Energy storage system fault management method, device, equipment and medium

Abstract

The present disclosure provides a method, device, equipment and medium for energy storage system fault management, which relates to the field of energy storage system fault diagnosis, and the method includes: establishing an expert knowledge base for a target energy storage system; based on the fault management information of each fault type point in the expert knowledge base, according to the real-time collected operation data of the target energy storage system, determining the fault point during the operation of the target energy storage system; based on the association rule mining algorithm, determining the association relationship between each fault point during the operation of the target energy storage system; according to the association relationship between each fault point during the operation of the target energy storage system, generating a fault diagnosis report and fault handling suggestions through a generative intelligent question-answering model. The present disclosure quickly and accurately generates a fault diagnosis report, accurately identifies the fault points in the energy storage system and their mutual associations, and improves the efficiency and accuracy of fault handling.

Description

Technical Field

[0001] The present disclosure relates to the field of energy storage system fault diagnosis, and in particular to energy storage system fault management methods, devices, equipment and media. Background Art

[0002] With the rapid development of renewable energy, energy storage systems are increasingly being used in power systems. Energy storage systems can effectively balance power supply and demand, improve the stability and reliability of power grids, and have become an indispensable and important part of modern power systems. However, energy storage systems may encounter a variety of faults during operation, such as battery failure, converter failure, communication failure, etc. If these faults cannot be diagnosed and handled in a timely and accurate manner, they will seriously affect the overall performance of the system and even endanger the safe operation of the power grid.

[0003] Related technologies mainly rely on expert experience and pre-set rules, usually by monitoring system operating parameters, combined with expert expertise and experience, to identify and handle faults. However, with the increase in the scale and complexity of energy storage systems, traditional fault diagnosis methods have gradually exposed their limitations, and the level of experts directly affects the accuracy and efficiency of fault diagnosis. Moreover, the complexity of energy storage systems and the emergence of new types of faults make it difficult to fully cover all possible fault conditions relying solely on expert experience. Therefore, there is an urgent need for a more intelligent and automated fault management method to improve the accuracy of fault diagnosis and processing efficiency of energy storage systems and ensure the stable operation of the system. Summary of the invention

[0004] The present disclosure provides a method, device, equipment and medium for energy storage system fault management, aiming to solve the problems existing in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present disclosure is implemented as follows:

[0006] In a first aspect, an embodiment of the present disclosure provides a method for managing energy storage system faults, the method comprising:

[0007] Establishing an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in the point table of the target energy storage system;

[0008] Based on the fault management information of each fault type point in the expert knowledge base, and according to the real-time collected operation data of the target energy storage system, determine the fault point during the operation of the target energy storage system;

[0009] Based on the association rule mining algorithm, determining the association relationship between various fault points during the operation of the target energy storage system;

[0010] According to the correlation between various fault points during the operation of the target energy storage system, a fault diagnosis report and fault handling suggestions are generated through a generative intelligent question-answering model.

[0011] Optionally, establishing an expert knowledge base of the target energy storage system includes:

[0012] Establishing an initial expert knowledge base, wherein the content of the initial expert knowledge base includes fault management information, wherein the fault management information includes equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and handling method;

[0013] Importing the point table of the target energy storage system into the initial expert knowledge base;

[0014] Based on the fault vocabulary library, keyword recognition is performed on the point content in the point table, all fault-type points are determined from the point table, and each fault-type point is matched with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system. The fault vocabulary library is pre-established based on the fault standards and fault descriptions of traditional energy storage systems.

[0015] Optionally, each fault type point is matched with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system, including:

[0016] Performing vocabulary matching between each fault type point and the fault vocabulary library;

[0017] According to the result of the vocabulary matching, all the fault type points in the point table and the fault management information in the initial expert knowledge base are integrated to obtain the expert knowledge base of the target energy storage system.

[0018] Optionally, the method of performing keyword recognition on the point contents in the point table based on the fault vocabulary library to determine all fault type points from the point table includes:

[0019] Based on the fault vocabulary library, performing text segmentation and part-of-speech tagging on the point content in the point table;

[0020] Based on the named entity recognition technology, the entities in the point content are determined, and each entity in the point content is classified into fault type points and other type points.

[0021] Optionally, the performing vocabulary matching between each fault type point and the fault vocabulary library includes:

[0022] Directly matching the words in the text of the fault type point with the entries in the fault vocabulary library;

[0023] In the case where the direct matching fails, similarity matching is performed between the words in the text of the fault type point and the entries in the fault vocabulary library to obtain a vocabulary matching score, wherein the similarity matching includes Jaccard similarity matching and cosine similarity matching;

[0024] In the case where the vocabulary matching score is greater than a preset matching threshold, the vocabulary matching is completed.

[0025] Optionally, after establishing the expert knowledge base of the target energy storage system, the method further includes:

[0026] Collecting operating data of the target energy storage system in real time, wherein the operating data includes voltage, current and temperature;

[0027] The operation data is preprocessed, and the preprocessing includes denoising and standardization.

[0028] Optionally, based on the fault management information of each fault type point in the expert knowledge base and according to the real-time collected operation data of the target energy storage system, determining the fault point during the operation of the target energy storage system includes:

[0029] Extracting fault characteristic parameters from target operating data using a machine learning model, wherein the machine learning model is pre-trained using historical operating data;

[0030] Comparing the fault characteristic parameters of the target operation data with the fault management information of each fault type point defined in the expert knowledge base;

[0031] When the fault characteristic parameters of the target operation data meet the fault judgment condition and fault threshold of any fault type point defined in the expert knowledge base, the point corresponding to the target operation data is determined to be the fault point.

[0032] Optionally, determining the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm includes:

[0033] Establishing multiple item sets based on various fault points during the operation of the target energy storage system;

[0034] Determine a target frequent item set from the multiple item sets based on an association rule mining algorithm;

[0035] Generate association rules according to the target frequent item set, and calculate the confidence and lift of each association rule;

[0036] Filter out the association rules that meet the preset rule evaluation threshold requirements, and determine the association rules with the highest confidence and lift as the target association rules;

[0037] The association relationship between various fault points during the operation of the target energy storage system is determined according to the target association rule.

[0038] Optionally, according to the correlation between various fault points during the operation of the target energy storage system, a fault diagnosis report and fault handling suggestions are generated by a generative intelligent question-answering model, including:

[0039] Using the fault management information in the expert knowledge base as a training corpus of a generative intelligent question answering model to pre-train the generative intelligent question answering model;

[0040] Inputting the correlation between various fault points during the operation of the target energy storage system into the pre-trained generative intelligent question-answering model;

[0041] Through the generative intelligent question-answering model, a fault diagnosis report is generated that includes fault cause analysis, impact scope assessment, and fault handling suggestions. The fault handling suggestions include repair steps, alternative solutions, and preventive measures.

[0042] Optionally, the method further comprises:

[0043] Record and store user feedback on generated fault diagnosis reports and processing suggestions;

[0044] The feedback information is used to regularly perform self-learning training on the generative intelligent question-answering model, and the content and algorithm parameters of the generative intelligent question-answering model are iteratively adjusted.

