Large model-based energy storage system maintenance method, apparatus and device, and medium

Abstract

The invention discloses an energy storage system maintenance method and device based on a large model, equipment and a medium, and relates to the technical field of energy storage system management.The method comprises the steps that obtained maintenance demand indication data is analyzed, and an analysis result is obtained; generating a maintenance request corresponding to the analysis result, and creating a request processing session corresponding to the maintenance request; inputting the maintenance request into the large model to obtain a target function output by the large model; analyzing retrieval data and a request state corresponding to the maintenance request by utilizing the target function to obtain a cue word; and inputting the cue word into the large model to obtain a response result which is output by the large model and corresponds to the maintenance demand indication data. The method is used for solving the problems of long consumed time, low fault response speed, poor fault positioning and the like when the energy storage system is maintained in the prior art, and real-time maintenance of the energy storage system and accurate positioning and quick response of the fault are realized.

Description

Technical Field

[0001] The present application relates to the technical field of energy storage system management, and in particular to a large model-based energy storage system maintenance method, device, equipment and medium. Background Art

[0002] With the rapid development of renewable energy and the deepening reform of the power market, energy storage technology has attracted widespread attention as a key means to balance the supply and demand of the power grid and improve energy efficiency. The energy storage system can not only smooth the fluctuations of renewable energy generation, but also provide additional power during peak hours, effectively alleviating the pressure on the power grid. However, the operation and maintenance of the energy storage system faces many challenges, especially in the distributed energy storage scenario, where the number of energy storage devices is huge and widely distributed, resulting in difficulty in fault location and high maintenance costs, which seriously affects the reliability and economy of the energy storage system.

[0003] At present, the maintenance of energy storage systems mainly relies on traditional manual inspections and fault reporting mechanisms. This method is not only time-consuming and labor-intensive, but also difficult to achieve rapid response and accurate handling of faults.

[0004] With the rise of big models, energy storage systems use big models to maintain energy storage systems. However, the existing big model-based energy storage system maintenance can only provide standard answers based on a preset knowledge base based on user questions, and cannot deeply understand the essence of the problem to provide effective solutions. Summary of the invention

[0005] In response to the above-mentioned problems and technical needs, the applicant has proposed a large-model-based energy storage system maintenance method, device, equipment and medium to solve the problems of long time consumption, slow fault response speed and poor fault location in the prior art when maintaining the energy storage system, and to achieve real-time maintenance of the energy storage system and accurate fault location and rapid response.

[0006] The embodiment of the present application provides a large model-based energy storage system maintenance method, the method comprising:

[0007] Acquiring maintenance demand indication data, and parsing the maintenance demand indication data to obtain parsing results;

[0008] generating a maintenance request corresponding to the parsing result, and creating a request processing session corresponding to the maintenance request, wherein the request processing session is used to record the request status of the maintenance request;

[0009] Inputting the maintenance request into a large model to obtain an objective function output by the large model;

[0010] Analyzing the search data corresponding to the maintenance request and the request status using the objective function to obtain a prompt word;

[0011] The prompt word is input into the large model to obtain a response result output by the large model corresponding to the maintenance requirement indication data.

[0012] According to an energy storage system maintenance method based on a large model in one embodiment of the present application, after inputting the maintenance request into the large model and obtaining the objective function output by the large model, the method further includes:

[0013] Based on the processing parameters corresponding to the objective function and the maintenance request, acquiring the search data from a preset knowledge base;

[0014] splicing the search data and the parsing result to obtain data to be processed;

[0015] The objective function is used to analyze the search data corresponding to the maintenance request and the request status to obtain prompt words, including:

[0016] The processing logic of the objective function is used to analyze the data to be processed and the request status to obtain the prompt word.

[0017] According to a large model-based energy storage system maintenance method according to an embodiment of the present application, a maintenance request corresponding to the analysis result is generated, and a request processing session corresponding to the maintenance request is created, including:

[0018] Generate a unique session identifier corresponding to the parsing result, and obtain a timestamp corresponding to the parsing result, a source channel of the maintenance requirement indication data, and user-related information;

[0019] Creating a request context based on the unique session identifier, the timestamp, the source channel, and the user-related information, and generating a maintenance request carrying the request context;

[0020] The maintenance request is input into a thread pool, and the request processing session and a tracking thread including the request processing session are obtained based on the thread pool.

