A new energy equipment defect cost quota calibration method and device

Abstract

The application relates to the technical field of new energy intelligent operation and maintenance, and discloses a new energy equipment defect cost quota calibration method and device, which comprises the following steps: constructing a knowledge graph for new energy equipment defect cost quota calibration; determining cost element reference parameters corresponding to a target defect according to the knowledge graph, and generating a basic cost quota of the target defect according to the cost element reference parameters; determining a scene correction coefficient corresponding to the target defect according to the knowledge graph, and correcting the basic cost quota by using the scene correction coefficient to obtain a scene-based cost quota; acquiring an actual processing cost of the target defect, and calibrating the knowledge graph according to the deviation between the actual processing cost and the scene-based cost quota to update the scene-based cost quota. The application solves the problem that the defect cost quota cannot be accurately dynamically calibrated due to the lack of knowledge graph association and actual cost deviation feedback of new energy equipment.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for new energy, specifically to a method and device for calibrating the cost quota for defects in new energy equipment. Background Technology

[0002] In existing technologies, the determination of cost quotas for handling defects in new energy equipment typically relies on historical operation and maintenance data and human experience. The correlation between defects, cost elements, and scenarios is not effectively integrated, and there is a lack of knowledge graphs to support the calibration of defect cost quotas. Therefore, when determining the benchmark parameters of cost elements corresponding to a target defect and generating basic cost quotas, it is difficult to reason based on unified correlation knowledge, resulting in a lack of reliable basis for determining basic cost quotas.

[0003] Furthermore, existing technologies lack a mechanism for determining scenario correction coefficients based on knowledge graphs, making it difficult to make scenario-based corrections to the basic cost quotas. Consequently, it is impossible to obtain scenario-based cost quotas that match the actual scenarios of the target defects. At the same time, existing technologies do not feed back the deviation between the actual processing cost and the scenario-based cost quotas to the knowledge graph, making it impossible to update the scenario-based cost quotas by calibrating the knowledge graph. This makes it difficult for the cost quotas to be dynamically adjusted with changes in actual processing costs. Summary of the Invention

[0004] This invention provides a method and apparatus for calibrating the defect cost quota of energy equipment, in order to solve the problem that the defect cost quota cannot be accurately calibrated due to the lack of knowledge graph association and actual cost deviation feedback.

[0005] In a first aspect, the present invention provides a method for calibrating the defect cost quota of new energy equipment, the method comprising: Construct a knowledge graph for calibration of defect cost quotas for new energy equipment; Based on the knowledge graph, determine the cost element benchmark parameters corresponding to the target defect, and generate the basic cost quota for the target defect based on the cost element benchmark parameters. Based on the knowledge graph, the scenario correction coefficient corresponding to the target defect is determined, and the basic cost quota is corrected using the scenario correction coefficient to obtain the scenario-based cost quota. Obtain the actual processing cost of the target defect, and calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota to update the scenario-based cost quota.

[0006] This invention provides a method for calibrating the cost quota of defects in new energy equipment. By constructing a knowledge graph for calibrating the cost quota of defects in new energy equipment, the benchmark parameters of cost elements and scenario correction coefficients corresponding to the target defect can be determined based on the knowledge graph, thus providing a unified basis for the generation of basic cost quotas and scenario-based cost quotas. Furthermore, the knowledge graph is calibrated based on the deviation between the actual processing cost and the scenario-based cost quota, and the scenario-based cost quota is updated accordingly, enabling the cost quota to be dynamically adjusted as the actual processing cost changes. This solves the problem in existing technologies where the defect cost quota cannot be accurately and dynamically calibrated due to the lack of knowledge graph association and feedback on actual cost deviations. In an optional embodiment, constructing the knowledge graph for calibrating the cost quota of defects in new energy equipment includes: Acquire multi-source heterogeneous data related to defect handling in new energy equipment; Multi-source heterogeneous data is preprocessed to obtain preprocessed data; The preprocessed data is mapped to a heterogeneous graph, and a heterogeneous graph attention network is used to align entities in the heterogeneous graph. Based on the entities after entity alignment, complete the missing attributes of the entities and extract the relationships between the entities to form triplet data; The triplet data is stored in a graph database to obtain a knowledge graph. The knowledge graph includes defect entities, cost element entities, scene entities, the relationship between defect entities and scene entities, the relationship between standard operating procedures and cost element entities, and the correspondence between scene entities and correction coefficients.

[0007] In the above technical solution, multi-source heterogeneous data related to defect handling of new energy equipment is acquired and preprocessed, mapped into a heterogeneous graph, and entity alignment is performed using a heterogeneous graph attention network. Then, missing attributes are filled in based on the entities after entity alignment, and the relationships between entities are extracted to form triples. Finally, the data is stored in a graph database to obtain a knowledge graph containing defect entities, cost element entities, scene entities, and corresponding relationships. This achieves the structured integration and association of multi-source heterogeneous data, providing a unified and queryable data foundation for subsequent determination of cost element benchmark parameters, scene correction coefficients, and calibration based on deviations.

[0008] In one optional implementation, a heterogeneous graph attention network is used to align entities in the heterogeneous graph, including: A heterogeneous graph attention network is used to train the heterogeneous graph. The features of different types of nodes and edges are learned through a multi-head attention mechanism to obtain entity fusion features. Based on entity fusion features, the similarity between entities from different data sources is calculated, and entities whose similarity meets the threshold are merged to obtain aligned entities.

[0009] In the above technical solution, a heterogeneous graph attention network is used to train the heterogeneous graph, and a multi-head attention mechanism is used to learn the features of different types of nodes and edges to obtain entity fusion features. Then, the similarity between entities in different data sources is calculated based on the entity fusion features, and entities with similarity that meet the threshold are merged. This enables entity alignment in multi-source data and provides an aligned entity foundation for the subsequent construction of knowledge graphs.

[0010] In one optional implementation, the cost element benchmark parameters corresponding to the target defect are determined based on the knowledge graph, including: Based on the attributes of the defective entity, a defect code for the target defect is generated; the defect code is a four-dimensional attribute code of component-level-type-operating condition. Based on the defect code, match the component entity and fault type corresponding to the target defect in the knowledge graph, and associate the standard operating procedure corresponding to the target defect. At least one of the following is extracted from the standard operating procedure: labor type, working hours, spare parts model and specifications, and resource shift quantity, as the benchmark parameter for cost elements.

[0011] In the above technical solution, a four-dimensional defect code of component-level-type-operating condition is generated based on the attributes of the defective entity. The component entity and fault type corresponding to the target defect are matched in the knowledge graph according to the defect code, and the corresponding standard operating procedure is associated. Then, at least one of the following is extracted from the standard operating procedure: labor type, working hours, spare parts model and specifications, and resource shift quantity as the cost element benchmark parameter. Thus, the determination process of the cost element benchmark parameter is based on the standardized defect code and the association relationship in the knowledge graph, realizing the accurate correspondence between the cost element benchmark parameter and the target defect and its standard operating procedure.

[0012] In one optional implementation, generating a basic cost quota for the target defect based on cost element benchmark parameters includes: Based on the cost element benchmark parameters, determine the labor cost, spare parts cost, resource cost and implicit cost corresponding to the target defect; The basic cost quota is obtained by adding up labor costs, spare parts costs, resource costs, and hidden costs.

[0013] In the above technical solution, the labor cost, spare parts cost, resource cost and implicit cost corresponding to the target defect are determined according to the cost element benchmark parameters, and the above four costs are added together to obtain the basic cost quota. Thus, the generation of the basic cost quota covers the cost composition of the four aspects of labor, spare parts, resources and implicit costs, and provides a clear cost decomposition and calculation basis for the basic cost quota.

