Executable suggestion generation method, device and equipment for hydropower station equipment maintenance

Abstract

The application provides a method, device and equipment for generating executable suggestions for hydropower station equipment maintenance, which comprises: quantitatively scoring candidate maintenance suggestions based on target scoring indicators including specificity, relevance, executability and conciseness of the candidate maintenance suggestions to obtain a comprehensive quality score; when the comprehensive quality score does not reach a preset quality threshold, revising the candidate maintenance suggestions based on the comprehensive quality score and rescorings until the comprehensive quality score reaches the preset quality threshold, and outputting the final candidate maintenance suggestions as executable maintenance suggestions. The application quantitatively scores candidate maintenance suggestions based on target scoring indicators including specificity, relevance, executability and conciseness, and revises the candidate maintenance suggestions in a loop until the conditions are met when the comprehensive quality score does not reach a preset quality threshold, thereby ensuring that the final output maintenance suggestions are subjected to multi-dimensional, standardized evaluation and closed-loop optimization, and improving the accuracy, executability and standardization of the maintenance suggestions.

Description

Technical Field

[0001] This invention relates to the field of suggestion generation technology, and in particular to an executable suggestion generation method, apparatus and equipment for the maintenance of hydropower station equipment. Background Technology

[0002] Hydropower station equipment, as the core carrier of power generation, basin scheduling, and safety assurance, encompasses all categories of key equipment, including turbines, generators, governors, excitation systems, gate control devices, hydraulic systems, monitoring sensors, and protection and control devices. Its daily maintenance, fault handling, hazard rectification, and preventative maintenance are characterized by high professionalism, high safety, low fault tolerance, and standardized processes. These aspects are crucial for ensuring the safe and stable operation of the power station, extending equipment lifespan, reducing the risk of unplanned outages, and improving overall power generation efficiency. As hydropower stations develop towards intelligence, unmanned operation, and intensification, textual information such as equipment operation data, fault records, maintenance feedback, and operation and maintenance logs is experiencing massive growth. How to quickly extract core issues from large-scale unstructured data and generate accurate and actionable maintenance decision recommendations has become a critical technological need for breakthroughs in the field of intelligent operation and maintenance of hydropower stations.

[0003] Currently, the generation of maintenance decisions and recommendations for hydropower station equipment mainly relies on two mainstream technical approaches. The first is the traditional rule-based and experience-driven approach. This approach centers on the professional experience of maintenance personnel, combining equipment manuals, industry regulations, and historical fault cases to conduct manual judgment and solution formulation. Maintenance guidance is formed through manual problem identification, matching of measures, and process arrangement, and is currently the most widely used traditional model in the industry. This method highly depends on the professional competence and experience of personnel. While it has some applicability in standardized routine maintenance, it generally suffers from significant drawbacks when facing complex faults, multi-equipment linkage anomalies, and massive data parallel processing scenarios, including low decision-making efficiency, difficulty in experience transfer, insufficient accuracy in problem localization, and a lack of executable and quantifiable basis for maintenance recommendations. The second approach is a generative technical approach based on a single large-scale language model. This approach directly inputs equipment maintenance logs, real-time fault data, online monitoring information, and historical operation and maintenance records into a single language model. Relying on the model's text understanding and generation capabilities, it outputs maintenance-related recommendations, enabling, to a certain extent, rapid processing of unstructured data and textual output of results, providing auxiliary reference for manual decision-making. However, single models have significant technical limitations when dealing with professional tasks such as the maintenance of hydropower station equipment, which are characterized by high reliability, strong standardization, and heavy practicality. They cannot meet the actual engineering requirements of "precision, feasibility, low cost, and high efficiency" in the maintenance of hydropower station equipment.

[0004] In recent years, large-scale language model frameworks based on multi-agent collaboration have been effectively validated in fields such as business review analysis and user recommendation systems. These frameworks, through the specialized division of labor and closed-loop collaboration among clustering agents, question extraction agents, suggestion generation agents, evaluation agents, and ranking agents, can transform large-scale unstructured text data into actionable decision suggestions, demonstrating significant advantages in improving the relevance, practicality, and systematic nature of the output results. However, existing multi-agent frameworks are designed for general scenarios, exhibiting poor domain adaptability. Summary of the Invention

[0005] This invention provides a method, apparatus, and equipment for generating executable suggestions for the maintenance of hydropower station equipment, which solves the shortcomings of existing technologies in hydropower station maintenance decision-making, which are highly dependent on human experience, resulting in low efficiency and difficulty in quantification, and the fact that suggestions generated by single models or general multi-agent systems often have factual biases and process errors, and cannot meet the requirements for engineering implementation.

[0006] This invention provides a method for generating executable suggestions for the maintenance of hydropower station equipment, comprising the following steps.

[0007] Obtain representative fault samples; Based on the representative fault samples, structured maintenance problems are generated; Generate at least one candidate maintenance suggestion for the structured maintenance problem; Based on the target scoring indicators, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility, and simplicity of the candidate maintenance suggestions. If the overall quality score does not reach the preset quality threshold, the candidate maintenance suggestion is corrected based on the overall quality score, and the step of quantifying the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score is re-executed until the overall quality score reaches the preset quality threshold. The final candidate maintenance suggestion is then used as an executable maintenance suggestion.

[0008] According to the present invention, an executable suggestion generation method for hydropower station equipment maintenance includes the step of obtaining representative fault samples, which comprises: Obtain the original hydropower station equipment maintenance text set and generate semantic vectors corresponding to each maintenance text in the original hydropower station equipment maintenance text set; Cluster the semantic vectors to obtain data clusters; Based on the center of each data cluster, representative text is selected from each data cluster; The representative fault samples are determined based on the representative text.

