Power equipment fault prediction method and device based on knowledge graph, terminal equipment and storage medium

Abstract

The invention discloses a power equipment fault prediction method and device based on a knowledge graph, terminal equipment and a storage medium, and belongs to the technical field of fault prediction.The method comprises the steps that hardware asset data and historical operation and maintenance data of power equipment are obtained, entities and the relation between the entities are extracted from the hardware asset data and the historical operation and maintenance data, and an asset knowledge graph is constructed; extracting a sub-graph of to-be-predicted target equipment from the asset knowledge graph; wherein the sub-graph is formed according to a target entity corresponding to the target device, an associated entity of an associated device associated with the target device, and a connection relationship between the target entity and the associated entity; inputting the sub-graph into a preset fault prediction model to obtain the occurrence probability of each fault type of the target equipment in a future preset time period; and finally determining a fault prediction result of the target equipment according to the probability. By implementing the method and the device, the problem that the accuracy of a fault prediction result is relatively low as only monitoring data of the target equipment is utilized for prediction in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and in particular to a method, apparatus, terminal equipment and storage medium for predicting power equipment faults based on knowledge graphs. Background Technology

[0002] With the rapid development of technologies such as the Internet of Things and big data, the power system has gradually formed an integrated monitoring system covering all aspects of power generation, transmission, transformation, distribution, and consumption. Power assets, as the core carriers of power grid operation, directly determine the quality and security of power supply. Therefore, the power monitoring system, as the "nerve center" of the smart grid, achieves real-time monitoring of the operational status of power assets through sensors, smart terminals, data acquisition and monitoring equipment deployed at various stages, generating massive amounts of multi-dimensional asset data.

[0003] Because asset data is mostly scattered across different systems and devices, forming data silos, it is difficult to achieve effective integration and sharing. Traditional fault prediction methods usually only analyze and predict the monitoring data of the target device itself. Since power equipment faults are mostly triggered by multiple factors, the monitoring data of a single device can only reflect the local state and cannot cover the complete causal link of the fault. Therefore, existing technologies have not effectively integrated and analyzed complex heterogeneous data and cannot deeply explore the correlation value between different data. As a result, their fault prediction results have low accuracy. Summary of the Invention

[0004] This invention provides a knowledge graph-based method, device, terminal equipment, and storage medium for predicting power equipment faults. It can solve the problem that existing technologies only use the monitoring data of the target equipment itself for analysis and prediction, without effectively integrating different heterogeneous data, resulting in low accuracy of fault prediction results.

[0005] One embodiment of the present invention provides a power equipment fault prediction method based on knowledge graphs, comprising: Acquire hardware asset data and historical operation and maintenance data of each power device in the power monitoring system; Based on the first preset data recognition model and the second preset data recognition model, entities and the relationships between entities are extracted from the above-mentioned hardware asset data and historical operation and maintenance data, and an asset knowledge graph is constructed based on the above-mentioned entities and relationships. Extract the subgraph of the target device to be predicted from the above asset knowledge graph; wherein the subgraph is composed of the target entity corresponding to the target device, the associated entity corresponding to the associated device associated with the target device, and the connection relationship between the target entity and the associated entity. Input the above subgraph into the preset fault prediction model to obtain the probability of each fault type occurring in the target equipment within a preset future time period; The probability of each fault type is compared with the corresponding preset probability threshold. If any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as the fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be no fault.

[0006] Furthermore, based on the first and second preset data recognition models, the entities and relationships between them are extracted from the aforementioned hardware asset data and historical operation and maintenance data, including: The aforementioned hardware asset data is input into a first preset data recognition model, so that the first preset data recognition model extracts the first entity of each power device and the first relationship between each first entity from the aforementioned hardware asset data; wherein, the aforementioned first entity includes: power device and device location; The aforementioned historical operation and maintenance data is input into the second preset data recognition model, so that the second preset data recognition model can extract the second entity and the second relationship between each second entity from the aforementioned historical operation and maintenance data; wherein, the aforementioned second entity includes: historical fault type and historical maintenance strategy; The first entity and the second entity are referred to as the entities, and the first relationship and the second relationship are referred to as the relationships.