[0045] Optionally, the determining a target frequent itemset from the multiple itemsets based on an association rule mining algorithm includes:

[0046] Establishing corresponding item sets for the fault points of the target energy storage system;

[0047] Generate all frequent itemsets based on the preset minimum support;

[0048] Generate a candidate item set through the frequent item set, scan the transaction database, and calculate the support of the candidate item set;

[0049] Based on the support of the candidate item set, a new frequent item set is generated;

[0050] Generate candidate item sets through new frequent item sets, scan the transaction database, and calculate the support of the candidate item sets;

[0051] Execution steps: generating new frequent item sets based on the support of the candidate item sets; until no new frequent item sets can be generated;

[0052] Construct the initial FP tree and generate the conditional pattern base for each frequent item set;

[0053] A target frequent item set is determined from the conditional pattern base.

[0054] In a second aspect, an embodiment of the present disclosure provides an energy storage system fault management device, the device comprising:

[0055] An establishment module is used to establish an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in the point table of the target energy storage system;

[0056] A fault point determination module, used to determine the fault point during the operation of the target energy storage system based on the fault management information of each fault type point in the expert knowledge base and the operating data of the target energy storage system collected in real time;

[0057] An association relationship determination module, used to determine the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm;

[0058] A generation module is used to generate a fault diagnosis report and fault handling suggestions through a generative intelligent question-answering model according to the correlation between various fault points during the operation of the target energy storage system.

[0059] Optionally, the establishing module includes:

[0060] An initial expert knowledge base establishment submodule is used to establish an initial expert knowledge base, the content of which includes fault management information, which includes equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and treatment method;

[0061] An import submodule, used for importing the point table of the target energy storage system into the initial expert knowledge base;

[0062] The matching submodule is used to perform keyword recognition on the point content in the point table based on the fault vocabulary library, determine all fault-type points from the point table, and match each fault-type point with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system. The fault vocabulary library is pre-established based on the fault standards and fault descriptions of the traditional energy storage system.

[0063] Optionally, the matching submodule includes:

[0064] A vocabulary matching unit, used for matching each fault type point with the fault vocabulary library;

[0065] An integration unit is used to integrate all fault type points in the point table and the fault management information in the initial expert knowledge base according to the result of the vocabulary matching, so as to obtain the expert knowledge base of the target energy storage system.

[0066] Optionally, the matching submodule includes:

[0067] A part-of-speech tagging unit, used for performing text segmentation and part-of-speech tagging on the point content in the point table based on the fault vocabulary library;

[0068] The entity determination unit is used to determine the entities in the point content based on the named entity recognition technology, and classify each entity in the point content into fault type points and other types of points.

[0069] Optionally, the vocabulary matching unit includes:

[0070] A direct matching subunit, used for directly matching the words in the text of the fault type point with the entries in the fault vocabulary library;

[0071] A similarity matching subunit is used to perform similarity matching on the words in the text of the fault type point and the entries in the fault vocabulary library to obtain a vocabulary matching score when the direct matching fails. The similarity matching includes Jaccard similarity matching and cosine similarity matching.

[0072] The judgment subunit is used to complete the vocabulary matching when the vocabulary matching score is greater than a preset matching threshold.

[0073] Optionally, the device further comprises:

[0074] A real-time acquisition module, used to acquire the operating data of the target energy storage system in real time, wherein the operating data includes voltage, current and temperature;

[0075] The preprocessing module is used to preprocess the operation data, and the preprocessing includes denoising and standardization.

[0076] Optionally, the fault point determination module includes:

[0077] An extraction submodule, for extracting fault characteristic parameters from target operation data using a machine learning model, wherein the machine learning model is pre-trained using historical operation data;

[0078] A comparison submodule, used for comparing the fault characteristic parameters of the target operation data with the fault management information of each fault type point defined in the expert knowledge base;

[0079] The fault point determination submodule is used to determine that the point corresponding to the target operation data is a fault point when the fault characteristic parameters of the target operation data meet the fault judgment conditions and fault thresholds of any fault type point defined in the expert knowledge base.

[0080] Optionally, the association relationship determination module includes:

[0081] An item set establishment submodule, used to establish multiple item sets based on various fault points during the operation of the target energy storage system;

[0082] A target frequent item determination submodule, used to determine a target frequent item set from the multiple item sets based on an association rule mining algorithm;

[0083] An association rule calculation submodule is used to generate association rules according to the target frequent item set and calculate the confidence and lift of each association rule;

[0084] The screening submodule is used to screen out association rules that meet the preset rule evaluation threshold requirements, and determine the association rules with the highest confidence and lift as target association rules;

[0085] The association relationship determination submodule is used to determine the association relationship between various fault points during the operation of the target energy storage system according to the target association rule.

[0086] Optionally, the generating module includes:

[0087] A pre-training submodule, used to use the fault management information in the expert knowledge base as a training corpus of the generative intelligent question-answering model to pre-train the generative intelligent question-answering model;

[0088] An input submodule, used to input the correlation relationship between various fault points during the operation of the target energy storage system into the pre-trained generative intelligent question-answering model;

[0089] The report generation submodule is used to generate a fault diagnosis report including fault cause analysis, impact scope assessment, and fault handling suggestions through the generative intelligent question-answering model. The fault handling suggestions include repair steps, alternative solutions, and preventive measures.

[0090] Optionally, the device further comprises:

[0091] The recording module is used to record and store the user's feedback information on the generated fault diagnosis report and processing suggestions;

[0092] The self-learning module is used to utilize the feedback information to regularly perform self-learning training on the generative intelligent question-answering model, and iteratively adjust the content and algorithm parameters of the generative intelligent question-answering model.

[0093] Optionally, the target frequent item determination submodule includes:

[0094] An establishing unit, used for establishing a corresponding item set for a fault point of the target energy storage system;

[0095] A first generating unit, used for generating all frequent item sets based on a preset minimum support;

[0096] A first computing unit, configured to generate a candidate item set from the frequent item set, scan a transaction database, and calculate the support of the candidate item set;

[0097] A second generating unit, used for generating a new frequent item set based on the support of the candidate item set;

[0098] A second calculation unit is used to generate a candidate item set through a new frequent item set, and scan the transaction database to calculate the support of the candidate item set;

[0099] The repeating execution unit is used to execute the steps of: generating a new frequent item set based on the support of the candidate item set; until no new frequent item set can be generated;

[0100] FP tree construction unit, used to construct the initial FP tree and generate a conditional pattern base for each frequent item set;

[0101] The target frequent item set determination unit is used to determine the target frequent item set from the conditional pattern base.

[0102] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the steps of the energy storage system fault management method when executed by the processor.

[0103] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the energy storage system fault management method are implemented.

[0104] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0105] The present disclosure uses an expert knowledge base and an association rule mining algorithm to accurately identify fault points and their interrelationships in the energy storage system, thereby quickly and accurately generating fault diagnosis reports, reducing reliance on manual judgment and improving the efficiency and accuracy of fault handling. Through a generative intelligent question-answering model, the present disclosure can automatically generate handling suggestions for specific faults, helping operation and maintenance personnel to quickly locate problems and take effective measures, reducing system downtime and ensuring the stable operation of the power grid. At the same time, the expert knowledge base is constantly updated and expanded to cover more fault types and handling experience, enhancing the system's ability to respond to new and complex faults, and ensuring that the energy storage system can maintain efficient and reliable operation under various operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for describing the embodiments of the present disclosure are briefly introduced below.