[0014] In one optional implementation, generating the basic cost quota for the target defect based on cost element benchmark parameters further includes: The processing of the target defect is simulated using a digital twin model to obtain the simulation cost; The deviation rate is calculated based on simulation costs and basic cost quotas; When the deviation rate is less than or equal to the preset threshold, the basic cost quota is determined to have passed the verification and the basic cost quota is output. When the deviation rate is greater than the preset threshold, determine the cause of the deviation, adjust the model parameters and / or standard operating procedures used to generate the basic cost quota, regenerate the basic cost quota and perform simulation verification until the deviation rate is less than or equal to the preset threshold.

[0015] In the above technical solution, the theoretical cost is obtained by simulating the processing of the target defect using a digital twin model. The deviation rate is calculated based on the theoretical cost and the basic cost quota. When the deviation rate is less than or equal to a preset threshold, the basic cost quota is output. When the deviation rate is greater than the preset threshold, the cause of the deviation is determined and the model parameters and / or standard operating procedures used to generate the basic cost quota are adjusted. The basic cost quota is regenerated and simulated for verification until the deviation rate meets the preset threshold. This achieves simulation verification and closed-loop correction of the basic cost quota, making the basic cost quota consistent with the theoretical cost.

[0016] In one optional implementation, the scene correction coefficient corresponding to the target defect is determined based on the knowledge graph, including: Based on the relationship between defect entities and scene entities, determine the scene corresponding to the target defect; Based on the correspondence between scene entities and correction coefficients, the correction coefficients corresponding to the scene are determined and used as the scene correction coefficients corresponding to the target defect.

[0017] In the above technical solution, the scene corresponding to the target defect is determined by the association between the defect entity and the scene entity, and the correction coefficient corresponding to the scene is determined by the correspondence between the scene entity and the correction coefficient as the scene correction coefficient corresponding to the target defect. Thus, the determination of the scene correction coefficient is based on the entity association and correspondence in the knowledge graph, and the accurate association between the target defect and its scene correction coefficient is realized.

[0018] In one optional implementation, the scene correction coefficients include a capacity correction coefficient, a terrain correction coefficient, a region correction coefficient, and a working condition correction coefficient. The basic cost quota is adjusted using a scenario-based adjustment factor to obtain scenario-based cost quotas, including: The basic cost quota is multiplied sequentially by the capacity correction factor, terrain correction factor, regional correction factor, and operating condition correction factor to obtain the scenario-based cost quota.

[0019] In the above technical solution, by multiplying the basic cost quota by the capacity correction coefficient, terrain correction coefficient, region correction coefficient and working condition correction coefficient in sequence, the basic cost quota can be corrected item by item according to the four dimensions of scenario factors: capacity, terrain, region and working condition. This results in a scenario-based cost quota corresponding to the specific scenario in which the target defect is located, and the cost quota is comprehensively adapted to multiple scenario factors.

[0020] In one optional implementation, the knowledge graph is calibrated based on the deviation between the actual processing cost and the contextualized cost quota to update the contextualized cost quota, including: The deviation rate is calculated based on the actual processing cost and the scenario-based cost quota. When the deviation rate exceeds the preset deviation threshold, the cause of the deviation is determined. Update at least one of the following in the knowledge graph based on the cause of the deviation: cost element entity, correction coefficient, association between standard operating procedures and cost element entity, and entity alignment result; The scenario-based cost quota for the target defect is recalculated based on the updated knowledge graph, resulting in the updated scenario-based cost quota.

[0021] In the above technical solution, the deviation rate is calculated based on the actual processing cost and the scenario-based cost quota. When the deviation rate is greater than the preset deviation threshold, the cause of the deviation is determined. Based on this, at least one of the following is updated in the knowledge graph: cost element entity, correction coefficient, association relationship between standard operating procedure and cost element entity, and entity alignment result. Then, the scenario-based cost quota of the target defect is recalculated based on the updated knowledge graph. This achieves the calibration of the knowledge graph based on the actual processing cost and the corresponding update of the scenario-based cost quota.

[0022] Secondly, the present invention provides a new energy equipment defect cost quota calibration device, the device comprising: The knowledge graph construction module is used to build a knowledge graph for calibration of defect cost quotas for new energy equipment. The basic cost quota generation module is used to determine the cost element benchmark parameters corresponding to the target defect based on the knowledge graph, and generate the basic cost quota of the target defect based on the cost element benchmark parameters. The scenario-based cost quota generation module is used to determine the scenario correction coefficient corresponding to the target defect based on the knowledge graph, and to use the scenario correction coefficient to correct the basic cost quota to obtain the scenario-based cost quota. The calibration module is used to obtain the actual processing cost of the target defect and to calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota in order to update the scenario-based cost quota.

[0023] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the above-described first aspect or any corresponding embodiment of the new energy equipment defect cost quota calibration method.

[0024] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute a method for calibrating the defect cost quota of new energy equipment as described in the first aspect or any corresponding embodiment.

[0025] Fifthly, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the above-described first aspect or any corresponding embodiment of the new energy equipment defect cost quota calibration method. Attached Figure Description

[0026] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first process of a method for calibrating the defect cost quota of new energy equipment according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the second process of a method for calibrating the defect cost quota of new energy equipment according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the third process of a new energy equipment defect cost quota calibration method according to an embodiment of the present invention; Figure 5 This is a structural block diagram of a new energy equipment defect cost quota calibration device according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0029] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0030] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0031] As an optional application scenario of this invention, such as Figure 1 As shown, the hardware system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0032] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0033] This invention provides a method for calibrating defect cost quotas for new energy equipment, applied to the cost management of defect handling for new energy equipment such as wind power and photovoltaics. By constructing a knowledge graph to associate defects, cost elements, and scenarios, basic cost quotas and scenario-based cost quotas are generated. The knowledge graph is then calibrated based on the deviation between the actual handling cost and the scenario-based cost quotas to update the scenario-based cost quotas, achieving dynamic calibration of defect handling cost quotas. According to this invention, an embodiment of the method for calibrating defect cost quotas for new energy equipment is provided. It should be noted that the steps shown in the flowchart can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] This embodiment provides a method for calibrating the defect cost quota of new energy equipment, which can be used in the aforementioned electronic equipment or terminal equipment. Figure 2 This is a flowchart of a method for calibrating the defect cost quota of new energy equipment according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Construct a knowledge graph for calibration of defect cost quotas for new energy equipment.

[0035] Among them, the calibration of the defect cost quota for new energy equipment refers to the process of pre-determining a standard quota (i.e., quota) for the cost that should be spent on handling defects in new energy equipment such as wind power and photovoltaics, and dynamically adjusting this standard quota according to the actual situation.

[0036] The calibration of defect cost quotas for new energy equipment is achieved using a knowledge graph. This knowledge graph includes defect entities, cost element entities, and scenario entities, as well as the relationships between defect entities and scenario entities, the relationships between standard operating procedures and cost element entities, and the correspondence between scenario entities and correction coefficients. The knowledge graph serves both as the basis for determining the correlation between cost element benchmark parameters and scenario correction coefficients, and as the object for receiving and updating feedback on actual cost deviations, thereby enabling the defect cost quotas for new energy equipment to be dynamically calibrated as actual processing costs change.

