Cross-domain collaborative information tracking and positioning auxiliary method and system

Abstract

The invention provides a cross-domain collaborative information tracking and positioning auxiliary method and system, and the method comprises the steps: firstly collecting an original data set generated by a target event in a plurality of associated domains, carrying out the cross-domain identifier matching processing to obtain a standardized event data set, and then carrying out the information association processing of the standardized event data set, according to the method, context semantic features and cross-domain time sequence behavior features of a target event are generated, then collaborative analysis is performed on the features based on a preset collaborative analysis model, a cross-domain collaborative tracking index set is generated, and a dynamic tracking path topology is constructed according to the cross-domain collaborative tracking index set. According to the method and the device, the target event is tracked and positioned, the positioning auxiliary strategy matched with the target event is generated, and finally the positioning auxiliary strategy is synchronized to the cross-domain tracking system to activate the cooperative positioning operation, so that the cooperative processing of multi-domain information is realized, and the target event tracking and positioning accuracy and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a cross-domain collaborative information tracking and positioning auxiliary method and system. Background Art

[0002] In today's complex social environment, target events often do not exist in isolation in a single field, but will have a wide range of impacts across multiple related fields. For example, in the power grid scenario, power grid failure events may involve multiple fields at the same time, and different fields will generate various types of raw data. These data formats, identifiers, and semantics are different. At present, when processing event tracking and positioning involving multi-field information, the data in each field are usually in a relatively independent state, lacking an effective integration and collaborative analysis mechanism. Existing technologies make it difficult to accurately associate and uniformly process the raw data related to target events generated in different fields, resulting in an inability to fully and deeply understand the comprehensive impact and development trend of target events in multiple fields, and thus making it difficult to accurately track target events and formulate effective positioning strategies, which seriously affects the ability to respond to complex events and the efficiency of processing. Summary of the Invention

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a cross-domain collaborative information tracking and positioning assistance method, the method comprising: Collecting the original data sets generated by the target event in multiple related fields, and performing cross-field identifier matching processing on the original data sets to obtain a standardized event data set; Performing information association processing on the standardized event data set to generate contextual semantic features and cross-domain temporal behavior features of the target event; Based on a preset collaborative analysis model, the contextual semantic features and the cross-domain temporal behavior features are collaboratively analyzed to generate a set of cross-domain collaborative tracking indicators; Constructing a dynamic tracking path topology according to the cross-domain collaborative tracking indicator set, and generating a positioning assistance strategy adapted to the dynamic tracking path topology; The positioning assistance strategy is synchronized to the cross-domain tracking system to activate the co-location operation.

[0004] On the other hand, an embodiment of the present invention also provides a cross-domain collaborative information tracking and positioning assistance system, including a processor and a machine-readable storage medium, the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0005] Based on the above aspects, the embodiment of the present invention collects the original data sets generated by the target event in multiple related fields and performs cross-domain identifier matching processing to obtain a standardized event data set, breaking down the barriers between data in different fields and achieving data integration and unified standardization. Information association processing is performed on the standardized event data set to generate the contextual semantic features and cross-domain temporal behavior features of the target event, which can comprehensively characterize the characteristics of the target event in multiple fields from the two key dimensions of semantics and temporal sequence. Based on the preset collaborative analysis model, the above features are collaboratively analyzed to generate a cross-domain collaborative tracking indicator set, which comprehensively considers the interaction of factors in multiple fields and makes the tracking indicators more comprehensive and accurate. Based on this indicator set, a dynamic tracking path topology is constructed and an adaptive positioning assistance strategy is generated, which can adjust the tracking and positioning methods in real time according to the dynamic changes of the target event in multiple fields. Finally, the positioning assistance strategy is synchronized to the cross-domain tracking system to activate the collaborative positioning operation, realizing the collaborative processing and efficient utilization of multi-domain information, greatly improving the accuracy and efficiency of tracking and positioning of target events, and enhancing the ability to deal with complex events. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a schematic diagram of the execution flow of the cross-domain collaborative information tracking and positioning assistance method provided by an embodiment of the present invention.

[0007] Figure 2 Schematic diagram of exemplary hardware and software components of a cross-domain collaborative information tracking and positioning assistance system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0008] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a cross-domain collaborative information tracking and positioning assistance method provided by an embodiment of the present invention. The cross-domain collaborative information tracking and positioning assistance method is introduced in detail below.

[0009] Step S110: collecting original data sets generated by target events in multiple related fields, and performing cross-field identifier matching processing on the original data sets to obtain standardized event data sets.

[0010] Taking the power grid scenario as an example, the target event is set as an abnormal harmonic interference event in the power grid system. This event involves multiple related areas, including the integration of new energy sources on the power generation side, monitoring of high-voltage transmission corridors, substation equipment operation and maintenance, the penetration of distributed power sources in low-voltage distribution networks, and the power consumption characteristics of large industrial users.

[0011] In the field of new energy access on the power generation side, as a large number of wind farms and solar photovoltaic power stations are connected to the power grid, raw data such as the pitch angle of wind turbines, the light intensity of photovoltaic panels, and the real-time fluctuation of new energy power generation are generated. The above data are combined into a set A, and the element a in set A represents a specific data record of new energy access on the power generation side.

[0012] In the field of high-voltage transmission corridor monitoring, in order to ensure the long-distance and efficient transmission of electricity, the contamination level of the insulators of the transmission lines, the dancing frequency of the conductors, the meteorological environment around the lines, etc. are monitored. These data form set B, where element b is a monitoring data record of a high-voltage transmission corridor.

[0013] In the field of substation equipment operation and maintenance, it is necessary to obtain real-time information such as the partial discharge of the transformer, the opening and closing time of the circuit breaker, and the temperature change of the busbar. These data constitute the set C, where element c is the substation equipment operation and maintenance data record.

[0014] In the field of distributed power penetration in low-voltage distribution networks, as more and more distributed power sources such as small wind turbines and household solar panels are connected to low-voltage distribution networks, data such as the access capacity of distributed power sources and the power exchange status with the main grid will be generated. This data is recorded as a set D, where element d represents the data record of distributed power penetration in low-voltage distribution networks.

[0015] In the area of ​​electricity consumption characteristics for large industrial users, data on the start-up and shutdown patterns of production equipment at large industrial enterprises, as well as peak and valley characteristics of electricity load, form set E, where element e is a record of electricity consumption characteristics for large industrial users. Aggregating sets A, B, C, D, and E forms the original data set F.

[0016] Data from different fields may use independent identifiers to identify the same or related information, which greatly complicates data integration and analysis. Therefore, it is necessary to perform cross-domain identifier matching on the original data set F, unifying the identifiers from different fields into standardized identifiers, thereby obtaining a standardized event data set.

[0017] Step S111: traverse each data record in the original data set, and extract the domain identifier and data format type of the data record.

[0018] In this embodiment, each data record in the original data set F is first traversed one by one. For element a in the data set A in the field of new energy access on the power generation side, its field identifier needs to be extracted when traversing the data record. The field identifier is used to accurately distinguish the specific field from which the data comes. For element a, its field identifier is "new energy access domain on the power generation side." At the same time, its data format type needs to be determined. Assuming that data set A is stored in an HBase distributed database, the data format type of element a is the HBase database format.

[0019] For element b in data set B for the high-voltage transmission corridor monitoring domain, during the traversal process, its domain identifier is extracted as "high-voltage transmission corridor monitoring domain." If data set B is transmitted and stored in real time using a Kafka message queue, the data format type of element b is the Kafka message format.

[0020] For element c in data set C related to substation equipment operation and maintenance, we traverse the data set and extract its domain identifier: "Substation Equipment Operation and Maintenance Domain." If data set C is stored in the InfluxDB time series database, the data format type of element c is the InfluxDB time series format.

[0021] For element d in data set D for the low-voltage distribution network distributed generation penetration domain, when traversing the data records, its domain identifier is extracted as "low-voltage distribution network distributed generation penetration domain." If data set D is stored and transmitted in batches as XML files, the data format type of element d is the XML file format.

[0022] For element e in the data set E of the large industrial user electricity consumption characteristics domain, after traversal, its domain identifier is extracted as "large industrial user electricity consumption characteristics domain". If the data set E is stored in CSV file format, the data format type of element e is CSV file format.

[0023] Step S112: matching the global unified identification code corresponding to the domain identifier according to a preset cross-domain mapping rule library, and converting the data format type into a target standardized format to obtain a converted data record.

[0024] The pre-defined cross-domain mapping rule base is a system of rules designed and maintained a priori. It establishes a precise correspondence between each domain identifier and the global unified identification code. This cross-domain mapping rule base can be stored in a relational database, such as MySQL, for easy query and update operations.

[0025] Once the domain identifier is extracted from the original data record, the corresponding global unified identification code can be found using the cross-domain mapping rule base. For example, for data element a in the power generation side new energy access domain, its domain identifier, "power generation side new energy access domain," is found to correspond to the global unified identification code "GRAE-001" by querying the cross-domain mapping rule base.

[0026] At the same time, different data formats need to be converted into a target standardized format. Assume that the target standardized format is Parquet, a columnar storage format for analytical services with efficient compression and query performance.