[0009] According to the present invention, an executable suggestion generation method for hydropower station equipment maintenance includes selecting representative text from each of the data clusters based on the center of each data cluster, comprising: Determine the cluster center vector of each of the data clusters; The maintenance text with the highest semantic similarity to the cluster center vector in each data cluster is selected as the representative text.

[0010] According to the present invention, an executable suggestion generation method for hydropower station equipment maintenance is provided, wherein the cluster center vector is obtained by averaging all semantic vectors in the data cluster.

[0011] According to the present invention, an executable suggestion generation method for hydropower station equipment maintenance is provided, wherein generating structured maintenance problems based on the representative fault samples includes: Determine the repair topics for the representative fault samples; From the representative fault samples, at least one maintenance item is extracted; The maintenance topic and the maintenance item are combined to generate the structured maintenance problem.

[0012] According to the present invention, a method for generating executable recommendations for the maintenance of hydropower station equipment, wherein the method further includes, after selecting the final candidate maintenance recommendations as executable maintenance recommendations, the method further comprises: Based on multiple executable maintenance suggestions that achieve the preset quality threshold in the comprehensive quality score, a candidate suggestion set is formed; Based on the business ranking dimension, each executable maintenance suggestion in the candidate suggestion set is scored for its business value, resulting in a business value score; the business ranking dimension includes implementation cost, practical difficulty, improvement effect, and safety compliance; The business value scores are sorted to obtain a sorting result, and the executable maintenance suggestion with the highest business value score in the sorting result is taken as the executable maintenance suggestion.

[0013] According to the present invention, a method for generating executable suggestions for the maintenance of hydropower station equipment includes, in which, based on a business ranking dimension, each executable maintenance suggestion in the candidate suggestion set is scored for its business value to obtain a business value score, including: Obtain standardized individual scores for the executable maintenance recommendations in the dimensions of implementation cost, practical difficulty, improvement effect, and safety compliance. Based on preset business weights, the standardized individual scores are weighted to obtain the business value score.

[0014] The present invention also provides an executable suggestion generation device for the maintenance of hydropower station equipment, comprising the following units: The acquisition unit is used to acquire representative fault samples. The first generation unit is used to generate structured maintenance problems based on the representative fault samples. The second generation unit is used to generate at least one candidate maintenance suggestion for the structured maintenance problem. The scoring unit is used to quantitatively score the candidate maintenance suggestions based on the target scoring indicators to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility and simplicity of the candidate maintenance suggestions. The correction unit is used to correct the candidate maintenance suggestion based on the overall quality score when the overall quality score does not reach the preset quality threshold, and to re-execute the step of quantifying the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score, until the overall quality score reaches the preset quality threshold, and then use the final candidate maintenance suggestion as an executable maintenance suggestion.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement an executable suggestion generation method for hydropower station equipment maintenance as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an executable suggestion generation method for hydropower station equipment maintenance as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements an executable suggestion generation method for hydropower station equipment maintenance as described above.

[0018] This invention provides a method, apparatus, and equipment for generating executable suggestions for hydropower station equipment maintenance. After acquiring representative fault samples and generating structured maintenance problems, candidate maintenance suggestions are generated. Then, based on target scoring indicators including specificity, relevance, executability, and simplicity, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score. If the comprehensive quality score does not reach a preset quality threshold, the candidate maintenance suggestions are corrected and re-scored until the comprehensive quality score reaches the preset quality threshold. Finally, the final candidate maintenance suggestions are output as executable maintenance suggestions. This invention ensures that the final output maintenance suggestions have undergone multi-dimensional, standardized evaluation and closed-loop optimization by quantitatively scoring the generated candidate maintenance suggestions based on target scoring indicators including specificity, relevance, executability, and simplicity, and by iteratively correcting them until the conditions are met when the comprehensive quality score does not reach the preset quality threshold. This improves the accuracy, executability, and standardization of maintenance suggestions, solving the problems of lack of executability and quantitative basis in traditional manual methods, and the uncontrollable quality and vague content when directly generated by a single model. This guarantees the scientific and reliable nature of hydropower station equipment maintenance decisions. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts illustrating the executable suggestion generation method for hydropower station equipment maintenance provided by the present invention.

[0021] Figure 2 This is the second flowchart of the executable suggestion generation method for hydropower station equipment maintenance provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the executable suggestion generation device for hydropower station equipment maintenance provided by the present invention.

[0023] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] This invention provides a method for generating executable suggestions for the maintenance of hydropower station equipment. It can generate accurate, implementable, and high-quality maintenance suggestions that are highly tailored to the specific needs of the hydropower station from massive amounts of unstructured maintenance data. This invention models the process of generating executable suggestions for hydropower station equipment maintenance as a unified functional expression: ; in, This represents the input set of text related to the maintenance of hydropower station equipment, including fault descriptions, maintenance feedback, operation and maintenance logs, and anomaly records. This represents a set of five intelligent agents. In this embodiment, each intelligent agent typically uses one or more artificial intelligence models (such as large language models) to implement its specific functions. This represents the internal trainable parameters of the artificial intelligence model used to implement the above-mentioned intelligent agent functions, as well as the set of weights used in this method flow to calculate the comprehensive quality score and business value score; This indicates the suggested overall quality score threshold; Indicates the maximum number of iterations; This indicates the final, executable troubleshooting recommendations.

[0026] The overall execution process is divided into six stages: inspection text input and preprocessing, clustering agent distillation, problem extraction agent structuring, suggestion generation agent initialization, evaluation agent iterative optimization, and ranking agent optimal output. Each agent has a clear division of labor, close-loop collaboration, and can complete the entire process calculation without human intervention.

[0027] Figure 1 This is one of the flowcharts illustrating the executable suggestion generation method for hydropower station equipment maintenance provided by the present invention, such as... Figure 1 As shown, the method includes the following: Step 110: Obtain representative fault samples.

