Converter station fault feature information expert database construction method

Abstract

The invention relates to the technical field of power system fault diagnosis, in particular to a converter station fault feature information expert database construction method, which comprises the following steps: firstly, acquiring data feature items of SER log events of a converter station, performing data preprocessing on the data feature items based on a DBSCAN clustering algorithm and an entropy method, reducing the number of events, reducing the SER data volume, and constructing a fault feature information expert database of the converter station; enhancing the incidence relation between the events to obtain a typical fault database; an item coding method is utilized to improve an Apriori algorithm to mine a typical fault database, SER event sets corresponding to different typical faults are obtained, and a screened SER associated event support group provides powerful support for fault diagnosis; constructing a convertor station fault feature information expert database based on an SQLite database by adopting SER event sets corresponding to different typical faults; a converter station fault feature information expert database is embedded into a program to assist in converter station fault analysis and diagnosis, so that the fault analysis efficiency is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault diagnosis, and in particular to a method for constructing an expert database of converter station fault feature information. Background Art

[0002] As the core of the DC transmission system, the converter station is prone to failures due to its complex equipment and long transmission distance. In order to improve the safe and stable operation of the DC transmission system, fast and accurate fault analysis and diagnosis are essential. However, the current converter station information system is not yet perfect, and the traditional fault analysis and diagnosis method mainly relies on manual work, which has the problems of low evaluation efficiency, long time consumption, and the accuracy of the results is affected by expert experience. In addition, there are few actual cases of DC faults, resulting in a long expert training cycle and a small number of experts, which further exacerbates the limitations of traditional methods.

[0003] As the complexity and intelligence of converter station equipment gradually increase, the operation and maintenance data of DC transmission systems show a trend of massive growth, which also puts forward higher requirements for the operation and maintenance level of converter stations. At present, the Sequential Event Recorder (SER), as the main source of converter station operation and maintenance data, can record the real-time operating status of each device and obtain important information such as event occurrence time, event level, event content, etc. Therefore, SER is one of the important means to extract fault features and intelligently diagnose converter stations.

[0004] In the actual operation of DC projects, there are fixed rules and processes for fault troubleshooting, analysis and emergency response. These rules and processes are essentially the accumulation of experience and knowledge in fault handling. Data mining technology can be used to extract fault features to form expert library samples, and then intelligent diagnosis of DC faults can be achieved based on expert library samples, greatly improving the efficiency of fault analysis and shortening the power outage time of DC transmission projects. Data mining technology has been applied in the operation and maintenance of converter stations. For example, association rule algorithms are used to mine SER event sets to diagnose whether there are missing events, which helps operation and maintenance personnel to promptly discover abnormal equipment movements and reduce the possibility of missing or misreading events when manually checking SER.

[0005] However, existing studies have only focused on manual analysis and mining results, and have not established a feature information expert database for massive SER events in converter stations to assist in intelligent fault diagnosis. Therefore, in order to overcome the limitations of traditional methods and adapt to the trend of massive growth in converter station operation and maintenance data, it is necessary to study a method for constructing a converter station fault feature information expert database based on SER preprocessing and improved association rule algorithm. Summary of the invention

[0006] The present invention provides a method for constructing an expert database of converter station fault feature information, which can greatly improve the efficiency of fault analysis and shorten the power outage time of direct current transmission projects, thereby better ensuring the safe and stable operation of the power system.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A method for constructing an expert database of converter station fault feature information comprises the following steps:

[0009] Obtain data feature items of SER log events of the converter station;

[0010] Based on the DBSCAN clustering algorithm and entropy method, data preprocessing is performed on the data feature items to obtain a typical fault database;

[0011] The apriori algorithm is improved by using the term coding method to mine the typical fault database, and the SER event set corresponding to different typical faults is obtained;

[0012] Using the SER event sets corresponding to the different typical faults, an expert database of converter station fault feature information based on the SQLite database is constructed;

[0013] The converter station fault feature information expert database is embedded into the program to assist in converter station fault analysis and diagnosis.

[0014] Furthermore, the data feature items are preprocessed based on the DBSCAN clustering algorithm and the entropy method, including:

[0015] The SER log events are clustered into time periods using the DBSCAN density clustering algorithm to remove event data within sparse event time periods;

[0016] The entropy method is used to calculate the information weight of each SER log event evaluation index in the event-intensive time period, and event data with low relevance is removed.

