-frequency sampling module reduces the data transmission and storage pressure by reducing the sampling frequency of non-critical devices; during the fault period, the high-frequency sampling captures the transient characteristics and improves the data effectiveness; for the aging device, the high-frequency sampling time is extended to accumulate the degradation trajectory data to support predictive maintenance, help the system to find potential faults in advance, and develop a preventive maintenance plan, thereby prolonging the service life of the device and ensuring the stable operation of the system.

[0063] Embodiment 2, please refer to Figure 1 , Figure 2 and Figure 8The high-frequency sampling module comprises a data sampling unit, an abnormality detection unit and a dynamic adjustment unit, the high-frequency sampling module is connected with a weight adjustment module through a signal, the weight adjustment module is used for dynamically adjusting the weight of log, index and call chain data in the fusion process; the weight adjustment module comprises a network construction unit, a feature aggregation unit and a weight adjustment unit; the network construction unit inputs log, index and call chain as node features into a heterogeneous graph neural network, defines different types of meta paths, captures high-order semantic relationships, and obtains complex dependencies between cross-modal data; the feature aggregation unit aggregates node features using an attention mechanism, dynamically identifies the association strength of nodes and neighbor nodes, assigns attention weights to different node features, and preliminarily identifies the relative importance of different node features in the fusion process; the weight adjustment unit assigns a confidence weight to the node feature based on the historical performance of the node feature, adjusts the attention weight of different node features using an adaptive fusion gate method, and generates a weight value.

[0064] Further, the network construction unit receives power operation and maintenance data from the data sampling unit, uses a Drain parser to structurally process log data, generates a unified log template, normalizes index data, processes the log data after processing as a feature of a node, processes the normalized index data as another feature of the node, and processes the call delay in the call chain data as a feature of an edge, and inputs the data into a heterogeneous graph neural network; according to the actual needs of the power operation and maintenance system, the abnormality detection unit defines different types of meta paths, wherein the meta path is a path describing the composite relationship between nodes, so as to capture the high-order semantic relationship between different types of data;

[0065] The feature aggregation unit uses a node-level attention mechanism, for each node in the heterogeneous graph neural network, uses a graph attention network to aggregate the features of its neighbor nodes, dynamically models the association strength of the node and its neighbors through a learnable attention coefficient, uses a semantic-level attention mechanism, calculates the importance of the node features on the meta path, maps the feature vectors of all nodes under the path to real values, performs a normalization operation to obtain a weighting coefficient, and assigns different attention weights to the neighbor nodes;

[0066] The weight adjustment unit identifies node features that perform more stably and accurately in past anomaly detection according to historical performance of the node features, such as anomaly detection effect of each node feature under different time windows, and configures a confidence weight for each node feature, wherein the confidence weight reflects the reliability and effectiveness of the node feature in historical tasks, for example, a certain log feature accurately reflects the device state in multiple past anomaly detections, and the feature is given a higher confidence weight. The weight adjustment unit dynamically adjusts the contribution of the node feature in the fusion process according to the confidence weight of the node feature using an adaptive fusion gating method, further adjusts the attention weight allocated by the feature aggregation unit, and generates the final weight value. The weight adjustment unit transmits the weight value to the high-frequency sampling module in real time, and the anomaly detection unit in the high-frequency sampling module performs weighted fusion on the log, index and call chain data according to the weight value.

[0067] Embodiment 3, please refer to Figure 3 and Figure 8 A high-efficiency power operation and maintenance data acquisition system, the high-frequency sampling module is connected with a self-organizing network communication module through a signal, and the self-organizing network communication module is used for stable transmission of power operation and maintenance data; the self-organizing network communication module comprises a communication evaluation unit, a network management unit and a push decision unit; the communication evaluation unit periodically evaluates the performance of the power line carrier and the micro-power wireless communication channel, uses a weighted grey correlation degree analysis method to sort the channel performance, and selects the channel with the best performance for communication; the network management unit monitors the state of each node in the network, constructs a network topology graph, detects node faults and communication interruptions, automatically reconfigures the network topology, and generates a channel switching strategy; the push decision unit performs rule matching on the collected and fused multi-source data, calculates the relevance of the multi-source data to the current operation and maintenance task, sets a push threshold according to a threshold update strategy, and compares the relevance and the push threshold.

