Stator partial discharge detection report generation method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a stator partial discharge detection report generation method and device, computer equipment, a computer readable storage medium and a computer program product, relates to the technical field of electric power, and can improve the stator partial discharge detection report generation efficiency. The method comprises the following steps: establishing a knowledge graph of stator partial discharge; obtaining a discharge characteristic corresponding to the real-time partial discharge data, comparing the discharge characteristic with entity characteristics of a plurality of entities of the knowledge graph, determining an entity matched with the discharge characteristic according to a comparison result, and determining a target fault mode according to the entity matched with the discharge characteristic; determining a decision support path comprising a fault causal chain and a fault maintenance strategy based on the target fault mode and an association relationship in the knowledge graph; and performing natural language description on the decision support path through a large language model to obtain a fault detection report of the partial discharge of the stator.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for generating a detection report of stator partial discharge. Background Art

[0002] With the development of intelligent monitoring technology for power equipment conditions, online partial discharge monitoring technology has emerged, combining multi-sensor data acquisition with machine learning algorithms. In the traditional process, high-frequency current transformers and ultra-high-frequency sensors collect stator discharge signals and generator operating parameters. Signal processing techniques such as wavelet transforms are then used to extract discharge characteristics. Machine learning models such as support vector machines are then used to identify the discharge type and output a preliminary report containing statistical data and simple alarms. However, the generated reports are simplistic, making it difficult to efficiently obtain a comprehensive fault analysis report that can directly assist in operational and maintenance decision-making. Summary of the Invention

[0003] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for generating a stator partial discharge detection report to address the above technical problems.

[0004] In a first aspect, the present application provides a method for generating a stator partial discharge detection report, comprising:

[0005] Establishing a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and association relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity;

[0006] Obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of a plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features;

[0007] Determining a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0008] The decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report.

[0009] In one embodiment, the discharge feature is compared with entity features of multiple entities in the knowledge graph, entities matching the discharge feature are determined based on the comparison results, and target fault modes are determined based on the entities matching the discharge feature, including:

[0010] converting the discharge features and the entity features of each of the entities into a unified feature vector space, and combining the converted discharge features with the converted entity features to obtain a joint feature;

[0011] In the feature vector space, calculating the association score of each of the joint features;

[0012] Determining the degree of association between each entity and the discharge feature relative to other entities based on the association score of each joint feature, and obtaining entities whose association degree meets a preset condition;

[0013] If the entity is the device parameter entity, determining a target fault mode that matches the discharge feature according to an association relationship related to the device parameter entity; and

[0014] If the entity is the fault mode entity, the fault mode indicated by the fault mode entity is used as the target fault mode.

[0015] In one embodiment, each of the entity features is obtained by the following steps:

[0016] Obtaining entity vectors of each entity and relationship vectors of association relationships between each entity;

[0017] For each group of the association relationship, the head entity and the tail entity associated with the association relationship, obtaining a sum of the association relationship vector corresponding to the association relationship and the head entity vector corresponding to the head entity, and determining a difference between an entity vector obtained by converting the head entity vector through the association relationship and the tail entity vector based on a difference between the sum and the tail entity vector corresponding to the tail entity;

[0018] A loss value is determined based on the multiple differences, and each entity vector and each relationship vector is iteratively adjusted based on the loss value until a preset iteration end condition is met, and each entity feature is obtained based on the current entity vector.

[0019] In one embodiment, after acquiring the real-time partial discharge data and extracting the discharge characteristics of the real-time partial discharge data, the method further includes:

[0020] Inputting the discharge characteristics into a pre-trained risk level determination model; the risk level determination model includes a probability dependency relationship between each risk level and each discharge characteristic; the probability dependency relationship indicates the conditional probability of the discharge characteristics at different risk levels, and the current probability of occurrence of each risk level;

[0021] The risk level determination model outputs a current risk level of the stator partial discharge according to the probability dependency and the discharge characteristics.

[0022] In one embodiment, the current risk level is determined according to a risk quantification parameter, including:

[0023] Based on preset weights, weighting the standardized discharge characteristics and the real-time partial discharge data corresponding to the standardized discharge characteristics to determine a risk score at the current risk level;

[0024] The degree of change of the risk score relative to the preset benchmark risk score is determined, and the risk quantification parameter corresponding to the current risk level is determined based on the degree of change.

[0025] In one embodiment, before the step of performing a natural language description of the decision support path using a large language model to obtain a stator partial discharge fault detection report, the step further includes:

[0026] Based on the discharge characteristics and the decision support path, determining keywords corresponding to fields of a preset structured template, and generating a structured fault detection report based on the keywords;

[0027] Before the large language model is used to describe the decision support path in natural language to obtain a stator partial discharge fault detection report, the method further includes:

[0028] A translation of the fault detection report into a target language is obtained, and the fault detection report is displayed based on the translation.

[0029] In a second aspect, the present application further provides a device for generating a detection report of stator partial discharge, comprising:

[0030] A partial discharge knowledge graph establishment module is used to establish a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and the association relationships between the multiple entities; the multiple entities include fault mode entities, equipment parameter entities, maintenance instance entities, and industry standard entities;

[0031] a target fault mode determination module, configured to obtain discharge features corresponding to real-time partial discharge data, compare the discharge features with entity features of the plurality of entities in the knowledge graph, determine entities matching the discharge features based on the comparison results, and determine a target fault mode based on the entities matching the discharge features;

[0032] A decision support path determination module, configured to determine a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0033] The fault detection report generation module is used to describe the decision support path in natural language through a large language model to obtain a stator partial discharge fault detection report.

[0034] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0035] Establishing a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and association relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity;

[0036] Obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of a plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features;

[0037] Determining a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0038] The decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0040] Establishing a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and association relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity;

[0041] Obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of a plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features;

[0042] Determining a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0043] The decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report.

[0044] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0045] Establishing a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and association relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity;

[0046] Obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of a plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features;

[0047] Determining a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0048] The decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report.

