A method, system and computer storage medium for fusion of power grid fault events

Abstract

The present invention relates to the field of power grid fault data processing, and more particularly to a power grid fault event fusion method, system, and computer storage medium. This method acquires multi-source heterogeneous data from a power system, standardizes fault events, and constructs a standardized event text structure. Original events are extracted from the data and integrated into a set of original event texts. Semantic parsing is performed on each original event text to extract the subject (equipment, component, or state quantity) and predicate (abnormal action or state change), which are then vectorized to obtain a dual-channel encoding vector. Based on this, a composite similarity function is constructed to generate a semantic similarity matrix between events. Similar event clusters are then merged from the bottom up to form a multi-granularity clustering result. A unique standardized event is assigned to each original event in the cluster to complete the fusion. The present invention accurately decouples the logical relationships between fault events, significantly improving semantic recognition accuracy and providing a highly interpretable decision-making basis for fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault data processing, and in particular to a power grid fault event fusion method, system and computer storage medium. Background Art

[0002] With the increasing development of intelligent power systems, the amount of fault event data generated during grid operation is growing exponentially. This data, including equipment alarm logs, sensor monitoring data, and manual maintenance reports, is multi-source, heterogeneous, semantically complex, and strongly correlated in time and space. Traditional fault event analysis methods rely heavily on manual experience or rule-based systems for event classification and correlation. These methods struggle to effectively mine the deep semantics of massive amounts of unstructured text data, resulting in inefficient fault root cause location and poor cross-system event fusion capabilities.

[0003] When it comes to event graph construction, existing technologies primarily focus on static entity relationship modeling, lacking sufficient causal logic analysis of dynamic event chains. As an advanced form of event graphs, causal graphs require precise depiction of the evolution of fault events and the laws of causal transmission. However, traditional methods suffer from numerous problems: Event chain extraction based on rule templates struggles to capture the complex and ever-changing fault scenarios in the power sector; and the event fusion process fails to consider the semantic hierarchical characteristics of the subject-predicate structure, resulting in high redundancy of event nodes in the graph, seriously impacting the accuracy of fault deduction and analysis.

[0004] Currently, power grid fault event processing technology faces the following bottlenecks: First, existing semantic clustering methods are mostly based on shallow text features, such as keyword matching and TF-IDF weight calculation, which cannot accurately grasp the contextual semantics of power professional terms and the logical connection between events. The differences in descriptions of the same equipment in different fault scenarios can easily lead to clustering bias. Second, the heterogeneity of multi-source event data makes it difficult for traditional methods to achieve efficient cross-modal information fusion, resulting in information fragmentation problems in event correlation analysis. In addition, there is a lack of effective modeling methods for the dynamic association of fault events with power grid topology and equipment knowledge bases, which limits the real-time assessment of the impact range of events and collaborative disposal decisions.

[0005] In recent years, large language models have demonstrated significant advantages in understanding textual semantics, but their application in the field of power failure remains limited. For one thing, general-purpose pre-trained models are insufficiently adaptable to specialized knowledge about power grids, resulting in low accuracy in event entity recognition and relationship extraction. Furthermore, existing methods fail to fully exploit the semantic spatial characteristics of large language models and fail to deeply collaborate with clustering algorithms, hindering their ability to discover complex event patterns. Therefore, combining the semantic representation capabilities of large language models with domain knowledge about power grid failures to construct an efficient event fusion framework has become a key challenge in improving the intelligent diagnosis of power system faults. Summary of the Invention

[0006] In order to solve the accuracy and logical consistency problems of event fusion in the construction of power grid fault event maps, as well as the misjudgment of semantically equivalent events due to differences in subject-predicate structures in power equipment fault texts, the present invention provides a power grid fault event fusion method, system and computer storage medium. Through the structured semantic parsing and hierarchical clustering mechanism driven by a large language model, the subject-predicate semantic decoupling and precise fusion of power fault events are achieved, the event correlation and domain adaptability of the event map are improved, and technical support is provided for tracing the root cause of power grid faults. Specifically, the method includes the following steps:

[0007] S1: Acquire multi-source heterogeneous data of the power system, standardize the definition of power fault events, construct a standardized event text structure, extract original events from the multi-source heterogeneous data, and integrate them according to the standardized event text structure to form an original event text set;

