Alarm analysis and positioning method and system for dispatching automation machine room equipment operation and maintenance

Abstract

The invention discloses an alarm analysis and positioning method for operation and maintenance of equipment in a dispatching automation machine room, which comprises the following steps of: establishing an alarm information knowledge graph template based on historical alarm information of the dispatching automation machine room, and establishing a real-time alarm information knowledge graph according to real-time alarm information; and matching the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph, and positioning fault equipment according to the real-time alarm information of the dispatching automation machine room and the fault type of the real-time fault alarm information. According to the invention, a large amount of alarm information in the dispatching automation machine room can be analyzed and positioned, and an analysis result is obtained for dispatching operation and maintenance personnel to refer.

Description

Technical Field

[0001] The present invention relates to the field of electric power dispatching automation operation and maintenance, and in particular to a method and system for alarm analysis and positioning of dispatching automation machine room equipment operation and maintenance. Background Art

[0002] With the continuous improvement of the intelligent operation capability of the power grid, the deep integration of advanced information and communication technology, control technology and all aspects and services of the power grid has been accelerated. At present, emerging technologies such as artificial intelligence have been widely used in all aspects of the power grid, but their application in optimizing the professional workflow and business data of dispatching automation is relatively small.

[0003] The operation and management of the dispatching automation room is a decentralized management mode of dispatching rooms in various places. There are mainly the following aspects in operation, maintenance, monitoring and management:

[0004] 1) There is a shortage of automation professional operation duty personnel in the province, which cannot meet the 24-hour duty requirements;

[0005] 2) The on-duty personnel patrol infrequently and are unable to promptly detect and handle faults in the computer room;

[0006] 3) Important information is often omitted due to the large amount of data, resulting in the inability to accurately identify abnormal faults in the system and equipment;

[0007] 4) Lack of means for alarm event analysis, fault diagnosis and auxiliary decision-making. Summary of the invention

[0008] The purpose of the present invention is to provide a method and system for alarm analysis and positioning of equipment operation and maintenance in a dispatching automation room, to analyze and locate a large amount of alarm information in the dispatching automation room, and to obtain analysis results for reference by dispatching operation and maintenance personnel.

[0009] To achieve this purpose, the present invention is designed to provide an alarm analysis and positioning method for dispatching automation equipment room operation and maintenance, which includes the following steps:

[0010] Step 1: Establish an alarm information knowledge graph template based on the historical alarm information of the dispatching automation room;

[0011] Step 2: extract the descriptive participles of the real-time fault events in the real-time alarm information text of the dispatching automation room as entities, and number them according to the extraction order of the extracted real-time fault event descriptive participles;

[0012] Step 3: Establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number;

[0013] Step 4: construct a real-time fault event alarm information triple according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triple;

[0014] Step 5: Match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph;

[0015] Step 6: Locate the faulty device according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

[0016] An alarm analysis and positioning system for dispatching automated equipment room operation and maintenance, comprising a knowledge graph template establishment module, a description word segmentation numbering module, a correlation matrix establishment module, a real-time knowledge graph establishment module, a fault type matching module and a fault equipment positioning module;

[0017] The knowledge graph template establishment module is used to establish an alarm information knowledge graph template based on the historical alarm information of the scheduling automation room;

[0018] The description word numbering module is used to extract the descriptive word of the real-time fault event in the real-time alarm information text of the dispatching automation room as an entity, and number the extracted real-time fault event description word according to the extraction order;

[0019] The correlation matrix establishment module is used to establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number;

[0020] The real-time knowledge graph establishment module is used to construct real-time fault event alarm information triples according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triples;

[0021] The fault type matching module is used to match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph;

[0022] The faulty equipment locating module is used to locate the faulty equipment according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

[0023] Beneficial effects of the present invention:

