Power distribution network system operation and maintenance method and device based on historical work order prediction

Abstract

The invention discloses a power distribution network system operation and maintenance method and device based on historical work order prediction, and relates to the field of power grid operation and maintenance management, and the method comprises the steps: carrying out the semantic analysis of fault description texts in a plurality of historical work orders of a power distribution network system to be operated and maintained, and constructing a fault feature data set, performing semantic analysis on the user description texts in the plurality of historical work orders to construct a plurality of user behavior characteristics; according to the current operation data of the power distribution network system, performing similarity analysis in combination with the fault feature data set, and determining a plurality of predicted faults; based on the first large model, determining a plurality of to-be-operated and to-be-maintained combinations of the plurality of predicted faults and the plurality of user behavior characteristics; and based on the first large model, generating an operation and maintenance scheme corresponding to each to-be-operated and maintained combination, and performing operation and maintenance on the power distribution network system according to the operation and maintenance scheme of each to-be-operated and maintained combination. According to the invention, intelligent operation and maintenance management can be carried out on the power distribution network system reasonably and accurately.

Description

Technical Field

[0001] This application relates to the field of power grid operation and maintenance management, and in particular to a method and apparatus for operation and maintenance of distribution network systems based on historical work order prediction. Background Technology

[0002] As the foundation of power supply, the safe and stable operation of the distribution network directly affects the power quality for grid users. Therefore, it is necessary to manage and maintain the distribution network to improve power supply quality. Currently, the distribution network contains a large number of terminal devices, and multiple operating and management systems operate in parallel. Consequently, the operation and maintenance of the distribution network system mainly relies on regular inspections and preventative testing, as well as reactive responses after faults occur. The latter, being reactive, lacks efficiency and initiative, and can only be repaired after the power system is damaged. Losses incurred between the onset of a power system fault and the completion of repairs are unavoidable. While the former involves proactive identification, it is difficult to comprehensively and efficiently monitor and maintain the entire distribution network. Furthermore, the former is usually based on the equipment's lifecycle and combined with human experience, lacking in-depth analysis and reasonable decision-making basis, leading to inaccurate actual operation and maintenance. All of the above existing operation and maintenance methods are manual and do not involve or consider intelligent operation and maintenance of the distribution network. Therefore, how to reasonably and accurately manage the intelligent operation and maintenance of the distribution network system to improve the power supply quality remains a pressing technical problem that needs to be solved in the current technology. Summary of the Invention

[0003] This application provides a method and apparatus for operation and maintenance of distribution network systems based on historical work order prediction, in order to solve the technical problem that the intelligent operation and maintenance management of existing distribution network systems lacks rationality and accuracy.

[0004] According to a first aspect of the embodiments of this application, a method for operation and maintenance of a distribution network system based on historical work order prediction is provided, including: Semantic analysis is performed on the fault description text in multiple historical work orders of the distribution network system to be maintained to construct a fault feature dataset. Semantic analysis is also performed on the user description text in the multiple historical work orders to construct multiple user behavior features. Based on the current operating data of the power distribution network system, and combined with the fault feature dataset, a similarity analysis is performed to identify multiple predicted faults. Based on the preset first major model, multiple combinations of predicted faults and user behavior characteristics to be maintained are determined. Based on the first large model, an operation and maintenance plan is generated for each combination to be operated and maintained, and the distribution network system is operated and maintained according to the operation and maintenance plan for each combination to be operated and maintained.

[0005] This application first performs semantic analysis on fault description text and user description text in multiple historical work orders of the distribution network system to construct a fault feature dataset and multiple user behavior features. This enables in-depth analysis of the fault characteristics of different faults when a fault occurs in the distribution network system, as well as the user behavior characteristics of different user behaviors that lead to faults in the distribution network system. This improves the rationality and accuracy of predicting and maintaining the distribution network system based on historical work orders. Then, based on the current operating data of the distribution network system, similarity analysis is performed in conjunction with the fault feature dataset to identify multiple predicted faults. Predicting and identifying multiple predicted faults through similarity analysis can quickly and accurately identify different faults in the current operating data that may lead to faults in the distribution network system. Next, based on a large model, multiple combinations of multiple predicted faults and multiple user behavior features are determined to generate maintenance plans corresponding to each combination. The large model assists in the selection of maintenance combinations and the decision-making of plans, improving the rationality and accuracy of maintenance combination selection and plan decisions. This improves the rationality of the operation and maintenance management of the distribution network system when subsequent operation and maintenance processing is carried out based on the maintenance plan of each combination.