[0027] For element a in data set A stored in the HBase database format, a specialized conversion program must be developed. This program first reads the data for element a from the HBase database, parses it according to the column family and column structure, then reorganizes the data into a columnar structure conforming to the Parquet format. Finally, the converted data is stored in the Hadoop Distributed File System (HDFS).

[0028] For element b in data set B stored in the Kafka message format, you need to write a Kafka consumer program to consume messages for element b from the Kafka message queue. The program parses the message content, extracts key data fields, and then encodes and stores the data according to the Parquet format, also in HDFS.

[0029] For element c in data set C stored in the InfluxDB time series format, use the InfluxDB query interface to retrieve the time series data for element c. Process the query results, organize the timestamps and data values ​​according to the Parquet format, and finally store them in HDFS.

[0030] For element d in data set D stored in XML file format, we need to develop an XML parser to parse the data content in the XML file, extract and convert the XML tags and data values, construct a columnar data structure according to the Parquet format, and store it in HDFS.

[0031] For element e in data set E stored in CSV file format, use the Python Pandas library to read the CSV file. Clean and transform the data, removing unnecessary spaces and special characters. Then, save the data in Parquet format and store it in HDFS. This process yields the transformed data record.

[0032] Step S113: Perform redundancy check on the converted data records, delete duplicate records and invalid fields, and then aggregate the verified data records in layers according to timestamps and field identifiers to form the standardized event data set with a unified structure; perform integrity verification on the standardized event data set so that the data coverage of each associated field meets the preset threshold.

[0033] Step S113 - 1 : extracting a key field set of each converted data record, wherein the key field set includes an event identification code, a timestamp, a domain identifier, and a content summary.

[0034] For each converted data record, its key field set needs to be extracted. Taking the converted data records in the field of new energy access on the power generation side as an example, it is necessary to accurately extract the event identification code, which is a unique identifier of the specific event associated with the data and can accurately distinguish different events. The timestamp records the specific moment when the data was generated, reflecting the timeliness of the data. The domain identifier clearly defines the domain to which the data belongs, such as the global unified identification code "GRAE-001" corresponding to the "new energy access domain on the power generation side." The content summary is a summary of the main content of the data. For example, for data records related to power generation fluctuations, the content summary may be "abnormal power generation fluctuations during a certain period of time."

[0035] Step S113 - 2 : constructing a hash index based on the event identification code, and detecting duplicate records of the same event in different fields according to the field identifier.

[0036] A hash index is constructed based on the event identifier. This index can quickly locate and search for data records with the same event identifier. Different fields may record the same event, and the field identifier can be used to distinguish data from different fields. For example, the renewable energy access field on the power generation side and the high-voltage transmission corridor monitoring field may both record the event of renewable energy power generation fluctuations. The combination of the field identifier and the event identifier can accurately identify these duplicate records.

[0037] Step S113 - 3 : performing a timestamp comparison on the detected duplicate records, retaining the record with the latest timestamp and deleting the remaining records.

[0038] When duplicate records are detected, their timestamps need to be compared. Because the latest record often contains the most accurate and complete information, the record with the latest timestamp is retained and the remaining records are deleted. For example, if two duplicate records are detected regarding a wind turbine failure in the renewable energy access area on the power generation side, one with an older timestamp and the other with a newer timestamp, the record with the newer timestamp is retained and the older one is deleted.

[0039] Step S113 - 4 : Calculate semantic similarity of the content summary, and mark it as redundant content if the similarity exceeds a preset threshold.

[0040] The semantic similarity of content summaries can be calculated using methods such as the cosine similarity algorithm in natural language processing. If the similarity of the content summaries of two data records exceeds a preset threshold, for example, the preset threshold is 0.8, and when the calculated similarity of the content summaries of the two records reaches 0.85, one of them will be marked as redundant.

[0041] Step S113 - 5 : Merge or delete fields of redundant content according to the marking results to generate a streamlined data record set.

[0042] Based on the results marked as redundant, appropriate processing is performed. If two redundant records contain complementary field information—that is, one record contains useful information that the other lacks—then the fields can be merged. If the field information is highly redundant and lacks additional useful information, one of the records is deleted. This process generates a streamlined set of data records.

[0043] Step S113 - 6 : The verified data records are hierarchically aggregated according to timestamps and domain identifiers to form the standardized event data set with a unified structure.

[0044] Data is divided into different time intervals based on timestamps. Within each time interval, data is further grouped according to domain identifiers. For example, within a specific hour, data from different domains, such as new energy access on the power generation side and high-voltage transmission corridor monitoring, can be grouped separately. Aggregation operations, such as averaging, maximum, and minimum values, are then performed on the data within each group to form data records with a unified structure, thereby forming a standardized event data set.

[0045] Step S113-7: Perform integrity verification on the standardized event data set so that the data coverage of each related field meets a preset threshold.

[0046] Set a preset data coverage threshold for each relevant domain. Calculate data coverage by counting the number of data records and the time span for each relevant domain. For example, data coverage for renewable energy access on the power generation side may be required to reach a certain percentage within a certain time period. If data coverage for a particular relevant domain does not meet the preset threshold, further investigation is required to determine if there are any issues with the data collection process or if there is any data loss, and appropriate supplementation and corrections should be made.

[0047] Step S120: performing information association processing on the standardized event data set to generate contextual semantic features and cross-domain temporal behavior features of the target event.

[0048] Step S121: performing word segmentation processing on the text information in the standardized event data set to obtain multiple semantic unit sets.

[0049] In power grid scenarios, standardized event data sets contain a large amount of textual information, such as equipment fault descriptions and textual records in operation and maintenance logs. For example, a text describing an equipment fault in the field of renewable energy access on the power generation side reads, "Slight cracks developed on wind turbine blades in strong winds." This text is processed using a word segmentation algorithm, such as a dictionary-based word segmentation method, to segment the text into multiple semantic units. The resulting semantic unit set includes "wind turbine blade," "strong winds," and "slight cracks," among others.

[0050] For operation and maintenance log text in the field of high-voltage transmission corridor monitoring, such as "Trees were found growing near transmission lines in mountainous areas, affecting line safety", the semantic unit set obtained after word segmentation is "transmission lines in mountainous areas", "tree growth", "line safety", etc.

[0051] Such word segmentation processing is performed on all text information in the entire standardized event data set, thereby obtaining multiple semantic unit sets.

[0052] Step S122: calling a pre-trained semantic encoder to perform context encoding processing on the semantic unit set, and extracting the topic distribution features of the target event in different related fields.

[0053] Step S122-1: Input the semantic unit set into the bidirectional recurrent neural network layer of the semantic encoder to obtain the hidden state vector of each semantic unit.

[0054] In the power grid scenario, the semantic units obtained through word segmentation contain key semantic information from various related domains. For example, the semantic unit set "wind turbine blades," "strong winds," and "minor cracks" from the field of renewable energy access on the power generation side are sequentially input into the bidirectional recurrent neural network (BRNN) layer of the pre-trained semantic encoder. The BRNN layer simultaneously considers the contextual information surrounding the semantic unit. Through computation and transmission within the network's neurons, it processes each input semantic unit and ultimately outputs a hidden state vector for each semantic unit. This hidden state vector contains the feature representation of the semantic unit in its context.

[0055] Step S122-2: Generate a set of local semantic features of the target event in a single domain based on the hidden layer state vector.

[0056] For each associated domain, a local semantic feature set is generated based on the hidden state vectors of the semantic units within that domain. Taking the new energy access domain on the power generation side as an example, the hidden state vectors of all semantic units within that domain are further processed and combined. These hidden state vectors can be integrated into a set of semantic features that represents the target event within that domain, using methods such as vector concatenation and weighted summation. This local semantic feature set reflects the semantic feature information related to the target event in the new energy access domain on the power generation side.

[0057] Step S122 - 3 : Mapping local semantic features of different domains to a shared semantic space through a cross-domain feature space alignment function.

[0058] Local semantic features from different domains may reside in different feature spaces, making it difficult to directly compare and analyze features across domains. Using a cross-domain feature space alignment function, local semantic features from domains such as renewable energy access on the power generation side, high-voltage transmission corridor monitoring, and substation equipment operation and maintenance are mapped into a shared semantic space. This shared semantic space provides a unified representation framework for features from different domains, enabling subsequent clustering and analysis within the same space.

[0059] Step S122 - 4 : performing clustering processing on the mapped local semantic feature sets of all domains to generate a set of topic clusters containing a unified semantic distribution across domains.

[0060] Clustering algorithms, such as K-means, are used to cluster the sets of local semantic features from all domains mapped to the shared semantic space. Clustering algorithms group similar semantic features together based on their similarity, forming topic clusters. Each topic cluster represents a set of similar semantic features, reflecting a unified semantic distribution across domains. For example, a topic cluster related to "power equipment failure" might be formed, containing semantic features related to equipment failure from different domains, such as the power generation, transmission, and substation sides.

[0061] Step S122-5: Calculate the semantic coverage and distribution matching degree of each topic cluster set between different related fields based on the cross-domain topic mapping relationship library, and generate the cross-domain topic transfer weight.

