Method and device for aligning operation time sequence data and maintenance text of power equipment

Abstract

The invention relates to a power equipment operation time sequence data and maintenance text alignment method and device. The method comprises the following steps: acquiring a standard structure path comprising a plurality of structure nodes; matching each structure node in the candidate channel pool to obtain a target channel set comprising a plurality of target channels; scoring each target channel according to the structure nodes and the target channels to obtain target channel scores, and selecting the target channels corresponding to the first k target channel scores from high to low as core channels; calling a time sequence data sequence in the approximate time window of each core channel, obtaining a mutation index of the time sequence data sequence, recording the moment when the mutation index exceeds a mutation threshold as a candidate anchor point, and taking the content corresponding to the time window in the time sequence data sequence as a candidate fragment; obtaining candidate fragment scores according to the data, and selecting a candidate fragment corresponding to the highest candidate fragment score as a target fragment; wherein the target fragment is associated and aligned with the maintenance text.

Description

Technical Field

[0001] This application relates to the field of power equipment technology, and in particular to a method and apparatus for aligning power equipment operation sequence data with maintenance text. Background Technology

[0002] During long-term operation, power equipment continuously generates a large amount of time-series data, covering multiple channels such as current, voltage, and temperature. The recording granularity varies and is affected by factors such as sampling packet loss, range switching, and equipment replacement. At the same time, the maintenance documents generated in the field operation and maintenance process are mostly unstructured and free-form descriptions, characterized by inconsistent terminology, disordered expression order, and the coexistence of synonyms and variant spellings.

[0003] In related technologies, it is common practice to compare historical curves and maintenance records with manual experience after the fact, or to map text to a small number of fixed measurement points using rule keywords, so as to achieve the association and alignment between runtime sequence data and maintenance text.

[0004] However, these two methods in related technologies are prone to problems such as path ambiguity, inaccurate channel pointing, and coarse time positioning in complex scenarios. This is especially true when dealing with multi-to-multi measurement point configurations on the same device, missing or unstable data in some channels, and slowly evolving signals before operation. Incorrect alignment and missing alignment coexist, making it difficult to form a structured alignment result that can be accumulated and reused. Therefore, how to accurately associate and align runtime sequence data with maintenance text has become an urgent technical problem to be solved. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for aligning runtime sequence data and maintenance text of power equipment, which can accurately associate and align runtime sequence data with maintenance text, in order to address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for aligning power equipment operation sequence data with maintenance text, the method comprising:

[0007] Obtain the maintenance text and extract the standard structure path from the maintenance text; wherein, the standard structure path includes multiple structure nodes;

[0008] A preset candidate channel pool is obtained, and each of the aforementioned structural nodes is matched in the candidate channel pool to obtain a target channel set; wherein, the target channel set includes multiple target channels;

[0009] Each target channel is scored according to the structure node and the target channel to obtain the target channel score. The target channel scores are sorted from high to low, and the target channels corresponding to the top k target channel scores are selected as core channels, where k is a positive integer.

[0010] Retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the moment when the mutation index exceeds the mutation threshold as a candidate anchor point, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as a candidate segment.

[0011] The candidate segments are scored based on the candidate anchor points, the core channel, the maintenance text, and the preset cross-channel segment synchronization index to obtain candidate segment scores. The candidate segment with the highest score is selected as the target segment. The target segment is associated and aligned with the maintenance text.

[0012] In one embodiment, obtaining the maintenance text and extracting the standard structure path from the maintenance text includes:

[0013] Obtain the maintenance text, parse and extract the maintenance text to obtain the preliminary structure path, perform position verification and structure reordering on the preliminary structure path to obtain the standard structure path.

[0014] In one embodiment, after scoring the candidate segments based on the candidate anchor points, the core channel, the maintenance text, and a preset cross-channel segment synchronization index to obtain candidate segment scores, the method further includes:

[0015] The candidate segments and their scores are jointly calibrated for engineering usability and structural consistency to obtain a candidate segment calibration score.

[0016] The candidate segment with the highest calibration score is selected as the target segment.

[0017] In one embodiment, after performing joint calibration of the candidate fragment and the candidate fragment score on engineering usability and structural consistency to obtain a candidate fragment calibration score, the method further includes:

[0018] The candidate anchor point corresponding to the highest candidate segment calibration score is selected as the target anchor point;

[0019] An index set is collected within a fixed neighborhood centered on the target anchor point; wherein, candidate anchor points within the index set are defined as the target anchor point.

[0020] Based on the target anchor point and the corresponding candidate segment calibration score, a weighted centroid is calculated to obtain the representative anchor point;

[0021] The candidate segment corresponding to the representative anchor point is selected as the target segment.

