Power transmission line fault prediction method and related device

Through the combination of approximate symbol aggregation algorithm and DeepAR model, the problem of low accuracy in prediction of faults in transmission lines is solved, accurate warning and management of potential faults during transmission lines is achieved, and the safety and stability of the power system is improved.

CN118940913APending Publication Date: 2024-11-12ELECTRIC POWER RES INST CHINA SOUTHERN POWER GRID CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411150583.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prediction accuracy of existing transmission line fault prediction schemes is low, and it is impossible to effectively identify potential faults, which affects the safe and stable operation of the power system.

Method used

The approximate symbol aggregation algorithm is used to reduce the dimensionality of multiple sets of line monitoring data, and abnormal data is identified and eliminated through the target curve similarity, forming a purified data set, and fault prediction is performed based on the DeepAR model.

Benefits of technology

It significantly improves data quality and prediction accuracy, achieves accurate warnings for potential faults during power transmission lines, improves the stability and reliability of power transmission lines, and provides strong guarantees for the safe and stable operation of the power system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940913A_ABST
    Figure CN118940913A_ABST
Patent Text Reader

Abstract

The invention provides a power transmission line fault prediction method and a related device, which can perform dimension reduction processing on large-scale original data by applying an approximate symbolic aggregation algorithm. And then, accurately identifying abnormal data in the original data after dimension reduction by adopting a similarity measurement technology and clearing the abnormal data so as to form a purified large-scale data set, and carrying out fault prediction on the power transmission line based on the purified data. Therefore, invalid data can be removed from a data source to reduce interference of the invalid data on a fault prediction result, and the utilization degree of the original data is guaranteed to the maximum extent, so that the data quality and the accuracy of subsequent analysis can be remarkably improved, the limitations of proper preprocessing and low prediction precision of the original data in the prior art are overcome, and the prediction accuracy of the fault prediction result is improved. Therefore, accurate early warning and management of potential faults in the operation process of the power transmission line can be realized, the operation stability and reliability of the power transmission line are improved, a powerful guarantee is provided for safe and stable operation of a power system, and the prediction accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of power transmission line monitoring, and in particular to a power transmission line fault prediction method and related devices. Background Art

[0002] As the core component of the power system, the operating status of power transmission and transformation equipment is directly related to the safety and stability of the power system. Since transmission lines are exposed to the natural environment for a long time, they need to face harsh operating conditions, such as strong winds, heavy rains, high temperatures, severe cold and other extreme climate tests. Against the background of increasingly complex and severe global climate change, the problems of corrosion, oxidation, deformation, loosening and so on faced by transmission lines are becoming more and more prominent. Accurate early warning of power transmission and transformation equipment operating failures, timely prediction and identification of potential failures, and then taking scientific and reasonable operation and maintenance measures have become the key to ensuring the safe and stable operation of power transmission and transformation equipment.

[0003] However, the inventors have found through research that the current power transmission line fault prediction scheme has the problem of low prediction accuracy. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defect of low prediction accuracy in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting a power transmission line fault, comprising:

[0006] Acquire multiple groups of line monitoring data corresponding to the power transmission line to be predicted, wherein the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data;

[0007] Performing standardization processing on the plurality of groups of line monitoring data, and obtaining each group of standardized monitoring data;

[0008] Using an approximate symbol aggregation algorithm, each group of the standardized monitoring data is subjected to dimensionality reduction processing, and multiple groups of N-dimensional dimensionality reduction monitoring data are obtained; wherein N is less than n;

[0009] For each dimensional monitoring data in each group of the reduced dimensionality monitoring data, the target curve similarity corresponding to the dimensional monitoring data is calculated according to the dimensional monitoring data and each group of target standardized data; if the dimensional monitoring data is determined to be abnormal data according to the target curve similarity, the dimensional monitoring data is removed from the reduced dimensionality monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the reduced dimensionality monitoring data to which the dimensional monitoring data belongs;

[0010] According to each group of reduced-dimensional monitoring data obtained after elimination, fault prediction is performed on the transmission line to be predicted, and a prediction result is obtained.

[0011] In one embodiment, for each dimension of monitoring data in each group of reduced-dimensional monitoring data, the target curve similarity corresponding to the dimension of monitoring data is calculated according to the dimension of monitoring data and each group of target standardized data, including:

[0012] For each dimensional monitoring data in each of the reduced dimensional monitoring data, the initial curve similarity between the dimensional monitoring data and each group of the target standardized data is calculated respectively, and the average calculation is performed on each of the initial curve similarities to obtain the target curve similarity corresponding to the dimensional monitoring data.

[0013] In one embodiment, respectively calculating the initial curve similarity between the dimensional monitoring data and each group of the target standardized data includes:

[0014] Based on the following expressions, the initial curve similarity between the dimension monitoring data and each group of the target standardized data is calculated respectively:

[0015]

[0016] In the formula, Monitor data for this dimension The initial curve similarity with a set of target standardized data Q, is the monitoring data of the i-th dimension in Q.

[0017] In one embodiment, for each dimension of monitoring data in each group of the reduced-dimensionality monitoring data, if the dimension monitoring data is determined to be abnormal data according to the target curve similarity, then the dimension monitoring data is removed from the reduced-dimensionality monitoring data, including:

[0018] For each dimensional monitoring data in each group of the reduced dimensionality monitoring data, at least one group of similar standardized data is screened out from each group of the target standardized data according to the similarities of each of the initial curves, and based on the dimensional monitoring data, each group of similar standardized data and the similarity of the target curve, the abnormal baseline value corresponding to the dimensional monitoring data is calculated; if the dimensional monitoring data is determined to be abnormal data based on the size relationship between the dimensional monitoring data and the abnormal baseline value, the dimensional monitoring data is eliminated from the reduced dimensionality monitoring data.

[0019] In one embodiment, the calculating of the abnormal reference value corresponding to the dimensional monitoring data based on the dimensional monitoring data, each group of similar standardized data and the target curve similarity includes:

[0020] Calculate the abnormal baseline value corresponding to the monitoring data of this dimension based on the following expression:

[0021]

[0022] In the formula, is the abnormal reference value corresponding to the monitoring data of this dimension, A is the similar standardized data of each group, Monitoring data for this dimension, is the target curve similarity.