[0028] Specifically, the first step is to obtain representative fault samples. These representative fault samples are selected from a large volume of highly redundant raw hydropower station equipment maintenance texts, such as fault records, maintenance feedback, and operation and maintenance logs, and are chosen to represent a typical type of fault or core maintenance problem. The purpose of obtaining representative fault samples is to "distill" the raw data, removing redundancy and noise while retaining core semantic information, thereby significantly improving the efficiency and accuracy of subsequent processing.

[0029] There are several ways to obtain representative fault samples. In one possible implementation, senior technical experts or maintenance engineers in the field of hydropower station equipment can manually select typical and frequently occurring fault descriptions from a large number of historical maintenance records, based on their professional knowledge and practical experience, to serve as representative fault samples. In another possible implementation, automated text processing technology can be used. For example, keyword frequency statistics can be performed on all original maintenance texts to filter out texts containing high-frequency fault keywords, such as "oil leak," "abnormal noise," and "abnormal temperature rise." Then, representative fault samples can be obtained from these filtered texts through random sampling or selection based on rules such as text length and completeness.

[0030] In a more preferred embodiment, unsupervised learning methods in natural language processing, such as text clustering, can be used to automatically group semantically similar fault description texts into one category, and then select one or more of the most representative texts from each category as representative fault samples. Compared with the aforementioned methods, this approach has a higher degree of automation and better ensures the comprehensiveness and typicality of the samples.

[0031] Step 120: Based on the representative fault samples, generate structured maintenance problems.

[0032] Specifically, after obtaining representative fault samples, structured maintenance problems can be generated based on these samples. Here, a structured maintenance problem refers to converting the representative fault samples into a standardized task description that can be understood and processed by a computer. This structured representation can usually clearly separate the theme and specific content of the maintenance task.

[0033] Step 130: Generate at least one candidate maintenance suggestion for the structured maintenance problem.

[0034] Specifically, after generating a structured maintenance problem, at least one candidate maintenance suggestion can be generated for the structured maintenance problem. The candidate maintenance suggestion reflects one or more preliminary, unassessed solutions proposed for the structured maintenance problem.

[0035] Here, generating candidate maintenance suggestions can rely on a knowledge base or generative model with expertise in hydropower stations. In one implementation, a maintenance scheme knowledge base can be constructed, storing a large number of "problem-solution" pairs. Upon receiving a structured maintenance problem, the system searches the knowledge base, matches the most similar problem, and uses its corresponding solution as a candidate maintenance suggestion.

[0036] In a preferred embodiment, this step is implemented through a suggestion-generating agent, which is a large language model fine-tuned from professional corpora including hydropower station maintenance procedures, safety regulations, and historical successful cases. Taking a structured maintenance problem as input, the suggestion-generating agent can follow the hydropower station's maintenance logic and safety constraints to generate candidate maintenance suggestions that are complete in content and clear in steps. For example, for the first... Representative questions The system can generate 3-4 initial candidate maintenance suggestions, which constitute the candidate maintenance suggestions. : ; in, This indicates a candidate maintenance suggestion. For the same structured maintenance problem This generates three different initial candidate maintenance suggestions.

[0037] During this generation process, a series of generation constraints must be followed to ensure the quality of candidate maintenance recommendations, specifically including: ① The steps must be executable and should not contain vague or unexecutable statements; ② Key information such as the implementing entity, required tools, timing of operation, and specific location must be clearly defined; ③ It must fully comply with the relevant safety regulations and operating procedures for hydropower station equipment; ④ It should not contain explanatory text or redundant descriptions that are irrelevant to solving the problem.

[0038] For example, in response to the above problem, a candidate maintenance suggestion might be generated: "1. Obtain a work permit and set up safety measures; 2. Depressurize and drain the oil stored in the air conditioning speed reducer oil tank; 3. Disassemble the oil tank flange and remove the aged seals; 4. Clean the flange sealing surface and replace the seals with new ones; 5. Reinstall and tighten the bolts to the specified torque; 6. Fill the oil tank with oil and build up pressure, and check for leaks." Step 140: Based on the target scoring indicators, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility and simplicity of the candidate maintenance suggestions.

[0039] Specifically, candidate maintenance recommendations can be quantitatively scored based on target scoring indicators to obtain a comprehensive quality score. These target scoring indicators can include the specificity, relevance, feasibility, and simplicity of the candidate maintenance recommendations.

[0040] Specificity (S) reflects whether the operational instructions of the candidate maintenance proposal are specific and clear, avoiding vague and generalized expressions. Specifically, specificity assesses whether the implementation details, such as time, personnel, tools, and location, are clearly defined. Relevance (R) reflects whether the content of the candidate maintenance proposal closely corresponds to the structured maintenance problem, i.e., whether it is an effective solution that directly addresses the target maintenance problem. Feasibility (A) reflects whether the candidate maintenance proposal has the conditions and feasibility for implementation in the actual hydropower station site environment. Specifically, feasibility assesses whether the proposal can be implemented on-site, whether the implementation cost is reasonable, and whether the operating procedures comply with safety production regulations. Conciseness (C) reflects whether the textual expression of the maintenance proposal is concise and easy to understand, i.e., whether there are redundant or irrelevant explanations or descriptions.

[0041] Specifically, the quantification scoring process can be completed by an independent evaluation agent (which can also be a large language model or a specialized classification model). The evaluation agent receives candidate maintenance suggestions and corresponding structured maintenance questions as input, and assigns a quantified score (e.g., 1-5 points) to each of the four indicators according to the built-in evaluation rules.

[0042] Subsequently, the scores from the four dimensions are weighted and summed to calculate the overall quality score. In the t-th iteration, the formula for calculating the overall quality score is: ; in, This indicates the overall quality score. , , and These are the weights for specificity, relevance, executability, and simplicity, respectively, with equal weights used by default. And satisfy .

