Electric power big data acquisition and processing method based on AI

Abstract

The invention relates to the technical field of data acquisition, in particular to an AI-based electric power big data acquisition and processing method. According to the method, numbering extraction is carried out on field contents and upper and lower positions of the power maintenance log and the transaction contract data, a propagation path between fields is constructed on a semantic level, and field groups with semantic offset and structural dislocation are subjected to bidirectional alignment after difference identification; according to the method, field semantic expressions and sequence positions are synchronously corrected in the association process, the accuracy of field connection between different types of data is improved, continuous judgment is conducted in combination with time parameters, connection between maintenance and contract fields better fits the actual business process, field contents and connection relations are further archived after field combinations are aligned, and the efficiency is improved. And a structured basis is provided for data reconciliation and scheduling collection, and the semantic consistency, the structure matching degree and the application availability of an electric power big data acquisition processing result are remarkably improved through the processing flow.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data collection, in particular to an AI-based power big data collection and processing method. BACKGROUND

[0002] The technical field of data collection relates to the collection, transmission and preliminary processing of various operation and maintenance data, transaction data and related management information in a power system.

[0003] Among them, the power big data collection and processing method refers to collecting the original data generated by specific devices such as electric energy meters, load monitoring equipment and transaction record terminals distributed in substations, distribution rooms and user sides through a centralized collection system, usually using periodic polling or event triggering methods with fixed time intervals for data grabbing, transmitting through serial communication or Ethernet, and then using rule-based script programs to preliminarily filter and structure the data source in the master station or data center, and using manually set field matching methods to arrange and archive the power data, electricity price information and contract parameters.

[0004] In the existing technology, the preliminary filtering and structured arrangement in the process of power big data collection and processing mainly rely on rule scripts and field matching methods, and the processing means lack dynamic adaptation ability to semantic differences and structural misplacement between fields, and when facing data with large differences in sources such as maintenance logs and transaction contracts, the fields often have inconsistent content expression or context position mismatch, which often leads to repeated, missed or mismatched problems in the archiving results, especially when there are non-standardized descriptions between contract clauses and maintenance records, the field semantics cannot be effectively corresponded, which affects the completeness of data reconciliation and the accuracy of dispatching and archiving, and if the archiving data is relied on before the execution of the dispatching task, it is easy to cause equipment instruction deviation or contract execution risk. SUMMARY

[0005] The application aims to solve the problems in the prior art and provides an AI-based power big data collection and processing method.

[0006] In order to achieve the above-mentioned purpose, the application adopts the following technical scheme: an AI-based power big data collection and processing method, comprising the following steps: S1: collecting power maintenance logs and power maintenance transaction contract data of power maintenance, extracting power maintenance log fields and transaction contract fields, and performing sequential numbering on the upper and lower field positions of the maintenance log fields and the transaction contract fields in their respective records to generate a field basic information set; S2: mapping semantic propagation paths for the maintenance log fields and the transaction contract fields in the field basic information set through a graph neural network to generate a semantic propagation path graph; S3: judging the difference between the maintenance log field and the transaction contract field in semantic expression respectively based on the semantic propagation path diagram, and screening a semantic offset field group; S4: performing bidirectional semantic propagation path correction on the semantic offset field group based on the up and down field positions of the maintenance log field and the transaction contract field, and generating a semantic migrated path diagram; S5: obtaining all maintenance log fields and transaction contract fields combined in the semantic migrated path diagram that have completed semantic mapping and are aligned with the up and down field positions, and obtaining a power big data collection and processing result.

[0007] As a further scheme of the present application, the field basic information set includes field content items, field position numbers, and field source identifiers, the semantic propagation path diagram includes up and down field sequence paths, field semantic correspondence relationships, and time correlation information, the semantic offset field group includes structure offset field pairs, semantic mismatch field pairs, and field position difference values, the semantic migrated path diagram includes path update results, field connection directions, and sequence position adjustment data, and the power big data collection and processing result is specifically field combination content and semantic mapping state.

[0008] As a further scheme of the present application, the field basic information set acquisition step is specifically: S111: collecting power maintenance log and power maintenance transaction contract data of power maintenance, extracting maintenance log fields and transaction contract fields therefrom, and generating a field extraction result; S112: based on the field extraction result, performing sequence numbering on the up and down field positions of the maintenance log fields and the transaction contract fields in their respective records respectively, calling position index values of the arrangement sequence of the fields in the original data, and binding the position index values with the corresponding field names, and generating field sequence annotation data; S113: integrating all numbered maintenance log fields and transaction contract fields according to the field sequence annotation data, and generating a field basic information set.

