Transformer operation inspection-oriented causal relationship enhanced knowledge graph construction method and system

Abstract

The invention belongs to the technical field of knowledge graph construction, and particularly relates to a transformer operation inspection-oriented causal relationship enhanced knowledge graph construction method and system. According to the method, information in a transformer fault processing text is deeply mined, a natural language processing technology and a deep learning algorithm are utilized, and accurate recognition and relation extraction of key entities such as fault description, reasons, phenomena, detection methods, detection results and operation and maintenance suggestions are achieved. Structured knowledge is organized and stored in a graph database, and a comprehensive, deep and easy-to-query transformer fault operation and maintenance knowledge graph is formed. By introducing a causal scientific theory, the causal relationship in the knowledge graph is remarkably enhanced, fault operation and maintenance causal knowledge with a more stable causal effect is formed, the accuracy and robustness of the causal relationship in the knowledge graph are improved, and operation and maintenance personnel can more intuitively understand complex association between fault phenomena and reasons.

Description

Technical Field

[0001] The present invention belongs to the technical field of knowledge graph construction, and specifically relates to a method and system for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance. Background Art

[0002] As the core hub of the power grid architecture, the healthy operation of the transformer is directly related to the overall stability and safety of the power system. Any transformer failure may quickly spread to the entire power grid, triggering a large-scale regional power outage, causing chain reactions such as damage to key equipment and stagnation of production lines, seriously disrupting the social and economic order and the daily life of residents. Therefore, the importance of transformer operation and maintenance management is self-evident, and it is the cornerstone of the power industry to maintain social stability and ensure power supply. As the concept of smart grid is deeply rooted in the hearts of the people, the power industry is undergoing a profound transformation from traditional to intelligent. In this wave of transformation, how to significantly improve the intelligence level of transformer operation and maintenance with the help of modern information technology has become a key issue that is of general concern and urgently needs to be solved within and outside the industry. Intelligent operation and maintenance requires not only the ability to monitor the operating status of the transformer in real time, but also accurate prediction and early intervention before the failure occurs, to ensure that the transformer can continue to serve the power grid efficiently and stably. In this context, the detailed records accumulated during the transformer operation and maintenance inspection process are becoming more and more valuable. These data are not only direct evidence reflecting the historical operating status of the equipment, but also an indispensable key data source for in-depth analysis of the causes of faults and formulation of targeted treatment strategies. Therefore, in-depth mining and analysis of these recorded data has become an important way to improve the level of intelligent operation and maintenance of transformers.

[0003] Transformer operation and maintenance is faced with complex and changeable working conditions and harsh operating environments, which requires that the decision results given by the adopted models or methods must have extremely high credibility. However, in the current fault operation and maintenance practice, people often focus on direct analysis of fault characteristics and quickly provide solutions for technical personnel to refer to, but ignore providing sufficient theoretical basis and explanation for these solutions. Although this approach can temporarily solve the problem, it is difficult to win the full trust of technical personnel, affecting their acceptance and implementation of fault solutions. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance, so as to solve one of the technical problems in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance, comprising the following steps: Obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; Based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is recognized and processed to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, the pre-trained ResNet-PCNN-ATT model is used to extract the relationship of the fault handling text to obtain the transformer fault operation and maintenance knowledge triple, and the transformer fault operation and maintenance knowledge triple is stored in the graph database to obtain the transformer operation and maintenance knowledge graph; Extract the transformer fault operation and maintenance related content from the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; Define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, calculate the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph, adjust the transformer operation and maintenance knowledge graph based on the causal relationship effect, and obtain the causal relationship enhanced knowledge graph.