[0021] According to a large model-based energy storage system maintenance method according to an embodiment of the present application, the maintenance request further includes: an application scenario;

[0022] Before inputting the maintenance request into the large model and obtaining the target function output by the large model, the method further includes:

[0023] Based on the application scenario, determining a target macro model corresponding to the maintenance request;

[0024] Inputting the maintenance request into the big model to obtain the objective function output by the big model includes:

[0025] The maintenance request is input into the target large model to obtain the target function output by the target large model.

[0026] According to a large model-based energy storage system maintenance method according to an embodiment of the present application, the large model includes: model attributes, the model attributes including: model magnitude and model understanding capability;

[0027] Based on the application scenario, determining a target macro model corresponding to the maintenance request includes:

[0028] When it is determined that the application scenario is a device control scenario, a large model with a model magnitude smaller than a preset magnitude is selected from a plurality of large models as the target large model;

[0029] When it is determined that the application scenario is a demand understanding consultation scenario, a large model whose model understanding ability is greater than a preset understanding ability is selected from multiple large models as the target large model.

[0030] According to a large-model-based energy storage system maintenance method according to an embodiment of the present application, the maintenance request is input into the large-model to obtain an objective function output by the large-model, including:

[0031] Input the maintenance request into the big model, and obtain and output the target function by matching the maintenance request with the description document of each function;

[0032] The description document includes the function's processing parameters, processing logic and processing scenarios.

[0033] According to an embodiment of the present application, the energy storage system maintenance method based on a large model further includes, after obtaining a response result corresponding to the maintenance demand indication data output by the large model:

[0034] The response result is sent to the smart terminal through any one or more of text, picture, video and control instructions.

[0035] The embodiment of the present application also provides a large model-based energy storage system maintenance device, the device comprising:

[0036] A parsing module, used to obtain maintenance demand indication data, and parse the maintenance demand indication data to obtain a parsing result;

[0037] a generating module, configured to generate a maintenance request corresponding to the parsing result, and to create a request processing session corresponding to the maintenance request, wherein the request processing session is used to record a request status of the maintenance request;

[0038] A first prediction module, used for inputting the maintenance request into a large model to obtain an objective function output by the large model;

[0039] An analysis module, configured to analyze the search data corresponding to the maintenance request and the request status using the objective function to obtain a prompt word;

[0040] The second prediction module is used to input the prompt word into the large model to obtain a response result output by the large model corresponding to the maintenance demand indication data.

[0041] An embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the large model-based energy storage system maintenance method as described in any one of the above items are implemented.

[0042] An embodiment of the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the large model-based energy storage system maintenance method as described in any one of the above items are implemented.

[0043] The energy storage system maintenance method, device, equipment and medium based on the big model provided in the embodiment of the present application obtain maintenance demand indication data and parse the maintenance demand indication data to obtain the parsing result; generate a maintenance request corresponding to the parsing result, and create a request processing session corresponding to the maintenance request. The present application pays attention to the request status of the maintenance request in real time through the request processing session, without manual maintenance, and the user can trigger the creation of the maintenance request based on his or her actual needs; then, the maintenance request is input into the big model to obtain the target function output by the big model; the retrieval data and request status corresponding to the maintenance request are analyzed by the target function to obtain the prompt word; the prompt word is input into the big model to obtain the response result corresponding to the maintenance demand indication data output by the big model. The present application determines the target function corresponding to the maintenance request through the big model to accurately analyze the retrieval data corresponding to the maintenance request, can accurately obtain the user's needs, and when the demand is a fault-type demand, can accurately analyze the fault and realize accurate positioning of the fault, and obtain the corresponding response result based on the big model, thereby realizing the purpose of real-time maintenance of the energy storage system and accurate positioning and rapid response of the fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1It is one of the flow diagrams of the large model-based energy storage system maintenance method provided in the embodiment of the present application;

[0046] Figure 2 This is the second flow chart of the large model-based energy storage system maintenance method provided in the embodiment of the present application;

[0047] Figure 3 It is a structural diagram of a maintenance system provided in an embodiment of the present application;

[0048] Figure 4 A schematic diagram of the structure of a large-model-based energy storage system maintenance device provided in an embodiment of the present application;

[0049] Figure 5 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Specifically, the application of existing large language model question answering systems in the field of energy storage still has the following limitations:

[0052] Single answer: Existing systems can only provide standard answers based on a preset knowledge base, lacking in-depth analysis and customized suggestions for specific issues of energy storage systems.

[0053] Shallow understanding: Although the large language model has powerful language processing capabilities, for complex operation and maintenance scenarios, such as battery status assessment and fault prediction, it is often difficult for the system to deeply understand the essence of the problem and provide effective solutions.