[0009] As a further scheme of the present application, the maintenance log field includes maintenance behavior description, maintenance equipment identification, and maintenance time.

[0010] As a further scheme of the present application, the transaction contract field includes contract task description, contract equipment clause, and contract performance time.

[0011] As a further scheme of the present application, the semantic propagation path diagram acquisition step is specifically: S211: Take the field basis information set as the connection basis in the graph neural network, construct the sequential connection path between the maintenance log field and the transaction contract field, set the edge connection relationship according to the front-back arrangement order of the fields in the original data, and generate a field sequential connection structure; S212: Based on the field sequential connection structure, extract the maintenance time and the contract performance time as time edge association information, extract the maintenance equipment identifier and the contract equipment clause execution semantics corresponding judgment, attach the semantic relationship of the matched field to the corresponding connection path, and generate semantic structure association information; S213: Construct a graph neural network graph structure according to the semantic structure association information, take the maintenance log field and the transaction contract field as nodes in the graph, take the connection path as an edge, and take the semantic and time association information as an edge attribute, to generate a semantic propagation path graph.

[0012] As a further scheme of the application, the obtaining step of the semantic offset field group is specifically: S311: Call the up-down field position index data of the maintenance log field and the transaction contract field in the semantic propagation path graph, extract the offset relationship based on the arrangement order of the positions of the up-down fields, identify the field combination with position corresponding offset in the record, and generate an up-down structure difference field pair; S312: Extract the semantic content of the maintenance equipment identifier and the contract equipment clause in the up-down structure difference field pair, perform key term matching and word meaning consistency judgment, identify the field combination with no common reference item in semantic expression, and record the semantic mismatch identifier on the corresponding path, to generate a semantic mismatch field pair; S313: Combine the up-down structure difference field pair and the semantic mismatch field pair, filter the field group with both semantic expression difference and structure offset, as the field combination with inconsistent semantic expression and sequential dislocation, and generate a semantic offset field group.

[0013] As a further scheme of the application, the obtaining step of the path graph after semantic migration is specifically: S411: Call the up-down field position of each maintenance log field and transaction contract field in the semantic offset field group, identify the field combination with up-down field position offset, and extract the time parameters corresponding to the field group as the maintenance time and the contract performance time, to generate a structure corresponding field relationship group; S412: According to the structure corresponding field relationship group, judge the connection state of the maintenance equipment identifier and the contract equipment clause in the semantic propagation path graph in each field group, detect whether there is a path connection relationship, and judge whether the time is continuous based on the arrangement order of the maintenance time and the contract performance time, filter the field group with established connection and time continuity in the semantic propagation path graph, and generate a path connection effective field set; S413: adjust the connection direction and the up and down field position number of the maintenance log field and the transaction contract field in the semantic propagation path graph respectively, update the connection order and the path direction of the original field in the graph, and generate the path graph after semantic migration.

[0014] As a further scheme of the present application, the acquisition step of the power big data collection processing result is specifically: S511: call the connection data of the maintenance log field and the transaction contract field in the path graph after semantic migration, filter all field combinations that have completed semantic mapping and up and down field position alignment, and generate an aligned field combination list; S512: according to the aligned field combination list, correspondingly arrange the content, up and down field position number and semantic connection direction of each group of fields, map and archive the combination relationship of the maintenance behavior description and the contract task description, the maintenance equipment identifier and the contract equipment clause in the structure diagram, and generate field comparison structure data; S513: based on the field comparison structure data, take the field content and the structure relationship as the input field of the power maintenance and transaction data reconciliation processing and dispatching archiving, and generate the power big data collection processing result.