[0023] In one embodiment, the fault prediction of the transmission line to be predicted is performed according to each group of reduced-dimensional monitoring data obtained after elimination, and a prediction result is obtained, including:

[0024] For each group of reduced-dimensionality monitoring data after elimination, if there are discontinuous point missing data in the group of reduced-dimensionality monitoring data, the discontinuous point missing data are filled based on the Hermite interpolation method;

[0025] For each group of reduced-dimensionality monitoring data after elimination, if there are missing data of continuous points in the group of reduced-dimensionality monitoring data, the missing data of continuous points are filled based on the piecewise cubic Hermite interpolation method;

[0026] According to each group of reduced-dimensional monitoring data obtained after filling, fault prediction is performed on the transmission line to be predicted, and the prediction result is obtained.

[0027] In one embodiment, the fault prediction of the transmission line to be predicted is performed according to each group of reduced-dimensional monitoring data obtained after filling, and the prediction result is obtained, including:

[0028] Each group of reduced-dimensional monitoring data obtained after filling is input into the pre-trained fault prediction model, and the prediction result output by the fault prediction model is obtained; wherein the fault prediction model is obtained based on the DeepAR model.

[0029] In one embodiment, the prediction result includes predicted fault type, predicted risk probability and priority processing suggestion; the method further includes:

[0030] According to the location information corresponding to the transmission line to be predicted, the predicted fault type, the predicted risk probability and the priority processing suggestion are displayed in a visual map.

[0031] In one embodiment, the plurality of groups of line monitoring data include line engineering data and line environment data.

[0032] In a second aspect, an embodiment of the present application provides a transmission line fault prediction device, comprising:

[0033] A line monitoring data acquisition module is used to acquire multiple groups of line monitoring data corresponding to the power transmission line to be predicted, wherein the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data;

[0034] A standardization processing module, used for performing standardization processing on the plurality of groups of line monitoring data and obtaining each group of standardized monitoring data;

[0035] A dimension reduction module, used to use an approximate symbol aggregation algorithm to perform dimension reduction processing on each group of the standardized monitoring data, and obtain multiple groups of N-dimensional dimension reduction monitoring data; wherein N is less than n;

[0036] A similarity measurement module is used to calculate the target curve similarity corresponding to each dimensional monitoring data in each group of the reduced dimensionality monitoring data according to the dimensional monitoring data and each group of target standardized data; if the dimensional monitoring data is determined to be abnormal data according to the target curve similarity, the dimensional monitoring data is removed from the reduced dimensionality monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the reduced dimensionality monitoring data to which the dimensional monitoring data belongs;

[0037] The fault prediction module is used to perform fault prediction on the transmission line to be predicted based on each group of reduced-dimensional monitoring data obtained after elimination, and obtain a prediction result.

[0038] In a third aspect, an embodiment of the present application provides a storage medium, which stores computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the transmission line fault prediction method described in any of the above embodiments.

[0039] In a fourth aspect, an embodiment of the present application provides a computer device, the computer device comprising: one or more processors, and a memory;

[0040] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the power transmission line fault prediction method described in any of the above embodiments are executed.

[0041] In the transmission line fault prediction method and related device provided in some embodiments of the present application, an approximate symbol aggregation algorithm can be used to reduce the dimension of large-scale original data (i.e., multiple groups of n-dimensional time series data). Then, a similarity measurement technology is used to accurately identify and remove abnormal data in the original data after dimensionality reduction, thereby forming a purified large-scale data set, and the fault prediction of the transmission line is performed based on each group of purified dimensionality reduction monitoring data. In this way, invalid data can be removed from the source of the data to reduce the interference of invalid data on the fault prediction results, and the utilization of the original data can be maximized, thereby significantly improving the data quality and the accuracy of subsequent analysis, overcoming the limitations of appropriate pre-processing of original data and low prediction accuracy in the prior art, and then realizing accurate early warning and management of potential faults in the operation of the transmission line, improving the stability and reliability of the operation of the transmission line, providing a strong guarantee for the safe and stable operation of the power system, and improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 One of the system architecture diagrams of a power transmission line fault prediction and management system in one embodiment;

[0044] Figure 2 This is a second system architecture diagram of a power transmission line fault prediction and management system in one embodiment;

[0045] Figure 3 A schematic diagram of a flow chart of a method for predicting a power transmission line fault in one embodiment;

[0046] Figure 4 A schematic diagram of a processing flow of multi-source data cleaning and “dirty” data repair in one embodiment;

[0047] Figure 5 A schematic diagram of a transmission line fault prediction framework in one embodiment;

[0048] Figure 6 A schematic diagram of the structure of a fault prediction model in one embodiment;

[0049] Figure 7 A schematic diagram of a visualization interface in one embodiment;

[0050] Figure 8is a schematic structural diagram of a power transmission line fault prediction device in one embodiment;

[0051] Fig. 9 FIG. 1 is a diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In one embodiment, the transmission line fault prediction method provided by the present application can be applied to Figure 1 The system adopts a hierarchical architecture and builds a multi-level data processing and analysis system to achieve forward-looking early warning and status monitoring management of the transmission line status. Through efficient data integration and artificial intelligence analysis, accurate assessment and early warning of transmission line operation risks can be achieved.

[0054] Among them, the perception module can integrate the collection of multi-source data, which can be used to obtain real-time environmental data of the environment in which the transmission line is located and real-time engineering data of the power system to which the transmission line belongs. This fusion of multi-source data can provide a comprehensive and accurate information basis for subsequent early warning analysis, so as to more comprehensively and accurately reflect the actual operating status of the transmission line.

[0055] Among them, real-time environmental data may include any one or any combination of key climate indicator data such as temperature data, humidity data, wind speed data, wind direction data, etc. Real-time engineering data may include any one or any combination of high-definition images of transmission lines, equipment surface temperature distribution data, tension data, etc. In one example, real-time environmental data may be provided by a network data platform.

[0056] Due to the large number of power equipment and their dispersed geographical locations, the preprocessing module can use advanced data cleaning and repair technology to preprocess the collected climate, tension and other data, improve data quality, reduce terminal processing pressure, and improve prediction accuracy. In some embodiments, the preprocessing module can clean, repair and standardize the collected data to eliminate noise and outliers, thereby improving data quality and prediction accuracy. For example, the data set can first be purified through steps such as missing value processing, outlier detection, deduplication and noise filtering. Subsequently, the data structure is optimized through transformation methods such as standardization / normalization, categorical variable encoding, feature construction and dimensionality reduction to adapt to the analysis algorithm.