[0017] Furthermore, the use of the DBSCAN density clustering algorithm to cluster the SER log events into time periods includes the following steps:

[0018] Data preparation, extracting timestamp data from data feature items of the SER log event;

[0019] Initialize parameters, set Eps (neighborhood radius) and MinPts (minimum number of points) according to the distribution and clustering requirements of the SER log events; appropriately select two parameters that directly affect the clustering results, where Eps determines the size of a neighborhood, and MinPts specifies how many points are required in a neighborhood to form a cluster;

[0020] Perform clustering, using the DBSCAN algorithm to cluster the timestamp data, clustering similar time points into the same cluster to form different time periods;

[0021] Specifically, first traverse the unprocessed points, that is, select a point that has not been processed from the database as the starting point of the new cluster;

[0022] Next, calculate the number of points in the neighborhood and the number of points in the Eps neighborhood of the selected point to see if MinPts is reached. Assuming that events A and B are two data points, the neighborhood of A is N Eps (A) For given parameters Eps and MinPts, if A and B satisfy the following formula, then it can be considered that event A to event B is density-direct. The specific formula is:

[0023]

[0024] Further determine the neighborhood density. If the number of points in the neighborhood is less than MinPts, mark the point as noise. If the number of points in the neighborhood is greater than or equal to MinPts, create a new cluster with the point as the core point.

[0025] Then add points to the new cluster, adding all points directly density-reachable from the core point to the cluster;

[0026] Further search and add density-reachable points, continue to find points in the Eps neighborhood of the newly added points, and determine whether these points can be added to the current cluster. If the number of points in the neighborhood of a point is also greater than or equal to MinPts, add these points to the cluster and continue this process;

[0027] Finally, the cluster division results are output. When no new points can be added to any cluster, the algorithm ends and the cluster division results are output.

[0028] Result Analysis,The obtained time period clusters are analyzed, and time periods with dense events and time periods with sparse events are identified.

[0029] Further, the entropy method is used to calculate the information weight of each SER log event evaluation index in the event-intensive time period, including the following steps:

[0030] Select evaluation indicators, based on the data feature items of the SER log event, select event level and alarm group as evaluation indicators of the entropy method; specifically select event level and alarm group as evaluation indicators, and divide the event into four levels: emergency, alarm, minor and normal according to the urgency and potential impact of the event, and divide the alarm into two categories: protection-related and others;

[0031] Data quantification processing: quantification of text data of event levels and alarm groups in corresponding dimensions;

[0032] Calculate information entropy and use entropy method to calculate the information entropy of each evaluation index;

[0033] Calculate the information weight, and calculate the information weight of each evaluation index according to the calculation result of the information entropy.

[0034] Specifically, we first clarify the decision criteria, namely, event level and alarm group, and express the attribute values ​​of each event under these two criteria in matrix form to form a decision matrix X, as shown below:

[0035]

[0036] Among them, x in the matrix ij The element represents the attribute value of the i-th event under the j-th criterion (ie, event level or alarm group).

[0037] Secondly, normalize the matrix X. The matrix normalization calculation formula is as follows:

[0038]

[0039] Among them, x max and x min They represent the maximum and minimum values ​​in the j-th column of the matrix X, respectively.

[0040] The normalized matrix is ​​used to calculate the entropy value of each evaluation index. The entropy value reflects the discrete degree of the index value, that is, the uncertainty of the information.

[0041] Calculate the entropy of the evaluation index. The calculation formula is as follows:

[0042]

[0043] Among them, k = 1 / lnm, k is a constant and ensures that the entropy value is in the range of [0,1], x ij is x′ in the matrix X ij Normalized value.

[0044] Calculate the entropy weight of the evaluation index. Calculate the weight of each evaluation index according to the entropy value. The weight is inversely proportional to the entropy value. That is, the smaller the entropy value, the greater the weight, indicating that the index is more important in the comprehensive evaluation. The calculation formula is as follows:

[0045]

[0046] Among them, n is the number of evaluation indicators.

[0047] Calculate the final weight of the event information, combine the weight of each evaluation index and the normalized attribute value of the event under each index, and calculate the comprehensive weight of each event. This weight directly reflects the importance of the event. The calculation formula is as follows:

[0048]

[0049] The method of using the term coding method to improve the apriori algorithm to mine the typical fault database includes the following steps:

[0050] Input the typical fault database, wherein each typical fault represents a SER log event set, including a plurality of the SER log events and data feature items thereof;

[0051] Set the minimum support threshold and determine a minimum support threshold to identify the frequency of item sets;

[0052] Construct candidate k-item sets, find out the frequent item sets of all single items, and generate candidate k-item sets based on the combination of frequent (k-1)-item sets;

[0053] Construct candidate (k+1)-item sets, count the candidate k-item sets, screen out candidate item sets whose support is not less than the minimum support threshold as frequent k-item sets, construct candidate (k+1)-item sets based on the frequent k-item sets, and repeat this process until no new frequent item sets can be generated.

[0054] The frequent itemsets with the largest support and the longest itemset length in the mining process are taken as the SER-associated event support group. This support group represents that under this typical fault, this type of SER event set can be considered to have a high probability of occurrence. The final frequent itemsets represent the SER event set with a high probability of occurrence under different typical faults.