[0015] Compared with the prior art, the present application has the following advantages and positive effects: In the present application, the field content and the up and down position of the power maintenance log and the transaction contract data are numbered and extracted, the propagation path between the fields is constructed at the semantic level, and after difference identification, the field groups with semantic deviation and structure misplacement are bidirectionally aligned, so that the field semantic expression and the sequential position are synchronized and corrected in the association process, the accuracy of the field connection between different types of data is improved, the continuity is judged in combination with the time parameter, the connection of the maintenance and contract fields is more in line with the actual business process, and after the alignment of the field combinations, the field content and the connection relationship are further archived, providing a structured basis for data reconciliation and dispatching collection. Through this processing process, the semantic consistency, structure matching degree and application availability of the power big data collection processing result are significantly improved, and the problems of field misalignment, reconciliation difficulty and the like caused by heterogeneous data sources of maintenance and transaction data are solved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The present application is a main step schematic diagram; Figure 2 The present application is a flowchart of step S1; Figure 3 The present application is a flowchart of step S2; Figure 4 The present application is a flowchart of step S3; Figure 5Flow chart for step S4 of the present application; Figure 6 Flow chart for step S5 of the present application. DETAILED DESCRIPTION

[0017] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0018] Referring to Figure 1 The present application provides a technical solution: an AI-based electric power big data collection and processing method, comprising the following steps: S1: collecting electric power maintenance logs and electric power maintenance transaction contract data of electric power maintenance, extracting electric power maintenance log fields and transaction contract fields therefrom, performing sequential numbering on the upper and lower field positions of the maintenance log fields and the transaction contract fields in their respective records, and generating a field basic information set; S2: mapping semantic propagation paths for the maintenance log fields and the transaction contract fields in the field basic information set through a graph neural network, and generating a semantic propagation path graph; S3: based on the semantic propagation path graph, respectively judging the differences in semantic expression of the maintenance log fields and the transaction contract fields, and screening a semantic offset field group; S4: based on the upper and lower field positions of the maintenance log fields and the transaction contract fields, performing bidirectional semantic propagation path correction on the semantic offset field group, and generating a path graph after semantic migration; S5: obtaining all maintenance log fields and transaction contract field combinations that have completed semantic mapping and upper and lower field position alignment in the path graph after semantic migration, and obtaining an electric power big data collection and processing result; The field basic information set includes field content items, field position numbers, and field source identifiers. The semantic propagation path graph includes upper and lower field sequential paths, field semantic correspondence relationships, and time association information. The semantic offset field group includes structure offset field pairs, semantic mismatch field pairs, and field position difference values. The path graph after semantic migration includes path update results, field connection directions, and sequential position adjustment data. The electric power big data collection and processing result is specifically field combination content and semantic mapping status.

[0019] Referring to Figure 2 The acquisition step of the field basic information set is specifically: S111: collecting electric power maintenance logs and electric power maintenance transaction contract data of electric power maintenance, extracting maintenance log fields and transaction contract fields therefrom, and generating a field extraction result; For the power maintenance log collected from the power maintenance server and the power maintenance transaction contract data collected from the contract management database, for the unstructured text of the power maintenance log, a preset keyword list containing "maintenance time", "maintenance personnel", "maintenance location", "maintenance equipment", "fault description" and "treatment process" is used to scan and match the log content line by line. Once a keyword in the list is matched, all characters from the keyword to the next keyword or to the end of the line text are extracted as the field content corresponding to the keyword. For the structured document of the power maintenance transaction contract, "contract number" and "signing date" are directly extracted from the "contract basic information" part, and "contract equipment clause" and "performance time" are extracted from the "contract clause" part according to the predefined document template. All field names and their contents extracted from the power maintenance log and the power maintenance transaction contract data are collected to generate field extraction results.

[0020] S112: Based on the field extraction results, the upper and lower field positions of the maintenance log fields and the transaction contract fields in their respective records are sequentially numbered, the arrangement order of the fields in the original data is called to establish the position index value, and the position index value is bound with the corresponding field name to generate field order annotation data; For the field extraction results, the maintenance log fields derived from a single power maintenance log record are assigned a consecutive integer starting from 1 as the position index value according to the order of their appearance from top to bottom in the original log text, and each position index value is bound with the corresponding field name to form a paired data containing the index and the name. Similarly, the transaction contract fields derived from a single power maintenance transaction contract are assigned a consecutive integer starting from 1 as the position index value according to their arrangement order in the contract document structure, and each position index value is bound with the corresponding field name. The above numbering and binding operations are completed for all maintenance log fields and transaction contract fields to generate field order annotation data.

[0021] S113: According to the field order annotation data, all numbered maintenance log fields and transaction contract fields are integrated to generate field basic information set; For the field order annotation data, a integration process is started, all the maintenance log fields and transaction contract fields annotated with position index values are traversed, each field is taken as an independent data entry, together with the source type of the field (i.e. maintenance log or transaction contract), the position index value of the field and the name of the field, and is completely moved into a newly created collection, the integration process does not modify any existing attributes of the field, only performs the collection action, until all the annotated fields in the record are included in the newly created collection, forming a field basic information collection containing all numbered maintenance log fields and transaction contract fields.