[0024] The present invention establishes an alarm information knowledge graph template based on historical alarm information, a real-time alarm information knowledge graph based on actual alarm information, and matches the fault type in the alarm information knowledge graph template based on the real-time alarm information knowledge graph, determines the faulty equipment, and obtains analysis results for reference by scheduling and operation personnel. Through the matching of knowledge graphs, specific fault modes can be quickly identified from a large amount of alarm information, avoiding the tediousness and inefficiency of manual one-by-one investigation. The structured representation of the knowledge graph makes it easy to cluster and associate similar or related alarm information, thereby improving the accuracy of fault identification. The nodes (such as devices, components) and edges (such as connection relationships, fault relationships) in the knowledge graph clearly show the system architecture and fault propagation path, which helps to quickly locate faulty equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Construct a flow chart for the knowledge graph of alarm information in the dispatch automation room;

[0026] Figure 2 This is a schematic diagram of the depth-first algorithm;

[0027] Figure 3 Determine the flow chart for the faulty equipment;

[0028] Figure 4 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0029] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0030] A method for analyzing and locating alarms for scheduling automated equipment room operation and maintenance, comprising the following steps;

[0031] Step 1: Establish an alarm information knowledge graph template based on the historical alarm information of the dispatching automation room;

[0032] Step 2: extract the descriptive participles of the real-time fault events in the real-time alarm information text of the dispatching automation room as entities, and number them according to the extraction order of the extracted real-time fault event descriptive participles;

[0033] Step 3: Establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number;

[0034] Step 4: construct a real-time fault event alarm information triple according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triple;

[0035] Step 5: Match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph;

[0036] Step 6: Locate the faulty device according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

[0037] The present invention studies the adjacency relationship of fault event description segmentations in alarm information. Each fault event description segmentation is an entity. A single alarm information generally contains multiple entities. The relationship between the entities forms a characteristic description of the fault event.

[0038] Alarm information is a description of equipment failure events in the dispatching automation room operation and maintenance system, so the alarm information knowledge graph belongs to the domain knowledge graph. Its entity meanings have clear naming specifications and do not require entity disambiguation and coreference resolution processes.

[0039] When extracting relations, a fault description word correlation matrix is ​​established based on the number of adjacent times between the previous fault word and the next fault word, and the number of adjacent times between fault words is used as the relationship between entities.

[0040] The alarm information is integrated to sort out the triple data of the alarm information, namely the head entity, the number of adjacent entities, and the tail entity. Using the Neo4j graph database, a knowledge graph based on the alarm information is established, such as Figure 1 shown.

[0041] A method for analyzing and locating alarms for scheduling automated equipment room operation and maintenance The specific method for establishing an alarm information knowledge graph template in step 1 is:

[0042] Step 1.1, extract the descriptive participles of historical fault events in the text of the historical alarm information of the dispatching automation room as entities, and number them according to the extraction order of the descriptive participles of the historical fault events;

[0043] Step 1.2: Establish a historical fault event correlation matrix based on the historical fault event description segmentations and corresponding numbers;

[0044] Step 1.3: Construct historical fault event alarm information triples based on the historical fault event correlation matrix, and form a historical alarm information knowledge graph template based on the historical fault event alarm information triples.

[0045] A method for alarm analysis and positioning of equipment operation and maintenance in a dispatching automation room, wherein in step 2, the descriptive participles of real-time fault events in the real-time alarm information text of the dispatching automation room are extracted as entities, and the specific method for numbering the extracted real-time fault event descriptive participles in the order of extraction is:

[0046] Step 2.1, based on the conventional dictionary and the Viterbi algorithm, the text of the real-time alarm information of the dispatching automation room is segmented;

[0047] Step 2.2, perform deduplication filtering on the dispatch automation room real-time alarm information text after word segmentation, build a real-time alarm information fault event description word segmentation dictionary, and number the real-time fault event description word segments appearing in the dictionary in chronological order.

