Distributed energy storage site survey evaluation method and system based on multi-model collaboration

Abstract

The invention relates to a distributed energy storage site surveying and evaluating method and system based on multi-model collaboration. The method comprises the steps that picture description and a structure questionnaire form are collected through a mobile terminal; matching according to the picture description and the structure questionnaire form to generate a first cue word; calling a large model, identifying setting information and electrical characteristics in the picture description, and obtaining a picture identification result and a text analysis result; constructing a feature request by combining the form data to generate evaluation model parameters; the evaluation model parameters, the existing energy storage equipment configuration and the load prediction data are fused, and the station exploration evaluation result of the machine room is output through the evaluation model; matching and generating a second cue word according to the station exploration evaluation result; calling the large model based on the matched second cue word, and generating an evaluation report according to the station exploration evaluation result; by applying the mode and the method of the invention, the processing efficiency is greatly improved and the data accuracy is improved for the design of a distributed energy storage site survey evaluation scene with a large site number scale.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage site survey and evaluation, and more specifically, to a distributed energy storage site survey and evaluation method and system based on multi-model collaboration. Background Art

[0002] Energy storage sites are typically selected from a large number of general base station sites. The selection criteria must comprehensively consider the power network structure, load conditions, site location, current equipment room layout and orientation, equipment location, availability of vacant space, bus power range, and other factors. Furthermore, factors such as energy storage capacity and cost are then combined to comprehensively determine whether a base station room can be considered an energy storage site.

[0003] Because the number of base station sites involved is too large, for example, China Tower alone has more than 8,000 sites in Dongguan City, Guangdong Province, and they are also relatively scattered. In the past, the method of manually reviewing the survey information of each site by providing it through a third party was not only labor-intensive and inefficient, but also prone to misjudgment. A distributed energy storage site survey and evaluation method and system based on multi-model collaboration that can better solve the above problems is needed. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a distributed energy storage site survey and evaluation method based on multi-model collaboration, and also provide a distributed energy storage site survey and evaluation system based on multi-model collaboration, in response to the above-mentioned defects of the prior art.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A distributed energy storage site survey and evaluation method based on multi-model collaboration is constructed, wherein the method comprises the following steps:

[0007] Use mobile terminals to collect picture descriptions and a structured questionnaire containing basic information about the computer room, spatial layout, equipment location, power interfaces, and the planned energy storage installation location;

[0008] Generate the first prompt word based on the picture description and the structured questionnaire form;

[0009] Based on the matched first prompt word, the large model is called to identify the setting information and electrical characteristics in the image description to obtain the image recognition result and text analysis result;

[0010] Combine the image recognition results and text parsing results with the form data to construct feature requests and generate evaluation model parameters;

[0011] The evaluation model parameters, existing energy storage equipment configuration, and load forecast data are integrated to output the site survey and evaluation results of the computer room through the evaluation model;

[0012] Generate a second prompt word based on the site survey and assessment results;

[0013] The large model is called based on the matched second prompt word, and an assessment report is generated according to the site survey and assessment results.

[0014] In the distributed energy storage site survey and assessment method based on multi-model collaboration of the present invention, the step of generating the first prompt word based on the matching of the image description and the structured questionnaire form includes:

[0015] Based on the commonly used prompt word framework and technology in energy storage site survey business scenarios, multiple prompt word templates are designed. When extracting assessment model features, prompt word templates related to image descriptions and structured questionnaire forms are automatically matched, and parameters are filled in according to the submitted data and context to obtain the final first prompt word.

[0016] In the distributed energy storage site survey and evaluation method based on multi-model collaboration of the present invention, the step of generating a second prompt word based on the site survey and evaluation results includes:

[0017] When generating an assessment report, the prompt word template related to the assessment report is automatically matched, and the parameters are filled in according to the submitted data and context to obtain the final second prompt word.

[0018] In the distributed energy storage site survey and assessment method based on multi-model collaboration described in the present invention, the prompt word template needs to be tested before formal application:

[0019] By designing the same scenario + the same data, multiple prompt word templates are used to call the large template;

[0020] Observe the calling results of different templates;

[0021] Continuously optimize and select the best template application.