[0022] In one embodiment, the step of scoring each target channel based on the structural node and the target channel to obtain a target channel score includes:

[0023] Obtain the node weight of each of the aforementioned structural nodes;

[0024] Obtain the path matching score between each of the structural nodes and the target channel;

[0025] Obtain the historical data stability score of the target channel;

[0026] The target channel score is obtained by multiplying and adding the node weight and path matching score of each structure node, and then adding the product of the preset adjustment factor and the historical data stability score.

[0027] In one embodiment, the step of scoring the candidate segments based on the candidate anchor points, the core channel, the maintenance text, and a preset cross-channel segment synchronization index to obtain a candidate segment score includes:

[0028] Obtain the mutation index of the candidate anchor points;

[0029] Obtain the semantic attribute matching weights between the core channel and the maintenance text;

[0030] The candidate fragment score is obtained by multiplying the mutation index, the attribute matching weight, and the cross-channel fragment synchronization index by their respective weights and then summing them.

[0031] In one embodiment, obtaining the mutation index of the time-series data sequence includes:

[0032] A composite mutation exponential function is constructed based on instantaneous slope variation, temporal perturbation degree, and historical stable offset;

[0033] The mutation index is obtained by calculating the time series data sequence based on the composite mutation index function.

[0034] Secondly, this application also provides an alignment device for power equipment operation sequence data and maintenance text, the device comprising:

[0035] The acquisition module is used to acquire maintenance text and extract a standard structural path from the maintenance text; wherein, the standard structural path includes multiple structural nodes;

[0036] A matching module is used to obtain a preset candidate channel pool, and match each of the structure nodes in the candidate channel pool to obtain a target channel set; wherein, the target channel set includes multiple target channels;

[0037] The scoring module is used to score each target channel according to the structure node and the target channel to obtain the target channel score, sort the target channel scores from high to low, and select the target channel corresponding to the highest target channel score as the core channel.

[0038] The retrieval module is used to retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the moment when the mutation index exceeds the mutation threshold as a candidate anchor point, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as a candidate segment.

[0039] The selection module is used to score the candidate segments based on the candidate anchor points, the core channel, the maintenance text, and a preset cross-channel segment synchronization index to obtain candidate segment scores, and select the candidate segment corresponding to the highest candidate segment score as the target segment; wherein, the target segment is associated and aligned with the maintenance text.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0042] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0043] The aforementioned methods, devices, computer equipment, computer-readable storage media, and computer program products for aligning power equipment operation sequence data and maintenance text achieve at least the following beneficial technical effects:

[0044] The core of the technical solution in this application lies in using a structural dictionary and equipment diagram to perform hierarchical parsing of maintenance text, obtaining a standard structural path that uniquely corresponds to the asset ledger, thus eliminating path ambiguity caused by free expression; subsequently, under the constraint of the standard structural path, a candidate channel pool is constructed by integrating multi-source configuration tables, and a candidate channel pool is selected by combining hierarchical weight and channel stability, prioritizing channels that are long-term usable and have a stronger relationship with end components, thus solving the inaccuracy of pointing under many-to-many mapping and data instability; in the time positioning stage, a composite mutation index that takes into account instantaneous slope, local disturbance, and historical offset is introduced, resulting in... The system integrates the correspondence between channel types and text semantics, as well as cross-channel synchronization, to generate candidate segments consistent with behavioral semantics and quantify their credibility, thus accommodating both drastic operations and slow evolution scenarios. During the result convergence phase, the candidate channel scores are calibrated using channel stability, structural hierarchy, and process reference duration. High-scoring segments within the time neighborhood are robustly aggregated, ultimately forming a single structured alignment result containing maintenance text, standard structural paths, representative anchor points, core channels, and target segments. This result can be directly written into the result library and called by the status assessment, maintenance verification, and sample accumulation modules. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart illustrating a method for aligning runtime sequence data and maintenance text for power equipment in one embodiment.

[0047] Figure 2 This is a flowchart illustrating a method for aligning power equipment runtime sequence data with maintenance text in another embodiment.

[0048] Figure 3 This is a flowchart illustrating a method for aligning power equipment runtime sequence data with maintenance text in another embodiment.

[0049] Figure 4 This is a structural block diagram of an alignment device for power equipment runtime sequence data and maintenance text in one embodiment. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0052] In one exemplary embodiment, reference is made to Figure 1 A method for aligning power equipment operating sequence data with maintenance text is provided. This method may include, but is not limited to, the following steps S101 to S105.

[0053] S101, Obtain the maintenance text and extract the standard structure path from the maintenance text; wherein, the standard structure path includes multiple structure nodes.

[0054] The goal of this step S101 is to parse the unstructured maintenance text t recorded in the inspection and maintenance system into a standard structured path that conforms to the hierarchical structure of power equipment. This is used to support subsequent matching with time-series data channels.

[0055] The extraction of standard structural paths must not only cover key semantic information such as the functional components, location nodes, and phases of the equipment, but also satisfy the structural order logic to ensure that the parsing results can establish a consistent relationship with the existing asset configuration table or equipment channel mapping table in the system.