[0048] A method for alarm analysis and positioning for scheduling automated equipment room operation and maintenance, characterized in that the specific method of establishing a real-time fault event description word segmentation correlation matrix based on real-time fault event description word segmentation and corresponding numbering in step 3 is:

[0049] Combine the real-time fault event description segment words with different numbers in pairs, and use the numbers of the real-time fault event description segment words as the row number and column number of the matrix in order to construct the real-time fault event description segment word correlation matrix;

[0050]

[0051] A nn represents the fault event description word segmentation correlation matrix, where the diagonal elements cor ii is 0; i,j=1,2,3,4,...,n, indicating the number of the fault event description word in the fault event description word dictionary, cor ij represents the number of times fault segmentation word j appears after fault segmentation word i. According to the characteristics of alarm information text data, it can be known that A nn is an asymmetric matrix;

[0052] When operating, first initialize the matrix, that is, cor ij =0(i,j=1,2,3,4,...,n), traverse the alarm information text after word segmentation, and when the fault description word numbered j appears once after the fault event description word numbered i, perform the following operation:

[0053] A nn =A' nn +E ij (2)

[0054] Among them, E ij Indicates that the value of the i-th row and j-th column of the matrix is ​​1, as shown in Formula 3, A' nn Represents the last traversal result matrix, A nn Represents the result matrix of the current traversal, which is used to construct the triple information of the knowledge graph. nn The value of the i-th row and j-th column in represents the number of times fault segmentation j appears after fault segmentation i, that is, the relationship of the corresponding triples;

[0055]

[0056] A method for analyzing and locating alarms for scheduling automated equipment room operation and maintenance In step 4, a real-time fault event alarm information triple is constructed according to the real-time fault event description word segmentation correlation matrix, and a specific method for forming a real-time alarm information knowledge graph according to the real-time fault event alarm information triple is as follows:

[0057] The data of the real-time fault event description word segmentation correlation matrix is ​​integrated, and the fault event description word segmentations that appear successively in the alarm information text after word segmentation are taken as the head entity and the tail entity respectively, and the adjacent frequencies between them are taken as the relationship between entities. The real-time fault event alarm information triples are constructed, and the real-time fault event alarm information triples are input into the Neo4j graph database, and then the real-time alarm information knowledge graph is output.

[0058] A method for alarm analysis and positioning for scheduling automated equipment room operation and maintenance, wherein the specific method for matching the fault type of real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph in step 5 is:

[0059] There are N sets of triple sequences (h, r, t) in the real-time alarm information knowledge graph, where h represents the head entity, t represents the tail entity, and r represents the adjacent frequency. The total adjacent frequency of alarm information is M:

[0060]

[0061] where r n is the adjacent frequency in the nth triple;

[0062] The similarity function Sim is defined as:

[0063]

[0064] In the formula, Represents the ath triplet in the real-time alarm information, with the header entity being h a , the adjacent frequency is r a , the tail entity is t a ; Indicates the bth triple in the jth template, with the head entity h b , the adjacent frequency is r b , the tail entity is t b ;

[0065]

[0066] Where ε j is the similarity between the real-time alarm information and the jth template, M is the total frequency of adjacent real-time alarm information, ∑Sim[A,Bj ] is the sum of similarity functions of all triplets in the real-time alarm information and all triplets in the j-th alarm information template;

[0067] The depth-first algorithm is used to traverse the real-time alarm information knowledge graph. The depth-first algorithm flow chart is as follows: Figure 2 , compare the N groups of triplets in the real-time alarm information knowledge graph with all templates one by one, and calculate the similarity between the N groups of triplets in the real-time alarm information knowledge graph and each template.

[0068] In order to improve the accuracy of the matching algorithm, the TF-IDF (term frequency–inverse document frequency) method is used to improve the algorithm. TF stands for term rate and IDF stands for inverse document frequency. The basic idea is that when a triple appears frequently in one type of template and rarely appears in other templates, it means that this triple can well represent the characteristics of this type of template. A higher weight is given to the triple with obvious characteristics to distinguish different templates. The calculation formula is as follows:

[0069]

[0070]

[0071] TF-IDF ij =TF ij ×IDF ij (9)

[0072] In equations (7), (8) and (9), TF ij For the triple L i The word rate in the j-th category of text, N ij For the triple L i The number of times it appears in the jth type of text, N kj is the sum of the number of occurrences of triple k in the j-th text, is the sum of all triples appearing in the j-th category of text, IDF i For the triple L i The inverse text frequency, N is the number of all texts in the data set, n i is a triple L i Number of texts; TF-IDF ij For the triple L i The TF-IDF value corresponding to the j-th type of text.