[0062] The cross-domain topic mapping relationship library records the association relationship between topics in different fields. For each topic cluster set, the semantic coverage and distribution matching between different related fields are calculated based on the relationship library. Semantic coverage reflects the scope and degree of appearance of the topic cluster in various fields, and distribution matching measures whether the distribution pattern of the topic cluster in different fields is similar. By comprehensively considering the semantic coverage and distribution matching, the cross-domain topic migration weight is generated. For example, if a topic cluster has a high frequency of appearance in both the field of new energy access on the power generation side and the field of high-voltage transmission corridor monitoring, and the distribution patterns are similar, then the cross-domain topic migration weight of the topic cluster between these two fields is high.

[0063] Step S122 - 6 : Obtain the domain identifier set corresponding to each of the subject cluster sets, and determine the domain coverage of the subject cluster set according to the domain identifier set.

[0064] For each topic cluster set, obtain its corresponding domain identifier set. The domain identifier specifies the associated fields involved in the topic cluster. Based on the domain identifier set, the domain coverage of the topic cluster set can be determined. If the domain identifier set corresponding to a topic cluster set includes the field of new energy access on the power generation side, the field of high-voltage transmission corridor monitoring, and the field of substation equipment operation and maintenance, then the field coverage of the topic cluster set involves these three fields.

[0065] Step S122-7: When the domain coverage exceeds a preset threshold, extract the high-frequency keyword sequence of the subject cluster set, and match the high-frequency keyword sequence with a preset cross-domain knowledge base to obtain a semantic extension vector of the subject cluster set.

[0066] In order to judge the importance and representativeness of a topic cluster set, a preset field coverage threshold is set. When the field coverage of a topic cluster set exceeds the threshold, it means that the topic cluster is of great significance in multiple fields. At this time, the high-frequency keyword sequence in the topic cluster set is extracted. High-frequency keywords are words that appear frequently in the topic cluster, and they represent the core semantics of the topic cluster. The high-frequency keyword sequence is matched with a preset cross-domain knowledge base, which contains professional knowledge and terminology in various fields of the power grid. Through the matching operation, the knowledge and information related to the high-frequency keywords are found, thereby obtaining the semantic extension vector of the topic cluster set. The semantic extension vector can further enrich the semantic information of the topic cluster set.

[0067] Step S122-8: Update the hidden state vector of the topic cluster set according to the semantic extension vector, assign cross-domain topic migration weights to the updated hidden state vector, and linearly superimpose multiple updated hidden state vectors based on the assigned weight values ​​to generate the topic distribution feature.

[0068] The information of the semantic expansion vector is incorporated into the hidden state vector of the topic cluster set, and the hidden state vector is updated to contain richer and more accurate semantic information. Then, the updated hidden state vector is weighted according to the previously generated cross-domain topic migration weight. Each hidden state vector is assigned a weight value according to its corresponding cross-domain topic migration weight. Finally, the weighted multiple hidden state vectors are linearly superimposed, that is, each weighted hidden state vector is added according to the set rules to obtain the final topic distribution feature. This topic distribution feature reflects the semantic topic distribution of the target event in different related fields.

[0069] Step S123: generating a cross-domain common semantic vector according to the topic distribution feature, and mapping the cross-domain common semantic vector to the contextual semantic feature.

[0070] The topic distribution features are analyzed and processed to extract common and representative semantic information across different related fields, generating cross-domain common semantic vectors. For example, the topic distribution features in the areas of new energy access on the power generation side, high-voltage transmission corridor monitoring, and substation equipment operation and maintenance all contain semantic information related to "power safety." This relevant semantic information is integrated and extracted to form a cross-domain common semantic vector.

[0071] The cross-domain common semantic vector is mapped into contextual semantic features using a mapping function. The mapping function can be a linear or nonlinear transformation function, selected based on specific needs and data characteristics. Through the mapping operation, the cross-domain common semantic vector is converted into a feature vector that reflects the contextual semantics of the target event, namely the contextual semantic feature.

[0072] Step S124: Generate an adaptive time series window division rule according to the time granularity and event frequency of each associated field, dynamically adjust the length of the time series window according to the field, perform cross-field time series window alignment processing on the timestamp information in the standardized event data set, and obtain multiple time series segment sets.

[0073] The time granularity and event frequency may vary across different domains. For example, power generation data for renewable energy integration on the power generation side may be collected every minute, while line pulsation data for high-voltage transmission corridor monitoring may be collected every hour. Based on the time granularity and event frequency of each domain, adaptive time series window partitioning rules are generated.

[0074] For the new energy access field on the power generation side, due to the high frequency of data collection, a smaller time window length can be set, such as 10 minutes; for the high-voltage transmission corridor monitoring field, due to the low frequency of data collection, a larger time window length can be set, such as 2 hours.

[0075] Dynamically adjust the time series window length by field. If, within a certain time period, the power generation in the renewable energy access area fluctuates significantly and the event frequency increases, the time series window length can be appropriately shortened to more accurately capture the event changes. If the line operation in the high-voltage transmission corridor monitoring area is relatively stable and the event frequency decreases, the time series window length can be appropriately increased to reduce the data processing workload.

[0076] Perform cross-domain time series window alignment on the timestamp information in the standardized event data set. Divide the data from different fields according to their respective time series windows, and then align the time series windows of each field. For example, align the 10-minute time series window of the new energy access field on the power generation side with the 2-hour time series window of the high-voltage transmission corridor monitoring field, so that the data from different fields are comparable in time. Through the alignment process, multiple time series segment sets are obtained, each of which contains data from different fields in the same time period.

[0077] Step S125: extracting the event triggering frequencies of each associated domain in the time series segment set, and constructing a cross-domain time series association matrix according to the event triggering frequencies.

[0078] For each time series segment, count the event triggering frequencies for each associated domain. For example, within a two-hour time series segment, count the number of wind turbine failure events triggered in the renewable energy access domain on the power generation side and the number of line anomaly events triggered in the high-voltage transmission corridor monitoring domain, and so on, to obtain the event triggering frequencies for each domain.

[0079] Based on these event trigger frequencies, a cross-domain temporal correlation matrix is ​​constructed. The rows and columns of the matrix represent different correlation domains, and the elements in the matrix represent the degree of correlation between the event trigger frequencies of two domains. For example, the element in row i and column j of the matrix represents the degree of correlation between the event trigger frequencies of the power generation side's new energy access domain and the high-voltage transmission corridor monitoring domain. The value of this element can be determined by calculating the correlation between the event trigger frequencies of the two domains.

[0080] Step S126: generating the cross-domain time series behavior feature based on the cross-domain time series association matrix.

[0081] Step S126 - 1 : Perform singular value decomposition on the cross-domain time series correlation matrix to obtain multiple time series pattern components.

[0082] Singular value decomposition (SVD) is an important matrix decomposition method that can decompose the cross-domain time series correlation matrix into the product of three matrices, from which multiple time series pattern components are obtained. In the power grid scenario, the cross-domain time series correlation matrix records the degree of correlation between the event triggering frequencies of different related domains such as power generation, transmission, substation, distribution, and user side. Through SVD processing, the cross-domain time series correlation matrix can be decomposed into multiple relatively simple time series pattern components, each of which represents a specific time series change pattern. For example, a time series pattern component reflecting the correlation between periodic fluctuations in power generation and changes in transmission line load may be obtained, as well as a time series pattern component reflecting the relationship between substation equipment failures and abnormal power consumption on the user side.

[0083] Step S126-2: Input the time series pattern components into a preset time series feature extractor to extract the periodic fluctuation characteristics and trend change slope of each time series pattern component.

[0084] The preset time series feature extractor is a trained model specifically designed to extract useful feature information from time series pattern components. When the time series pattern components are input into the extractor, it analyzes each component. For periodic fluctuation features, the extractor will identify whether there is a periodic change pattern in the time series pattern components, such as whether the generated power fluctuates at certain time intervals, as well as information such as the amplitude and frequency of the fluctuations. The trend change slope reflects whether the change trend of the time series pattern component over a period of time is rising, falling, or remains stable, as well as the rate of change. For example, for the current change time series pattern component of the transmission line, the extractor can calculate whether the current is rising or falling, and the speed of rise or fall.

[0085] Step S126-3: Construct a cross-domain time series correlation map based on periodic fluctuation characteristics and trend change slope.

[0086] A cross-domain time series association map is constructed using the extracted periodic fluctuation characteristics and trend change slopes. In this cross-domain time series association map, the periodic fluctuation characteristics and trend change slopes of each time series pattern component are used as node attributes, and the association relationship between different time series pattern components is used as an edge. For example, if there is a strong correlation between a time series pattern component reflecting power generation fluctuations and another time series pattern component reflecting transmission line voltage changes, then there will be an edge connecting the two nodes in the cross-domain time series association map, and the attributes of the edge can represent the strength and direction of the association. In this way, the cross-domain time series association map intuitively displays the time series association relationship between different fields.

[0087] Step S126 - 4 : performing sub-graph division processing on the cross-domain temporal association graph to obtain multiple temporal behavior clusters.

