A Distributed New Energy Operation and Maintenance Decision Support Method and System Based on Knowledge Graph

Abstract

This invention discloses a distributed new energy operation and maintenance decision support method and system based on knowledge graphs, relating to the field of operation and maintenance management technology. The method includes: instantiating and constraining a knowledge graph event view set, and completing verifiable annotation by referencing evidence chains one by one; simultaneously pruning the feasible domain of the state view to obtain a guardrail constraint set; performing spatiotemporal inference and quantile statistics on the guardrail constraint set based on the knowledge graph to generate a lower bound, median, and upper bound of the quantile sequence, while defining quantile bandwidths and summarizing them into a quantile bandwidth table; calculating backup costs and emergency repair benefits on the knowledge graph based on the quantile bandwidth table, generating backup suggestions and task candidates, and assigning values ​​to the state view in conjunction with the guardrail constraint set to output a time-series scheduling scheme; adding evidence descriptions and time constraints to the time-series scheduling scheme, and uniformly summarizing them into an execution delivery list. This invention improves the adaptability of the decision scheme to complex operation and maintenance environments.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management technology, and in particular to a distributed new energy operation and maintenance decision support method and system based on knowledge graphs. Background Technology

[0002] With the ongoing global energy transition, the large-scale integration of distributed renewable energy sources has brought new challenges to power system operation and maintenance. Current research and practice in this field focus on using data-driven methods for equipment condition monitoring and fault early warning, and on building digital models to support operation and maintenance decisions. Conventional methods typically rely on time-series databases and relational data models to clean, align, and store collected multi-source observation data. Then, based on preset rules or statistical analysis models, abnormal states are identified, and preliminary operation and maintenance strategy suggestions are generated, aiming to improve the timeliness and relevance of operation and maintenance work.

[0003] However, existing conventional methods still have room for improvement in addressing the complex interrelationships and uncertainties of distributed renewable energy operation and maintenance. On the one hand, the spatiotemporal correlations between multi-source data and domain knowledge are not fully formalized and utilized, leading to a reliance on empirical judgment for root cause analysis and impact scope analysis of abnormal events, lacking explicit and verifiable evidence chains, thus affecting the transparency and credibility of the decision-making process. On the other hand, when generating operation and maintenance strategies, conventional methods typically use deterministic thresholds or single scenarios for optimization to address the uncertainties caused by equipment status fluctuations and changes in constraints, making it difficult to dynamically quantify and evaluate the costs and benefits of different decision-making schemes at different confidence levels. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a knowledge graph-based distributed new energy operation and maintenance decision support method to solve the problems of insufficient decision transparency and limited ability to cope with uncertainty in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a distributed new energy operation and maintenance decision support method based on knowledge graphs, which includes, Collect multi-source observation data and form a state view using a knowledge graph as a carrier. Standardize the state view by anomaly identification to obtain a graph event view set. Instantiate constraints on the graph event view set, and complete verifiable annotations by referencing the evidence chain index one by one. At the same time, perform feasible domain pruning on the state view to obtain the guardrail constraint set. The guardrail constraint set is subjected to spatiotemporal inference and quantile statistics based on knowledge graphs to generate the lower bound, median and upper bound of the quantile sequence. At the same time, the quantile bandwidth is defined and summarized into a quantile bandwidth table. Based on the quantile bandwidth table, the backup cost and emergency repair benefit are calculated in the knowledge graph, backup suggestions and task candidates are generated, and the state view is assigned values ​​in conjunction with the guardrail constraint set to output the time-series scheduling scheme. Add evidence and time constraints to the timing scheduling scheme, and summarize them into an execution delivery list.

[0007] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the steps for forming the state view are as follows: Collect multi-source observation data and record the observation sources, establish a unified timeline, and extract the hourly time; Multi-source observation data are mapped to the nearest integer point and piecewise linear interpolation is performed on missing integer points. The units of multi-source observation data are unified, and the conversion coefficients and conversion directions are recorded to obtain aligned multi-source observation data. The system will add device identifiers and observation sources to the aligned multi-source observation data, associate the aligned multi-source observation data with the operation and maintenance objects and multi-domain relationships, and generate attribute snapshots based on the knowledge graph, with the operation and maintenance objects as entity nodes and multi-domain relationships as relationship edges, forming a state view with the entity nodes and relationship edges.

[0008] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the process of obtaining the graph event view set refers to extracting the aligned multi-source observation data observation values ​​of the state view, comparing them, marking abnormal fluctuation records and data mutation records, merging them into graph standardized events, and summarizing them into a graph event view set.

[0009] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the steps of instantiating and constraining the knowledge graph event view set and completing verifiable annotation by referencing the evidence chain index one by one are as follows: Based on the graph event view set, locate the entity nodes and relationship edges of the operation and maintenance objects in the status view, read the hourly time and attribute snapshots, and determine the scope of effect; register the constraint direction and limit range on the entity nodes and relationship edges and reference the evidence link; merge the graph standardized event entries according to the hourly time and the operation and maintenance objects to obtain the constraint record; Based on the standardized events and evidence links in the constraint record graph, an evidence chain index is generated, written to the attribute snapshot, and a searchable reference is established to complete the verifiable annotation.