[0055] Furthermore, the construction of the converter station fault feature information expert database based on the SQLite database is to store the mined fault feature information into the SQLite database and form the converter station feature information expert database, and the key information recorded in the SQLite database includes the fault name, host, alarm group, event and event number;

[0056] The converter station fault feature information expert database is embedded into the program, and the data in the converter station fault feature information expert database is accessed by calling the corresponding API, and a fault diagnosis rule or model is formed according to the access information, and fault matching and diagnosis are performed with the data in the converter station fault feature information expert database.

[0057] The method for constructing an expert database of converter station fault feature information based on SER preprocessing and improved association rule algorithm combines DBSCAN clustering, entropy method and association rule algorithm based on term coding technology, which reduces the dimension and amount of SER data and is conducive to mining key information from SER events. Then, the SER data generated by typical faults in the converter station are mined to form the feature information of the fault diagnosis system, and finally, an expert database of converter station feature information based on SQLite database is established. It can effectively mine fault feature information from a large number of SER events and assist in completing fault diagnosis.

[0058] A device for constructing an expert database of fault characteristic information of a converter station, comprising:

[0059] A data acquisition module, used to acquire data feature items of SER log events of a converter station;

[0060] A data preprocessing module performs data preprocessing on the data feature items based on the DBSCAN clustering algorithm and the entropy method;

[0061] The data mining module uses the term encoding method to improve the apriori algorithm to mine the data feature items of the preprocessed SER log events;

[0062] The database construction module uses the mined fault feature information to build a converter station fault feature information expert database based on the SQLite database;

[0063] A program embedding and diagnosis module is used to embed the converter station fault feature information expert library into the program, access the data in the converter station fault feature information expert library by calling the corresponding API, form a fault diagnosis rule or model based on the access information, and perform fault matching and diagnosis with the data in the converter station fault feature information expert library.

[0064] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method for constructing an expert database of fault characteristic information of a converter station when executing the computer program.

[0065] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for constructing an expert database of converter station fault feature information.

[0066] A computer program product includes a computer program, which, when executed by a processor, implements the steps of the method for constructing an expert database of converter station fault feature information.

[0067] The beneficial effects of the present invention are:

[0068] 1. The present invention creatively proposes a method for constructing an expert database of converter station fault feature information based on SER preprocessing and improved association rule algorithm. The existing technologies are integrated and improved on this basis to assist in realizing intelligent fault diagnosis.

[0069] 2. The present invention further proposes a data preprocessing method based on DBSCAN clustering and entropy method. Compared with other data processing methods, it can more effectively reduce the number of alarms, reduce the amount of alarm data, exclude a large amount of irrelevant information, and provide a set of events with high importance and close correlation for subsequent association rule mining.

[0070] 3. The present invention further proposes an improved association rule algorithm based on item coding, which reduces the number of database traversals and the number of candidate item sets during data mining, improves the speed of association rule mining, and is beneficial to the mining of massive SER data in actual engineering.

[0071] 4. The present invention further proposes the construction of a converter station feature information expert database based on SQLite3, which effectively reduces the possibility of missing or misreading events caused by manual SER inspection, greatly improves the efficiency of fault analysis, and shortens the power outage time of DC transmission projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the embodiments of the present invention 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 recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0073] Figure 1 It is a flow chart of the method for constructing an expert database of converter station fault feature information in the present invention;

[0074] Figure 2 It is a flow chart of converter station event clustering in the expert database construction method of the present invention;

[0075] Figure 3 A visualization diagram of clustering results in the expert database construction method of the present invention;

[0076] Figure 4 This is a schematic diagram of the results of the converter station SER preprocessing part in the expert database construction method of the present invention;

[0077] Figure 5 It is a graph showing the number of SER preprocessing events in the expert database construction method of the present invention;

[0078] Figure 6 It is a visualization diagram of the feature information expert database in the expert database construction method of the present invention;

[0079] Figure 7 It is a flowchart of calling the converter station characteristic information expert database in the expert database construction method of the present invention;

[0080] Figure 8 It is a schematic diagram of the main interface of the fault diagnosis system in the expert database construction method of the present invention. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0082] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.

[0083] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0084] In the present invention, the Sequential Event Recording (SER) system can record the real-time operating status of each device in the converter station and obtain important information such as event occurrence time, event level, event content, etc. Therefore, the SER system becomes an important data source for fault feature extraction and intelligent diagnosis of converter stations.

[0085] Further use of data mining technology, such as association rule algorithms, to mine information from SER event sets and diagnose whether there are any missing events; this helps operation and maintenance personnel to promptly detect abnormal equipment behavior and improve the accuracy and efficiency of fault analysis and diagnosis.

[0086] Although data mining technology has been applied in converter station operation and maintenance, existing research has mainly remained at the stage of manual analysis of mining results. No expert database of feature information has been established for massive SER events in converter stations to assist in the intelligent diagnosis of faults.