[0057] The data transmission module can use efficient and reliable communication protocols to transmit pre-processed data to the big data cloud in real time for centralized processing and analysis. This ensures the real-time nature of data and the efficiency of processing, providing support for rapid response to potential failures.

[0058] The artificial intelligence analysis module can be set up in the data processing center. It can use artificial intelligence algorithms to deeply analyze and fuse multi-source data, and deeply mine and analyze the fused perception data, images, text and other information to build a comprehensive risk warning model for power transmission equipment. By building complex mathematical models and algorithms, the artificial intelligence analysis module can accurately predict and warn of possible fault risks in transmission lines, issue suggestions for concurrent processing of multiple fault points, and realize real-time monitoring and evaluation of the operating status of transmission lines, thereby providing strong guarantees for the safe and stable operation of the power system.

[0059] The terminal interactive module can adopt a visual interactive display mode to accurately display the cloud analysis, evaluation and warning results in a layered map-like manner, including the status risk warning prompts of the transmission line and its various status real-time monitoring pictures and data. In this way, it can provide an intuitive and reliable way to globally display and locate potential risk points of the transmission line.

[0060] Through this transmission line status assessment and early warning method based on an integrated artificial intelligence algorithm, a multi-level and refined data processing and analysis system is constructed to achieve forward-looking early warning and management of the transmission line status. This can not only improve the safety and stability of power system operation, but also provide strong support for the sustainable development of the power industry.

[0061] In one embodiment, the transmission line fault prediction method provided by the present application can be applied to Figure 2 The transmission line fault prediction and management system shown in the figure includes a state perception terminal, a data integration module, a data server, a cloud platform, a terminal interaction module, a remote terminal and a host computer.

[0062] Among them, the state perception terminal is used to obtain various characteristic information that can be collected during the operation of the transmission line. The data integration module can collect and summarize the publicly available meteorological data in real time through data integration classification technology, and match the meteorological data with the state characteristics of the transmission line at the corresponding time and space scale. The state perception device and the integrated collected meteorological data are fused and corresponding, packaged and classified, and sent to the data server.

[0063] The data server records and stores all the operating status information related to the transmission line. The cloud platform connects the state perception terminal and the data server through the network and provides an access interface to the outside world. The remote terminal can access the cloud platform to establish a state warning visualization. The host computer is connected to the state perception terminal device by signal for local control. The information exchange between the host computer and the data server provides a backup communication channel for the state perception terminal device.

[0064] In one example, the state sensing module may include an image sensing device, a tension sensing device, and a temperature sensing device. The image sensing device may include a high-definition camera, a communication module, and a battery module. The image sensing device is used to obtain high-definition images of the operation of the transmission line and upload them to a data server in real time. Furthermore, the image sensing device may also include a fill light, which may be set close to the image acquisition camera. The fill light is started synchronously with the high-definition camera, so as to provide a shooting environment in complex environments such as weak light and backlight, thereby supplementing the lighting environment for the high-definition camera to ensure the clarity of the image captured by the high-definition camera.

[0065] The tension sensing device may include a tension sensor, a communication module and a battery module, and the tension sensing device is used to obtain the load changes during the operation of the transmission line and upload them to the data server in real time. Furthermore, the tension sensor can be installed at a fixed point of the transmission line, and the tension on the conductor or ground wire is converted into an electrical signal using the principle of a strain gauge.

[0066] The temperature sensing device may include an infrared camera / temperature sensor, a communication module and a battery module, and the temperature sensing device is used to measure the temperature information of key parts of the transmission line and upload it to the data server in real time. Furthermore, the temperature sensor can use the principles of thermocouples, thermistors or infrared technology to convert the temperature changes on the surface of the line into electrical signals, which are then converted into temperature readings.

[0067] Furthermore, the state perception device may also include a flash memory module for local storage and backup of data. The communication module described herein may include a WIFI module, a cellular network, a Bluetooth module and / or an NFC (Near Field Communication) chip. The state perception terminal device uses a WIFI module and a cellular network to complete communication with the cloud platform, and the state perception device uses a Bluetooth module and an NFC chip to complete communication with the host computer; the state perception device camera obtains the real-time operation status image of the transmission line, and transmits the image to the server background through the built-in 5G communication.

[0068] The state-aware device can use the standard Android system and perform 5G communication through the inserted SIM (Subscriber Identity Module) card. When the 5G signal is poor, it can automatically switch to 4G network communication to ensure efficient and uninterrupted remote communication.

[0069] The present invention monitors the operating status of the transmission line in real time through the state sensing device, stores the engineering status information through the data server, uploads the data information in real time through the communication module, integrates the climate data through the cloud platform to trace and compare the key information and give fault warnings, thus realizing the digitization, precision and standardization of the whole process of real-time sensing and warning of the transmission line status.

[0070] In one embodiment, Figure 3 As shown, the present application provides a method for predicting transmission line faults, comprising the following steps:

[0071] S302: Acquire multiple groups of line monitoring data corresponding to the power transmission line to be predicted, where the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data.

[0072] The line monitoring data refers to the monitoring data of the transmission line to be predicted, which can be used to reflect the operating parameters of the transmission line to be predicted and / or the environmental parameters of the environment in which the transmission line to be operated is located. In this step, the line monitoring data of the transmission line to be predicted under multiple different monitoring indicators can be obtained, and multiple groups of line monitoring data can be obtained. For example, the multiple groups of line monitoring data may include line temperature data, line tension data, line voltage data, and line current data of the transmission line to be predicted.