[0068] Further, the communication evaluation unit periodically collects communication performance data from the power line carrier and the micro-power wireless communication channel, covering key indicators such as signal strength change rate, network transmission delay and packet loss rate, and uses a weighted grey correlation degree analysis method to comprehensively evaluate the performance indicators of multiple channels, wherein the performance indicators of each channel are used as matrix elements to calculate a weight vector of each performance indicator to reflect the importance of different indicators in comprehensive evaluation, and then a weighted grey correlation degree vector is constructed, the comprehensive ranking of channel performance is obtained by comparing the correlation degrees of each channel and the ideal channel, and based on the results of the weighted grey correlation degree analysis, the communication evaluation unit finally selects the channel with the best performance for communication;

[0069] The network management unit continuously monitors the status of each node in the network of the selected channel by the communication evaluation unit, including the online or offline status of the node, the communication quality, the data transmission rate, etc. Based on the results of the node status monitoring, the network management unit dynamically constructs and updates the network topology graph, which reflects the connection relationship and communication path between each node in the network, including the distribution of master nodes, relay nodes and terminal nodes. When a node failure or communication interruption is detected, the network management unit immediately performs fault detection and diagnosis. Combined with the channel performance ranking results provided by the communication evaluation unit, the network management unit generates a channel switching strategy, automatically reconfigures the network topology, bypasses the failed node by adjusting the position of the relay node, adding a backup path or enabling a backup node, and restores the connectivity and communication quality of the network. When the performance of the current communication channel decreases or fails, a channel with better performance is selected for communication according to the switching strategy.

[0070] During the data transmission process between the system, module and unit through the communication evaluation unit, the push decision unit combines the network node status of the network management unit. In the case of network stability, the network management unit pushes the operation and maintenance data to the operation and maintenance personnel. During the pushing process, the communication evaluation unit formulates rules in the rule library based on the actual needs and business logic of power operation and maintenance, and matches the multi-source data delivered by the high-frequency sampling module one by one, locates the operation and maintenance tasks and business scenarios corresponding to the data, calculates the relevance of multi-source data and current operation and maintenance tasks according to the timeliness, accuracy, importance and direct impact degree of the data on the operation and maintenance tasks, sets the push threshold value by using the threshold update strategy, for example, for the urgent equipment failure warning task, the push threshold value is relatively low, so as to ensure that the relevant data can be pushed in time; for non-urgent statistical analysis tasks, the push threshold value is set higher; after setting the push threshold value in the push decision unit, the calculated relevance is compared with the push threshold value, if the relevance is higher than the push threshold value, it means that the data has high importance to the current operation and maintenance task and needs to be pushed; if the relevance is lower than the push threshold value, it is considered that the importance of the data to the current operation and maintenance task is low, and it is temporarily not pushed, so that the operation and maintenance personnel can obtain relevant information in time and process it, avoiding the operation and maintenance personnel receiving a large amount of irrelevant information, improving the efficiency and accuracy of operation and maintenance work.

[0071] Embodiment 4, please refer to Figure 1 , Figure 4 and Figure 8The high-frequency sampling module comprises a data sampling unit, an abnormality detection unit and a dynamic adjustment unit; the high-frequency sampling module is connected with an abnormality evaluation module through a signal, and the abnormality evaluation module is used for dynamically allocating an abnormality level; the abnormality evaluation module comprises an intelligent word segmentation unit, an entity recognition unit and a level evaluation unit; the intelligent word segmentation unit identifies text information of multi-source fusion data, performs intelligent word segmentation on the text describing the equipment abnormality in the power operation and maintenance environment by using a deep learning model, and divides the word boundary; the entity recognition unit adopts a named entity recognition technology, combines an entity type library, automatically identifies and labels key entities in the abnormality text, and constructs the correlation between the key entities; the level evaluation unit extracts features for level evaluation in the abnormality text, sets initial weights of the features, identifies influences of time, equipment type and operation environment on the feature weights, dynamically adjusts the feature weights, and performs level evaluation on the abnormality by using a deep neural network.