[0006] Furthermore, based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is identified and processed to obtain the final result of key entity labeling of the fault handling text: The ALBERT-BiLSTM-CRF model includes an ALBERT layer, a BiLSTM layer and a CRF layer; The ALBERT layer extracts a high-level feature vector representation of the text based on the input initial vector sequence; the high-level feature vector representation of the text is input into the BiLSTM layer, and the BiLSTM layer captures the contextual language features of each word in the text and outputs a vector sequence that incorporates the contextual information; The high-level feature vector representation output by the ALBERT layer and the vector sequence output by the BiLSTM layer that incorporates context information are fused in the feature concatenation layer, and the fusion result is processed by the fully connected layer to obtain a score sequence for each key entity; wherein the score sequence is composed of the respective scores of all labels corresponding to the key entity; The score sequence of each key entity is input into the CRF layer, and a comprehensive score is calculated in the CRF layer according to the score of each label and the transfer score between different labels; based on the comprehensive score, the CRF layer outputs a label path as the final result of key entity labeling; wherein the label path has the largest comprehensive score among all possible label sequences.

[0007] Furthermore, the conditional probability relational expression is used to represent the conditional probability of the association relationship between entities after intervention adjustment; wherein the intervention adjustment is specifically: Determine the nodes to be intervened and adjusted in the fault operation and maintenance knowledge structure causal graph; which includes nodes R, S and T; T is an intermediary node that can send directional connections to R and S at the same time; The do operator is used to implement do intervention to eliminate the association between nodes R and T.

[0008] Furthermore, before marking the fault handling text, the following steps are also included: Based on a pre-built stop word list, noise words in the transformer fault handling text are filtered out.

[0009] Furthermore, the BIO annotation system is used to annotate the fault handling text.

[0010] Perform word segmentation on the fault handling text that identifies the key entities, and convert the word segmentation of the fault handling text into an initial vector sequence through the Embedding layer; Furthermore, the conditional probability relation is expressed as follows: P(S|do(R))=

[0011] Among them, P(S|do(R)) represents the conditional probability of S occurring after the intervention on node R; represents the conditional probability of S occurring given R and t; is the probability of the latent variable or hidden variable.

[0012] Furthermore, the content related to transformer fault operation and maintenance is extracted from the transformer operation and maintenance knowledge graph to obtain a causal graph of the fault operation and maintenance knowledge structure, including: based on the transformer operation and maintenance knowledge graph, query and search are performed according to preset search statements related to transformer fault operation and maintenance to obtain a causal graph of the fault operation and maintenance knowledge structure.

[0013] In a second aspect, the present invention provides a device for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance, comprising: A data acquisition module is used to obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; A graph construction module is used to identify and process the fault handling text with key entities identified based on a pre-trained ALBERT-BiLSTM-CRF model to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, a pre-trained ResNet-PCNN-ATT model is used to extract relations from the fault handling text to obtain transformer fault operation and maintenance knowledge triples, and the transformer fault operation and maintenance knowledge triples are stored in a graph database to obtain a transformer operation and maintenance knowledge graph; The data extraction module is used to extract the transformer fault operation and maintenance related content in the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; A graph enhancement module is used to define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph is calculated, and based on the causal relationship effect, the transformer operation and maintenance knowledge graph is adjusted to obtain a causal relationship enhanced knowledge graph.

[0014] In a third aspect of the present invention, an electronic device is provided, comprising a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the above-mentioned method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance.