[0008] S2: Based on the original event set, semantically parse each original event text to extract the subject and predicate of the event, where the subject includes the equipment, components, or state quantities involved in the fault event, and the predicate includes the abnormal action or state change of the fault event;

[0009] S3: Vectorize the subject and predicate of the event to obtain dual-channel encoding vectors of the subject and predicate;

[0010] S4: calculating the similarity of the dual-channel encoding vectors of the subject and the predicate, constructing a composite similarity function according to the similarity, and calculating the semantic similarity matrix between events based on the composite similarity function;

[0011] S5: merging similar event clusters from bottom to top based on the semantic similarity matrix, and generating multi-granularity clustering results according to a preset clustering granularity level;

[0012] S6: Based on the multi-granularity clustering result, generate multiple candidate standardized events for each cluster, assign a unique standardized event to each original event in the cluster from the multiple candidate standardized events, and complete the power grid fault event fusion.

[0013] In one embodiment of the present invention, in S5, the method for generating multi-granularity clustering results according to the preset clustering granularity level is as follows:

[0014] S51: Use the dual-channel encoding vectors of the subject and predicate extracted in S3 to update the original event text set to obtain the updated fault event set , where each event is the vector representation of the subject and predicate after dual-channel semantic encoding; and, according to the composite similarity function, the semantic similarity matrix between events in the updated fault event set is calculated. , where the matrix elements Representing an event and events The semantic similarity between

[0015] S52: Cluster each of the updated fault events separately to construct an initial cluster set ,The number of clusters is equal to the number of failure events;

[0016] S53: Iterate the calculation starting from the initial cluster set to calculate the inter-cluster similarity matrix at the current time t , for cluster and clusters , use the following formula to calculate the similarity between them :

[0017] ,in, and Represent clusters and clusters The number of events in Represents the event calculated according to the composite similarity function and events similarity between

[0018] S54: Based on the inter-cluster similarity matrix , get the current cluster similarity The largest cluster pair ,Right now: ;

[0019] S55: Determine whether the number of clusters after the current iterative update reaches the target value K or the maximum inter-cluster similarity is less than or equal to the similarity threshold Conditions:

[0020] If not, return to step S53;

[0021] If yes, compare the inter-cluster similarity with the preset similarity threshold If the inter-cluster similarity is greater than the similarity threshold , then the cluster and clusters Merge into a new cluster , and update the cluster set , Indicates a removal operation;

[0022] S56: According to the preset clustering granularity level, gradually construct a dendrogram for each iterative merged cluster, cut the dendrogram, and obtain a multi-granularity clustering result .

[0023] In one embodiment of the present invention, in S56, according to the preset clustering granularity level, a dendrogram is gradually constructed for each iterative merged cluster, and the dendrogram is cut to obtain a multi-granularity clustering result. The method is as follows:

[0024] The initial single event cluster is used as the root node, the new clusters generated by each merging are used as child nodes, and the similarity during merging is used as the node height, and a dendrogram is gradually constructed. Each node of the dendrogram represents a cluster, and the height of the node represents the similarity during merging.

[0025] According to the preset clustering granularity level, the corresponding cutting height in the dendrogram is determined; horizontal cutting is performed at the specified height of the dendrogram, and the dendrogram is divided into multiple subtrees according to the intersection of the cutting line and the dendrogram branches. Each subtree corresponds to a clustering result. By cutting at different heights, multi-granularity clustering results are obtained. .

[0026] In one embodiment of the present invention, in S1, a set of original event texts is formed. The method is as follows:

[0027] The equipment, components, or state quantities involved in the fault event are defined as the subject, and the abnormal action or state change of the fault event is defined as the predicate. A binary array is constructed based on the subject and the predicate to represent the text structure of the fault event, as follows:

[0028] ,

[0029] Extract original events from the multi-source heterogeneous data and integrate them according to the standardized event text structure to form an original event text set :

[0030] ,in, Represents the subject array, represents an array of predicates, represents the original event set of the k-th data source, represents the kth data source, k=1,...,K, and K represents the total amount of the multi-source heterogeneous data.