[0073] When TF-IDF ij When it is not 0, let TF-IDF ij For the triple L iThe average value of all corresponding non-zero weight values;

[0074]

[0075] In the formula For the triple L i The sum of all corresponding non-zero weight values;

[0076] By weighting the triples using formula (10), the original matching algorithm needs to be modified as follows:

[0077]

[0078] when hour,

[0079]

[0080] Where r a is the adjacent frequency corresponding to the ath triplet in the real-time alarm information, TF-IDF nj For L n The TF-IDF value corresponding to the j-th type of text;

[0081] when hour,

[0082]

[0083] Where M j is the equivalent adjacent total frequency when the real-time alarm information matches the jth template, ∑Sim'[A,B j ] is the sum of the weight values ​​when the real-time alarm information matches the jth template;

[0084] Since there are many triplets in the template, it is found during the research process that when the alarm information text is short, all its triplets may be included in multiple templates at the same time, resulting in the same matching similarity with multiple sets of templates. At this time, the sum of the weight values ​​corresponding to each template ∑Sim'[A,B j ] is used to judge. When the similarity with multiple templates is the same, the sum of the weight values ​​is compared, and the template with the largest sum of weight values ​​is output. The template type is the type of alarm information, and the objective function W of the matching algorithm is obtained:

[0085]

[0086] The depth-first algorithm is used to traverse the real-time alarm information knowledge graph, and the N groups of triples in the real-time alarm information knowledge graph are compared with all templates one by one. The specific method for calculating the similarity between the N groups of triples in the real-time alarm information knowledge graph and each template is as follows:

[0087] like Figure 3 As shown, starting from the starting node of the real-time fault alarm information knowledge graph, the adjacent nodes are continuously visited along a path of the real-time fault alarm information knowledge graph until there are no unvisited adjacent nodes, and then backtracking to the previous node, continue to visit other adjacent nodes until the entire real-time fault alarm information knowledge graph is visited. The node is an entity of the real-time fault alarm information knowledge graph, and the path refers to the path of the real-time fault alarm information knowledge graph. The depth-first algorithm steps are as follows:

[0088] Step 5.1, input the system name of the real-time fault alarm information knowledge graph into the set Z;

[0089] Step 5.2: Take out the first entity in set Z and set it to entity N, and delete the first entity in Z.

[0090] Step 5.3, let the ordered set R be an empty set;

[0091] Step 5.4, use the depth-first search method to search for an unsearched real-time fault alarm information knowledge graph path N downwards, and store all the similarities between the triples in the real-time fault alarm information knowledge graph contained therein and the matching of each template into the set P;

[0092] Step 5.5, continue searching for new paths in the real-time fault alarm information knowledge graph until an ordered set R obtained by arranging all entities on a path is found, all triples in the real-time fault alarm information knowledge graph in the set R are marked, and the set R is cleared;

[0093] Step 5.6: Continue to search for a new real-time fault alarm information knowledge graph path starting from entity N, return to step 5.3, and continue searching;

[0094] Step 5.7: When all real-time fault alarm information knowledge graph paths have been searched, return to step 5.2;

[0095] Step 5.8: When all systems are searched, the set Z is an empty set, and the similarity between the triple set in the knowledge graph of all real-time fault alarm information searched in the set P and the matching of each template is output.

[0096] A method for analyzing and locating alarms in a scheduling automation room equipment operation and maintenance. In step 6, the specific method for locating the faulty equipment according to the fault type of the real-time alarm information and the real-time fault alarm information of the scheduling automation room is:

[0097] The prediction function is used to perform similarity matching between each suspicious device and the fault type template corresponding to the alarm information, so as to more accurately determine the faulty device.