[0087] Therefore, the scheme of the present invention aims to construct an expert database of converter station fault feature information based on SER preprocessing and improved association rule algorithm, which is specifically achieved through the following three key steps: first, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering and entropy method are used for data preprocessing to reduce the number of events and the amount of SER data, while enhancing the correlation between events; second, the Apriori algorithm is improved by using the term coding method to mine the associated event support groups and strongly associated events in the converter station SER; finally, based on the mining results, a converter station fault feature information expert database is constructed and embedded into the program to assist in completing the fault analysis and diagnosis of the converter station.

[0088] By providing a method for constructing an expert database of converter station fault feature information, the efficiency of fault analysis is greatly improved and the power outage time of DC transmission projects is shortened. Rapid and accurate fault analysis and diagnosis helps operation and maintenance personnel take timely measures to deal with faults and prevent the scope of faults from further expanding, thereby ensuring the safe and stable operation of the power system.

[0089] Specifically, the present invention discloses a method for constructing an expert database of converter station fault feature information, comprising the following steps:

[0090] Obtain data feature items of SER log events of the converter station;

[0091] Based on the DBSCAN clustering algorithm and entropy method, the data feature items are preprocessed to obtain a typical fault database;

[0092] The apriori algorithm is improved by using the item coding method to mine the typical fault database and obtain the SER event set corresponding to different typical faults.

[0093] Using the SER event sets corresponding to different typical faults, an expert database of converter station fault feature information based on the SQLite database is constructed;

[0094] The expert database of converter station fault feature information is embedded into the program to assist in converter station fault analysis and diagnosis.

[0095] See also Figure 1 As shown in the figure, the construction method is based on SER preprocessing and improved association rule algorithm, including using the idea of ​​density clustering, selecting the density clustering algorithm DBSCAN to complete the clustering of time periods for SER events with discrete information, finding time periods with dense events and converting events in sparse time periods into noise and deleting them from the original SER data set.

[0096] Furthermore, the event level (including emergency, alarm, minor and normal) and alarm group (protection related and other) are used as evaluation indicators of the entropy method to calculate the event information weight, and the corresponding dimensions of its text data are quantified to reflect the importance of the event. The interference of low-relevance information is eliminated to reduce the amount of event data.

[0097] Furthermore, the SER data of the converter station is mined. The improved association rule algorithm is used to analyze the SER-related event support group and the strong-related events when a typical fault occurs in the converter station. The frequent item set with the largest support and the longest item set length in the mining process is taken as the SER-related event support group, which represents that under this typical fault, this type of SER event set can be regarded as having a high probability of occurrence, and the SER event set corresponding to different typical faults is obtained.

[0098] Finally, the relevant attributes of the converter station event, including fault name, host, alarm group, event, etc., are entered into the SQLite database to form a converter station feature information expert library. By calling the corresponding API to access various data information in the feature information library, appropriate methods are used to fully utilize the feature information at the application layer, such as forming rules, to realize the application of the feature information expert library.

[0099] Example 1

[0100] In this embodiment, data preprocessing is performed on data feature items based on the DBSCAN clustering algorithm and the entropy method, including:

[0101] The DBSCAN density clustering algorithm is used to cluster SER log events into time periods, and event data in sparse event time periods are removed. The entropy method is used to calculate the information weight of each SER log event evaluation index in the event-intensive time period, and event data with low relevance are removed.

[0102] The DBSCAN clustering algorithm is a density-based clustering method that can identify clusters of any shape and has good robustness to noisy data. In SER data, similar events can be clustered together through DBSCAN clustering to form event clusters, thereby reducing the number of events and the amount of data. The DBSCAN density clustering algorithm is used to cluster the SER log events into time periods, including the following steps:

[0103] Data preparation: extracting timestamp data from the data feature items of the SER log events; this is the key data for time period clustering.

[0104] Initialize parameters, set Eps (neighborhood radius) and MinPts (minimum number of points) according to the distribution and clustering requirements of the SER log events; appropriately select two parameters that directly affect the clustering results, which need to be tuned through experiments or experience; Eps determines the size of a neighborhood, and MinPts specifies how many points are required in a neighborhood to form a cluster;

[0105] Perform clustering, using the DBSCAN algorithm to cluster the timestamp data, clustering similar time points into the same cluster to form different time periods;

[0106] Specifically, first traverse the unprocessed points, that is, select a point that has not been processed from the database as the starting point of the new cluster;

[0107] Next, calculate the number of points in the neighborhood and the number of points in the Eps neighborhood of the selected point to see if MinPts is reached. Assuming that events A and B are two data points, the neighborhood of A is N Eps (A) For given parameters Eps and MinPts, if A and B satisfy the following formula, then it can be considered that event A to event B is density-direct. The specific formula is:

[0108]

[0109] Further determine the neighborhood density. If the number of points in the neighborhood is less than MinPts, mark the point as noise. If the number of points in the neighborhood is greater than or equal to MinPts, create a new cluster with the point as the core point.