[0043] Step 150: If the overall quality score does not reach the preset quality threshold, the candidate maintenance suggestion is corrected based on the overall quality score, and the step of quantifying and scoring the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score is re-executed until the overall quality score reaches the preset quality threshold. The final candidate maintenance suggestion is then used as an executable maintenance suggestion.

[0044] Specifically, if the overall quality score does not reach the preset quality threshold, the candidate maintenance suggestions are revised based on the overall quality score, and the steps of quantifying the candidate maintenance suggestions based on the target scoring index are re-executed to obtain the overall quality score, until the overall quality score reaches the preset quality threshold, and the final candidate maintenance suggestions are used as executable maintenance suggestions.

[0045] The preset quality threshold is a pre-defined score line. For example, a suggested passing threshold can be set. 3.5 out of 5. Only candidate maintenance recommendations that achieve or exceed this threshold in overall quality score are considered acceptable.

[0046] If the overall quality score calculated in the previous step does not reach the preset quality threshold, the system will trigger a correction process. This iterative generation process can be represented by the following formula: In the t-th iteration, the optimization suggestion The agent is generated by suggestions. Based on the evaluation feedback from the previous round (round t-1) generated .

[0047] The feedback from this round of evaluations This is determined by the evaluation agent. Regarding the current recommendations Based on the set of evaluation rules The results of the evaluation .

[0048] Among them, evaluation feedback Specifically, this may include scores for each dimension and necessary textual feedback, such as indicating "low executability score because the required tool model was not specified" or "insufficient specificity, the operation steps are not clearly defined".

[0049] After receiving the correction instruction, the suggestion generation agent will generate a new version of the candidate repair suggestion. This new suggestion will then re-enter step 140 for scoring. This "scoring-feedback-correction-regeneration" loop will continue until a preset iteration stopping condition is met. The preset iteration stopping condition can be expressed as: ; This means that the overall quality score for the current round reaches or exceeds the passing threshold. or number of iterations Reaching the preset maximum number of iterations (For example, =5), to avoid an infinite loop.

[0050] When the loop terminates, the system will select the suggestion that satisfies the preset iteration stopping condition as the optimal suggestion. The selection rules can be expressed as follows: ; That is, choosing a suggestion that satisfies the stopping condition at the minimum number of iterations T. This final recommendation, which meets the quality requirements, will be formally designated as an actionable maintenance recommendation. Actionable maintenance recommendations are reliably guaranteed in quality due to undergoing multiple rounds of rigorous, standardized evaluation and optimization.

[0051] The method provided in this invention obtains representative fault samples and generates structured maintenance problems, then generates candidate maintenance suggestions. Based on target scoring indicators including specificity, relevance, executability, and simplicity, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score. If the comprehensive quality score does not reach a preset quality threshold, the candidate maintenance suggestions are corrected based on the comprehensive quality score and re-scored until the comprehensive quality score reaches the preset quality threshold. Finally, the final candidate maintenance suggestions are output as executable maintenance suggestions. This invention, by quantitatively scoring the generated candidate maintenance suggestions based on target scoring indicators including specificity, relevance, executability, and simplicity, and by iteratively correcting the comprehensive quality score until the conditions are met, ensures that the final output maintenance suggestions have undergone multi-dimensional, standardized evaluation and closed-loop optimization. This improves the accuracy, executability, and standardization of maintenance suggestions, solving the problems of lack of executability and quantitative basis in traditional manual methods, and the uncontrollable quality and vague content when directly generated by a single model. This ensures the scientific and reliable nature of hydropower station equipment maintenance decisions.

[0052] Based on the above embodiments, the step of obtaining the representative fault sample includes: Step 210: Obtain the original hydropower station equipment maintenance text set and generate the semantic vector corresponding to each maintenance text in the original hydropower station equipment maintenance text set. Step 220: Cluster the semantic vectors to obtain data clusters; Step 230: Based on the center of each data cluster, select representative text from each data cluster; Step 240: Determine the representative fault sample based on the representative text.

[0053] Specifically, firstly, the original hydropower station equipment maintenance text set is obtained, and semantic vectors corresponding to each maintenance text in the original hydropower station equipment maintenance text set are generated.

[0054] Specifically, the original hydropower station equipment maintenance text dataset (S), which serves as the initial data source, is cleaned, deduplicated, standardized for symbols, and filtered for outliers to obtain a clean maintenance text dataset. This process can be represented by a text cleaning function: ; in, This is a text cleaning function that retains only the core semantic information such as fault phenomena, maintenance problems, and handling results in each maintenance text after processing, providing high-quality input for subsequent processing.

[0055] Furthermore, semantic vectors can be generated from the cleaned maintenance texts. For each maintenance text in the clean maintenance text dataset, a corresponding semantic vector is generated. Semantic vectors, also known as word embeddings or text embeddings, are a mathematical method that represents text as a series of fixed-length real-valued vectors. Semantic vectors can capture the deep semantic information of text, making the vectors of texts with similar meanings more spatially similar.

[0056] Specifically, each cleaning and inspection text Semantic vectorization is achieved by converting a pre-trained language model into a dense vector of fixed dimensions. This process can be represented as: ; in, Represent a 3D text embedding vector; This indicates the semantic embedding model used, such as BERT (Bidirectional Encoder Representations from Transformers).

[0057] Then, the semantic vectors can be clustered to obtain data clusters. After converting all maintenance texts into semantic vectors, density clustering, such as DBSCAN (Density-Based Spatial Clustering of Applications with Noise) or hierarchical clustering algorithms, is used to perform clustering operations on these semantic vectors. The purpose is to automatically group semantically similar maintenance texts into the same group.

[0058] The resulting data clusters refer to a set of vectors formed by the clustering algorithm. All embedded vectors are divided into K semantically similar clusters, forming a cluster set. Each data cluster corresponds to a specific, recurring, typical fault or maintenance problem. For any text... The cluster affiliation satisfies ,in Represents the cluster allocation function, Indicates the first All maintenance texts contained in each cluster.