[0107] Figure 1 is a schematic diagram of steps of a method for managing energy storage system faults according to an embodiment of the present disclosure;

[0108] Figure 2 It is a structural block diagram of an energy storage system fault management device shown in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0109] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0110] The fault data of modern energy storage systems are complex and diverse, involving multiple points and fault types. The generative artificial intelligence model is pre-trained with a large amount of text data and can understand complex language patterns and semantic relationships. In the disclosed embodiment, the generative intelligent question-answering model can effectively process these complex data and semantic relationships and provide accurate diagnosis and analysis. And based on the generative intelligent question-answering model, a detailed fault diagnosis report is dynamically generated according to real-time data and fault association relationships, which has good application prospects for energy storage system alarms and troubleshooting.

[0111] Figure 1 FIG. 1 is a schematic diagram of the steps of a method for managing a fault of an energy storage system according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0112] Step S101 : establishing an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in a point table of the target energy storage system.

[0113] Step S102 : Based on the fault management information of each fault type point in the expert knowledge base and according to the real-time collected operation data of the target energy storage system, the fault point during the operation of the target energy storage system is determined.

[0114] Step S103: determining the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm.

[0115] Step S104, generating a fault diagnosis report and fault handling suggestions through a generative intelligent question-answering model according to the correlation between various fault points during the operation of the target energy storage system.

[0116] Involving step S101, the energy storage system in the embodiment of the present disclosure, that is, the energy storage power station, is a technical device that temporarily stores electrical energy through electrical energy storage equipment and releases it when needed. The energy storage system can be used to balance electricity supply and demand and improve the stability and reliability of the power grid. The target energy storage system is the energy storage system object that the present disclosure aims to monitor and troubleshoot. Taking into account that various faults may occur in the target energy storage system during operation, the expert knowledge base contains detailed information about these faults, including fault manifestations, causes and treatment methods, to help the system quickly and accurately identify and handle faults. Establishing an expert knowledge base is equivalent to providing a reference framework for operation and maintenance personnel, systematizing scattered fault information, and facilitating data analysis and fault prediction.

[0117] The point table is a table that records various monitoring point information in the energy storage system. Each point represents a monitoring parameter or state in the energy storage system, such as voltage, current, temperature, etc. In the embodiment of the present disclosure, the content in the point table includes the point number and point content obtained by telemetry and telesignaling. Each point can be classified into fault type points and other types of points according to its point content. The expert knowledge base defines the fault management information of each fault type point. According to the fault management information, the manifestation and cause of the fault can be analyzed, and a preliminary diagnosis can be made to provide subsequent fault handling suggestions.

[0118] In step S102, as described in step S101, the expert knowledge base contains detailed information on all potential fault points in the energy storage system. This information includes descriptions of the fault points, fault manifestations, possible fault causes, and recommended fault handling methods. Based on the fault management information, the characteristics of each fault point can be specifically known, such as the parameter range during normal operation, the manifestation of abnormal conditions, possible alarm thresholds, fault occurrence frequency, historical fault data, etc.

[0119] The target energy storage system is equipped with various sensors and monitoring devices, which continuously collect various operating data of the target energy storage system, such as voltage, current, temperature, frequency and other parameters. The real-time collected operating data is transmitted to the central monitoring system through the communication network and stored in the database for real-time analysis and processing. The real-time collected operating data is compared with the normal operating parameter range in the expert knowledge base. If the data of a certain point in the operating data exceeds the normal range, it is preliminarily judged that there may be a fault at this point. For data that exceeds the normal range, its specific manifestation is further analyzed and matched with the known fault features in the knowledge base. For example, if the fluctuation frequency and amplitude of a certain voltage point match the characteristics of a certain fault in the knowledge base, it can be determined that the point is a fault point. Based on the matched fault features and historical data, combined with the fault cause information in the knowledge base, the possible cause of the fault can be inferred. For example, an abnormal increase in temperature may be related to battery overheating or a heat dissipation system failure. After completing the fault feature matching and cause inference, the existence of the fault point during the operation of the target energy storage system is confirmed.

[0120] Involving step S103, in the fault diagnosis of the target energy storage system, the association rule algorithm is used to discover the relationship and rules between various fault points, so as to help identify potential fault modes and chains. The present disclosure takes into account that in the energy storage system, different fault points may not occur independently, but have a certain correlation. Understanding the correlation between fault points helps to locate and diagnose faults more quickly and accurately. For example, if a fault at point A is usually accompanied by a fault at point B, then when an abnormality occurs at point A, point B can be checked in particular.

[0121] Involving step S104, based on the fault point association relationship mined in step S103, a detailed fault diagnosis report and fault handling suggestions are generated through a generative intelligent question-answering model. Assume that there are the following fault points and their association relationships in the target energy storage system: Point A: battery temperature is too high; Point B: battery voltage is abnormal; Point C: cooling system failure; the association rule is that points A and C often appear at the same time, and point B usually appears after point A. Through the generative intelligent question-answering model, a detailed fault diagnosis report is generated according to the results of association rule mining. The fault diagnosis report can be shown in Table 1:

[0122]

[0123] Table 1 Fault diagnosis report

[0124] According to the diagnosis report, specific troubleshooting suggestions can be generated as shown in Table 2:

[0125]

[0126] Table 2 Troubleshooting suggestions

[0127] The present disclosure uses an expert knowledge base and an association rule mining algorithm to accurately identify fault points and their interrelationships in the energy storage system, thereby quickly and accurately generating fault diagnosis reports, reducing reliance on manual judgment and improving the efficiency and accuracy of fault handling. Through a generative intelligent question-answering model, the present disclosure can automatically generate handling suggestions for specific faults, helping operation and maintenance personnel to quickly locate problems and take effective measures, reducing system downtime and ensuring the stable operation of the power grid. At the same time, the expert knowledge base is constantly updated and expanded to cover more fault types and handling experience, enhancing the system's ability to respond to new and complex faults, and ensuring that the energy storage system can maintain efficient and reliable operation under various operating conditions.

[0128] Exemplarily, the establishment of an expert knowledge base for a target energy storage system comprises: establishing an initial expert knowledge base, the content of the initial expert knowledge base comprising fault management information, the fault management information comprising equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and processing method; importing a point table of the target energy storage system into the initial expert knowledge base; based on a fault vocabulary library, performing keyword recognition on the content of the points in the point table, determining all fault-type points from the point table, and matching each fault-type point with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system, wherein the fault vocabulary library is pre-established based on the fault standards and fault descriptions of traditional energy storage systems.

[0129] According to the existing national standards, industry standards and various specifications for energy storage power stations, an initial expert knowledge base is constructed. The content of the initial expert knowledge base includes the following parts: equipment type, the type of target energy storage system equipment, such as batteries, converters, cooling systems, etc.; alarm items, the alarm items that need to be monitored and detected by the target energy storage system, such as abnormal voltage, excessive temperature, communication failure, etc.; alarm level, the severity classification of the alarm, such as general alarm, important alarm, emergency alarm; fault judgment conditions, based on which conditions it is judged as a fault, such as exceeding a certain threshold, abnormal frequency, etc.; fault threshold, specific numerical limit, such as voltage exceeding 300V or temperature exceeding 70°C; fault manifestation, specific manifestation when the fault occurs, such as voltage fluctuation, sharp temperature rise, equipment shutdown, etc.; fault cause, various reasons that may lead to the fault, such as poor heat dissipation, battery aging, loose connection, etc.; processing method, the processing steps and methods for dealing with the fault, such as replacing parts, adjusting parameters, cleaning the radiator, etc.