[0022] Referring to Figure 3 , the semantic propagation path diagram acquisition step is specifically: S211: Take the field basic information collection as the connection basis in the graph neural network, construct the order connection path between the maintenance log fields and the transaction contract fields, set the edge connection relationship according to the front and back arrangement order of the fields in the original data, and generate the field order connection structure; According to the field basic information collection, each field information in the collection is taken as an independent node, between the nodes derived from the same maintenance log record, a directed connection edge is established from the node with a smaller position index value to the node with a larger position index value adjacent thereto, and the same operation is performed between the nodes derived from the same transaction contract, that is, a directed connection path is established between the nodes according to the order of the position index values from small to large, and in the process of establishing the connection, no connection across different maintenance log records or different transaction contracts is created, and no connection between the maintenance log field nodes and the transaction contract field nodes is established, thereby generating the field order connection structure.

[0023] S212: Based on the field order connection structure, extract the maintenance time and the contract performance time as time edge association information, extract the maintenance equipment identifier and the contract equipment clause execution semantic corresponding judgment, attach the semantic relationship of the matched field to the corresponding connection path, and generate semantic structure association information; In the field order connection structure, first find all the nodes with the name "maintenance time" and all the nodes with the name "contract performance time", extract the time information contained in these two nodes, and mark these time information as time edge association information, then extract the content of all "maintenance equipment identifier" nodes and the content of all "contract equipment clause" nodes, and perform semantic corresponding judgment on the content strings of these two types of nodes, which is completed by calculating the Jaccard similarity coefficient, and the calculation formula is: The specific meanings of each letter in the formula are as follows: : Represents the Jaccard similarity coefficient between the content of the "Maintenance Equipment Identifier" field and the "Contract Equipment Terms" field. It is a final result that quantifies the degree of similarity between two text strings. : This represents a set of characters formed after deduplication of the "Maintenance Equipment Identifier" field content from the power maintenance log. Each element in the set is a unique character from the field content. : This represents a set of characters formed after deduplication of the "Contract Equipment Terms" field content from the power maintenance transaction contract. Each element in the set is a unique character from the field content. : Represents a set With sets The number of elements in the intersection, sign This represents the intersection operation, which filters out elements that exist in both sets. and set All common characters, symbols This indicates the total number of characters contained in the intersection. : Represents a set With sets The number of elements in a union, sign This represents the union operation, which merges sets. and set All characters and remove duplicates, symbols This represents the total number of characters contained in the union of sets. For example, if the content of "Maintenance Equipment Identifier" is "#3 Main Transformer", then its character set... Given {'#','3','main','transformer','voltage','device'}, if the content of the "Contract Equipment Terms" is "Maintenance of #3 Main Transformer", then its character set is... For {'#','3','main','transformer','voltage','equipment','maintenance','protection'}, first calculate the intersection. ={'#','3','main','transformer','voltage','device'}, its number of elements The value is 6, then the union is calculated. ={'#','3','Main','Transformer','Voltage','Electrical','Maintenance','Protection'}, its number of elements The value is 8. Substitute the value into the formula: Set a similarity threshold for semantic correspondence judgment. The setting process of the threshold value is as follows: a verification data set containing a plurality of device description field pairs is prepared, and "corresponding" or "not corresponding" is manually marked for each field pair by a power expert, then the Jaccard similarity coefficient of each field pair in the data set is calculated, and then the true positive rate and false positive rate under each threshold value from 0.1 to 0.9 are calculated, and finally the threshold value that makes the difference between the true positive rate and the false positive rate maximum is selected as the final threshold value. Since the calculated similarity coefficient 0.75 is greater than the threshold value 0.7, it is determined that the semantics of the two fields match, and the matching conclusion is attached to the semantic relationship of the connection path that may be established in the future to generate semantic structure association information.