[0049] The above-mentioned stator partial discharge detection report generation method, device, computer equipment, computer-readable storage medium and computer program product establish a knowledge graph of stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and the association relationships between each entity; the multiple entities include fault mode entities, equipment parameter entities, maintenance instance entities and industry standard entities; obtain discharge characteristics corresponding to real-time partial discharge data, compare the discharge characteristics with the entity characteristics of multiple entities in the knowledge graph, determine the entities matching the discharge characteristics based on the comparison results, and determine the target fault mode based on the entities matching the discharge characteristics; based on the target fault mode and the association relationship in the knowledge graph, determine a decision support path including a fault causal chain and a fault repair strategy; describe the decision support path in natural language using a large language model to obtain a stator partial discharge fault detection report. In this application, a solid knowledge foundation for fault analysis is established by constructing a knowledge graph containing fault modes and their associations. Subsequently, the real-time extracted discharge features are intelligently matched with the entities in the graph to accurately identify the target fault mode that matches the current operating conditions. Furthermore, the inherent association logic of the knowledge graph is further utilized to deduce a decision-making support path, revealing the fault evolution mechanism or recommended response strategy. Finally, a large language model is used to automatically convert the fault handling path into an easy-to-understand natural language detection report, completing the closed loop from raw data to intelligent detection and improving the quality and report generation efficiency of stator partial discharge detection reports. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 1 is a flow chart of a method for generating a stator partial discharge detection report in one embodiment;

[0052] Figure 2 1 is a flow chart of a method for generating a stator partial discharge detection report in another embodiment;

[0053] Figure 3 1 is a flow chart of a method for generating a stator partial discharge detection report in yet another embodiment;

[0054] Figure 4 1 is a structural block diagram of a device for generating a detection report of stator partial discharge in one embodiment;

[0055] Figure 5FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0057] It should be noted that the terms "first", "second", etc. used in this application may be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "including" and "having" used in this application and any variations thereof are intended to cover non-exclusive inclusions. The term "plurality" used in this application refers to two or more. The term "and / or" used in this application refers to one of the solutions or any combination of multiple solutions.

[0058] In one embodiment, Figure 1 As shown, a method for generating a stator partial discharge detection report is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] Step S102: establishing a knowledge graph of stator partial discharge.

[0060] The knowledge graph includes, but is not limited to, multiple entities related to stator partial discharge faults and the relationships between these entities. The multiple entities include, but are not limited to, failure mode entities, equipment parameter entities, maintenance instance entities, and industry standard entities.

[0061] A knowledge graph is a semantic network based on a graph data structure. It stores knowledge in the form of nodes and edges. Nodes represent real-world entities—anything distinguishable from other things, such as a specific failure mode, a device parameter, or an industry standard. Edges represent objective relationships between entities. For example, the failure mode entity "insulation aging" can have a failure consequence relationship with the device parameter entity "overheated winding temperature." The basic building block of a knowledge graph is typically a triple consisting of a head entity, a relationship, and a tail entity.

[0062] Exemplarily, the knowledge graph can be constructed using expert preset rules. For example, one or more experts in the field of generator fault diagnosis first jointly define the model layer of the knowledge graph, including determining the entity types (such as the aforementioned fault mode entities, equipment parameter entities, maintenance instance entities, and industry standard entities) and the types of associations (such as fault results, associated environmental factors, repair plans, and standard violations) that need to be included in the graph. Subsequently, based on professional knowledge and existing documentation, a team of experts or knowledge engineers uses semi-automated tools to enter specific knowledge instances (such as fault mode: discharge in stator slots; equipment parameter: rated voltage 35kV) into the knowledge graph database in the form of triplets, forming the initial knowledge graph.

[0063] Step S104: obtain the discharge characteristics corresponding to the real-time partial discharge data, compare the discharge characteristics with the entity characteristics of multiple entities in the knowledge graph, determine the entities matching the discharge characteristics based on the comparison results, and determine the target fault mode based on the entities matching the discharge characteristics.

[0064] Discharge characteristics can be one or more sets of quantitative indicators extracted from the raw, continuous, real-time partial discharge data waveform collected by the sensor to facilitate quantitative analysis and comparison. These characteristics can be numerical values, vectors, or patterns derived through signal processing algorithms (such as one or more fast Fourier transforms (FFTs), wavelet transforms, and Hilbert transforms) and statistical analysis. Exemplary discharge characteristics include, but are not limited to, discharge pulse amplitude, pulse count, signal rise time, energy integration, statistical characteristics of the discharge phase distribution spectrum (such as skewness and kurtosis), and the spectral energy distribution of the discharge signal.

[0065] Entity features can be a set of descriptive attributes or values associated with each entity node. These features define the specific characteristics of the entity in the domain knowledge. In the embodiments of the present application, entity features can be symbolic descriptions (such as text labels), values, or more complex structured data (such as a typical feature template). For example, for a fault mode entity (such as discharge in the stator slot), its entity features can be a set of typical baseline values or ranges of discharge features associated with the fault mode; for an equipment parameter entity (such as generator rated voltage), its entity features are the specific values or categories of the parameter (for example, 35kV).

[0066] For example, the terminal continuously receives raw electrical signal waveforms from a high-frequency current transformer (HFCT) or ultra-high-frequency (UHF) sensor coupled to the generator stator. The signal processing unit within the terminal processes this waveform data in real time. For example, it uses wavelet packet decomposition to extract energy characteristics of different frequency bands, or uses phase analysis techniques to construct a discharge phase distribution map. From this map, statistical features such as shape factor and asymmetry are calculated, resulting in a real-time discharge feature vector containing multiple dimensional features.

[0067] Optionally, the entity feature of each fault mode entity in the knowledge graph can be a standardized fault feature template. This template can be a multidimensional vector, with each dimension corresponding to a typical discharge feature value. After the terminal obtains the real-time discharge feature vector, it calculates the distance or similarity between the real-time vector and each fault feature template stored in the knowledge graph. This distance can be Euclidean distance or Mahalanobis distance, or the similarity can be cosine similarity. The terminal identifies the fault mode entity corresponding to the fault feature template with the smallest distance (or highest similarity) as a matching entity and sets it as the target fault mode.