[0054] Limited interaction form: Current question-and-answer systems are mainly based on text communication, and lack support for multimedia information such as images and sounds commonly used in energy storage system operation and maintenance, and cannot meet users' needs for intuitive understanding of equipment status.

[0055] Passive response: Most existing operation and maintenance systems adopt a passive response mode, that is, they only process after the user raises a specific problem. They lack the ability to actively collect and analyze equipment operation data, making it difficult to achieve preventive maintenance.

[0056] In order to solve the above problems, the embodiment of the present application provides a large model-based energy storage system maintenance method, which can be applied to a server, an intelligent terminal, or an energy storage system maintenance system. Some other descriptions in the embodiment of the present application are for illustrative purposes only and are not intended to limit the scope of protection of the present application. They will not be described one by one in the following. The specific implementation of the method is as follows: Figure 1 As shown:

[0057] Step 101, obtaining maintenance requirement indication data, and parsing the maintenance requirement indication data to obtain parsing results.

[0058] The maintenance requirement indication data includes maintenance requirement indication data input by the user based on the client (data actively input by the user), and maintenance requirement indication data automatically triggered based on abnormal data of the energy storage system (automatically collected data).

[0059] Specifically, the maintenance requirement indication data is input into the large language model for understanding and semantic parsing, and the parsing result output by the large language model is obtained.

[0060] Step 102: Generate a maintenance request corresponding to the parsing result, and create a request processing session corresponding to the maintenance request.

[0061] The request processing session is used to record the request status of the maintenance request.

[0062] The request processing session describes the entire processing process from receiving a maintenance request to returning a response.

[0063] Step 103, input the maintenance request into the large model to obtain the target function output by the large model.

[0064] Among them, the large model is a pre-created large model of the energy storage field.

[0065] Step 104: Analyze the search data and request status corresponding to the maintenance request using the objective function to obtain prompt words.

[0066] Step 105, input the prompt word into the large model, and obtain the response result corresponding to the maintenance demand indication data output by the large model.

[0067] The energy storage system maintenance method based on the big model provided in the embodiment of the present application obtains maintenance demand indication data and parses the maintenance demand indication data to obtain a parsing result; generates a maintenance request corresponding to the parsing result, and creates a request processing session corresponding to the maintenance request. The present application pays attention to the request status of the maintenance request in real time through the request processing session, and no manual maintenance is required. The user can trigger the creation of the maintenance request based on his or her actual needs; then, the maintenance request is input into the big model to obtain the target function output by the big model; the retrieval data and request status corresponding to the maintenance request are analyzed using the target function to obtain a prompt word; the prompt word is input into the big model to obtain a response result corresponding to the maintenance demand indication data output by the big model. The present application determines the target function corresponding to the maintenance request through the big model to accurately analyze the retrieval data corresponding to the maintenance request, and can accurately obtain the user's needs. When the demand is a fault-type demand, the fault can be accurately analyzed to achieve accurate positioning of the fault, and a corresponding response result is obtained based on the big model, thereby achieving the purpose of real-time maintenance of the energy storage system and accurate positioning and rapid response of the fault.

[0068] In a specific embodiment, the specific implementation of obtaining the maintenance requirement indication data includes:

[0069] Obtain maintenance demand indication data input by a user based on a client; and / or obtain maintenance demand indication data automatically triggered based on abnormal data of the energy storage system.

[0070] Among them, the input of the present application includes a multimodal input interface, which can receive information such as pictures, text, and voice. This information can be actively input by the user or automatically collected.

[0071] In a specific embodiment, the specific implementation of generating a maintenance request corresponding to the parsing result and creating a request processing session corresponding to the maintenance request includes:

[0072] Generate a unique session identifier corresponding to the parsing result, and obtain the timestamp corresponding to the parsing result, the source channel of the maintenance requirement indication data, and user-related information; create a request context based on the unique session identifier, timestamp, source channel, and user-related information, and generate a maintenance request carrying the request context; input the maintenance request into the thread pool, and obtain the request processing session based on the thread pool, as well as the tracking thread including the request processing session.

[0073] The timestamp is the timestamp corresponding to when the analysis result circulates in the maintenance system.

[0074] The user-related information includes: user information and device information of the energy storage device corresponding to the user information.

[0075] Among them, the source channels include: data sources based on user input from the client and data sources automatically triggered by the energy storage system based on abnormal data.