[0098] Step 6.1, define the estimation function Fore(x):

[0099]

[0100] Where x is the alarm information set containing suspicious devices, M j is the equivalent adjacent total frequency of the knowledge graph template obtained by formula (11). The prediction function is used to perform similarity matching between each suspicious device and the fault type template corresponding to the alarm information, so as to more accurately determine the faulty device.

[0101] Step 6.2, screening out suspicious devices a, b, c... according to the real-time alarm information of the dispatching automation room;

[0102] Step 6.3: Extract the row information a1, a2, ...a where the suspicious device is located n ,...c1,c2,...c n ,...and save as sub-alarm information set a={a1,a2,...a n}, ..., c = {c1, c2, ... c n},...;

[0103] Step 6.4: Different sub-alarm information sets are respectively introduced into the estimation function (16), and the suspicious device corresponding to the sub-alarm information set with the largest output of the estimation function is output as the faulty device.

[0104] An alarm analysis and positioning system for dispatching automation room equipment operation and maintenance Figure 4 As shown, it includes a knowledge graph template building module, a description word segmentation numbering module, a correlation matrix building module, a real-time knowledge graph building module, a fault type matching module and a fault equipment positioning module;

[0105] The knowledge graph template establishment module is used to establish an alarm information knowledge graph template based on the historical alarm information of the scheduling automation room;

[0106] The description word numbering module is used to extract the descriptive word of the real-time fault event in the real-time alarm information text of the dispatching automation room as an entity, and number the extracted real-time fault event description word according to the extraction order;

[0107] The correlation matrix establishment module is used to establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number;

[0108] The real-time knowledge graph establishment module is used to construct real-time fault event alarm information triples according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triples;

[0109] The fault type matching module is used to match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph;

[0110] The faulty equipment locating module is used to locate the faulty equipment according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

[0111] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for alarm analysis and positioning for scheduling automated computer room equipment operation and maintenance.

[0112] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

Claims

1. A method for analyzing and locating alarms for operation and maintenance of equipment in an automated computer room, characterized in that: It includes the following steps; Step 1: Establish an alarm information knowledge graph template based on the historical alarm information of the dispatching automation room; Step 2: extract the descriptive participles of the real-time fault events in the real-time alarm information text of the dispatching automation room as entities, and number them according to the extraction order of the extracted real-time fault event descriptive participles; Step 3: Establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number; Step 4: construct a real-time fault event alarm information triple according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triple; Step 5: Match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph; Step 6: Locate the faulty device according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

2. According to claim 1, the alarm analysis and positioning method for dispatching automation equipment room operation and maintenance is characterized in that: The specific method for establishing the alarm information knowledge graph template in step 1 is: Step 1.1, extract the descriptive participles of historical fault events in the text of the historical alarm information of the dispatching automation room as entities, and number them according to the extraction order of the descriptive participles of the historical fault events; Step 1.2: Establish a historical fault event correlation matrix based on the historical fault event description segmentations and corresponding numbers; Step 1.3: Construct historical fault event alarm information triples based on the historical fault event correlation matrix, and form a historical alarm information knowledge graph template based on the historical fault event alarm information triples.

3. The method for analyzing and locating alarms for operation and maintenance of equipment in an automated computer room according to claim 1, characterized in that: In the step 2, the real-time fault event descriptive participles in the real-time alarm information text of the dispatching automation room are extracted as entities, and the specific method for numbering the extracted real-time fault event descriptive participles in the order of extraction is: Step 2.1, based on the conventional dictionary and the Viterbi algorithm, the text of the real-time alarm information of the dispatching automation room is segmented; Step 2.2, perform deduplication filtering on the dispatch automation room real-time alarm information text after word segmentation, build a real-time alarm information fault event description word segmentation dictionary, and number the real-time fault event description word segments appearing in the dictionary in chronological order.