[0007] Furthermore, the asset knowledge graph constructed based on the aforementioned entities and relationships includes: From the aforementioned hardware asset data and historical operation and maintenance data, we can extract the attribute information, historical fault occurrence time, and historical maintenance personnel information of each power equipment. Using the aforementioned entities as nodes, the aforementioned relationships as edges, and the aforementioned attribute information, historical fault occurrence time, and historical maintenance personnel information as the node attributes of the corresponding nodes, the aforementioned asset knowledge graph is constructed.

[0008] Furthermore, the training of the aforementioned first preset data recognition model includes: Obtain several hardware asset description texts with a first real label; wherein the first real label is used to represent the first real entity and the first real relationship in the hardware asset description text; The above hardware asset description text is input into the first data recognition model to be trained for iterative training until the first loss function converges, and the trained first preset data recognition model is obtained. In each iteration, the current predicted entity and predicted relationship are obtained based on the current hardware asset description text; the current first loss function is calculated based on the current predicted first entity, predicted first relationship and corresponding first true label, and it is determined whether the current first loss function has converged; if it has converged, the current first data recognition model is used as the above-mentioned first preset data recognition model; otherwise, the model parameters in the current first data recognition model are adjusted and training continues.

[0009] Furthermore, the training of the aforementioned second preset data recognition model includes: Obtain several operation and maintenance knowledge sample data with second real labels; wherein, the second real labels are used to represent the real second entities and real second relationships in the operation and maintenance knowledge sample data; The above-mentioned operation and maintenance knowledge sample data is input into the second data recognition model to be trained for iterative training until the second damage function converges, and the trained second preset data recognition model is obtained. In each iteration, the current predicted second entity and predicted second relation are obtained based on the current operation and maintenance knowledge sample data. Based on the current predicted second entity, predicted second relation and corresponding second true label, the current second loss function is calculated, and it is determined whether the current second loss function has converged. If it has converged, the current second data recognition model is used as the above-mentioned second preset data recognition model. Otherwise, the model parameters in the current second data recognition model are adjusted, and training continues.

[0010] Furthermore, the training of the aforementioned pre-defined fault prediction model includes: Obtain several sub-graph samples with third real labels; wherein, the third real labels are used to indicate whether the power equipment corresponding to the sub-graph samples has experienced various types of faults within the preset time period; The above subgraph samples are input into the fault prediction model to be trained for iterative training until the third loss function converges, and the trained preset fault prediction model is obtained. In each iteration, the current predicted probability for each fault type is obtained based on the current subgraph samples; the current third damage function is calculated based on the current predicted probability and the corresponding third true label, and it is determined whether the current third loss function has converged; if it has converged, the current fault prediction model is used as the above-mentioned preset fault prediction model; otherwise, the model parameters in the current fault prediction model are adjusted and training continues.

[0011] Furthermore, after comparing the probability of each fault type with the corresponding preset probability threshold, and if any probability is greater than the corresponding preset probability threshold, then the corresponding fault type is taken as the fault prediction result of the target device, the method further includes: From the above asset knowledge graph, the target historical maintenance strategy corresponding to the fault type of the above fault prediction result is extracted; The target equipment shall be maintained in accordance with the aforementioned historical maintenance strategy.