[0059] Furthermore, representative texts can be selected from each data cluster based on its central element. The purpose of this step is to identify the text that best represents the core idea of ​​each identified fault category (i.e., data cluster).

[0060] Here, clustering agent It is responsible for transforming massive, highly redundant maintenance texts into a small number of highly representative fault samples, achieving corpus distillation, deduplication, and retention of core issues, thus solving the bottleneck of single models being unable to handle large-scale data.

[0061] Finally, representative fault samples can be determined based on representative texts. This is done by calculating the number of texts within each cluster. Sort the clusters from largest to smallest, select the representative samples corresponding to the top M high-frequency problem clusters, and summarize them to obtain the representative fault samples. The representative fault samples can be represented as: ; in, This represents a representative fault sample. This represents the first high-frequency problem cluster. This represents the second high-frequency problem cluster. This represents the Mth high-frequency problem cluster.

[0062] This step allows tens of thousands of original maintenance records to be efficiently compressed into dozens of the most representative samples, thereby significantly reducing data redundancy and greatly improving the efficiency of subsequent processing.

[0063] The method provided in this invention converts the original maintenance text set into semantic vectors and uses a clustering algorithm to divide the data into clusters, thereby achieving automated summarization and sorting of massive, highly redundant unstructured maintenance data. This solves the shortcomings of traditional manual methods in dealing with massive amounts of data, such as low efficiency and difficulty in discovering potential correlations and core issues.

[0064] Based on the above embodiments, step 230 includes: Step 230-1: Determine the cluster center vector of each data cluster; Step 230-2: Select the inspection text with the highest semantic similarity to the cluster center vector in each data cluster as the representative text.

[0065] Specifically, the first step is to determine the cluster center vector for each data cluster. The cluster center vector is a "center point" of the data cluster in the vector space, and mathematically, it represents the average semantic direction or core semantics of all semantic vectors within that data cluster.

[0066] After determining the cluster center vector, the semantic similarity between the semantic vector of each maintenance text in the data cluster and the cluster center vector is calculated. Semantic similarity is an indicator that measures the directional proximity between the semantic vector of each maintenance text and the cluster center vector; a higher semantic similarity value indicates that the texts corresponding to the two vectors are more semantically similar.

[0067] Finally, within each data cluster, the inspection text with the highest semantic similarity to the cluster's center vector is selected as the representative text of the data cluster. This selection process uses cosine similarity as a metric, which can be specifically expressed as: ; in, For the selected representative sample, Represents a data cluster, This represents the semantic vector of each maintenance text in the data cluster. This represents the cluster center vector.

[0068] Based on the above embodiments, the cluster center vector is obtained by averaging all semantic vectors in the data cluster.

[0069] In this embodiment, the cluster center vector is obtained by taking the arithmetic mean of all semantic vectors in the data cluster. The formula for calculating the cluster center vector is as follows: ; in, Let be the center vector of the k-th cluster. The number of maintenance texts within the k-th cluster. Represents a data cluster, This represents the semantic vector of each maintenance text in the data cluster.

[0070] Based on the above embodiments, step 120 includes: Step 120-1: Determine the repair topic of the representative fault sample; Step 120-2: Extract at least one maintenance item from the representative fault sample; Step 120-3: Combine the maintenance topic with the maintenance item to generate the structured maintenance problem.

[0071] Specifically, this method involves extracting an intelligent agent from a problem. The execution process aims to perform in-depth analysis of representative samples, extracting structured information of "maintenance topic + specific maintenance items" to avoid generating overly fragmented or generalized suggestions. Specifically, this includes: First, the maintenance topics for representative fault samples are determined. These maintenance topics refer to a high-level, standardized classification of the maintenance work content for hydropower station equipment. For example, maintenance topics can be preset to include equipment cleaning and maintenance, component wear and replacement, parameter anomaly calibration, leakage handling, sensor failure, and control system anomalies; however, this embodiment of the invention does not impose specific limitations on these. By mapping fault samples described in natural language to standard topics, preliminary normalization of the problems is achieved.

[0072] Then, at least one maintenance item can be extracted from the representative fault sample. The maintenance item is used to reflect the core problem description that is extracted from the original text and points to a specific equipment component and fault phenomenon.

[0073] In the specific extraction process, preset extraction rules can be followed. For example, for each maintenance topic, no more than 5 specific, non-repeating, and solvable maintenance problems can be extracted from its corresponding representative fault samples to form a set of maintenance items. ; in, It is for maintenance topics The set of maintenance items, This refers to the number of entries.

[0074] Finally, maintenance topics and maintenance items can be combined to generate structured maintenance problems, forming a structured key-value pair of "topic-item". The structured key-value pair is the structured maintenance problem.

[0075] The final output is a set of structured pairs consisting of multiple topics and their corresponding maintenance items: ; in, Q This results in the final set of structured maintenance problems. This is the first maintenance topic. This is the first maintenance item. This is the second maintenance topic. This is the second maintenance item. For the first M One maintenance topic. For the first M One maintenance item.

[0076] Based on the above embodiments, step 150, which involves selecting the final candidate maintenance suggestion as an executable maintenance suggestion, further includes: Step 151: Based on multiple executable maintenance suggestions whose comprehensive quality scores reach the preset quality threshold, a candidate suggestion set is formed; Step 152: Based on the business ranking dimension, score the business value of each executable maintenance suggestion in the candidate suggestion set to obtain a business value score; the business ranking dimension includes implementation cost, practical difficulty, improvement effect and safety compliance; Step 153: Sort the business value scores of each business to obtain the sorting results, and take the executable maintenance suggestion with the highest business value score in the sorting results as the executable maintenance suggestion.

[0077] Specifically, this embodiment aims to select the most practically valuable optimal suggestion from multiple technically qualified suggestions, further considering the actual operational needs of the hydropower station. This step is performed by a sorting agent. implement.