[0076] The request context refers to the collection of all information and status related to the request during the request processing, including request parameters, session information, user identity, request header, request body, etc.

[0077] The thread pool creates a certain number of threads in advance and puts them into a "pool". When a task needs to be executed, the system will take an idle thread from the thread pool to execute the task. When the task is completed, the thread will return to the thread pool and wait for the next task to be executed.

[0078] Specifically, the maintenance request data is encapsulated into a task, which is submitted to the thread pool and executed by a thread in the thread pool, which is defined as a tracking thread in this application. After the tracking thread is assigned to the task, it starts to execute the operations and business logic in the task object.

[0079] In a specific embodiment, the maintenance request is input into the big model, and the specific implementation of the objective function of the big model output includes:

[0080] The maintenance request is input into the big model, and the target function is obtained and output by matching the maintenance request with the description document of each function.

[0081] The description document includes the function's processing parameters, processing logic, and processing scenarios.

[0082] Different functions correspond to different description documents. For example, the functions corresponding to the after-sales module are used for fault diagnosis and maintenance guidance; the functions corresponding to the equipment control module are used to execute specific control instructions; the functions corresponding to the pre-sales module are used to provide product information and suggestions; the functions corresponding to the active monitoring module are used for data analysis and anomaly detection; etc.

[0083] Among them, processing scenarios include: after-sales scenarios, equipment control scenarios, pre-sales scenarios and active monitoring scenarios.

[0084] Specifically, a processing scenario is determined based on the information carried in the maintenance request, and then the maintenance requirements and the processing scenario are matched with the description document of the function to obtain the target function.

[0085] In a specific embodiment, the maintenance request further includes: an application scenario.

[0086] Before inputting the maintenance request into the large model and obtaining the target function output by the large model, a target large model corresponding to the maintenance request is determined based on the application scenario. Then, the maintenance request is input into the target large model and the target function output by the target large model is obtained.

[0087] In a specific embodiment, the large model includes: model attributes, and the model attributes include: model magnitude and model comprehension ability.

[0088] Based on the application scenario, the specific implementation of determining the target large model corresponding to the maintenance request includes:

[0089] When it is determined that the application scenario is a device control scenario, a large model with a model magnitude smaller than a preset magnitude is selected from multiple large models as the target large model; when it is determined that the application scenario is a demand understanding consultation scenario, a large model with a model understanding capability greater than a preset understanding capability is selected from multiple large models as the target large model.

[0090] When the application scenario is determined to be a system monitoring scenario, the preset standard large model is used as the target large model.

[0091] Among them, application scenarios include processing scenarios. Equipment control scenarios include: equipment control scenarios; demand understanding consultation scenarios include: after-sales scenarios and pre-sales scenarios; system monitoring scenarios include: active monitoring scenarios.

[0092] Specifically, when the application scenario is determined to be a device control scenario, a lightweight big model is used as the target big model; when the application scenario is determined to be a demand understanding consultation scenario, a big model with complex understanding capabilities is determined as the target big model.

[0093] This application can accurately and quickly obtain the objective function by determining the large model corresponding to the maintenance request, and then obtain an accurate response result.

[0094] In addition, for device control requests (the corresponding application scenario is a device control scenario request), a special processing mechanism is adopted. When the user issues a control instruction, the legitimacy of the instruction will be verified first, and then the specific instruction will be sent to the energy storage device through the Internet of Things interface. The execution status will be monitored and fed back in real time throughout the process to ensure the reliability and security of the control operation. The decision layer 304 also implements a strict authority management mechanism to finely control the operating authority of the energy storage device, especially when it comes to device control and sensitive data access. At the same time, ensure the encryption and identity authentication of the data transmission process to ensure the security of the entire processing process.

[0095] This modular design architecture makes the maintenance system highly scalable. When new business scenarios need to be supported, it is only necessary to develop new functional modules and write corresponding description documents without changing the existing architecture.

[0096] Specifically, the objective function may also be obtained through a pre-created mapping table, where the mapping table is used to characterize the mapping relationship between the maintenance request (user intention) and the objective function.

[0097] In a specific embodiment, after inputting the maintenance request into the big model and obtaining the target function output by the big model, based on the processing parameters corresponding to the target function and the maintenance request, retrieve data is obtained from the preset knowledge base; the retrieved data and the parsing results are spliced ​​to obtain the data to be processed. Then, the processing logic of the target function is used to analyze the data to be processed and the request status to obtain the prompt word.