[0012] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments; This invention provides a knowledge graph-based power equipment fault prediction device, comprising: The system includes a data acquisition module, a graph construction module, a subgraph extraction module, a probability prediction module, and a fault prediction module. The aforementioned data acquisition module is used to acquire hardware asset data of each power device in the power monitoring system, as well as historical operation and maintenance data; The aforementioned graph construction module is used to extract entities and the relationships between entities from the aforementioned hardware asset data and historical operation and maintenance data based on the first preset data recognition model and the second preset data recognition model, and to construct an asset knowledge graph based on the aforementioned entities and relationships. The aforementioned subgraph extraction module is used to extract a subgraph of the target device to be predicted from the aforementioned asset knowledge graph; wherein, the aforementioned subgraph is composed of the target entity corresponding to the aforementioned target device, the associated entity corresponding to the associated device associated with the aforementioned target device, and the connection relationship between the aforementioned target entity and the aforementioned associated entity; The aforementioned probability prediction module is used to input the aforementioned subgraph into a preset fault prediction model to obtain the probability of each fault type occurring in the aforementioned target equipment within a preset future time period; The aforementioned fault prediction module is used to compare the probability of each fault type with the corresponding preset probability threshold. If any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as the fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be fault-free.

[0013] Based on the above method embodiments, the present invention provides a corresponding terminal device embodiment; The present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the knowledge graph-based power equipment fault prediction method described in any embodiment of the present invention.

[0014] Based on the above method embodiments, the present invention provides a corresponding storage medium embodiment; The present invention provides a storage medium including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the knowledge graph-based power equipment fault prediction method described in any embodiment of the present invention.

[0015] The embodiments of the present invention have the following beneficial effects: This invention provides a method, apparatus, terminal device, and storage medium for power equipment fault prediction based on a knowledge graph. The method includes: acquiring hardware asset data and historical operation and maintenance data of each power device in a power monitoring system; then, extracting entities and relationships between entities from the hardware asset data and historical operation and maintenance data according to a first preset data recognition model and a second preset data recognition model, and constructing an asset knowledge graph based on the entities and relationships; then, extracting a subgraph of the target device to be predicted from the asset knowledge graph; wherein the subgraph is composed of the target entity corresponding to the target device, the associated entity corresponding to the associated device, and the connection relationship between the target entity and the associated entity; then, inputting the subgraph into a preset fault prediction model to obtain the probability of each fault type occurring in the target device within a preset future time period; finally, comparing the probability of each fault type with the corresponding preset probability threshold; if any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as the fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be no fault. Therefore, this invention first constructs a unified asset knowledge graph for each device in the power monitoring system. Then, based on this graph, it extracts subgraphs containing relevant information data of the device to be predicted and information data of devices associated with the device to be predicted. The invention then predicts the fault prediction result of the device in the future within a preset time period. This process integrates all data related to the device to be predicted, and the fault prediction result is obtained after integrating various relevant information, thus improving the accuracy of the prediction result. Attached Figure Description

[0016] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a knowledge graph-based method for predicting power equipment faults, provided in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of a power equipment fault prediction device based on a knowledge graph, provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0021] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0024] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0025] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0026] See Figure 1 To address the problem that existing technologies rely solely on monitoring data of the target equipment itself for analysis and prediction, without effectively integrating heterogeneous data, resulting in low accuracy in fault prediction, this invention provides a knowledge graph-based method for predicting power equipment faults, comprising: Step S101: Obtain hardware asset data of each power device in the power monitoring system, as well as historical operation and maintenance data; Specifically, the aforementioned hardware asset data includes: the physical location and attribute information of each power device. The attribute information includes: device identification (e.g., device ID, serial number), device type (e.g., transformer, circuit breaker, server), device model, and rated parameters (e.g., voltage, current, capacity). The aforementioned historical maintenance data includes: historical fault records and historical repair records. Historical fault records include: faulty device identification, fault type, and fault occurrence time. Historical repair records include: repair personnel, repair strategy, and repaired device identification.

[0027] Step S102: Based on the first preset data recognition model and the second preset data recognition model, extract entities and the relationships between entities from the above hardware asset data and historical operation and maintenance data, and construct an asset knowledge graph based on the above entities and relationships. Specifically, the aforementioned first preset data recognition model is a neural network model, which contains a basic encoding layer based on the BERT model, a BiLSTM+CRF layer for predicting entity label sequences, and a fully connected layer. The basic encoding layer is used to convert the input data into contextual semantic vectors. The BiLSTM+CRF layer predicts entity label sequences, and then entity pairs are extracted based on the entity label sequences. The semantic vectors of the entity pairs are concatenated and input into the fully connected layer for relation classification.