[0056] The maintenance text is entered by on-site maintenance personnel after inspection or troubleshooting via mobile work terminal or back-end input system. The data is usually stored in the operation and maintenance system database and can be accessed and retrieved through REST API or SQL interface.

[0057] As an example, the maintenance text t might be: "The temperature of the B-phase cable head in switchgear #2 is too high; replace it." This type of maintenance text exhibits significant unstructured characteristics, with components, locations, phases, and actions mixed together. It lacks standardized labeling rules and is prone to abbreviations, omissions, or position swaps due to human error. Therefore, after obtaining the maintenance text, it is necessary to parse and extract it to obtain a preliminary structural path. This preliminary structural path then undergoes position verification and structural reordering to obtain a standard structural path.

[0058] Optionally, step S101 may include: obtaining inspection text, parsing and extracting the inspection text to obtain a preliminary structural path, and performing position verification and structural reordering on the preliminary structural path to obtain the standard structural path.

[0059] As an example, a path parsing module based on multi-level dictionary matching and regular expression template parsing mechanism is used. The inspection text is parsed and extracted to obtain a preliminary structural path. As shown in the following formula:

[0060]

[0061] in, This represents the preliminary structural path extracted from the inspection text. It represents an identifiable equipment structural element, such as "2# switch cabinet", "B phase", and "cable head"; the dictionary is automatically generated from existing equipment information in the operation and maintenance knowledge base, including component names (such as busbars, bushings, terminals, etc.), location names (such as "East Zone", "2nd Floor", "4# cabinet"), and phase fields (A, B, C phases), etc.; the matching method is based on regular expression branching and fuzzy matching, and adapts to various expression orders.

[0062] Taking the maintenance report "Severe temperature rise at the outgoing connector of phase B cabinet No. 2" as an example, the analysis results are as follows:

[0063] (Location device) (Separation information) (Functional components).

[0064] In the initial structural path After extraction, since the order in the representation may not match the actual connection order of the devices, further structural graph normalization is required using the structural graph normalization module. For the preliminary structural path Position verification and structural reordering are performed to obtain the standard structural path. As shown in the following formula:

[0065] ;

[0066] in, These are the structural nodes after graph verification, exhibiting valid order, clear hierarchy, and physical channel correspondence capability. (Structure Graph Regularization Module) It is a directed graph verification system built based on the static structural hierarchy diagram derived from the substation asset configuration. Its main functions include: verifying whether adjacent units in the structural path have physical connection relationships, correcting sequence errors (such as "busbar > disconnector > switchgear" being standardized as "switchgear > disconnector > busbar"), and retaining valid sub-paths when there are invalid jumps in the path (such as "cable head > busbar").

[0067] This structural diagram originates from equipment ledger files provided by the power grid company, such as CAD (Computer-Aided Design) diagrams and asset topology XML. It has been uniformly structured into a graph, and path verification utilizes graph traversal to confirm the order. This module ensures that every item in the output path... All of them can be mapped to a unique device identifier (such as "SB00124") so that they can be bound to the signal channel later.

[0068] This method allows for the path from the inspection text t to the standard structure. The parsing process is built entirely on engineering knowledge and system configuration rules, which can adapt to common equipment descriptions in the power industry and improve compatibility with complex and non-standard descriptions, providing operable standard equipment path inputs for subsequent steps.

[0069] S102, obtain a preset candidate channel pool, match each structure node in the candidate channel pool to obtain a target channel set; wherein, the target channel set includes multiple target channels.

[0070] The system builds a candidate channel pool from the device channel configuration table imported during the initial deployment. Candidate channel pool This includes multiple candidate channels. The entries are derived from the SCADA system point table, the intelligent terminal's internal configuration files (such as telemetry channel definition XML), and the manual mapping table of equipment-measurement points provided by the operation and maintenance knowledge platform.

[0071] To construct the mapping from the standard structural path to the candidate channel set, we first target each structural node. Based on semantic fuzzy matching and path prefix matching mechanisms, candidate channels corresponding to each structural node are obtained and defined as target channels. The target channels corresponding to all structural nodes constitute the target channel set. Candidate channels include channel numbers with similar names (e.g., “outgoing terminal” and “outgoing connector T021”) and consistent location prefixes (e.g., “2# switch cabinet” matches the channel prefix “SB002”).

[0072] S103, score each target channel according to the structural nodes and target channels to obtain the target channel score, sort the target channel scores from high to low, and select the target channels corresponding to the top k target channel scores as the core channels.

[0073] To improve mapping accuracy, a structural hierarchy weighting mechanism and a channel state scoring function are introduced to jointly construct a channel scoring function. The core idea is that the final score of a channel is determined not only by its matching hierarchy in the structural path, but also by its operational stability and semantic distance to each node in the structural path.