[0130] Import the point table of the target energy storage system into the initial expert knowledge base. As mentioned above, it includes the point number and content of telesignaling and telemetry. The fault vocabulary library is pre-established based on the fault standards and fault descriptions of traditional energy storage systems, and contains keywords and phrases related to common faults, such as "overvoltage", "overtemperature", "short circuit", "fault", etc. Based on the fault vocabulary library, keyword recognition is performed on the point content in the point table, all fault-type points are determined from the point table, and each fault-type point is matched with the fault management information in the initial expert knowledge base, and the fault manifestation, fault cause and treatment method of the initial expert library are supplemented and improved during the matching process. Finally, the expert knowledge base of the target energy storage system is completely established, including detailed information such as the matched and improved equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and treatment method.

[0131] Exemplarily, each fault type point is matched with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system, including: vocabulary matching of each fault type point and the fault vocabulary base; according to the result of the vocabulary matching, all the fault type points in the point table and the fault management information in the initial expert knowledge base are integrated to obtain the expert knowledge base of the target energy storage system.

[0132] Match each fault type point with the fault vocabulary library. For example, the point table contains three points: point 1: battery 1 voltage is high; point 2: converter temperature is abnormal; point 3: cooling system failure. The fault vocabulary library includes the following items: high voltage, failure.

[0133] For the match of point 1, the keywords are identified as battery, voltage, and high, which are directly matched with "high voltage" in the fault vocabulary library and marked as the fault point 1. For the match of point 2, the keywords are identified as converter, temperature, and abnormal. There is no matching entry related to "temperature abnormality" in the fault vocabulary library. However, according to further judgment, the machine learning model or manual auxiliary judgment can be used here to determine that point 2 is a fault point, and it is still marked as a fault point.

[0134] Integrate the fault point and the fault management information in the initial expert knowledge base. For point 1, since the match is successful, it is assumed that the fault management information related to "high voltage" in the initial expert knowledge base is already detailed and does not need to be further supplemented and improved. However, for point 2, since no entry related to "abnormal temperature" is matched in the fault vocabulary library, the description, cause and treatment method of the "abnormal converter temperature" fault point are integrated into the expert knowledge base of the target energy storage system.

[0135] Through the above detailed keyword identification, matching and information integration steps, the expert knowledge base of the target energy storage system finally established can comprehensively cover various fault points in the system, provide detailed fault management information, and help improve the efficiency and accuracy of fault diagnosis and processing.

[0136] Exemplarily, based on the fault vocabulary library, keyword recognition is performed on the point content in the point table to determine all fault-type points from the point table, including: based on the fault vocabulary library, text segmentation and part-of-speech tagging are performed on the point content in the point table; based on named entity recognition technology, entities in the point content are determined, and each entity in the point content is classified into fault-type points and other types of points.

[0137] First, clarify the content of the fault vocabulary. Assume that the fault vocabulary contains the following entries: High voltage

[0138] , abnormal temperature, fault, low voltage, excessive current, communication interruption. The content of point table 1 is assumed to be as follows: battery 1 has high voltage. Next, the text is segmented and part-of-speech tagged for each point content. Segmentation is to split the text into words or phrases, and part-of-speech tagging is to mark the parts of speech of these words or phrases. For point 1, the segmentation results are: battery 1 / noun, voltage / noun, high / adjective. Named entity recognition (NER) is the process of identifying words or phrases with specific meanings in text. In the above example, entities related to faults need to be identified. The entity recognition results for point 1 are: battery 1 (equipment), high voltage (fault), and the entities in the point content are classified into fault-type points and other types of points. Specifically, the process of point content classification can be divided into seven main steps: "data preparation", "feature extraction", "selection of NER model", "model training", "model evaluation", "model adjustment and optimization" and "entity classification". The following is a detailed description of these seven main steps.

[0139] The first step is to collect samples containing a large amount of point table point text data. These samples should contain texts of fault descriptions and normal descriptions for training and testing named entity recognition (NER) models to accurately identify fault points in the energy storage system.

[0140] The second step is to extract text features in data preprocessing. Optionally, methods such as Bag of Words, TF-IDF, and part-of-speech features are used to extract representative text features for building model input to help the model better understand the information in the text and continue text segmentation and part-of-speech tagging.

[0141] In the third step, according to specific needs, choose rule-based methods, traditional machine learning models (such as conditional random fields CRF), or deep learning models (such as BiLSTM-CRF, Transformer, etc.) to perform named entity recognition tasks. Rule-based methods tend to be simple fault identification in specific fields, while deep learning models are more suitable for processing complex and diverse text descriptions.

[0142] In the fourth step, the prepared point table point text data is used to train the selected NER model. For deep learning models, hyperparameters need to be adjusted to obtain the best performance. During the training process, the model will learn how to extract fault-related entities from the text and classify them as fault-type points or other category points.

[0143] In the fifth step, the trained model is evaluated using an independent test set. Evaluation indicators include accuracy, recall, and F1 value, which are used to measure the overall performance of the model, especially its performance in the fault point recognition task. The evaluation results can reflect the effect of the model in real application scenarios and guide subsequent model optimization.

[0144] The sixth step is to adjust and optimize the model based on the evaluation results. For example, the performance of the model can be improved by increasing the amount of data, improving the feature extraction method, adjusting the model structure or hyperparameters, etc. Continuous model optimization can ensure that the model can still maintain efficient and accurate recognition capabilities when facing new point table point text.

[0145] Finally, in the seventh step, the new point table point text is input into the trained NER model, and the model will classify the entities in the text to determine whether these entities belong to fault-type points. For example, the text "single battery temperature is too high" and "overvoltage protection" are identified as fault-type points by the model, while descriptions such as "battery cell temperature" are classified as other categories.

[0146] Through text segmentation, part-of-speech tagging and named entity recognition technology, the point content in the point table is classified to determine all fault-type points. The steps from text processing to entity recognition and then to fault point classification in this disclosure ensure that key information related to the fault can be accurately extracted and identified from the point table, providing data support for subsequent fault management and processing.

[0147] Exemplarily, the vocabulary matching of each fault type point and the fault vocabulary library includes: directly matching the vocabulary in the text of the fault type point with the entries in the fault vocabulary library; when the direct matching fails, similarity matching of the vocabulary in the text of the fault type point with the entries in the fault vocabulary library to obtain a vocabulary matching score, wherein the similarity matching includes Jaccard similarity matching and cosine similarity matching; when the vocabulary matching score is greater than a preset matching threshold, the vocabulary matching is completed.

[0148] Directly match the words in the text of the fault point with the entries in the fault vocabulary library, that is, match directly through entry search. When direct matching fails, that is, in the case of incomplete matching, use similarity matching algorithms, such as Jaccard similarity and cosine similarity, to determine the vocabulary matching score. Assume that a new point table content is added, point 4: battery pack voltage abnormality, there is no "voltage abnormality" entry in the fault vocabulary library, so direct matching fails, and similarity matching is performed at this time.