[0088] Graph partitioning algorithms, such as spectral clustering, are used to partition the cross-domain temporal correlation graph into subgraphs. The spectral clustering algorithm analyzes the eigenvectors of the Laplacian matrix of the cross-domain temporal correlation graph and divides it into multiple relatively independent subgraphs, each representing a temporal behavior cluster. In power grid scenarios, these temporal behavior clusters may correspond to different grid operating states or event types. For example, one temporal behavior cluster may contain a series of events and associations related to power fluctuations of renewable energy generation, while another temporal behavior cluster may involve temporal information related to routine maintenance and fault handling of substation equipment. Through subgraph partitioning, complex cross-domain temporal correlation graphs can be decomposed into multiple substructures that are easy to analyze and process.

[0089] Step S126 - 5 : extract the core path features and branch correlation of each temporal behavior cluster, and generate cross-domain temporal behavior features based on the core path features and branch correlation.

[0090] For each temporal behavior cluster, we first extract its core path features. These represent the main threads and key processes of event development within the cluster. By analyzing the causal relationships and sequence between events within the cluster, we identify key event nodes on the core path and the relationships connecting these nodes. These nodes are then quantitatively described as a core path feature vector. This vector contains information such as the type of events on the core path, the time interval between occurrences, and the intensity of the events.

[0091] The branch correlation reflects the degree of association between events outside the core path and the core path. By calculating the correlation indicators between branch events and each event on the core path, such as temporal correlation and causal correlation, the branch correlation value of each branch event and the core path is obtained. The branch correlation values ​​of all branch events are combined into a vector, which serves as the branch correlation feature of the temporal behavior cluster.

[0092] Finally, the core path feature vector and branch correlation feature vector of each temporal behavior cluster are concatenated to form a characteristic representation of the temporal behavior cluster. The characteristic representations of all temporal behavior clusters are then integrated, for example using a weighted average method. Different weights are assigned to each temporal behavior cluster based on its importance or stability, and the characteristic representations of each temporal behavior cluster are weighted averaged to obtain the final cross-domain temporal behavior feature. This cross-domain temporal behavior feature can comprehensively reflect the temporal behavior relationships between different related domains in the power grid.

[0093] Step S130: Based on a preset collaborative analysis model, collaboratively analyze the contextual semantic features and the cross-domain temporal behavior features to generate a set of cross-domain collaborative tracking indicators.

[0094] The pre-designed and trained collaborative analysis model is used to collaboratively analyze contextual semantic features and cross-domain temporal behavior features. This model mainly consists of a multi-head attention layer and a fully connected layer.

[0095] Step S131: Input the contextual semantic features and the cross-domain temporal behavior features into the multi-head attention layer of the collaborative analysis model to generate multiple attention weight distribution maps.

[0096] The multi-head attention layer is a key component of the collaborative analysis model, enabling multi-dimensional attention and analysis of contextual semantic features and cross-domain temporal behavioral features. When input into the multi-head attention layer, these features are first linearly transformed, mapping them into different subspaces. Within each subspace, the importance of each feature element is determined by calculating the correlation between the query, key, and value. The query represents the information of interest, the key represents the information used to match the query, and the value represents the specific feature information. By calculating the similarity between the query and key, the attention weight for each feature element is determined. For example, when analyzing contextual semantic features, keywords closely related to the core semantics of the target event will receive a relatively high attention weight. When analyzing cross-domain temporal behavioral features, key event nodes along the core path will also receive a high attention weight.

[0097] The multi-head attention layer performs multiple such attention calculations in parallel, each of which is called a head. Each head generates an attention weight distribution map, which reflects the importance distribution of the feature elements that head focuses on. The attention weight distribution maps of multiple heads are combined to analyze and focus on input features from different angles and dimensions.

[0098] Step S132: Determine the feature contribution of different related fields according to the attention weight distribution map.

[0099] After obtaining multiple attention weight distribution maps, it is necessary to determine the feature contribution of different associated fields based on these attention weight distribution maps. For each associated field, in each attention weight distribution map, the attention weights of the characteristic elements in the field are summed to obtain the total weight of the field under the attention head. Then, the total weights under all attention heads are averaged to obtain the comprehensive attention weight of the associated field. The higher the comprehensive attention weight, the greater the contribution of the features of the associated field in the collaborative analysis. For example, in the power grid scenario, if the comprehensive attention weight of the new energy access field on the power generation side is high, it means that the features of this field play an important role in the analysis and tracking of the target event.

[0100] Step S133: dynamically weighting and splicing the contextual semantic features and the cross-domain temporal behavior features based on the feature contribution to generate a fusion feature vector.

[0101] In this embodiment, for each element in the contextual semantic features and cross-domain temporal behavior features, a weight is assigned to it according to the feature contribution of the associated field to which it belongs. For example, for the feature elements of the new energy access field on the power generation side, a corresponding weight is assigned to it according to the feature contribution of the field. Then, the weighted contextual semantic features and cross-domain temporal behavior features are spliced ​​to form a fused feature vector. The splicing method can be to connect the two feature vectors in sequence, while taking into account the consistency of the dimension and dimensionality of the features, to ensure that the spliced ​​fused feature vector can accurately reflect the comprehensive information of the contextual semantics and cross-domain temporal behavior.

[0102] Step S134: calling the fully connected layer of the collaborative analysis model to perform nonlinear transformation processing on the fused feature vector to obtain a preliminary tracking indicator set.

[0103] The fully connected layer is another key component of the collaborative analysis model, performing nonlinear transformations on the fused feature vector. When the fused feature vector is input to the fully connected layer, it multiplies the input feature vector by a set of learnable weight matrices and adds a bias term. The result of this multiplication and bias addition is then processed through a nonlinear activation function to produce a transformed feature vector. This transformed feature vector serves as the initial set of tracking indicators. The purpose of the nonlinear activation function is to introduce nonlinear factors, enabling the model to learn more complex feature relationships. For example, a commonly used nonlinear activation function is the ReLU function, which sets values ​​less than zero to zero and leaves values ​​greater than zero unchanged, thereby enhancing the model's expressive power.

[0104] Step S135: extracting the domain dependency parameter of each indicator in the preliminary tracking indicator set, and determining the distribution deviation of the indicator among multiple related fields according to the domain dependency parameter.

[0105] For each indicator in the initial tracking indicator set, its domain dependency parameter needs to be extracted. The domain dependency parameter reflects the degree of dependence between the indicator and its associated domains. By analyzing the indicator's generation process and feature sources, the associated domains on which the indicator primarily depends can be determined. For example, an indicator may primarily rely on the areas of renewable energy access on the power generation side and high-voltage transmission corridor monitoring. Its domain dependency parameter can be determined by calculating the correlation between the indicator and the characteristics of these two domains.

[0106] Based on the domain dependency parameters, the distribution deviation of the indicator across multiple related domains is determined. Distribution deviation can be calculated by calculating the difference between the actual distribution of the indicator in each related domain and the theoretical distribution. The theoretical distribution can be determined based on the characteristic contribution of each related domain and the indicator generation mechanism. For example, if the theoretical distribution ratio of an indicator in the field of renewable energy access on the power generation side is a certain value, but the actual distribution ratio differs significantly from this value, then this indicates that the indicator has a distribution deviation in the field of renewable energy access on the power generation side.

[0107] Step S136: When the distribution deviation exceeds a preset tolerance threshold, the indicator is determined to be an abnormal indicator.

[0108] To determine whether an indicator is abnormal, a preset tolerance threshold is required. This threshold is determined based on actual conditions and experience, and represents the acceptable range of distribution deviation for an indicator across various related fields. When the distribution deviation of an indicator exceeds the preset tolerance threshold, it indicates that the distribution of the indicator differs significantly from normal conditions, and the indicator is considered abnormal. For example, if the distribution deviation of an indicator in a related field exceeds the upper limit of the preset tolerance threshold, it may indicate an abnormal event in that field or a problem with data collection.

[0109] Step S137: Perform cross-domain correlation verification on the non-abnormal indicators, calculate their trigger frequency consistency and semantic matching in the cross-domain temporal behavior characteristics, and when the trigger frequency consistency and the semantic matching meet the preset conditions, add the indicators to the cross-domain collaborative tracking indicator set.

[0110] For non-abnormal indicators, cross-domain correlation verification is required. Cross-domain correlation verification is mainly carried out from two aspects, namely trigger frequency consistency and semantic matching.

[0111] Trigger frequency consistency refers to whether an indicator's trigger frequency is consistent across different related domains. In cross-domain temporal behavior characteristics, each indicator has a corresponding trigger frequency. Trigger frequency consistency can be calculated by comparing the trigger frequencies of indicators across different related domains. For example, if an indicator's trigger frequency shows similar trends in both the power generation-side renewable energy access domain and the high-voltage transmission corridor monitoring domain, then this indicates high trigger frequency consistency across these two domains.

[0112] Semantic matching refers to the degree of match between an indicator's semantics and the contextual semantic features. This matching is calculated by comparing the indicator's semantic information with the contextual semantic features. For example, if an indicator's semantics closely matches the keyword "power safety" in the contextual semantic features, then the indicator has a high semantic matching degree.