[0110] Then add points to the new cluster, adding all points directly density-reachable from the core point to the cluster;

[0111] Further search and add density-reachable points, continue to find points in the Eps neighborhood of the newly added points, and determine whether these points can be added to the current cluster. If the number of points in the neighborhood of a point is also greater than or equal to MinPts, add these points to the cluster and continue this process;

[0112] Finally, the cluster division results are output. When no new points can be added to any cluster, the algorithm ends and the cluster division results are output.

[0113] Result analysis: Analyze the obtained time period clusters and identify time periods with dense events and time periods with sparse events; based on the analysis results, remove event data in sparse time periods to reduce the impact of noise on subsequent analysis.

[0114] To summarize, this embodiment can reduce the amount of data and improve the efficiency of subsequent data analysis and mining by removing event data in sparse time periods; event data in sparse time periods may contain more noise and outliers, and removing these data helps to improve the accuracy of subsequent fault diagnosis models; focusing on key time periods and retaining data in time periods where events occur intensively can more focusedly analyze the fault characteristics and patterns in these time periods.

[0115] In this embodiment, the density clustering algorithm DBSCAN is first used to cluster the time periods to identify the time periods with dense and sparse events; on this basis, in order to further streamline the data and highlight key information, the entropy method is introduced to evaluate the importance of the event. The entropy method is used to calculate the information weight of each SER log event evaluation index in the event-intensive time period, including the following steps:

[0116] Select evaluation indicators, based on the data feature items of the SER log event, select event level and alarm group as evaluation indicators of the entropy method; specifically select event level and alarm group as evaluation indicators, and divide the event into four levels: emergency, alarm, minor and normal according to the urgency and potential impact of the event, and divide the alarm into two categories: protection-related and others;

[0117] Data quantification processing: quantify the text data of event levels and alarm groups in corresponding dimensions, encode each event according to the event level and alarm group, and form numerical evaluation indicators;

[0118] Calculate information entropy and use entropy method to calculate the information entropy of each evaluation index;

[0119] Calculate the information weight, and calculate the information weight of each evaluation indicator according to the calculation result of the information entropy. The larger the weight, the greater the role of the indicator in evaluating the importance of the event.

[0120] Specifically, we first clarify the decision criteria, namely, event level and alarm group, and express the attribute values ​​of each event under these two criteria in matrix form to form a decision matrix X, as shown below:

[0121]

[0122] The xij element in the matrix represents the attribute value of the ith event under the jth criterion (ie, event level or alarm group).

[0123] Secondly, normalize the matrix X. In order to eliminate the influence of different dimensions, the decision matrix X is normalized. The normalization formula takes into account the maximum and minimum values ​​of each column of data to ensure that all attribute values ​​fall within the interval [0,1]. The matrix normalization calculation formula is as follows:

[0124]

[0125] Among them, x max and x min They represent the maximum and minimum values ​​in the j-th column of the matrix X, respectively.

[0126] The normalized matrix is ​​used to calculate the entropy value of each evaluation index. The entropy value reflects the discrete degree of the index value, that is, the uncertainty of the information.

[0127] Calculate the entropy of the evaluation index. The calculation formula is as follows:

[0128]

[0129] Among them, k = 1 / lnm, k is a constant and ensures that the entropy value is in the range of [0,1], x ij is x′ in the matrix X ij Normalized value.

[0130] Calculate the entropy weight of the evaluation index. Calculate the weight of each evaluation index according to the entropy value. The weight is inversely proportional to the entropy value. That is, the smaller the entropy value, the greater the weight, indicating that the index is more important in the comprehensive evaluation. The calculation formula is as follows:

[0131]

[0132] Among them, n is the number of evaluation indicators.

[0133] Calculate the final weight of the event information, combine the weight of each evaluation index and the normalized attribute value of the event under each index, and calculate the comprehensive weight of each event. This weight directly reflects the importance of the event. The calculation formula is as follows:

[0134]

[0135] In summary, the entropy method can identify time periods with dense events based on density clustering, and can also quantitatively evaluate the importance of events, more objectively reflect the value of each event in fault diagnosis, eliminate the interference of low-correlation information, and help reduce data noise and improve data quality. When processing large-scale, discrete SER event data, further reducing the amount of data provides a more valuable data basis for subsequent fault feature information extraction and fault diagnosis, while improving processing efficiency.

[0136] Example 2

[0137] This embodiment improves the association rule algorithm and mines and analyzes the SER data under different typical faults to obtain the corresponding SER event sets. These event sets help the staff to more accurately identify the fault type, locate the fault cause, and take corresponding treatment measures.

[0138] The apriori algorithm is improved by using the item coding method to mine the typical fault database, including the following steps:

[0139] A typical fault database is input, where each typical fault represents a SER log event set, which includes a series of items (ie, SER log events).

[0140] Set a minimum support threshold and determine a minimum support threshold to identify the frequency of item sets; support represents the frequency of item sets in the transaction database. This threshold can be adjusted according to actual conditions to ensure that the mined frequent item sets are representative.