[0078] First, a candidate suggestion set is formed based on multiple executable maintenance suggestions whose comprehensive quality scores reach a preset quality threshold. During the iterative optimization process, multiple (e.g., three) executable maintenance suggestions, all with comprehensive quality scores reaching the preset quality threshold, may be generated for the same structured maintenance problem. These executable maintenance suggestions are then aggregated to form a candidate suggestion set. Each executable suggestion in the candidate suggestion set is considered qualified in terms of specificity, relevance, executability, and simplicity.

[0079] Then, based on the business ranking dimension, the business value of each executable maintenance suggestion in the candidate suggestion set can be scored. Here, the business ranking dimension is used to reflect the key business indicators that, in addition to technical feasibility, need to be considered in the actual production and operation scenario of a hydropower station and affect the decision-making priority.

[0080] The business ranking dimensions include implementation cost, practical difficulty, improvement effect, and safety compliance. Implementation cost refers to the cost of manpower, materials, time, and other resources required to implement the maintenance recommendation; lower costs are better. Practical difficulty refers to the technical complexity of implementing the maintenance recommendation on-site and the skill requirements for personnel; lower difficulty is better. Improvement effect refers to the expected results of implementing the maintenance recommendation in resolving faults, improving equipment performance, or eliminating safety hazards; more significant effects are better. Safety compliance refers to whether all operational steps of the maintenance recommendation fully comply with the hydropower station's safety production regulations and industry standards; full compliance is preferred.

[0081] Here, the business value score is a comprehensive score used to quantify the overall business value of a technically competent maintenance recommendation.

[0082] Finally, the business value scores are ranked to obtain the ranking results. The executable maintenance suggestion with the highest business value score in the ranking results is selected as the executable maintenance suggestion. This selection process can be represented as: ; in, For the candidate suggestion set, This indicates that the maintenance recommendations are feasible.

[0083] The method provided in this invention, after screening out technically qualified suggestions, further introduces business ranking dimensions including implementation cost, practical difficulty, improvement effect, and safety compliance. Multiple qualified suggestions are then scored and ranked based on their business value. This elevates maintenance decisions from a "technically feasible" level to a "business-optimal" level, addressing the shortcomings of existing technical solutions that focus solely on technical implementation while neglecting multiple constraints such as cost, efficiency, and safety in actual engineering. This ensures that the final maintenance suggestions are not only technically accurate and executable but also possess comprehensive engineering application characteristics such as direct implementation, cost control, safety compliance, and clear results, greatly enhancing the practical value of the decision.

[0084] Based on the above embodiments, step 152 includes: Step 1521: Obtain the standardized individual scores of the executable maintenance recommendations on the dimensions of implementation cost, practical difficulty, improvement effect, and safety compliance. Step 1522: Based on the preset business weights, the standardized individual scores are weighted to obtain the business value score.

[0085] Specifically, firstly, standardized individual scores can be obtained for executable maintenance recommendations in terms of implementation cost, practical difficulty, improvement effect, and safety compliance.

[0086] Then, based on preset business weights, the scores of each standardized item can be weighted to obtain a business value score.

[0087] Among them, business weight The configuration can be flexibly adjusted according to the actual management needs of the hydropower station. The formula for calculating the comprehensive ranking score can be expressed as: ; in, Indicates business value score. Indicates implementation costs. Indicates the difficulty of practical operation. Indicates the improvement effect. This indicates that security and compliance are ensured, and each business weight meets the constraints: .

[0088] Based on any of the above embodiments Figure 2 This is the second flowchart illustrating the executable suggestion generation method for hydropower station equipment maintenance provided by the present invention, as shown below. Figure 2As shown, the process first obtains raw data from a hydropower station equipment maintenance text data source and preprocesses it using a hydropower vertical domain dataset construction module. This preprocessing includes text clearing, deduplication, standardization, and filtering of invalid characters to build a clean and standardized dataset. Next, a clustering agent and a text distillation module perform text embedding generation on the preprocessed data, converting the text into vectors. Then, vector clustering and cluster center calculation are performed, and finally, a highly condensed representative maintenance sample set is output through representative sample selection. This sample set is then fed into a problem extraction agent and a structured parsing module. This module extracts maintenance categories and specific problems through topic classification, and after deduplication and merging, forms a structured output, namely <maintenance topic, The inspection item is correct. Subsequently, the suggestion generation agent receives the structured question and, under the constraint of the inspection procedure, generates 3-4 initial suggestions. These suggestions undergo compliance screening to obtain initial candidate inspection suggestions. These suggestions enter the evaluation agent and the iterative optimization closed-loop module. This module calculates a comprehensive score based on four dimensions: specificity, relevance, executability, and simplicity, and determines whether the criteria are met. If the criteria are met, it outputs a high-quality suggestion after iterative optimization. If the criteria are not met, feedback correction triggers the regeneration of suggestions, and re-evaluation is performed until the criteria are met. Finally, the high-quality suggestions are sent to the ranking agent and the optimal suggestion output module. This module comprehensively scores the suggestions based on business dimensions such as cost, difficulty, effectiveness, and compliance, and selects the optimal solution to output the final executable inspection suggestion.