[0073] In one example, multiple groups of line monitoring data may include line engineering data of the transmission line to be predicted and line environment data of the environment in which the transmission line to be predicted is located. Among them, the line environment data can gather global real-time meteorological data and can be focused on local areas, including but not limited to environmental temperature data, environmental humidity data, wind speed data, ultraviolet intensity data, rainfall data and other characteristic data that may affect the operating status of the transmission line. The line engineering data can be the line operating parameters of the transmission line to be predicted, including but not limited to key data such as line temperature data, line tension data, line voltage data, and line current data. In this example, multi-source data can be integrated to predict the fault of the transmission line, and the multi-source data includes but is not limited to environmental parameters, equipment status information, and real-time monitoring data. By comprehensively analyzing multi-source data, the problem of a single data source in traditional prediction methods can be significantly overcome, and the possible operating hazards and accurate probabilities of overhead transmission lines can be discovered and predicted in a timely manner, thereby realizing comprehensive and accurate monitoring of the operating status of overhead transmission lines. In this way, this application can be based on the perception fusion data, starting from the overall and focus data, and layered and line evaluation to predict the overall and local individual operating status and failure probability. The richness and comprehensiveness of the data broadens the normalization and scalability of the original prediction data, and protects the accuracy of the prediction from the root.

[0074] S304: performing standardization processing on multiple groups of line monitoring data, and obtaining each group of standardized monitoring data.

[0075] In this step, each group of line monitoring data can be standardized to eliminate the influence of quantitative data of different dimensions on subsequent analysis. After the standardization, each group of standardized monitoring data can be obtained, and each group of standardized monitoring data is n-dimensional data.

[0076] Take the dimension reduction of a set of standardized monitoring data X as an example. , where x1 represents the first dimension data in the standardized monitoring data X, x2 represents the second dimension data in the standardized monitoring data X, and x n Represents the nth dimension data in the standardized monitoring data X. The average value is 0 and the standard deviation is 1, and each group of standardized monitoring data is processed uniformly. The unified monitoring data obtained after the unified processing of X is adopted Indicates, where c1 represents the first dimension data in the standardized monitoring data C, c2 represents the second dimension data in the standardized monitoring data C, and c n Represents the n-th dimension data in the standardized monitoring data C. and Represent the mean and standard deviation of X respectively. The following expression can be used to represent the i-th dimension element in C:

[0077]

[0078] S306: using an approximate symbol aggregation algorithm to perform dimensionality reduction processing on each group of standardized monitoring data, and obtain multiple groups of N-dimensional reduced-dimensionality monitoring data; wherein N is less than n.

[0079] In this step, the approximate symbol aggregation algorithm can be applied to reduce the dimension of each group of n-dimensional standardized monitoring data to N dimensions, and obtain each group of reduced-dimensional monitoring data. In order to process the time series data C, the dimension reduction method can be adopted to reduce the dimension of the time series data from n to N. Represents the N-dimensional time series data after dimensionality reduction. 1 / t is used to represent the interval length of each segment, where Indicates the interval compression ratio of each segment. It represents the mean of the i-th segment among the N segments of the standardized monitoring data, which can be expressed as:

[0080]

[0081] Through the above steps, each group of n-dimensional standardized monitoring data can be reduced to N dimensions, and each group of N-dimensional reduced-dimensional monitoring data can be obtained.

[0082] S308: For each dimensional monitoring data in each group of dimensionality reduction monitoring data, calculate the target curve similarity corresponding to the dimensional monitoring data based on the dimensional monitoring data and each group of target standardized data. If the dimensional monitoring data is determined to be abnormal data based on the target curve similarity, the dimensional monitoring data is removed from the dimensionality reduction monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the dimensionality reduction monitoring data to which the dimensional monitoring data belongs.

[0083] In this step, considering that the monitoring data may lose some important information of the original data during the dimensionality reduction process, thereby affecting the calculation results of the similarity, therefore, in order to improve the accuracy of the similarity assessment, the present application can compare and correct the dimensionality reduction data with the original data before dimensionality reduction, and accurately classify the dimensionality reduction data accordingly, so as to identify normal state data and abnormal data in the N-dimensional dimensionality reduction data, and extract the abnormal data from the source of the data, so as to eliminate the interference of abnormal data on fault prediction, and maximize the utilization of the original data. Among them, abnormal data can include abnormal state data and "dirty" data. Exemplarily, abnormal data includes but is not limited to deviations, omissions and redundant data.

[0084] Specifically, for each set of reduced dimension monitoring data Each dimension of monitoring data This application can be used for Execute steps A1 to A5:

[0085] Step A1: According to and each group of target standardized data, respectively, The corresponding target curve similarity. For example, step S302 obtains 3 groups of line monitoring data. After processing in steps 304 to 306, 3 groups of standardized monitoring data and 3 groups of reduced-dimensionality monitoring data can be obtained. Among them, the first group of reduced-dimensionality monitoring data is obtained by reducing the dimension of the first group of standardized monitoring data, the second group of reduced-dimensionality monitoring data is obtained by reducing the dimension of the second group of standardized monitoring data, and the third group of reduced-dimensionality monitoring data is obtained by reducing the dimension of the third group of standardized monitoring data. If is a dimension data in the first set of dimension reduction monitoring data, then each set of target standardized data can be the second set of standardized data and the third set of standardized data. Similarly, if is a dimension data in the second set of reduced dimension monitoring data, then each set of target standardized data can be the first set of standardized data and the third set of standardized data. is a dimension data in the third set of reduced dimension monitoring data, then each set of target standardized data can be the first set of standardized data and the second set of standardized data. and corresponds to The target standardized data of each group is calculated The corresponding target curve similarity.

[0086] Step A3: According to The corresponding target curve similarity is judged Whether it is abnormal data;

[0087] Step A5: If If it is abnormal data, from Remove from the list, otherwise, keep .

[0088] S310: According to each group of reduced-dimensional monitoring data obtained after elimination, fault prediction is performed on the predicted transmission line, and a prediction result is obtained.

[0089] In this step, after the abnormal data are removed from each group of reduced-dimensional monitoring data, the fault prediction of the transmission line can be performed based on each group of reduced-dimensional monitoring data obtained after the abnormal data are removed, and the prediction result can be obtained.

[0090] In this application, the approximate symbol aggregation algorithm can be used to reduce the dimension of large-scale original data (i.e., multiple groups of n-dimensional time series data). Then, the similarity measurement technology is used to accurately identify and remove abnormal data in the original data after dimensionality reduction, thereby forming a purified large-scale data set, and the fault prediction of the transmission line is performed based on each group of purified dimensionality reduction monitoring data. In this way, invalid data can be removed from the source of the data to reduce the interference of invalid data on the fault prediction results, and maximize the utilization of the original data, thereby significantly improving the data quality and the accuracy of subsequent analysis, overcoming the limitations of appropriate preprocessing of original data and low prediction accuracy in the prior art, and then realizing accurate early warning and management of potential faults in the operation of the transmission line, improving the stability and reliability of the operation of the transmission line, and providing strong guarantees for the safe and stable operation of the power system.