4. The alarm analysis and positioning method for dispatching automation room equipment operation and maintenance according to claim 1 is characterized in that: The specific method of establishing the real-time fault event description segmentation correlation matrix based on the real-time fault event description segmentation and the corresponding number in step 3 is: Combine the real-time fault event description segment words with different numbers in pairs, and use the numbers of the real-time fault event description segment words as the row number and column number of the matrix in order to construct the real-time fault event description segment word correlation matrix; A nn represents the real-time fault event description word segmentation correlation matrix, where the diagonal elements cor ii is 0; i,j=1,2,3,4,...,n, indicating the number of the fault event description word in the fault event description word dictionary, cor ij represents the number of times fault segmentation word j appears after fault segmentation word i. According to the characteristics of alarm information text data, it can be known that A nn is an asymmetric matrix; When operating, first initialize the matrix, that is, cor ij = 0, i, j = 1, 2, 3, 4, ..., n, traverse the alarm information text after word segmentation, when the fault description word numbered j appears once after the fault event description word numbered i, perform the following operation: THE nn =A' nn +E ij (2) Among them, E ij Indicates that the value of the i-th row and j-th column of the matrix is ​​1, as shown in Formula 3, A' nn Represents the last traversal result matrix, A nn Represents the result matrix of the current traversal; 5. The method for analyzing and locating alarms for dispatching automation room equipment operation and maintenance according to claim 1, characterized in that: In step 4, the specific method of constructing a real-time fault event alarm information triple according to the real-time fault event description word segmentation correlation matrix, and forming a real-time alarm information knowledge graph according to the real-time fault event alarm information triple is: The real-time fault event description word segmentation correlation matrix is ​​integrated to sort out the real-time fault event alarm information triple data, namely the head entity, the adjacent number relationship between entities, and the tail entity. The real-time fault event alarm information triple is input into the Neo4j graph database, and then the real-time alarm information knowledge graph is output.

6. The alarm analysis and positioning method for dispatching automation room equipment operation and maintenance according to claim 1 is characterized in that: The specific method of matching the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph in step 5 is: There are N sets of triple sequences (h, r, t) in the real-time alarm information knowledge graph, where h represents the head entity, t represents the tail entity, and r represents the adjacent frequency. The total adjacent frequency of alarm information is M: where r n is the adjacent frequency in the nth triplet, n = 1, 2, 3...N; The similarity function Sim is defined as: In the formula, Represents the ath triplet in the real-time alarm information, with the header entity being h a , the adjacent frequency is r a , the tail entity is t a ; Indicates the bth triple in the jth template, with the head entity h b , the adjacent frequency is r b , the tail entity is t b ; Where ε j is the similarity between the real-time alarm information and the jth template, M is the total frequency of adjacent real-time alarm information, ∑Sim[A,B j ] is the sum of similarity functions of all triplets in the real-time alarm information and all triplets in the j-th alarm information template; The depth-first algorithm is used to traverse the real-time alarm information knowledge graph, and the N groups of triples in the real-time alarm information knowledge graph are compared with all templates one by one. The similarity between the N groups of triples in the real-time alarm information knowledge graph and each template is calculated, and the template with the highest similarity is selected; In order to improve the accuracy of the matching algorithm, the TF-IDF method is used to improve the algorithm. TF represents the word rate, and IDF represents the inverse document frequency. The calculation formula is as follows: TF-IDF ij =TF ij ×IDF ij (9) In formula (7), formula (8) and formula (9), TF ij For the triple L i The word rate in the j-th category of text, N ij For the triple L i The number of times it appears in the jth type of text, N kj is the sum of the number of occurrences of triple k in the j-th text, is the sum of all triples appearing in the j-th category of text, IDF i For the triple L i The inverse text frequency, N is the number of all texts in the data set, n i is a triple L i Number of texts; TF-IDF ij For the triple L i The TF-IDF value corresponding to the j-th type of text; When TF-IDF ij When it is not 0, let it be a triplet L i The average of all corresponding non-zero weight values; In the formula For the triple L i The sum of all corresponding non-zero weight values; By weighting the triples using formula (10), the original matching algorithm needs to be modified as follows: when hour, Where r a is the adjacent frequency corresponding to the ath triplet in the real-time alarm information, TF-IDF nj For L n The TF-IDF value corresponding to the j-th type of text; when hour, Where M j is the equivalent adjacent total frequency when the real-time alarm information matches the jth template, ∑Sim'[A,B j ] is the sum of the weight values ​​when the real-time alarm information matches the jth template; When the real-time alarm information has the same similarity with multiple sets of template matches, the sum of the weight values ​​of the corresponding templates ∑Sim'[A,B j ] is judged, and the template with the largest sum of weight values ​​is output, thereby obtaining the objective function W of the matching algorithm: The template type corresponding to the objective function W is the fault type of the real-time alarm information.