[0068] Optionally, the matching process is not limited to directly matching fault mode entities. If the terminal determines that a device parameter entity (for example, entity feature: current winding temperature > 30°C) closely matches the real-time discharge feature, the terminal will further query the knowledge graph to find a fault mode entity (for example, insulation thermal aging) that has a causal or strong correlation relationship with the device parameter entity and use it as the inferred target fault mode.

[0069] Step S106: Based on the target failure mode and the association relationship in the knowledge graph, a decision support path including a failure causal chain and a failure repair strategy is determined.

[0070] The decision support path can be a subgraph extracted or inferred from the main knowledge graph or a structured data object (such as JSON or XML). This path uses the target fault mode entity as the core node and includes a series of upstream (fault cause) and downstream (fault repair strategy) entities and relationship chains connected to this core node through specific associations. This provides operations and maintenance personnel with a comprehensive view of the fault, greatly assisting in fault confirmation, root cause analysis, and repair decision-making. In one embodiment, for example, the decision support could be: increased discharge pulse density → [fault consequence] → insulation aging → [violation] → industry standard A → [repair plan] → replacement of insulation material.

[0071] Exemplarily, the terminal uses a series of pre-set query templates in structured graph query languages (such as SPARQL or Cypher). Upon receiving the target fault mode, the terminal populates the corresponding query template with the fault mode identifier. For example, a template can be designed to query all entity paths connected to the [target fault mode] through the fault result relationship chain, and simultaneously query all entities connected to the [target fault mode] through the repair solution relationship. The terminal then executes this fully populated query statement on the knowledge graph database. The query result returned by the database is itself a structured dataset containing all eligible nodes and path information, which the terminal directly organizes into a decision support path.

[0072] In a specific embodiment, the terminal starts from the insulation aging entity node in the knowledge graph and performs a reverse causal tracing. It can find a fault-result relationship pointing to insulation aging and discover that its upstream entity is increased discharge pulse density. This chain from cause / phenomenon to result (increased discharge pulse density → [fault result] → insulation aging) constitutes the fault causal chain portion of the decision support path. This reveals to the user the direct manifestation of the fault result, insulation aging.

[0073] In some embodiments, the terminal can also start from the insulation aging entity node and conduct a forward association exploration. First, a violation relationship is found, pointing to an industry standard entity, namely industry standard A. This indicates that the current insulation aging condition may no longer meet the requirements of the industry standard. Then, the terminal can start from the insulation aging node or the industry standard A node and find a repair solution relationship, which points to a maintenance instance entity or a specific maintenance action, which can be replacing insulation materials. This chain from fault to standard to solution (insulation aging → [violation] → industry standard A → [repair solution] → replacement of insulation materials) constitutes the fault repair strategy part of the decision support path.

[0074] Step S108 : Using a large language model to describe the decision support path in natural language to obtain a stator partial discharge fault detection report.

[0075] Among them, the large language model can be a model based on deep learning, such as a neural network model with a Transformer architecture. It is characterized by having a large number of model parameters (usually in the billions or even more), and through pre-training on a very large text corpus, it has mastered a wealth of language rules, factual knowledge, and certain reasoning capabilities. In this application, examples that can be used include but are not limited to models based on architectures such as GPT (Generative Pre-trained Transformer) or BART / T5, which have been fine-tuned or have contextual learning capabilities. Through their natural language generation (NLG) capabilities, they understand the logic, causal, and policy information contained in decision support paths input in a specific format (such as structured text, JSON, or vectorized representation), and generate natural language text that conforms to grammatical specifications and has clear logic based on this information.

[0076] For example, the terminal first serializes the entire decision support path, converting it into a detailed, descriptive text string or a JSON object. The terminal then uses this serialized string as context to construct a complete instruction prompt, such as: "You are a generator fault diagnosis expert. Please write a detailed fault detection report based on the following structured information: [insert serialized decision support path here]." This prompt is directly sent to a general-purpose large language model. The large language model understands the entire input information at once and directly generates a complete, all-inclusive fault detection report text in an end-to-end manner.

[0077] In this example, a solid knowledge foundation for fault analysis is established by constructing a knowledge graph encompassing fault modes and their associations. Subsequently, the real-time extracted discharge features are intelligently matched with entities in this graph to accurately identify the target fault mode that matches the current operating conditions. Furthermore, the inherent association logic of the knowledge graph is leveraged to deduce a decision-making support path, revealing the fault evolution mechanism and recommended response strategies. Finally, a large language model is used to automatically convert the fault handling path into an easy-to-understand natural language detection report, completing the closed loop from raw data to intelligent detection and improving the quality and efficiency of stator partial discharge detection reports.

[0078] In an exemplary embodiment, the discharge feature is compared with entity features of multiple entities in the knowledge graph, entities matching the discharge feature are determined based on the comparison results, and target fault modes are determined based on the entities matching the discharge feature.

[0079] The discharge characteristics and the entity characteristics of each entity are converted into a unified feature vector space, and the converted discharge characteristics are combined with the converted entity characteristics to obtain a joint characteristic; in the feature vector space, the correlation score of each joint characteristic is calculated; based on the correlation score of each joint characteristic, the degree of correlation between each entity and the discharge characteristics relative to other entities is determined, and entities whose correlation degree meets preset conditions are obtained; if the entity is a device parameter entity, the target fault mode that matches the discharge characteristics is determined based on the association relationship related to the device parameter entity; and if the entity is a fault mode entity, the fault mode indicated by the fault mode entity is used as the target fault mode.

[0080] Among them, the feature vector space can be a multi-dimensional mathematical space, in which each dimension corresponds to a specific quantitative feature, so that information with different physical meanings and data types (such as real-time discharge features from sensors and entity features from knowledge graphs) can be converted into a unified format (i.e., feature vectors), thereby supporting standardized quantitative comparisons and algebraic operations.