[0024] S213: constructing a graph neural network graph according to the semantic structure association information, maintaining the log fields and the transaction contract fields as nodes in the graph, the connection paths as edges, and the semantic and time association information as edge attributes, and generating a semantic propagation path graph; Using the semantic structure association information, first, all the maintenance log fields and the transaction contract fields in the field basic information set are taken as nodes in the graph structure, and the sequential connection paths in the field sequential connection structure are taken as the basic edges between the nodes. Then, for each group of "maintenance device identifier" nodes and "contract device clause" nodes marked as semantic matching in the semantic structure association information, a semantic association edge is added between the two nodes, and the "semantic relationship: matching" attribute is assigned to the added edge. Meanwhile, the "maintenance time" nodes and the "contract performance time" nodes in the same original record as the two matching nodes are found, and an edge is added between the two time nodes, and the extracted time information is taken as the "time association" attribute of the time edge, thereby generating a semantic propagation path graph.

[0025] Referring to Figure 4 , the obtaining step of the semantic offset field group is as follows: S311: calling the upper and lower field position index data of the maintenance log fields and the transaction contract fields in the semantic propagation path graph, extracting the offset relationship based on the arrangement order of the positions of the upper and lower fields, identifying the field combination with position corresponding offset in the record, and generating the upper and lower structure difference field pair; Based on the semantic propagation path graph and the field sequential annotation data, first, all pairs of maintenance log field nodes and transaction contract field nodes directly connected by edges with the "semantic relationship: matching" attribute in the semantic propagation path graph are screened, for each pair of nodes screened, the respective upper and lower field position index values are queried from the field sequential annotation data, and then the two position index values are compared, if the numerical values of the two position index values are not equal, the field pair is identified as a combination with position corresponding offset, and all the identified field combinations with position corresponding offset are collected to generate the upper and lower structure difference field pair.

[0026] S312: Extract the semantic content of the maintenance equipment identification and contract equipment clause in the upper and lower structure difference field pair, perform key term matching and word meaning consistency judgment, identify the field combination with no common reference item in semantic expression, and record the semantic mismatch identification on the corresponding path to generate the semantic mismatch field pair; For each pair of "maintenance equipment identification" field and "contract equipment clause" field in the upper and lower structure difference field pair, extract the original semantic content of each field, quantize the key terms in the field content into numerical vectors through a word vector model pre-trained on a large corpus of power industry documents, and take the element-level average of all key term vectors as the semantic vector of the entire field for field content containing multiple key terms. Then, calculate the cosine distance between the semantic vectors of the two fields to determine whether the semantics are mismatched, and the calculation formula is: and , where the specific meanings of each letter are as follows: : represents the cosine distance between vectors and , which is the final calculation result for measuring the degree of semantic dissimilarity between the two fields. : represents the semantic vector obtained after the "maintenance equipment identification" field is converted by the word vector model, which is a set of numerical values capturing the specific meaning of the field in the power industry. : represents the semantic vector obtained after the "contract equipment clause" field is converted by the same word vector model, which is also a set of numerical values used to mathematically express the semantics of the field. : represents the specific numerical component of semantic vector in the th dimension, and each dimension corresponds to an abstract semantic feature learned by the model. : represents the specific numerical component of semantic vector in the th dimension, corresponding to the same semantic feature dimension as . : represents the index number of the vector dimension, starting from 1 and up to the total number of vector dimensions , used to traverse each component in the vector. : represents the total number of dimensions of the vector space defined by the word vector model, which is a fixed integer, such as 300, determining the number of numerical components included in each semantic vector. : represents the summation operation, indicating that the expression immediately following it is summed from to ​The summation of all calculation results. Taking an example, if the average vector of the field "Replace #3 transformer bushing" is... for The average vector of the field "Inspecting C-line switch" for First, calculate the dot product: ,calculate Norm: ,calculate Norm: Calculate the cosine similarity: Finally, calculate the cosine distance: Set a distance threshold for semantic mismatch. The threshold is set to 0.6. The basis for setting the threshold is to perform distance calculation on a validation set containing a large number of power term pairs to form distance distributions of two types of term pairs: "relevant" and "unrelevant". The distance value at the intersection of the two distribution curves is selected as the threshold because the point can minimize the misjudgment of the two types of terms. Since the calculated distance of 0.6087 is greater than the threshold of 0.6, it is determined that the semantic expressions of these two fields have no common referents. A semantic mismatch identifier is added to the path connecting these two fields in the semantic propagation path graph. All the fields that have gone through the above judgment process and have been recorded with semantic mismatch identifiers are combined to generate semantic mismatch field pairs.