[0028] Specifically, the aforementioned second preset data recognition model is a BERT-BiLSTM model, which internally includes a BERT encoder, a BiLSTM layer, a CRF decoding layer, and a relation classification module. The BERT layer is used to encode the semantics of the input descriptive text (e.g., distinguishing between "loose terminal" and "burnt terminal"). Then, the BiLSTM layer is used to capture contextual dependencies (e.g., identifying the hierarchical relationship between devices and components in "communication port failure of relay protection cabinet A"). Next, the CRF decoding layer is used to output the entities corresponding to historical operation and maintenance data. Finally, the relation classification module is used to generate triples based on an attention mechanism (e.g., [device] - fault -> [type]).

[0029] In a preferred embodiment, the extraction of entities and the relationships between entities from the hardware asset data and historical operation and maintenance data based on the first preset data recognition model and the second preset data recognition model includes: The aforementioned hardware asset data is input into a first preset data recognition model, so that the first preset data recognition model extracts the first entity of each power device and the first relationship between each first entity from the aforementioned hardware asset data; wherein, the aforementioned first entity includes: power device and device location; Specifically, the aforementioned first relationship includes at least one of the following: positional relationship, connection relationship, hierarchical relationship, and redundancy relationship.

[0030] The aforementioned historical operation and maintenance data is input into the second preset data recognition model, so that the second preset data recognition model can extract the second entity and the second relationship between each second entity from the aforementioned historical operation and maintenance data; wherein, the aforementioned second entity includes: historical fault type and historical maintenance strategy; Specifically, the second relationship mentioned above includes the relationship between fault and maintenance strategy.

[0031] The first entity and the second entity are referred to as the entities, and the first relationship and the second relationship are referred to as the relationships.

[0032] In this preferred embodiment, a first entity and a first relationship are extracted from hardware asset data using a first preset data identification model, and a second entity and a second relationship are extracted from historical operation and maintenance data using a second preset data identification model.

[0033] In another preferred embodiment, the asset knowledge graph constructed based on the aforementioned entities and relationships includes: From the aforementioned hardware asset data and historical operation and maintenance data, we can extract the attribute information, historical fault occurrence time, and historical maintenance personnel information of each power equipment. Using the aforementioned entities as nodes, the aforementioned relationships as edges, and the aforementioned attribute information, historical fault occurrence time, and historical maintenance personnel information as the node attributes of the corresponding nodes, the aforementioned asset knowledge graph is constructed.

[0034] Specifically, when constructing an asset knowledge graph, the uniqueness of the equipment identifier can be used to match and align the various power equipment. In the constructed asset knowledge graph, the relationships include at least one of the following: equipment-fault relationship, fault-maintenance strategy relationship, equipment location relationship, equipment connection relationship, hierarchical relationship, and redundancy relationship.

[0035] In this preferred embodiment, an asset knowledge graph is constructed based on entities and relationships.

[0036] In another preferred embodiment, the training of the first preset data recognition model includes: Obtain several hardware asset description texts with a first real label; wherein the first real label is used to represent the first real entity and the first real relationship in the hardware asset description text; The above hardware asset description text is input into the first data recognition model to be trained for iterative training until the first loss function converges, and the trained first preset data recognition model is obtained. In each iteration, the current predicted entity and predicted relationship are obtained based on the current hardware asset description text; the current first loss function is calculated based on the current predicted first entity, predicted first relationship and corresponding first true label, and it is determined whether the current first loss function has converged; if it has converged, the current first data recognition model is used as the above-mentioned first preset data recognition model; otherwise, the model parameters in the current first data recognition model are adjusted and training continues.

[0037] Specifically, the hardware asset description text used for model training contains the same content as the hardware asset data, and the first real entities in the first real labels include at least two entity types: power equipment and equipment location. The types of the first real relationships include at least location relationships, connection relationships, hierarchical relationships, and redundant relationships. During training, the backpropagation algorithm is used to optimize the corresponding model parameters until the first loss function converges. The trained model can identify entities and their corresponding relationships from the input data.