[0081] A joint feature can be formed by combining two or more independent feature vectors to form a new feature vector. An alternative combination method is vector concatenation, which connects two vectors end-to-end to form a new vector with a longer dimension that contains all the information of the two original vectors.

[0082] The relevance score can be an unnormalized scalar value, an intermediate result in the relevance calculation process, reflecting the degree of match between a knowledge graph entity and the current real-time discharge feature. A larger value generally indicates a higher degree of match. The relevance degree can be a normalized value, typically ranging from 0 to 1, with the sum of all relevance degrees for a comparison set (e.g., all related neighbor entities) being 1. It is used to distinguish the degree of similarity between different entities relative to the discharge feature.

[0083] Exemplarily, the terminal converts the discharge features and the entity features of each entity into a unified feature vector space, and concatenates the converted discharge features with the converted entity features to obtain joint features, and applies a set of optimized parameterized mathematical operations to each joint feature to calculate its association score. After obtaining the original association scores of all entities paired with the current discharge features, the terminal performs a normalization step, for example, converting all the original scores through a function that can highlight the relative size (such as the Softmax function) to generate a set of standardized association weights. The terminal then obtains one or more entities with the highest association or that meet a preset threshold as the matching result. In a specific embodiment, the degree of association between the discharge feature vector and any entity feature vector can be realized by the attention mechanism formula (1) Calculation.

[0084]

[0085] in, is a shared linear transformation matrix used to transform features into a unified feature vector space, is the discharge eigenvector, is the entity feature vector, is a learnable attention parameter vector, is the set of neighbor nodes, are other neighbor nodes, is the entity feature vector of other neighbor nodes. The operation as a whole constitutes the computation of the joint features, and the result is a raw correlation score, which is the sum of the raw correlation scores calculated for all neighbor entities. The entire fractional structure is a Softmax function used for normalization to calculate the final correlation degree of the node relative to all other neighbor nodes.

[0086] After obtaining the entity with the highest degree of association or that meets the preset conditions through any of the above methods, perform the following logical judgment:

[0087] If the entity that meets the conditions is a device parameter entity (for example, a matching winding temperature entity with a status of "too high"), the terminal will further query the knowledge graph to find a failure mode entity with a preset association (such as a failure result or high correlation) with the device parameter entity. The terminal will then return an insulation thermal aging entity. The terminal will then use this inferred insulation thermal aging as the final target failure mode.

[0088] If the entity that meets the conditions is itself a fault mode entity (for example, directly matched to the end corona discharge entity), the terminal directly uses the fault mode indicated by the entity (i.e., end corona discharge) as the target fault mode of this detection.

[0089] In this embodiment, through a matching method based on feature vector space and combined with an attention mechanism, it is possible to capture deeper, nonlinear associations between features, and dynamically learn the importance of different features and neighboring entities, thereby significantly improving the accuracy and robustness of fault pattern recognition.

[0090] In an exemplary embodiment, each entity feature is obtained by the following steps:

[0091] Obtain the entity vector of each entity and the relationship vector of the association relationship between each entity; for each group of association relationships, the head entity and the tail entity associated with the association relationship, obtain the sum of the association relationship vector corresponding to the association relationship and the head entity vector corresponding to the head entity, and determine the difference between the entity vector obtained by converting the head entity vector through the association relationship and the tail entity vector based on the difference between the summation result and the tail entity vector corresponding to the tail entity; determine the loss value based on multiple differences, iteratively adjust each entity vector and each relationship vector based on the loss value until the preset iteration end condition is met, and obtain each entity feature based on the current entity vectors.

[0092] Specifically, the terminal traverses all entity and relationship types in the knowledge graph and initializes a vector of a specified dimension for each entity and relationship. For any correct triple in the knowledge graph (e.g., head entity h, relationship r, tail entity t), their corresponding vectors (h, r, t) should approximately satisfy a simple translation relationship in the feature vector space, that is, the head entity vector plus the relationship vector should be approximately equal to the tail entity vector, as shown in Formula (2).

[0093]

[0094] in, Represents the knowledge graph.

[0095] For each correct triple, the terminal calculates the sum of the head entity vector and the relationship vector, and then calculates the difference or distance between the sum and the tail entity vector. This difference or distance represents the difference in the model's prediction of this correct fact. Ideally, this difference should be close to zero. Specifically, a total loss value is calculated using a preset loss function, as shown in formula (3). Then, the terminal uses an optimization algorithm (for example, stochastic gradient descent) to make slight adjustments to each entity vector and relationship vector involved in the calculation based on the size and gradient direction of the loss value, so that the adjusted vector can make the total loss value The process continues until the preset end-of-iteration conditions are met (for example, the loss value no longer decreases significantly, or the preset number of iterations is reached). When the iteration process meets the end conditions, all current entity vectors are considered stable and optimized. They will be solidified as the final entity features for subsequent steps such as fault mode identification.

[0096]

[0097] in, It is actually a preset interval parameter, which the terminal can minimize through optimization algorithm. , and iteratively adjust each entity vector and relationship vector in this process.

[0098] In this embodiment, the symbolic, discrete knowledge graph is converted into a continuous vector space model that can perform mathematical operations. This not only enables efficient real-time reasoning of the knowledge graph, but more importantly, it provides the ability to process unknown faults and predict missing relationships based on semantic similarity, thereby improving detection efficiency.

[0099] In an exemplary embodiment, after acquiring real-time partial discharge data and extracting discharge features of the real-time partial discharge data, the method further includes:

[0100] The discharge characteristics are input into a pre-trained risk level determination model; the risk level determination model includes a probabilistic dependency relationship between each risk level and each discharge characteristic; the probabilistic dependency relationship indicates the conditional probability of the discharge characteristics under different risk levels, as well as the current probability of occurrence of each risk level; the risk level determination model outputs the current risk level of stator partial discharge based on the probabilistic dependency relationship and the discharge characteristics.