[0072] Further, the intelligent word segmentation unit receives multi-source fusion data and a root cause hypothesis list from the abnormality detection unit, contains a large amount of text information describing equipment abnormality, loads a pre-trained deep learning model, performs intelligent word segmentation on the text by using the deep learning model, dynamically divides the word boundary according to the context information, and in the word segmentation process, the intelligent word segmentation unit identifies and labels the start and end positions of each word to form a word sequence;

[0073] The entity recognition unit receives the text data processed by the intelligent word segmentation unit, loads a pre-defined entity type library, the library contains various entity types that may appear in the field of power equipment operation and maintenance, such as equipment name, abnormal type, operation environment, etc., adopts a named entity recognition technology, scans the text sentence by sentence, identifies the entity conforming to the definition in the entity type library, assigns a corresponding label to the identified entity, further analyzes the relationship between the entities, constructs the correlation between the key entities according to the context relationship and business logic between the entities, forms a complete abnormality information network, and when the intelligent word segmentation unit receives the abnormality text “a transformer appears abnormal heat, the temperature reaches 120℃, and the abnormal heat occurs in a transformer area of a certain transformer substation No. 1 on January 1, 2025”, the entity recognition unit identifies “transformer” as an equipment name entity, “abnormal heat” as a fault type entity, “120℃” as a temperature value entity, “January 1, 2025” as a time entity, and “a transformer area of a certain transformer substation No. 1” as a location entity, and by establishing the correlation between the entities, it is understood that a specific equipment has a certain fault at a specific time and location, thereby providing a comprehensive and accurate information basis for subsequent abnormality processing;

[0074] The grade evaluation unit extracts features for grade evaluation from the abnormal text identified and labeled by the entity recognition unit, including the severity of the fault, the impact range, the duration, etc., sets initial weights for the features extracted according to expert experience or historical data, reflects the importance of different features in grade evaluation, considers the influence of factors such as time, device type and operating environment on feature weights, dynamically adjusts the feature weights, for example, certain fault phenomena may be more serious in a specific device type or operating environment, so the weight should be increased accordingly; and realizes accurate identification and evaluation of power operation and maintenance abnormalities through the abnormality evaluation module.

[0075] Embodiment 5, please refer to Figure 1 、 Figure 5 and Figure 8 , an efficient power operation and maintenance data acquisition system, the high-frequency sampling module includes a data sampling unit, an anomaly detection unit and a dynamic adjustment unit, the high-frequency sampling module is connected with a data warehouse module through a signal, and the data warehouse module is used for storage and retrieval of operation and maintenance data; the data warehouse module includes a storage strategy unit, an index optimization unit and a cycle management unit; the storage strategy unit combines the sampling frequency information of the power operation and maintenance data, analyzes the data volume growth trend, predicts the future storage demand of the power operation and maintenance data, and adjusts the storage proportion of the distributed file system and the columnar database; the index optimization unit constructs real-time indexes for the newly added data stream with improved sampling frequency according to the sampling frequency information, generates query logs according to the results of user access and query of power operation and maintenance data, and identifies query frequency; the cycle management unit sets the storage level of the power operation and maintenance data, generates data migration strategies according to the query frequency information and the sampling frequency information of the power operation and maintenance data.