[0010] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the method of obtaining the guardrail constraint set refers to performing feasible domain pruning on entity nodes and relation edges in the state view according to the constraint direction and restriction range to obtain the guardrail feasible domain, and summarizing the guardrail feasible domain according to the hourly time to obtain the guardrail constraint set.

[0011] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the step of summarizing into a quantile bandwidth table is as follows: Generate an ordered set of hourly times based on the hourly time. Based on the set of guardrail constraints, read the feasible region of the guardrail. Group the feasible region of the guardrail according to the constraint direction. Then, merge the feasible region of the guardrail into a candidate set of feasible regions of the guardrail according to the limited range. Spatiotemporal inference is performed on the candidate set of feasible regions for guardrails. The operation and maintenance objects are compared with the candidate set of feasible regions for guardrails along the multi-domain relationship. The consistent intervals and conflict markers of the guardrails are registered. The consistent intervals of the guardrails are globally merged to obtain the union set and trim the feasible regions of the guardrails. Read the lower boundary and upper boundary of the feasible region of the union clipping fence, perform quantile statistics to obtain the lower and upper boundaries of the quantile sequence, define the quantile bandwidth and the median value of the quantile sequence, and summarize them into a quantile bandwidth table.

[0012] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the steps for generating backup suggestions and task candidates are as follows: The quantile bandwidth of the quantile bandwidth table is divided into uncertainties; Calculate the interval consistency ratio and the proportion of observation sources to obtain the credibility of the operation and maintenance object. Based on the conflict marker, obtain the corrected credibility by combining the credibility of the operation and maintenance object with the interval consistency ratio and the proportion of observation sources. Based on the uncertainty level classification of quantile bandwidth, the backup cost and emergency repair benefit are calculated by using quantile bandwidth and correction confidence. The entity nodes and relation edges are sorted according to the backup cost and the repair benefit, respectively, to obtain backup suggestions and task candidates.

[0013] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the steps of the output time-series scheduling scheme are as follows: Assign values ​​to the status view at the hourly time set and mark the assignment status as complete; All operation and maintenance objects and all hourly time sets are assigned values ​​and integrated to generate a time-series scheduling scheme.

[0014] As a preferred embodiment of the knowledge graph-based distributed new energy operation and maintenance decision support method of the present invention, the step of summarizing into an execution delivery list is as follows: According to the time-series scheduling scheme, the effective range of the feasible region of the guardrail is retrieved from the guardrail constraint set, and the intersection of this range with the set of hourly times is used as the time limit constraint. The effective range of the feasible domain of the guardrail is retrieved, the inclusion relationship and intersection relationship are compared, the guardrail consistency statement and intersection statement are generated, and the guardrail consistency statement and intersection statement are summarized into evidence statement; Match the time constraints with the evidence and combine them into an execution delivery list.

[0015] Secondly, the present invention provides a distributed new energy operation and maintenance decision support system based on knowledge graph, including a data acquisition and mapping module, which collects multi-source observation data and forms a state view using knowledge graph as a carrier, performs anomaly identification and standardization on the state view, and obtains a graph event view set; The event constraint pruning module instantiates constraints on the graph event view set and completes verifiable annotation by referencing the evidence chain index one by one. At the same time, it prunes the feasible domain of the state view to obtain the guardrail constraint set. The spatiotemporal quantile statistics module performs spatiotemporal inference and quantile statistics on the guardrail constraint set based on knowledge graphs, generates the lower bound, median value and upper bound of the quantile sequence, and defines the quantile bandwidth, which is then summarized into a quantile bandwidth table. The backup repair assignment module, based on the quantile bandwidth table, calculates the backup cost and repair benefit in the knowledge graph, generates backup suggestions and task candidates, and assigns values ​​to the state view in conjunction with the guardrail constraint set, and outputs a time-series scheduling scheme. The delivery description and time limit module adds evidence and time limit constraints to the time-series scheduling scheme, and summarizes them into an execution delivery list.

[0016] The beneficial effects of this invention are as follows: by generating a set of guardrail constraints, the transparency and reliability of operation and maintenance decisions are improved, and the auditability and credibility of the anomaly handling process are enhanced; by generating a quantile bandwidth table, the uncertainty range of the system state is dynamically reflected, providing robust data support for subsequent cost and benefit calculations, and improving the adaptability of the decision-making scheme to complex operation and maintenance environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1This is a flowchart of a knowledge graph-based distributed new energy operation and maintenance decision support method.

[0019] Figure 2 This is a schematic diagram of a distributed new energy operation and maintenance decision support system based on knowledge graphs.

[0020] Figure 3 A flowchart for obtaining the guardrail constraint set.

[0021] Figure 4 This is a flowchart for outputting the timing scheduling scheme. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a distributed new energy operation and maintenance decision support method based on knowledge graphs, including the following steps: S1. Collect multi-source observation data and form a state view using a knowledge graph as a carrier. Standardize the state view by identifying anomalies to obtain a graph event view set.