[0015] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance as described above is implemented.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The causal relationship enhanced knowledge graph construction method of the present invention realizes accurate identification and relationship extraction of key entities such as fault description, cause, phenomenon, detection method, detection result and operation and maintenance suggestions by deeply mining the information in the transformer fault handling text, using natural language processing technology and deep learning algorithm. Structured knowledge is organized and stored in the graph database to form a comprehensive, in-depth and easy-to-query transformer fault operation and maintenance knowledge graph. By introducing the causal science theory, the causal relationship in the knowledge graph is significantly enhanced to form fault operation and maintenance causal knowledge with more stable causal effects, which improves the accuracy and robustness of the causal relationship in the knowledge graph, so that the operation and maintenance personnel can more intuitively understand the complex relationship between the fault phenomenon and the cause, thereby realizing in-depth tracking and efficient solution of the root cause of the fault. Through the causal relationship enhanced knowledge graph, the operation and maintenance personnel can quickly locate the source of the fault, formulate targeted maintenance plans, shorten the fault handling time, reduce the operation and maintenance cost, and improve the overall stability and reliability of the power system. A causal relationship enhanced knowledge graph construction device, electronic device and computer-readable storage medium for transformer operation and inspection provided by the present invention also solve the problems raised in the background technology part. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 A schematic diagram of a process of constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to an embodiment of the present invention; Figure 2 A schematic diagram of the principle of a method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to an embodiment of the present invention; Figure 3 Schematic diagram of intervention adjustment strategy of structural causal diagram in an embodiment of the present invention Figure 4 This is a structural block diagram of a device for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to an embodiment of the present invention; Figure 5 The present invention is a structural block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0019] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0020] Example 1 In today's industrial field, with the increasing complexity of power systems and the high dependence of critical infrastructure on power supply stability, the stable operation of transformers has become an important part of ensuring the security of power supply. However, as the core equipment for power transmission and distribution, transformer failures are often sudden, complex and diverse, which brings huge challenges to operation and maintenance. Traditional fault prediction and maintenance methods often rely on experience judgment and regular maintenance, which are not only inefficient, but also difficult to accurately capture the root cause of the fault, resulting in a long fault handling cycle and high cost. By constructing a knowledge graph with enhanced causal relationships, the present invention not only promotes the intelligent upgrade of transformer fault prediction and maintenance, but also provides strong technical support for in-depth tracking and efficient resolution of the root cause of the fault.

[0021] The current transformer operation and maintenance field mainly focuses on the knowledge representation and management of industrial text data, which has played a significant role in promoting the transformation of operation and maintenance data into valuable operation and maintenance knowledge. However, current research generally ignores the mining of deep causal relationships in operation and maintenance texts, and a large amount of prior knowledge and causal information implicit in operation and maintenance data have not been fully developed and utilized. These prior knowledge and causal information are important cornerstones to ensure the interpretability and robustness of tasks such as production safety, fault diagnosis, and preventive maintenance in the transformer operation and maintenance process.

[0022] Currently, in the field of transformer operation and maintenance, there are software tools for complex fault analysis based on equipment fault trees. These softwares play an important role in the initial stage of troubleshooting. By building a detailed fault tree model, they systematically enumerate and classify the various causes that may cause equipment failures. However, the limitations of this method lie in its static and preset nature. Fault tree models are often built based on past experience and expert knowledge. Although they can cover a wide range of fault causes, they are unable to cope with new situations and new problems that emerge in actual detection. It is unable to conduct real-time, in-depth analysis and judgment of the specific path and possibility of fault occurrence based on the dynamic changes of actual detection data.

[0023] With its powerful semantic association and structured representation capabilities, knowledge graphs provide strong support for the analysis of equipment failure mechanisms and the formulation of operation and maintenance decisions. However, there are still many challenges in the construction and maintenance of knowledge graphs. On the one hand, data annotation and automatic extraction are one of the main means of constructing knowledge graphs. Although this method can reduce the manual burden to a certain extent, its accuracy is highly dependent on the quality of the annotated data and the performance of the extraction model, and the effect of automatic extraction is often difficult to guarantee for complex and changeable equipment failure conditions. On the other hand, manual organization, as another common method, can ensure the accuracy and completeness of knowledge, but it is labor-intensive, time-consuming, and difficult to ensure real-time updating of knowledge. In the rapidly changing operation and maintenance environment, both methods face the dilemma of poor real-time knowledge acquisition, and it is difficult to meet the urgent needs of operation and maintenance work for efficient and accurate knowledge.

[0024] like Figure 1 and Figure 2 As shown in the figure, a causal relationship enhanced knowledge graph construction method for transformer operation and maintenance includes: S1. Obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities.