[0038] For illustration, the descriptive text could be: Server A, located in the main control room, is connected to switch C in substation B via redundant links. Then, the BIOES annotation system is used to annotate this descriptive text with the first real-world label, for example: labeling the entity corresponding to the power equipment: Server A → Equipment; labeling the entity corresponding to the equipment location: Main Control Room → Location.

[0039] Preferably, after the first data recognition model is trained, its performance can be further verified. When the predetermined performance indicators are achieved in entity recognition and relation extraction tasks, this model is then saved as the aforementioned first preset data recognition model.

[0040] In this preferred embodiment, the first data recognition model was trained using several hardware asset description texts with first real labels.

[0041] In another preferred embodiment, the training of the second preset data recognition model includes: Obtain several operation and maintenance knowledge sample data with second real labels; wherein, the second real labels are used to represent the real second entities and real second relationships in the operation and maintenance knowledge sample data; Specifically, the operation and maintenance knowledge sample data also includes historical fault records and historical maintenance records of the corresponding power equipment.

[0042] The above-mentioned operation and maintenance knowledge sample data is input into the second data recognition model to be trained for iterative training until the second damage function converges, and the trained second preset data recognition model is obtained. In each iteration, the current predicted second entity and predicted second relation are obtained based on the current operation and maintenance knowledge sample data. Based on the current predicted second entity, predicted second relation and corresponding second true label, the current second loss function is calculated, and it is determined whether the current second loss function has converged. If it has converged, the current second data recognition model is used as the above-mentioned second preset data recognition model. Otherwise, the model parameters in the current second data recognition model are adjusted, and training continues.

[0043] Specifically, for operation and maintenance knowledge sample data, the following are used as secondary real labels: labeling the words describing the fault type (e.g., "insulation breakdown" or "communication interruption") and binding them with the fault occurrence time attribute; labeling the description of the maintenance strategy (e.g., "replace terminal" or "firmware upgrade") and binding it with the maintenance personnel attribute; labeling the equipment identifier associated with the fault or maintenance (e.g., "protection device P1"); labeling the attribution relationship between power equipment and fault type (e.g., "[protection device P1] - fault occurred -> [communication interruption]"); and labeling the handling relationship between maintenance strategy and fault type ("[replace terminal] - repair -> [communication interruption]").

[0044] In this preferred embodiment, the second preset data recognition model was trained using a number of operation and maintenance knowledge sample data with second real labels.

[0045] Step S103: Extract the subgraph of the target device to be predicted from the above asset knowledge graph; wherein, the subgraph is composed of the target entity corresponding to the target device, the associated entity corresponding to the associated device associated with the target device, and the connection relationship between the target entity and the associated entity. Specifically, since each node in the asset knowledge graph has corresponding node attributes, each node in the extracted subgraph also has corresponding node attributes. During subgraph extraction, the entity type of the target device (electrical equipment) is first determined. Then, based on this, other types of entities corresponding to the target device are extracted and included as the target entity, such as the device location, historical fault types, and historical maintenance strategies. Next, related entities of other electrical equipment that have edge connections with the target device's electrical equipment entity are extracted. Finally, the edges between these entities and their corresponding node attributes are extracted, resulting in the subgraph.

[0046] Step S104: Input the above subgraph into the preset fault prediction model to obtain the probability of each fault type occurring in the target equipment within a preset time period in the future; Specifically, the aforementioned pre-defined fault prediction model is a neural network model, which includes an input layer, a graph neural network module, a temporal feature extraction module, and a joint output layer. The graph neural network module (e.g., a GNN network) encodes the topological relationship features between devices in the subgraph, the temporal feature extraction module (e.g., an LSTM network) encodes the temporal patterns of device fault history, and the joint output layer predicts the probability of each fault type occurring within a pre-defined time period based on the output features of the graph neural network module and the temporal feature extraction module.