[0026] Multi-source observation data were collected at the edge of multiple sites and the observation sources were recorded. A unified timeline was established and the hourly time was extracted.

[0027] Furthermore, multi-source observation data is collected and the observation sources are recorded. A unified timeline is established for the multi-source observation data, and the hourly time is extracted.

[0028] It should be noted that multi-source observation data includes operational data, environmental data, and sensing data.

[0029] It should be noted that environmental data includes temperature data, radiation data, and wind speed data.

[0030] Multi-source observation data are mapped to the nearest integer point and piecewise linear interpolation is performed on missing integer points. The units of the multi-source observation data are unified, and the conversion coefficients and conversion directions are recorded to obtain aligned multi-source observation data.

[0031] Furthermore, the observation time of each multi-source observation data is extracted, and the time difference between the observation time and the two consecutive integer points is calculated. The multi-source observation data is then mapped to the integer point with the smallest time difference. If the two time differences are equal, the multi-source observation data is mapped to the previous integer point. After mapping, for missing integer points, piecewise linear interpolation is performed based on the multi-source observation data of the two adjacent aligned integer points and the time ratio between the missing integer point and the two adjacent aligned integer points. The original units of the interpolated multi-source observation data are converted to a unified unit, and the conversion coefficient and conversion direction are recorded to obtain aligned multi-source observation data.

[0032] It should be noted that converting the original units of the multi-source observation data that has been interpolated into a unified unit means unifying units of different magnitudes into the same magnitude. For example, temperature data is converted to degrees Celsius, irradiance data is converted to watts per square meter, and wind speed data is converted to meters per second.

[0033] The system will add device identifiers and observation sources to the aligned multi-source observation data, associate the aligned multi-source observation data with the operation and maintenance objects and multi-domain relationships, and generate attribute snapshots based on the knowledge graph, with the operation and maintenance objects as entity nodes and multi-domain relationships as relationship edges, forming a state view with the entity nodes and relationship edges.

[0034] Furthermore, the device number or tag number of the operational data and the device number or tag number corresponding to the geographic coordinates of the sensing data are used as device identifiers; device identifiers and observation sources are added to the aligned multi-source observation data.

[0035] Each aligned multi-source observation data is associated with its corresponding operation and maintenance object along a unified timeline, with the operation and maintenance object as an entity node and the multi-domain relationship as a relation edge. Using the hourly time as an index, the aligned multi-source observation data observation value corresponding to the hourly time is registered as an attribute snapshot of the entity node. Based on the entity node, relation edge, and attribute snapshot, a state view based on a knowledge graph is constructed, and the state view is divided into knowledge graph fragments according to the operation and maintenance object.

[0036] It should be noted that the operation and maintenance objects refer to the objects involved in operation and maintenance, including equipment, sites, lines, thermal loops, energy storage units, and market nodes; multi-domain relationships refer to the associations between operation and maintenance objects, including electricity, heat, gas, control, geography, maintenance dependencies, and contractual coupling.

[0037] It should be noted that the attribute snapshot includes the hourly time on a unified timeline, device identifier, observation source, unified unit, conversion factor, conversion direction, and aligned multi-source observation data observations.

[0038] The observations of the aligned multi-source observation data in the state view are extracted and compared. Records of abnormal fluctuations and data mutations are marked, merged into spectral standardized events, and summarized into a spectral event view set.

[0039] Furthermore, the aligned multi-source observation data of the same maintenance object in the status view at consecutive hourly intervals are extracted and compared to mark abnormal fluctuation records and data mutation records. The abnormal fluctuation records and data mutation records are merged into graph standardized events, and the hourly time, maintenance object, equipment identifier, observation source, unified unit, conversion factor, conversion direction and corresponding aligned multi-source observation data of each graph standardized event are determined, forming graph standardized event entries with maintenance objects as entity nodes and hourly time as index. All graph standardized event entries are sorted by hourly time and maintenance object and summarized into a graph event view set.

[0040] It should be noted that abnormal fluctuation records refer to situations in which the aligned multi-source observation data values ​​change direction back and forth or fluctuate in value multiple times within consecutive whole hours; data mutation records refer to situations in which the aligned multi-source observation data values ​​of the same operation and maintenance object show a single jump with a significantly increased amplitude between two adjacent whole hours.

[0041] S2. Instantiate constraints on the graph event view set, and complete verifiable annotations by referencing the evidence chain index one by one. At the same time, perform feasible domain pruning on the state view to obtain the guardrail constraint set.

[0042] Based on the graph event view set, locate the entity nodes and relationship edges of the operation and maintenance objects in the status view, read the hourly time and attribute snapshots, and determine the effective scope; register the constraint direction and limit range on the entity nodes and relationship edges and reference the evidence link; merge the graph standardized event entries according to the hourly time and the operation and maintenance objects to obtain the constraint record.