[0006] In some embodiments of this application, the step of performing semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, and constructing a fault feature dataset, specifically includes: The fault description text in each historical work order of the power distribution network system is segmented into words to obtain the fault feature vector of each historical work order; Based on the fault feature vector of each historical work order, determine the fault type corresponding to each historical work order; Based on a pre-defined fault semantic rule base, the fault feature vector of each historical work order is semantically labeled to obtain the fault semantic label of each historical work order. Cluster analysis is performed on the fault feature vector, fault type, and fault semantic label of each historical work order to determine multiple fault feature clusters, and a fault feature dataset is constructed based on the multiple fault feature clusters.

[0007] This application first performs word segmentation on the fault description text in each historical work order to obtain a fault feature vector, then determines the fault type corresponding to each historical work order, and performs semantic tagging based on the fault semantic rule base to obtain the fault semantic label corresponding to each historical work order. Then, by performing cluster analysis on the fault feature vector, fault type and fault semantic label corresponding to each historical work order, the comprehensiveness of the cluster analysis can be improved, and the multiple fault feature clusters obtained by clustering can accurately reflect the fault characteristics of different faults when a fault occurs in the distribution network system. Thus, when constructing a fault feature dataset based on multiple fault feature clusters, the accuracy of the fault feature dataset is improved.

[0008] In some embodiments of this application, the step of performing semantic analysis on the user description text in the multiple historical work orders to construct multiple user behavior features specifically includes: Intent recognition is performed on the user description text in the multiple historical work orders to obtain the user intent of each historical work order; Based on the first major model, multiple predicted user behaviors are predicted for the user intent corresponding to each historical work order. Cluster analysis is performed on multiple predicted user behaviors for each historical work order to determine multiple predicted behavior clusters, and multiple user behavior features are obtained based on the multiple predicted behavior clusters.

[0009] This application first performs intent recognition on the user description text in each historical work order to obtain the user intent, and then predicts multiple predicted user behaviors corresponding to the user intent of each historical work order based on a large model, which can accurately predict multiple possible user behaviors for different user intents; then, by performing cluster analysis on the multiple predicted user behaviors of each historical work order, the multiple predicted behavior clusters obtained by clustering accurately reflect the user behavior characteristics of different user behaviors that lead to faults in the distribution network system, thereby improving the accuracy of each user behavior feature when obtaining multiple user behavior features based on multiple predicted behavior clusters.

[0010] In some embodiments of this application, the step of determining multiple predicted faults by performing similarity analysis based on the current operating data of the distribution network system and the fault feature dataset specifically includes: Based on the current operating data of the power distribution network system, the predicted abnormal operating data of the power distribution network system are determined; Based on a preset fault semantic rule base, the predicted abnormal operation data is semantically labeled to obtain predicted fault semantic tags. Calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and determine multiple predicted faults based on the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

[0011] This application first determines the predicted abnormal operation data based on the current operation data of the distribution network system, and obtains the predicted fault semantic label by semantic tagging based on the fault semantic rule base. Then, it calculates the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, thereby determining multiple predicted faults. This can quickly and accurately identify different faults in the current operation data that may cause faults in the distribution network system, thus providing a predictive data basis for the subsequent generation of operation and maintenance plans.

[0012] In some embodiments of this application, determining multiple combinations of predicted faults and multiple user behavior characteristics to be maintained based on a preset first large model specifically includes: The multiple predicted faults and the multiple user behavior features are fully combined to obtain multiple combinations to be predicted; Based on the first large model, the trigger probability of each combination to be predicted is predicted, and multiple combinations to be maintained are determined according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

[0013] This application first obtains multiple combinations to be predicted based on the full combination of multiple predicted faults and multiple user behavior characteristics. Then, it predicts the trigger probability of each combination to be predicted based on a large model, thereby determining multiple combinations to be maintained. The large model assists in the selection of maintenance combinations and the decision-making of solutions, thereby improving the rationality and accuracy of the selection of maintenance combinations and the decision-making of solutions.