[0027] S313: Combining the structural difference field pairs with the semantic mismatch field pairs, filter the field groups that have both semantic expression differences and structural offsets, and generate semantic offset field groups as a combination of fields with inconsistent semantic expression and misaligned order. Based on the structural difference field pairs and semantic mismatch field pairs, a filtering process is initiated. Each field combination in the structural difference field pairs is traversed, and it is checked whether the currently traversed field combination also exists in the set of semantic mismatch field pairs. If a field combination satisfies both the conditions of existing in the structural difference field pairs and existing in the semantic mismatch field pairs, that is, the field combination has both differences in position index values ​​and semantic mismatch, then the field combination is filtered out. All the filtered field combinations are used to generate a semantic offset field group.

[0028] Please see Figure 5 The specific steps for obtaining the semantically transferred path graph are as follows: S411: Call the upper and lower field positions of each maintenance log field and transaction contract field in the semantic offset field group, identify the field combinations with upper and lower field position offsets, and extract the time parameters corresponding to the field group as maintenance time and contract performance time to generate the structure corresponding field relationship group; The semantic offset field group, wherein each group maintains the internal structure analysis of the content of the log field and the transaction contract field, identifies and extracts the lower-level fields with the same name existing in the content, and judges whether the relative positions of these lower-level fields in the content of the respective upper-level fields exist offset. If offset is identified, perform position alignment operation according to the arrangement order in the upper-level field content to keep the relative positions consistent; then, if there is a combination of lower-level fields with the same name and the relative position alignment has been completed in a group of maintenance log fields and transaction contract fields, extract the time parameters of these lower-level fields and define them as maintenance time and contract performance time respectively. All field combinations that complete structure alignment and successfully extract time parameters through the above-mentioned way generate structure corresponding field relationship groups.

[0029] S412: According to the structure corresponding field relationship group, judge the connection state of maintenance equipment identifier and contract equipment clause in the semantic propagation path graph in each group of fields, detect whether there is a path connection relationship, and judge whether the time is continuous based on the arrangement order of maintenance time and contract performance time, filter the field groups that have established connection and have time continuity in the semantic propagation path graph, and generate a path connection effective field set; For the structure corresponding field relationship group, where each pair of maintenance equipment identifier field and contract equipment clause field in the group is judged for connection state in the semantic propagation path graph. The judgment process is to take one field node as the starting point in the graph and perform path search to check whether there is a path composed of edges that can reach another field node. If such a path exists, it is determined that the path connection relationship exists. Then, the maintenance time and contract performance time corresponding to the field group are judged for time continuity. The judgment process is to convert the two time values into comparable numerical formats and compare their sizes. If the numerical value of the maintenance time is not less than the numerical value of the contract performance time, it is determined that the time is continuous. Only those field groups that have established connection and have time continuity in the semantic propagation path graph are filtered to generate a path connection effective field set.

[0030] S413: Adjust the connection direction and up-down field position number of the maintenance log field and the transaction contract field in the semantic propagation path graph in the path connection effective field set respectively, update the connection order and path direction of the original fields in the graph, and generate the semantic migrated path graph; For each field group in the path connection effective field set, an adjustment operation is performed in the data structure of the semantic propagation path graph. First, the edge connecting the maintenance log field node and the transaction contract field node in the field group is located, the direction attribute of the edge is accessed, and if the direction attribute is from the maintenance log field to the transaction contract field or is undirected, the attribute is modified to be from the transaction contract field to the maintenance log field. The adjustment establishes the logical leading relationship of the power maintenance transaction contract to the power maintenance log. Then, the position index value attribute of the transaction contract field node in the field group is obtained, and then the maintenance log field node is located, and the position index value attribute thereof is updated to the position index value of the transaction contract field just obtained. The operation makes the two nodes completely aligned in the structure order, avoiding misalignment caused by different data sources. The above edge direction modification and node position index value updating operations are repeatedly performed on all field groups in the path connection effective field set, so as to update the connection order and path direction of the original fields in the graph, and generate the path graph after semantic migration.