[0091] In one embodiment, for each dimension of monitoring data in each set of reduced dimension monitoring data, the target curve similarity corresponding to the dimension monitoring data is calculated according to the dimension monitoring data and each set of target standardized data, including:

[0092] For each dimension of monitoring data in each dimensionality-reduced monitoring data, the initial curve similarity between the dimensional monitoring data and each group of target standardized data is calculated respectively, and the average calculation of each initial curve similarity is performed to obtain the target curve similarity corresponding to the dimensional monitoring data.

[0093] In this embodiment, for each set of dimensionality reduction monitoring data Each dimension of monitoring data , this application can be calculated separately The initial curve similarity between each set of target standardized data is calculated, and the average value of the initial curve similarity is obtained. The corresponding target curve similarity. In this way, the accuracy of abnormal data identification can be further improved to further improve the prediction accuracy.

[0094] In one embodiment, respectively calculating the initial curve similarity between the dimensional monitoring data and each set of target standardized data includes: respectively calculating the initial curve similarity between the dimensional monitoring data and each set of target standardized data based on the following expression:

[0095]

[0096] In the formula, Monitor data for this dimension The initial curve similarity with a set of target standardized data Q, is the monitoring data of the i-th dimension in Q.

[0097] In this way, the target standardized data Q and the monitoring data in the reduced dimension monitoring data can be accurately calculated. The similarity of the curves between them can further improve the accuracy of abnormal data identification and prediction accuracy.

[0098] In one embodiment, for each dimension of monitoring data in each set of reduced-dimensionality monitoring data, if the dimension of monitoring data is determined to be abnormal data according to the target curve similarity, the dimension of monitoring data is removed from the reduced-dimensionality monitoring data, including:

[0099] For each dimensional monitoring data in each group of reduced dimensionality monitoring data, at least one group of similar standardized data is screened out from each group of target standardized data according to the similarity of each initial curve, and based on the dimensional monitoring data, each group of similar standardized data and the similarity of the target curve, the abnormal baseline value corresponding to the dimensional monitoring data is calculated. If the dimensional monitoring data is determined to be abnormal data based on the size relationship between the dimensional monitoring data and the abnormal baseline value, the dimensional monitoring data is eliminated from the reduced dimensionality monitoring data.

[0100] In this embodiment, for each set of dimensionality reduction monitoring data Each dimension of monitoring data , this application can execute steps B1~B7:

[0101] Step B1: According to The corresponding initial curve similarity is At least one group of similar standardized data is selected from each group of target standardized data; wherein similar standardized data refers to The target standardized data with a relatively high similarity, for example, the target standardized data corresponding to the initial curve similarity greater than or equal to a preset similarity threshold, can be used as each group of similar standardized data.

[0102] Step B3: Monitor data based on this dimension , The corresponding groups of similar standardized data and The corresponding target curve similarity is calculated The abnormal baseline value is used to judge Whether it is the threshold of abnormal data.

[0103] Step B5: Compare and The size relationship between the corresponding abnormal benchmark values, and judge based on the size relationship Is it abnormal data? For example, if Greater than The corresponding abnormal baseline value can be determined is abnormal data, otherwise, it can be determined It is normal data.

[0104] Step B7: If If it is abnormal data, from Remove from the list, otherwise, keep .

[0105] Furthermore, in one example, after approximate symbol aggregation calculation and similarity measurement, the present application can use the weighted adjustment method (i.e., the curve fitting method) to calculate w groups of similar standardized data, and thereby obtain a reference curve corresponding to the line monitoring data X. If there are missing values ​​in the line monitoring data X, a reference curve can be obtained through multiple weighted calculations to supplement the data and compare and analyze abnormal data situations.

[0106] In this embodiment, the initial curves can be screened out according to their similarity. Similar standardized data of each group are similar, and the similar standardized data of each group are used for judgment. Whether it is an abnormal baseline value of abnormal data, this can improve the accuracy of abnormal data identification, thereby improving the accuracy of data cleaning and the accuracy of fault prediction.

[0107] In one embodiment, based on the dimensional monitoring data, each group of similar standardized data and the target curve similarity, calculating the abnormal reference value corresponding to the dimensional monitoring data includes:

[0108] Calculate the abnormal baseline value corresponding to the monitoring data of this dimension based on the following expression:

[0109]

[0110] In the formula, is the abnormal baseline value corresponding to the monitoring data of this dimension, A is the similar standardized data of each group, Monitoring data for this dimension, is the target curve similarity.

[0111] In this embodiment, in order to more accurately identify abnormal data, the present application can determine the abnormality of the data midpoint by adjusting the maximum threshold (i.e., the abnormal baseline value). In the formula, A is each group of similar standardized data. The present application can calculate the monitoring data of each dimension of each group of similar standardized data and The maximum value among the differences is .

[0112] In one embodiment, according to each group of reduced-dimensional monitoring data obtained after elimination, fault prediction is performed on the predicted transmission line, and prediction results are obtained, including:

[0113] For each group of reduced-dimensionality monitoring data after elimination, if there are discontinuous point missing data in the group of reduced-dimensionality monitoring data, the discontinuous point missing data are filled based on the Hermite interpolation method;

[0114] For each group of reduced-dimensionality monitoring data after elimination, if there are missing data of continuous points in the group of reduced-dimensionality monitoring data, the missing data of continuous points are filled based on the piecewise cubic Hermite interpolation method;

[0115] According to each group of reduced-dimensional monitoring data obtained after filling, fault prediction is performed on the predicted transmission line and the prediction results are obtained.

[0116] Specifically, after using the above approximate symbol aggregation algorithm to clean the data, the reduced-dimensional data inevitably has data omissions, which damages the integrity of the data, which is not conducive to further analysis and use of the data. Therefore, it is necessary to reconstruct the small missing data in the reduced-dimensional data in time series to complete the data repair.

[0117] For the analysis of missing data, it can be found that it can generally be divided into two cases: discontinuous points and continuous points. For discontinuous points, the Hermite interpolation method can be directly used to construct the interpolation function and fill in the missing data points. For continuous points, in order to further improve the accuracy of data filling, the piecewise cubic Hermite interpolation method is used to construct the interpolation function and fill in the missing data points in sequence.