[0113] When both trigger frequency consistency and semantic matching meet the preset conditions, the indicator is reasonable and effective in terms of cross-domain relevance and semantic expression, and the indicator is added to the cross-domain collaborative tracking indicator set. The preset conditions can be set based on actual needs and data characteristics. For example, thresholds for trigger frequency consistency and semantic matching can be set. Only when the trigger frequency consistency and semantic matching of the indicator exceed the corresponding thresholds will the indicator be added to the cross-domain collaborative tracking indicator set.

[0114] Step S140: constructing a dynamic tracking path topology according to the cross-domain collaborative tracking indicator set, and generating a positioning assistance strategy adapted to the dynamic tracking path topology.

[0115] Step S141: Mapping the cross-domain collaborative tracking indicator set to multiple tracking nodes, and generating connection edges between the nodes according to the association relationships in the cross-domain collaborative tracking indicator set.

[0116] The cross-domain collaborative tracking indicator set includes multiple indicators related to the target event, each of which can be mapped to a tracking node. For example, in a power grid scenario, one indicator in the cross-domain collaborative tracking indicator set is the "generation-side renewable energy power fluctuation indicator," which can be mapped to a tracking node. By performing this mapping on all indicators, multiple tracking nodes are obtained.

[0117] Edges are generated between nodes based on the relationships within the cross-domain collaborative tracking indicator set. These relationships can be causal, temporal, or similar. For example, if a change in the "power generation side renewable energy power fluctuation indicator" causes a change in the "high-voltage transmission corridor line load change indicator," a connecting edge is generated between the two corresponding tracking nodes. The direction of the connecting edge can be determined based on the directionality of the relationship. For example, if it is a causal relationship, the direction is from the node corresponding to the cause indicator to the node corresponding to the result indicator.

[0118] Step S142: extracting the weight coefficient and direction attribute of each connecting edge, and constructing an initial tracking path topology graph.

[0119] For each generated edge, its weight coefficient and direction attribute need to be extracted. The weight coefficient reflects the closeness of the association between two tracking nodes. The weight coefficient can be determined by analyzing factors such as the correlation and influence between indicators. For example, if the correlation between two indicators is high, the weight coefficient of the edge connecting the corresponding tracking nodes will be larger. The direction attribute specifies the direction of the connection. As mentioned above, it is determined based on the directionality of the association relationship.

[0120] Based on the weight coefficients and direction attributes of the tracking nodes and connecting edges, an initial tracking path topology graph is constructed. The initial tracking path topology graph is a directed graph, in which nodes represent tracking indicators, edges represent the associations between indicators, and the weight coefficients and direction attributes of the edges reflect the strength and direction of the association.

[0121] Step S143: performing redundant edge elimination processing on the initial tracking path topology graph, and merging adjacent nodes whose weight coefficients are lower than a threshold.

[0122] The initial tracing path topology graph may contain some redundant connection edges. These edges do not contribute substantially to the construction and analysis of the tracing path, but instead increase the complexity of the graph. Therefore, the initial tracing path topology graph needs to be processed to eliminate redundant edges. The redundant nature of an edge can be determined by analyzing the edge weight coefficient and the relationship between nodes. For example, if an edge has a very small weight coefficient and the two nodes it connects can also establish a strong relationship through other paths, then this edge can be considered redundant and removed from the graph.

[0123] At the same time, adjacent nodes with weight coefficients below a threshold can be merged. Adjacent nodes are nodes directly connected by edges in the graph. When the weight coefficient of the edge connecting adjacent nodes is below the threshold, it indicates that the connection between the two nodes is weak, and they can be merged into a single node. When merging nodes, the attributes and relationships of the two nodes must be integrated to ensure that the merged node accurately reflects the information of the two original nodes.

[0124] Step S144: identifying all possible core path candidate sets in the merged topology graph, and calculating the connectivity strength and execution efficiency score of each candidate path in the core path candidate set.

[0125] In the merged topology graph, all possible core path candidate sets need to be identified. Core paths are paths that play a key role in tracking the target event, reflecting the event's main development context and key processes. Graph search algorithms, such as depth-first search or breadth-first search, can be used to traverse the topology graph, identify all possible paths, and then select the core path candidate set based on their importance and relevance.

[0126] For each candidate path in the core path candidate set, its connectivity strength and execution efficiency score are calculated. Connectivity strength reflects the closeness of the connections between nodes on the path and can be calculated by calculating the sum of the weight coefficients of the connecting edges on the path. The execution efficiency score takes into account factors such as the execution difficulty and time cost of the path. For example, if a path requires multiple complex steps or takes a long time to execute, its execution efficiency score will be relatively low. An evaluation model can be established to comprehensively consider various factors of the path to calculate the execution efficiency score.

[0127] Step S145: Sort the candidate paths according to the connectivity strength and the execution efficiency score, and select the top N paths as core paths.

[0128] Based on the calculated connectivity strength and execution efficiency scores, the candidate paths in the core path candidate set are sorted. The sorting can be performed in descending order based on the combined connectivity strength and execution efficiency scores. The combined score is obtained by taking a weighted sum of the connectivity strength and execution efficiency scores, with the weights set based on actual needs and data characteristics.

[0129] Select the top N paths after sorting as core paths. The value of N can be determined based on actual conditions and requirements, representing the number of core paths to focus on when tracking the target event. Core paths can provide key clues and directions for tracking the target event.

[0130] Step S146: extracting path segments that overlap with the core path from the remaining candidate paths, and generating backup path priorities according to the number of intersections of the path segments.

[0131] After selecting the core path, the remaining candidate paths need to be extracted for any path segments that overlap with the core path. An overlap refers to the overlap between two paths at certain nodes or edges. By comparing the remaining candidate paths with the core path, these overlapping path segments can be identified.

[0132] The backup path priority is generated based on the number of intersections in a path segment. A higher number of intersections indicates a closer relationship between the path segment and the core path. This increases the feasibility and effectiveness of the path segment as a backup path in the event of a core path failure, and therefore its backup path priority. The backup paths can be prioritized by sorting the path segments in descending order based on the number of intersections.

[0133] Step S147: Select a backup path based on the backup path priority, and integrate the core path and the backup path into the dynamic tracking path topology.

[0134] Select the appropriate backup path based on its priority. A set number of backup paths can be selected based on actual needs and circumstances to ensure timely switching to backup paths when an anomaly occurs on the core path, ensuring the continuity and effectiveness of tracking the target event.

[0135] Integrate core and backup paths into a dynamic tracking path topology. This is a comprehensive path topology that includes core and backup paths, providing multiple viable path options for tracking target events. During this integration process, ensure the accuracy and completeness of the connectivity and attribute information between core and backup paths so that these paths can be accurately used for tracking in real-world applications.

[0136] Step S148: Generate a positioning assistance strategy that is adapted to the dynamic tracking path topology.

[0137] Step S148 - 1 : According to a preset cross-domain resource type mapping table, the resource requirement type of each node in the dynamic tracking path topology is converted into a standardized resource identifier, and the corresponding execution constraint condition is parsed.

[0138] The preset cross-domain resource type mapping table is a pre-established mapping relationship table that defines the correspondence between resource requirement types in different domains and standardized resource identifiers. For each node in the dynamic tracking path topology, the resource requirement type is converted to a standardized resource identifier by querying the cross-domain resource type mapping table. For example, in a power grid scenario, a tracking node's resource requirement type is "power generation equipment maintenance resources." This can be converted to the corresponding standardized resource identifier using the mapping table.

[0139] At the same time, the execution constraints corresponding to each standardized resource identifier are parsed. Execution constraints include resource availability, usage time limits, and operating specifications. For example, for a "power generation equipment maintenance resource," its execution constraints might include the maintenance personnel's work schedule and the inventory status of the maintenance equipment.

[0140] Step S148 - 2 : Match the available resource pool in the cross-domain tracking system according to the standardized resource identifier, and generate a resource allocation plan.

[0141] The converted standardized resource identifier is then matched against the available resource pool in the cross-domain tracking system. The available resource pool is a database that stores information about various available resources, including their type, quantity, and status. The available resource pool is queried to find available resources that match the standardized resource identifier.

[0142] Generate a resource allocation plan based on the matching results. This plan needs to consider factors such as the resource requirements of each node and the execution order. For example, for a tracking node that requires multiple resources, different types of resources must be properly allocated to ensure normal node execution. Furthermore, resource allocation timing must be properly scheduled based on the execution order of the nodes to avoid resource conflicts.

[0143] Step S148-3: Perform feasibility verification on the resource allocation plan based on the execution constraint conditions, adjust the resource allocation ratio and time window, and obtain an adjusted resource allocation plan.

[0144] After obtaining the resource allocation plan, it is necessary to perform a feasibility check on it based on the previously analyzed execution constraints. The purpose of the feasibility check is to ensure that the resource allocation plan can meet various constraints during actual execution and ensure the smooth progress of the tracking operation.

[0145] For example, consider the allocation of maintenance resources for power generation equipment in a power grid scenario. Assume that execution constraints include maintenance personnel's work schedules, equipment inventory availability, and safety regulations for maintenance operations. For a resource allocation plan that schedules maintenance for a specific power generation equipment unit during a specific time period, it's necessary to check whether a sufficient number of maintenance personnel with the appropriate skills are available during that time period. This can be done by querying the maintenance personnel schedule to determine the number of personnel available during that time period. If the number of personnel is insufficient, the plan is unfeasible in terms of personnel resources.