[0031] Please refer to Figure 6 The acquisition step of the power big data collection and processing result is specifically: S511: The connection data of the maintenance log field and the transaction contract field in the path graph after semantic migration is called, all field combinations that have completed semantic mapping and have completed alignment of upper and lower field positions are filtered, and an aligned field combination list is generated. In the path graph after semantic migration, a filtering program is started. The program traverses each edge in the graph and performs a series of checks on each edge. First, the two nodes connected by the edge are checked. The type attribute of one node must be “maintenance log field”, and the type attribute of the other node must be “transaction contract field”. If not, the current edge is skipped. Second, the attribute of the edge is checked. It must contain a “semantic relationship” attribute with a value of “match” to confirm that the two fields have been verified and confirmed as related at the semantic level. If not, it is also skipped. Finally, the “position index value” attributes of the two nodes connected by the edge are obtained respectively, and whether the two attribute values are completely equal is compared to confirm whether the two fields have been aligned in the structure order. If not, it is also skipped. Only when an edge passes the above three checks, the maintenance log field and the transaction contract field connected by the edge are extracted as an effective combination. All extracted effective combinations are added to a new list, and finally an aligned field combination list is generated.

[0032] S512: According to the aligned field combination list, the content of each field group, the position number of upper and lower fields, and the semantic connection direction are correspondingly arranged, the combination relationship of the maintenance behavior description and the contract task description, the maintenance equipment identifier and the contract equipment clause in the structure graph is mapped and archived, and field comparison structure data is generated. For each set of fields in the aligned field combination list, a structured sorting and archiving process is performed, which creates a new data record for each set of fields, in the new record, the content of the power maintenance log field node is extracted and stored in the "maintenance behavior description" field, the content of the power maintenance transaction contract field node is extracted and stored in the "contract task description" field, then the aligned "position index value" is obtained from any node and stored in the "structural position number" field, and the direction attribute of the edge connecting the two nodes is obtained and stored in the "semantic connection direction" field. In this way, the combination relationship between maintenance behavior description and contract task description, maintenance equipment identification and contract equipment clause in the structure diagram is mapped one by one, and these structured data records containing complete comparison information are archived to generate field comparison structure data.

[0033] S513: Based on the field comparison structure data, the field content and the structure relationship are taken as input fields for power maintenance and transaction data reconciliation processing and scheduling archiving to generate power big data collection processing results. The field comparison structure data is taken as input for power maintenance and transaction data reconciliation processing and scheduling archiving. In the reconciliation processing process, the automated program reads the records in the field comparison structure data one by one, directly compares the contents of the "maintenance behavior description" and "contract task description" fields, and if the contents are consistent under the preset rules, such as identical characters or matching key technical parameters, the record is marked as "reconciliation successful", otherwise it is marked as "to be manually reviewed". In the scheduling archiving process, the program uses the "structural position number" in the record to uniformly sort the fields from the power maintenance log and the power maintenance transaction contract, and establishes logical links between the archived data according to the "semantic connection direction". The originally scattered maintenance log data and contract data are integrated into a structured, logical and traceable archive collection. The output results of the two processes together constitute the power big data collection processing results.

[0034] The above is only a preferred embodiment of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification of the above embodiments based on the technical essence of the present application without departing from the technical solution content of the present application shall fall within the protection scope of the present application.