[0149] In the disclosed embodiment, similarity matching includes Jaccard similarity matching and cosine similarity matching. First, the Jaccard similarity matching method is described. The Jaccard similarity matching calculation formula is:

[0150]

[0151] In the formula, A and B are the entries in the fault vocabulary and the vocabulary set in the text, respectively. The segmentation results of point 4 are: battery pack / noun, voltage / noun, abnormal / adjective. The similarity calculation is performed with "high voltage" in the fault vocabulary. The set of segmentation results is: {battery pack, voltage, abnormal}, and the segmentation result of the fault vocabulary "high voltage" is: {voltage, high}. Jaccard similarity = |{voltage}| / |{battery pack, voltage, abnormal} ∪ {voltage, high}| = 1 / 4 = 0.25.

[0152] Next, the cosine similarity matching method is described. The Jaccard similarity calculation formula is:

[0153]

[0154] In the formula, A and B are two vectors, is their dot product, and is their norm. The point 4 vocabulary is vectorized into vector A: [0, 1, 1] (battery pack, voltage, abnormality), and the fault vocabulary "high voltage" is vectorized into vector B: [0, 1, 1] (battery pack, voltage, high), cosine similarity = = 0.5. Optionally, the final vocabulary matching score is obtained by weighting the matching results of Jaccard similarity and cosine similarity:

[0155]

[0156] In the formula, and is the weight of Jaccard similarity and cosine similarity, which can be adjusted according to the actual situation, and is generally set to 0.5. If calculated according to the weight ratio of 0.5 and 0.5, the vocabulary matching score of point 4 is 0.25 0.5+0.5 0.5=0.375. Assuming the preset matching threshold is 0.3, the vocabulary matching score of point 4 is higher than the preset matching threshold, and the vocabulary matching is successful. Point 4 "battery pack voltage abnormality" is marked as a fault point, and the fault-related vocabulary "voltage high" is matched.

[0157] The present disclosure can more accurately identify faulty points in the energy storage system point table by combining direct matching and similarity matching methods. In the event of direct matching failure, similarity matching provides a powerful tool to ensure that faulty points can be accurately identified and processed even when the textual expressions are slightly different. The use of multi-level matching methods improves the sensitivity and accuracy of fault detection, providing a reliable data basis for subsequent fault diagnosis and processing.

[0158] Exemplarily, after establishing the expert knowledge base of the target energy storage system, the method further includes:

[0159] The operating data of the target energy storage system is collected in real time, wherein the operating data includes voltage, current and temperature; and the operating data is preprocessed, wherein the preprocessing includes denoising and standardization.

[0160] After establishing the expert knowledge base of the target energy storage system, the operating data of the target energy storage system is collected in real time, and the collected data is preprocessed, including denoising and standardization.

[0161] Assume that the operating data collected in real time during the operation of the target energy storage system are as follows:

[0162] Time point 1: voltage = 400V, current = 50A, temperature = 25°C; time point 2: voltage = 450V, current = 52A, temperature = 26°C; time point 3: voltage = 398V, current = 51A, temperature = 25.5°C. De-noising is to remove noise and outliers in the operating data to make the data smoother and more stable. The data collected at time point 2 has an abnormal spike value, and the voltage jumps to 450V instantly, which is obviously an outlier. Optionally, the denoising process uses the average moving method. The denoised operating data is: time point 1: voltage = 400V, current = 50A, temperature = 25°C; time point 2: voltage = 401V, current = 51A, temperature = 25.5°C; time point 3: voltage = 398V, current = 51A, temperature = 25.5°C. Next, the denoised operating data is standardized. The purpose of standardization is to convert data of different dimensions into the same dimensional range for easy comparison and analysis. Optionally, Z-score standardization and Min-Max standardization are used. Taking Z-score standardization as an example, the operating data is converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. First, the mean and standard deviation of each parameter in the denoised operating data are calculated. The voltage mean is 399.67V, and the standard deviation is 1.25V; the current mean is 50.67A, and the standard deviation is 0.47A; the temperature mean is 25.33°C, and the standard deviation is 0.29°C.

[0163] Then Z-score standardization is performed. Time point 1: voltage Z-score = (400-399.67) / 1.25 = 0.26, current Z-score = (50-50.67) / 0.47 = -1.43, temperature Z-score = (25-25.33) / 0.29 = -1.14; Time point 2: voltage Z-score = (401-399.67) / 1.25 = 1.06, current Z-score = (51-50.67) / 0.47 = 0.70, temperature Z-score = (25.5-25.33) / 0.29 = 0.59; Time point 3: voltage Z-score = (398-399.67) / 1.25 = -1.33, current Z-score = (51-50.67) / 0.47 = 0.70, temperature Z-score = (25.5-25.33) / 0.29 = 0.59. Finally, the final pre-processed operating data can be obtained from the results of the standardization process.

[0164] Exemplarily, based on the fault management information of each fault type point in the expert knowledge base, according to the operating data of the target energy storage system collected in real time, the fault point during the operation of the target energy storage system is determined, including: using a machine learning model to extract fault characteristic parameters from the target operating data, the machine learning model is pre-trained using historical operating data; comparing the fault characteristic parameters of the target operating data with the fault management information of each fault type point defined in the expert knowledge base; when the fault characteristic parameters of the target operating data meet the fault judgment conditions and fault thresholds of any fault type point defined in the expert knowledge base, determining that the point corresponding to the target operating data is a fault point.

[0165] Using a pre-trained machine learning model, fault feature parameters are extracted from the operating data of the target energy storage system collected in real time. The machine learning model is trained based on historical operating data and fault data as training samples. In the disclosed embodiment, the target operating data is the operating data that is currently being monitored and needs to be judged for faults. It is assumed that the pre-processed target operating data includes voltage: 390V, current: 55A, temperature: 30°C, and it is assumed that the fault management information definition of the expert knowledge base fully established in the previous text has the relevant content of fault type A: voltage characteristics: lower than 395V; current characteristics: higher than 50A; temperature characteristics: higher than 28°C; fault judgment conditions: voltage, temperature, and current meet the above characteristics at the same time; fault threshold: voltage <395V, current >50A, temperature >28°C.

[0166] The fault characteristic parameters extracted from the target operation data are: extracted voltage characteristic: 390V; extracted current characteristic: 55A; extracted temperature characteristic: 30°C. Comparing the fault characteristic parameters extracted from the target operation data with the fault management information in the expert knowledge base, the voltage characteristic: 390V<395V (meets the fault condition); the current characteristic: 55A>50A (meets the fault condition); the temperature characteristic: 30°C>28°C (meets the fault condition). According to the comparison results, the fault characteristic parameters (voltage, temperature, current) of the target operation data simultaneously meet the fault judgment conditions and fault thresholds of fault type A, so it can be determined that the point corresponding to the target operation data is the fault point.

[0167] Exemplarily, the method of determining the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm includes: establishing multiple item sets based on various fault points during the operation of the target energy storage system; determining a target frequent item set from the multiple item sets based on an association rule mining algorithm; generating association rules according to the target frequent item set, and calculating the confidence and lift of each association rule; screening out association rules that meet the requirements of a preset rule evaluation threshold, and determining the association rule with the highest confidence and lift as the target association rule; and determining the association relationship between various fault points during the operation of the target energy storage system according to the target association rule.

[0168] Exemplarily, the association rule mining algorithm is based on determining a target frequent item set from the multiple item sets, including: establishing a corresponding item set for the fault point of the target energy storage system; generating all frequent item sets based on a preset minimum support; generating a candidate item set through the frequent item set, scanning a transaction database, and calculating the support of the candidate item set; generating a new frequent item set based on the support of the candidate item set; generating a candidate item set through the new frequent item set, scanning the transaction database, and calculating the support of the candidate item set; executing the steps of: generating a new frequent item set based on the support of the candidate item set; until a new frequent item set cannot be generated; constructing an initial FP tree, generating a conditional pattern base for each frequent item set; and determining a target frequent item set from the conditional pattern base.