[0043] Furthermore, the graph event view set is instantiated and constrained in the knowledge graph shards. The entity nodes and related relationship edges of the corresponding operation and maintenance objects are located in the state view. The hourly times and attribute snapshots of the entity nodes on a unified timeline are read. The hourly times of the graph-standardized event entries in the graph event view set are used as the effective scope, and the constraint direction and limitation range are registered for the entity nodes and relationship edges within this scope. Evidence links are also referenced for the entity nodes and relationship edges. Graph-standardized event entries belonging to the same operation and maintenance object at the same hourly time are merged to form constraint records indexed by the hourly time. These constraint records maintain a one-to-one correspondence with the graph event view set, supporting retrieval and tracing in the state view by entity nodes, relationship edges, hourly times, and evidence links.

[0044] It should be noted that evidence links refer to references that can be located within the status view, aligning snapshots of observations and attributes from multi-source observation data.

[0045] It should be noted that the constraint direction is determined based on the type of the graph normalization event and the direction of change of consecutive integer points. If the type of the graph normalization event is a data mutation record, when the data mutation record shows an upward jump between two adjacent integer points, the constraint direction is set downward; when the data mutation record shows a downward jump between two adjacent integer points, the constraint direction is set upward. If the type of the graph normalization event is an abnormal fluctuation record, the first integer point after the end of the abnormal fluctuation record is designated as the index integer point. The direction of change of the index integer point and the integer point preceding the index integer point is recorded, and the constraint direction is set in the opposite direction to the direction of change of the index integer point and the integer point preceding the index integer point.

[0046] It should be noted that the limiting range is calculated by taking the absolute difference between the current hourly aligned multi-source observation data and the previous hourly aligned multi-source observation data and the next hourly aligned multi-source observation data, under the same unit, and selecting the larger absolute difference as the limiting range.

[0047] Based on the standardized events and evidence links in the constraint record graph, an evidence chain index is generated, written to the attribute snapshot, and a searchable reference is established to complete the verifiable annotation.

[0048] Furthermore, the hourly time, device identifier, observation source, unified unit, conversion factor, conversion direction, and aligned multi-source observation data of the standardized event entries in the graph are extracted and merged with the evidence links to form an evidence chain index, which is numbered according to the hourly time. Attribute snapshots of the entity nodes and relational edges of the maintenance object are written into the evidence chain index, and the entity nodes, relational edges, and hourly time are registered within the evidence chain index, forming a searchable reference that can be retrieved bidirectionally by entity nodes, relational edges, and hourly time. When constraint records are formed at the same hourly time, they are numbered according to the evidence chain index within the constraint records. Multiple evidence chain indexes are appended in ascending order, and the hourly time, device identifier, and observation source corresponding to each evidence chain index are retained in the constraint record. The corresponding evidence chain index number is registered in the graph standardization event entry, and the corresponding graph standardization event entry's evidence chain index number is registered in the constraint record to ensure a one-to-one correspondence between the evidence chain index and the graph standardization event entry. The correspondence between the evidence chain index, constraint record, and graph standardization event entry is uniformly written into the state view, and the hourly time, entity node, and relation edge are simultaneously registered in the searchable reference to ensure bidirectional positioning, thus completing the verifiable annotation.

[0049] In the state view, perform feasible region clipping on entity nodes and relation edges according to constraint direction and limit magnitude to obtain the fence feasible region. Summarize the fence feasible region according to the hour to obtain the fence constraint set.

[0050] Furthermore, at the same hourly time in the status view and constraint record, feasible region pruning is performed on the entity nodes and relation edges of each operation and maintenance object. When the constraint direction is set downward, the difference between the aligned multi-source observation data observation value and the constraint amplitude is calculated to obtain the lower bound value of the feasible region. The range between the aligned multi-source observation data observation value and the lower bound value of the feasible region is defined as the feasible region. When the constraint direction is set upward, the summation between the aligned multi-source observation data observation value and the constraint amplitude is calculated to obtain the upper bound value of the feasible region. The range between the aligned multi-source observation data observation value and the upper bound value of the feasible region is defined as the guardrail feasible region. The guardrail feasible regions are summarized according to the hourly time to obtain the guardrail constraint set.

[0051] S3. Perform spatiotemporal inference and quantile statistics on the guardrail constraint set based on knowledge graphs to generate the lower bound, median, and upper bound of the quantile sequence. At the same time, define the quantile bandwidth and summarize it into a quantile bandwidth table.

[0052] Generate an ordered set of hourly times based on the hourly time. Based on the set of guardrail constraints, read the feasible region of the guardrail. Group the feasible region of the guardrail according to the constraint direction. Then, merge the feasible regions of the guardrail into a candidate set of feasible regions of the guardrail according to the limited range.

[0053] Furthermore, an ordered set of time points is first generated at each hour, and then the feasible regions of the guardrail and their corresponding evidence chains are read hourly in the knowledge graph segment using the entity nodes and relation edges of the operation and maintenance objects as indexes. Within the same hour, the feasible regions of the guardrail are grouped according to the constraint direction based on the different types of aligned multi-source observation data in each operation and maintenance object. The feasible regions of the guardrail within each group are sorted according to their defined ranges. The feasible regions of the guardrail with intersecting or contiguous defined ranges are merged into a single feasible region's defined range, and the evidence chain indexes are merged. The feasible regions of the guardrail with completely identical defined ranges are deduplicated to form a candidate set of feasible regions of the guardrail at each hour.