[0022] In the distributed energy storage site survey and evaluation method based on multi-model collaboration described in the present invention, the training of the evaluation model includes:

[0023] Extract characteristic variables from raw data according to business and compliance requirements to obtain regularized features;

[0024] Select feature variables from the original data using feature selection tools and derive feature variables from the original data to obtain algorithmic features;

[0025] Input regularized features into business rules, perform rule iteration of the evaluation model and output rule matching results / scores;

[0026] Input algorithmic features into the evaluation model, perform model training and output model results / scores;

[0027] Evaluate the model based on rule matching results / scores and model results / scores.

[0028] In the distributed energy storage site survey and assessment method based on multi-model collaboration described in the present invention, the specifications for collecting images include:

[0029] Open the collection app and enter the site survey option;

[0030] Select the shooting business type and enter the corresponding shooting mode;

[0031] Take photos of the object according to the instructions given in the shooting mode;

[0032] Submit photos with business type information included in the photo title.

[0033] A distributed energy storage site survey and evaluation system based on multi-model collaboration is used to implement the distributed energy storage site survey and evaluation method based on multi-model collaboration as described above, wherein the system includes an acquisition subsystem, an evaluation subsystem and an intelligent agent subsystem;

[0034] The collection subsystem is used to collect picture descriptions and a structured questionnaire containing basic information about the computer room, spatial layout, equipment location, power interface, and the location where the energy storage is to be installed through a mobile terminal, and send the information to the intelligent agent subsystem via the evaluation subsystem;

[0035] The intelligent agent subsystem is configured to generate a first prompt word based on a match between the image description and the structured questionnaire form, and to call a large model based on the matched first prompt word. The large model recognizes the setting information and electrical characteristics in the image description to obtain an image recognition result and a text parsing result. The image recognition result and the text parsing result are combined with the form data to construct a feature request to generate evaluation model parameters, which are then sent to the evaluation subsystem.

[0036] The evaluation subsystem is used to integrate the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, output the site survey and evaluation results of the computer room through the evaluation model, and send them to the intelligent subsystem;

[0037] The intelligent agent subsystem is further configured to generate a second prompt word according to the site survey and evaluation results, call the large model based on the matched second prompt word, generate an evaluation report based on the site survey and evaluation results, and feed the report back to the evaluation subsystem.

[0038] The distributed energy storage site survey and evaluation system based on multi-model collaboration described in the present invention, wherein:

[0039] The acquisition subsystem includes a data acquisition module, an image acquisition SDK, and a questionnaire management module; the data acquisition module is used to acquire data through structured questionnaire forms and on-site photos; the image acquisition SDK is used to standardize the angle, size, watermark, and business type of on-site photos to facilitate large model image recognition; and the questionnaire management module is used to manage the acquired data;

[0040] The evaluation subsystem includes a reporting module and a decision-making evaluation module; the decision-making evaluation module integrates the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, and outputs the site survey and evaluation results of the computer room through the evaluation model; the reporting module is used for report output management;

[0041] The intelligent agent subsystem includes a prompt word generation module, a domain knowledge base module and a large model; the prompt word generation module is used to match and generate corresponding prompt words based on the received data; the domain knowledge base module is used to provide basic data support for the prompt word generation module; the large model is used to identify the setting information and electrical characteristics in the image description to obtain image recognition results and text parsing results, and to generate an assessment report based on the site survey and assessment results.

[0042] The distributed energy storage site survey and evaluation system based on multi-model collaboration of the present invention further includes a training and labeling subsystem; the training and labeling subsystem includes an evaluation model training module, a manual labeling module, and a subsequent operation and maintenance result feedback module;

[0043] The manual annotation module is used to annotate and describe information on site photos, facilitating fine-tuning of the visual model of the large model in the field of site survey business;

[0044] The evaluation model training module is used to design a multi-dimensional energy storage site survey and evaluation model based on the site power network, load, space, scalable modules, and energy storage capacity dimensions;

[0045] The subsequent operation and maintenance result feedback module is used to obtain subsequent operation and maintenance data of the installed sites to facilitate the continuous optimization and closed-loop evaluation model.