[0031] In one embodiment of the present invention, in S2, the method for extracting the subject and predicate of the event is as follows:

[0032] Construct an event parsing function: , where e represents the fault event, s is the subject, and p is the predicate;

[0033] The event parsing function is guided by text instructions to perform semantic parsing on each original event text, output a fault event text in JSON format, and extract the subject and predicate of the event from the fault event text in JSON format.

[0034] In one embodiment of the present invention, in S2, the method for extracting the subject and predicate of an event further includes: using a regular expression to make the parsing result of the event parsing function completely conform to the JSON format:

[0035] ,

[0036] in, Indicates output data, Represents a regular expression, JSON represents data consisting of attribute-value pairs, Represents a constraint.

[0037] In one embodiment of the present invention, in S3, the method for obtaining the dual-channel encoding vectors of the subject and the predicate is as follows:

[0038] Perform dual-channel encoding of structured events, semantically embed the parsed subject s and predicate p respectively, and use the pre-trained SBERT model to encode the structured input. The subject s and the predicate p share the same encoder to ensure vector space alignment, and generate dual-channel encoding vectors of the subject and predicate:

[0039] , ,

[0040] in, represents the subject channel encoding vector, represents the predicate channel encoding vector, Represents the dimension, and d is the output dimension size of the SBERT model.

[0041] In one embodiment of the present invention, in S4, the method for constructing a composite similarity function based on the similarity is as follows:

[0042] Calculating failure events and fault events The weighted cosine similarity of the subject and predicate dual-channel encoding vectors is calculated, and a composite similarity function is constructed based on the weighted cosine similarity. :

[0043] ,

[0044] Among them, α∈[0,1] is the adjustable weight coefficient, Indicates a fault event The subject channel encoding vector, Indicates a fault event The subject channel encoding vector, Indicates a fault event The predicate channel encoding vector of Indicates a fault event The predicate channel encoding vector of .

[0045] Based on the same inventive concept, the present invention further provides a power grid fault event fusion system for implementing the steps of the power grid fault event fusion method. The power grid fault event fusion system includes the following modules:

[0046] A data standardization module is used to obtain multi-source heterogeneous data of the power system, standardize the definition of power fault events, build a standardized event text structure, extract original events from the multi-source heterogeneous data, and integrate them according to the standardized event text structure to form an original event text set;

[0047] An event semantic parsing module, configured to perform semantic parsing on each original event text based on the original event set, and extract the subject and predicate of the event, wherein the subject includes the equipment, component, or state quantity involved in the fault event, and the predicate includes the abnormal action or state change of the fault event;

[0048] The event vectorization representation module is used to vectorize the subject and predicate of the event respectively, and obtain the dual-channel encoding vectors of the subject and predicate;

[0049] A semantic similarity calculation module, configured to calculate the similarity of the dual-channel encoding vectors of the subject and the predicate, construct a composite similarity function based on the similarity, and calculate a semantic similarity matrix between events based on the composite similarity function;

[0050] A multi-granularity clustering result acquisition module is used to merge similar event clusters from bottom to top based on the semantic similarity matrix and generate multi-granularity clustering results according to a preset clustering granularity level;

[0051] An event fusion module is used to generate multiple candidate standardized events for each cluster based on the multi-granularity clustering results, and assign a unique standardized event to each original event in the cluster from the multiple candidate standardized events to complete power grid fault event fusion.

[0052] The present invention also provides a computer storage medium storing a computer software product. The computer software product includes several instructions for enabling a computer device to execute the power grid fault event fusion method.

[0053] The above technical solution of the present invention has the following advantages over the prior art:

[0054] 1. Accurate semantic parsing capability: Structured semantic parsing technology based on a large language model, combined with the characteristics of domain data, effectively decouples the subject-predicate logical relationship of fault events, significantly improving the semantic recognition accuracy of device entities and state actions, and overcoming the limitations of traditional methods in capturing the contextual associations of professional terms.

[0055] 2. Multi-dimensional semantic representation optimization: Using dual-channel semantic encoding technology, independent vector spaces are constructed for device entities and state actions respectively. Through the fusion of multi-dimensional semantic features, the differentiated representation capability of "synonymous and heterogeneous" texts is enhanced, providing more accurate semantic support for event correlation analysis.