[0146] Also, check whether the inventory of equipment needed for maintenance is sufficient. For example, repairing a certain power generation equipment may require specific types of tools and parts. Check whether the inventory system has sufficient quantities of these items. If inventory is insufficient, consider adjusting resource allocation, such as relocating equipment from other locations or postponing the maintenance plan.

[0147] To ensure safety regulations for maintenance operations, it's necessary to check whether the maintenance process and operational arrangements in the resource allocation plan comply with safety standards. For example, whether sufficient safety equipment is in place and whether the proper sequence for equipment power outage maintenance is followed. If not, the resource allocation plan needs to be modified to ensure operational safety.

[0148] If a resource allocation plan proves unworkable, the resource allocation ratio and time window need to be adjusted. Adjusting the resource allocation ratio can be achieved by reallocating the quantities of different resource types. For example, if the required maintenance personnel for a particular power generation unit are insufficient, personnel can be redeployed from other maintenance tasks while reducing the manpower allocated to these other tasks. Adjusting the time window involves altering the timing of resource allocation. For example, if both maintenance personnel and equipment are insufficient within a certain time period, the maintenance plan can be postponed to a time period with sufficient resources.

[0149] By continuously performing feasibility checks and adjustments based on execution constraints, an adjusted resource allocation plan is eventually obtained. This plan can meet the execution constraints in terms of resource allocation and time scheduling and has high feasibility.

[0150] Step S148 - 4 : Allocate a unique resource identifier to each node in the dynamic tracking path topology, and establish a mapping relationship between the resource identifier and the available resource pool.

[0151] To accurately manage and allocate resources, each node in the dynamic tracing path topology needs to be assigned a unique resource identifier. This resource identifier uniquely identifies the resource required by the node and is used to identify and manage resources throughout the cross-domain tracing system.

[0152] Taking the power grid scenario as an example, a representative power generation equipment monitoring node in the dynamic tracking path topology is assigned a unique resource identifier, such as "GD-001." This identifier is unique across the entire system and can accurately correspond to the node.

[0153] After assigning a resource identifier to each node, a mapping relationship needs to be established between the resource identifier and the available resource pool. The available resource pool stores detailed information about various available resources, including resource type, quantity, and status. This mapping relationship makes it easy to find the corresponding available resource based on the node's resource identifier.

[0154] For example, for a power generation equipment monitoring node with the resource identifier "GD-001," the required resources may include monitoring equipment and monitoring personnel. In the available resource pool, the resource information related to the monitoring equipment and monitoring personnel is searched and mapped to "GD-001." This mapping relationship can be implemented by creating a mapping table in the database. Each row in the mapping table records a resource identifier and the corresponding available resource information.

[0155] In this way, when performing resource allocation and scheduling, the required available resources can be quickly located according to the node's resource identifier, thereby improving the efficiency and accuracy of resource management.

[0156] Step S148-5: Divide the resource scheduling phases according to the time window in the adjusted resource allocation plan, and set a priority tag for each phase.

[0157] The resource scheduling process is divided into different phases based on the time windows in the adjusted resource allocation plan. The time windows specify the start and end times of each resource allocation task. Based on this time information, the entire resource scheduling process can be divided into multiple consecutive time periods, each of which is a resource scheduling phase.

[0158] For example, in a power grid scenario, suppose the adjusted resource allocation plan includes three different power generation equipment maintenance tasks, each scheduled for different time windows. Based on these time windows, the resource scheduling process can be divided into three phases, with each phase corresponding to the resource scheduling of a maintenance task.

[0159] Set a priority label for each resource scheduling stage. The priority label reflects the importance and urgency of the stage in the entire resource scheduling process. The priority setting can be determined based on a variety of factors, such as the urgency of the task and the degree of impact on grid operation. For example, for a power generation equipment maintenance task that affects the power supply to important users, the corresponding resource scheduling stage priority is relatively high and can be set to "high priority"; while for some maintenance tasks that have less impact on grid operation, the resource scheduling stage priority can be set to "medium priority" or "low priority".

[0160] By dividing resource scheduling stages and setting priority tags, resources can be better managed and scheduled, ensuring that high-priority tasks receive priority resource support, improving resource utilization efficiency and tracking operation effectiveness.

[0161] Step S148-6: Align the resource scheduling phase with the node execution order in the dynamic tracking path topology to generate a binding relationship table.

[0162] To ensure that resources are accurately allocated to each node in the dynamic tracking path topology, it is necessary to align the resource scheduling phase with the node execution order. The node execution order is determined by the association relationship between nodes in the dynamic tracking path topology and the logical order of tracking tasks.

[0163] Taking the power grid scenario as an example, a dynamic tracking path topology may contain multiple nodes, such as power generation equipment monitoring nodes, transmission line inspection nodes, and substation equipment maintenance nodes. These nodes have a set execution order. The resource scheduling phase also has a corresponding time sequence. Aligning the resource scheduling phase with the node execution order ensures that when a node needs resources, it can be allocated resources from the corresponding resource scheduling phase in a timely manner.

[0164] For example, if a power generation equipment monitoring node needs to start monitoring tasks at a certain time, then the node's execution time must be matched with the time allocated to the monitoring resources during the resource scheduling phase. If the time allocated to the monitoring resources during the resource scheduling phase does not match the node's execution time, adjustments must be made to align the two.

[0165] After alignment, a binding table is generated. This table records the binding relationship between each node and its corresponding resource scheduling phase. A row in the table records a node ID, the corresponding resource scheduling phase ID, and related resource allocation information. This binding table clearly shows which resources each node requires and when.

[0166] Step S148-7: Generate resource scheduling instructions and time synchronization instructions including resource identifiers, scheduling phases, and priority tags based on the binding relationship table to obtain the positioning assistance strategy.

[0167] According to the binding relationship table, resource scheduling instructions and time synchronization instructions are generated. Resource scheduling instructions are used to guide resource allocation and scheduling operations in the cross-domain tracking system. They contain information such as resource identifiers, scheduling stages, and priority tags.

[0168] For example, in a power grid scenario, for a power generation equipment monitoring node recorded in the binding relationship table, the corresponding resource scheduling instruction will clearly indicate the node's resource identifier (such as "GD-001"), the resource scheduling phase it belongs to (such as "Phase 1"), and the priority tag of that phase (such as "High Priority"). The resource scheduling instruction also details the type and quantity of resources to be allocated to the node, as well as the time and location of resource allocation.

[0169] Time synchronization instructions ensure that each node's execution time is consistent with the resource scheduling phase. They specify the specific times each node should start and end its tasks, as well as the time synchronization requirements with other nodes. For example, a power generation equipment monitoring node needs to begin monitoring at a specific time. A time synchronization instruction precisely tells the node that time and requires it to synchronize with other relevant nodes, ensuring the coordination and accuracy of the entire tracking operation.

[0170] Combining resource scheduling instructions with time synchronization instructions yields a positioning assistance strategy. This strategy provides detailed resource allocation and scheduling guidance for cross-domain tracking systems, enabling them to more efficiently coordinate positioning operations and improve the accuracy and efficiency of tracking target events.

[0171] Step S150: Synchronize the positioning assistance strategy to the cross-domain tracking system to activate the collaborative positioning operation.

[0172] For example, step S150 includes: Step S151: Decompose the positioning assistance strategy into multiple sub-strategy units, each sub-strategy unit corresponding to a tracking subsystem in an associated field.

[0173] The location-assistance strategy is a comprehensive one, encompassing resource scheduling and collaborative operations across multiple domains. To facilitate its implementation within a cross-domain tracking system, it needs to be broken down into multiple sub-strategy units. Each sub-strategy unit corresponds to a tracking subsystem in a specific domain, allowing each tracking subsystem to focus on its specific tasks.

[0174] In the power grid scenario, related areas include renewable energy integration on the power generation side, high-voltage transmission corridor monitoring, substation equipment operation and maintenance, distributed power penetration in low-voltage distribution networks, and the power consumption characteristics of large industrial users. The positioning assistance strategy is broken down into sub-strategy units corresponding to each of these areas. For example, for renewable energy integration on the power generation side, the sub-strategy unit would include resource scheduling instructions and time synchronization instructions for each node within that area, such as resource allocation and time scheduling for power generation equipment monitoring nodes and renewable energy power generation power control nodes.

[0175] Step S152: constructing a policy execution sequence according to the execution dependency relationship of the sub-policy units, and setting a cross-domain communication protocol to synchronize the execution status.

[0176] Different sub-strategy units may have execution dependencies, meaning that the execution of one sub-strategy unit depends on the completion of other sub-strategy units. For example, in a power grid scenario, a sub-strategy unit responsible for high-voltage transmission corridor monitoring may not begin line inspections until a sub-strategy unit responsible for renewable energy access on the power generation side has completed power generation adjustment.