[0054] Spatiotemporal inference is performed on the candidate set of feasible regions for guardrails. The operation and maintenance objects are compared with the candidate set of feasible regions for guardrails along the multi-domain relationship. The consistent intervals and conflict markers of the guardrails are registered. The consistent intervals of the guardrails are globally merged to obtain the union set and trim the feasible regions of the guardrails.

[0055] Furthermore, spatiotemporal inference is performed on the candidate set of feasible fence regions. At the same hour, the candidate set of feasible fence regions for the operation and maintenance object is extracted, and the candidate set of feasible fence regions corresponding to the entity nodes or relation edges associated with the operation and maintenance object are compared with the candidate set of feasible fence regions corresponding to the entity nodes or relation edges associated with the operation and maintenance object through the multi-domain relationship. If there is an intersection between the candidate set of feasible fence regions for the operation and maintenance object and the candidate set of feasible fence regions corresponding to the entity nodes or relation edges associated with the operation and maintenance object, the intersection is recorded as the fence consistency interval. If there is no intersection between the candidate set of feasible fence regions for the operation and maintenance object and the candidate set of feasible fence regions corresponding to the entity nodes or relation edges associated with the operation and maintenance object, the evidence chain index is registered as a conflict marker at the hour. At the hour, the fence consistency intervals are globally merged to obtain the union trimmed feasible fence region.

[0056] Read the lower boundary and upper boundary of the feasible region of the union clipping fence, perform quantile statistics to obtain the lower and upper boundaries of the quantile sequence, define the quantile bandwidth and the median value of the quantile sequence, and summarize them into a quantile bandwidth table.

[0057] Furthermore, on the hourly time set, the feasible regions of the guardrails, truncated by union, are extracted for each maintenance object. At each hour, the lower and upper boundaries of the truncated guardrail feasible regions are read. Based on the hourly time, the lower and upper boundaries of the truncated guardrail feasible regions are analyzed separately to obtain the lower boundary sequence and the upper boundary sequence of the truncated guardrail feasible regions. Perform quantile statistics to obtain the lower bound of the quantile sequence; perform quantile statistics on the upper boundary sequence of the feasible region of the union clipped guardrail to obtain the upper bound of the quantile sequence; define the quantile bandwidth based on the range of the lower and upper bounds of the quantile sequence, and take the median value of the quantile bandwidth to obtain the median value of the quantile sequence; summarize the quantile bandwidth, maintenance object, entity node, relation edge, hourly time set, device identifier, observation source, unified unit, conversion factor, conversion direction, and evidence chain index into a quantile bandwidth table.

[0058] S4. Based on the quantile bandwidth table, calculate the backup cost and emergency repair benefit in the knowledge graph, generate backup suggestions and task candidates, and assign values ​​to the state view in conjunction with the guardrail constraint set to output the time-series scheduling scheme.

[0059] The quantile bandwidth of the quantile bandwidth table is divided into uncertainties.

[0060] Furthermore, under the same operation and maintenance object and the same set of whole-point times, the quantile bandwidths of different entity nodes or relation edges of the operation and maintenance object are extracted and arranged in ascending order of quantile bandwidth to obtain a quantile bandwidth sequence. The uncertainty of the quantile bandwidth sequence is divided into three levels: strong, medium and weak. The quantile bandwidth sequence is then divided into three equal parts: the left part of the quantile bandwidth sequence is the quantile bandwidth sequence interval with a weak uncertainty, the middle part of the quantile bandwidth sequence is the quantile bandwidth sequence interval with a medium uncertainty, and the right part of the quantile bandwidth sequence is the quantile bandwidth sequence interval with a strong uncertainty.

[0061] Calculate the interval consistency ratio and the proportion of observation sources to obtain the credibility of the operation and maintenance object. Based on the conflict marker, obtain the corrected credibility by combining the credibility of the operation and maintenance object with the interval consistency ratio and the proportion of observation sources.

[0062] Furthermore, in the knowledge graph shards, the evidence chain indexes corresponding to the operation and maintenance objects, entity nodes, relation edges, and hourly time sets are aggregated. Aligned multi-source observation data observations are read one by one from each evidence chain index. The total number of aligned multi-source observation data observations is counted. For each aligned multi-source observation data observation, it is determined whether it falls within the quantile bandwidth range. The number of aligned multi-source observation data observations falling within the quantile bandwidth range is counted. The ratio of the number of aligned multi-source observation data observations falling within the quantile bandwidth range to the total number of aligned multi-source observation data observations is used as the interval consistency ratio. Each aligned multi-source observation data observation has an observation source. The number of observation source types for each multi-source observation data observation is counted. The ratio of the number of observation source types for each multi-source observation data observation to the total number of aligned multi-source observation data observations is used as the observation source proportion. The mean of the interval consistency ratio and the observation source proportion is calculated to obtain the credibility of the operation and maintenance object.