[0074] Optionally, the step S103, which involves scoring each target channel based on the structural node and the target channel to obtain a target channel score, may include: obtaining the node weight of each structural node; obtaining the path matching score between each structural node and the target channel; obtaining the historical data stability score of the target channel; multiplying and summing the node weight and path matching score of each structural node, and then adding the product of a preset adjustment factor and the historical data stability score to obtain the target channel score. As shown in the following formula:

[0075] ;

[0076] in, Indicates the target channel Target channel score, Represents structural nodes The node weights are typically set so that nodes at the end of the path (such as "cable heads") have higher weights. Represents structural nodes With the target channel The path matching score between them is calculated using a variant of edit distance with decay, such as when and A score of 0.8 is given when the structural component names differ by only 1 character, and a score of 1 is given when they are completely identical. It is the target channel The historical data stability score is defined as the proportion of data continuity for that channel within the most recent T days, ranging from [0,1]. For example, if there is valid data for 28 out of the past 30 days, then... =28 / 30; This is a moderating factor for the stability score, with an empirical range of 0.3 to 0.6; the more emphasis is placed on channel quality, the higher the value. The higher.

[0077] The second term in this formula It is specifically designed for scenarios where channels may be abnormal under long-term operating conditions of power equipment. By using signal quality as an important factor in channel screening, it effectively avoids the risk of channels that are "configured but do not actually have data" being incorrectly selected into the alignment process, demonstrating strong engineering adaptability and creativity.

[0078] After obtaining the scores of all target channels Then, the target channel scores are sorted from highest to lowest, and the target channels corresponding to the top k scores are selected as the core channels. Core Channel Set The value of k is typically between 3 and 5, dynamically adjusted based on the complexity of the device structure. All selected channels possess actual acquired runtime sequence data, which can be used for subsequent mutation identification.

[0079] For example, if the standard structural path The system is configured as "Distribution Cabinet No. 3 > Phase A > Incoming Terminal". The core channels matched by the system include the current channel "IA034" and the temperature channel "TA034". However, "IA034" has a high data loss rate over the past 30 days. =0.2), its total score is low, so it will be excluded from the core channel set C, while "TA034" has good stability ( =0.95), and was selected as the core channel.

[0080] The core channel set C will be returned with the channel number as the primary key. The format is uniform, and all channels have runtime sequence data that can be retrieved from the system database, allowing direct integration into the next step for behavior recognition. This core channel set not only comes from structural path matching results but also incorporates channel status information. It represents a channel selection mechanism focused on availability assurance and precise alignment, demonstrating a deep understanding of the complexity of industrial environments and system-level design capabilities in multimodal alignment tasks.

[0081] Through the above steps S101 to S103, based on the extraction of the maintenance text structure path, the parsed standard structure path is mapped to a set of runtime sequence data channels that actually exist in the power equipment monitoring system, for subsequent behavior mutation identification and alignment fragment generation. Based on the traditional "structure noun → data channel" lookup table mode, structural optimization was carried out for three common problems in the power equipment scenario: (1) There are problems such as non-standard equipment naming, inconsistent levels, and mixed abbreviations in the actual configuration table; (2) The mapping relationship between components and data channels in the structure path is not one-to-one, but many-to-many mapping, such as one outgoing terminal may correspond to multiple measurement points; (3) There may be channels with historical failures, acquisition failures, or unstable accuracy in the running data channels.

[0082] S104, retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the time when the mutation index exceeds the mutation threshold as candidate anchor points, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as candidate segments.

[0083] For each core channel Retrieve time-series data sequences within an approximate time window from the historical data management system. The time range covers a fixed interval before and after the timestamp of the text record (e.g., ±12 hours), and the sampling frequency is automatically adjusted according to the channel type (e.g., temperature 1 min / point, current 5 s / point).

[0084] Optionally, obtaining the mutation index of the time series data sequence in step S104 may include: constructing a composite mutation index function based on instantaneous slope variation, temporal perturbation degree, and historical stable offset; and calculating the mutation index based on the composite mutation index function on the time series data sequence.

[0085] Considering that abrupt changes in power signals often manifest not only as numerical jumps but also as local fluctuation energy anomalies, a composite abrupt change exponential function combining instantaneous slope variation, temporal perturbation degree, and historical stability offset was designed. The abrupt change exponential function is used to calculate the abrupt change exponential index on the time-series data sequence, which is then used to accurately identify anchor points of candidate behaviors. As shown in the following formula:

[0086] ;

[0087] in, Time series data sequence The mutation index. The time independent variable (or sampling index) representing time series data is used to represent the time series data sequence. Sampling time, express forward Local variance within the time window; Calculate the step size for the slope. core channel The historical average for the same time period is calculated specifically according to the criteria of "same equipment, same channel, same time period / operating condition". A common practice is to take the average value of the channel within the same time window (e.g., bandwidth at the same time of the same day, or the same load / shift interval) over the past several days as the baseline; , These control the response intensity to local disturbances and background shifts, respectively, and their values ​​can be adjusted within the range of [0.2, 0.5] according to the scene.