Claims

1. An AI-based power big data collection and processing method, characterized in that, The method comprises the following steps: S1: collecting power maintenance logs and power maintenance transaction contract data of power maintenance, extracting power maintenance log fields and transaction contract fields therefrom, performing sequential numbering on the upper and lower field positions of the maintenance log fields and the transaction contract fields in their respective records, and generating a field basic information set; S2: mapping semantic propagation paths for the maintenance log fields and the transaction contract fields in the field basic information set through a graph neural network, and generating a semantic propagation path graph; S3: judging the differences in semantic expression of the maintenance log fields and the transaction contract fields based on the semantic propagation path graph, and screening a semantic offset field group; S4: performing bidirectional semantic propagation path correction on the semantic offset field group based on the upper and lower field positions of the maintenance log fields and the transaction contract fields, and generating a path graph after semantic migration; S5: obtaining all maintenance log fields and transaction contract field combinations that have completed semantic mapping and upper and lower field position alignment in the path graph after semantic migration, and obtaining a power big data collection and processing result. 2.The AI-based electric power big data collection and processing method of claim 1, wherein The field basic information set comprises field content items, field position numbers, and field source identifiers, the semantic propagation path graph comprises upper and lower field sequential paths, field semantic corresponding relationships, and time association information, the semantic offset field group comprises structural offset field pairs, semantic mismatch field pairs, and field position difference values, the path graph after semantic migration comprises path update results, field connection directions, and sequential position adjustment data, and the power big data collection and processing result specifically comprises field combination content and semantic mapping states. 3.The AI-based electric power big data collection and processing method of claim 1, wherein The acquisition step of the field basic information set specifically comprises: S111: collecting power maintenance logs and power maintenance transaction contract data, extracting maintenance log fields and transaction contract fields therefrom, and generating a field extraction result; S112: performing sequential numbering on the upper and lower field positions of the maintenance log fields and the transaction contract fields in their respective records based on the field extraction result, calling position index values of the fields in the original data according to the arrangement order of the fields, binding the position index values and corresponding field names, and generating field sequential annotation data; S113: integrating all numbered maintenance log fields and transaction contract fields according to the field sequential annotation data, and generating a field basic information set. 4.The AI-based electric power big data collection and processing method of claim 3, wherein The maintenance log fields comprise maintenance behavior descriptions, maintenance equipment identifiers, and maintenance times. 5.The AI-based electric power big data collection and processing method of claim 3, wherein, The transaction contract fields comprise contract task descriptions, contract equipment clauses, and contract performance times. 6.The AI-based electric power big data collection and processing method of claim 3, wherein The acquisition step of the semantic propagation path graph specifically comprises: S211: taking the field basic information set as a connection basis in a graph neural network, constructing sequential connection paths between the maintenance log fields and the transaction contract fields, setting edge connection relationships according to the front and back arrangement order of the fields in the original data, and generating a field sequential connection structure; S212: extracting maintenance times and contract performance times as time edge association information based on the field sequential connection structure, extracting maintenance equipment identifiers and contract equipment clauses to perform semantic corresponding judgment, attaching semantic relationships of matched fields to corresponding connection paths, and generating semantic structure association information; S213: Construct a graph neural network graph structure according to the semantic structure association information, maintain the log field and the transaction contract field as nodes in the graph, the connection path as the edge, and the semantic and time association information as the edge attribute, and generate a semantic propagation path graph. 7.The AI-based electric power big data collection and processing method of claim 6, wherein, The obtaining step of the semantic offset field group is specifically: S311: Call the up and down field position index data of the maintenance log field and the transaction contract field in the semantic propagation path graph, extract the offset relationship based on the arrangement order of the up and down field positions, identify the field combination with position corresponding offset in the record, and generate an up and down structure difference field pair; S312: Extract the semantic content of the maintenance equipment identifier and the contract equipment clause in the up and down structure difference field pair, perform key term matching and word meaning consistency judgment, identify the field combination with no common reference item in semantic expression, and record the semantic mismatch identifier on the corresponding path, and generate a semantic mismatch field pair; S313: Combine the up and down structure difference field pair and the semantic mismatch field pair, filter the field group with both semantic expression difference and structure offset, as the field combination with inconsistent semantic expression and sequential dislocation, and generate a semantic offset field group. 8.The AI-based electric power big data collection and processing method of claim 7, wherein, The obtaining step of the path graph after semantic migration is specifically: S411: Call the up and down field positions of the maintenance log field and the transaction contract field in each group of the semantic offset field group, identify the field combination with up and down field position offset, and extract the time parameters corresponding to the field group as the maintenance time and the contract performance time, and generate a structure corresponding field relationship group; S412: According to the structure corresponding field relationship group, judge the connection state of the maintenance equipment identifier and the contract equipment clause in each group of fields in the semantic propagation path graph, detect whether there is a path connection relationship, and judge whether the time is continuous based on the arrangement order of the maintenance time and the contract performance time, filter the field group that has established connection and has time continuity in the semantic propagation path graph, and generate a path connection effective field set; S413: Adjust the connection direction and the up and down field position number of the maintenance log field and the transaction contract field in the semantic propagation path graph in the path connection effective field set respectively, update the connection order and the path direction of the original field in the graph, and generate a path graph after semantic migration. 9.The AI-based electric power big data collection and processing method of claim 8, wherein, The obtaining step of the power big data collection processing result is specifically: S511: Call the connection data of the maintenance log field and the transaction contract field in the semantic migration path graph, filter the field combination that has completed semantic mapping and up and down field position alignment, and generate an aligned field combination list; S512: According to the aligned field combination list, correspondingly arrange the content, up and down field position number and semantic connection direction of each group of fields, map and archive the combination relationship of maintenance behavior description and contract task description, maintenance equipment identifier and contract equipment clause in the structure graph, and generate field comparison structure data; S513: Based on the field comparison structure data, take the field content and the structure relationship as the input field of the power maintenance and transaction data reconciliation processing and dispatching archiving, and generate a power big data collection processing result.

Images