[0118] Specifically, in the process of using the piecewise cubic Hermite interpolation method for data filling, the typical two-point cubic Hermite interpolation method can be selected according to the Hermite interpolation principle and its application requirements. There is a unique Hermite interpolation polynomial , at the node and The following conditions are met:

[0119]

[0120] In the formula, is the interpolation point, To correspond to The function value of To correspond to The function value of for Corresponds to In this way, the “dirty” data repair based on the piecewise cubic Hermite interpolation technology can avoid data waste, and innovatively fill in the missing data in the original fusion data packet scientifically, correct and improve the data quality to a certain extent, and further improve the prediction accuracy.

[0121] In one example, the data cleaning and “dirty” data repair process for raw data is as follows: Figure 4 As shown, these steps not only enhance the generalization ability and prediction accuracy of the model, improve data processing efficiency, enhance data interpretability, and improve the adaptability of the analysis method and the versatility of the algorithm. This application significantly improves the data quality and the accuracy of subsequent analysis by systematically cleaning the fused data set and repairing "dirty" data.

[0122] In one embodiment, according to each group of reduced-dimensional monitoring data obtained after filling, fault prediction is performed on the predicted transmission line, and prediction results are obtained, including:

[0123] Each group of reduced-dimensional monitoring data obtained after filling is input into the pre-trained fault prediction model, and the prediction result output by the fault prediction model is obtained; wherein the fault prediction model is obtained based on the DeepAR model.

[0124] In this embodiment, the cleaned and repaired data can be subjected to deep cloud mining and learning. In terms of data fusion and intelligent analysis, this application can use artificial intelligence algorithms to deeply mine the intrinsic connections between multi-source data based on the cleaned and repaired fusion data to further improve the accuracy of prediction.

[0125] In this embodiment, the fault prediction model can be obtained based on the DeepAR model. Among them, the fault prediction model based on DeepAR has strong adaptability, and can effectively learn and mine data in both extreme scenarios of excessive data and scarce data. The addition of attention mechanism and dynamic L2 regularization strategy enables it to accurately predict the type of transmission line fault and warn its probability in the form of percentage, which improves the accuracy of transmission line fault perception in a leapfrog manner, breaks through the prediction confusion and prediction failure limitations of conventional models in extreme cases, and ensures the stable operation of the power system at the core level. Based on neural network technology combined with practical factors such as the geographical location of multiple fault points, the importance of the points, and realistic factors such as personnel configuration and traffic conditions in specific scenarios, it intelligently recommends the optimal processing path and plan to assist relevant personnel in preventing and handling hidden dangers of transmission line faults.

[0126] The fault prediction model based on DeepAR can combine the meteorological and engineering data and standard data under historical fault conditions, and adopt a series of related optimization strategies to improve the prediction performance of the model, so as to achieve accurate prediction of the probability of transmission line fault risk. The overall framework structure is as follows: Figure 5As shown in Figure 1. In view of the nonlinear and non-stationary characteristics of transmission line operation data and meteorological data, the nonlinear fitting ability of the DeepAR model is used to construct a transmission line fault probability prediction model based on DeepAR. To further improve the model, the attention mechanism and dynamic L2 regularization strategy are used to achieve high-precision prediction of the transmission line fault risk probability. The structure of the transmission line fault risk probability prediction model based on DeepAR is shown in Figure 1. Figure 6 shown.

[0127] See also Figure 6 , during the fault prediction model training process, the model input represents the transmission line failure risk probability at time t, and It represents the value of the i-th influencing factor at time t. Network is the basic neural unit of DeepAR, which is selected autonomously according to specific needs. At time t, the input of the network is , And the network output at the last moment , and the output , it can be expressed as:

[0128]

[0129] Then the output of the network Used to calculate the likelihood function Parameters In this paper, Gaussian distribution is selected as the likelihood function, and its form is:

[0130]

[0131] In the formula, , denote the expectation and standard deviation respectively, then the model parameters , where expectation It is obtained by the mapping function output by the network, namely:

[0132]

[0133] In the formula, , They represent mapping weight and mapping bias respectively.

[0134] In order to ensure the standard deviation If it is greater than 0, select the softplus activation function to map the network output and get the standard deviation for:

[0135]

[0136] When making predictions, assume that the prediction interval is , and The historical data before the moment is input into the network to obtain the output of the network Then, the sampled values ​​are obtained from the joint distribution represented by Monte Carlo sampling. These sampled values ​​are used as the input of the model at the next moment, and the above process is repeated. Through the calculation of this series of sampled values, the predicted values ​​of various types of transmission line fault risks and corresponding probabilities can be obtained.

[0137] In one embodiment, the prediction result includes the predicted fault type, predicted risk probability and priority processing suggestions. The transmission line fault prediction method of the present application may also include: displaying the predicted fault type, predicted risk probability and priority processing suggestions in a visual map according to the location information corresponding to the transmission line to be predicted.

[0138] This embodiment can display the results in layers in the form of a visual map, and generate intelligent processing solutions in combination with real-life scenarios, thereby effectively avoiding the occurrence of potential faults, improving the stability of power system operation, and providing timely and accurate decision support for power system operation and maintenance personnel. In this way, the limitation of the non-intuitive display results in the prior art can be overcome.

[0139] The terminal interactive platform is based on artificial intelligence and communicates with the data server in real time to build a transmission line status risk visualization management platform. It coordinates the time axis horizontally and the layered focused visualization map vertically. It combines active and passive dynamic display of transmission line fault types, fault risk probabilities, and concurrent processing solutions for multiple fault points. The time axis allows convenient query of past and future status information, and progressively controls the operating status of regional transmission lines from the global and local areas. It assists relevant personnel in more efficiently perceiving the health status of transmission lines, improves the efficiency of problem handling in the power system, and proactively ensures the stable operation of the power system.