[0177] Based on the execution dependencies of the sub-policy units, a policy execution sequence is constructed. This specifies the execution order of each sub-policy unit, ensuring that each sub-policy unit is executed at the appropriate time. This can be constructed using a topological sorting algorithm from graph theory. The sub-policy units are considered nodes in a graph, and the execution dependencies are considered directed edges between nodes. Topological sorting is used to determine the execution order of the nodes.

[0178] Furthermore, to ensure real-time synchronization of execution status between tracking subsystems, a cross-domain communication protocol is required. This protocol defines the communication rules and data formats between the various tracking subsystems. For example, it specifies how execution status information is sent and received between subsystems, and how errors and exceptions are handled. Through this protocol, each tracking subsystem can keep up to date with the execution progress of other subsystems, ensuring the coordination of the entire collaborative positioning operation.

[0179] Step S153: deploying a policy execution agent module in the cross-domain tracking system to monitor the execution progress and resource consumption of the sub-policy units in real time.

[0180] In order to effectively manage and monitor the execution process of co-location operations, a policy execution agent module is deployed in the cross-domain tracking system. The policy execution agent module is a dedicated software module responsible for real-time monitoring of the execution progress and resource consumption of each sub-policy unit.

[0181] In power grid scenarios, the policy execution agent communicates with each tracking subsystem to obtain execution status information for each sub-policy unit. For example, for a sub-policy unit involved in renewable energy access on the power generation side, the policy execution agent monitors in real time whether power generation equipment monitoring nodes start and complete monitoring tasks on time, and whether resources are allocated and used as planned.

[0182] Regarding resource consumption, the policy execution agent module monitors the quantity and type of resources used by each sub-policy unit, such as manpower, equipment, and electricity. By monitoring resource consumption in real time, it can promptly identify abnormal resource usage, such as waste or insufficient resources, so that appropriate measures can be taken to adjust them.

[0183] Step S154: When an abnormality in the execution of a sub-policy unit is detected, a dynamic adjustment mechanism is triggered to reallocate resources or switch to a backup path.

[0184] Step S154 - 1 : Generate adjustment candidate solutions based on the standardized resource identifier of the current abnormal sub-policy unit and the remaining available resource pool capacity.

[0185] When the policy execution agent module detects an execution anomaly of a sub-policy unit, it needs to trigger the dynamic adjustment mechanism. First, it generates adjustment candidate solutions based on the standardized resource identifier of the current abnormal sub-policy unit and the remaining available resource pool capacity.

[0186] In a power grid scenario, assume that a maintenance sub-strategy unit for power generation equipment in the renewable energy access area on the power generation side is performing abnormally, such as a shortage of maintenance personnel, which leads to delayed maintenance progress. Based on the standardized resource identifier of this sub-strategy unit (such as "GD-002"), the remaining available resource pool is queried to determine the current number of maintenance personnel and equipment available. Based on this information, candidate adjustment plans are generated. For example, plan one might involve redeploying maintenance personnel from other regions; plan two might involve extending the maintenance period while utilizing existing maintenance personnel to expedite maintenance of other equipment to balance the overall maintenance tasks.

[0187] Step S154 - 2 : Evaluate the impact weight of each candidate solution on the overall positioning progress and calculate the adjusted expected execution efficiency.

[0188] For each generated adjustment candidate, its impact on the overall co-location progress needs to be assessed. This weight reflects the impact of the solution on the overall co-location operation progress. This weight can be determined by analyzing factors such as the solution's implementation time, resource requirements, and its relevance to other sub-strategy units.

[0189] For example, regarding the candidate adjustment options for the power generation equipment maintenance sub-strategy unit mentioned above, Option 1 might require a longer deployment of maintenance personnel from other regions, which would cause a certain delay in the overall positioning progress, but would ensure that the maintenance task is completed on time. Option 2, while extending the maintenance time, might affect the execution of subsequent related sub-strategies, but would allow the maintenance task to continue without increasing excessive resources. Based on these analyses, each option is assigned an impact weight.

[0190] At the same time, the expected execution efficiency after adjustment is calculated. This takes into account factors such as the execution time and resource utilization efficiency of the sub-strategy unit after the plan is implemented. This can be calculated by establishing an evaluation model that comprehensively considers various factors. For example, for Plan 1, the expected completion time and resource utilization efficiency of the adjusted sub-strategy unit are calculated, taking into account the time required to deploy personnel and the adaptation period for new personnel. For Plan 2, the expected execution efficiency is calculated based on the extended time and the arrangement of existing manpower.

[0191] Step S154 - 3 : Select the candidate solution with the highest expected execution efficiency as the optimal adjustment strategy, update the resource allocation plan and dynamically track the path topology.

[0192] Based on the impact weights and expected execution efficiency, the candidate with the highest expected execution efficiency is selected as the optimal adjustment strategy. In the above example of the power generation equipment maintenance sub-strategy unit, if the expected execution efficiency of option 2 is higher than that of option 1, then option 2 is selected as the optimal adjustment strategy.

[0193] After selecting the optimal adjustment strategy, the resource allocation plan and dynamic tracking path topology need to be updated. For the resource allocation plan, resources are reallocated based on the optimal adjustment strategy, such as adjusting the allocation of maintenance personnel and the scheduling of maintenance times. For the dynamic tracking path topology, the relationships between nodes and path information are updated based on the adjusted resource allocation and execution order. For example, if Option 2 is selected to extend the maintenance time, the execution time and resource requirements of subsequent related sub-strategy units may need to be adjusted, and the information of these nodes in the dynamic tracking path topology will be updated accordingly.

[0194] Step S154 - 4 : Synchronize the updated resource allocation plan and topology to the relevant tracking subsystem, and reset the priority tag of the policy execution sequence.

[0195] Synchronize the updated resource allocation plan and dynamic tracking path topology to the relevant tracking subsystems. In power grid scenarios, this updated information is sent to relevant tracking subsystems, such as those involved in renewable energy access on the power generation side and monitoring high-voltage transmission corridors, ensuring that each subsystem is aware of the adjusted resource allocation and path information.

[0196] At the same time, the priority tags of the policy execution sequence are reset. Due to changes in resource allocation and execution order, the priorities of each sub-policy unit need to be reassessed. For example, after adjusting the power generation equipment maintenance sub-policy unit, the urgency and importance of subsequent related sub-policy units may be affected, and the priority tags need to be reset based on the new situation.

[0197] Step S154-5: Record the key parameter change log during the adjustment process and use it for subsequent strategy optimization analysis.

[0198] During the dynamic adjustment process, changes to key parameters are recorded. These include resource allocation quantity, execution time, priority tags, and more. These logs provide detailed information about the adjustment process and are crucial for subsequent strategy optimization analysis.

[0199] In power grid scenarios, the system records information such as the number of maintenance personnel deployed, the duration of maintenance time extensions, and changes in priority tags during adjustments to power generation equipment maintenance sub-strategy units. By analyzing these logs, lessons learned from the adjustments can be identified, including the causes of sub-strategy unit execution anomalies and the effectiveness and shortcomings of the adjustment strategies. Based on these analysis results, the positioning assistance strategy can be optimized to improve the stability and efficiency of collaborative positioning operations.

[0200] Step S155: Summarize the execution results of each sub-strategy unit to the central node, generate a global collaborative positioning report and feed it back to the user terminal.

[0201] After each sub-strategy unit is executed, its execution results are summarized to the central node. The central node is the core node in the cross-domain tracking system, responsible for collecting and processing information from each tracking subsystem.

[0202] In the power grid scenario, the central node will collect the execution results of sub-strategy units in various fields, such as new energy access on the power generation side, high-voltage transmission corridor monitoring, and substation equipment operation and maintenance, including task completion status, resource usage, exception handling, etc.

[0203] Based on the aggregated execution results, a global co-location report is generated. This report provides a comprehensive summary and analysis of the entire co-location operation, including information such as the tracking status of target events, the execution results of each related area, and resource utilization efficiency. The report can be presented in the form of charts, tables, and text descriptions, making the information more intuitive and easy to understand.

[0204] Finally, the global collaborative positioning report is fed back to the user terminal. This can be a computer, mobile phone, or other device used by grid managers. By receiving the report, users can promptly understand the results of the collaborative positioning operation, evaluate its effectiveness, and make appropriate decisions based on the information in the report, such as further optimizing positioning assistance strategies and adjusting grid operation plans.

[0205] Figure 2 The following diagram illustrates exemplary hardware and software components of an information tracking and positioning assistance system 100 for cross-domain collaboration, which can implement the concepts of the present application, according to some embodiments of the present application. For example, the processor 120 can be used in the information tracking and positioning assistance system 100 for cross-domain collaboration and to perform the functions of the present application.

[0206] The cross-domain collaborative information tracking and positioning assistance system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the cross-domain collaborative information tracking and positioning assistance method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0207] For example, the cross-domain collaborative information tracking and positioning assistance system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the cross-domain collaborative information tracking and positioning assistance system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The cross-domain collaborative information tracking and positioning assistance system 100 also includes an I / O interface 150 between the computer and other input and output devices.

[0208] For ease of explanation, only one processor is described in the cross-domain collaborative information tracking and positioning assistance system 100. However, it should be noted that the cross-domain collaborative information tracking and positioning assistance system 100 in this application may also include multiple processors, so the steps performed by one processor described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the cross-domain collaborative information tracking and positioning assistance system 100 executes steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0209] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above-mentioned cross-domain collaborative information tracking and positioning assistance method is implemented.