[0101] Among them, the pre-trained risk level determination model is a machine learning model based on probabilistic graph theory. Specifically, it can be a Bayesian network, which is used to perform probabilistic reasoning based on the learned knowledge when receiving real-time discharge characteristics to infer the actual risk level.

[0102] The probabilistic dependency relationship is a concrete mathematical expression of the model's internal knowledge. It formally defines the mutual influence and correlation strength between the various variables in the system (risk levels and discharge characteristics). This specifically consists of two parts: the first is the conditional probability, which quantifies the probability of a specific discharge characteristic occurring at a given risk level, reflecting the impact of the risk level on the discharge characteristic; the second is the current probability of occurrence of each risk level (also known as the prior probability in Bayesian theory). This represents the inherent probability of occurrence of each risk level based on historical statistics or expert experience, without any real-time observation data.

[0103] In one specific embodiment, after extracting real-time discharge features, the terminal performs a risk level determination step. The pre-trained risk level determination model is specifically a Bayesian network. The network's topology is designed so that nodes representing the four predefined risk levels of low, medium, high, and emergency serve as parent nodes, while nodes representing individual discharge features extracted from real-time data (e.g., pulse density and skewness in the feature vector F) serve as child nodes. Directed edges from risk level nodes to discharge feature nodes represent the probabilistic dependency between them.

[0104] During the pre-training phase, the terminal analyzes historical data to populate a conditional probability table for each discharge feature node. This table stores the conditional probability of the discharge feature taking different values (such as high, medium, and low) at each specific risk level. It also determines the prior probability for each risk level node.

[0105] During real-time detection, when the terminal obtains a new set of discharge features, it uses the Bayesian network to perform reasoning, calculates the posterior probability of each risk level, and outputs the risk level with the highest posterior probability as the current risk level. Specifically, as shown in Formula (4), the posterior probability of the risk level being high is calculated.

[0106]

[0107] Where F is the discharge eigenvector, is the conditional probability of risk level being high or low, is the probability that the current risk level is high, Indicates the optional risk level, such as low, medium, high, and emergency. is the conditional probability under risk level c, is the probability that the current risk level is c.

[0108] In this embodiment, by using prior knowledge based on historical statistics or expert experience (i.e., the prior probability of the risk level) and real-time evidence from the current operating conditions (i.e., the discharge characteristics), a probabilistic model such as a Bayesian network is used to accurately map the discharge characteristics to a specific risk level, which is used to instruct operation and maintenance personnel to better handle the current risk fault.

[0109] In an exemplary embodiment, the current risk level is determined according to the risk quantification parameter, including:

[0110] Based on preset weights, the standardized discharge characteristics and the real-time partial discharge data corresponding to the standardized discharge characteristics are weighted to determine the risk score under the current risk level; the degree of change in the deviation of the risk score from the preset benchmark risk score is determined, and based on the degree of change, the risk quantification parameter corresponding to the current risk level is determined.

[0111] The risk score can be a comprehensive scalar value obtained by weighted summation of multiple standardized risk influencing factors (including discharge characteristics and equipment status parameters). Its function is to compress and integrate multi-dimensional, fragmented risk information into a single, uniformly assessable value. The preset baseline risk score can be a key threshold pre-calibrated through statistical analysis of historical data. It physically represents the critical point where risk transitions from one magnitude to another and serves as a reference for measuring the current risk score. The risk quantification parameter can be a continuous value, usually normalized to a specific interval (such as 0 to 1). It is not a discrete level label, but rather a quantitative description of the severity of the risk. In some embodiments, it is the degree of membership or confidence of the current state in a risk set.

[0112] Specifically, first, the terminal executes a risk score determination step. This step performs a linear weighted summation of multiple standardized risk factors based on preset weights to determine a risk score. Risk factors exemplarily include standardized discharge characteristics (such as pulse density, skewness, etc.) and standardized equipment status parameters (such as winding temperature, load rate, etc.). Subsequently, the terminal executes a risk quantification parameter determination step. In an exemplary embodiment, this step is implemented through a nonlinear membership function, which is intended to determine the deviation of the risk score from a preset baseline risk score, and based on the degree of the deviation and a sensitivity parameter that controls the degree of change, ultimately determines the risk quantification parameter ( In one embodiment, the calculation process can be represented by the following formula (5):

[0113]

[0114] in, This is the risk score calculated in the previous step. Score the preset baseline risk. To control the sensitivity parameters of the steepness of the curve, both parameters are predetermined by analyzing and calibrating historical data.

[0115] In this embodiment, through the above steps, the fuzzy risk level is represented by quantitative risk parameters, so that operation and maintenance personnel can understand the fault risk more objectively, and the entire system has the ability of early warning and fault prediction, thereby improving the processing efficiency of fault handling.

[0116] In an exemplary embodiment, before obtaining a stator partial discharge fault detection report by using a large language model to describe the decision support path in natural language, the method further includes:

[0117] Based on the discharge characteristics and decision support path, the keywords corresponding to the fields of the preset structured template are determined, and a structured fault detection report is generated based on the keywords.

[0118] The preset structured template can be a predefined report document framework, technically a text file with placeholders, Extensible Markup Language (XML), or JSON object. It provides a unified, standardized format and structure for the final fault detection report, ensuring standardized and consistent report content organization. Keywords can be core information fragments extracted from discharge characteristics and decision support paths that can be accurately filled into the structured template placeholders, such as a specific fault mode name, a quantitative risk level value, or a brief description of the repair measure.

[0119] In a specific embodiment, the terminal parses the determined decision support path, quickly extracts key information such as fault location (LOC) and risk level (RISK), and fills these keywords into a predefined report framework containing placeholders such as {LOC} and {RISK}, thereby quickly generating a structured summary part of the report.