[0169] For each fault point during the operation of the target energy storage system, multiple item sets are established. Each item set contains one or more fault points. The target frequent item set is determined from the multiple established item sets using the association rule mining algorithm. These frequent item sets are screened out from the item set under the condition of meeting the preset minimum support. According to the preset rule evaluation threshold, the association rules that meet the requirements are screened out, and the association rules with the highest confidence and lift are determined as the target association rules. According to the target association rules, the association relationship between the various fault points during the operation of the target energy storage system is determined.

[0170] For example, assume that the target energy storage system includes two fault points A and B, and establish item sets {A} and {B}. Based on the preset minimum support (assuming it is 40%), the following five fault events are recorded during the operation of the energy storage system: event 1 {A}, event 2 {B}, event 3 {A, B}, event 4 {A}, event 5 {A, B}. Calculate the support of each item set: {A} appears 4 times, and the support is 4 / 5 (number of occurrences in the event / total number of events) = 80%; {B} appears 3 times, and the support is 3 / 5 = 60%. Since the minimum support is 40%, {A} and {B} are both frequent item sets. Using the frequent item set, generate a new frequent item set {A, B} as a candidate item set. Scan the transaction database and calculate the support of the candidate two-item set. Support ({A, B}) = 2 / 5 = 40%. Since the support of {A, B} is equal to the minimum support, {A, B} is also a frequent item set. Since it is impossible to generate new frequent item sets at this time, build an initial FP tree (Frequent Pattern Tree). Arrange all transactions in descending order of frequency, insert transactions into the FP tree one by one, and build a compact data structure to represent the pattern of frequent item sets. In this example, because no frequent two-item sets are generated, the construction of the FP tree stops. Assuming that an FP tree is generated in a more complex situation, the conditional pattern base can be extracted through the path of the frequent item set. After constructing the FP tree and generating the conditional pattern base, the target frequent item set is determined by analyzing the conditional pattern base. In this example, {A, B} is the target frequent item set, and association rules are generated from the target frequent item set: A → B, B → A. The confidence and lift of A → B are: confidence = support({A, B}) / support({A}) = (2 / 5) / (4 / 5) = 50%, lift = confidence(A → B) / support({B}) = 50% / 60% ≈ 0.83. The confidence and lift of B → A are: confidence = support({A, B}) / support({B}) = (2 / 5) / (3 / 5) = 66.67%, lift = confidence(B → A) / support({A}) = 66.67% / 80% ≈ 0.83.

[0171] Assume that the rule evaluation thresholds are confidence at least 50% and lift at least 1.0. The confidence of rule A → B is 50%, but the lift is 0.83, which does not meet the lift threshold. The confidence of rule B → A is 66.67%, but the lift is 0.83, which also does not meet the lift threshold. Therefore, according to the lift threshold, neither rule will be screened out. Since no rule meets the lift threshold, it is impossible to determine from this set of data that there is a significant association between A and B. Although no significant association was found in this simple example, in actual applications, by adjusting the minimum support and rule evaluation threshold, more valuable association rules can be mined to help personnel better understand and manage faults in energy storage systems.

[0172] Exemplarily, according to the correlation between various fault points during the operation of the target energy storage system, a fault diagnosis report and fault handling suggestions are generated through a generative intelligent question and answer model, including: inputting the correlation between various fault points during the operation of the target energy storage system into the generative intelligent question and answer model; through the generative intelligent question and answer model, according to the fault management information defined in the expert knowledge base and the classification results of each entity in the point content, a fault diagnosis report including fault cause analysis, impact scope assessment, and fault handling suggestions is generated, and the fault handling suggestions include repair steps, alternative plans and preventive measures.

[0173] The pre-training of the generative intelligent question-answering model uses the information in the expert knowledge base as training data to conduct preliminary training on the model, so that the model can understand and process natural language input related to energy storage system failures, thereby having the ability to generate detailed diagnostic reports. The trained generative intelligent question-answering model can automatically process complex fault-related information, infer the cause of the fault, assess the scope of impact, and provide appropriate handling suggestions.

[0174] The mined fault point association relationship is input into the pre-trained generative intelligent question answering model. Assume that the generative intelligent question answering model uses a generative language model that can process natural language input and generate a detailed report. The following is an example of a simplified text instruction input into a generative language model in an embodiment of the present disclosure:

[0175] <The relationship between the fault points is as follows: Point A (battery voltage abnormality) → Point B (inverter temperature abnormality), confidence 70%, improvement 1.2. The fault management information of the expert knowledge base is as follows: Battery voltage abnormality: fault cause (poor heat dissipation, battery aging), treatment method (replace parts, adjust parameters); Inverter temperature abnormality: fault cause (poor heat dissipation, cooling system failure), treatment method (clean radiator, check cooling system). Please generate a fault diagnosis report and treatment suggestions. >, the generative intelligent question answering model will generate a detailed fault diagnosis report according to the text instructions. The display content can be in the form shown in Table 1 above, and can further include fault cause analysis, impact range assessment, and fault handling suggestions. Specifically, for the above example, the content of the fault cause analysis output by the generative intelligent question answering model can be: <There is a strong correlation between battery voltage abnormality (point A) and inverter temperature abnormality (point B) (confidence 70%). The main reason may be poor heat dissipation, which causes the temperature of the battery and inverter to rise at the same time. >

[0176] The impact range assessment output by the generative intelligent question-answering model can be: <When the battery voltage is abnormal, the converter temperature may also be abnormal, which will affect the operating stability of the entire energy storage system and may cause system shutdown or reduced efficiency. >

[0177] The fault handling suggestions output by the generative intelligent question answering model can be:

[0178] Repair steps: Turn off the battery power to ensure safety; replace aging or faulty batteries; adjust the battery voltage to the normal range. Check the inverter radiator and clean the dust; check the cooling system and repair or replace faulty parts.

[0179] Alternative solution: Add backup batteries and converters to the system to cope with unexpected failures. Install an independent temperature monitoring system to monitor and warn of temperature anomalies in real time.

[0180] Preventive measures: Regularly check and maintain batteries and converters, clean radiators, and ensure the normal operation of the system. Train operation and maintenance personnel, conduct regular emergency drills, and improve the response speed and accuracy of handling faults. >

[0181] Through association rule mining and generative intelligent question-answering models, fault diagnosis reports and processing suggestions can be efficiently generated, which not only improves the accuracy and timeliness of fault diagnosis, but also provides a systematic reference framework for operation and maintenance personnel to ensure the stable operation of the energy storage system.

[0182] Exemplarily, the method also includes: recording and storing user feedback information on the generated fault diagnosis report and processing suggestions; using the feedback information to regularly perform self-learning training on the generative intelligent question and answer model, and iteratively adjust the content and algorithm parameters of the generative intelligent question and answer model.

[0183] Feedback information is used to train the generative intelligent question-answering model. Based on the feedback information, the model's content generation and algorithm parameters are optimized to make it more in line with user needs. Through training with a large amount of feedback information, the model parameters are adjusted, including enhancing the ability to identify and process common keywords in user comments, optimizing the content of the preventive measures section in the generated report, and adjusting the model's generation strategy. Regular self-learning training is carried out, and the performance of the generative intelligent question-answering model is gradually improved through continuous iteration and optimization.