[0140] This application can achieve full-process management of fault detection, and realize real-time viewing of fused data on risk-free days; accurate prediction of line faults and probabilities on risky days, and concurrent priority processing recommendations for multiple fault points. Based on the real-time transmission and recording of meteorological and engineering data by the communication module and the big data cloud, operators can view the original line data and corrected data in real time through the terminal interactive platform. In some examples, this application can sort the transmission line faults and risks by time and area based on a priority algorithm, and display them in real time through a visual map. Its framework can be as follows: Figure 7 At the same time, when multiple risks occur concurrently in the same region, the artificial intelligence module can issue priority treatment and operation planning suggestions based on the region’s personnel configuration, risk probability, traffic conditions, climate conditions and other realistic factors to achieve the optimal solution for transmission line risk treatment.

[0141] The following describes a power transmission line fault prediction device provided in an embodiment of the present application. The power transmission line fault prediction device described below and the power transmission line fault prediction method described above can refer to each other.

[0142] In one embodiment, the present application provides a transmission line fault prediction device 800. Figure 8 As shown, the device 800 includes:

[0143] The line monitoring data acquisition module 810 is used to acquire multiple groups of line monitoring data corresponding to the power transmission line to be predicted, wherein the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data;

[0144] A standardization processing module 820, used to perform standardization processing on the multiple groups of line monitoring data and obtain each group of standardized monitoring data;

[0145] A dimension reduction module 830 is used to use an approximate symbol aggregation algorithm to perform dimension reduction processing on each group of the standardized monitoring data, and obtain multiple groups of N-dimensional reduced dimension monitoring data; wherein N is less than n;

[0146] The similarity measurement module 840 is used to calculate the target curve similarity corresponding to each dimensional monitoring data in each group of the reduced dimensionality monitoring data according to the dimensional monitoring data and each group of target standardized data. If the dimensional monitoring data is determined to be abnormal data according to the target curve similarity, the dimensional monitoring data is removed from the reduced dimensionality monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the reduced dimensionality monitoring data to which the dimensional monitoring data belongs;

[0147] The fault prediction module 850 is used to perform fault prediction on the transmission line to be predicted based on each group of reduced-dimensional monitoring data obtained after elimination, and obtain a prediction result.

[0148] In one embodiment, the similarity measurement module 840 of the present application includes:

[0149] The similarity calculation unit is used to calculate the initial curve similarity between each dimensional monitoring data in each of the reduced dimensional monitoring data and each group of the target standardized data, and to average the initial curve similarities to obtain the target curve similarity corresponding to the dimensional monitoring data.

[0150] In one embodiment, the similarity calculation unit of the present application includes:

[0151] The initial curve similarity calculation unit is used to calculate the initial curve similarity between the dimensional monitoring data and each group of the target standardized data based on the following expression:

[0152]

[0153] In the formula, Monitor data for this dimension The initial curve similarity with a set of target standardized data Q, is the monitoring data of the i-th dimension in Q.

[0154] In one embodiment, the similarity measurement module 840 of the present application further includes:

[0155] The abnormal data elimination unit is used to screen out at least one group of similar standardized data from each group of target standardized data for each dimensional monitoring data in each group of reduced dimensionality monitoring data according to the similarities of each initial curve, and calculate the abnormal reference value corresponding to the dimensional monitoring data based on the dimensional monitoring data, each group of similar standardized data and the similarity of the target curve; if the dimensional monitoring data is determined to be abnormal data based on the size relationship between the dimensional monitoring data and the abnormal reference value, the dimensional monitoring data is eliminated from the reduced dimensionality monitoring data.

[0156] In one embodiment, the abnormal data elimination unit of the present application includes:

[0157] The abnormal reference value calculation unit is used to calculate the abnormal reference value corresponding to the monitoring data of this dimension based on the following expression:

[0158]

[0159] In the formula, is the abnormal reference value corresponding to the monitoring data of this dimension, A is the similar standardized data of each group, Monitoring data for this dimension, is the target curve similarity.

[0160] In one embodiment, the fault prediction module 850 of the present application includes:

[0161] A first filling unit is used for filling the missing data of discontinuous points based on the Hermite interpolation method for each group of reduced-dimensionality monitoring data after elimination, if the group of reduced-dimensionality monitoring data has missing data of discontinuous points;

[0162] A second filling unit is used for filling the missing data of continuous points based on the piecewise cubic Hermite interpolation method for each group of reduced-dimensionality monitoring data after elimination, if the group of reduced-dimensionality monitoring data has missing data of continuous points;

[0163] The prediction unit is used to perform fault prediction on the transmission line to be predicted based on each group of reduced-dimensional monitoring data obtained after filling, and obtain the prediction result.

[0164] In one embodiment, the prediction unit of the present application includes:

[0165] The model prediction unit is used to input each group of reduced-dimensional monitoring data obtained after filling into the pre-trained fault prediction model, and obtain the prediction result output by the fault prediction model; wherein the fault prediction model is obtained based on the DeepAR model.

[0166] In one embodiment, the prediction result includes the predicted fault type, predicted risk probability and priority processing suggestions. The transmission line fault prediction device 800 of the present application also includes:

[0167] A visualization module is used to display the predicted fault type, the predicted risk probability and the priority processing suggestion in a visualization map according to the location information corresponding to the transmission line to be predicted.

[0168] In one embodiment, the plurality of groups of line monitoring data include line engineering data and line environment data.

[0169] In one embodiment, the present application further provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the power transmission line fault prediction method in any embodiment.

[0170] In one embodiment, the present application further provides a computer device having computer-readable instructions stored therein. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the power transmission line fault prediction method in any embodiment.

[0171] Indicatively, Fig. 9 The internal structure diagram of a computer device provided in an embodiment of the present application is shown in FIG. 1 . In one example, the computer device may be a server. Fig. 9 , the computer device 900 includes a processing component 902, which further includes one or more processors, and a memory resource represented by a memory 901, for storing instructions that can be executed by the processing component 902, such as an application. The application stored in the memory 901 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 902 is configured to execute instructions to perform the steps of the method described in any of the above embodiments.

[0172] The computer device 900 may further include a power supply component 903 configured to perform power management of the computer device 900, a wired or wireless network interface 904 configured to connect the computer device 900 to a network, and an input / output (I / O) interface 905. The computer device 900 may operate based on an operating system stored in the memory 901, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0173] Those skilled in the art will understand that the internal structure of the computer device shown in the present application is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0174] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. Herein, "one", "one", "said", "the" and "it" may also include plural forms, unless the context clearly indicates another way. A plurality refers to at least two cases, such as 2, 3, 5 or 8, etc. "And / or" includes any and all combinations of the relevant listed items.