[0089] This invention deeply integrates a multi-agent collaborative architecture with hydropower maintenance scenarios, comprehensively improving the quality, efficiency, and engineering practicality of equipment maintenance suggestions. Core indicators such as the feasibility, specificity, and non-redundancy of maintenance suggestions achieve near-perfect scores of 92–95 / 100. The illusion problem of large language models is effectively suppressed, reducing the occurrence rate of illusions by more than 60%. The system can process massive amounts of maintenance text into usable suggestions in seconds, improving decision-making efficiency by more than 10 times compared to traditional manual methods, and significantly shortening the fault response and solution formulation cycle. This invention maintains stable output on both large and medium parameter language models, with the overall score difference between different models being less than 2 points. The model robustness is significantly better than a single LLM (Large Language Model) solution. The generated maintenance suggestions have clear steps, strong operability, and controllable implementation costs, and can be directly applied to daily maintenance, fault rectification, and preventive maintenance work in hydropower stations, possessing extremely high field implementation value. Meanwhile, this invention has excellent versatility across various fields, and can be stably adapted to the maintenance scenarios of all types of hydropower equipment, such as turbines, generators, governors, excitation systems, gate control devices, and monitoring sensors. The system supports independent deployment in resource-constrained environments such as power station field servers and edge terminals, and can complete the entire process calculation without relying on large cloud models, effectively reducing deployment and inference costs, and ensuring the privacy and security of sensitive information such as power station equipment operation and fault data.

[0090] The executable suggestion generation device for hydropower station equipment maintenance provided by the present invention will be described below. The executable suggestion generation device for hydropower station equipment maintenance described below can be referred to in correspondence with the executable suggestion generation method for hydropower station equipment maintenance described above.

[0091] Based on any of the above embodiments, the present invention provides an executable suggestion generation device for the maintenance of hydropower station equipment. Figure 3 This is a schematic diagram of the executable suggestion generation device for hydropower station equipment maintenance provided by the present invention, as shown below. Figure 3 As shown, the device includes: Acquisition unit 310 is used to acquire representative fault samples; The first generation unit 320 is used to generate structured maintenance problems based on the representative fault samples. The second generation unit 330 is used to generate at least one candidate maintenance suggestion for the structured maintenance problem. The scoring unit 340 is used to quantitatively score the candidate maintenance suggestions based on the target scoring indicators to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility and simplicity of the candidate maintenance suggestions. The correction unit 350 is used to correct the candidate maintenance suggestion based on the overall quality score when the overall quality score does not reach the preset quality threshold, and to re-execute the step of quantifying the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score, until the overall quality score reaches the preset quality threshold, and then use the final candidate maintenance suggestion as an executable maintenance suggestion.

[0092] The apparatus provided in this invention acquires representative fault samples and generates structured maintenance problems, then generates candidate maintenance suggestions. Based on target scoring indicators including specificity, relevance, executability, and simplicity, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score. If the comprehensive quality score does not reach a preset quality threshold, the candidate maintenance suggestions are corrected and re-scored until the comprehensive quality score reaches the preset quality threshold. Finally, the final candidate maintenance suggestions are output as executable maintenance suggestions. This invention, by quantitatively scoring the generated candidate maintenance suggestions based on target scoring indicators including specificity, relevance, executability, and simplicity, and by iteratively correcting the comprehensive quality score until the preset quality threshold is met, ensures that the final output maintenance suggestions have undergone multi-dimensional, standardized evaluation and closed-loop optimization. This improves the accuracy, executability, and standardization of maintenance suggestions, solving the problems of lack of executability and quantitative basis in traditional manual methods, and the uncontrollable quality and vague content when directly generated by a single model. This ensures the scientific and reliable nature of hydropower station equipment maintenance decisions.

[0093] Based on any of the above embodiments, the acquisition unit 310 specifically includes: A semantic vector unit is generated to obtain the original hydropower station equipment maintenance text set and generate the semantic vector corresponding to each maintenance text in the original hydropower station equipment maintenance text set. Clustering units are used to cluster the semantic vectors to obtain data clusters; A selection unit is used to select representative text from each of the data clusters based on the center of each data cluster; A determining unit is used to determine the representative fault sample based on the representative text.

[0094] Based on any of the above embodiments, the selection unit is specifically used for: Determine the cluster center vector of each of the data clusters; The maintenance text with the highest semantic similarity to the cluster center vector in each data cluster is selected as the representative text.

[0095] Based on any of the above embodiments, the cluster center vector is obtained by averaging all semantic vectors in the data cluster.

[0096] Based on any of the above embodiments, the first generating unit 320 is specifically used for: Determine the repair topics for the representative fault samples; From the representative fault samples, at least one maintenance item is extracted; The maintenance topic and the maintenance item are combined to generate the structured maintenance problem.

[0097] Based on any of the above embodiments, a sorting unit is further included, wherein the sorting unit specifically includes: The constituent unit is used to form a candidate suggestion set based on multiple executable maintenance suggestions that reach the preset quality threshold according to the comprehensive quality score; The scoring unit is used to score the business value of each executable maintenance suggestion in the candidate suggestion set based on the business ranking dimension, and obtain the business value score; the business ranking dimension includes implementation cost, practical difficulty, improvement effect and safety compliance; The sorting subunit is used to sort the business value scores of each business, obtain the sorting result, and take the executable maintenance suggestion with the highest business value score in the sorting result as the executable maintenance suggestion.

[0098] Based on any of the above embodiments, the scoring unit is specifically used for: Obtain standardized individual scores for the executable maintenance recommendations in the dimensions of implementation cost, practical difficulty, improvement effect, and safety compliance. Based on preset business weights, the standardized individual scores are weighted to obtain the business value score.

[0099] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can call logic instructions in the memory 430 to execute an executable suggestion generation method for hydropower station equipment maintenance. This method includes: acquiring representative fault samples; generating structured maintenance problems based on the representative fault samples; generating at least one candidate maintenance suggestion for the structured maintenance problem; quantifying and scoring the candidate maintenance suggestion based on a target scoring index to obtain a comprehensive quality score; the target scoring index includes the specificity, relevance, executability, and simplicity of the candidate maintenance suggestion; if the comprehensive quality score does not reach a preset quality threshold, correcting the candidate maintenance suggestion based on the comprehensive quality score, and re-executing the step of quantifying and scoring the candidate maintenance suggestion based on the target scoring index to obtain a comprehensive quality score, until the comprehensive quality score reaches the preset quality threshold, and finally using the final candidate maintenance suggestion as an executable maintenance suggestion.