[0098] Among them, the preset knowledge base includes: an offline knowledge base and an online knowledge base.

[0099] The offline knowledge base mainly stores relatively stable professional knowledge, including: energy storage equipment manuals, troubleshooting guides, standard specifications, etc. The above content is structured for raw data and imported into the maintenance system in batches through the knowledge import tool.

[0100] The online knowledge base is synchronized with the business database through a real-time interface, and can obtain dynamic information such as the operating status of energy storage equipment, user feedback, and fault records in a timely manner.

[0101] The maintenance system uses a vector database to store the above data. Each piece of data is converted into a high-dimensional vector, and then an efficient retrieval index is established. The vectorized data is obtained using a text embedding model, which accurately captures the semantic features of the data and supports similarity reduction and semantic matching.

[0102] The present application can quickly obtain the required retrieval data through a preset knowledge base, and then process the obtained data to be processed and the request status based on the processing logic of the target function to obtain the function data result (prompt word).

[0103] In a specific embodiment, after obtaining the response result corresponding to the maintenance requirement indication data output by the large model, the response result is sent to the smart terminal through any one or more of text, pictures, videos and control instructions.

[0104] Specifically, the present application pre-creates different output templates corresponding to different processing scenarios. After obtaining the response result, the output module corresponding to the processing scenario is obtained, and the response result is output based on the output template.

[0105] Among them, the output template includes how to display core information, as well as the tone, professionalism and expression of the output.

[0106] Specifically, by continuously collecting user feedback to optimize the output template, the system's response quality can be continuously improved, so that the output results are more in line with user expectations and the user experience is improved.

[0107] Next, through Figure 2 The application of this application to the maintenance system of the energy storage system is used as an example for illustration, where Figure 3 As shown, the maintenance system includes: an input layer 301, a processing layer 302, a model layer 303, a planning layer 304 and an output layer 305.

[0108] Step 201, obtaining the parsing result corresponding to the maintenance requirement indication data.

[0109] Specifically, the input layer 301 serves as a data input port of the maintenance system of the energy storage system, for example, receives various forms of data input from users through a client, and processes multi-modal input data (maintenance demand indication data) using the input layer 301 .

[0110] When the user inputs text, the text is encoded and recognized, and the preset special characters and sensitive words are filtered and detected. The super-long text (text longer than the preset text length) can be automatically segmented. In addition, the maintenance system will retain the original data entered by the user (including the original punctuation and typesetting format), and further understand the semantics based on the original data to obtain the parsing results.

[0111] When the user inputs via voice, real-time streaming processing technology is used to filter environmental noise through a noise reduction algorithm, and the voice is converted into text through a speech recognition engine. The converted text is then processed in the above text processing method to obtain a parsing result.

[0112] When users input images, the images are formatted in a unified manner, for example, IPG, PNG, and WEBP formats are converted into preset standard formats, and the converted images are compressed and resized to ensure the efficiency of subsequent processing. The obtained images are intelligently analyzed, for example, scene recognition, text extraction, and face detection, to obtain analysis results.

[0113] When the user inputs any one or more of text, voice and pictures, the input data is processed based on the above processing method, and a unified session context is established to associate different types of input data in time sequence to ensure semantic coherence and obtain the parsing result.

[0114] Step 202, generating a unique session identifier corresponding to the parsing result, and obtaining a timestamp, source channel, and user-related information corresponding to the parsing result.

[0115] The unique session identifier is generated by the processing layer 302 of the maintenance system, and the timestamp corresponding to the time when the parsing result is transmitted from the input layer 301 to the processing layer 302 is recorded.

[0116] Step 203: Create a request context based on the unique session identifier, timestamp, source channel and user-related information, generate a maintenance request carrying the request context, and input the maintenance request into the thread pool to obtain a tracking thread.

[0117] Specifically, a request context may be created based on a unique session identifier, a timestamp, a source channel, user-related information, and a processing priority, a maintenance request carrying the request context may be generated, and the maintenance request may be input into a thread pool to obtain a tracking thread.

[0118] Among them, timestamp, source channel, user-related information and processing priority serve as metadata.

[0119] Among them, the working threads in the thread pool are event-driven, and each thread maintains an independent processing pipeline.

[0120] The processing priority is used to indicate the processing priority of the maintenance request, which can be divided into three types: high, medium, and low. The maintenance request is processed based on the processing priority.

[0121] Step 204: Create a request processing session corresponding to the maintenance request.