[0088] Calculation by sliding window on each core channel The mutation index exceeds the mutation threshold. The moment was recorded as a candidate anchor point , with each Construct a fixed-length time window centered on the time series data, and use the content corresponding to the time window in the time series data sequence as candidate segments. All candidate segments constitute a candidate segment set. In one example, adjacent candidate anchor points that are too close together are merged to avoid repeated triggering: if the interval between two adjacent candidate anchor points is less than the merging threshold (it is recommended to be no less than half the window width), then segments within that interval are retained. The single candidate anchor point corresponding to the peak value is discarded, while the others are retained; candidate anchor points that are further apart are retained.

[0089] S105: Based on the candidate anchor point, core channel, maintenance text, and preset cross-channel segment synchronization index, the candidate segment is scored to obtain the candidate segment score, and the candidate segment corresponding to the highest candidate segment score is selected as the target segment; wherein, the target segment is associated and aligned with the maintenance text.

[0090] Relying solely on physical signal mutations is insufficient to confirm whether these candidate segments are related to the content described in the maintenance text t. Therefore, a semantic alignment mechanism is needed for further screening. The introduced scoring function integrates three dimensions of evaluation: signal physical characteristics, channel semantic correspondence, and cross-channel behavioral consistency. Specifically, candidate segments are scored based on candidate anchor points, core channels, maintenance text, and preset cross-channel segment synchronization indicators to obtain candidate segment scores.

[0091] Optionally, step S105, which scores the candidate segments based on the candidate anchor points, the core channel, the maintenance text, and a preset cross-channel segment synchronization index to obtain a candidate segment score, may include: obtaining the mutation index of the candidate anchor points; obtaining the semantic attribute matching weights between the core channel and the maintenance text; and multiplying the mutation index, attribute matching weights, and cross-channel segment synchronization index by their respective weights and then summing them to obtain the candidate segment score. As shown in the following formula:

[0092] ;

[0093] in, Indicates candidate fragments The candidate segment score. Candidate anchor points The corresponding mutation index, , and These represent the weighting coefficients for mutation intensity, semantic matching degree, and cross-channel synchronization, respectively, with default values ​​of approximately 0.4, 0.3, and 0.3. Indicates core channel The semantic attribute matching weights of the inspection text t are calculated based on the knowledge base correspondence between channel type (such as temperature, current) and text keywords (such as "heating" and "tripping"). "Indicates cross-channel segment synchronization index", reflected in the current candidate segment. Within the given time interval, whether abrupt changes also occur in other channels, and whether they are concentrated at downstream nodes of the structural path, is defined as follows:

[0094] ;

[0095] in, It is an index function, when , If it is 1, then it is 0. Indicates channel At any moment The mutation index value calculated here comes from the composite mutation index function defined in the above steps. This indicates the threshold for cross-channel synchronization determination. This represents the number of channels in set C. For channel The hierarchical weights in the structural path. This indicator reflects the core idea of ​​combining structural path with timing synchronization: a real equipment operation or failure phenomenon often produces a chain reaction in multiple channels of the structural path, rather than just an isolated mutation.

[0096] Furthermore, after calculating the candidate segment scores for all candidate segments, the candidate segment with the highest score is selected as the target segment. The target segment is then aligned with the inspection text.

[0097] The technical solution of this application embodiment involves acquiring maintenance text and extracting a standard structural path from it. The standard structural path includes multiple structural nodes. A preset candidate channel pool is obtained, and each structural node is matched within the pool to obtain a target channel set. This target channel set includes multiple target channels. Each target channel is scored based on its structural nodes and target channel, resulting in a target channel score. The target channel scores are sorted from highest to lowest, and the target channels corresponding to the top k scores are selected as core channels. A time-series data sequence within an approximate time window of each core channel is retrieved, and the mutation index of the time-series data sequence is obtained. Moments where the mutation index exceeds a mutation threshold are recorded as candidate anchor points. A time window is constructed centered on each candidate anchor point, and the content corresponding to the time window in the time-series data sequence is selected as candidate segments. The candidate segments are scored based on the candidate anchor points, core channels, maintenance text, and a preset cross-channel segment synchronization index, resulting in candidate segment scores. The candidate segment with the highest score is selected as the target segment; the target segment is associated and aligned with the maintenance text. This achieves accurate association and alignment between candidate segments in runtime time-series data and maintenance text.

[0098] It should be noted that finding candidate segments aligned with the maintenance text is to locate the behavioral segments most likely to correspond to the maintenance text description in a large amount of runtime sequence data, determine the specific data segments corresponding to the events described in the maintenance text, and ensure that the segments carry complete signal change characteristics, thereby reflecting the entire process of operation or fault and achieving precise correlation between time, channel and semantics.