[0037] Specifically, when constructing a knowledge graph for calibrating the cost quota for defects in new energy equipment, multi-source heterogeneous data related to defect handling of new energy equipment is started. After preprocessing, the data is mapped into a heterogeneous graph. Entities in the heterogeneous graph are aligned using a heterogeneous graph attention network. Then, based on the aligned entities, missing attributes are filled in and the relationships between entities are extracted to form triplet data, which is stored in the graph database. This results in a knowledge graph containing defect entities, cost element entities, scene entities, and the relationships between defect entities and scene entities, the relationships between standard operating procedures and cost element entities, and the correspondence between scene entities and correction coefficients. This provides a basis for determining the benchmark parameters of cost elements and scene correction coefficients in the future.

[0038] Step S202: Based on the knowledge graph, determine the cost element benchmark parameters corresponding to the target defect, and generate the basic cost quota of the target defect based on the cost element benchmark parameters.

[0039] Here, the target defect refers to the specific defect object that needs to be calibrated for cost quotas, which is the defect that this method targets when it runs. For example, a specific fault occurring in a unit of a new energy power station, such as a general mechanical fault in the gearbox of a 6MW coastal wind farm (high salt spray condition), corresponds to a defect entity in the knowledge graph.

[0040] Cost element benchmark parameters refer to the baseline values ​​of various cost components when a target defect is handled under standard operating procedures. In this embodiment, these parameters are determined based on a knowledge graph: according to the defect code of the target defect, the corresponding component entity and fault type are matched in the knowledge graph, the corresponding standard operating procedure is associated, and then the types of workers, working hours, spare parts models and specifications, and resource shifts are extracted from the standard operating procedure as cost element benchmark parameters. It is equivalent to the input basis for calculating basic cost quotas and is used to further determine labor costs, spare parts costs, resource costs, and implicit costs.

[0041] The basic cost quota refers to the sum of labor costs, spare parts costs, resource costs, and implicit costs determined based on the baseline parameters of cost elements for the target defect, before considering specific scenario adjustments. This sum serves as the standard cost benchmark for handling the defect. In this embodiment, it is generated based on the relationship between defects, cost elements, and standard operating procedures in a knowledge graph. It reflects the cost level under the standard operating procedures and baseline parameters, and does not change with specific scenarios such as capacity, terrain, region, or operating conditions. The basic cost quota is then adjusted using scenario adjustment coefficients to obtain a scenario-based cost quota. Therefore, the basic cost quota is not a data filtering process, but rather a benchmark for subsequent scenario-based cost quota calculations.

[0042] Specifically, for a target defect, the baseline values ​​of cost elements such as labor, spare parts, resources and hidden costs under the standard operating procedures can be obtained based on the relationship between the defect and the standard operating procedures and cost elements in the knowledge graph. Based on this, the basic cost quota of the defect without considering specific scenario corrections can be calculated and summarized, which serves as the benchmark for subsequent scenario-based corrections.

[0043] Step S203: Based on the knowledge graph, determine the scenario correction coefficient corresponding to the target defect, and use the scenario correction coefficient to correct the basic cost quota to obtain the scenario-based cost quota.

[0044] This refers to the coefficient used to adjust the basic cost quota based on the specific scenario in which the target defect is located. This adjustment coefficient can include dimensions such as capacity, terrain, region, and operating conditions. It is used to adjust the basic cost quota item by item that does not consider scenario differences, so that it can be adapted to the actual scenario in which the target defect is located, thereby obtaining a scenario-based cost quota.

[0045] Scenario-based cost quotas refer to cost quotas obtained by adjusting the basic cost quotas using a scenario correction coefficient corresponding to the scenario in which the target defect occurs. The basic cost quotas reflect the baseline cost under standard operating procedures and do not consider scenario differences such as capacity, terrain, region, and working conditions. After determining the scenario correction coefficient corresponding to the target defect through a knowledge graph, the basic cost quotas are adjusted to obtain scenario-based cost quotas that match the actual scenario of the target defect. These quotas serve as the basis for subsequent comparison with actual processing costs and dynamic calibration.

[0046] Specifically, the target defect is associated with a corresponding scenario in the knowledge graph, and each scenario has a corresponding correction coefficient. Based on this, the scenario correction coefficient corresponding to the target defect can be determined, and the basic cost quota can be corrected using this coefficient to obtain a scenario-based cost quota that matches the actual scenario of the defect.

[0047] Step S204: Obtain the actual processing cost of the target defect, and calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota to update the scenario-based cost quota.

[0048] The actual processing cost refers to the cost incurred during the actual handling of the target defect, which typically includes the actual labor costs, spare parts costs, resource costs, and implicit costs arising from the handling of the defect. This cost can be obtained from actual operation and maintenance records, used to compare with scenario-based cost quotas, calculate the deviation, and calibrate the knowledge graph accordingly.

[0049] Specifically, after the target defect is processed, its actual processing cost is obtained and compared with the scenario-based cost quota. The knowledge graph is calibrated based on the deviation between the two, and the scenario-based cost quota is updated based on the calibrated knowledge graph, so that the cost quota can be dynamically adjusted as the actual processing cost changes.

[0050] This embodiment provides a method for calibrating the cost quota of defects in new energy equipment, aiming to solve the problem that the cost quota of defects in new energy equipment cannot be accurately and dynamically calibrated due to the lack of knowledge graph association and actual cost deviation feedback in the prior art. The method includes: constructing a knowledge graph for calibrating the cost quota of defects in new energy equipment; determining the cost element benchmark parameters corresponding to the target defect based on the knowledge graph, and generating a basic cost quota for the target defect based on the cost element benchmark parameters; determining the scenario correction coefficient corresponding to the target defect based on the knowledge graph, and correcting the basic cost quota using the scenario correction coefficient to obtain a scenario-based cost quota; obtaining the actual processing cost of the target defect, and calibrating the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota to update the scenario-based cost quota. The above scheme determines the cost element benchmark parameters and scenario correction coefficients based on the knowledge graph, providing a unified basis for the generation of basic cost quotas and scenario-based cost quotas. Furthermore, the knowledge graph is calibrated and the scenario-based cost quotas are updated by the deviation between the actual processing cost and the scenario-based cost quotas, so that the cost quotas can be dynamically adjusted with the changes in actual processing cost, thereby achieving the effect of accurate dynamic calibration of defect cost quotas.

[0051] This embodiment provides a method for calibrating the defect cost quota of new energy equipment, which can be used in the aforementioned electronic equipment or terminal equipment. Figure 3 This is a flowchart of a method for calibrating the defect cost quota of new energy equipment according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Construct a knowledge graph for calibration of defect cost quotas for new energy equipment.

[0052] Specifically, step S301 includes: Step S3011: Obtain multi-source heterogeneous data related to defect handling of new energy equipment.

[0053] Specifically, historical maintenance work orders (text format) for new energy equipment, spare parts procurement contracts (structured tables), resource shift records (semi-structured data), meteorological monitoring data (time series data), and industry standard documents (text format) such as NB / T 10574-2021 are collected as multi-source heterogeneous data.

[0054] Step S3012: Preprocess the multi-source heterogeneous data to obtain preprocessed data.

[0055] The data is cleaned (redundant and erroneous data is removed), format is standardized (data of different formats is uniformly converted into a processable format), and attributes are extracted (entity attributes are extracted from text data, such as defective parts and fault types from work orders), to obtain preprocessed multi-source heterogeneous data.

[0056] Step S3013: Map the preprocessed data to a heterogeneous graph, and use a heterogeneous graph attention network to align entities in the heterogeneous graph.

[0057] In an optional implementation, step S3013 includes: Step a1: The heterogeneous graph attention network is used to train the heterogeneous graph. The features of different types of nodes and edges are learned through the multi-head attention mechanism to obtain entity fusion features.