[0047] In a preferred embodiment, training the aforementioned preset fault prediction model includes: Obtain several sub-graph samples with third real labels; wherein, the third real labels are used to indicate whether the power equipment corresponding to the sub-graph samples has experienced various types of faults within the preset time period; The above subgraph samples are input into the fault prediction model to be trained for iterative training until the third loss function converges, and the trained preset fault prediction model is obtained. In each iteration, the current predicted probability for each fault type is obtained based on the current subgraph samples; the current third damage function is calculated based on the current predicted probability and the corresponding third true label, and it is determined whether the current third loss function has converged; if it has converged, the current fault prediction model is used as the above-mentioned preset fault prediction model; otherwise, the model parameters in the current fault prediction model are adjusted and training continues.

[0048] Specifically, each sub-graph sample corresponds to the state of the power equipment at a specific historical moment (e.g., January 1, 2023). Its content type is the same as the sub-graphs mentioned above, the difference being that the sub-graphs are extracted around the target equipment to be predicted, while the sub-graph samples are extracted around the corresponding power equipment. The sub-graph samples are labeled with a binary label indicating whether the power equipment will experience a specific fault type within a preset future time period (e.g., the next 30 days), serving as the third true label. For illustration, if the corresponding specific fault type occurs, it is labeled as 1; otherwise, it is labeled as 0.

[0049] Specifically, when the sub-image samples are input into the model for iterative training, the aforementioned third true label is used as a supervision signal to train the model until the third damage function converges, so that the trained model has the ability to predict the faults of the target device.

[0050] In this preferred embodiment, a preset fault prediction model is trained using several subgraph samples with third real labels.

[0051] Step S105: Compare the probability of each fault type with the corresponding preset probability threshold. If any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as the fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be fault-free.

[0052] Specifically, if the probability of all fault types is not greater than the corresponding preset probability threshold, it means that the target device will not fail within the preset time period in the future. However, if the probability of any fault type is greater than the corresponding preset probability threshold, it means that the target device will fail within the preset time period in the future. Therefore, the judgment result is that the target device will fail within the corresponding fault type.

[0053] In a preferred embodiment, after comparing the probability of each fault type with a corresponding preset probability threshold, and if any probability is greater than the corresponding preset probability threshold, then taking the corresponding fault type as the fault prediction result of the target device, the method further includes: From the above asset knowledge graph, the target historical maintenance strategy corresponding to the fault type of the above fault prediction result is extracted; The target equipment shall be maintained in accordance with the aforementioned historical maintenance strategy.

[0054] Specifically, historical maintenance strategies corresponding to the fault prediction results are extracted from the asset knowledge graph and used as the target historical maintenance strategies. Maintenance suggestions for the target equipment are then randomly generated according to these strategies to achieve equipment maintenance.

[0055] In this preferred embodiment, after determining that the target device will fail within a preset period of time in the future, the target device is maintained based on the historical maintenance strategy corresponding to the failure type.

[0056] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0057] like Figure 2 As shown, one embodiment of the present invention provides a power equipment fault prediction device based on knowledge graph, comprising: The system includes a data acquisition module, a graph construction module, a subgraph extraction module, a probability prediction module, and a fault prediction module. The aforementioned data acquisition module is used to acquire hardware asset data of each power device in the power monitoring system, as well as historical operation and maintenance data; The aforementioned graph construction module is used to extract entities and the relationships between entities from the aforementioned hardware asset data and historical operation and maintenance data based on the first preset data recognition model and the second preset data recognition model, and to construct an asset knowledge graph based on the aforementioned entities and relationships. The aforementioned subgraph extraction module is used to extract a subgraph of the target device to be predicted from the aforementioned asset knowledge graph; wherein, the aforementioned subgraph is composed of the target entity corresponding to the aforementioned target device, the associated entity corresponding to the associated device associated with the aforementioned target device, and the connection relationship between the aforementioned target entity and the aforementioned associated entity; The aforementioned probability prediction module is used to input the aforementioned subgraph into a preset fault prediction model to obtain the probability of each fault type occurring in the aforementioned target equipment within a preset future time period; The aforementioned fault prediction module is used to compare the probability of each fault type with the corresponding preset probability threshold. If any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as the fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be fault-free.