[0056] 3. Intelligent event fusion architecture: This architecture integrates hierarchical clustering and generative language modeling technologies, automatically divides event clusters through multi-granularity semantic analysis, and generates standardized event descriptions that comply with industry standards. This significantly improves the efficiency of event graph construction and the systematic nature of knowledge accumulation, providing highly interpretable decision-making basis for fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0058] Figure 1 This is a flow chart of a method for merging power grid fault events provided in an embodiment of the present invention;

[0059] Figure 2 It is a schematic diagram of a process for generating multi-granularity clustering results provided in an embodiment of the present invention;

[0060] Figure 3 1 is a schematic diagram of the structure of a power grid fault event fusion system provided in an embodiment of the present invention;

[0061] Explanation of the accompanying figures in the specification: 100, data standardization module; 200, event semantic analysis module; 300, event vectorization representation module; 400, semantic similarity calculation module; 500, multi-granularity clustering result acquisition module; 600, event fusion module. DETAILED DESCRIPTION

[0062] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0063] Example 1:

[0064] See also Figure 1 As shown, the present invention provides a method for merging power grid fault events, which includes the following steps:

[0065] S1: Standardize the definition of power failure events and build a standardized event text structure. Collect and integrate raw events from various data sources in the power system (such as equipment alarm logs, sensor monitoring records, manual maintenance reports, etc.) to form a collection of raw event texts, providing a unified format data foundation for subsequent processing.

[0066] S2: Based on the original event text collection, a large language model is used to perform semantic analysis on each original event text, accurately extracting the subject (i.e., the equipment, component, or state quantity involved in the fault) and predicate (such as state change, abnormal action, etc.) in the event, and constructing standardized semantic features to prepare for subsequent precise clustering analysis;

[0067] S3: Using dual-channel semantic encoding technology, we vectorize the subject and predicate of an event, obtain dual-channel encoding vectors for the subject and predicate, and then construct a multi-dimensional semantic feature space for the event, reflecting the semantic information of the event more comprehensively and accurately.

[0068] S4: Calculate the similarity between the dual-channel encoding vectors of the subject and predicate, and construct a composite similarity function based on the dynamic weight allocation mechanism. Based on this composite similarity function, calculate the semantic similarity between events, generate a semantic similarity matrix, and quantify the degree of semantic association between events;

[0069] S5: A hierarchical clustering algorithm is used to iteratively merge similar event clusters in a bottom-up manner based on the semantic similarity matrix. The clustering granularity is controlled and adjusted according to the preset level to generate clustering results with multi-granularity characteristics, providing a reasonable classification basis for event fusion.

[0070] S6: Based on the multi-granularity clustering results, a large language model is used to generate multiple candidate standardized events for each cluster. Through the semantic matching algorithm, a unique standardized event is accurately assigned to each original event within the cluster from these candidate standardized events, completing the fusion of power grid fault events to meet actual application needs.

[0071] Furthermore, in S1, the original event text set is formed The method is as follows:

[0072] The equipment, components, or state quantities involved in the fault event are defined as the subject, and the abnormal action or state change of the fault event is defined as the predicate. A binary array is constructed based on the subject and the predicate to represent the text structure of the fault event, as follows:

[0073] ,

[0074] Extract original events from the multi-source heterogeneous data and integrate them according to the standardized event text structure to form an original event text set :

[0075] ,in, Represents the subject array, represents an array of predicates, represents the original event set of the k-th data source, represents the kth data source, k=1,...,K, and K represents the total amount of the multi-source heterogeneous data.

[0076] Furthermore, in S2, the method for extracting the subject and predicate of the event is as follows:

[0077] Construct an event parsing function: , where e represents the fault event, s is the subject, and p is the predicate;

[0078] The event parsing function is guided by the text instruction Prompt to perform semantic parsing on each original event text, output the fault event text in JSON format, and extract the subject and predicate of the event from the fault event text in JSON format.

[0079] Assume that the dedicated prompt is designed as follows:

[0080] You are an expert at extracting the subject and predicate of power grid fault events. Given an event, identify the subject and predicate, and return a collection of these elements. The subject of an event is the device or state that represents the event, and the predicate is the abnormal action or state change that represents the event. Here are some examples:

[0081] Input: Deflector aging

[0082] Output: {"event": "Aging of guide vane", "subject": "guide vane", "predicate": "Aging"}".