[0184] Through the above-mentioned self-learning training and iteration, the generative intelligent question-answering model is continuously optimized and improved, which can better meet user needs and improve the efficiency and accuracy of fault diagnosis and processing.

[0185] Figure 2 is a structural block diagram of an energy storage system fault management device shown in an embodiment of the present disclosure, such as Figure 2 As shown, the device comprises:

[0186] Establishing module 201, used to establish an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in the point table of the target energy storage system;

[0187] A fault point determination module 202 is used to determine the fault point during the operation of the target energy storage system based on the fault management information of each fault type point in the expert knowledge base and the operating data of the target energy storage system collected in real time;

[0188] The association relationship determination module 203 is used to determine the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm;

[0189] The generation module 204 is used to generate a fault diagnosis report and fault handling suggestions through a generative intelligent question-answering model according to the correlation between various fault points during the operation of the target energy storage system.

[0190] Exemplarily, the establishment module includes:

[0191] An initial expert knowledge base establishment submodule is used to establish an initial expert knowledge base, the content of which includes fault management information, which includes equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and treatment method;

[0192] An import submodule, used for importing the point table of the target energy storage system into the initial expert knowledge base;

[0193] The matching submodule is used to perform keyword recognition on the point content in the point table based on the fault vocabulary library, determine all fault-type points from the point table, and match each fault-type point with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system. The fault vocabulary library is pre-established based on the fault standards and fault descriptions of the traditional energy storage system.

[0194] Exemplarily, the matching submodule includes:

[0195] A vocabulary matching unit, used for matching each fault type point with the fault vocabulary library;

[0196] An integration unit is used to integrate all fault type points in the point table and the fault management information in the initial expert knowledge base according to the result of the vocabulary matching, so as to obtain the expert knowledge base of the target energy storage system.

[0197] Exemplarily, the matching submodule includes:

[0198] A part-of-speech tagging unit, used for performing text segmentation and part-of-speech tagging on the point content in the point table based on the fault vocabulary library;

[0199] The entity determination unit is used to determine the entities in the point content based on the named entity recognition technology, and classify each entity in the point content into fault type points and other types of points.

[0200] Exemplarily, the vocabulary matching unit includes:

[0201] A direct matching subunit, used for directly matching the words in the text of the fault type point with the entries in the fault vocabulary library;

[0202] A similarity matching subunit is used to perform similarity matching on the words in the text of the fault type point and the entries in the fault vocabulary library to obtain a vocabulary matching score when the direct matching fails. The similarity matching includes Jaccard similarity matching and cosine similarity matching.

[0203] The judgment subunit is used to complete the vocabulary matching when the vocabulary matching score is greater than a preset matching threshold.

[0204] Exemplarily, the device further includes:

[0205] A real-time acquisition module, used to acquire the operating data of the target energy storage system in real time, wherein the operating data includes voltage, current and temperature;

[0206] The preprocessing module is used to preprocess the operation data, and the preprocessing includes denoising and standardization.

[0207] Exemplarily, the fault point determination module includes:

[0208] An extraction submodule, for extracting fault characteristic parameters from target operation data using a machine learning model, wherein the machine learning model is pre-trained using historical operation data;

[0209] A comparison submodule, used for comparing the fault characteristic parameters of the target operation data with the fault management information of each fault type point defined in the expert knowledge base;

[0210] The fault point determination submodule is used to determine that the point corresponding to the target operation data is a fault point when the fault characteristic parameters of the target operation data meet the fault judgment conditions and fault thresholds of any fault type point defined in the expert knowledge base.

[0211] Exemplarily, the association relationship determination module includes:

[0212] An item set establishment submodule, used to establish multiple item sets based on various fault points during the operation of the target energy storage system;

[0213] A target frequent item determination submodule, used to determine a target frequent item set from the multiple item sets based on an association rule mining algorithm;

[0214] An association rule calculation submodule is used to generate association rules according to the target frequent item set and calculate the confidence and lift of each association rule;

[0215] The screening submodule is used to screen out association rules that meet the preset rule evaluation threshold requirements, and determine the association rules with the highest confidence and lift as target association rules;

[0216] The association relationship determination submodule is used to determine the association relationship between various fault points during the operation of the target energy storage system according to the target association rule.

[0217] Exemplarily, the generating module includes:

[0218] A pre-training submodule, used to use the fault management information in the expert knowledge base as a training corpus of the generative intelligent question-answering model to pre-train the generative intelligent question-answering model;

[0219] An input submodule, used to input the correlation relationship between various fault points during the operation of the target energy storage system into the pre-trained generative intelligent question-answering model;

[0220] The report generation submodule is used to generate a fault diagnosis report including fault cause analysis, impact scope assessment, and fault handling suggestions through the generative intelligent question-answering model. The fault handling suggestions include repair steps, alternative solutions, and preventive measures.

[0221] Exemplarily, the device further includes:

[0222] The recording module is used to record and store the user's feedback information on the generated fault diagnosis report and processing suggestions;

[0223] The self-learning module is used to utilize the feedback information to regularly perform self-learning training on the generative intelligent question-answering model, and iteratively adjust the content and algorithm parameters of the generative intelligent question-answering model.

[0224] Exemplarily, the target frequent item determination submodule includes:

[0225] An establishing unit, used for establishing a corresponding item set for a fault point of the target energy storage system;

[0226] A first generating unit, used for generating all frequent item sets based on a preset minimum support;

[0227] A first computing unit, configured to generate a candidate item set from the frequent item set, scan a transaction database, and calculate the support of the candidate item set;

[0228] A second generating unit, used for generating a new frequent item set based on the support of the candidate item set;

[0229] A second calculation unit is used to generate a candidate item set through a new frequent item set, and scan the transaction database to calculate the support of the candidate item set;

[0230] The repeating execution unit is used to execute the steps of: generating a new frequent item set based on the support of the candidate item set; until no new frequent item set can be generated;

[0231] FP tree construction unit, used to construct the initial FP tree and generate a conditional pattern base for each frequent item set;

[0232] The target frequent item set determination unit is used to determine the target frequent item set from the conditional pattern base.

[0233] The embodiment of the present disclosure also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, the various processes in the above-mentioned embodiment of the energy storage system fault management method are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0234] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process in the above-mentioned energy storage system fault management method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the embodiments can be referred to each other.

[0235] It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, equipment and media. Therefore, the embodiments of the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present disclosure may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0236] The embodiments of the present disclosure are described with reference to the flowcharts and / or block diagrams of the methods and devices according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1These computer program instructions can also be loaded into a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0237] Although the preferred embodiments of the present disclosure have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present disclosure.