[0058] It should be noted that the device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without creative effort. The above schematic diagrams are merely examples of a knowledge graph-based power equipment fault prediction device and do not constitute a limitation on a knowledge graph-based power equipment fault prediction device. It may include more or fewer components than illustrated, or combine certain components, or use different components.

[0059] Based on the above method embodiments, the present invention provides corresponding terminal device embodiments.

[0060] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the knowledge graph-based power equipment fault prediction method described in any embodiment of the present invention.

[0061] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the device. The aforementioned terminal devices may be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. These devices may include, but are not limited to, processors and memory. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the device, connecting various parts of the device via various interfaces and lines. The aforementioned memory can be used to store the aforementioned computer programs and / or modules. The aforementioned processor implements various functions of the aforementioned device by running or executing the computer programs and / or modules stored in the aforementioned memory, and by calling data stored in the memory. The aforementioned memory may mainly include a program storage area and a data storage area, wherein the program storage area may store the operating system, at least one application program required for a function, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0062] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0063] Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute the knowledge graph-based power equipment fault prediction method described in any embodiment of the present invention.

[0064] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0065] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A power equipment fault prediction method based on a knowledge graph, characterized in that, The method comprises: obtaining hardware asset data of each power equipment in a power monitoring system and historical operation and maintenance data; extracting entities and relationships between the entities from the hardware asset data and the historical operation and maintenance data according to a first preset data recognition model and a second preset data recognition model, and constructing an asset knowledge graph according to the entities and the relationships; extracting a subgraph of a target equipment to be predicted from the asset knowledge graph; wherein the subgraph is composed of a target entity corresponding to the target equipment, an associated entity corresponding to an associated equipment associated with the target equipment, and a connection relationship between the target entity and the associated entity; inputting the subgraph into a preset fault prediction model to obtain probabilities of each fault type occurring in a future preset period of time of the target equipment; comparing the probabilities of each fault type with corresponding preset probability thresholds respectively, and if any probability is greater than the corresponding preset probability threshold, taking the corresponding fault type as a fault prediction result of the target equipment; otherwise, determining that the fault prediction result of the target equipment is no fault. 2.The power equipment fault prediction method based on a knowledge graph according to claim 1, wherein, The method comprises: inputting the hardware asset data into the first preset data recognition model, so that the first preset data recognition model extracts first entities of each power equipment and first relationships between the first entities from the hardware asset data; wherein the first entities include power equipment and equipment location; inputting the historical operation and maintenance data into the second preset data recognition model, so that the second preset data recognition model extracts second entities and second relationships between the second entities from the historical operation and maintenance data; wherein the second entities include historical fault types and historical maintenance strategies; taking the first entities and the second entities as the entities and taking the first relationships and the second relationships as the relationships. 3.The power equipment fault prediction method based on the knowledge graph of claim 2, characterized in that, The method comprises: extracting attribute information, historical fault occurrence time, and historical maintenance personnel information of each power equipment from the hardware asset data and the historical operation and maintenance data; constructing the asset knowledge graph by taking the entities as nodes, the relationships as edges, and the attribute information, the historical fault occurrence time, and the historical maintenance personnel information as node attributes of corresponding nodes.

4. The power equipment fault prediction method based on a knowledge graph according to claim 3, characterized in that, The training of the first preset data recognition model comprises: obtaining a plurality of hardware asset description texts with first real labels; wherein the first real labels are used to represent first real entities and first real relationships in the hardware asset description texts; inputting the hardware asset description texts into a first data recognition model to be trained for iterative training until a first loss function converges, to obtain a trained first preset data recognition model; Wherein, in each iteration, according to the current hardware asset description text, the current predicted entity and the predicted relationship are obtained; according to the current predicted first entity, the predicted first relationship and the corresponding first real label, the current first loss function is calculated, and whether the current first loss function converges is judged; if converges, the current first data recognition model is taken as the first preset data recognition model; otherwise, after adjusting the model parameters in the current first data recognition model, the training is continued.