[0014] According to a second aspect of the embodiments of this application, a distribution network system operation and maintenance device based on historical work order prediction is provided, including a historical work order analysis module, a fault prediction analysis module, an operation and maintenance combination determination module, and a scheme operation and maintenance processing module. The historical work order analysis module is used to perform semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, to construct a fault feature dataset, and to perform semantic analysis on the user description text in the multiple historical work orders, to construct multiple user behavior features. The fault prediction and analysis module is used to perform similarity analysis based on the current operating data of the power distribution network system and the fault feature dataset to determine multiple predicted faults. The operation and maintenance combination determination module is used to determine multiple operation and maintenance combinations from the full combination of the multiple predicted faults and the multiple user behavior features based on a preset first major model. The operation and maintenance processing module is used to generate an operation and maintenance plan corresponding to each combination to be operated and maintained based on the first large model, and to perform operation and maintenance processing on the distribution network system according to the operation and maintenance plan of each combination to be operated and maintained.

[0015] In some embodiments of this application, the historical work order analysis module includes a fault feature analysis submodule; the fault feature analysis submodule includes a fault feature processing unit, a fault type determination unit, a fault semantic tagging unit, and a fault clustering analysis unit; The fault feature processing unit is used to perform word segmentation on the fault description text in each historical work order of the distribution network system to obtain the fault feature vector of each historical work order. The fault type determination unit is used to determine the fault type corresponding to each historical work order based on the fault feature vector of each historical work order. The fault semantic tagging unit is used to semantically tag the fault feature vector of each historical work order based on a preset fault semantic rule base, so as to obtain the fault semantic tag of each historical work order. The fault clustering analysis unit is used to perform clustering analysis on the fault feature vector, fault type and fault semantic label of each historical work order, determine multiple fault feature clusters, and construct a fault feature dataset based on the multiple fault feature clusters.

[0016] In some embodiments of this application, the historical work order analysis module includes a user feature analysis submodule; the user feature analysis submodule includes a user intent recognition unit, a user behavior prediction unit, and a behavior clustering analysis unit; The user intent recognition unit is used to perform intent recognition on the user description text in the multiple historical work orders to obtain the user intent of each historical work order. The user behavior prediction unit is used to predict multiple predicted user behaviors corresponding to the user intent of each historical work order based on the first large model. The behavior clustering analysis unit is used to perform clustering analysis on multiple predicted user behaviors for each historical work order, determine multiple predicted behavior clusters, and obtain multiple user behavior features based on the multiple predicted behavior clusters.

[0017] In some embodiments of this application, the fault prediction and analysis module includes an abnormal data prediction unit, a fault tag prediction unit, and a fault prediction determination unit; The abnormal data prediction unit is used to determine the predicted abnormal operating data of the distribution network system based on the current operating data of the distribution network system. The fault label prediction unit is used to semantically label the predicted abnormal operation data based on a preset fault semantic rule base to obtain predicted fault semantic labels. The predicted fault determination unit is used to calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and to determine multiple predicted faults according to the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

[0018] In some embodiments of this application, the operation and maintenance combination determination module includes a prediction combination construction unit and an operation and maintenance combination determination unit; The prediction combination construction unit is used to perform a full combination of the multiple predicted faults and the multiple user behavior features to obtain multiple combinations to be predicted. The operation and maintenance combination determination unit is used to predict the trigger probability of each combination to be predicted based on the first large model, and determine multiple combinations to be operated and maintained according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

[0019] This application first performs semantic analysis on fault description text and user description text in multiple historical work orders of the distribution network system to construct a fault feature dataset and multiple user behavior features. This enables in-depth analysis of the fault characteristics of different faults when a fault occurs in the distribution network system, as well as the user behavior characteristics of different user behaviors that lead to faults in the distribution network system. This improves the rationality and accuracy of predicting and maintaining the distribution network system based on historical work orders. Then, based on the current operating data of the distribution network system, similarity analysis is performed in conjunction with the fault feature dataset to identify multiple predicted faults. Predicting and identifying multiple predicted faults through similarity analysis can quickly and accurately identify different faults in the current operating data that may lead to faults in the distribution network system. Next, based on a large model, multiple combinations of multiple predicted faults and multiple user behavior features are determined to generate maintenance plans corresponding to each combination. The large model assists in the selection of maintenance combinations and the decision-making of plans, improving the rationality and accuracy of maintenance combination selection and plan decisions. This improves the rationality of the operation and maintenance management of the distribution network system when subsequent operation and maintenance processing is carried out based on the maintenance plan of each combination. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a distribution network system operation and maintenance method based on historical work order prediction, as shown in some embodiments of this application. Figure 2 This is a modular structure diagram of a power distribution network system operation and maintenance device based on historical work order prediction, as shown in some embodiments of this application. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below in conjunction with the accompanying drawings are exemplary and are only used to explain some embodiments of this application, and should not be construed as limiting the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments shown in this application without inventive effort are within the protection scope of this application.