[0083] Specifically, in S2, based on the original event set, semantically parsing each original event text to extract the subject and predicate of the event further includes: using regular expressions to make the parsing result of the event parsing function fully conform to the JSON format to ensure robustness:

[0084] ,

[0085] in, Indicates output data, Represents a regular expression, JSON represents data consisting of attribute-value pairs, Indicates the constraint item. The above formula indicates that the input JSON data (e) is processed according to the constraint item. Validates or processes the regular expression rules defined in and returns the result.

[0086] Furthermore, in S3 of this embodiment, the subject and predicate of the event are respectively vectorized to obtain dual-channel encoding vectors of the subject and predicate as follows:

[0087] Perform dual-channel encoding of structured events, semantically embed the parsed subject s and predicate p respectively, and use the pre-trained SBERT model to encode the structured input. The subject s and the predicate p share the same encoder to ensure vector space alignment, and generate dual-channel encoding vectors of the subject and predicate:

[0088] , ,

[0089] in, represents the subject channel encoding vector, represents the predicate channel encoding vector, represents the dimension, d is the output dimension size of the SBERT model, and is usually set to 768.

[0090] Furthermore, in S4, the dual-channel encoding vectors of the subject and predicate are calculated and The method for constructing a composite similarity function based on the similarity is as follows:

[0091] Calculating failure events and fault events The weighted cosine similarity of the subject and predicate dual-channel encoding vectors is calculated, and a composite similarity function is constructed based on the weighted cosine similarity. :

[0092] ,

[0093] Among them, α∈[0,1] is an adjustable weight coefficient, which dynamically adjusts the contribution ratio of the subject and predicate through domain knowledge (such as equipment criticality). Indicates a fault event The subject channel encoding vector, Indicates a fault event The subject channel encoding vector, Indicates a fault event The predicate channel encoding vector of Indicates a fault event The predicate channel encoding vector of .

[0094] Further, see Figure 2 As shown, in S5, the method for generating multi-granularity clustering results according to the preset clustering granularity level is as follows:

[0095] S51: Use the dual-channel encoding vectors of the subject and predicate extracted in S3 to update the original event text set to obtain the updated fault event set , where each event is the vector representation of the subject and predicate after dual-channel semantic encoding; and, according to the composite similarity function, the semantic similarity matrix between events in the updated fault event set is calculated. , where the matrix elements Representing an event and events The semantic similarity between

[0096] S52: Cluster each of the updated fault events separately to construct an initial cluster set ,The number of clusters is equal to the number of failure events;

[0097] S53: Iterate the calculation starting from the initial cluster set to calculate the inter-cluster similarity matrix at the current time t , for cluster and clusters , use the following formula to calculate the similarity between them :

[0098] ,in, and Represent clusters and clusters The number of events in Represents the event calculated according to the composite similarity function and events similarity between

[0099] S54: Based on the inter-cluster similarity matrix , get the current cluster similarity The largest cluster pair ,Right now: ;

[0100] S55: Determine whether the number of clusters after the current iterative update reaches the target value K or the maximum inter-cluster similarity is less than or equal to the similarity threshold Conditions:

[0101] If not, return to step S53;

[0102] If yes, compare the inter-cluster similarity with the preset similarity threshold If the inter-cluster similarity is greater than the similarity threshold , then the cluster and clusters Merge into a new cluster , and update the cluster set , Indicates a removal operation;

[0103] S56: According to the preset clustering granularity level, gradually construct a dendrogram for each iterative merged cluster, cut the dendrogram, and obtain a multi-granularity clustering result .

[0104] Furthermore, in S56, according to the preset clustering granularity level, a dendrogram is gradually constructed for each iteratively merged cluster, and the dendrogram is cut to obtain a multi-granularity clustering result. The method is as follows:

[0105] The initial single event cluster is used as the root node, the new clusters generated by each merging are used as child nodes, and the similarity during merging is used as the node height, and a dendrogram is gradually constructed. Each node of the dendrogram represents a cluster, and the height of the node represents the similarity during merging.

[0106] According to the preset clustering granularity level, the corresponding cutting height in the dendrogram is determined. The higher the cutting height, the fewer clusters are generated and the coarser the granularity; the lower the cutting height, the more clusters are generated and the finer the granularity.