[0046] The distributed energy storage site survey and evaluation system based on multi-model collaboration of the present invention further includes a data processing subsystem;

[0047] The data processing subsystem includes a preprocessing module and a feature management module; the preprocessing module is used for storing and processing survey data; the feature management module is used for extracting model feature data and providing qualified training data for the training and annotation subsystem.

[0048] The application has the advantages that: the method of the application is applied to the survey and evaluation of distributed energy storage sites with a large number of sites, and a large model is used to compress the average time for single-site evaluation from 1 hour to 5 minutes, and batch parallel processing is supported. As the business expands across provinces and cities and the number of energy storage sites continues to increase, the efficiency advantage of the scheme is increasingly reflected, greatly improving the processing efficiency and improving the data accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be further described below with reference to the drawings and embodiments. The drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0050] Figure 1 is a flowchart of a distributed energy storage site survey and evaluation method based on multi-model cooperation of a preferred embodiment of the present application;

[0051] Figure 2 is a flowchart of an evaluation model training process of a distributed energy storage site survey and evaluation method based on multi-model cooperation of a preferred embodiment of the present application;

[0052] Figure 3 is a flowchart of feature extraction and model design based on business understanding and machine learning of a distributed energy storage site survey and evaluation method based on multi-model cooperation of a preferred embodiment of the present application;

[0053] Figure 4 is a flowchart of a photographing process of a distributed energy storage site survey and evaluation method based on multi-model cooperation of a preferred embodiment of the present application;

[0054] Figure 5 is a flowchart of prompt word generation of a distributed energy storage site survey and evaluation method based on multi-model cooperation of a preferred embodiment of the present application;

[0055] Figure 6 is a principle block diagram of a distributed energy storage site survey and evaluation system based on multi-model cooperation of a preferred embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0057] The multi-model cooperation-based distributed energy storage station survey evaluation method of the preferred embodiment of the present application, as shown in Figure 1 , and referring to Figure 2-Figure 5 , comprises the steps of:

[0058] S01: Collecting picture description and structure questionnaire forms containing basic information of machine room, space layout, equipment position, power interface, and energy storage installation position information through a mobile terminal;

[0059] S02: Generating a first prompt word according to the picture description and the structure questionnaire forms;

[0060] S03: Calling a large model based on the matched first prompt word, identifying the set information and electrical characteristics in the picture description, and obtaining picture recognition results and text analysis results;

[0061] S04: Constructing feature request evaluation model parameters by combining picture recognition results and text analysis results with form data;

[0062] S05: Fusing the evaluation model parameters, existing energy storage device configuration, and load prediction data, and outputting the station survey evaluation results of the machine room through the evaluation model;

[0063] S06: Generating a second prompt word according to the station survey evaluation results;

[0064] S07: Calling a large model based on the matched second prompt word, and generating an evaluation report according to the station survey evaluation results;

[0065] The application method of the present application is designed for distributed energy storage station survey evaluation scenarios with large number of stations, and through a large model, the average time for single station evaluation is compressed from 1 hour to 5 minutes, and batch parallel processing is supported. With the expansion of cross-provincial and cross-city business and the continuous increase of energy storage stations, the efficiency advantage of the present application is increasingly reflected, the processing efficiency is greatly improved, and the data accuracy is improved.

[0066] Preferably, as shown in Figure 5 , the generation of the prompt word adopts the principle of:

[0067] 1. Based on common prompt word frameworks (such as ICIO and CRISPE) and technologies (such as few-shot prompts), a plurality of prompt word templates are designed in combination with existing business scenarios.

[0068] 2. When generating a prompt word, a corresponding prompt word template is automatically selected according to the business scenario, and parameters are filled in according to the submitted data and context to obtain the final prompt word.

[0069] 3. The large model is called using the prompt word.

[0070] The principle of prompt word testing is:

[0071] 1. Before the prompt word template is officially applied, the prompt word needs to be tested.