[0141] Construct candidate k-item sets, find out the frequent item sets of all single items, and generate candidate k-item sets based on the combination of frequent (k-1)-item sets;

[0142] Construct candidate (k+1)-item sets, count the candidate k-item sets, screen out candidate item sets whose support is not less than the minimum support threshold as frequent k-item sets, construct candidate (k+1)-item sets based on frequent k-item sets, and repeat this process until no new frequent item sets can be generated.

[0143] The filtered frequent itemsets are sorted from large to small according to their support, and the itemsets with the largest support and the longest itemset length are selected as the SER-related event support groups. These support groups provide valuable information for fault diagnosis, that is, when a typical fault occurs, which SER events are likely to occur simultaneously.

[0144] In summary, in this embodiment, the apriori algorithm is improved by the term coding method to improve the accuracy and efficiency of mining associated events in the SER data of the converter station. The screened SER associated event support group provides strong support for fault diagnosis, which helps the staff to respond to and handle faults more quickly. The SER event sets corresponding to different typical faults provide valuable data resources for the operation and maintenance management of the converter station.

[0145] Example 3

[0146] This embodiment constructs a converter station fault feature information expert database based on the SQLite3 database. The expert database can effectively mine fault feature information from a large number of SER event sets. By querying the fault feature information in the expert database, the cause of the fault can be quickly located, and the efficiency and accuracy of fault diagnosis can be improved, thereby improving the operating reliability and safety of the converter station.

[0147] Constructing a converter station fault feature information expert database based on the SQLite database is to store the mined fault feature information into the SQLite database and form a converter station feature information expert database. The key information recorded in the SQLite database includes fault name, host, alarm group, event and event number.

[0148] Specifically, first determine the relevant attributes of the converter station events to record data information, including fault name, host, alarm group, event, etc. These attributes are the basis for building an expert database and can fully reflect the characteristics of converter station faults. Design the database table structure based on the attributes to ensure data integrity and consistency. For example, a table containing fields such as fault name, host, alarm group, event, etc. can be designed to store converter station event information.

[0149] Secondly, the SER events generated during the trial operation of a flexible DC project are used as data sources. These data have practical application backgrounds and can be used to verify the effectiveness and accuracy of the expert database. The data response of the data layer and the program call of the application layer are realized through SQLite data interface encapsulation. This ensures the security and reliability of the data and facilitates the subsequent application development.

[0150] Finally, the SER event set corresponding to the typical faults obtained after preprocessing is imported into the SQLite database, which can be achieved by writing a data import script or program; performing optimization operations such as indexing and compression on the database can further improve query efficiency and storage performance.

[0151] In this embodiment, the converter station fault characteristic information expert library is embedded in the program, and the data in the converter station fault characteristic information expert library is accessed by calling the corresponding API. A fault diagnosis rule or model is formed according to the access information, and fault matching and diagnosis are performed with the data in the converter station fault characteristic information expert library.

[0152] Specifically, according to the characteristics of the SQLite database and application requirements, the corresponding API interface is designed to facilitate the application to access various data information in the feature information library. By calling the API interface, the application can obtain the required converter station fault feature information, which may include fault name, host status, alarm group information, etc., and perform subsequent processing and analysis.

[0153] Furthermore, according to the results of data analysis, corresponding fault diagnosis rules, prediction models, etc. are formed, and a rule matching program is written according to the formed fault diagnosis rules or models. The program can automatically match the faults occurring in the converter station with the records in the feature information library and give the corresponding diagnosis results. In the rule matching program, the feature information expert library is called to obtain the required feature information for matching, which can be achieved by directly querying the database or using the API interface.

[0154] In summary, this embodiment uses the SER event set corresponding to the different typical faults to build a converter station fault feature information expert library based on the SQLite database to achieve rapid diagnosis and location of converter station faults. At the same time, by calling the feature information expert library and embedding it into the application, these feature information can be conveniently used at the application layer for fault diagnosis, prediction and other related applications, to achieve automated and intelligent fault diagnosis and processing, and further improve the operational reliability and safety of the converter station.

[0155] Example 4

[0156] In the specific implementation process, the data preprocessing method based on DBSCAN clustering and entropy method includes the following steps:

[0157] ① Collect the original SER data of the converter station, as shown in Table 1 below. These data include six feature dimensions: time, host, system alarm, event level, alarm group, and event list.

[0158]

[0159] Table 1. SER data feature dimensions

[0160] ②Retain the time dimension, which can reflect the time period with dense alarm time. The correlation between SER events in the dense time period is often closer, and the clustering analysis of converter station events can be carried out based on this.

[0161] ③ Use the density clustering algorithm DBSCAN to complete the clustering of time periods, find the time periods with dense events, and transform the events in the sparse time periods into noise and delete them from the original SER data set, such as Figure 2 shown.

[0162] ④ Select the clustering radius Eps as 0.01, then the clustering algorithm has a suitable number of clusters and event sets. The clustering results are as follows Figure 3 As shown in the figure, there are 72 clusters representing dense event segments and 59 clusters representing sparse time segments after clustering.