[0058] Constructing a heterogeneous graph model: The preprocessed multi-source heterogeneous data is mapped to a heterogeneous graph. Nodes in the heterogeneous graph are various entities (such as defect entities, component entities, etc.), and edges are potential relationships between entities (such as the "use" relationship between defects in work orders and spare parts in spare parts procurement contracts). Node features include the attribute features of entities (such as the model and material of component entities), text features (such as the text vector describing defects in work orders), and structural features (such as the frequency of occurrence of entities in the dataset).

[0059] The heterogeneous graph is input into the HGAT (Heterogeneous Graph Attention Network) model. This model uses a multi-head attention mechanism to weightedly learn the features of different types of entities and the relationships between them, adaptively capturing both global semantic features and local structural features to obtain entity fusion features. Specifically, the heterogeneous graph attention network model designs a dedicated attention coefficient calculation method for different types of nodes and edges in the heterogeneous graph, using the following formula: (1); in, For the type of node or edge, For type The corresponding attention scoring function, For type The feature transformation matrix, , They are nodes , eigenvectors, For nodes For nodes Attention score.

[0060] The attention score is normalized using the softmax function to obtain the node. For nodes Normalized attention coefficient: (2); in, For nodes Similar types Set of neighboring nodes; In type Next, node For nodes The normalized attention coefficients represent the node's attention coefficients. When aggregating neighbor features, assign them to nodes. The weight, For normalization function, It is an exponential function. For type Next, node For neighboring nodes The original attention score.

[0061] Finally, the entity fusion feature vector of the node is obtained by weighted summation: (3); in, This is an activation function (such as the ReLU function).

[0062] Step a2: Based on the entity fusion features, calculate the similarity between entities in different data sources, merge entities whose similarity meets the threshold, and obtain aligned entities.

[0063] Based on the entity fusion feature vector output by the heterogeneous graph attention network model, the similarity of entities in different data sources is calculated (using cosine similarity or Euclidean distance). A similarity threshold is set (e.g., 0.85). When the similarity of two entities is higher than the similarity threshold, they are determined to be the same entity and merged to achieve entity alignment.

[0064] Step S3014: Based on the entities after entity alignment, complete the missing entity attributes and extract the relationships between entities to form triplet data.

[0065] By utilizing the aligned entity relationships, rule-based reasoning and machine learning methods (such as a relation extraction model based on BERT (Bidirectional Encoder Representations from Transformers)) are used to complete the missing entity attributes, extract the five-element relations between entities, and form triples (head entity-relation-tail entity).

[0066] Step S3015: Store the triplet data in the graph database to obtain the knowledge graph; the knowledge graph includes defect entities, cost element entities, scene entities, the relationship between defect entities and scene entities, the relationship between standard operating procedures and cost element entities, and the correspondence between scene entities and correction coefficients.

[0067] The triplet data is stored in a graph database (such as Neo4j), and the GraphScope tool is used to realize the visualization query and management of the graph, supporting knowledge extraction, fusion and reasoning functions.

[0068] The core entities of the knowledge graph include defect entities, component entities, cost element entities, scenario entities, and meteorological entities. Among them, defect entities include four-dimensional attributes: component, level, type, and operating condition; cost element entities include four types of sub-entities: labor, spare parts, resources, and implicit costs; scenario entities include four types of sub-entities: capacity, terrain, region, and operating condition; and meteorological entities include attributes such as wind speed, power coefficient, and temperature.

[0069] The knowledge graph relational model establishes a five-element relational model: defect-binding-SOP (Standard Operating Procedure), SOP-consumption-cost element, defect-adaptation-scenario, scenario-correspondence-correction coefficient, and weather-impact-implicit cost.

[0070] Step S302: Based on the knowledge graph, determine the cost element benchmark parameters corresponding to the target defect, and generate the basic cost quota of the target defect based on the cost element benchmark parameters.

[0071] Specifically, step S302 includes: Step S3021: Generate the defect code of the target defect based on the attributes of the defect entity; the defect code is a code of four-dimensional attributes: component-level-type-working condition.

[0072] Specifically, based on the four-dimensional attributes of defect entities in the atlas, a 14-digit defect code is automatically generated, representing the component, level, type, and operating condition. The coding rule is XX-XX-XX-XX. Specifically, the first two digits are the component code (01=blade, 02=gearbox, 03=generator, 04=converter); the third and fourth digits are the level code (01=urgent, 02=general, 03=minor); the fifth and sixth digits are the type code (01=mechanical failure, 02=electrical failure, 03=control failure); and the seventh and eighth digits are the operating condition code (01=normal, 02=high salt spray, 03=high altitude, 04=low temperature). Example: The code for a general mechanical failure (high salt spray condition) in the gearbox of a 6MW coastal wind farm is 02-02-01-02.

[0073] Step S3022: Based on the defect code, match the component entity and fault type corresponding to the target defect in the knowledge graph, and associate the standard operating procedure corresponding to the target defect.

[0074] This step involves intelligent matching of standard operating procedures (SOPs) and cost elements. Through the rule-based reasoning engine of the knowledge graph, the reasoning process is automatically triggered after the defect code is input: First, match the component entity and fault type corresponding to the target defect; second, associate the SOP with the defect; third, extract the baseline parameters of cost elements such as the type of worker, working hours, spare parts model and specifications, and resource shift quantity corresponding to the SOP; fourth, call up meteorological entity data.

[0075] Step S3023: Extract at least one of the following from the standard operating procedure: labor type, working hours, spare parts model and specifications, and resource shift quantity, as the cost element benchmark parameter.

[0076] This step determines the scope and dimensions of data extraction. From the four-dimensional integrated knowledge graph of defects, costs, scenarios, and weather, more than 1,000 complete defect handling case data are extracted. Each data point must cover all information in the feature dimension and tag dimension to ensure data integrity and relevance.

[0077] Feature dimensions (input items): Core identifier: a 14-digit defect code (e.g., 02-02-01-02, including component, grade, type, and operating condition information); Standard operating procedure parameters: work procedures (e.g., disassembly → inspection → replacement → assembly → adjustment), manpower allocation (job type, number of people, standard working hours), spare parts requirements (model, specifications, quantity), resource allocation (equipment type, number of shifts, usage time); Auxiliary features: Defect occurrence time (season, month), equipment running time, frequency of historical processing records.

[0078] Step S3024: Based on the cost element benchmark parameters, determine the labor cost, spare parts cost, resource cost, and implicit cost corresponding to the target defect; add up the labor cost, spare parts cost, resource cost, and implicit cost to obtain the basic cost quota.

[0079] Explicit costs include labor costs, spare parts costs, and resource costs, which are calculated as follows: Labor costs (unit price × working hours × number of people), spare parts costs (unit price × quantity + transportation cost), and resource costs (equipment shift cost + fuel / electricity cost).

[0080] Calculate implicit costs using the following formula: (4); In the formula, For hidden costs, For the equipment's rated power, Standard operating hours This represents the wind speed power coefficient corresponding to the meteorological entity in the map.

[0081] The specific process for determining the basic cost quota includes: (1) Data preprocessing workflow: a) Cleaning: Remove data with missing key parameters (such as no clear spare part model or incomplete work hour records) and outliers (such as costs that are more than 3 times higher than similar cases), and finally retain 800-1000 valid data. b) Standardization: Coding standardization: Convert defect codes into numerical features (e.g., component codes 01→1, 02→2, grade codes 01→3, 02→2, 03→1, weighted according to urgency). Cost standardization: unify the currency unit (e.g., convert all to yuan), eliminate non-core cost items such as taxes and preferential treatment, and ensure consistent cost standards; c) Dataset partitioning: Randomly partition the training set (640-800 records) and the test set (160-200 records) in an 8:2 ratio. The training set is used for model parameter fitting, and the test set is used to verify the model's generalization ability. When partitioning, ensure that the distribution of each defect type and scenario is uniform to avoid data skew.