[0122] Among them, the tracking thread corresponds to the request processing session, which runs through the entire request life cycle. During the entire processing process, the request status will be recorded in real time, including: current processing stage, completed steps, pending steps and other information.

[0123] This application implements a complete request tracking mechanism, which can view the request status of any request in real time. In addition, for requests whose processing time exceeds the preset time threshold, an alarm mechanism will be triggered and an alternative processing flow will be started.

[0124] Among them, the backup processing flow is used to quickly complete the execution of the request.

[0125] Step 205 , based on the application scenario, determine the target large model corresponding to the maintenance request, and input the maintenance request into the target large model to obtain the target function output by the target large model.

[0126] The model layer 303 adopts a distributed microservice architecture, and manages and schedules multiple large language models through a model gateway. The model grid serves as a unified entry point, realizing the intelligent selection of large language models and the load balancing adjustment of the operation and maintenance system.

[0127] The processing layer 302 transmits the maintenance request to the model layer 303. When the model layer 303 receives the maintenance request, the model gateway will route it based on the regional information of the request processing session to ensure that the data compliance requirements of different regions are met. For example, for maintenance requests in region A, they will be routed to the model service deployed in region A first to avoid sensitive data issues in region A.

[0128] At the same time, the model gateway determines the corresponding target large model based on the application scenario of the maintenance request, and processes the maintenance request through the target large model.

[0129] The model gateway is a component used to manage and deploy machine learning models. It is usually responsible for functions such as model loading, reasoning, monitoring, and updating to ensure that the model can run efficiently and stably in the production environment.

[0130] Step 206, based on the processing parameters and maintenance request corresponding to the objective function, retrieve data is obtained from the preset knowledge base; the retrieve data and the parsing result are combined to obtain the data to be processed.

[0131] The model layer 303 also includes a knowledge base system, which adopts a hybrid storage architecture, namely, offline knowledge base storage and online knowledge base storage.

[0132] In addition, in terms of the connection between the model layer 303 and the decision layer 304, the model layer adopts a standardized service interface, and the decision layer 304 converts the input data of the model layer into a structured decision request.

[0133] Specifically, the decision layer 304 is a key link in the maintenance system that is responsible for core business processing. The decision layer 304 accurately matches the data input by the model layer 303 based on the description documents of the pre-loaded specific modules.

[0134] Specifically, after the big model matches the maintenance request to the corresponding functional module (objective function), the functional module will obtain the required information through a unified data access interface according to the business logic requirements. This information may come from multiple data sources, including: real-time data, historical operation data, user information and knowledge base. In special cases, it is also necessary to call a third-party interface to obtain weather, electricity price and other data.

[0135] Step 207: Analyze the data to be processed and the request status using the processing logic of the target function to obtain prompt words.

[0136] The functional module processes the acquired information to obtain a processing result, and structures the processing result to obtain a prompt word.

[0137] Step 208, returning the prompt word to the large model, and obtaining a response result through the large model.

[0138] Specifically, the large model generates more natural and understandable answers based on the prompt words.

[0139] Below, this application is described in detail through a specific scenario.

[0140] Scenario setting: The temperature of the battery pack of a certain energy storage device rises abnormally.

[0141] For input layer 301:

[0142] User input: The maintenance personnel took a photo of the battery pack through the mobile phone app and attached a voice description: "The temperature of this battery pack is very high. I wonder if there is a problem."

[0143] Initial processing: After receiving the input, the client performs initial processing on the photo to identify the model and location of the battery pack. At the same time, the voice message is converted into text: "The temperature of this battery pack is very high. I wonder if there is a problem."

[0144] For processing layer 302:

[0145] Task distribution: The input content is passed to the Thread pool for task distribution, ensuring that each input end has an independent processing mechanism.

[0146] Preprocessing: Perform image enhancement processing on the image to ensure clarity; clean the text information and remove irrelevant content.

[0147] For model layer 303:

[0148] Knowledge base search: Based on the recognized semantics, the model layer searches for relevant fault diagnosis information and processing methods from the knowledge base, which contains a large number of equipment maintenance manuals, fault cases, and expert experience.

[0149] Model selection: Select the appropriate model for processing based on the region and application characteristics of the input information. For example, a model for the energy storage field may contain more battery management knowledge.

[0150] For decision layer 304:

[0151] Module matching: The large model matches the recognized semantic information with the different modules in the decision layer to determine the most suitable processing module. In this case, the "fault diagnosis" module is matched.

[0152] Data query: The fault diagnosis module queries the database to obtain detailed information about the battery pack, including the equipment serial number, historical operation records, temperature change trends, etc.