[0022] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, unless otherwise explicitly specified, "a plurality of" or "several" means two or more.

[0023] Currently, the operation and maintenance of distribution network systems mainly relies on regular inspections and preventative testing, as well as reactive responses after faults. The latter, being reactive, lacks efficiency and initiative, only addressing damage after it occurs, and losses incurred between the fault and repair completion are unavoidable. While the former involves proactive identification, it struggles to comprehensively and efficiently monitor and maintain the entire distribution network. Furthermore, it typically relies on equipment lifecycle analysis and human experience, lacking in-depth analysis and sound decision-making criteria, leading to inaccurate actual maintenance. All existing maintenance methods are manual and do not address or consider intelligent operation and maintenance of distribution networks. Therefore, how to rationally and accurately manage the intelligent operation and maintenance of distribution network systems to improve power supply quality remains a pressing technical problem to be solved.

[0024] Based on the above technical background, please refer to Figure 1 This application provides a distribution network system operation and maintenance method based on historical work order prediction, including steps S101 to S104, each step as follows: Step S101: Perform semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, construct a fault feature dataset, and perform semantic analysis on the user description text in the multiple historical work orders to construct multiple user behavior features.

[0025] In some embodiments of this application, the step of performing semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, and constructing a fault feature dataset, specifically includes: The fault description text in each historical work order of the power distribution network system is segmented into words to obtain the fault feature vector of each historical work order; Based on the fault feature vector of each historical work order, determine the fault type corresponding to each historical work order; Based on a pre-defined fault semantic rule base, the fault feature vector of each historical work order is semantically labeled to obtain the fault semantic label of each historical work order. Cluster analysis is performed on the fault feature vector, fault type, and fault semantic label of each historical work order to determine multiple fault feature clusters, and a fault feature dataset is constructed based on the multiple fault feature clusters.

[0026] Specifically, when semantically labeling fault feature vectors based on a fault semantic rule base, each component of the fault feature vector is usually fully matched with the fault features in the fault semantic rule base, and the semantic labels corresponding to the fault features that have passed the full match in the fault semantic rule base are determined as the fault semantic labels of the corresponding fault feature vector components.

[0027] In some embodiments of this application, when performing cluster analysis on fault feature vectors, fault types and fault semantic labels, unsupervised clustering is usually adopted, including but not limited to K-means clustering algorithm, density-based DBSCAN algorithm and hierarchical clustering analysis algorithm, with hierarchical clustering analysis algorithm being the preferred embodiment.

[0028] This application first performs word segmentation on the fault description text in each historical work order to obtain a fault feature vector, then determines the fault type corresponding to each historical work order, and performs semantic tagging based on the fault semantic rule base to obtain the fault semantic label corresponding to each historical work order. Then, by performing cluster analysis on the fault feature vector, fault type and fault semantic label corresponding to each historical work order, the comprehensiveness of the cluster analysis can be improved, and the multiple fault feature clusters obtained by clustering can accurately reflect the fault characteristics of different faults when a fault occurs in the distribution network system. Thus, when constructing a fault feature dataset based on multiple fault feature clusters, the accuracy of the fault feature dataset is improved.

[0029] In some embodiments of this application, the step of performing semantic analysis on the user description text in the multiple historical work orders to construct multiple user behavior features specifically includes: Intent recognition is performed on the user description text in the multiple historical work orders to obtain the user intent of each historical work order; Based on the first major model, multiple predicted user behaviors are predicted for the user intent corresponding to each historical work order. Cluster analysis is performed on multiple predicted user behaviors for each historical work order to determine multiple predicted behavior clusters, and multiple user behavior features are obtained based on the multiple predicted behavior clusters.

[0030] Specifically, when performing intent recognition on user description text, it can be based on traditional deep learning models, such as intent recognition based on the BERT model; or it can be based on large models, especially domain-specific large models obtained by fine-tuning a basic large model using historical power grid work order data.