5. The power equipment fault prediction method based on a knowledge graph according to claim 4, characterized in that, The training of the second preset data recognition model comprises: Obtaining a plurality of operation and maintenance knowledge sample data with second real labels; wherein, the second real label is used to represent the real second entity and the real second relationship in the operation and maintenance knowledge sample data; The operation and maintenance knowledge sample data is input into the second data recognition model to be trained for iterative training until the second loss function converges, and a trained second preset data recognition model is obtained; Wherein, in each iteration, according to the current operation and maintenance knowledge sample data, the current predicted second entity and the predicted second relationship are obtained; according to the current predicted second entity, the predicted second relationship and the corresponding second real label, the current second loss function is calculated, and whether the current second loss function converges is judged; if converges, the current second data recognition model is taken as the second preset data recognition model; otherwise, after adjusting the model parameters in the current second data recognition model, the training is continued.

6. The power equipment fault prediction method based on a knowledge graph according to claim 5, characterized in that, The training of the preset fault prediction model comprises: Obtaining a plurality of subgraph samples with third real labels; wherein, the third real label is used to represent whether the power equipment corresponding to the subgraph sample has each fault type of fault in the preset period; The subgraph sample is input into the fault prediction model to be trained for iterative training until the third loss function converges, and a trained preset fault prediction model is obtained; Wherein, in each iteration, according to the current subgraph sample, the current prediction probability of each fault type is obtained; according to the current prediction probability and the corresponding third real label, the current third loss function is calculated, and whether the current third loss function converges is judged; if converges, the current fault prediction model is taken as the preset fault prediction model; otherwise, after adjusting the model parameters in the current fault prediction model, the training is continued.

7. The power equipment fault prediction method based on a knowledge graph according to claim 6, characterized in that, After the probability of each fault type is compared with the corresponding preset probability threshold value, if any probability is greater than the corresponding preset probability threshold value, the corresponding fault type is taken as the fault prediction result of the target equipment, the method further comprises: From the asset knowledge graph, the target historical maintenance strategy of the fault type corresponding to the fault prediction result is extracted; The target device is maintained according to the target historical maintenance strategy.

8. A power equipment fault prediction device based on a knowledge graph, characterized in that, Comprise: Data acquisition module, graph construction module, subgraph extraction module, probability prediction module and fault prediction module; The data acquisition module is used for acquiring hardware asset data and historical operation and maintenance data of each power equipment in the power monitoring system; The graph construction module is configured to extract entities and relationships between the entities from the hardware asset data and the historical operation and maintenance data according to a first preset data recognition model and a second preset data recognition model, and construct an asset knowledge graph according to the entities and the relationships. The subgraph extraction module is configured to extract a subgraph of a target device to be predicted from the asset knowledge graph, wherein the subgraph is composed of a target entity corresponding to the target device, an associated entity corresponding to an associated device associated with the target device, and a connection relationship between the target entity and the associated entity. The probability prediction module is configured to input the subgraph into a preset fault prediction model to obtain probabilities of each fault type of the target device occurring in a future preset time period. The fault prediction module is configured to compare the probabilities of each fault type with corresponding preset probability thresholds, and if any probability is greater than the corresponding preset probability threshold, the corresponding fault type is taken as a fault prediction result of the target device; otherwise, the fault prediction result of the target device is determined to be no fault.

9. A terminal device, comprising: A storage medium includes a stored computer program, wherein when the computer program runs, the storage medium controls a device where the storage medium is located to perform a power equipment fault prediction method based on a knowledge graph according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium includes a stored computer program, wherein when the computer program runs, the storage medium controls a device where the storage medium is located to perform a power equipment fault prediction method based on a knowledge graph according to any one of claims 1 to 7.

Images