[0063] The evidence chain index is searched to determine if there are conflict markers registered in the candidate set of feasible fence domains. If a conflict marker exists and it is within the same operational object, entity node, relation edge, or time frame as the operational object's credibility, the operational object's credibility is compared with the interval consistency ratio, and the smaller value between the operational object's credibility and the interval consistency ratio is selected as the corrected credibility. If no conflict marker exists, the operational object's credibility is used as the corrected credibility. If a conflict marker exists and it appears on an entity node or relation edge connected to the operational object's credibility through a multi-domain relation, the operational object's credibility is compared with the proportion of observation sources, and the smaller value between the operational object's credibility and the proportion of observation sources is selected as the corrected credibility.

[0064] Based on the uncertainty level classification of quantile bandwidth, the backup cost and emergency repair benefit are calculated by using quantile bandwidth and corrected confidence.

[0065] Furthermore, under the same maintenance object, the same entity node or relation edge, and the same set of whole-hour times, the quantile bandwidth is read and the backup cost and emergency repair benefit are calculated under a unified unit. When the uncertainty of the quantile bandwidth is weak, the quantile bandwidth and the corrected reliability are multiplied to obtain the backup cost; when the uncertainty of the quantile bandwidth is moderate, the quantile bandwidth and the corrected reliability are multiplied to obtain the backup cost, and the complement of the quantile bandwidth and the corrected reliability is multiplied to obtain the emergency repair benefit; when the uncertainty of the quantile bandwidth is strong, the quantile bandwidth and the corrected reliability are multiplied to obtain the backup cost, and the complement of the quantile bandwidth and the corrected reliability is multiplied to obtain the emergency repair benefit.

[0066] The entity nodes and relation edges are sorted according to the backup cost and the repair benefit, respectively, to obtain backup suggestions and task candidates.

[0067] Furthermore, within the same maintenance object and the same set of whole-hour times, the corresponding entity nodes and relational edges are arranged from largest to smallest according to the backup cost, forming backup suggestions with fields such as quantile bandwidth, correction credibility, backup cost, and evidence chain index; at the same time, the corresponding entity nodes or relational edges are arranged from largest to smallest according to the emergency repair benefit, forming task candidates with fields such as quantile bandwidth, correction credibility, emergency repair benefit, and evidence chain index.

[0068] Assign values ​​to the state view at the hourly time set and mark the assignment status as complete.

[0069] Furthermore, the entity nodes or relation edges in the state view are assigned values ​​at the hourly time set. One alternative suggestion or task candidate is retrieved sequentially according to a predetermined order as a candidate. The lower bound, upper bound, quantile sequence, quantile bandwidth, corrected confidence level, and evidence chain index of the candidate are read. The feasible region of the guardrail corresponding to the entity node or relation edge at the hourly time set is retrieved from the guardrail constraint set. It is determined whether the quantile bandwidth range of the candidate intersects with the feasible region of the guardrail. If the quantile bandwidth range does not intersect with the feasible region of the guardrail, the incompatibility reason is recorded according to the evidence chain index and the item is skipped. In the intersection, the candidate options are first marked as comparison items, and the quantile bandwidth ranges of other candidate options at the same time are checked to see if they also intersect with the feasible region of the guardrail. All candidate options that meet the intersection condition are included in the comparison items. From the comparison items, the candidate option with the larger modified confidence value is selected, and the entity node or relation edge is assigned a value in the hourly time set. The intersection of the quantile bandwidth range and the feasible region of the guardrail is used as the assignment range, and the lower bound of the quantile sequence, the upper bound of the quantile sequence, the quantile bandwidth, the modified confidence value and the evidence chain index are recorded. At the same time, the assignment status of the entity node or relation edge in the hourly time set is marked as completed.

[0070] All operation and maintenance objects and all hourly time sets are assigned values ​​and integrated to generate a time-series scheduling scheme.

[0071] Furthermore, all operation and maintenance objects and all hourly time sets are assigned values. The assignment range of each entity node or relation edge in each hourly time set is summarized in time order with the corresponding lower bound of the quantile sequence, upper bound of the quantile sequence, quantile bandwidth, correction credibility and evidence chain index. The results are then uniformly integrated according to the operation and maintenance objects, entity nodes or relation edges and hourly time sets to generate a time-series scheduling scheme.

[0072] S5. Add evidence and time constraints to the timing scheduling scheme and summarize them into an execution delivery list.

[0073] According to the time-series scheduling scheme, the effective range of the feasible region of the guardrail is retrieved from the guardrail constraint set, and the intersection of this range with the hourly time set is used as the time limit constraint.

[0074] Furthermore, within the hourly time set, the operation and maintenance objects and their corresponding entity nodes or relational edges in the time-series scheduling scheme are read in chronological order. The assignment range, lower bound of the quantile sequence, upper bound of the quantile sequence, quantile bandwidth, median value of the quantile sequence, correction credibility, evidence chain index, backup suggestions and task candidates are retained one by one. From the guardrail constraint set, the effective range of the feasible domain of the guardrail that is consistent with the operation and maintenance object, entity node or relational edge and the hourly time set is retrieved, and the intersection of the effective range and the hourly time set is used as the time limit constraint.