[0025] Specifically, before the fault handling text is annotated, the method further includes filtering out noise words in the transformer fault handling text based on a pre-built stop word list.

[0026] Preferably, the BIO annotation system is used to annotate the fault handling text. The fault handling text with the key entity identified is segmented, and the segmented words of the fault handling text are converted into an initial vector sequence through the Embedding layer.

[0027] S2. Based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is recognized and processed to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, the pre-trained ResNet-PCNN-ATT model is used to extract relations from the fault handling text to obtain transformer fault operation and maintenance knowledge triples, and the transformer fault operation and maintenance knowledge triples are stored in the graph database to obtain the transformer operation and maintenance knowledge graph.

[0028] Specifically, in step S2: The ALBERT-BiLSTM-CRF model includes an ALBERT layer, a BiLSTM layer and a CRF layer; The ALBERT layer extracts a high-level feature vector representation of the text based on the input initial vector sequence; the high-level feature vector representation of the text is input into the BiLSTM layer, and the BiLSTM layer captures the contextual language features of each word in the text and outputs a vector sequence that incorporates the contextual information; The high-level feature vector representation output by the ALBERT layer and the vector sequence output by the BiLSTM layer that incorporates context information are fused in the feature concatenation layer, and the fusion result is processed by the fully connected layer to obtain a score sequence for each key entity; wherein the score sequence is composed of the respective scores of all labels corresponding to the key entity; The score sequence of each key entity is input into the CRF layer, and a comprehensive score is calculated in the CRF layer according to the score of each label and the transfer score between different labels; based on the comprehensive score, the CRF layer outputs a label path as the final result of key entity labeling; wherein the label path has the largest comprehensive score among all possible label sequences.

[0029] S3. Extract the transformer fault operation and maintenance related content from the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph.

[0030] Specifically, the content related to transformer fault operation and maintenance is extracted from the transformer operation and maintenance knowledge graph to obtain a causal graph of the fault operation and maintenance knowledge structure, including: based on the transformer operation and maintenance knowledge graph, query and search are performed according to preset search statements related to transformer fault operation and maintenance to obtain a causal graph of the fault operation and maintenance knowledge structure.

[0031] S4. Define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein, the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, calculate the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph, adjust the transformer operation and maintenance knowledge graph based on the causal relationship effect, and obtain the causal relationship enhanced knowledge graph.

[0032] Specifically, the conditional probability relational expression is used to represent the conditional probability of the association relationship between entities after intervention adjustment; the conditional probability relational expression is expressed as follows: P(S|do(R))=

[0033] Among them, P(S|do(R)) represents the conditional probability of S occurring after the intervention on node R; represents the conditional probability of S occurring given R and t; is the probability of the latent variable or hidden variable.

[0034] The intervention adjustment is specifically as follows: determining the nodes to be intervened and adjusted in the fault operation and maintenance knowledge structure causal graph; including nodes R, S and T; T is an intermediate node, which can send directional connections to R and S at the same time; using the do operator to implement do intervention to eliminate the correlation between nodes R and T.

[0035] Among the options, we first build a knowledge graph. By deeply mining the information in materials such as transformer fault handling texts and industrial maintenance documents, we use natural language processing technology and deep learning algorithms to accurately identify and extract relationships of key entities such as fault descriptions, causes, phenomena, detection methods, detection results, and operation and maintenance recommendations. These structured knowledge are organized and stored in the Neo4j graph database to form a comprehensive, in-depth, and easy-to-query transformer fault operation and maintenance knowledge graph.

[0036] Among the options, causal relationship analysis is also enhanced. By introducing causal science theory, the causal relationship in the knowledge graph is significantly enhanced, the implicit causal semantic relationship between nodes in the transformer fault operation and maintenance causal event knowledge graph is completed, and fault operation and maintenance causal knowledge with more stable causal effects is formed. This not only improves the accuracy and robustness of the causal relationship in the knowledge graph, but also enables operation and maintenance personnel to more intuitively understand the complex relationship between fault phenomena and causes, thereby achieving in-depth tracking and efficient resolution of the root causes of the fault. Through the knowledge graph enhanced by causal relationships, operation and maintenance personnel can quickly locate the source of the fault and formulate targeted maintenance plans, which greatly shortens the fault handling time, reduces operation and maintenance costs, and improves the overall stability and reliability of the power system.