[0082] (2) XGBoost (eXtreme Gradient Boosting) model training and base cost quota prediction: 1) Model initialization configuration: a) Core parameter settings: The learning rate (eta) = 0.2 controls the step size of weight updates for each decision tree, balancing model training speed and accuracy. Tree depth ( The tree has 6 layers, which limits the complexity of a single tree and avoids overfitting. Number of iterations ( =100 times, generating 100 decision trees for ensemble prediction; Other parameters: minimum sample weights and ( ) = 1, to prevent abnormal node splitting caused by insufficient sample size; regularization coefficient ( To reduce model complexity, the objective function is set to the squared error regression (reg:squarederror) to adapt to cost numerical prediction scenarios.

[0083] b) Feature engineering optimization: Feature combination: Combine defect code, SOP working hours, spare parts quantity, and resource shifts into cross features to enhance the impact of related features on cost; Feature selection: By using XGBoost's built-in feature importance evaluation (gain value), redundant features with an importance of less than 0.01 (such as case features with a historical processing frequency of less than 3 times) are removed, thereby improving model training efficiency.

[0084] 2) Model training and optimization iteration: Training process: Input the preprocessed training set into the model, and train it according to the process of building a decision tree → calculating error → updating sample weights → iteratively generating the next tree. Validate the root mean square error (RMSE) of the model on the test set every 20 iterations. Early stopping mechanism: If the root mean square error of the test set does not decrease after 10 consecutive iterations (the decrease is <0.5%), the iteration is stopped to avoid overtraining; Parameter fine-tuning: If the root mean square error of the training set is low but the root mean square error of the test set is high (overfitting), then reduce the tree depth (e.g., adjust to 5 layers) or increase the regularization coefficient; if the root mean square errors of both are high (underfitting), then increase the learning rate (e.g., adjust to 0.3) or increase the number of iterations (e.g., adjust to 120 times), ultimately ensuring that the model's prediction error on the test set is ≤5%.

[0085] 3) Basic quota calculation output: Single-case quota prediction: Input the coding of new defects and standard operating procedure parameters into the trained model, and the model outputs four predicted values: labor, spare parts, resources, and hidden costs. Basic cost quota integration: Summing according to the following formula yields the basic quota for handling this defect (standard cost without considering scenario correction). (5); In the formula, Basic cost quota, For labor costs, For spare parts costs, For resource costs, This is a hidden cost.

[0086] Results storage: Information such as basic cost quotas, input features, and prediction errors are synchronized to the knowledge graph, and the corresponding defect codes and standard operating procedures are associated to form a record of defect-feature-quota association, providing basic data for subsequent scenario-based calculations.

[0087] Step S3025: Simulate the processing of the target defect using a digital twin model to obtain the theoretical cost; calculate the deviation rate based on the theoretical cost and the basic cost quota; when the deviation rate is less than or equal to a preset threshold, determine that the basic cost quota has passed the verification and output the basic cost quota; when the deviation rate is greater than the preset threshold, determine the cause of the deviation, adjust the model parameters and / or standard operating procedures used to generate the basic cost quota, regenerate the basic cost quota and perform simulation verification until the deviation rate is less than or equal to the preset threshold.

[0088] This step involves cross-validation of the digital twin model. The specific steps are as follows: 1) Digital twin model building, core component modeling: Equipment 3D modeling: Import the 3D CAD models of defect-related components (such as gearboxes and blades) and surrounding related components to restore the component dimensions, installation positions, and connection relationships, with an accuracy error ≤0.1mm; Scenario modeling: Build a virtual environment consistent with the actual operation and maintenance scenario (such as the terrain and equipment placement of onshore wind farms, and the salt spray environment simulation of coastal wind farms), and import meteorological entity data (such as wind speed and temperature during the operation period). Process modeling: Based on the standard operating procedure parameters in the graph, the action logic of each process of "disassembly-replacement-installation-adjustment" is broken down (e.g., disassembling the gearbox requires first fixing the engine compartment, removing the oil pipes, lifting the old parts, cleaning the mounting surface, lifting the new parts, connecting the pipelines, and debugging the operation). The time threshold and resource usage rules for each action are defined (e.g., crane lifting time = part weight / lifting speed + safety redundancy time).

[0089] 2) Virtual simulation and simulation cost calculation: a) Simulation Execution: Import the standard operating procedure parameters (such as 4 aerial workers, 8 hours, 1 crane) and component 3D models from the XGBoost model into the virtual scene, and automatically execute the entire defect handling process according to the disassembled process logic. Record the process in real time during the simulation. Time consumption: The actual execution time of each process (e.g., disassembly takes 2.5 hours, installation takes 3 hours); Resource consumption: equipment operating time (e.g., crane working for 3 hours), consumable consumption (e.g., quantity of bolts, lubricating oil); Implicit cost association: During the simulation process, virtual meteorological data is read synchronously, and implicit costs are calculated according to formula (4), such as power generation loss during shutdown. For example, if a 6MW generator set operates for 8 hours and the wind speed power coefficient is 0.2, then the implicit cost = 6000kW × 8h × 0.2 × electricity price.

[0090] b) Simulation cost Calculation: Based on the time, resources, and consumables data recorded in the virtual simulation, and combined with the labor unit price, equipment shift fee, and spare parts unit price stored in the graph, the total labor cost, spare parts cost, resource cost, and implicit cost are recalculated to obtain the simulation cost.

[0091] 3) Deviation determination and iterative optimization: The formula for calculating the rate of variation is as follows: (6).

[0092] If the deviation rate is ≤5%, the basic quota passes the verification and can be directly output for subsequent scenario-based calculations; Deviation optimization (when the deviation rate is >5%) involves root cause localization: that is, analyzing the causes of deviations through knowledge graph reasoning.

[0093] like >: Investigate whether the model underestimates the time required for complex processes (such as increased difficulty in disassembling parts under high salt spray conditions) or fails to consider the cost of using special tools; like Check whether the model overestimates spare parts consumption (e.g., no additional consumables are needed for actual replacement) or whether the standard operating procedure parameters are redundant (e.g., too much manual configuration).

[0094] Optimize execution: Model optimization: Adjust XGBoost model parameters (such as increasing feature weights for high salt spray conditions), supplement simulation data to the training set, and retrain the model; Standard operating procedure optimization: Based on the process bottlenecks in the simulation (such as excessive disassembly time), optimize the standard operating procedure (such as adding one assistant worker and adjusting the work sequence) and update the standard operating procedure related parameters in the graph; Loop verification: Recalculate after optimization and Continue until the deviation rate is ≤5%, ensuring that the basic quotas conform to both data patterns and actual operating scenarios.

[0095] Step S303: Based on the knowledge graph, determine the scenario correction coefficient corresponding to the target defect, and use the scenario correction coefficient to correct the basic cost quota to obtain the scenario-based cost quota.

[0096] Specifically, the scene correction coefficients include capacity correction coefficients, terrain correction coefficients, area correction coefficients, and operating condition correction coefficients; step S303 above includes: Step S3031: Determine the scene corresponding to the target defect based on the association between the defect entity and the scene entity; determine the correction coefficient corresponding to the scene based on the correspondence between the scene entity and the correction coefficient, and use it as the scene correction coefficient corresponding to the target defect.