[0163] ⑤ Select event level and alarm group as the evaluation indicators of the entropy method to calculate the event information weight, and quantify the corresponding dimensions. The quantification results of event level and alarm group are shown in Table 2.

[0164]

[0165] Table 2. Event level and alarm group quantification

[0166] ⑥ The relative importance of alarm information is evaluated by weighting event information through entropy method. By calculating the information entropy and weight of the criteria, the weights of multiple criteria are converted into a comprehensive weight, thereby achieving comprehensive evaluation and ranking of different goals, and finally obtaining the partial results of converter station SER preprocessing, such as Figure 4 shown.

[0167] ⑦ Such as Figure 5 As shown in the figure, it means that the number of SER data after preprocessing is greatly reduced compared with the original data. It can be seen that after the converter station SER data preprocessing adopted in this paper, the number of alarms can be effectively reduced, the amount of alarm data can be reduced, a large amount of irrelevant information can be excluded, and a set of events with high importance and close correlation can be provided for subsequent association rule mining.

[0168] Furthermore, using the improved association rule algorithm based on item coding to mine data includes the following steps:

[0169] ① Taking the converter grounding fault as an example, the SER system recorded a total of 2134 SER events within about 120 seconds from 16:44:03 on July 4, 2018 to 16:45:57 on July 4, 2018 when the fault occurred.

[0170] ② Set the minimum support of the Apriori algorithm to 0.03 and the minimum confidence to 0.7. The frequent itemsets generated during the mining process are statistically sorted by item set length and support. Some results are shown in Table 3.

[0171]

[0172] Table 3. Some mining results

[0173] ③ From the above table, we can see that the event set {9, 10, 11} has the highest support in the frequent item set mined, and can be used as the SER association event support group. According to the support group, we search in all association rules and get the association rules as follows: Figure 6 shown.

[0174] ④ The event set {9, 10, 11} is used as the support group of SER associated events, and its subset can be used as the rule antecedent of the association rule, while the rule consequent events 12 and 13 are used as SER strong associated events. The support group events and strong associated events are mapped from event IDs to specific SER events, and the mining results are shown in Table 4.

[0175]

[0176] Table 4. Converter ground fault correlation mining results

[0177] Furthermore, building a converter station feature information expert database based on SQLite3 includes the following steps:

[0178] ① Construct a converter station feature information expert database to record data information mainly related to converter station events, including fault name, host, alarm group, event, etc. The detailed fields of the database are shown in Table 5.

[0179]

[0180] Table 4. Feature information database event information table

[0181] ② Taking the SER events generated during the trial operation of a flexible DC project as the data source, the data layer data response and application layer program call are realized through SQLite data interface encapsulation to complete the construction of the feature information expert database. The data layer visualization results are as follows: Figure 6 shown.

[0182] ③ Call the feature information expert library and embed it into the application program. By calling the corresponding API to access various data information in the feature information library, use appropriate methods to make full use of the feature information at the application layer, such as forming rules, etc., to realize the application of the feature information expert library. The calling process is as follows: Figure 7 shown.

[0183] ④ According to the above principle, write the rule matching program and call the feature information expert library. The result is as follows: Figure 8 shown.

[0184] In summary, this embodiment provides a method for constructing an expert database of fault feature information of converter stations based on SER preprocessing and improved association rule algorithm, which adopts entropy method and DBSCAN clustering technology. First, the cluster boundary is determined by defining density reachability and density connection, thereby reducing the amount of SER transaction data and finding the intensive time period. Secondly, based on the concept of information entropy, the information entropy and weight of the criterion are calculated, and the weights of multiple criteria are converted into a comprehensive weight, thereby realizing comprehensive evaluation and ranking of different objectives. Finally, a set of SER events with high importance and close association is generated.

[0185] Furthermore, in this embodiment, an algorithm optimization method based on the term coding method is adopted to realize the purpose that the algorithm only needs to scan the database once and greatly reduce the number of candidate sets, thereby improving the processing speed of the algorithm when facing massive SER data. The present invention adopts the converter station feature information expert library technology based on SQLite3, embeds the feature information expert library into the application, accesses various data information in the feature information library by calling the corresponding API, and uses appropriate methods to fully utilize the feature information at the application layer, such as forming rules, etc., to realize the application of the feature information expert library.

[0186] In this embodiment, a device for constructing an expert database of fault feature information of a converter station is provided, comprising:

[0187] A data acquisition module, used to acquire data feature items of SER log events of a converter station;

[0188] Data preprocessing module, which performs data preprocessing on data feature items based on DBSCAN clustering algorithm and entropy method;

[0189] The data mining module uses the term encoding method to improve the apriori algorithm to mine the data feature items of the preprocessed SER log events;

[0190] The database construction module uses the mined fault feature information to build a converter station fault feature information expert database based on the SQLite database;

[0191] The program embedding and diagnosis module is used to embed the converter station fault feature information expert library into the program, access the data in the converter station fault feature information expert library by calling the corresponding API, form fault diagnosis rules or models based on the access information, and perform fault matching and diagnosis with the data in the converter station fault feature information expert library.