[0031] In some embodiments of this application, when performing cluster analysis on multiple predicted user behaviors, unsupervised clustering is typically adopted, including but not limited to K-means clustering algorithm, density-based DBSCAN algorithm and hierarchical clustering analysis algorithm, with hierarchical clustering analysis algorithm being the preferred embodiment.

[0032] This application first performs intent recognition on the user description text in each historical work order to obtain the user intent, and then predicts multiple predicted user behaviors corresponding to the user intent of each historical work order based on a large model, which can accurately predict multiple possible user behaviors for different user intents; then, by performing cluster analysis on the multiple predicted user behaviors of each historical work order, the multiple predicted behavior clusters obtained by clustering accurately reflect the user behavior characteristics of different user behaviors that lead to faults in the distribution network system, thereby improving the accuracy of each user behavior feature when obtaining multiple user behavior features based on multiple predicted behavior clusters.

[0033] Step S102: Based on the current operating data of the power distribution network system and the fault feature dataset, perform similarity analysis to identify multiple predicted faults.

[0034] In some embodiments of this application, the step of determining multiple predicted faults by performing similarity analysis based on the current operating data of the distribution network system and the fault feature dataset specifically includes: Based on the current operating data of the power distribution network system, the predicted abnormal operating data of the power distribution network system are determined; Based on a preset fault semantic rule base, the predicted abnormal operation data is semantically labeled to obtain predicted fault semantic tags. Calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and determine multiple predicted faults based on the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

[0035] Specifically, when calculating the similarity between the predicted fault semantic label and each fault feature cluster, the similarity algorithm used includes, but is not limited to, cosine similarity, Pearson correlation coefficient, Euclidean distance and TF-IDF algorithm, with cosine similarity being the preferred implementation.

[0036] This application first determines the predicted abnormal operation data based on the current operation data of the distribution network system, and obtains the predicted fault semantic label by semantic tagging based on the fault semantic rule base. Then, it calculates the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, thereby determining multiple predicted faults. This can quickly and accurately identify different faults in the current operation data that may cause faults in the distribution network system, thus providing a predictive data basis for the subsequent generation of operation and maintenance plans.

[0037] Step S103: Based on the preset first major model, determine multiple combinations of the multiple predicted faults and the multiple user behavior characteristics to be maintained.

[0038] In some embodiments of this application, the first large model is specifically a domain large model obtained by fine-tuning the basic large model using historical work order data of the power grid. The basic large model includes, but is not limited to, Zhipu Qingyan, Qwen model, Deepseek, GPT-4, Gemini and Doubao large model.

[0039] In some embodiments of this application, determining multiple combinations of predicted faults and multiple user behavior characteristics to be maintained based on a preset first large model specifically includes: The multiple predicted faults and the multiple user behavior features are fully combined to obtain multiple combinations to be predicted; Based on the first large model, the trigger probability of each combination to be predicted is predicted, and multiple combinations to be maintained are determined according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

[0040] This application first obtains multiple combinations to be predicted based on the full combination of multiple predicted faults and multiple user behavior characteristics. Then, it predicts the trigger probability of each combination to be predicted based on a large model, thereby determining multiple combinations to be maintained. The large model assists in the selection of maintenance combinations and the decision-making of solutions, thereby improving the rationality and accuracy of the selection of maintenance combinations and the decision-making of solutions.

[0041] Step S104: Based on the first large model, generate an operation and maintenance plan corresponding to each combination to be operated and maintained, and perform operation and maintenance processing on the power distribution network system according to the operation and maintenance plan of each combination to be operated and maintained.

[0042] This application first performs semantic analysis on fault description text and user description text in multiple historical work orders of the distribution network system to construct a fault feature dataset and multiple user behavior features. This enables in-depth analysis of the fault characteristics of different faults when a fault occurs in the distribution network system, as well as the user behavior characteristics of different user behaviors that lead to faults in the distribution network system. This improves the rationality and accuracy of predicting and maintaining the distribution network system based on historical work orders. Then, based on the current operating data of the distribution network system, similarity analysis is performed in conjunction with the fault feature dataset to identify multiple predicted faults. Predicting and identifying multiple predicted faults through similarity analysis can quickly and accurately identify different faults in the current operating data that may lead to faults in the distribution network system. Next, based on a large model, multiple combinations of multiple predicted faults and multiple user behavior features are determined to generate maintenance plans corresponding to each combination. The large model assists in the selection of maintenance combinations and the decision-making of plans, improving the rationality and accuracy of maintenance combination selection and plan decisions. This improves the rationality of the operation and maintenance management of the distribution network system when subsequent operation and maintenance processing is carried out based on the maintenance plan of each combination.