[0037] The above solution has strong adaptability and scalability, and can be customized and optimized according to different transformer models, fault types and data characteristics. At the same time, with the continuous advancement of technology and the continuous accumulation of data, the solution can also be continuously upgraded and expanded to adapt to more complex and diverse application scenarios. This flexibility and scalability make this technical solution have broad application prospects and promotion value.

[0038] In an optional solution, a causal relationship enhanced knowledge graph construction method for transformer operation and maintenance is also provided, including transformer operation and maintenance knowledge graph construction and causal relationship enhanced knowledge graph. The following is a detailed introduction: (1) Construction of transformer operation and maintenance knowledge graph ① Raw data preprocessing Since the relevant texts of transformer fault handling are mainly recorded manually, which are often accompanied by unstructured characteristics and noise interference, a preprocessing step is required to improve the quality of subsequent data processing and analysis. This process mainly includes two steps: Stop word cleaning: Given that power fault handling texts are often interspersed with noise words that are of no substantial help to the named entity recognition task, such as specific English abbreviations and various non-text symbols, a special stop word list is constructed through detailed analysis to filter out these invalid information and purify the text content.

[0039] Text annotation: To facilitate subsequent processing, the BIO annotation system is used to carefully annotate the fault handling text. Among them, the B label is used to identify the starting position of the entity (such as fault location, fault type, etc.), the I label is used to mark the internal continuation of the entity to the end part, and the O label represents non-entity text. Taking "220kV×××Substation #1 Transformer Flange Oil Leakage" as an example, the annotation result will be: "O, O, O, O, O, O, O, O, O, O, O, O, O, O, B-fault location, I-fault location, I-fault location, B-fault type, I-fault type". This annotation method helps to more accurately identify and process key information in the text.

[0040] ② Named entity recognition based on improved ALBERT-BiLSTM-CRF After the fault handling text annotated by manual BIO is segmented, these segmented words are converted into a vector sequence x that can be understood by the neural network through the Embedding layer. Subsequently, the vector sequence is sent to the ALBERT pre-trained model to extract the high-level feature vector representation of the input text. Next, the feature sequence output from the ALBERT layer is further input to the BiLSTM layer, which is designed to capture the contextual language features of each word in the text and generate a vector sequence that incorporates contextual information.

[0041] After obtaining the output vectors of ALBERT and BiLSTM, a feature concatenation layer effectively fuses the vector outputs of these two sources according to formula (1). The fused result is processed by a fully connected layer and converted into a score sequence for each entity, which is then input into the CRF (conditional random field) layer.

[0042] (1) At the CRF layer, the system calculates a comprehensive score based on the score P of each label and the transfer score A between different labels using formula (2). Finally, the CRF layer outputs a label path that has the largest comprehensive score among all possible label sequences, thereby achieving accurate labeling of entities in the fault handling text.

[0043] (2) ③Relation extraction based on ResNet-PCNN-ATT First, each word in the sentence is encoded through the pre-trained Bert algorithm model, and the relative position distance between the entity pairs is spliced ​​and converted into a vector as the input of the model; then, the deep residual learning (ResNet) and the segmented convolutional neural network (PCNN) are combined to extract effective semantic features to solve the noise error caused by incorrect labeling. Finally, the multi-level attention mechanism is used to calculate the correlation between the corresponding entity and the context word to better increase the weight of the keyword to improve the accuracy of the model relationship extraction. After entity recognition and relationship extraction, the transformer fault operation and maintenance knowledge triple is finally extracted, stored in the graph database and visualized as a transformer fault operation and maintenance event map.