[0072] 2. By designing the same scenario + the same data, use multiple prompt word templates to call the large template.

[0073] 3. Observe the calling results of different templates.

[0074] 4. Continuously optimize and select the best template application.

[0075] like Figure 2 and Figure 3 As shown, the training content of the evaluation model includes:

[0076] Regularized features: feature variables extracted based on business requirements and compliance requirements.

[0077] Algorithmic features: Feature variables selected by feature selection tools (such as Scikit-learn) and feature variables derived from original features.

[0078] Regularized features are input into business rules and the rule matching results / scores are output;

[0079] Algorithmic features, input to the model, output model results / scores;

[0080] Business rules are developed and implemented by developers based on business needs and policies (may be iterated due to business changes);

[0081] The model is obtained by developers through continuous iterative training using machine learning.

[0082] like Figure 4 As shown, the specifications for collecting pictures include:

[0083] Open the collection app and enter the site survey option;

[0084] Select the shooting business type and enter the corresponding shooting mode;

[0085] Take photos of the object according to the instructions given in the shooting mode;

[0086] Photo submission, with the photo title containing business type information;

[0087] We use our self-developed front-end SDK to standardize the angles and sizes of photos taken for different businesses, and then use parameterized prompt word engineering to improve the sensitivity of large model image feature recognition.

[0088] A distributed energy storage site survey and evaluation system based on multi-model collaboration is used to implement the distributed energy storage site survey and evaluation method based on multi-model collaboration as mentioned above, such as Figure 6 As shown, the system includes a collection subsystem 10, an evaluation subsystem 11 and an intelligent agent subsystem 12;

[0089] The collection subsystem 10 is used to collect picture descriptions and a structured questionnaire containing basic information about the computer room, spatial layout, equipment location, power interface, and the proposed location of the energy storage installation through a mobile terminal, and send the information to the intelligent agent subsystem via the evaluation subsystem;

[0090] The intelligent agent subsystem 12 is configured to generate a first prompt word based on a match between the image description and the structured questionnaire form, and to invoke the large model based on the matched first prompt word. The large model identifies the setting information and electrical characteristics in the image description to obtain an image recognition result and a text parsing result. The image recognition result and the text parsing result are combined with the form data to construct a feature request to generate evaluation model parameters, which are then sent to the evaluation subsystem.

[0091] The evaluation subsystem 11 is used to integrate the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, output the site survey and evaluation results of the computer room through the evaluation model, and send them to the intelligent subsystem;

[0092] The intelligent agent subsystem 12 is further configured to generate a second prompt word according to the site survey and evaluation results, call the large model based on the matched second prompt word, generate an evaluation report based on the site survey and evaluation results, and feed the report back to the evaluation subsystem;

[0093] The system applied in this application is designed for distributed energy storage site survey and evaluation scenarios with a large number of sites. Through a large model, the average time for single-site evaluation is compressed from 1 hour to 5 minutes, and batch parallel processing is supported. As cross-provincial and municipal business expands and energy storage sites continue to increase, this solution increasingly demonstrates its efficiency advantages, greatly improving processing efficiency and data accuracy.

[0094] More specifically, each module can be divided into:

[0095] The collection subsystem 10 includes a data collection module 100, an image collection SDK 101 (front-end), and a questionnaire management module 102. The data collection module is used to obtain data through structured questionnaire forms and on-site photos. The image collection SDK is used to standardize the angle, size, watermark, and business (equipment) type of on-site photos to facilitate large model image recognition. The questionnaire management module is used to manage the obtained data.