[0097] Specifically, the fourth step involves constructing the correction coefficient matrix: based on the relationship between scene entities and correction coefficients in the knowledge graph, a capacity correction coefficient is established. Terrain correction factor Regional correction coefficient Operating condition correction factor The correction coefficient matrix is ​​constructed according to the following rules: Capacity correction factors: 1.5MW=1.0, 3MW=1.5, 6MW=2.2, 8MW=2.8; Terrain correction factors: Land = 1.0, Coastal = 1.8, Plateau = 1.3, Mountain = 1.4; Regional correction factors: ordinary areas = 1.0, developed areas = 1.2, remote areas = 1.5; Operating condition correction factors: normal operating conditions = 1.0, high salt spray = 1.3, low temperature = 1.2, high altitude = 1.4.

[0098] Step S3032: Multiply the basic cost quota by the capacity correction factor, terrain correction factor, regional correction factor and working condition correction factor in sequence to obtain the scenario-based cost quota.

[0099] By using a path reasoning engine based on a knowledge graph, the system matches the scene entities and correction coefficients corresponding to the target defect, and then uses weighted calculation to obtain the scene-based cost quota. The calculation formula is as follows: (7); Example: Gearbox failure quota for a 6MW remote coastal high-salt-spray operating condition, calculated in ten thousand yuan.

[0100] The formula for calculating the gearbox failure quota in remote coastal areas with high salt spray conditions is as follows: .

[0101] Step S304: Obtain the actual processing cost of the target defect, and calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota to update the scenario-based cost quota.

[0102] Specifically, step S304 includes: Step S3041: Calculate the deviation rate based on the actual processing cost and the scenario-based cost quota; when the deviation rate is greater than the preset deviation threshold, determine the cause of the deviation.

[0103] Set a quota deviation threshold Actual cost of defect handling Calculate the deviation rate: (8); when At that time, the calibration process is triggered.

[0104] Root cause analysis based on indicator graphs: Using fishbone diagram analysis and a conflict detection engine in the knowledge graph, the causes of deviations are located from five dimensions: people, machines, materials, methods, and environment. If the cause is market price fluctuation, the entity attributes of cost elements are updated; if the cause is change in scenario parameters, the correction coefficient matrix is ​​supplemented; if the cause is process optimization, the standard operating procedure relationships are updated; if the cause is knowledge graph data error (such as entity alignment error), it is fed back to the entity alignment module of the heterogeneous graph attention network model to re-optimize the model parameters and correct the entity alignment.

[0105] Step S3042: Update at least one of the following in the knowledge graph based on the cause of the deviation: cost element entity, correction coefficient, association between standard operating procedure and cost element entity, and entity alignment result.

[0106] The calibrated quota data, new defect cases, correction coefficients, and corrected entity alignment results are updated to the knowledge graph to achieve dynamic accumulation of knowledge.

[0107] Step S3043: Recalculate the scenario-based cost quota of the target defect based on the updated knowledge graph to obtain the updated scenario-based cost quota.

[0108] Specifically, after the knowledge graph is calibrated, the cost elements, correction coefficients, standard operating procedure relationships, or entity alignment results are updated. At this point, based on the updated knowledge graph, the cost element benchmark parameters and scenario correction coefficients corresponding to the target defect are re-determined, and the basic cost quota is regenerated and the scenario is corrected in the same way as before, so that the updated scenario-based cost quota can be obtained, making the quota consistent with the calibrated knowledge graph.

[0109] The dynamically calibrated scenario-based cost quota is embedded into the intelligent operation and maintenance scheduling system for new energy equipment. Combined with meteorological entity data in the map, the system selects low-wind-speed operation and maintenance windows and outputs the operation and maintenance plan with the lowest cost and the least power generation loss.

[0110] It should be noted that the dynamic calibration system for defect cost quotas of new energy equipment includes: 1. Knowledge Graph Construction and Management Module: The core includes an entity alignment submodule based on a heterogeneous graph attention network model, an entity and relation definition submodule, a data filling submodule, and a graph storage and visualization submodule. The entity alignment submodule is responsible for entity alignment and knowledge fusion of multi-source heterogeneous data; it supports the import of industry standard specifications, and can convert defect classification and operation and maintenance procedures from standards such as NB / T10574-2021 and NB / T 31047-2013 into graph knowledge rules; the GraphScope tool enables graph visualization and querying.

[0111] 2. Defect coding and standard operating procedure association module: Automatically generates four-dimensional defect codes based on knowledge graphs and intelligently matches the corresponding standard operating procedures and cost element benchmark parameters.

[0112] 3. Quota generation and verification module: It has a built-in XGBoost algorithm model and digital twin simulation engine. It generates basic quotas by inputting feature data in the map and completes cross-validation.

[0113] 4. Correction coefficient management module: Based on knowledge graph, a four-dimensional correction coefficient matrix is ​​constructed, which supports intelligent matching of scenarios and correction coefficients.

[0114] 5. Dynamic Calibration Module: Collects actual operation and maintenance cost data, calculates quota deviation rate, triggers root cause analysis and quota calibration, and updates the knowledge graph synchronously; it has a knowledge conflict detection function, which can automatically identify inconsistent cost data, correction coefficients and entity alignment errors in the graph, and push early warning information.

[0115] 6. Quota Application Module: It interfaces with the intelligent operation and maintenance scheduling system for new energy equipment to output scenario-based and accurate cost quotas, and combines meteorological data to screen low-wind-speed operation and maintenance windows to support operation and maintenance decisions.

[0116] This embodiment also provides the following alternative solutions: Entity alignment model alternatives: In addition to heterogeneous graph attention networks, other entity alignment models based on graph neural networks can be used, such as heterogeneous graph convolutional networks and graph attention network models based on attention mechanisms (adapted to heterogeneous graphs). By adjusting the feature extraction method and attention calculation logic of the model, the entity alignment accuracy can be ensured to be ≥90%, and the effective fusion of multi-source heterogeneous data can still be achieved.

[0117] Knowledge graph construction tool alternatives: In addition to GraphScope, other tools that support large-scale graph construction, reasoning, and visualization, such as Neo4j Bloom and TigerGraph, can be used. The core entity definition, relation model, and heterogeneous graph attention network entity alignment process remain unchanged, and knowledge graph construction and management can still be achieved.

[0118] Defect coding bit length adjustment: The four-dimensional defect coding of component-level-type-operating condition can be adjusted to twelve bits (XX-XX-XX-XX, two bits per segment) or sixteen bits (XXX-XXX-XXX-XXX, three bits per segment) according to the specific needs of actual application scenarios. The core attribute classification logic of component, level, type and operating condition corresponding to the coding remains unchanged, and the standardized classification of defects and intelligent association with standard operating procedures can still be achieved.

[0119] Algorithm model replacement: The extreme gradient boosting algorithm can be replaced with other machine learning algorithms with high-precision regression prediction capabilities, such as random forest and lightweight gradient boosting machine. By adjusting the model parameters (such as the number of decision trees in random forest and the learning rate of lightweight gradient boosting machine), the prediction error of the basic cost quota can be ensured to be ≤5%. This can be combined with the cross-validation process of the digital twin model to achieve accurate generation of the basic cost quota.

[0120] Adjustment of correction coefficient values: The specific values ​​of the correction coefficients for capacity, terrain, region, and operating conditions can be adapted and adjusted according to the industry standards and market conditions of different regions (e.g., the correction coefficient for remote areas can be adjusted to 1.4 or 1.6). The four-dimensional structure of the correction coefficient matrix and the scenario matching logic remain unchanged, and accurate calculation of cross-scenario cost quotas can still be achieved.

[0121] Root cause analysis method alternative: Fishbone diagram analysis can be replaced by five-question analysis, which combines entity relationships in knowledge graphs to trace the root cause of deviations layer by layer from five dimensions: people, machines, materials, methods, and environment. It can still achieve accurate positioning and calibration of cost quota deviations.