[0107] Perform horizontal cutting at the specified height of the dendrogram, and divide the dendrogram into multiple subtrees according to the intersection of the cutting line and the dendrogram branches. Each subtree corresponds to a clustering result. By cutting at different heights, multi-granularity clustering results are obtained. .

[0108] Event Collection For example, through the iterative merging process, merge and , forming clusters ,merge and , forming clusters ,merge and , forming the final cluster .

[0109] If the cut is made at a height of 2, the resulting cluster division is: ; If the cut is made at a height of 1, the resulting cluster division is: ; If the cut is made at a height of 3, the resulting cluster division is: .

[0110] Furthermore, in S6, based on the multi-granularity clustering results, multiple candidate standardized events are generated for each cluster, and a unique standardized event is assigned to each original event in the cluster from the multiple candidate standardized events to complete the power grid fault event fusion. The specific method is as follows:

[0111] The final cluster division results obtained by S5 hierarchical clustering , each cluster Contains semantically similar events;

[0112] For each cluster , design a special text instruction prompt to guide the large language model to generate standardized event descriptions. The prompt example is as follows:

[0113] You are an expert in fusing power equipment failure events. I am building a graph of power equipment failure events, and you will help me group multiple events in an event cluster into standardized events. These events may be similar, or they may share certain features and semantics. Your task is to group them into one or more standardized events. Each standardized event description should contain as much key information as possible to ensure clear meaning. The following is a list of events in the event cluster:

[0114] Event 1: [Detailed description of Event 1]

[0115] Event 2: [Detailed description of Event 2] ...

[0116] Please generate n candidate standardized event descriptions (n ≥ 1), each of which should be concise and include key information.

[0117] Input the above prompt into the large language model to generate n candidate normalization events (n ≥ 1): .

[0118] For each original event and each candidate normalization event , m=1,...,n, use the pre-trained SBERT model for semantic embedding to obtain the semantic vector and , calculate for each original event e and each candidate standardized event The semantic similarity between them is measured using the cosine similarity formula: ;

[0119] For each original event , select the candidate standardized event with the greatest semantic similarity as the only standardized event of the event, that is: ;

[0120] Assign each original event e its corresponding unique standardized event , forming a semantic mapping relationship M: M→ .

[0121] For each cluster , outputs a standardized event set after semantic mapping The standardized event descriptions within each cluster are consistent and concise, facilitating subsequent fault diagnosis and event graph construction. Thus, the fusion of power grid fault events is complete.

[0122] Example 2:

[0123] Based on the same inventive concept as that of the first embodiment, the present invention also provides a power grid fault event fusion system for implementing the steps of the power grid fault event fusion method described in the first embodiment. Figure 3 As shown, the power grid fault event fusion system includes the following modules:

[0124] The data standardization module 100 is used to obtain multi-source heterogeneous data of the power system, standardize the definition of power fault events, construct a standardized event text structure, extract original events from the multi-source heterogeneous data, and integrate them according to the standardized event text structure to form an original event text set;

[0125] An event semantic parsing module 200 is configured to perform semantic parsing on each original event text based on the original event set, extracting the subject and predicate of the event, wherein the subject includes the equipment, component, or state quantity involved in the fault event, and the predicate includes the abnormal action or state change of the fault event;

[0126] An event vectorization representation module 300 is used to vectorize the subject and predicate of an event to obtain dual-channel encoding vectors of the subject and predicate;

[0127] A semantic similarity calculation module 400 is used to calculate the similarity of the dual-channel encoding vectors of the subject and the predicate, construct a composite similarity function based on the similarity, and calculate the semantic similarity matrix between events based on the composite similarity function;

[0128] A multi-granularity clustering result acquisition module 500 is used to merge similar event clusters from bottom to top based on the semantic similarity matrix and generate a multi-granularity clustering result according to a preset clustering granularity level;

[0129] The event fusion module 600 is used to generate multiple candidate standardized events for each cluster based on the multi-granularity clustering results, and assign a unique standardized event to each original event in the cluster from the multiple candidate standardized events to complete power grid fault event fusion.