[0096] The evaluation subsystem 11 includes a reporting module 110 and a decision-making evaluation module 111; the decision-making evaluation module integrates the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, and outputs the site survey and evaluation results of the computer room through the evaluation model; the reporting module is used for report output management;

[0097] The intelligent subsystem 12 includes a prompt word generation module 120, a domain knowledge base module 121 and a large model 122 (using open AIP, supporting the docking of large models such as deepseek, openAI, and Qianwen); the prompt word generation module 120 is used to match and generate corresponding prompt words based on the received data; the domain knowledge base module 121 is used to provide basic data support for the prompt word generation module; the large model 122 is used to identify the setting information and electrical characteristics in the image description to obtain image recognition results and text parsing results, and generate an evaluation report based on the survey station evaluation results; this subsystem is also one of the core of this platform. It is responsible for the collaboration of other subsystems, collecting and receiving real-time survey data, generating large model prompt words, calling large model intelligent photo processing and semi-structured text processing, calling the evaluation model to obtain evaluation results, and calling the large model to generate the final evaluation report.

[0098] The system may also include a training and labeling subsystem 13; the training and labeling subsystem 13 includes an evaluation model training module 130, a manual labeling module 131 and a subsequent operation and maintenance result feedback module 132; the manual labeling module 131 is used to annotate information on on-site photos to facilitate fine-tuning of the visual model of the large model in the site survey business field; the evaluation model training module 130 is used by the company's business experts to design a multi-dimensional energy storage site survey and evaluation model based on the site power network, load, space, scalable modules, and energy storage capacity dimensions; the subsequent operation and maintenance result feedback module 132 is used to obtain subsequent operation and maintenance operation data of the installed sites, which facilitates the continuous optimization and closed-loop of the evaluation model, thereby improving the reliability and accuracy of the model.

[0099] The system includes a data processing subsystem 14; the data processing subsystem 14 includes a preprocessing module 140 and a feature management module 141; the preprocessing module 140 is used for storing and processing survey data; the feature management module 141 is used for extracting model feature data and providing qualified training data for the training and labeling subsystem.

[0100] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A distributed energy storage site survey and evaluation method based on multi-model collaboration, characterized in that: The method comprises the steps of: Use mobile terminals to collect picture descriptions and a structured questionnaire containing basic information about the computer room, spatial layout, equipment location, power interfaces, and the planned energy storage installation location; Generate the first prompt word based on the picture description and the structured questionnaire form; Based on the matched first prompt word, the large model is called to identify the setting information and electrical characteristics in the image description to obtain the image recognition result and text analysis result; Combine the image recognition results and text parsing results with the form data to construct feature requests and generate evaluation model parameters; The evaluation model parameters, existing energy storage equipment configuration, and load forecast data are integrated to output the site survey and evaluation results of the computer room through the evaluation model; Generate a second prompt word based on the site survey and assessment results; The large model is called based on the matched second prompt word, and an assessment report is generated according to the survey and assessment results.

2. The distributed energy storage site survey and evaluation method based on multi-model collaboration according to claim 1 is characterized in that: Generating the first prompt word according to the matching of the picture description and the structured questionnaire form includes: Based on the commonly used prompt word framework and technology in energy storage site survey business scenarios, multiple prompt word templates are designed. When extracting assessment model features, prompt word templates related to image descriptions and structured questionnaire forms are automatically matched, and parameters are filled in according to the submitted data and context to obtain the final first prompt word.

3. The distributed energy storage site survey and evaluation method based on multi-model collaboration according to claim 2 is characterized in that: Generating a second prompt word according to the site survey and assessment results includes: When generating an assessment report, the prompt word template related to the assessment report is automatically matched, and the parameters are filled in according to the submitted data and context to obtain the final second prompt word.

4. The distributed energy storage site survey and evaluation method based on multi-model collaboration according to claim 2 is characterized in that: Before the prompt word template is officially used, the prompt words need to be tested: By designing the same scenario + the same data, multiple prompt word templates are used to call the large template; Observe the calling results of different templates; Continuously optimize and select the best template application.

5. The distributed energy storage site survey and evaluation method based on multi-model collaboration according to claim 1 is characterized in that: The training of the evaluation model includes: Extract characteristic variables from raw data according to business and compliance requirements to obtain regularized features; Select feature variables from the original data using feature selection tools and derive feature variables from the original data to obtain algorithmic features; Input regularized features into business rules, perform rule iteration of the evaluation model and output rule matching results / scores; Input algorithmic features into the evaluation model, perform model training and output model results / scores; Evaluate the model based on rule matching results / scores and model results / scores.