[0044] (2) Construct a causal relationship enhanced knowledge graph for root cause analysis of transformer failure operation and maintenance.

[0045] ① Structural causal diagram of transformer fault knowledge network Although the transformer fault operation and maintenance event graph contains rich causal relationship data, the complex information structure still lacks clear interrelated semantic information between transformer components, fault phenomena, fault causes and fault response measures. In order to deeply explore the potential causal semantic relationship chain between graph nodes, based on the insights provided by Pearl causal theory, a causal graph of fault operation and maintenance knowledge structure is constructed to strengthen the logic and causal expression between nodes.

[0046] Specifically, the graph database query statement is used to query the causal relationship network between transformer components, fault phenomena, fault causes, fault response measures, and detection results in the transformer fault operation and maintenance event map, and obtain the fault operation and maintenance knowledge structure causal graph. The detection result is the result obtained by detecting the transformer related information; the detection result may indicate a fault or not.

[0047] ②Causal effect estimation learning model based on ISPN structural causal graph model This section mainly defines the conditional probability formula and introduces the method of intervention adjustment.

[0048] Develop a strategy to intervene and adjust the causal association knowledge graph model through a backdoor adjustment mechanism. The specific strategy is as follows: Figure 3 As shown in the figure. There is a backdoor path between node variables R and S, that is, R←T→S, which means that there is an intermediary node T that sends directional connections to both R and S. In other words, this is a path connecting R and S, but its direction does not start from R, but indirectly connects the two through T. In the intervention adjustment strategy, the do operator is used to represent the intervention behavior, which essentially cuts off all relationship links directly pointing to the target node variable. After the do intervention is implemented, the correlation between nodes R and T is eliminated, and the two become independent of each other.

[0049] For example, the expression P(S=s|do(R=r)) represents the conditional probability that S takes the value of s when the value of R is set to r under the intervention operation. If node T satisfies the backdoor adjustment condition for the (R,S) pair, then under this intervention framework, the probability of the causal effect of R on S, P(S|do(R)), can be defined as follows: P(S|do(R))= = = =

[0050] Among them, P(S|do(R)) is the conditional probability of S occurring after intervening in R, that is, forcing R to occur. is the conditional probability of S occurring given R and t. It is a latent variable or hidden variable that may affect S and R. is the probability of S occurring when R occurs without intervention. is the probability of t occurring given R occurs; is the probability of a latent variable or hidden variable. In the above formula, the subscript m represents the intermediate process value.

[0051] Each item on the right side of the above formula is based on the original probability distribution before the do intervention is executed. The right side of the second equal sign is derived from Bayes' rule, while the third equal sign is based on the independence assumption of each variable under specific conditions. Although traditional causal Bayesian networks are widely used in causal reasoning, they have scalability challenges when dealing with complex causal problems. However, by applying the backdoor adjustment formula, intervention calculations in the causal association knowledge graph can be effectively performed using only the fault data observed during the transformer operation and maintenance process.

[0052] ③Causal relationship enhancement learning method based on intervention SPN (ISPN) The characteristics of the structural causal graph model are distributed and organized according to the three-level causal hierarchy (Pearl CausalHierarchy, PCH) proposed by Pearl. The three-level causal hierarchy consists of three levels: L1 relevance, L2 intervention, and L3 counterfactuals, where the latter two levels are directly related to the process of causal inference. The distribution of arbitrary interventions on causal knowledge is simulated (the second level of PCH). For any causal knowledge intervention query Q at the L2 level, by performing the intervention operation (denoted as the do operation), the conditional probability P(S|do(R)) of the result S can be calculated under the given intervention do(R).