[0120] Optionally, for analytical content that needs to be elaborated in the report, such as a detailed interpretation of the fault causal chain, the terminal will call a pre-trained text generation model based on the Transformer architecture, as shown in Formula (6). This model receives the graph encoding vector of the knowledge graph path as its conditional input and generates text based on the autoregressive principle.

[0121]

[0122] Specifically, the process of generating report paragraphs by this model follows the probabilistic chain rule, that is, Next, predict and generate the next word in the report text one by one ( ), whose generation probability Depends on two core pieces of information, the generated previous content ( ) to ensure smooth writing, and the initial input vector as a factual basis ( ) to ensure content accuracy. By continuously selecting and concatenating the most probable tokens at each step, the model generates logically coherent, detailed natural language paragraphs. Ultimately, the terminal combines the structured information populated by the template with the narrative paragraphs generated by deep learning to produce a complete and appropriately detailed final fault detection report.

[0123] In this embodiment, a deterministic keyword filling method is used for key factual data such as fault name and risk level that cannot be wrong. This can completely avoid the risk of information errors and omissions that may occur during the generation process of large language models, and ensure the accuracy of the core conclusions of the report.

[0124] Accordingly, in an exemplary embodiment, after the decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report, the following steps are further included:

[0125] Obtain the translation content of the fault detection report corresponding to the target language, and display the fault detection report based on the translation content.

[0126] The target language is the human language into which the user wishes to convert the original fault detection report, such as English, German, or Japanese. The choice of target language is often determined by the language preference of the report's final recipient or the language requirements of the work environment. The translated content can be all or part of the text in the original fault detection report, resulting in an equivalent semantic expression in the target language after translation.

[0127] Specifically, the terminal needs to determine the target language, which can be selected by the user through the interface or automatically determined based on a preset regional configuration. After determining the target language, the terminal takes the full text of the generated Chinese fault detection report or key paragraphs as input, and appends explicit translation instructions, such as "Translate the following into [target language name, such as English]: [Chinese report content]. The large language model then outputs the corresponding translation text based on the instructions.

[0128] Optionally, in order to ensure the accuracy and consistency of the translation of professional terms in the field of power equipment and avoid ambiguity that may be caused by general translation models, the terminal can further combine a preset domain terminology library when executing any of the above translation methods. The terminology library contains professional vocabulary in the fields of power engineering, fault diagnosis, etc. and their standard translation methods in different target languages. During the translation process or after the translation is completed, the terminal can use this terminology library to proofread and correct the translation results. For example, the general translation can be translated as aging, and it can be accurately corrected to the professional expression corresponding to insulation degradation or insulation aging in the target language according to the terminology library.

[0129] After obtaining the translation content corresponding to the target language, the terminal finally displays the fault detection report based on the translation content, for example, displaying the translated report text on the user interface, or generating a target language version of the report file for the user to download or view.

[0130] In this embodiment, by combining the domain term library, it is possible to effectively avoid mistranslation or inconsistent translation of professional terms by general translation tools, thereby ensuring the professionalism and rigor of the technical report, which is helpful in avoiding operational errors caused by misunderstandings.

[0131] In an exemplary embodiment, Figure 2 As shown, the present application provides a method for generating a stator partial discharge detection report. First, a knowledge graph is constructed. A knowledge graph in the field of generator stator partial discharge is constructed based on multi-source heterogeneous data (online monitoring data, maintenance cases, standard documents), supporting incremental knowledge updates. After obtaining the knowledge graph, a graph neural network and a rule engine are combined to realize fault pattern recognition and causal chain reasoning. Based on the realized fault mode and causal chain, the knowledge graph path and template constraints are integrated to generate a fault report containing data evidence and reasoning logic.

[0132] Specifically, if Figure 3 Shown, including:

[0133] Step S301: Knowledge graph construction and entity feature learning. The terminal first constructs a knowledge graph specifically for the field of generator stator partial discharge. This graph systematically defines various entity types such as fault modes, equipment parameters, maintenance examples, and industry standards, and clarifies the logical relationships between them. In order to enable this symbolic knowledge to be effectively utilized by subsequent models, the terminal further adopts graph embedding technology to learn a low-dimensional, dense vector representation for each entity and relationship in the graph. This learning process is iteratively optimized so that the algebraic operations of entity vectors and relationship vectors in the vector space can accurately reflect their true logical connections in the graph. The final stable entity vector is used as the entity feature in the subsequent steps.

[0134] S302: Real-time partial discharge data acquisition and corresponding discharge feature extraction. After the system enters the online monitoring state, the terminal acquires the raw partial discharge monitoring data stream from sensors installed on the generator stator (such as high-frequency current transformers or ultra-high-frequency sensors) in real time. Next, the terminal applies a series of signal processing algorithms, such as wavelet transforms, Fourier analysis, or statistical moment calculations, to extract from these raw, complex signal waveforms a set of discharge feature vectors that quantitatively characterize the key characteristics of the current partial discharge phenomenon. These characteristics may include pulse amplitude, number, phase distribution characteristics, energy, and waveform parameters.

[0135] S303, risk level assessment and quantification. After obtaining the real-time discharge characteristics, the terminal will input them into a pre-trained risk level determination model (such as a Bayesian network). The model internally stores the conditional probability of various discharge characteristics appearing under different risk levels (such as low, medium, high, and emergency) and the prior probability of occurrence of each risk level. By combining these probabilistic dependencies with the currently observed discharge characteristics, the model performs probabilistic reasoning and outputs a preliminary discrete current risk level. To further refine the assessment, the terminal will also calculate a continuous risk quantification parameter: first, the standardized discharge characteristics and related equipment status parameters are weighted to obtain a risk score; then, through a nonlinear function calibrated based on historical data, the degree of deviation of this risk score from a preset benchmark risk score and its changing trend are evaluated, thereby obtaining a more refined and continuous risk quantification value.