[0122] The new energy equipment defect cost quota calibration method provided in this embodiment introduces an entity alignment method based on heterogeneous graph attention network. It uses the attention mechanism to adaptively capture the attribute features and semantic associations of entities in multi-source heterogeneous data, effectively solving the problem that traditional entity alignment methods are difficult to handle the complexity of new energy operation and maintenance data. This improves the entity alignment accuracy to over 95%, reduces entity redundancy, attribute missing and relationship errors, and provides high-quality and highly reliable knowledge support for cost quota calculation.

[0123] Based on a high-quality knowledge graph, semantic associations are achieved among defect, cost, scenario, and meteorological data. The cost quota calculation process relies on precise entity associations and reasoning rules, possesses a complete traceability link and clear explanatory logic, and effectively improves the interpretability and credibility of the cost quota scheme.

[0124] By integrating extreme gradient boosting algorithms with digital twin models and cross-validating through algorithm training and virtual simulation, the deviation of basic cost quotas is reduced to ≤5%. The four-dimensional maintenance positive coefficient matrix of capacity, terrain, region, and working condition fully covers multi-capacity, multi-terrain, and multi-working condition scenarios, reducing cross-scenario cost deviations to within ±10% and accurately adapting to the cost accounting needs of defect handling in different scenarios.

[0125] Establish a closed-loop mechanism of deviation monitoring, root cause analysis, and calibration update, combined with the dynamic iteration of knowledge graphs (including the correction of entity alignment results), so that cost quotas can quickly respond to changes such as market price fluctuations, technology iterations, and process optimizations, and maintain long-term adaptability to actual conditions, avoiding the solidification and disconnection of cost quotas.

[0126] By deeply integrating the dynamically calibrated cost quota with the intelligent operation and maintenance scheduling system, and combining meteorological data to screen low-wind-speed operation and maintenance windows, the optimal operation and maintenance plan can be output, which can reduce the annual power generation loss of a single unit by 5%-8% and the operation and maintenance cost by 20%-30%, providing a scientific and reliable cost benchmark for the intelligent operation and maintenance of new energy equipment.

[0127] As a specific application embodiment of the present invention, combined with Figure 4 The present invention provides a further detailed description of the method for calibrating the defect cost quota of new energy equipment, such as... Figure 4 As shown, the specific process is as follows: Step S401: Knowledge Graph Construction: Input multi-source heterogeneous operation and maintenance data and industry standards, construct a four-in-one knowledge graph covering defects, costs, scenarios and weather through Heterogeneous Graph Attention Network (HGAT) and five-element relation model, and output the four-in-one knowledge graph.

[0128] Step S402: Linking defect coding with standard operating procedures: Input defect attribute characteristics, perform four-dimensional defect coding and associate it with standard operating procedures (SOPs) to generate defect codes, intelligently match SOPs, extract cost element benchmark parameters, and output four-dimensional defect codes and cost element benchmark parameters.

[0129] Step S403: Generation and Verification of Basic Cost Quotas Input defect codes and baseline parameters, and use XGBoost to predict basic cost quotas. And utilize the simulation cost of digital twin simulation Calculate the deviation rate between the two. Determine if the deviation rate is less than or equal to 5%. If so, output the basic cost quota that has passed verification; If not, then adjust the XGBoost parameters to... Re-predict based on the baseline until the deviation rate meets the requirements.

[0130] Step S404: Calculation of scenario-based cost quotas: Input the basic cost quota and scenario parameters (capacity / terrain / region / operating condition), match the four-dimensional maintenance positive coefficient matrix, and calculate the implicit cost. The calculation formula is: The final output is a precise, scenario-based cost quota.

[0131] Step S405: Dynamic calibration of scenario-based cost quotas: Input actual operation and maintenance cost data, calculate the quota deviation rate, and perform intelligent trigger correction coefficient matrix for environmental changes. Determine whether the deviation rate is greater than the threshold or whether an environmental change has occurred: if so, trigger the calibration process, feed back to the knowledge graph for iterative updates, and output the calibrated quota and the updated knowledge graph; if not, directly output the calibrated quota and the updated knowledge graph.

[0132] Step S406: Intelligent Operation and Maintenance Application: Input the calibrated quota and real-time meteorological data, perform cost-benefit game analysis, generate the optimal operation and maintenance plan, make a decision on whether to carry out immediate repairs or postpone maintenance, and finally output the optimal operation and maintenance plan.

[0133] This embodiment also provides a new energy equipment defect cost quota calibration device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0134] This embodiment provides a calibration device for the defect cost quota of new energy equipment, such as Figure 5 As shown, it includes: Knowledge graph construction module 501 is used to construct a knowledge graph for calibration of defect cost quotas for new energy equipment. The basic cost quota generation module 502 is used to determine the cost element benchmark parameters corresponding to the target defect based on the knowledge graph, and generate the basic cost quota of the target defect based on the cost element benchmark parameters. The scenario-based cost quota generation module 503 is used to determine the scenario correction coefficient corresponding to the target defect based on the knowledge graph, and to correct the basic cost quota using the scenario correction coefficient to obtain the scenario-based cost quota. The calibration module 504 is used to obtain the actual processing cost of the target defect and to calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota in order to update the scenario-based cost quota.

[0135] In some optional implementations, the knowledge graph construction module 501 is specifically used for: Acquire multi-source heterogeneous data related to defect handling in new energy equipment; preprocess the multi-source heterogeneous data to obtain preprocessed data; map the preprocessed data into a heterogeneous graph, and use a heterogeneous graph attention network to align entities in the heterogeneous graph; based on the aligned entities, complete the missing attributes of the entities, and extract the relationships between entities to form triplet data; store the triplet data in a graph database to obtain a knowledge graph; the knowledge graph includes defect entities, cost element entities, scene entities, the relationships between defect entities and scene entities, the relationships between standard operating procedures and cost element entities, and the correspondence between scene entities and correction coefficients.

[0136] In some optional implementations, the knowledge graph construction module 501 is further used for: A heterogeneous graph attention network is used to train the heterogeneous graph. The features of different types of nodes and edges are learned through a multi-head attention mechanism to obtain entity fusion features. Based on the entity fusion features, the similarity between entities in different data sources is calculated. Entities with similarity that meet the threshold are merged to obtain aligned entities.

[0137] In one optional implementation, the basic cost quota generation module 502 is specifically used for: Based on the attributes of the defective entity, a defect code for the target defect is generated; the defect code is a four-dimensional attribute code of component-level-type-operating condition; based on the defect code, the component entity and fault type corresponding to the target defect are matched in the knowledge graph, and the standard operating procedure corresponding to the target defect is associated; at least one of the following is extracted from the standard operating procedure: labor type, working hours, spare parts model and specifications, and resource shift quantity, as the benchmark parameter of cost element.

[0138] In an optional implementation, the basic cost quota generation module 502 is further configured to: Based on the cost element benchmark parameters, determine the labor cost, spare parts cost, resource cost, and implicit cost corresponding to the target defect; add up the labor cost, spare parts cost, resource cost, and implicit cost to obtain the basic cost quota.

[0139] The process of handling the target defect is simulated using a digital twin model to obtain the simulation cost; the deviation rate is calculated based on the simulation cost and the basic cost quota; when the deviation rate is less than or equal to a preset threshold, the basic cost quota is verified and output; when the deviation rate is greater than the preset threshold, the cause of the deviation is determined, the model parameters and / or standard operating procedures used to generate the basic cost quota are adjusted, the basic cost quota is regenerated and simulated for verification, until the deviation rate is less than or equal to the preset threshold.