[0075] The effective range of the fence's feasible domain is retrieved, the inclusion and intersection relationships are compared, and fence consistency and intersection statements are generated. The fence consistency and intersection statements are then summarized into evidence statements.

[0076] Furthermore, within the hourly time set, each group of maintenance objects and their corresponding entity nodes or relational edges in the time-series scheduling scheme are read sequentially. The assignment range, lower bound of the quantile sequence, upper bound of the quantile sequence, quantile bandwidth, quantile sequence median value, correction confidence, and evidence chain index are collected, and the evidence chain index corresponding to the backup suggestion and the task candidate is associated. The feasible region of the guardrail is retrieved, and the effective range and evidence chain index of the feasible region are read. The assignment range and the effective range are compared for inclusion and intersection relationships. When the assignment range is completely contained by the effective range, it is registered as a guardrail consistency statement. Simultaneously, the evidence chain index of the feasible region of the guardrail and the evidence chain index corresponding to the assignment range are referenced; when the assignment range and the effective range partially overlap, it is registered as an intersection description and the evidence chain index of the feasible region of the guardrail and the evidence chain index corresponding to the assignment range are referenced; the operation and maintenance object, entity node or relation edge, set of whole hour time, assignment range, lower bound of quantile sequence, upper bound of quantile sequence, quantile bandwidth, median value of quantile sequence, correction credibility, evidence chain index, backup suggestion, task candidate, effective range of the feasible region of the guardrail, guardrail consistency description and intersection description are summarized into evidence description according to the set of whole hour time.

[0077] Using the operation and maintenance object, entity node or relationship edge and the set of hourly times as indexes, the time limit constraints and evidence descriptions are matched one by one and merged into an execution delivery list.

[0078] This embodiment also provides a knowledge graph-based distributed new energy operation and maintenance decision support system, including: The data acquisition and mapping module collects multi-source observation data and forms a state view using a knowledge graph as a carrier. The state view is then standardized by anomaly identification to obtain a graph event view set. The event constraint pruning module instantiates constraints on the graph event view set and completes verifiable annotation by referencing the evidence chain index one by one. At the same time, it prunes the feasible domain of the state view to obtain the guardrail constraint set. The spatiotemporal quantile statistics module performs spatiotemporal inference and quantile statistics on the guardrail constraint set based on knowledge graphs, generates the lower bound, median value and upper bound of the quantile sequence, and defines the quantile bandwidth, which is then summarized into a quantile bandwidth table. The backup repair assignment module, based on the quantile bandwidth table, calculates the backup cost and repair benefit in the knowledge graph, generates backup suggestions and task candidates, and assigns values ​​to the state view in conjunction with the guardrail constraint set, and outputs a time-series scheduling scheme. The delivery description and time limit module adds evidence and time limit constraints to the time-series scheduling scheme, and summarizes them into an execution delivery list.

[0079] In summary, this invention improves the transparency and reliability of operation and maintenance decisions and enhances the auditability and credibility of the anomaly handling process by generating a guardrail constraint set; and dynamically reflects the uncertainty range of the system state by generating a quantile bandwidth table, providing robust data support for subsequent cost and benefit calculations and improving the adaptability of the decision-making scheme to complex operation and maintenance environments.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distributed new energy operation and maintenance decision support method based on knowledge graphs, characterized in that: include, Collect multi-source observation data and form a state view using a knowledge graph as a carrier. Standardize the state view by anomaly identification to obtain a graph event view set. Instantiate constraints on the graph event view set, and complete verifiable annotations by referencing the evidence chain index one by one. At the same time, perform feasible domain pruning on the state view to obtain the guardrail constraint set. The guardrail constraint set is subjected to spatiotemporal inference and quantile statistics based on knowledge graphs to generate the lower bound, median and upper bound of the quantile sequence. At the same time, the quantile bandwidth is defined and summarized into a quantile bandwidth table. Based on the quantile bandwidth table, the backup cost and emergency repair benefit are calculated in the knowledge graph, backup suggestions and task candidates are generated, and the state view is assigned values ​​in conjunction with the guardrail constraint set to output the time-series scheduling scheme. Add evidence and time constraints to the timing scheduling scheme, and summarize them into an execution delivery list.

2. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 1, characterized in that: The steps for forming the state view are as follows: Collect multi-source observation data and record the observation sources, establish a unified timeline, and extract the hourly time; Multi-source observation data are mapped to the nearest integer point and piecewise linear interpolation is performed on missing integer points. The units of multi-source observation data are unified, and the conversion coefficients and conversion directions are recorded to obtain aligned multi-source observation data. The system will add device identifiers and observation sources to the aligned multi-source observation data, associate the aligned multi-source observation data with the operation and maintenance objects and multi-domain relationships, and generate attribute snapshots based on the knowledge graph, with the operation and maintenance objects as entity nodes and multi-domain relationships as relationship edges, forming a state view with the entity nodes and relationship edges.

3. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 2, characterized in that: The process of obtaining the spectral event view set involves extracting and comparing the aligned multi-source observation data from the state view, marking abnormal fluctuation records and data mutation records, merging them into spectral standardized events, and summarizing them into the spectral event view set.

4. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 3, characterized in that: The steps for instantiating and constraining the graph event view set, and for completing verifiable annotations by referencing the evidence chain index one by one, are as follows: Based on the graph event view set, locate the entity nodes and relationship edges of the operation and maintenance objects in the status view, read the hourly time and attribute snapshots, and determine the scope of effect; register the constraint direction and limit range on the entity nodes and relationship edges and reference the evidence link; merge the graph standardized event entries according to the hourly time and the operation and maintenance objects to obtain the constraint record; Based on the standardized events and evidence links in the constraint record graph, an evidence chain index is generated, written to the attribute snapshot, and a searchable reference is established to complete the verifiable annotation.

5. The distributed new energy operation and maintenance decision support method based on knowledge graph as described in claim 4, characterized in that: The process of obtaining the guardrail constraint set involves performing feasible domain clipping on entity nodes and relation edges in the state view according to the constraint direction and limit magnitude to obtain the guardrail feasible domain, and then summarizing the guardrail feasible domain according to the hour to obtain the guardrail constraint set.

6. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 5, characterized in that: The process of summarizing into a quantile bandwidth table is as follows: Generate an ordered set of hourly times based on the hourly time. Based on the set of guardrail constraints, read the feasible region of the guardrail. Group the feasible region of the guardrail according to the constraint direction. Then, merge the feasible region of the guardrail into a candidate set of feasible regions of the guardrail according to the limited range. Spatiotemporal inference is performed on the candidate set of feasible regions for guardrails. The operation and maintenance objects are compared with the candidate set of feasible regions for guardrails along the multi-domain relationship. The consistent intervals and conflict markers of the guardrails are registered. The consistent intervals of the guardrails are globally merged to obtain the union set and trim the feasible regions of the guardrails. Read the lower boundary and upper boundary of the feasible region of the union clipping fence, perform quantile statistics to obtain the lower and upper boundaries of the quantile sequence, define the quantile bandwidth and the median value of the quantile sequence, and summarize them into a quantile bandwidth table.

7. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 6, characterized in that: The steps for generating alternative suggestions and task candidates are as follows: The quantile bandwidth of the quantile bandwidth table is divided into uncertainties; Calculate the interval consistency ratio and the proportion of observation sources to obtain the credibility of the operation and maintenance object. Based on the conflict marker, obtain the corrected credibility by combining the credibility of the operation and maintenance object with the interval consistency ratio and the proportion of observation sources. Based on the uncertainty level classification of quantile bandwidth, the backup cost and emergency repair benefit are calculated by using quantile bandwidth and correction confidence. The entity nodes and relation edges are sorted according to the backup cost and the repair benefit, respectively, to obtain backup suggestions and task candidates.

8. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 7, characterized in that: The output timing scheduling scheme comprises the following steps: Assign values ​​to the status view at the set of times on the hour, and mark the assignment status as completed; All operation and maintenance objects and all hourly time sets are assigned values ​​and integrated to generate a time-series scheduling scheme.

9. The knowledge graph-based distributed new energy operation and maintenance decision support method as described in claim 8, characterized in that: The summary is compiled into an execution delivery list, and the steps are as follows. According to the time-series scheduling scheme, the effective range of the feasible region of the guardrail is retrieved from the guardrail constraint set, and the intersection of this range with the set of hourly times is used as the time limit constraint. The effective range of the feasible domain of the guardrail is retrieved, the inclusion relationship and intersection relationship are compared, the guardrail consistency statement and intersection statement are generated, and the guardrail consistency statement and intersection statement are summarized into evidence statement; Match the time constraints with the evidence and combine them into an execution delivery list.

10. A knowledge graph-based distributed new energy operation and maintenance decision support system, based on the knowledge graph-based distributed new energy operation and maintenance decision support method according to any one of claims 1 to 9, characterized in that: include, The data acquisition and mapping module collects multi-source observation data and forms a state view using a knowledge graph as a carrier. The state view is then standardized by anomaly identification to obtain a graph event view set. The event constraint pruning module instantiates constraints on the graph event view set and completes verifiable annotation by referencing the evidence chain index one by one. At the same time, it prunes the feasible domain of the state view to obtain the guardrail constraint set. The spatiotemporal quantile statistics module performs spatiotemporal inference and quantile statistics on the guardrail constraint set based on knowledge graphs, generates the lower bound, median value and upper bound of the quantile sequence, and defines the quantile bandwidth, which is then summarized into a quantile bandwidth table. The backup repair assignment module, based on the quantile bandwidth table, calculates the backup cost and repair benefit in the knowledge graph, generates backup suggestions and task candidates, and assigns values ​​to the state view in conjunction with the guardrail constraint set, and outputs a time-series scheduling scheme. The delivery description and time limit module adds evidence and time limit constraints to the time-series scheduling scheme, and summarizes them into an execution delivery list.

Images