[0043] For a method corresponding to the one described above, please refer to [link to relevant documentation]. Figure 2 The present application provides a distribution network system operation and maintenance device based on historical work order prediction, including a historical work order analysis module 210, a fault prediction analysis module 220, an operation and maintenance combination determination module 230, and a scheme operation and maintenance processing module 240. The historical work order analysis module 210 is used to perform semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, to construct a fault feature dataset, and to perform semantic analysis on the user description text in the multiple historical work orders, to construct multiple user behavior features. The fault prediction and analysis module 220 is used to perform similarity analysis based on the current operating data of the power distribution network system and the fault feature dataset to determine multiple predicted faults. The operation and maintenance combination determination module 230 is used to determine multiple operation and maintenance combinations from the full combination of the multiple predicted faults and the multiple user behavior features based on a preset first large model. The operation and maintenance processing module 240 is used to generate an operation and maintenance plan corresponding to each combination to be operated and maintained based on the first large model, and to perform operation and maintenance processing on the power distribution network system according to the operation and maintenance plan of each combination to be operated and maintained.

[0044] In some embodiments of this application, the historical work order analysis module 210 includes a fault feature analysis submodule; the fault feature analysis submodule includes a fault feature processing unit, a fault type determination unit, a fault semantic labeling unit, and a fault clustering analysis unit; The fault feature processing unit is used to perform word segmentation on the fault description text in each historical work order of the distribution network system to obtain the fault feature vector of each historical work order. The fault type determination unit is used to determine the fault type corresponding to each historical work order based on the fault feature vector of each historical work order. The fault semantic tagging unit is used to semantically tag the fault feature vector of each historical work order based on a preset fault semantic rule base, so as to obtain the fault semantic tag of each historical work order. The fault clustering analysis unit is used to perform clustering analysis on the fault feature vector, fault type and fault semantic label of each historical work order, determine multiple fault feature clusters, and construct a fault feature dataset based on the multiple fault feature clusters.

[0045] In some embodiments of this application, the historical work order analysis module 210 includes a user feature analysis submodule; the user feature analysis submodule includes a user intent recognition unit, a user behavior prediction unit, and a behavior clustering analysis unit; The user intent recognition unit is used to perform intent recognition on the user description text in the multiple historical work orders to obtain the user intent of each historical work order. The user behavior prediction unit is used to predict multiple predicted user behaviors corresponding to the user intent of each historical work order based on the first large model. The behavior clustering analysis unit is used to perform clustering analysis on multiple predicted user behaviors for each historical work order, determine multiple predicted behavior clusters, and obtain multiple user behavior features based on the multiple predicted behavior clusters.

[0046] In some embodiments of this application, the fault prediction and analysis module 220 includes an abnormal data prediction unit, a fault tag prediction unit, and a fault prediction determination unit; The abnormal data prediction unit is used to determine the predicted abnormal operating data of the distribution network system based on the current operating data of the distribution network system. The fault label prediction unit is used to semantically label the predicted abnormal operation data based on a preset fault semantic rule base to obtain predicted fault semantic labels. The predicted fault determination unit is used to calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and to determine multiple predicted faults according to the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

[0047] In some embodiments of this application, the operation and maintenance combination determination module 230 includes a prediction combination construction unit and an operation and maintenance combination determination unit; The prediction combination construction unit is used to perform a full combination of the multiple predicted faults and the multiple user behavior features to obtain multiple combinations to be predicted. The operation and maintenance combination determination unit is used to predict the trigger probability of each combination to be predicted based on the first large model, and determine multiple combinations to be operated and maintained according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