[0136] S304, fault pattern recognition based on knowledge graph. This is the core diagnosis stage, in which the real-time discharge features extracted by the terminal are compared with the entity features of each entity in the knowledge graph. Specifically, the terminal first maps the real-time discharge feature vector and each entity feature vector to a unified feature vector space, and then combines the real-time discharge feature vector with each entity feature vector to form a joint feature. Next, the terminal processes these joint features through a set of adjustable preset parameters optimized by historical data, and calculates the original correlation score corresponding to each joint feature. These scores are then normalized to obtain the final degree of correlation between each entity and the current discharge feature relative to other entities. The terminal selects the entity with the highest degree of correlation or that meets specific conditions, and finally determines one or more target fault modes based on the type of the matching entity (for example, directly matching to the fault mode entity, or matching to the device parameter entity and then inferring through the graph relationship).

[0137] S305: Build a decision support path. After determining the target failure mode, the terminal uses this as a starting point to conduct in-depth mining and path construction within the knowledge graph. Following predefined relationships in the knowledge graph (such as failure outcome, cause, solution, and reference standard), it conducts reverse causal tracing to construct a causal chain, and forward solution exploration to develop a repair strategy. These two components together form a structured decision support path that comprehensively explains the potential causes and development logic of the target failure mode, as well as recommended countermeasures or reference standards.

[0138] S306, generation and display of the fault detection report. This process may first include a template filling step: based on the discharge characteristics and decision support path, the terminal extracts key information as keywords, fills in the corresponding fields of a preset structured template, and quickly generates a structured summary of the report. Subsequently, the terminal calls a large language model, takes the obtained complete decision support path as the core input, and uses autoregressive text generation to describe its content in a detailed and coherent natural language form. The generated natural language description is integrated with the aforementioned structured information to form the final fault detection report. If necessary, the terminal can also translate this report into the user-specified target language for display, thereby realizing a complete and intelligent fault detection and reporting process.

[0139] Optionally, in order to verify the effectiveness of the present application, an exemplary test was conducted. In this test, the method of the present application was applied to a set of actual generator stator partial discharge cases and compared with the traditional diagnostic method to evaluate the improvement in diagnostic accuracy. The calculation method for the improvement in diagnostic accuracy is shown in formula (7). Specifically, the number of correctly diagnosed cases of the new method ( ) and the number of correctly diagnosed cases with the old method ( ), divided by the total number of test cases ( ), and then multiply by 100%. After actual measurement, the method of the present invention has improved the diagnostic accuracy by ( ) reached more than 25%.

[0140]

[0141] This increase in accuracy means that fault modes can be identified more reliably, misjudgments and missed judgments can be reduced, and a more solid technical guarantee is provided for subsequent equipment maintenance and operation decisions.

[0142] In terms of report generation efficiency, this embodiment also shows significant advantages. The total time for report generation ( ), as shown in formula (8), mainly consists of two parts: one is the fault matching and decision path generation time ( ), and secondly, the natural language report generation time ( In the test, the time between fault matching and decision path generation was controlled within 1 second ( <1s, and the generation time of natural language report is controlled within 5 seconds ( <5s). As a result, the entire fault detection report can be generated in seconds, in stark contrast to traditional methods that typically require hours or even longer for manual analysis and report writing. This significantly improves the response speed of fault diagnosis and reporting, saving valuable time for rapid decision-making and timely troubleshooting.

[0143]

[0144] In this embodiment, no human intervention is required from data collection to report generation, and the response time is shortened from the traditional several hours to minutes, which enables timely detection and handling of faults, effectively prevents equipment damage and downtime accidents caused by faults, and improves the reliability and stability of the equipment. At the same time, by associating fault mechanisms with maintenance suggestions through knowledge graphs, the report logic is clear and conforms to the cognitive habits of operation and maintenance personnel.

[0145] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless clearly stated herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps. It is understandable that the various steps in different embodiments can be freely combined as needed, and the various non-contradictory schemes formed by the combination all fall within the scope of protection of this application.

[0146] Based on the same inventive concept, embodiments of the present application further provide a stator partial discharge detection report generation device for implementing the aforementioned stator partial discharge detection report generation method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the stator partial discharge detection report generation device provided below can be found in the aforementioned limitations of the stator partial discharge detection report generation method and are not further elaborated here.

[0147] In an exemplary embodiment, Figure 4 As shown, a stator partial discharge detection report generation device is provided, comprising: a partial discharge knowledge graph establishment module 410, a target fault mode determination module 420, a decision support path determination module 430, and a fault detection report generation module 440, wherein:

[0148] A partial discharge knowledge graph establishment module 410 is configured to establish a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity;

[0149] a target fault mode determination module 420 for obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of the plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features;

[0150] A decision support path determination module 430 is configured to determine a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph;

[0151] The fault detection report generating module 440 is configured to describe the decision support path in natural language using a large language model to obtain a stator partial discharge fault detection report.

[0152] In one embodiment, the target fault mode determination module 420 is further used to convert the discharge characteristics and the entity characteristics of each of the entities into a unified feature vector space, and combine the converted discharge characteristics with the converted entity characteristics to obtain a joint characteristic; in the feature vector space, calculate the correlation score of each of the joint characteristics; according to the correlation score of each of the joint characteristics, determine the degree of correlation between each of the entities and the discharge characteristics relative to other entities, and obtain entities whose correlation degree meets preset conditions; if the entity is the device parameter entity, determine the target fault mode that matches the discharge characteristic based on the association relationship related to the device parameter entity; and if the entity is the fault mode entity, use the fault mode indicated by the fault mode entity as the target fault mode.

[0153] In one embodiment, the partial discharge knowledge graph establishment module 410 is also used to obtain the entity vector of each of the entities and the relationship vector of the association relationship between each of the entities; for each group of the association relationship, the head entity and the tail entity associated with the association relationship, the sum of the association relationship vector corresponding to the association relationship and the head entity vector corresponding to the head entity is obtained, and the difference between the entity vector obtained by converting the head entity vector through the association relationship and the tail entity vector is determined based on the difference between the summation result and the tail entity vector corresponding to the tail entity; a loss value is determined based on the multiple differences, and each of the entity vectors and each of the relationship vectors is iteratively adjusted based on the loss value until a preset iteration end condition is met, and each of the entity features is obtained based on the current each of the entity vectors.