[0076] Further, the storage strategy unit receives the sampling frequency information of the power operation and maintenance data from the high-frequency sampling module, identifies the sampling frequency difference of different data sources, predicts the future data change trend based on historical storage information, evaluates the future storage resource demand according to the prediction result, dynamically adjusts the storage architecture, distributes high sampling frequency data to the columnar database to ensure the real-time and availability of the data, and migrates low sampling frequency to the distributed file system to reduce the cost;

[0077] The index optimization unit identifies high-frequency data streams according to the sampling frequency information of the high-frequency sampling module, triggers a real-time index construction mechanism, and combines the types of data such as current, voltage, device state logs, etc. to mark the index priority and preferentially allocate index resources to the data of key devices; the index optimization unit collects user query behaviors, generates structured logs, and identifies high-frequency query patterns through association rules, such as 60% query combination of "substation ID + time range + current value", and constructs a joint index accordingly. For data items with high query frequency, the index optimization unit increases the depth or breadth of the index to improve query efficiency; and for data items with low query frequency, the index optimization unit reduces the occupied space of the index to optimize storage resources; the index optimization unit automatically cleans up invalid indexes of migrated data according to the data migration strategy of the period management unit;

[0078] The period management unit generates a data migration strategy according to the importance, access frequency and other factors of the power operation and maintenance data. For example, data with high frequency of access and high real-time requirement will be migrated to the hot data layer. The period management unit sets the storage hierarchy and data migration strategy to realize efficient storage and fast access of power operation and maintenance data. Combined with the query frequency and sampling frequency information of the data, the storage location of the data is dynamically adjusted to balance the storage cost and access performance, thereby improving the access speed of the data and reducing the storage cost. It provides a strong guarantee for the long-term preservation and efficient use of power operation and maintenance data.

[0079] Embodiment 6, please refer to Figure 4 、 Figure 6 、 Figure 7 and Figure 8 , an efficient power operation and maintenance data acquisition system, the abnormality evaluation module comprises: an intelligent word segmentation unit, an entity recognition unit and a level evaluation unit, the abnormality evaluation module is connected with a data enhancement module through a signal, and the data enhancement module is used for generating high-quality enhanced samples from abnormal texts; data enhancement includes: a text enhancement unit and a quality control unit; the text enhancement unit sorts the importance description in the abnormal text, adopts a masking strategy, reconstructs the masked abnormal text, and generates enhanced samples; the quality control unit calculates the difference between the enhanced samples and the abnormal text, selects enhanced samples with low repetition, integrates the enhanced samples and the abnormal text, and generates an expanded data set.

[0080] Further, the text enhancement unit receives the abnormal text processed by the intelligent word segmentation unit of the anomaly evaluation module, masks the top 30% to 50% high-weight fields according to the text length and field density, reconstructs the masked text using a pre-trained power field GPT model, restores the content of the masked part, maintains the overall semantic consistency of the text, and generates an enhanced sample; the quality control unit obtains the enhanced sample generated by the text enhancement unit and the abnormal text processed by the intelligent word segmentation unit, calculates the difference between the enhanced sample and the abnormal text using a cosine distance measurement method, and after completing the difference calculation, the quality control unit selects the enhanced sample with low repetition according to the preset threshold, thereby ensuring the diversity of the data set, and then outputs the diversified enhanced text. The expanded abnormal sample data set screened by the quality control unit enables the anomaly evaluation module to make more accurate judgments when evaluating the actual operation and maintenance of abnormal conditions, thereby improving the efficiency and reliability of the entire power operation and maintenance data collection system.

[0081] Working principle: the high-frequency sampling module sets the corresponding sampling frequency according to the characteristics and operation and maintenance requirements of the equipment, the weight adjustment module fuses the multi-source power operation and maintenance data collected by the high-frequency sampling module, the high-frequency sampling module performs anomaly detection on the fused multi-source data, and dynamically adjusts the sampling frequency of the abnormal equipment according to the real-time abnormal information of the equipment;

[0082] When the high-frequency sampling module detects abnormal information, the anomaly evaluation module and the data enhancement module identify the abnormal text information in the power operation and maintenance data, evaluate the power operation and maintenance anomaly, identify the abnormal level, and synchronize the analyzed abnormal level to the high-frequency sampling module in real time;

[0083] When the modules and units in the system transmit data, stable communication is performed through the communication channel selected by the ad hoc network communication module, and important operation and maintenance data is pushed to the operation and maintenance personnel in a timely manner through the ad hoc network communication module. The data warehouse module generates an index and generates a migration strategy according to the query frequency of the operation and maintenance personnel.