[0130] A power grid fault event fusion system proposed in this embodiment is used to implement the aforementioned power grid fault event fusion method. Therefore, the specific implementation methods of the power grid fault event fusion system can be seen in the embodiment part of the aforementioned power grid fault event fusion method. For example, the data standardization module 100, the event semantic analysis module 200, the event vectorization representation module 300, the semantic similarity calculation module 400, the multi-granularity clustering result acquisition module 500 and the event fusion module 600 are respectively used to implement steps S1, S2, S3, S4, S5 and S6 in the power grid fault event fusion method in Example 1. Therefore, its specific implementation methods can refer to the descriptions of the corresponding embodiments of each part. In order to avoid redundancy, they will not be repeated here.

[0131] Example 3:

[0132] The present invention also provides a computer storage medium storing a computer software product. The computer software product includes several instructions for enabling a computer device to execute the power grid fault event fusion method described in the first embodiment.

[0133] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0137] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for fusion of power grid fault events, characterized in that: The following steps are involved: S1: Acquire multi-source heterogeneous data of the power system, standardize the definition of power fault events, construct a standardized event text structure, extract original events from the multi-source heterogeneous data, and integrate them according to the standardized event text structure to form an original event text set; S2: Based on the original event set, semantically parse each original event text to extract the subject and predicate of the event, where the subject includes the equipment, components, or state quantities involved in the fault event, and the predicate includes the abnormal action or state change of the fault event; S3: Vectorize the subject and predicate of the event to obtain dual-channel encoding vectors of the subject and predicate; S4: calculating the similarity of the dual-channel encoding vectors of the subject and the predicate, constructing a composite similarity function according to the similarity, and calculating the semantic similarity matrix between events based on the composite similarity function; S5: merging similar event clusters from bottom to top based on the semantic similarity matrix, and generating multi-granularity clustering results according to a preset clustering granularity level; S6: Based on the multi-granularity clustering result, generate multiple candidate standardized events for each cluster, assign a unique standardized event to each original event in the cluster from the multiple candidate standardized events, and complete the power grid fault event fusion.

2. The power grid fault event fusion method according to claim 1, characterized in that: In S5, the method for generating multi-granularity clustering results according to the preset clustering granularity level is as follows: S51: Use the dual-channel encoding vectors of the subject and predicate extracted in S3 to update the original event text set to obtain the updated fault event set , where each event is the vector representation of the subject and predicate after dual-channel semantic encoding; and, according to the composite similarity function, the semantic similarity matrix between events in the updated fault event set is calculated. , where the matrix elements Representing an event and events The semantic similarity between S52: Cluster each of the updated fault events separately to construct an initial cluster set ,The number of clusters is equal to the number of failure events; S53: Iterate the calculation starting from the initial cluster set to calculate the inter-cluster similarity matrix at the current time t , for cluster and clusters , use the following formula to calculate the similarity between them : ,in, and Represent clusters and clusters The number of events in Represents the event calculated according to the composite similarity function and events similarity between S54: Based on the inter-cluster similarity matrix , get the current cluster similarity The largest cluster pair ,Right now: ; S55: Determine whether the number of clusters after the current iterative update reaches the target value K or the maximum inter-cluster similarity is less than or equal to the similarity threshold Conditions: If not, return to step S53; If yes, compare the inter-cluster similarity with the preset similarity threshold If the inter-cluster similarity is greater than the similarity threshold , then the cluster and clusters Merge into a new cluster , and update the cluster set , Indicates a removal operation; S56: According to the preset clustering granularity level, gradually construct a dendrogram for each iterative merged cluster, cut the dendrogram, and obtain a multi-granularity clustering result .

3. The power grid fault event fusion method according to claim 2, characterized in that: In S56, according to the preset clustering granularity level, a dendrogram is gradually constructed for each iterative merged cluster, and the dendrogram is cut to obtain a multi-granularity clustering result. The method is as follows: The initial single event cluster is used as the root node, the new clusters generated by each merging are used as child nodes, and the similarity during merging is used as the node height, and a dendrogram is gradually constructed. Each node of the dendrogram represents a cluster, and the height of the node represents the similarity during merging. According to the preset clustering granularity level, the corresponding cutting height in the dendrogram is determined; horizontal cutting is performed at the specified height of the dendrogram, and the dendrogram is divided into multiple subtrees according to the intersection of the cutting line and the dendrogram branches. Each subtree corresponds to a clustering result. By cutting at different heights, multi-granularity clustering results are obtained. .