[0238] Finally, it should be noted that, in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "include..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements. The above describes in detail the energy storage system fault management method, device, equipment and medium provided by the present disclosure. Specific examples are used in this article to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method and its core idea of ​​the present disclosure; at the same time, for those of ordinary skill in the art, according to the idea of ​​the present disclosure, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

Claims

1. A method for managing energy storage system faults, characterized in that: The method comprises: Establishing an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in the point table of the target energy storage system; Based on the fault management information of each fault type point in the expert knowledge base, and according to the real-time collected operation data of the target energy storage system, determine the fault point during the operation of the target energy storage system; Based on the association rule mining algorithm, determining the association relationship between various fault points during the operation of the target energy storage system; According to the correlation between various fault points during the operation of the target energy storage system, a fault diagnosis report and fault handling suggestions are generated through a generative intelligent question-answering model; The establishing of the expert knowledge base of the target energy storage system comprises: Establishing an initial expert knowledge base, wherein the content of the initial expert knowledge base includes fault management information, wherein the fault management information includes equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and handling method; Importing the point table of the target energy storage system into the initial expert knowledge base, wherein the content of the point table includes the point number and point content obtained by telemetry and telesignaling; Based on the fault vocabulary library, keyword recognition is performed on the point content in the point table, all fault-type points are determined from the point table, and each fault-type point is matched with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system, wherein the fault vocabulary library is pre-established according to the fault standards and fault descriptions of the traditional energy storage system; According to the correlation between various fault points during the operation of the target energy storage system, a fault diagnosis report and fault handling suggestions are generated through a generative intelligent question-answering model, including: Using the fault management information in the expert knowledge base as a training corpus of a generative intelligent question answering model to pre-train the generative intelligent question answering model; Inputting the correlation between various fault points during the operation of the target energy storage system into the pre-trained generative intelligent question-answering model; Through the generative question-answering model, a fault diagnosis report is generated that includes fault cause analysis, impact scope assessment, and fault handling suggestions. The fault handling suggestions include repair steps, alternative solutions, and preventive measures.

2. The method according to claim 1, characterized in that Each fault type point is matched with the fault management information in the initial expert knowledge base to obtain the expert knowledge base of the target energy storage system, including: Performing vocabulary matching between each fault type point and the fault vocabulary library; According to the result of the vocabulary matching, all the fault type points in the point table and the fault management information in the initial expert knowledge base are integrated to obtain the expert knowledge base of the target energy storage system.

3. The method according to claim 1, characterized in that Based on the fault vocabulary library, keyword recognition is performed on the point content in the point table to determine all fault type points from the point table, including: Based on the fault vocabulary library, performing text segmentation and part-of-speech tagging on the point content in the point table; Based on the named entity recognition technology, the entities in the point content are determined, and each entity in the point content is classified into fault type points and other type points.

4. The method according to claim 2, characterized in that: The step of matching each fault type point with the fault vocabulary library includes: Directly matching the words in the text of the fault type point with the entries in the fault vocabulary library; In the case where the direct matching fails, similarity matching is performed between the words in the text of the fault type point and the words in the fault vocabulary library to obtain a vocabulary matching score, wherein the similarity matching includes Jaccard similarity matching and cosine similarity matching; In the case where the vocabulary matching score is greater than a preset matching threshold, the vocabulary matching is completed.

5. The method according to claim 1, characterized in that After establishing an expert knowledge base of the target energy storage system, the method further includes: Collecting operating data of the target energy storage system in real time, wherein the operating data includes voltage, current and temperature; The operation data is preprocessed, and the preprocessing includes denoising and standardization.

6. The method according to claim 1, characterized in that Based on the fault management information of each fault type point in the expert knowledge base, and according to the real-time collected operation data of the target energy storage system, the fault point during the operation of the target energy storage system is determined, including: Extracting fault characteristic parameters from target operating data using a machine learning model, wherein the machine learning model is pre-trained using historical operating data; Comparing the fault characteristic parameters of the target operation data with the fault management information of each fault type point defined in the expert knowledge base; When the fault characteristic parameters of the target operation data meet the fault judgment condition and fault threshold of any fault type point defined in the expert knowledge base, the point corresponding to the target operation data is determined to be the fault point.

7. The method according to claim 1, characterized in that The method of determining the association relationship between various fault points during the operation of the target energy storage system based on the association rule mining algorithm includes: Establishing multiple item sets based on various fault points during the operation of the target energy storage system; Determine a target frequent item set from the multiple item sets based on an association rule mining algorithm; Generate association rules according to the target frequent item set, and calculate the confidence and lift of each association rule; Filter out the association rules that meet the preset rule evaluation threshold requirements, and determine the association rules with the highest confidence and lift as the target association rules; The association relationship between various fault points during the operation of the target energy storage system is determined according to the target association rule.

8. The method according to claim 1, characterized in that The method further comprises: Record and store user feedback on generated fault diagnosis reports and processing suggestions; The feedback information is used to regularly perform self-learning training on the generative intelligent question-answering model, and the content and algorithm parameters of the generative intelligent question-answering model are iteratively adjusted.

9. The method according to claim 7, characterized in that: The step of determining a target frequent item set from the multiple item sets based on an association rule mining algorithm includes: Establishing corresponding item sets for the fault points of the target energy storage system; Generate all frequent itemsets based on the preset minimum support; Generate a candidate item set through the frequent item set, scan the transaction database, and calculate the support of the candidate item set; Based on the support of the candidate item set, a new frequent item set is generated; Generate candidate item sets through new frequent item sets, scan the transaction database, and calculate the support of the candidate item sets; Execution steps: generating new frequent item sets based on the support of the candidate item sets; until no new frequent item sets can be generated; Construct the initial FP tree and generate the conditional pattern base for each frequent item set; A target frequent item set is determined from the conditional pattern base.

10. An energy storage system fault management device, characterized in that: The device comprises: An establishment module is used to establish an expert knowledge base of a target energy storage system, wherein the content of the expert knowledge base includes fault management information of each fault type point in the point table of the target energy storage system; A fault point determination module, used to determine the fault point during the operation of the target energy storage system based on the fault management information of each fault type point in the expert knowledge base and the operating data of the target energy storage system collected in real time; An association relationship determination module, used to determine the association relationship between various fault points during the operation of the target energy storage system based on an association rule mining algorithm; A generation module, configured to generate a fault diagnosis report and fault handling suggestions through a generative intelligent question-answering model according to the correlation between various fault points during the operation of the target energy storage system; The establishment module includes: An initial expert knowledge base establishment submodule is used to establish an initial expert knowledge base, the content of which includes fault management information, which includes equipment type, alarm item, alarm level, fault judgment condition, fault threshold, fault manifestation, fault cause and treatment method; An import submodule, used to import the point table of the target energy storage system into the initial expert knowledge base, wherein the content of the point table includes the point number and point content obtained by telemetry and telesignaling; A matching submodule, for performing keyword recognition on the point contents in the point table based on a fault vocabulary library, determining all fault-type points from the point table, and matching each fault-type point with the fault management information in the initial expert knowledge base to obtain an expert knowledge base of the target energy storage system, wherein the fault vocabulary library is pre-established based on the fault standards and fault descriptions of the traditional energy storage system; The generation module comprises: A pre-training submodule, used to use the fault management information in the expert knowledge base as a training corpus of the generative intelligent question-answering model to pre-train the generative intelligent question-answering model; An input submodule, used to input the correlation relationship between various fault points during the operation of the target energy storage system into the pre-trained generative intelligent question-answering model; The report generation submodule is used to generate a fault diagnosis report including fault cause analysis, impact scope assessment, and fault handling suggestions through the generative question-answering model. The fault handling suggestions include repair steps, alternative solutions, and preventive measures.

11. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the energy storage system fault management method according to any one of claims 1 to 9.

12. A computer-readable medium, characterized in that A computer program is stored thereon, wherein when the computer program is executed by a processor, the energy storage system fault management method as described in any one of claims 1 to 9 is implemented.

Images