[0154] In one embodiment, the target fault mode determination module 420 is further used to input the discharge characteristics into a pre-trained risk level determination model; the risk level determination model includes a probability dependency relationship between each risk level and each discharge characteristic; the probability dependency relationship indicates the conditional probability of the discharge characteristic at different risk levels, and the current occurrence probability of each risk level; the risk level determination model outputs the current risk level of the stator partial discharge based on the probability dependency relationship and the discharge characteristics.

[0155] In one embodiment, the target failure mode determination module 420 is further configured to weight the standardized discharge characteristics and the real-time partial discharge data corresponding to the standardized discharge characteristics based on preset weights to determine a risk score under the current risk level; determine a degree of change in the deviation of the risk score from a preset baseline risk score, and determine a risk quantification parameter corresponding to the current risk level based on the degree of change.

[0156] In one embodiment, the fault detection report generation module 440 is further used to determine the keywords corresponding to the fields of the preset structured template based on the discharge characteristics and the decision support path, and generate a structured fault detection report based on the keywords; obtain the translation content of the fault detection report corresponding to the target language, and display the fault detection report based on the translation content.

[0157] Each module in the stator partial discharge detection report generating device can be implemented in whole or in part via software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0158] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 5As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via wired or wireless communication, and the wireless communication can be achieved via Wi-Fi, a mobile cellular network, near field communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for generating a stator partial discharge detection report.

[0159] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0160] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0161] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0162] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0163] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0164] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0165] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0166] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating a stator partial discharge detection report, characterized in that: The method comprises: Establishing a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and association relationships between the entities; the multiple entities include a fault mode entity, an equipment parameter entity, a maintenance instance entity, and an industry standard entity; Obtaining discharge features corresponding to real-time partial discharge data, comparing the discharge features with entity features of a plurality of entities in the knowledge graph, determining entities matching the discharge features based on the comparison results, and determining a target fault mode based on the entities matching the discharge features; Determining a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph; The decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report.

2. The method according to claim 1, characterized in that The discharge feature is compared with entity features of the plurality of entities in the knowledge graph, and entities matching the discharge feature are determined according to the comparison results, and target fault modes are determined according to the entities matching the discharge feature. converting the discharge features and the entity features of each of the entities into a unified feature vector space, and combining the converted discharge features with the converted entity features to obtain a joint feature; In the feature vector space, calculating the association score of each of the joint features; Determining the degree of association between each entity and the discharge feature relative to other entities based on the association score of each joint feature, and obtaining entities whose association degree meets a preset condition; If the entity is the device parameter entity, determining a target fault mode that matches the discharge feature according to an association relationship related to the device parameter entity; and If the entity is the fault mode entity, the fault mode indicated by the fault mode entity is used as the target fault mode.

3. The method according to claim 1, characterized in that Each of the entity features is obtained through the following steps: Obtaining entity vectors of each entity and relationship vectors of association relationships between each entity; For each group of the association relationship, the head entity and the tail entity associated with the association relationship, obtaining a sum of the association relationship vector corresponding to the association relationship and the head entity vector corresponding to the head entity, and determining a difference between an entity vector obtained by converting the head entity vector through the association relationship and the tail entity vector based on a difference between the sum and the tail entity vector corresponding to the tail entity; A loss value is determined based on the multiple differences, and each entity vector and each relationship vector is iteratively adjusted based on the loss value until a preset iteration end condition is met, and each entity feature is obtained based on the current entity vector.

4. The method according to claim 1, wherein After acquiring the real-time partial discharge data and extracting the discharge characteristics of the real-time partial discharge data, the method further includes: Inputting the discharge characteristics into a pre-trained risk level determination model; the risk level determination model includes a probability dependency relationship between each risk level and each discharge characteristic; the probability dependency relationship indicates the conditional probability of the discharge characteristics at different risk levels, and the current probability of occurrence of each risk level; The risk level determination model outputs a current risk level of the stator partial discharge according to the probability dependency and the discharge characteristics.

5. The method according to claim 4, characterized in that The current risk level is determined by corresponding risk quantification parameters, including: Based on preset weights, weighting the standardized discharge characteristics and the real-time partial discharge data corresponding to the standardized discharge characteristics to determine a risk score at the current risk level; The degree of change of the risk score relative to the preset benchmark risk score is determined, and the risk quantification parameter corresponding to the current risk level is determined based on the degree of change.

6. The method according to any one of claims 1 to 5, characterized in that Before the decision support path is described in natural language by a large language model to obtain a stator partial discharge fault detection report, the method further includes: Based on the discharge characteristics and the decision support path, determining keywords corresponding to fields of a preset structured template, and generating a structured fault detection report based on the keywords; After the decision support path is described in natural language using a large language model to obtain a stator partial discharge fault detection report, the method further includes: Obtain translation content of the fault detection report corresponding to the target language, and display the fault detection report based on the translation content.

7. A stator partial discharge detection report generation device, characterized in that: The device comprises: A partial discharge knowledge graph establishment module is used to establish a knowledge graph for stator partial discharge; the knowledge graph includes multiple entities related to stator partial discharge faults and the association relationships between the multiple entities; the multiple entities include fault mode entities, equipment parameter entities, maintenance instance entities, and industry standard entities; a target fault mode determination module, configured to obtain discharge features corresponding to real-time partial discharge data, compare the discharge features with entity features of the plurality of entities in the knowledge graph, determine entities matching the discharge features based on the comparison results, and determine a target fault mode based on the entities matching the discharge features; A decision support path determination module, configured to determine a decision support path including a fault causal chain and a fault repair strategy based on the target fault mode and the association relationship in the knowledge graph; The fault detection report generation module is used to describe the decision support path in natural language through a large language model to obtain a stator partial discharge fault detection report.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images