[0099] In one exemplary embodiment, reference is made to Figure 2 In step S105, candidate segments are scored based on candidate anchor points, core channels, maintenance text, and preset cross-channel segment synchronization indicators. After obtaining the candidate segment scores, the alignment method for power equipment operation sequence data and maintenance text may also include, but is not limited to, the following steps S206 to S207. The alignment method for power equipment operation sequence data and maintenance text may include the following steps S201 to S207.

[0100] S201, Obtain the maintenance text and extract the standard structure path from the maintenance text; the standard structure path includes multiple structure nodes.

[0101] S202, obtain a preset candidate channel pool, match each structure node in the candidate channel pool to obtain a target channel set; wherein, the target channel set includes multiple target channels.

[0102] S203, score each target channel according to the structural nodes and target channels to obtain the target channel score, sort the target channel scores from high to low, and select the target channels corresponding to the top k target channel scores as the core channels.

[0103] S204, retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the time when the mutation index exceeds the mutation threshold as candidate anchor points, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as candidate segments.

[0104] S205: Based on the candidate anchor point, core channel, maintenance text, and preset cross-channel segment synchronization index, the candidate segment is scored to obtain the candidate segment score.

[0105] S206, perform joint calibration of candidate segments and candidate segment scores for engineering usability and structural consistency to obtain candidate segment calibration scores.

[0106] First, a joint calibration of engineering usability and structural consistency is performed on each candidate segment and its corresponding candidate segment score to obtain the candidate segment calibration score. As shown in the following formula:

[0107] ;

[0108] in, This indicates the score for the candidate segment. Indicates candidate fragments The corresponding core channel, Indicates core channel Historical data stability score. Indicates core channel Hierarchical weights in the structural path. The window length is indicated by the window truncation strategy recorded in step S104. This indicates the reference window length, with recommended intervals of [0.2, 0.6], [0.2, 0.6], and [0.05, 0.3]. The event type is provided by the operations and maintenance knowledge base. When the event type cannot be obtained, the default value based on the channel type is used. The adjustment logic is as follows: if the average duration of candidate segments is systematically short, then increase the value. Conversely, reduce the size to maintain the same window construction diameter as in step S104. , , This is the hyperparameter group, default. . , Strengthen long-term channel availability and prioritize terminal-side infrastructure. Suppress candidate segments whose durations significantly deviate from typical process times. If quality inspection reveals a recent degradation in the channel's condition (i.e., ... If the weighting is significantly reduced, the candidate segment will be automatically downweighted. The corresponding channel belongs to the standard structure path. End-point critical components (i.e.) Higher scores will receive bonus points for structural consistency; if If a candidate fragment deviates from the reference range, it will be penalized to prevent excessively short or long candidate fragments from affecting downstream availability.

[0109] S207, select the candidate segment corresponding to the highest candidate segment calibration score as the target segment.

[0110] Based on the above, the technical solution of this application embodiment, by jointly calibrating the candidate segments and candidate segment scores for engineering usability and structural consistency, obtains the candidate segment calibration score, and selects the candidate segment corresponding to the highest candidate segment calibration score as the target segment, further improving the accuracy of associating and aligning runtime sequence data with maintenance text.

[0111] In one exemplary embodiment, reference is made to Figure 3 After performing joint calibration of engineering usability and structural consistency on the candidate segments and candidate segment scores in step S201 to obtain the candidate segment calibration scores, the alignment method for power equipment operation sequence data and maintenance text may also include, but is not limited to, the following steps S307 to S310. The alignment method for power equipment operation sequence data and maintenance text may include the following steps S301 to S310.

[0112] S301, Obtain the maintenance text and extract the standard structure path from the maintenance text; wherein, the standard structure path includes multiple structure nodes.

[0113] S302, obtain a preset candidate channel pool, match each structure node in the candidate channel pool to obtain a target channel set; wherein, the target channel set includes multiple target channels.

[0114] S303: Score each target channel according to the structural nodes and target channels to obtain the target channel score. Sort the target channel scores from high to low and select the target channel corresponding to the highest target channel score as the core channel.

[0115] S304, retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the time when the mutation index exceeds the mutation threshold as candidate anchor points, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as candidate segments.

[0116] S305: Based on the candidate anchor point, core channel, maintenance text, and preset cross-channel segment synchronization index, the candidate segment is scored to obtain the candidate segment score.

[0117] S306, perform joint calibration of candidate segments and candidate segment scores for engineering usability and structural consistency to obtain candidate segment calibration scores.

[0118] S307, select the candidate anchor point corresponding to the highest candidate segment calibration score as the target anchor point.

[0119] S308, collect an index set within a fixed neighborhood centered on the target anchor point; wherein, the candidate anchor points within the index set are defined as the target anchor point.