4. The power grid fault event fusion method according to claim 1, characterized in that: In S1, a collection of original event texts is formed The method is as follows: The equipment, components, or state quantities involved in the fault event are defined as the subject, and the abnormal action or state change of the fault event is defined as the predicate. A binary array is constructed based on the subject and the predicate to represent the text structure of the fault event, as follows: ， Extract original events from the multi-source heterogeneous data and integrate them according to the standardized event text structure to form an original event text set : ,in, Represents the subject array, represents an array of predicates, represents the original event set of the k-th data source, represents the kth data source, k=1,...,K, and K represents the total amount of the multi-source heterogeneous data.

5. The power grid fault event fusion method according to claim 1, characterized in that: In S2, the method for extracting the subject and predicate of an event is as follows: Construct an event parsing function: , where e represents the fault event, s is the subject, and p is the predicate; The event parsing function is guided by text instructions to perform semantic parsing on each original event text, output a fault event text in JSON format, and extract the subject and predicate of the event from the fault event text in JSON format.

6. The power grid fault event fusion method according to claim 5, characterized in that: In S2, the method for extracting the subject and predicate of the event further includes: using a regular expression to make the parsing result of the event parsing function completely conform to the JSON format: ， in, Indicates output data, Represents a regular expression, JSON represents data consisting of attribute-value pairs, Represents a constraint.

7. The power grid fault event fusion method according to claim 1, characterized in that: In S3, the method for obtaining the dual-channel encoding vectors of the subject and predicate is as follows: Perform dual-channel encoding of structured events, semantically embed the parsed subject s and predicate p respectively, and use the pre-trained SBERT model to encode the structured input. The subject s and the predicate p share the same encoder to ensure vector space alignment, and generate dual-channel encoding vectors of the subject and predicate: ， ， in, represents the subject channel encoding vector, represents the predicate channel encoding vector, Represents the dimension, and d is the output dimension size of the SBERT model.

8. The power grid fault event fusion method according to claim 1, characterized in that: In S4, the method for constructing a composite similarity function based on the similarity is as follows: Calculating failure events and fault events The weighted cosine similarity of the subject and predicate dual-channel encoding vectors is calculated, and a composite similarity function is constructed based on the weighted cosine similarity. : ， Among them, α∈[0,1] is the adjustable weight coefficient, Indicates a fault event The subject channel encoding vector, Indicates a fault event The subject channel encoding vector, Indicates a fault event The predicate channel encoding vector of Indicates a fault event The predicate channel encoding vector of .

9. A power grid fault event fusion system, characterized in that: For implementing the steps of the power grid fault event fusion method according to any one of claims 1 to 8, the power grid fault event fusion system comprises the following modules: A data standardization module is used to obtain multi-source heterogeneous data of the power system, standardize the definition of power fault events, build a standardized event text structure, extract original events from the multi-source heterogeneous data, and integrate them according to the standardized event text structure to form an original event text set; An event semantic parsing module, configured to perform semantic parsing on each original event text based on the original event set, and extract the subject and predicate of the event, wherein the subject includes the equipment, component, or state quantity involved in the fault event, and the predicate includes the abnormal action or state change of the fault event; The event vectorization representation module is used to vectorize the subject and predicate of the event respectively, and obtain the dual-channel encoding vectors of the subject and predicate; A semantic similarity calculation module, configured to calculate the similarity of the dual-channel encoding vectors of the subject and the predicate, construct a composite similarity function based on the similarity, and calculate a semantic similarity matrix between events based on the composite similarity function; A multi-granularity clustering result acquisition module is used to merge similar event clusters from bottom to top based on the semantic similarity matrix and generate multi-granularity clustering results according to a preset clustering granularity level; An event fusion module is used to generate multiple candidate standardized events for each cluster based on the multi-granularity clustering results, and assign a unique standardized event to each original event in the cluster from the multiple candidate standardized events to complete power grid fault event fusion.

10. A computer storage medium, characterized in that The computer storage medium stores a computer software product, and the computer software product includes several instructions for enabling a computer device to execute the power grid fault event fusion method according to any one of claims 1 to 8.

Images