[0140] In one optional implementation, the scenario-based cost quota generation module 503 is specifically used for: Based on the relationship between defect entities and scene entities, the scene corresponding to the target defect is determined; based on the correspondence between scene entities and correction coefficients, the correction coefficient corresponding to the scene is determined, which serves as the scene correction coefficient for the target defect.

[0141] In one optional implementation, the scene correction coefficients include capacity correction coefficients, terrain correction coefficients, area correction coefficients, and operating condition correction coefficients; the scene-based cost quota generation module 503 is also used for: The basic cost quota is multiplied sequentially by the capacity correction factor, terrain correction factor, regional correction factor, and operating condition correction factor to obtain the scenario-based cost quota.

[0142] In one alternative implementation, the calibration module 504 is specifically used for: The deviation rate is calculated based on the actual processing cost and the scenario-based cost quota. When the deviation rate is greater than the preset deviation threshold, the cause of the deviation is determined. Based on the cause of the deviation, at least one of the following is updated in the knowledge graph: cost element entity, correction coefficient, association between standard operating procedures and cost element entities, and entity alignment result. Based on the updated knowledge graph, the scenario-based cost quota of the target defect is recalculated to obtain the updated scenario-based cost quota.

[0143] The new energy equipment defect cost quota calibration device provided in this embodiment of the invention can execute the new energy equipment defect cost quota calibration method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0144] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0145] The following is a detailed reference. Figure 6 This diagram illustrates a suitable structural design for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 601, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 602 or a program loaded from memory 608 into random access memory (RAM) 603. RAM 603 also stores various programs and data required for the operation of the electronic device. The processor 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0146] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0147] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a memory 608, or installed from a ROM 602. When the computer program is executed by the processor 601, it performs the functions defined in the new energy equipment defect cost quota calibration method of the embodiments of the present invention.

[0148] Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0149] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the new energy equipment defect cost quota calibration method shown in the above embodiments is implemented.

[0150] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0151] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for calibrating the defect cost quota of new energy equipment, characterized in that, The method includes: Construct a knowledge graph for calibration of defect cost quotas for new energy equipment; Based on the knowledge graph, the cost element benchmark parameters corresponding to the target defect are determined, and the basic cost quota of the target defect is generated based on the cost element benchmark parameters. Based on the knowledge graph, the scenario correction coefficient corresponding to the target defect is determined, and the basic cost quota is corrected using the scenario correction coefficient to obtain the scenario-based cost quota. The actual processing cost of the target defect is obtained, and the knowledge graph is calibrated based on the deviation between the actual processing cost and the scenario-based cost quota in order to update the scenario-based cost quota.

2. The method according to claim 1, characterized in that, The construction of the knowledge graph for the calibration of defect cost quotas for new energy equipment includes: Acquire multi-source heterogeneous data related to defect handling in new energy equipment; The multi-source heterogeneous data is preprocessed to obtain preprocessed data; The preprocessed data is mapped to a heterogeneous graph, and a heterogeneous graph attention network is used to align entities in the heterogeneous graph. Based on the entities after entity alignment, complete the missing attributes of the entities and extract the relationships between the entities to form triplet data; The triplet data is stored in a graph database to obtain the knowledge graph; the knowledge graph includes defect entities, cost element entities, scene entities, the relationship between defect entities and scene entities, the relationship between standard operating procedures and cost element entities, and the correspondence between scene entities and correction coefficients.

3. The method according to claim 2, characterized in that, The step of using a heterogeneous graph attention network to align entities in the heterogeneous graph includes: The heterogeneous graph is trained using a heterogeneous graph attention network, and the features of different types of nodes and edges are learned through a multi-head attention mechanism to obtain entity fusion features; Based on the entity fusion features, the similarity between entities in different data sources is calculated, and entities whose similarity meets the threshold are merged to obtain aligned entities.

4. The method according to claim 2, characterized in that, The step of determining the cost element benchmark parameters corresponding to the target defect based on the knowledge graph includes: Based on the attributes of the defective entity, a defect code for the target defect is generated; the defect code is a code of four-dimensional attributes: component-level-type-operating condition. Based on the defect code, the component entity and fault type corresponding to the target defect are matched in the knowledge graph, and the standard operating procedure corresponding to the target defect is associated. At least one of the following is extracted from the standard operating procedure: labor type, working hours, spare parts model and specification, and resource shift quantity, as the benchmark parameter of the cost element.

5. The method according to claim 2, characterized in that, The step of generating the basic cost quota for the target defect based on the cost element benchmark parameters includes: Based on the aforementioned cost element benchmark parameters, determine the labor cost, spare parts cost, resource cost, and implicit cost corresponding to the target defect; The basic cost quota is obtained by adding up the labor cost, spare parts cost, resource cost and implicit cost.

6. The method according to claim 5, characterized in that, The step of generating the basic cost quota for the target defect based on the cost element benchmark parameters further includes: The processing of the target defect is simulated using a digital twin model to obtain the simulation cost; The deviation rate is calculated based on the simulation cost and the basic cost quota. When the deviation rate is less than or equal to a preset threshold, the basic cost quota is determined to have passed the verification, and the basic cost quota is output. When the deviation rate is greater than the preset threshold, the cause of the deviation is determined, the model parameters and / or the standard operating procedure used to generate the basic cost quota are adjusted, the basic cost quota is regenerated and simulation verification is performed until the deviation rate is less than or equal to the preset threshold.

7. The method according to claim 2, characterized in that, The step of determining the scene correction coefficient corresponding to the target defect based on the knowledge graph includes: Based on the association between the defective entity and the scene entity, the scene corresponding to the target defect is determined; Based on the correspondence between the scene entities and the correction coefficients, the correction coefficients corresponding to the scene are determined and used as the scene correction coefficients corresponding to the target defect.

8. The method according to claim 1, characterized in that, The scene correction coefficients include capacity correction coefficients, terrain correction coefficients, area correction coefficients, and operating condition correction coefficients; The step of adjusting the basic cost quota using the scenario correction coefficient to obtain a scenario-based cost quota includes: The basic cost quota is multiplied sequentially by the capacity correction coefficient, the terrain correction coefficient, the region correction coefficient, and the operating condition correction coefficient to obtain the scenario-based cost quota.

9. The method according to claim 1, characterized in that, The step of calibrating the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota, in order to update the scenario-based cost quota, includes: The deviation rate is calculated based on the actual processing cost and the scenario-based cost quota. When the deviation rate is greater than a preset deviation threshold, the cause of the deviation is determined; Update at least one of the following in the knowledge graph based on the cause of the deviation: cost element entity, correction coefficient, association between standard operating procedures and cost element entity, and entity alignment result; The scenario-based cost quota of the target defect is recalculated based on the updated knowledge graph to obtain the updated scenario-based cost quota.

10. A calibration device for defect cost quota of new energy equipment, characterized in that, The device includes: The knowledge graph construction module is used to build a knowledge graph for calibration of defect cost quotas for new energy equipment. The basic cost quota generation module is used to determine the cost element benchmark parameters corresponding to the target defect based on the knowledge graph, and generate the basic cost quota of the target defect based on the cost element benchmark parameters. The scenario-based cost quota generation module is used to determine the scenario correction coefficient corresponding to the target defect based on the knowledge graph, and to correct the basic cost quota using the scenario correction coefficient to obtain the scenario-based cost quota. A calibration module is used to obtain the actual processing cost of the target defect, and to calibrate the knowledge graph based on the deviation between the actual processing cost and the scenario-based cost quota, so as to update the scenario-based cost quota.

Images