7. The method for analyzing and locating alarms for operation and maintenance of equipment in an automated computer room according to claim 1, characterized in that: The depth-first algorithm is used to traverse the real-time alarm information knowledge graph, and the N groups of triples in the real-time alarm information knowledge graph are compared with all templates one by one. The specific method for calculating the similarity between the N groups of triples in the real-time alarm information knowledge graph and each template is as follows: Step 5.1, input the system name of the real-time fault alarm information knowledge graph into the set Z; Step 5.2, take out the first entity in set Z, set it as entity N, and delete the first entity in Z; Step 5.3, let the ordered set R be an empty set; Step 5.4, use the depth-first search method to search for an unsearched real-time fault alarm information knowledge graph path N downwards, and store all the similarities between the triples in the real-time fault alarm information knowledge graph contained therein and the matching of each template into the set P; Step 5.5, continue searching for new paths in the real-time fault alarm information knowledge graph until an ordered set R obtained by arranging all entities on a path is found, all triples in the real-time fault alarm information knowledge graph in the set R are marked, and the set R is cleared; Step 5.6: Continue to search for a new real-time fault alarm information knowledge graph path starting from entity N, return to step 5.3, and continue searching; Step 5.7: When all real-time fault alarm information knowledge graph paths have been searched, return to step 5.2; Step 5.8: When all systems are searched, the set Z is an empty set, and the similarity between the triple set in the knowledge graph of all real-time fault alarm information searched in the set P and the matching of each template is output.

8. The method for analyzing and locating alarms for operation and maintenance of equipment in an automated computer room according to claim 1, characterized in that: The specific method of locating the faulty device according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room is as follows: Step 6.1, define the estimation function Fore(x): Where x is the alarm information set containing suspicious devices, M j is the equivalent adjacent total frequency of the knowledge graph template obtained by formula (11); Step 6.2, screening out suspicious devices a, b, c... according to the real-time alarm information of the dispatching automation room; Step 6.3: Extract the row information a1, a2, ...a where the suspicious device is located n ,...c1,c2,...c n ,...and save as sub-alarm information set a={a1,a2,...a n }, ..., c = {c1, c2, ... c n },...; Step 6.4: Different sub-alarm information sets are respectively introduced into the estimation function (16), and the suspicious device corresponding to the sub-alarm information set with the largest output of the estimation function is output as the faulty device.

9. An alarm analysis and positioning system for dispatching automated equipment room operation and maintenance, characterized by: It includes a knowledge graph template building module, a description word segmentation numbering module, a correlation matrix building module, a real-time knowledge graph building module, a fault type matching module and a fault equipment location module; The knowledge graph template establishment module is used to establish an alarm information knowledge graph template based on the historical alarm information of the scheduling automation room; The description word numbering module is used to extract the descriptive word of the real-time fault event in the real-time alarm information text of the dispatching automation room as an entity, and number the extracted real-time fault event description word according to the extraction order; The correlation matrix establishment module is used to establish a real-time fault event description word segmentation correlation matrix based on the real-time fault event description word segmentation and the corresponding number; The real-time knowledge graph establishment module is used to construct real-time fault event alarm information triples according to the real-time fault event description word segmentation correlation matrix, and form a real-time alarm information knowledge graph according to the real-time fault event alarm information triples; The fault type matching module is used to match the fault type of the real-time fault alarm information in the alarm information knowledge graph template according to the real-time alarm information knowledge graph; The faulty equipment locating module is used to locate the faulty equipment according to the fault type of the real-time alarm information and the real-time fault alarm information of the dispatching automation room.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images