[0100] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the executable suggestion generation method for hydropower station equipment maintenance provided by the above methods. The method includes: obtaining representative fault samples and generating structured maintenance problems based on the representative fault samples; generating at least one candidate maintenance suggestion for the structured maintenance problem; quantifying and scoring the candidate maintenance suggestion based on a target scoring index to obtain a comprehensive quality score; the target scoring index includes the specificity, relevance, executability, and simplicity of the candidate maintenance suggestion; if the comprehensive quality score does not reach a preset quality threshold, correcting the candidate maintenance suggestion based on the comprehensive quality score, and re-executing the step of quantifying and scoring the candidate maintenance suggestion based on the target scoring index to obtain a comprehensive quality score, until the comprehensive quality score reaches the preset quality threshold, and using the final candidate maintenance suggestion as an executable maintenance suggestion.

[0102] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for generating executable suggestions for the maintenance of hydropower station equipment provided by the methods described above. This method includes: acquiring representative fault samples; generating a structured maintenance problem based on the representative fault samples; generating at least one candidate maintenance suggestion for the structured maintenance problem; quantifying and scoring the candidate maintenance suggestion based on a target scoring index to obtain a comprehensive quality score; the target scoring index includes the specificity, relevance, executability, and simplicity of the candidate maintenance suggestion; if the comprehensive quality score does not reach a preset quality threshold, correcting the candidate maintenance suggestion based on the comprehensive quality score, and re-executing the step of quantifying and scoring the candidate maintenance suggestion based on the target scoring index to obtain a comprehensive quality score, until the comprehensive quality score reaches the preset quality threshold, and finally using the final candidate maintenance suggestion as an executable maintenance suggestion.

[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating executable suggestions for the maintenance of hydropower station equipment, characterized in that, include: Obtain representative fault samples; Based on the representative fault samples, structured maintenance problems are generated; Generate at least one candidate maintenance suggestion for the structured maintenance problem; Based on the target scoring indicators, the candidate maintenance suggestions are quantitatively scored to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility, and simplicity of the candidate maintenance suggestions. If the overall quality score does not reach the preset quality threshold, the candidate maintenance suggestion is corrected based on the overall quality score, and the step of quantifying the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score is re-executed until the overall quality score reaches the preset quality threshold. The final candidate maintenance suggestion is then used as an executable maintenance suggestion.

2. The method for generating executable suggestions for the maintenance of hydropower station equipment according to claim 1, characterized in that, The steps for obtaining the representative fault samples include: Obtain the original hydropower station equipment maintenance text set and generate semantic vectors corresponding to each maintenance text in the original hydropower station equipment maintenance text set; Cluster the semantic vectors to obtain data clusters; Based on the center of each data cluster, representative text is selected from each data cluster; The representative fault samples are determined based on the representative text.

3. The method for generating executable suggestions for the maintenance of hydropower station equipment according to claim 2, characterized in that, The step of selecting representative text from each data cluster based on the center of each data cluster includes: Determine the cluster center vector of each of the data clusters; The maintenance text with the highest semantic similarity to the cluster center vector in each data cluster is selected as the representative text.

4. The method for generating executable suggestions for the maintenance of hydropower station equipment according to claim 3, characterized in that, The cluster center vector is obtained by averaging all semantic vectors in the data cluster.

5. The method for generating executable recommendations for the maintenance of hydropower station equipment according to any one of claims 1 to 4, characterized in that, The generation of structured maintenance problems based on the representative fault samples includes: Determine the repair topics for the representative fault samples; From the representative fault samples, at least one maintenance item is extracted; The maintenance topic and the maintenance item are combined to generate the structured maintenance problem.

6. The method for generating executable recommendations for the maintenance of hydropower station equipment according to any one of claims 1 to 4, characterized in that, The step of selecting the final candidate maintenance recommendation as an executable maintenance recommendation further includes: Based on multiple executable maintenance suggestions that achieve the preset quality threshold in the comprehensive quality score, a candidate suggestion set is formed; Based on the business ranking dimension, each executable maintenance suggestion in the candidate suggestion set is scored for its business value, resulting in a business value score; the business ranking dimension includes implementation cost, practical difficulty, improvement effect, and safety compliance; The business value scores are sorted to obtain a sorting result, and the executable maintenance suggestion with the highest business value score in the sorting result is taken as the executable maintenance suggestion.

7. The method for generating executable suggestions for the maintenance of hydropower station equipment according to claim 6, characterized in that, The step of scoring the business value of each executable maintenance suggestion in the candidate suggestion set based on the business ranking dimension to obtain a business value score includes: Obtain standardized individual scores for the executable maintenance recommendations in the dimensions of implementation cost, practical difficulty, improvement effect, and safety compliance. Based on preset business weights, the standardized individual scores are weighted to obtain the business value score.

8. An executable suggestion generation device for the maintenance of hydropower station equipment, characterized in that, include: The acquisition unit is used to acquire representative fault samples. The first generation unit is used to generate structured maintenance problems based on the representative fault samples. The second generation unit is used to generate at least one candidate maintenance suggestion for the structured maintenance problem. The scoring unit is used to quantitatively score the candidate maintenance suggestions based on the target scoring indicators to obtain a comprehensive quality score; the target scoring indicators include the specificity, relevance, feasibility and simplicity of the candidate maintenance suggestions. The correction unit is used to correct the candidate maintenance suggestion based on the overall quality score when the overall quality score does not reach the preset quality threshold, and to re-execute the step of quantifying the candidate maintenance suggestion based on the target scoring index to obtain the overall quality score, until the overall quality score reaches the preset quality threshold, and then use the final candidate maintenance suggestion as an executable maintenance suggestion.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the executable suggestion generation method for hydropower station equipment maintenance as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the executable suggestion generation method for hydropower station equipment maintenance as described in any one of claims 1 to 7.

Images