[0192] In this embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for constructing an expert database of fault characteristic information of converter stations are implemented.

[0193] In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for constructing an expert database of fault characteristic information of converter stations are implemented.

[0194] In this embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of the method for constructing an expert database of converter station fault feature information.

[0195] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0196] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0197] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0198] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing an expert database of converter station fault feature information, characterized in that: The following steps are involved: Obtain data feature items of SER log events of the converter station; Based on the DBSCAN clustering algorithm and entropy method, data preprocessing is performed on the data feature items to obtain a typical fault database; The apriori algorithm is improved by using the term coding method to mine the typical fault database, and the SER event set corresponding to different typical faults is obtained; Using the SER event sets corresponding to the different typical faults, an expert database of converter station fault feature information based on the SQLite database is constructed; The converter station fault feature information expert database is embedded into the program to assist in converter station fault analysis and diagnosis.

2. The method for constructing a converter station fault feature information expert database according to claim 1, characterized in that: The data preprocessing of the data feature items based on the DBSCAN clustering algorithm and the entropy method includes: The SER log events are clustered into time periods using the DBSCAN density clustering algorithm to remove event data within sparse event time periods; The entropy method is used to calculate the information weight of each SER log event evaluation index in the event-intensive time period, and event data with low relevance is removed.

3. The method for constructing a converter station fault feature information expert database according to claim 2, characterized in that: The method of clustering the SER log events into time periods using the DBSCAN density clustering algorithm comprises the following steps: Data preparation, extracting timestamp data from data feature items of the SER log event; Initialize parameters and set Eps (neighborhood radius) and MinPts (minimum number of points) according to the distribution and clustering requirements of the SER log events; Perform clustering, using the DBSCAN algorithm to cluster the timestamp data, clustering similar time points into the same cluster to form different time periods; Result Analysis,The obtained time period clusters are analyzed, and time periods with dense events and time periods with sparse events are identified.

4. The method for constructing a converter station fault feature information expert database according to claim 2, characterized in that: The method of using the entropy method to calculate the information weight of each SER log event evaluation index in the event-intensive time period includes the following steps: Selecting evaluation indicators, based on the data feature items of the SER log event, selecting event level and alarm group as evaluation indicators of the entropy method; Data quantification processing: quantification of text data of event levels and alarm groups in corresponding dimensions; Calculate information entropy and use entropy method to calculate the information entropy of each evaluation index; Calculate the information weight, and calculate the information weight of each evaluation index according to the calculation result of the information entropy.

5. The method for constructing a converter station fault feature information expert database according to claim 1, characterized in that: The method of using the term coding method to improve the apriori algorithm to mine the typical fault database includes the following steps: Input the typical fault database, wherein each typical fault represents a SER log event set; Set the minimum support threshold and determine a minimum support threshold to identify the frequency of item sets; Construct candidate k-item sets, find out the frequent item sets of all single items, and generate candidate k-item sets based on the combination of frequent (k-1)-item sets; Construct candidate (k+1)-item sets, count the candidate k-item sets, screen out candidate item sets whose support is not less than the minimum support threshold as frequent k-item sets, construct candidate (k+1)-item sets based on the frequent k-item sets, and repeat this process until no new frequent item sets can be generated.

6. The method for constructing a converter station fault feature information expert database according to claim 1, characterized in that: The construction of the converter station fault feature information expert database based on the SQLite database is to store the mined fault feature information into the SQLite database to form the converter station feature information expert database, and the key information recorded in the SQLite database includes the fault name, host, alarm group, event and event number; The converter station fault feature information expert database is embedded into the program, and the data in the converter station fault feature information expert database is accessed by calling the corresponding API, and a fault diagnosis rule or model is formed according to the access information, and fault matching and diagnosis are performed with the data in the converter station fault feature information expert database.

7. A device for constructing an expert database of fault feature information of a converter station, characterized in that: include: A data acquisition module, used to acquire data feature items of SER log events of a converter station; A data preprocessing module performs data preprocessing on the data feature items based on the DBSCAN clustering algorithm and the entropy method; The data mining module uses the term encoding method to improve the apriori algorithm to mine the data feature items of the preprocessed SER log events; The database construction module uses the mined fault feature information to build a converter station fault feature information expert database based on the SQLite database; A program embedding and diagnosis module is used to embed the converter station fault feature information expert library into the program, access the data in the converter station fault feature information expert library by calling the corresponding API, form a fault diagnosis rule or model based on the access information, and perform fault matching and diagnosis with the data in the converter station fault feature information expert library.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for constructing an expert database of converter station fault feature information according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing an expert database of converter station fault feature information according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for constructing an expert database of converter station fault feature information described in any one of claims 1 to 6 are implemented.

Images