[0053] Specifically, the first level (L1) of PCH calculates the statistical probability P(x) of the data based on the structural causal graph model, and these calculation results are then used as input to support the estimation of the causal relationship effect P(S|do(R)) based on the structural causal graph model of ISPN in the second level. The probability statistic P(S|R) is learned from the transformer fault data containing causal relationships using the Bayesian rule, and then the conditional probability P(S|do(R)) of the causal association knowledge is indirectly derived by performing intervention and product network calculations. Among them, P(S|do(R)) is the result of the causal association effect estimation, which is directly given by the calculation value of ISPN (intervention and product network) at the root node n1. Through the estimation and learning of the association effect between the causal knowledge of equipment failures, the relationship between the causal knowledge extracted from the knowledge graph forms a strong and weak hierarchical division, thereby generating a more robust causal event knowledge graph. By completing the implicit causal semantic relationship between the nodes of the knowledge graph in the field of transformer fault operation and maintenance, and utilizing the fault operation and maintenance causal knowledge with more stable causal effects, a more reliable source of knowledge is provided for operation and maintenance, providing important basic support for the research on root cause analysis tasks of fault operation and maintenance.

[0054] Example 2 like Figure 4 As shown, based on the same inventive concept as the above embodiment, the present invention also provides a causal relationship enhanced knowledge graph construction device for transformer operation and inspection, which is characterized by comprising: A data acquisition module is used to obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; A graph construction module is used to identify and process the fault handling text with key entities identified based on a pre-trained ALBERT-BiLSTM-CRF model to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, a pre-trained ResNet-PCNN-ATT model is used to extract relations from the fault handling text to obtain transformer fault operation and maintenance knowledge triples, and the transformer fault operation and maintenance knowledge triples are stored in a graph database to obtain a transformer operation and maintenance knowledge graph; The data extraction module is used to extract the transformer fault operation and maintenance related content in the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; A graph enhancement module is used to define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph is calculated, and based on the causal relationship effect, the transformer operation and maintenance knowledge graph is adjusted to obtain a causal relationship enhanced knowledge graph.

[0055] Example 3 like Figure 5 As shown, the present invention also provides an electronic device 100 for implementing a method for constructing a causal relationship enhanced knowledge graph for transformer operation and inspection; The electronic device 100 includes a memory 101 , at least one processor 102 , a computer program 103 stored in the memory 101 and executable on the at least one processor 102 , and at least one communication bus 104 .

[0056] The memory 101 can be used to store a computer program 103. The processor 102 implements the steps of a method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance in Example 1 by running or executing the computer program stored in the memory 101 and calling data stored in the memory 101.

[0057] The memory 101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0058] At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0059] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for constructing a causal relationship enhanced knowledge graph for transformer operation and inspection, and the processor 102 can execute the plurality of instructions to implement: Obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; Based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is recognized and processed to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, the pre-trained ResNet-PCNN-ATT model is used to extract the relationship of the fault handling text to obtain the transformer fault operation and maintenance knowledge triple, and the transformer fault operation and maintenance knowledge triple is stored in the graph database to obtain the transformer operation and maintenance knowledge graph; Extract the transformer fault operation and maintenance related content from the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; Define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, calculate the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph, adjust the transformer operation and maintenance knowledge graph based on the causal relationship effect, and obtain the causal relationship enhanced knowledge graph.

[0060] Example 4 If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0061] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0062] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0063] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0065] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance, characterized in that: These include: Obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; Based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is recognized and processed to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, the pre-trained ResNet-PCNN-ATT model is used to extract the relationship of the fault handling text to obtain the transformer fault operation and maintenance knowledge triple, and the transformer fault operation and maintenance knowledge triple is stored in the graph database to obtain the transformer operation and maintenance knowledge graph; Extract the transformer fault operation and maintenance related content from the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; Define a conditional probability relationship between nodes in a causal graph of a fault operation and maintenance knowledge structure; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; Based on the intervention sum-product network and the conditional probability relationship, the causal effect between the nodes in the causal graph of the fault operation and maintenance knowledge structure is calculated, and the transformer operation and maintenance knowledge graph is adjusted based on the causal effect to obtain the causal enhanced knowledge graph.

2. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 1 is characterized in that: Based on the pre-trained ALBERT-BiLSTM-CRF model, the fault handling text with key entities identified is identified and processed to obtain the final result of key entity labeling of the fault handling text: The ALBERT-BiLSTM-CRF model includes an ALBERT layer, a BiLSTM layer and a CRF layer; The ALBERT layer extracts a high-level feature vector representation of the text based on the input initial vector sequence; Input the high-level feature vector representation of the text into the BiLSTM layer, which captures the contextual language features of each word in the text and outputs a vector sequence that incorporates the contextual information; The high-level feature vector representation output by the ALBERT layer and the vector sequence output by the BiLSTM layer that incorporates context information are fused in the feature concatenation layer, and the fusion result is processed by the fully connected layer to obtain a score sequence for each key entity; wherein the score sequence is composed of the respective scores of all labels corresponding to the key entity; The score sequence of each key entity is input into the CRF layer, and a comprehensive score is calculated in the CRF layer according to the score of each label and the transfer score between different labels; based on the comprehensive score, the CRF layer outputs a label path as the final result of key entity labeling; wherein the label path has the largest comprehensive score among all possible label sequences.

3. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 1 is characterized in that: The conditional probability relational expression is used to express the conditional probability of the association relationship between entities after intervention adjustment; wherein the intervention adjustment is specifically: Determine the nodes to be intervened and adjusted in the fault operation and maintenance knowledge structure causal graph; which includes nodes R, S and T; T is an intermediary node that can send directional connections to R and S at the same time; The do operator is used to implement do intervention to eliminate the association between nodes R and T.

4. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 1 is characterized in that: Before annotating the fault handling text, the following steps are also included: Based on a pre-built stop word list, noise words in the transformer fault handling text are filtered out.

5. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 1 is characterized in that: Use BIO annotation system to annotate fault handling text; The fault handling text with key entities identified is segmented, and the segmented words of the fault handling text are converted into an initial vector sequence through the Embedding layer.

6. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 1 is characterized in that: The conditional probability relationship is expressed as follows: P(S|do(R))= Among them, P(S|do(R)) represents the conditional probability of S occurring after the intervention on node R; represents the conditional probability of S occurring given R and t; is the probability of the latent variable or hidden variable.

7. The method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance according to claim 6 is characterized in that: Extract the transformer fault operation and maintenance related content in the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph, including: based on the transformer operation and maintenance knowledge graph, perform query and search according to preset search statements related to transformer fault operation and maintenance to obtain the fault operation and maintenance knowledge structure causal graph.

8. A device for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance, characterized in that: include: A data acquisition module is used to obtain the fault handling text of the transformer, annotate the fault handling text, and identify key entities; A graph construction module is used to identify and process the fault handling text with key entities identified based on a pre-trained ALBERT-BiLSTM-CRF model to obtain the final result of key entity labeling of the fault handling text; based on the final result of key entity labeling, a pre-trained ResNet-PCNN-ATT model is used to extract relations from the fault handling text to obtain transformer fault operation and maintenance knowledge triples, and the transformer fault operation and maintenance knowledge triples are stored in a graph database to obtain a transformer operation and maintenance knowledge graph; The data extraction module is used to extract the transformer fault operation and maintenance related content in the transformer operation and maintenance knowledge graph to obtain the fault operation and maintenance knowledge structure causal graph; A graph enhancement module is used to define the conditional probability relationship between nodes in the fault operation and maintenance knowledge structure causal graph; wherein the conditional probability relationship is used to represent the conditional probability of the association relationship between entities after intervention adjustment; based on the intervention formula and product network and the conditional probability relationship, the causal relationship effect between each node in the fault operation and maintenance knowledge structure causal graph is calculated, and based on the causal relationship effect, the transformer operation and maintenance knowledge graph is adjusted to obtain a causal relationship enhanced knowledge graph.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, it implements the method for constructing a causal relationship enhanced knowledge graph for transformer operation and maintenance as described in any one of claims 1 to 7.

Images