[0153] Output content generation: Generate specific fault diagnosis reports and processing suggestions based on the queried data.

[0154] For example:

[0155] The text suggests: "According to historical data, the temperature of this battery pack has continued to rise in the past 2 hours. It is recommended to disconnect the inverter immediately and shut it down, and check whether the cooling system is working properly."

[0156] Picture guide: "Please refer to the following cooling system inspection steps: [insert picture]"

[0157] Control command: "The system will automatically send a disconnect and shutdown command to the battery pack. Please confirm the operation: [Confirm button]"

[0158] For output layer 305:

[0159] Content optimization: The generated content is sent to the big model again for optimization to ensure smooth language and clear logic.

[0160] Multimedia output: The optimized content is presented to operation and maintenance personnel in multimedia form through mobile phone APP, including text, pictures, videos and control instructions.

[0161] Control command execution: After the operation and maintenance personnel click the confirmation button, the system sends a disconnect and shutdown command to the battery pack through the IoT interface to achieve remote control.

[0162] This application integrates large models, multimodal information processing technology and customized objective functions to achieve efficient management and automated execution of complex operation and maintenance tasks. It can not only accurately understand user needs, but also generate responses in various forms including text, images and voice according to actual conditions, and even directly send control instructions to quickly solve problems, thereby improving system operation and maintenance efficiency and service quality.

[0163] The embodiment of the present application also provides a large model-based energy storage system maintenance device, the specific implementation of which can refer to the description of the large model-based energy storage system maintenance method, and the repeated parts will not be repeated. Figure 4 As shown, the device comprises:

[0164] The parsing module 401 is used to obtain the maintenance demand indication data and parse the maintenance demand indication data to obtain the parsing result;

[0165] A generating module 402, configured to generate a maintenance request corresponding to the parsing result, and to create a request processing session corresponding to the maintenance request, wherein the request processing session is used to record the request status of the maintenance request;

[0166] The first prediction module 403 is used to input the maintenance request into the large model to obtain the target function output by the large model;

[0167] An analysis module 404 is used to analyze the search data and request status corresponding to the maintenance request using the objective function to obtain a prompt word;

[0168] The second prediction module 405 is used to input the prompt word into the large model to obtain the response result corresponding to the maintenance demand indication data output by the large model.

[0169] In a specific embodiment, the device also includes a retrieval module, which is used to obtain retrieval data from a preset knowledge base based on the processing parameters and maintenance requests corresponding to the target function; splicing the retrieval data and the analysis results to obtain the data to be processed; and an analysis module 404, which is used to use the processing logic of the target function to analyze the data to be processed and the request status to obtain prompt words.

[0170] In a specific embodiment, the generation module 402 is used to generate a unique session identifier corresponding to the parsing result, and obtain the timestamp corresponding to the parsing result, the source channel of the maintenance requirement indication data, and user-related information; create a request context based on the unique session identifier, timestamp, source channel, and user-related information, and generate a maintenance request carrying the request context; input the maintenance request into the thread pool, and obtain a request processing session based on the thread pool, as well as a tracking thread including the request processing session.

[0171] In a specific embodiment, the maintenance request further includes: an application scenario. The device further includes a determination module, which is used to determine a target macro model corresponding to the maintenance request based on the application scenario. A first prediction module 403 is used to input the maintenance request into the target macro model to obtain an objective function output by the target macro model.

[0172] In a specific embodiment, the big model includes: model attributes, which include: model magnitude and model understanding ability. A determination module is used to select a big model with a model magnitude less than a preset magnitude from multiple big models as a target big model when the application scenario is determined to be a device control scenario; and to select a big model with a model understanding ability greater than a preset understanding ability from multiple big models as a target big model when the application scenario is determined to be a demand understanding consultation scenario.

[0173] In a specific embodiment, the first prediction module 403 is used to input the maintenance request into the large model, and obtain and output the target function by matching the maintenance request with the description document of each function.

[0174] The description document includes the function's processing parameters, processing logic, and processing scenarios.

[0175] In a specific embodiment, the device includes an output module for sending the response result to the smart terminal through any one or more of text, pictures, videos and control instructions.

[0176] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 501, a communications interface 502, a memory 503 and a communication bus 504, wherein the processor 501, the communications interface 502 and the memory 503 communicate with each other via the communication bus 504. The processor 501 may call the logic instructions in the memory 503 to execute the maintenance method of the energy storage system based on the large model.