[0210] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A cross-domain collaborative information tracking and positioning auxiliary method, characterized in that: The method comprises: Collecting the original data sets generated by the target event in multiple related fields, and performing cross-field identifier matching processing on the original data sets to obtain a standardized event data set; Performing information association processing on the standardized event data set to generate contextual semantic features and cross-domain temporal behavior features of the target event; Based on a preset collaborative analysis model, the contextual semantic features and the cross-domain temporal behavior features are collaboratively analyzed to generate a set of cross-domain collaborative tracking indicators; Constructing a dynamic tracking path topology according to the cross-domain collaborative tracking indicator set, and generating a positioning assistance strategy adapted to the dynamic tracking path topology; The positioning assistance strategy is synchronized to the cross-domain tracking system to activate the co-location operation.

2. The cross-domain collaborative information tracking and positioning assistance method according to claim 1, characterized in that: The performing cross-domain identifier matching processing on the original data set to obtain a standardized event data set includes: Traversing each data record in the original data set, and extracting the domain identifier and data format type of the data record; Matching the global unified identification code corresponding to the domain identifier according to a preset cross-domain mapping rule library, and converting the data format type into a target standardized format to obtain a converted data record; Performing redundancy check on the converted data records, deleting duplicate records and invalid fields, and then hierarchically aggregating the verified data records according to timestamps and field identifiers to form the standardized event data set with a unified structure; The standardized event data set is verified for integrity so that the data coverage of each associated field meets a preset threshold.

3. The cross-domain collaborative information tracking and positioning assistance method according to claim 2, characterized in that: The redundancy check processing is performed on the converted data records to delete duplicate records and invalid fields, including: Extracting a key field set of each converted data record, wherein the key field set includes an event identification code, a timestamp, a domain identifier, and a content summary; Building a hash index based on the event identification code and detecting duplicate records of the same event in different fields according to the field identifier; Compare the timestamps of the detected duplicate records, retain the record with the latest timestamp and delete the rest; Calculating semantic similarity of the content summary, and marking it as redundant content if the similarity exceeds a preset threshold; Based on the marking results, redundant content is merged or deleted to generate a streamlined set of data records.

4. The cross-domain collaborative information tracking and positioning assistance method according to claim 1, characterized in that: The performing information association processing on the standardized event data set to generate contextual semantic features and cross-domain temporal behavior features of the target event includes: Performing word segmentation processing on the text information in the standardized event data set to obtain multiple semantic unit sets; Calling a pre-trained semantic encoder to perform context encoding processing on the semantic unit set to extract the topic distribution characteristics of the target event in different related fields; generating a cross-domain common semantic vector according to the topic distribution feature, and mapping the cross-domain common semantic vector to the contextual semantic feature; Generate an adaptive time series window division rule based on the time granularity and event frequency of each associated field, dynamically adjust the length of the time series window by field, and perform cross-field time series window alignment processing on the timestamp information in the standardized event data set to obtain multiple time series segment sets; Extracting event triggering frequencies of various related fields in the set of time series segments, and constructing a cross-field time series correlation matrix based on the event triggering frequencies; The cross-domain time series behavior feature is generated based on the cross-domain time series association matrix.

5. The cross-domain collaborative information tracking and positioning assistance method according to claim 4, characterized in that: The calling of the pre-trained semantic encoder to perform context encoding processing on the semantic unit set to extract the topic distribution features of the target event in different related fields includes: Inputting the semantic unit set into the bidirectional recurrent neural network layer of the semantic encoder to obtain a hidden state vector of each semantic unit; Generating a local semantic feature set of the target event in a single domain according to the hidden layer state vector; The local semantic features of different domains are mapped to a shared semantic space through a cross-domain feature space alignment function. The local semantic feature sets of all domains after mapping are clustered to generate a set of topic clusters with a unified semantic distribution across domains. According to the cross-domain topic mapping relationship library, the semantic coverage and distribution matching degree of each topic cluster set between different related fields are calculated to generate the cross-domain topic transfer weight; Obtaining a domain identifier set corresponding to each of the subject cluster sets, and determining the domain coverage of the subject cluster set according to the domain identifier set; When the domain coverage exceeds a preset threshold, extract the high-frequency keyword sequence of the subject cluster set, and match the high-frequency keyword sequence with a preset cross-domain knowledge base to obtain the semantic extension vector of the subject cluster set; The hidden state vector of the topic cluster set is updated according to the semantic extension vector, and the cross-domain topic migration weight is assigned to the updated hidden state vector, and multiple updated hidden state vectors are linearly superimposed based on the assigned weight values ​​to generate the topic distribution feature.

6. The cross-domain collaborative information tracking and positioning assistance method according to claim 4, characterized in that: Generating the cross-domain time series behavior feature based on the cross-domain time series association matrix includes: Performing singular value decomposition on the cross-domain time series correlation matrix to obtain multiple time series pattern components; Inputting the time series pattern components into a preset time series feature extractor to extract the periodic fluctuation characteristics and trend change slope of each time series pattern component; Constructing a cross-domain time series correlation map based on the periodic fluctuation characteristics and the trend change slope; Performing sub-graph division processing on the cross-domain temporal association graph to obtain multiple temporal behavior clusters; The core path features and branch correlations of each of the temporal behavior clusters are extracted, and the cross-domain temporal behavior features are generated according to the core path features and the branch correlations.

7. The cross-domain collaborative information tracking and positioning assistance method according to claim 1, characterized in that: The collaborative analysis model based on the preset collaborative analysis model performs collaborative analysis on the contextual semantic features and the cross-domain temporal behavior features to generate a set of cross-domain collaborative tracking indicators, including: Inputting the contextual semantic features and the cross-domain temporal behavior features into the multi-head attention layer of the collaborative analysis model to generate multiple attention weight distribution maps; Determining the feature contributions of different associated fields according to the attention weight distribution map; Based on the feature contribution, the contextual semantic features and the cross-domain temporal behavior features are dynamically weighted and spliced ​​to generate a fused feature vector; Calling the fully connected layer of the collaborative analysis model to perform nonlinear transformation processing on the fused feature vector to obtain a preliminary tracking indicator set; Extracting a domain dependency parameter of each indicator in the preliminary tracking indicator set, and determining a distribution deviation of the indicator among a plurality of related domains based on the domain dependency parameter; When the distribution deviation exceeds a preset tolerance threshold, determining the indicator as an abnormal indicator; The non-abnormal indicators are verified for cross-domain correlation, and their trigger frequency consistency and semantic matching in the cross-domain temporal behavior characteristics are calculated. When both the trigger frequency consistency and the semantic matching meet the preset conditions, the indicators are added to the cross-domain collaborative tracking indicator set.

8. The cross-domain collaborative information tracking and positioning assistance method according to claim 1, characterized in that: The constructing of a dynamic tracking path topology according to the cross-domain collaborative tracking indicator set includes: Mapping the cross-domain collaborative tracking indicator set to a plurality of tracking nodes, and generating connection edges between the nodes according to the association relationships in the cross-domain collaborative tracking indicator set; Extracting the weight coefficient and direction attribute of each connecting edge to construct an initial tracking path topology graph; Eliminating redundant edges on the initial tracking path topology graph and merging adjacent nodes whose weight coefficients are lower than a threshold; Identifying all possible core path candidate sets in the merged topology graph, and calculating the connectivity strength and execution efficiency score of each candidate path in the core path candidate set; sorting the candidate paths according to the connectivity strength and the execution efficiency score, and selecting the top N paths as core paths; Extracting path segments that overlap with the core path from the remaining candidate paths, and generating backup path priorities based on the number of intersections of the path segments; A backup path is selected based on the backup path priority, and the core path and the backup path are integrated into the dynamic tracking path topology.

9. The cross-domain collaborative information tracking and positioning assistance method according to claim 1, characterized in that: The generating of a positioning assistance strategy adapted to the dynamic tracking path topology includes: According to a preset cross-domain resource type mapping table, the resource requirement type of each node in the dynamic tracking path topology is converted into a standardized resource identifier, and the corresponding execution constraint conditions are parsed; Matching available resource pools in a cross-domain tracking system according to the standardized resource identifiers and generating a resource allocation plan; Performing feasibility verification on the resource allocation plan based on the execution constraint conditions, adjusting the resource allocation ratio and time window, and obtaining an adjusted resource allocation plan; Assigning a unique resource identifier to each node in the dynamic tracking path topology and establishing a mapping relationship between the resource identifier and an available resource pool; Divide resource scheduling phases according to the time windows in the adjusted resource allocation plan, and set a priority label for each phase; Aligning the resource scheduling phase with the node execution order in the dynamic tracking path topology to generate a binding relationship table; A resource scheduling instruction and a time synchronization instruction including a resource identifier, a scheduling phase and a priority tag are generated based on the binding relationship table to obtain the positioning assistance strategy.

10. A cross-domain collaborative information tracking and positioning assistance system, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the cross-domain collaborative information tracking and positioning assistance method described in any one of claims 1 to 9.

Images