[0175] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0176] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting transmission line faults, It is characterized in that include: Acquire multiple groups of line monitoring data corresponding to the transmission line to be predicted, wherein the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data; Performing standardization processing on the plurality of groups of line monitoring data, and obtaining each group of standardized monitoring data; Using an approximate symbol aggregation algorithm, each group of the standardized monitoring data is subjected to dimensionality reduction processing, and multiple groups of N-dimensional dimensionality reduction monitoring data are obtained; wherein N is less than n; For each dimensional monitoring data in each group of the reduced dimensionality monitoring data, the target curve similarity corresponding to the dimensional monitoring data is calculated according to the dimensional monitoring data and each group of target standardized data; if the dimensional monitoring data is determined to be abnormal data according to the target curve similarity, the dimensional monitoring data is removed from the reduced dimensionality monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the reduced dimensionality monitoring data to which the dimensional monitoring data belongs; According to each group of reduced-dimensional monitoring data obtained after elimination, fault prediction is performed on the transmission line to be predicted, and a prediction result is obtained.

2. The method according to claim 1, It is characterized in that For each dimension of monitoring data in each group of the reduced-dimensional monitoring data, the target curve similarity corresponding to the dimension of monitoring data is calculated according to the dimension of monitoring data and each group of target standardized data, including: For each dimensional monitoring data in each of the reduced dimensional monitoring data, the initial curve similarity between the dimensional monitoring data and each group of the target standardized data is calculated respectively, and the average calculation is performed on each of the initial curve similarities to obtain the target curve similarity corresponding to the dimensional monitoring data.

3. The method according to claim 2, It is characterized in that The respectively calculating the initial curve similarity between the dimensional monitoring data and each group of the target standardized data comprises: Based on the following expressions, the initial curve similarity between the dimension monitoring data and each group of the target standardized data is calculated respectively: In the formula, Monitor data for this dimension The initial curve similarity with a set of target standardized data Q, is the monitoring data of the i-th dimension in Q.

4. The method according to claim 2, It is characterized in that For each dimension of monitoring data in each group of the reduced-dimensional monitoring data, if the dimension of monitoring data is determined to be abnormal data according to the target curve similarity, the dimension of monitoring data is removed from the reduced-dimensional monitoring data, including: For each dimensional monitoring data in each group of the reduced dimensionality monitoring data, at least one group of similar standardized data is screened out from each group of the target standardized data according to the similarities of each of the initial curves, and based on the dimensional monitoring data, each group of similar standardized data and the similarity of the target curve, the abnormal baseline value corresponding to the dimensional monitoring data is calculated; if the dimensional monitoring data is determined to be abnormal data based on the size relationship between the dimensional monitoring data and the abnormal baseline value, the dimensional monitoring data is eliminated from the reduced dimensionality monitoring data.

5. The method according to claim 4, It is characterized in that The calculating, based on the dimensional monitoring data, each group of similar standardized data and the target curve similarity, the abnormal reference value corresponding to the dimensional monitoring data comprises: Calculate the abnormal baseline value corresponding to the monitoring data of this dimension based on the following expression: In the formula, is the abnormal reference value corresponding to the monitoring data of this dimension, A is the similar standardized data of each group, Monitoring data for this dimension, is the target curve similarity.

6. The method according to any one of claims 1 to 5, It is characterized in that According to each group of reduced-dimensional monitoring data obtained after elimination, fault prediction is performed on the transmission line to be predicted, and a prediction result is obtained, including: For each group of reduced-dimensionality monitoring data after elimination, if there are discontinuous point missing data in the group of reduced-dimensionality monitoring data, the discontinuous point missing data are filled based on the Hermite interpolation method; For each group of reduced-dimensionality monitoring data after elimination, if there are missing data of continuous points in the group of reduced-dimensionality monitoring data, the missing data of continuous points are filled based on the piecewise cubic Hermite interpolation method; According to each group of reduced-dimensional monitoring data obtained after filling, fault prediction is performed on the transmission line to be predicted, and the prediction result is obtained.

7. The method according to claim 6, It is characterized in that According to each group of reduced-dimensional monitoring data obtained after filling, fault prediction is performed on the transmission line to be predicted, and the prediction result is obtained, including: Each group of reduced-dimensional monitoring data obtained after filling is input into the pre-trained fault prediction model, and the prediction result output by the fault prediction model is obtained; wherein the fault prediction model is obtained based on the DeepAR model.

8. The method according to any one of claims 1 to 5, It is characterized in that The prediction result includes predicted fault type, predicted risk probability and priority processing suggestions; the method also includes: According to the location information corresponding to the transmission line to be predicted, the predicted fault type, the predicted risk probability and the priority processing suggestion are displayed in a visual map.

9. The method according to any one of claims 1 to 5, It is characterized in that The multiple groups of line monitoring data include line engineering data and line environment data.

10. A transmission line fault prediction device, It is characterized in that include: A line monitoring data acquisition module is used to acquire multiple groups of line monitoring data corresponding to the power transmission line to be predicted, wherein the multiple groups of line monitoring data correspond to different monitoring indicators, and each group of line monitoring data is n-dimensional time series data; A standardization processing module, used for performing standardization processing on the plurality of groups of line monitoring data and obtaining each group of standardized monitoring data; A dimension reduction module, used to use an approximate symbol aggregation algorithm to perform dimension reduction processing on each group of the standardized monitoring data, and obtain multiple groups of N-dimensional dimension reduction monitoring data; wherein N is less than n; A similarity measurement module is used to calculate the target curve similarity corresponding to each dimensional monitoring data in each group of the reduced dimensionality monitoring data according to the dimensional monitoring data and each group of target standardized data; if the dimensional monitoring data is determined to be abnormal data according to the target curve similarity, the dimensional monitoring data is removed from the reduced dimensionality monitoring data, wherein the target standardized data is the remaining groups of standardized monitoring data except the source monitoring data, and the source monitoring data is the standardized monitoring data corresponding to the reduced dimensionality monitoring data to which the dimensional monitoring data belongs; The fault prediction module is used to perform fault prediction on the transmission line to be predicted based on each group of reduced-dimensional monitoring data obtained after elimination, and obtain a prediction result.