[0120] In steps S307 to S308 of some embodiments, the candidate segment calibration scores are sorted from high to low, and the highest score index corresponding to the highest candidate segment calibration score is selected. , the highest score index The corresponding candidate anchor point is defined as the target anchor point. To handle the situation where multiple candidate segments coexist in the neighborhood at the same time, the target anchor point is used. Collect an index set within a fixed neighborhood centered on the target. Candidate anchor points within the index set are defined as target anchor points. The neighborhood radius, consistent with the window construction in step S104, is given by system parameters.

[0121] S309, calculate the weighted centroid based on the target anchor point and the corresponding candidate segment calibration score to obtain the representative anchor point.

[0122] S310, select the candidate segment corresponding to the anchor point as the target segment.

[0123] In steps S309 to S310 of some embodiments, a weighted centroid is calculated based on the target anchor point and the corresponding candidate fragment calibration score to obtain the representative anchor point. As shown in the following formula:

[0124] ;

[0125] in, Indicates the candidate fragment calibration score. Indicates index The corresponding target anchor point. Select a representative anchor point. The corresponding candidate segment is the target segment.

[0126] The technical solution of this application embodiment collects an index set and selects representative anchor points to smooth out the time jitter caused by abrupt changes in multiple channels or neighboring channels, making the output time points more robust, thereby making the target segment more representative and interpretable.

[0127] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0128] Based on the same inventive concept, this application also provides an apparatus for aligning power equipment runtime sequence data and maintenance text, which implements the above-described method for aligning power equipment runtime sequence data and maintenance text. The solution provided by this apparatus is similar to the implementation described in the above-described method. Therefore, the specific limitations of one or more embodiments of the apparatus for aligning power equipment runtime sequence data and maintenance text provided below can be found in the limitations of the alignment method for power equipment runtime sequence data and maintenance text described above, and will not be repeated here.

[0129] In one exemplary embodiment, such as Figure 4 As shown, an apparatus for aligning power equipment runtime sequence data and maintenance text is provided, comprising:

[0130] The acquisition module 410 is used to acquire maintenance text and extract a standard structural path from the maintenance text; wherein the standard structural path includes multiple structural nodes.

[0131] The matching module 420 is used to obtain a preset candidate channel pool, and match each of the structure nodes in the candidate channel pool to obtain a target channel set; wherein the target channel set includes multiple target channels.

[0132] The scoring module 430 is used to score each target channel according to the structure node and the target channel to obtain the target channel score, sort the target channel scores from high to low, and select the target channel corresponding to the previous target channel score as the core channel.

[0133] The retrieval module 440 is used to retrieve the time series data sequence within the approximate time window of each core channel, obtain the mutation index of the time series data sequence, record the moment when the mutation index exceeds the mutation threshold as a candidate anchor point, construct a time window with each candidate anchor point as the center, and take the content corresponding to the time window in the time series data sequence as a candidate segment.

[0134] The selection module 450 is used to score the candidate segments based on the candidate anchor point, the core channel, the maintenance text, and a preset cross-channel segment synchronization index to obtain a candidate segment score, and select the candidate segment corresponding to the highest candidate segment score as the target segment; wherein, the target segment is associated and aligned with the maintenance text.

[0135] In an exemplary embodiment, the acquisition module 410 is used to acquire maintenance text, parse and extract the maintenance text to obtain a preliminary structural path, and perform position verification and structural reordering on the preliminary structural path to obtain the standard structural path.

[0136] In an exemplary embodiment, a selection module 450 is used to perform joint calibration of the candidate fragment and the candidate fragment score for engineering usability and structural consistency to obtain a candidate fragment calibration score; the candidate fragment corresponding to the highest candidate fragment calibration score is selected as the target fragment.

[0137] In an exemplary embodiment, the selection module 450 is used to select the candidate anchor point corresponding to the highest candidate segment calibration score as the target anchor point; collect an index set in a fixed neighborhood centered on the target anchor point; wherein, the candidate anchor points in the index set are defined as target anchor points; perform weighted centroid calculation based on the target anchor point and the corresponding candidate segment calibration score to obtain a representative anchor point; and select the candidate segment corresponding to the representative anchor point as the target segment.

[0138] In an exemplary embodiment, the scoring module 430 is used to obtain the node weight of each of the structural nodes; obtain the path matching score between each of the structural nodes and the target channel; obtain the historical data stability score of the target channel; multiply and add the node weight and the path matching score of each of the structural nodes, and add the product of a preset adjustment factor and the historical data stability score to obtain the target channel score.

[0139] In an exemplary embodiment, module 450 is selected to obtain the mutation index of the candidate anchor point; obtain the semantic attribute matching weight between the core channel and the maintenance text; and multiply the mutation index, the attribute matching weight, and the cross-channel segment synchronization index by their respective weights and then add them together to obtain the candidate segment score.

[0140] In an exemplary embodiment, the retrieval module 440 is configured to construct a composite mutation index function based on instantaneous slope variation, temporal perturbation degree, and historical stable offset; and to calculate the mutation index based on the composite mutation index function on the time series data sequence.