[0048] This application first performs semantic analysis on fault description text and user description text in multiple historical work orders of the distribution network system to construct a fault feature dataset and multiple user behavior features. This enables in-depth analysis of the fault characteristics of different faults when a fault occurs in the distribution network system, as well as the user behavior characteristics of different user behaviors that lead to faults in the distribution network system. This improves the rationality and accuracy of predicting and maintaining the distribution network system based on historical work orders. Then, based on the current operating data of the distribution network system, similarity analysis is performed in conjunction with the fault feature dataset to identify multiple predicted faults. Predicting and identifying multiple predicted faults through similarity analysis can quickly and accurately identify different faults in the current operating data that may lead to faults in the distribution network system. Next, based on a large model, multiple combinations of multiple predicted faults and multiple user behavior features are determined to generate maintenance plans corresponding to each combination. The large model assists in the selection of maintenance combinations and the decision-making of plans, improving the rationality and accuracy of maintenance combination selection and plan decisions. This improves the rationality of the operation and maintenance management of the distribution network system when subsequent operation and maintenance processing is carried out based on the maintenance plan of each combination.

[0049] It should be understood that the device provided in the embodiments of this application corresponds to the aforementioned method. The distribution network system operation and maintenance device based on historical work order prediction provided in the embodiments of this application can realize the distribution network system operation and maintenance method based on historical work order prediction provided in any embodiment of this application.

[0050] Adaptively, embodiments of this application also provide a computer device and a computer-readable storage medium.

[0051] The computer device includes: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; The processor executes the computer program to implement the power distribution network system operation and maintenance method based on historical work order prediction of this application.

[0052] The computer-readable storage medium stores multiple instructions, which are adapted for a processor to load and execute a distribution network system operation and maintenance method based on historical work order prediction according to this application.

[0053] The above description represents some embodiments of this application, providing a further detailed explanation of the purpose, technical solution, and beneficial effects of this application. It should be understood that the above-described embodiments of this application should not be construed as limiting this application. In particular, any changes, modifications, equivalent substitutions, and variations made by those skilled in the art within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for operation and maintenance of a distribution network system based on historical work order prediction, characterized in that, include: Semantic analysis is performed on the fault description text in multiple historical work orders of the distribution network system to be maintained to construct a fault feature dataset. Semantic analysis is also performed on the user description text in the multiple historical work orders to construct multiple user behavior features. Based on the current operating data of the power distribution network system, and combined with the fault feature dataset, a similarity analysis is performed to identify multiple predicted faults. Based on the preset first major model, multiple combinations of predicted faults and user behavior characteristics to be maintained are determined. Based on the first large model, an operation and maintenance plan is generated for each combination to be operated and maintained, and the distribution network system is operated and maintained according to the operation and maintenance plan for each combination to be operated and maintained.

2. The distribution network system operation and maintenance method based on historical work order prediction according to claim 1, characterized in that, The fault description texts in multiple historical work orders of the distribution network system to be maintained are semantically analyzed to construct a fault feature dataset, which specifically includes: The fault description text in each historical work order of the power distribution network system is segmented into words to obtain the fault feature vector of each historical work order; Based on the fault feature vector of each historical work order, determine the fault type corresponding to each historical work order; Based on a pre-defined fault semantic rule base, the fault feature vector of each historical work order is semantically labeled to obtain the fault semantic label of each historical work order. Cluster analysis is performed on the fault feature vector, fault type, and fault semantic label of each historical work order to determine multiple fault feature clusters, and a fault feature dataset is constructed based on the multiple fault feature clusters.

3. The distribution network system operation and maintenance method based on historical work order prediction according to claim 1, characterized in that, The semantic analysis of the user description text in the multiple historical work orders is performed to construct multiple user behavior features, specifically including: Intent recognition is performed on the user description text in the multiple historical work orders to obtain the user intent of each historical work order; Based on the first major model, multiple predicted user behaviors are predicted for the user intent corresponding to each historical work order. Cluster analysis is performed on multiple predicted user behaviors for each historical work order to determine multiple predicted behavior clusters, and multiple user behavior features are obtained based on the multiple predicted behavior clusters.