6. The distributed energy storage site survey and evaluation method based on multi-model collaboration according to claim 1 is characterized in that: The specifications for collecting images include: Open the collection app and enter the site survey option; Select the shooting business type and enter the corresponding shooting mode; Take photos of the object according to the instructions given in the shooting mode; Submit photos with business type information included in the photo title.

7. A distributed energy storage site survey and evaluation system based on multi-model collaboration, used to implement the distributed energy storage site survey and evaluation method based on multi-model collaboration as described in any one of claims 1 to 6, characterized in that: The system includes an acquisition subsystem, an evaluation subsystem and an intelligent agent subsystem; The collection subsystem is used to collect picture descriptions and a structured questionnaire containing basic information about the computer room, spatial layout, equipment location, power interface, and the location where the energy storage is to be installed through a mobile terminal, and send the information to the intelligent agent subsystem via the evaluation subsystem; The intelligent agent subsystem is used to generate a first prompt word based on the matching of the image description and the structured questionnaire form, and call the large model based on the matched first prompt word. The large model recognizes the setting information and electrical characteristics in the image description to obtain the image recognition result and the text parsing result; The image recognition result and the text parsing result are combined with the form data to construct a feature request to generate evaluation model parameters and send them to the evaluation subsystem; The evaluation subsystem is used to integrate the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, output the site survey and evaluation results of the computer room through the evaluation model, and send them to the intelligent subsystem; The intelligent agent subsystem is further configured to generate a second prompt word according to the site survey and evaluation results, call the large model based on the matched second prompt word, generate an evaluation report based on the site survey and evaluation results, and feed the report back to the evaluation subsystem.

8. The distributed energy storage site survey and evaluation system based on multi-model collaboration according to claim 7 is characterized in that: The acquisition subsystem includes a data acquisition module, an image acquisition SDK, and a questionnaire management module; the data acquisition module is used to acquire data through structured questionnaire forms and on-site photos; the image acquisition SDK is used to standardize the angle, size, watermark, and business type of on-site photos to facilitate large model image recognition; and the questionnaire management module is used to manage the acquired data; The evaluation subsystem includes a reporting module and a decision-making evaluation module; the decision-making evaluation module integrates the evaluation model parameters, the existing energy storage equipment configuration, and the load forecast data, and outputs the site survey and evaluation results of the computer room through the evaluation model; the reporting module is used for report output management; The intelligent agent subsystem includes a prompt word generation module, a domain knowledge base module and a large model; the prompt word generation module is used to match the received data and generate corresponding prompt words; The domain knowledge base module is used to provide basic data support for the prompt word generation module; The large model is used to identify the setting information and electrical characteristics in the image description to obtain image recognition results and text parsing results, and to generate an assessment report based on the site survey and assessment results.

9. The distributed energy storage site survey and evaluation system based on multi-model collaboration according to claim 7 is characterized in that: The distributed energy storage site survey and evaluation system based on multi-model collaboration also includes a training and labeling subsystem; the training and labeling subsystem includes an evaluation model training module, a manual labeling module and a subsequent operation and maintenance result feedback module; The manual annotation module is used to annotate and describe information on site photos, facilitating fine-tuning of the visual model of the large model in the field of site survey business; The evaluation model training module is used to design a multi-dimensional energy storage site survey and evaluation model based on the site power network, load, space, scalable modules, and energy storage capacity dimensions; The subsequent operation and maintenance result feedback module is used to obtain subsequent operation and maintenance data of the installed sites to facilitate the continuous optimization and closed-loop evaluation model.

10. The distributed energy storage site survey and evaluation system based on multi-model collaboration according to claim 9 is characterized in that: The distributed energy storage site survey and evaluation system based on multi-model collaboration also includes a data processing subsystem; The data processing subsystem includes a pre-processing module and a feature management module; The pre-processing module is used for storing and processing survey data; The feature management module is used to extract model feature data and provide qualified training data for the training and labeling subsystem.

Images