[0141] The modules in the aforementioned alignment device for power equipment runtime sequence data and maintenance text can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the operations corresponding to each module.

[0142] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0143] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0144] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0145] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0146] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0147] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for aligning power equipment operation sequence data and maintenance text, characterized in that, The method comprises: acquiring a maintenance text, and extracting a standard structure path from the maintenance text; wherein the standard structure path comprises a plurality of structure nodes; acquiring a preset candidate channel pool, and matching each structure node in the candidate channel pool to obtain a target channel set; wherein the target channel set comprises a plurality of target channels; scoring each target channel according to the structure node and the target channel to obtain a target channel score, sorting the target channel scores from high to low, and selecting the target channels corresponding to the top k target channel scores as core channels, wherein k is a positive integer; acquiring a time series data sequence within an approximate time window of each core channel, acquiring a mutation index of the time series data sequence, recording a time point at which the mutation index exceeds a mutation threshold as a candidate anchor point, constructing a time window centered on each candidate anchor point, and taking the content corresponding to the time window in the time series data sequence as a candidate segment; scoring the candidate segments according to the candidate anchor points, the core channels, the maintenance text, and a preset cross-channel segment synchronicity indicator to obtain a candidate segment score, and selecting a candidate segment corresponding to the highest candidate segment score as a target segment; wherein the target segment is associated and aligned with the maintenance text.

2. The method of claim 1, wherein, The method comprises: acquiring a maintenance text, and extracting a standard structure path from the maintenance text; wherein the standard structure path comprises a plurality of structure nodes; 3. The method of claim 1, wherein, acquiring a maintenance text, and extracting a standard structure path from the maintenance text; wherein the standard structure path comprises a plurality of structure nodes; The method further comprises: jointly calibrating the candidate segments and the candidate segment scores for engineering usability and structural consistency to obtain a candidate segment calibration score; 4. The method of claim 3, wherein, selecting a candidate segment corresponding to the highest candidate segment calibration score as a target segment. The method further comprises: selecting a candidate anchor point corresponding to the highest candidate segment calibration score as a target anchor point; collecting an index set within a fixed neighborhood centered on the target anchor point; wherein a candidate anchor point in the index set is defined as a target anchor point; performing weighted centroid calculation on the target anchor point and the corresponding candidate segment calibration score to obtain a representative anchor point; 5. The method of claim 1, wherein, selecting a candidate segment corresponding to the representative anchor point as the target segment. The method further comprises: acquiring a node weight of each structure node; acquiring a path matching score of each structure node and the target channel; acquiring a historical data stability score of the target channel; The node weight of each structure node, the path matching score are multiplied and added, and the product of a preset adjustment factor and the historical data stability score is added to obtain the target channel score.

6. The method of claim 1, wherein, The candidate segment is scored according to the candidate anchor point, the core channel, the maintenance text and a preset cross-channel segment synchronicity index to obtain a candidate segment score. The mutation index of the candidate anchor point is obtained. The attribute matching weight of the core channel and the maintenance text in semantics is obtained. The mutation index, the attribute matching weight and the cross-channel segment synchronicity index are multiplied by corresponding weights and then added to obtain the candidate segment score.

7. The method of claim 1, wherein, The mutation index of the time sequence data sequence is obtained, including: A composite mutation index function is constructed according to instantaneous slope variation, time domain disturbance degree and historical stability offset; The time sequence data sequence is calculated according to the composite mutation index function to obtain the mutation index.

8. An alignment device for power equipment operation sequence data and maintenance text, characterized in that, The device comprises: An acquisition module is configured to acquire a maintenance text and extract a standard structure path from the maintenance text, wherein the standard structure path comprises a plurality of structure nodes; A matching module is configured to acquire a preset candidate channel pool, match each structure node in the candidate channel pool to obtain a target channel set, wherein the target channel set comprises a plurality of target channels; The scoring module is configured to score each of the target channels according to the structure nodes and the target channels, to obtain target channel scores, to sort the target channel scores from high to low, and to select target channels corresponding to the top target channel scores as core channels. A calling module is configured to call time sequence data sequence in a core channel approximate time window, obtain a mutation index of the time sequence data sequence, record a time point at which the mutation index exceeds a mutation threshold as a candidate anchor point, and construct a time window with each candidate anchor point as a center to obtain a candidate segment corresponding to the time window in the time sequence data sequence; A selection module is configured to score the candidate segment according to the candidate anchor point, the core channel, the maintenance text and a preset cross-channel segment synchronicity index to obtain a candidate segment score, and select a candidate segment corresponding to the highest candidate segment score as a target segment, wherein the target segment is associated and aligned with the maintenance text.

9. A computer device, comprising: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the alignment method for power equipment operation time sequence data and maintenance text according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the alignment method for power equipment operation time sequence data and maintenance text according to any one of claims 1 to 7.

Images