[0177] In addition, the logic instructions in the above-mentioned memory 503 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0178] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the maintenance method of the energy storage system based on the large model provided by the above methods.

[0179] On the other hand, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the large model-based energy storage system maintenance method provided in the above embodiments.

[0180] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0181] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0182] Finally, it should be noted that the above is only the preferred implementation of the present application, and the present application is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the scope of protection of the present application.

Claims

1. A large model-based energy storage system maintenance method, characterized in that: The method comprises: Acquiring maintenance demand indication data, and parsing the maintenance demand indication data to obtain parsing results; generating a maintenance request corresponding to the parsing result, and creating a request processing session corresponding to the maintenance request, wherein the request processing session is used to record the request status of the maintenance request; Inputting the maintenance request into a large model to obtain an objective function output by the large model; Analyzing the search data corresponding to the maintenance request and the request status using the objective function to obtain a prompt word; The prompt word is input into the large model to obtain a response result output by the large model corresponding to the maintenance requirement indication data.

2. The large model-based energy storage system maintenance method according to claim 1, characterized in that: After inputting the maintenance request into the large model and obtaining the target function output by the large model, the method further includes: Based on the processing parameters corresponding to the objective function and the maintenance request, acquiring the search data from a preset knowledge base; splicing the search data and the parsing result to obtain data to be processed; The objective function is used to analyze the search data corresponding to the maintenance request and the request status to obtain prompt words, including: The processing logic of the objective function is used to analyze the data to be processed and the request status to obtain the prompt word.

3. The large model-based energy storage system maintenance method according to claim 1, characterized in that: Generating a maintenance request corresponding to the parsing result and creating a request processing session corresponding to the maintenance request, including: Generate a unique session identifier corresponding to the parsing result, and obtain a timestamp corresponding to the parsing result, a source channel of the maintenance requirement indication data, and user-related information; Creating a request context based on the unique session identifier, the timestamp, the source channel, and the user-related information, and generating a maintenance request carrying the request context; The maintenance request is input into a thread pool, and the request processing session and a tracking thread including the request processing session are obtained based on the thread pool.

4. The large model-based energy storage system maintenance method according to any one of claims 1 to 3, characterized in that: The maintenance request also includes: application scenarios; Before inputting the maintenance request into the large model and obtaining the target function output by the large model, the method further includes: Based on the application scenario, determining a target macro model corresponding to the maintenance request; Inputting the maintenance request into the big model to obtain the objective function output by the big model includes: The maintenance request is input into the target large model to obtain the target function output by the target large model.

5. The large model-based energy storage system maintenance method according to claim 4, characterized in that: The large model includes: model attributes, wherein the model attributes include: model magnitude and model understanding ability; Based on the application scenario, determining a target macro model corresponding to the maintenance request includes: When it is determined that the application scenario is a device control scenario, a large model with a model magnitude smaller than a preset magnitude is selected from a plurality of large models as the target large model; When it is determined that the application scenario is a demand understanding consultation scenario, a large model whose model understanding ability is greater than a preset understanding ability is selected from multiple large models as the target large model.

6. The large model-based energy storage system maintenance method according to any one of claims 1 to 3, characterized in that: Inputting the maintenance request into the big model to obtain the objective function output by the big model includes: Input the maintenance request into the big model, and obtain and output the target function by matching the maintenance request with the description document of each function; The description document includes the function's processing parameters, processing logic and processing scenarios.

7. The large model-based energy storage system maintenance method according to any one of claims 1 to 3, characterized in that: After obtaining the response result output by the large model corresponding to the maintenance requirement indication data, the method further includes: The response result is sent to the smart terminal through any one or more of text, picture, video and control instructions.

8. A maintenance device for an energy storage system based on a large model, characterized in that: include: A parsing module, used to obtain maintenance demand indication data, and parse the maintenance demand indication data to obtain a parsing result; a generating module, configured to generate a maintenance request corresponding to the parsing result, and to create a request processing session corresponding to the maintenance request, wherein the request processing session is used to record a request status of the maintenance request; A first prediction module, used for inputting the maintenance request into a large model to obtain an objective function output by the large model; An analysis module, configured to analyze the search data corresponding to the maintenance request and the request status using the objective function to obtain a prompt word; The second prediction module is used to input the prompt word into the large model to obtain a response result output by the large model corresponding to the maintenance demand indication data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the large model-based energy storage system maintenance method as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the large model-based energy storage system maintenance method as described in any one of claims 1 to 7 are implemented.

Images