4. The distribution network system operation and maintenance method based on historical work order prediction according to claim 1, characterized in that, The step of determining multiple predicted faults by performing similarity analysis based on the current operating data of the distribution network system and the fault feature dataset specifically includes: Based on the current operating data of the power distribution network system, the predicted abnormal operating data of the power distribution network system are determined; Based on a preset fault semantic rule base, the predicted abnormal operation data is semantically labeled to obtain predicted fault semantic tags. Calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and determine multiple predicted faults based on the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

5. The distribution network system operation and maintenance method based on historical work order prediction according to claim 1, characterized in that, The first preset model determines multiple combinations of predicted faults and user behavior characteristics to be maintained, specifically including: The multiple predicted faults and the multiple user behavior features are fully combined to obtain multiple combinations to be predicted; Based on the first large model, the trigger probability of each combination to be predicted is predicted, and multiple combinations to be maintained are determined according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

6. A distribution network system operation and maintenance device based on historical work order prediction, characterized in that, It includes a historical work order analysis module, a fault prediction analysis module, an operation and maintenance combination determination module, and a solution operation and maintenance processing module; The historical work order analysis module is used to perform semantic analysis on the fault description text in multiple historical work orders of the distribution network system to be maintained, to construct a fault feature dataset, and to perform semantic analysis on the user description text in the multiple historical work orders, to construct multiple user behavior features. The fault prediction and analysis module is used to perform similarity analysis based on the current operating data of the power distribution network system and the fault feature dataset to determine multiple predicted faults. The operation and maintenance combination determination module is used to determine multiple operation and maintenance combinations from the full combination of the multiple predicted faults and the multiple user behavior features based on a preset first major model. The operation and maintenance processing module is used to generate an operation and maintenance plan corresponding to each combination to be operated and maintained based on the first large model, and to perform operation and maintenance processing on the distribution network system according to the operation and maintenance plan of each combination to be operated and maintained.

7. A distribution network system operation and maintenance device based on historical work order prediction according to claim 6, characterized in that, The historical work order analysis module includes a fault feature analysis submodule; The fault feature analysis submodule includes a fault feature processing unit, a fault type determination unit, a fault semantic labeling unit, and a fault clustering analysis unit; The fault feature processing unit is used to perform word segmentation on the fault description text in each historical work order of the distribution network system to obtain the fault feature vector of each historical work order. The fault type determination unit is used to determine the fault type corresponding to each historical work order based on the fault feature vector of each historical work order. The fault semantic tagging unit is used to semantically tag the fault feature vector of each historical work order based on a preset fault semantic rule base, so as to obtain the fault semantic tag of each historical work order. The fault clustering analysis unit is used to perform clustering analysis on the fault feature vector, fault type and fault semantic label of each historical work order, determine multiple fault feature clusters, and construct a fault feature dataset based on the multiple fault feature clusters.

8. A distribution network system operation and maintenance device based on historical work order prediction according to claim 6, characterized in that, The historical work order analysis module includes a user feature analysis submodule; the user feature analysis submodule includes a user intent recognition unit, a user behavior prediction unit, and a behavior clustering analysis unit. The user intent recognition unit is used to perform intent recognition on the user description text in the multiple historical work orders to obtain the user intent of each historical work order. The user behavior prediction unit is used to predict multiple predicted user behaviors corresponding to the user intent of each historical work order based on the first large model. The behavior clustering analysis unit is used to perform clustering analysis on multiple predicted user behaviors for each historical work order, determine multiple predicted behavior clusters, and obtain multiple user behavior features based on the multiple predicted behavior clusters.

9. A distribution network system operation and maintenance device based on historical work order prediction according to claim 6, characterized in that, The fault prediction and analysis module includes an abnormal data prediction unit, a fault tag prediction unit, and a fault prediction determination unit. The abnormal data prediction unit is used to determine the predicted abnormal operating data of the distribution network system based on the current operating data of the distribution network system. The fault label prediction unit is used to semantically label the predicted abnormal operation data based on a preset fault semantic rule base to obtain predicted fault semantic labels. The predicted fault determination unit is used to calculate the similarity between the predicted fault semantic label and each fault feature cluster in the fault feature dataset, and to determine multiple predicted faults according to the sorting sequence of the similarity between each fault feature cluster and the predicted fault semantic label from high to low.

10. A distribution network system operation and maintenance device based on historical work order prediction according to claim 6, characterized in that, The operation and maintenance combination determination module includes a prediction combination construction unit and an operation and maintenance combination determination unit; The prediction combination construction unit is used to perform a full combination of the multiple predicted faults and the multiple user behavior features to obtain multiple combinations to be predicted. The operation and maintenance combination determination unit is used to predict the trigger probability of each combination to be predicted based on the first large model, and determine multiple combinations to be operated and maintained according to the sorting sequence of the trigger probabilities of the combinations to be predicted from high to low.

Images