[0084] It is apparent to those skilled in the art that the present application is not limited to the details of the foregoing exemplary embodiments, but can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and not restrictive, and the scope of the present application is defined by the appended claims rather than the foregoing description, and therefore all changes falling within the meaning and range of equivalents of the elements of the claims are intended to be embraced therein. Any reference signs in the claims should not be considered as limiting the claims involved.

Claims

1. A high-efficiency power operation and maintenance data acquisition system, characterized in that: The high-frequency sampling module is used for dynamically adjusting a sampling frequency, and acquiring key data in a power operation and maintenance process. The high-frequency sampling module comprises a data sampling unit, an anomaly detection unit and a dynamic adjustment unit, the data sampling unit is connected with the anomaly detection unit and the dynamic adjustment unit through signals, and the anomaly detection unit is connected with the dynamic adjustment unit through signals. The data sampling unit sets different sampling requirements according to the characteristics and operation and maintenance requirements of various key devices in the power operation and maintenance environment, performs high-frequency power operation and maintenance data sampling, and collects log data, index data and call chain data of the device. The anomaly detection unit performs multi-source data fusion on the log, index and call chain data, calculates an anomaly score through reconstruction loss and adversarial loss, constructs a fault propagation model, locates a core anomaly node, and outputs a root cause hypothesis list. The dynamic adjustment unit identifies the requirement of adjusting the sampling frequency according to the root cause hypothesis list, generates an adjustment strategy, and records the time of each sampling frequency adjustment and the sampling frequency before and after the adjustment.

2. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The high-frequency sampling module is connected with a weight adjustment module through signals, and the weight adjustment module is used for dynamically adjusting the weight of the log, index and call chain data in the fusion process. The weight adjustment module comprises a network construction unit, a feature aggregation unit and a weight adjustment unit, the network construction unit is connected with the feature aggregation unit and the data sampling unit through signals, the feature aggregation unit is connected with the weight adjustment unit through signals, and the weight adjustment unit is connected with the anomaly detection unit through signals. The network construction unit inputs the log, index and call chain as node features into a heterogeneous graph neural network, defines different types of meta-paths, captures high-order semantic relationships, and obtains complex dependencies between cross-modal data. The feature aggregation unit aggregates node features by using an attention mechanism, dynamically identifies the association strength between nodes and neighbor nodes, assigns attention weights to different node features, and preliminarily identifies the relative importance of different node features in the fusion process. The weight adjustment unit assigns a confidence weight to the node features based on the historical performance of the node features, adjusts the attention weights of different node features by using an adaptive fusion gating method, and generates a weight value.

3. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The high-frequency sampling module is connected with a self-organizing network communication module through signals, and the self-organizing network communication module is used for stably transmitting power operation and maintenance data. The self-organizing network communication module comprises a communication evaluation unit, a network management unit and a push decision unit, the communication evaluation unit is connected with the network management unit and the push decision unit through signals, and the network management unit is connected with the push decision unit through signals. The communication evaluation unit regularly evaluates the performance of power line carrier and micro-power wireless communication channels, sorts the channel performance by using a weighted grey correlation degree analysis method, and selects the channel with the optimal performance for communication. The network management unit monitors the state of each node in the network, constructs a network topology graph, detects node faults and communication interruptions, automatically reconfigures the network topology, and generates a channel switching strategy. The push decision unit performs rule matching on the collected and fused multi-source data, calculates the relevance of the multi-source data to the current operation and maintenance task, sets a push threshold according to a threshold updating strategy, and compares the relevance with the push threshold.

4. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The high-frequency sampling module is connected with an abnormality evaluation module through a signal, and the abnormality evaluation module is used for dynamically allocating abnormality levels. The abnormality evaluation module comprises an intelligent word segmentation unit, an entity recognition unit and a level evaluation unit, the intelligent word segmentation unit is connected with the entity recognition unit through a signal, the entity recognition unit is connected with the level evaluation unit through a signal, and the level evaluation unit is connected with an abnormality detection unit through a signal. The intelligent word segmentation unit identifies text information of the multi-source fusion data, intelligently segments words of the text describing device abnormalities in the power operation and maintenance environment by using a deep learning model, and divides word boundaries. The entity recognition unit adopts a named entity recognition technology, automatically recognizes and labels key entities in the abnormal text in combination with an entity type library, and constructs a correlation relationship between the key entities. The level evaluation unit extracts features for level evaluation in the abnormal text, sets initial weights of the features, identifies influences of time, device types and operation environments on the feature weights, dynamically adjusts the feature weights, and performs level evaluation on the abnormality by using a deep neural network.

5. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The high-frequency sampling module is connected with a data warehouse module through a signal, and the data warehouse module is used for storage and retrieval of operation and maintenance data. The data warehouse module comprises a storage strategy unit, an index optimization unit and a period management unit, the storage strategy unit is connected with the index optimization unit through a signal, and the index optimization unit is connected with the period management unit through a signal. The storage strategy unit analyzes a data volume growth trend in combination with sampling frequency information of the power operation and maintenance data, predicts storage requirements of future power operation and maintenance data, and adjusts a storage proportion of a distributed file system and a columnar database. The index optimization unit constructs real-time indexes for newly added data streams with improved sampling frequencies according to the sampling frequency information, generates query logs according to results of user access and query of the power operation and maintenance data, and identifies query frequencies. The period management unit sets storage levels of the power operation and maintenance data, and generates a data migration strategy according to query frequency information and sampling frequency information of the power operation and maintenance data.

6. The high-efficiency electric power operation and maintenance data acquisition system according to claim 4, characterized in that: The abnormality evaluation module is connected with a data enhancement module through a signal, and the data enhancement module is used for generating high-quality enhanced samples from the abnormal text. The data enhancement module comprises a text enhancement unit and a quality control unit, the text enhancement unit is connected with the quality control unit through a signal, and the quality control unit is connected with the entity recognition unit through a signal. The text enhancement unit sorts importance descriptions in the abnormal text, adopts a masking strategy, reconstructs the abnormal text after masking, and generates enhanced samples. The quality control unit calculates differences between the enhanced samples and the abnormal text, selects enhanced samples with low repetition degrees, integrates the enhanced samples and the abnormal text, and generates an expanded data set.

7. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The sampling requirements in the data sampling unit include: Device type characteristic sampling requirements: key devices prone to faults in power operation and maintenance are sampled in priority, and non-key devices are sampled regularly. Sampling requirements of operation and maintenance needs: set high sampling frequency for key data that needs real-time monitoring; Sampling requirements of environmental factor characteristics: set high sampling frequency under harsh environmental conditions; Sampling requirements of historical data: set high sampling frequency for devices that have frequently appeared abnormities in history.

8. The high-efficiency electric power operation and maintenance data acquisition system according to claim 1, characterized in that: The adjustment strategy in the dynamic adjustment unit is to increase the sampling frequency of the device that detects the abnormality.

9. The high-efficiency electric power operation and maintenance data acquisition system according to claim 4, characterized in that: The key entities in the entity recognition unit are entities related to device abnormality, including device name, abnormal type, running environment, and abnormal location.

10. The high-efficiency electric power operation and maintenance data acquisition system of claim 5, wherein: The data migration strategy in the cycle management unit is to divide power operation and maintenance data into high-frequency, medium-frequency, and low-frequency data according to their sampling and query frequency, and migrate them to hot data layer, warm data layer, and cold data layer respectively.

Images