An intelligent algorithm development auxiliary tool for a spirit pivot plasma control system

Abstract

The application discloses an intelligent algorithm development auxiliary tool for a spirit pivot plasma control system, belongs to the technical field of plasma control systems, and comprises a visual module, an AI question and answer module and a WebView panel; the visual module is used for analyzing a PCS project directory, generating an interactive view based on the WebView, and supporting architecture analysis, file generation, code navigation and bidirectional editing synchronization; the AI question and answer module adopts a locally deployed search enhancement generation system, combines a PCS knowledge vector database, and provides field-specific query and answers; the WebView panel is used as an interactive page of a user and the visual module and the AI question and answer module, renders a tree view and an organization structure diagram of the visual module, and displays questions and answers of the AI question and answer module. The application simplifies complex parameter configuration and multi-file management, and significantly reduces operation time.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of plasma control system (PCS), and particularly relates to an intelligent algorithm development auxiliary tool for a plasma control system. BACKGROUND

[0002] The plasma control system (PCS) is the core technology of the tokamak device to realize controllable nuclear fusion, which is used for accurately controlling the discharge process of the plasma to ensure the stability, safety and high efficiency of the experiment. The plasma control system is specially designed for the China Fusion Engineering Test Reactor (CFETR) and verified on the superconducting tokamak EAST device. The distributed control framework based on components is adopted, the core control logic is realized by C++, the interface description language (IDL) defines the user interface, and the extensible markup language (XML) stores the parameter configuration. The control algorithm development involves complex parameter definition, multi-file management and real-time framework integration, and the algorithm needs to be registered to the data engine to support efficient data access and real-time processing.

[0003] Under the traditional development mode, the developer needs to manually edit multi-file configuration and parameter definition, and the development process is tedious and prone to errors, which leads to low development efficiency. On the one hand, the existing tools do not provide a user-friendly visual interface, and the developer has difficulty in intuitively understanding the hierarchical relationship of the plasma control system architecture (including control category, control algorithm and data item subset), the development cycle is long, and the learning threshold is high. On the other hand, there is a lack of highly automated and intelligent development tools to assist in control algorithm development, and parameter configuration, file generation and code integration rely on manual operation, which has much repetitive work and increases the risk of errors, and it is difficult to meet the rapid iteration requirements. The developer needs to refer to a large amount of documents or rely on expert guidance, the knowledge acquisition efficiency is low, and the development progress is affected. The existing development mode limits the efficiency and engineering level of the control algorithm development in the plasma control system. SUMMARY

[0004] To solve the above technical problems, the application adopts the following technical solutions:

[0005] An intelligent algorithm development auxiliary tool for a plasma control system, comprising: a visual module, an AI question and answer module, and a WebView panel.

[0006] The visualization module is used for parsing a PCS project directory, generating a WebView-based interactive view, supporting architecture parsing, generating files, code navigation, and bidirectional editing synchronization; the visualization module comprises a PCS architecture tree construction module, a change content parsing module, and an automatic generation content module;

[0007] The AI question and answer module adopts a locally deployed retrieval augmented generation system, combines a PCS knowledge vector database, and provides domain-specific query and answer;

[0008] The WebView panel is used as an interactive page of a user with the visualization module and the AI question and answer module, renders a tree view and an organization architecture diagram of the visualization module, and displays a question and answer of the AI question and answer module.

[0009] The present application has the following beneficial effects:

[0010] The present application solves the problem of lack of intuitive visual interface through code and view interaction function. The present application realizes jumping from a variable of a custom control type in an algorithm file to corresponding content of an extensible markup language (XML) file defining the variable in a visual studio code (VS Code) through a language service protocol (LSP), and provides an "open in plasma X" function to position the variable in the code to an architecture node in the visual view, realizes seamless interaction between code and view, improves interaction efficiency, and provides more intuitive development experience.

[0011] The present application solves the problem of insufficient automation degree through an automatic file generation method. The present application integrates an IDL (interactive data language) compiler and an algorithm template, automatically generates C++ source code, an IDL file and a configuration file, and registers variables to a data engine. A user can generate a complete code framework by inputting parameters in a WebView, and the generation process takes only a few seconds, which saves more than 90% time compared with manually writing IDL and code, and significantly improves engineering level.

[0012] The present application solves the problem of lack of domain-specific intelligent support through an RAG (retrieval augmented generation) driven intelligent question and answer function. The present application provides targeted technical guidance based on a locally deployed vector database and a large language model (DeepSeek-R1) and in combination with PCS domain documents (such as technical manuals and experiment reports). The user inputs a query (such as "how to configure plasma current parameters"), and the system generates an answer in real time, which is several times more efficient than manually consulting documents, and significantly reduces the learning threshold.

[0013] The application solves the problems of complex parameter configuration and multi-file management through the visual parameter editing and architecture management method. Based on the VS Code embedded web view interface (WebView API), the application provides interactive view rendering PCS architecture, supports form parameter editing and keyword filtering. After the user edits the parameters or architecture in WebView, the plug-in automatically updates the file. Compared with the traditional manual editing process, PlasmaX simplifies complex parameter configuration and multi-file management, and significantly reduces operation time. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 It is an overall function architecture module interaction diagram of the application;

[0015] Figure 2 It is a structure and workflow diagram of the visualization module of the application;

[0016] Figure 3 It is a logic flow diagram of code and view interaction in the visualization module of the application; (a) is a logic flow diagram of code navigation of the visualization module, (b) is a logic flow diagram of jump definition in the visualization module, and (c) is a logic flow diagram of positioning in the plug-in view of the visualization module;

[0017] Figure 4 It is a structure and workflow diagram of the AI question and answer module of the application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical scheme and advantages of the application more clear and understandable, the application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application. In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as they do not conflict with each other.

[0019] The application proposes an intelligent algorithm development auxiliary tool PlasmaX (hereinafter referred to as plug-in) for a spirit pivot plasma control system, which is a VS Code (‌Visual Studio Code, a free and open source cross-platform code editor) plug-in for the development of a spirit pivot plasma control system (PCS) control algorithm, integrates a visualization module and a retrieval augmented generation (RAG) driven intelligent question and answer module, and solves the traditional development problems through a visual interface, an automated process and intelligent support. The technical scheme of the application is described in detail below in combination with the flow chart and principle block diagram.

[0020] The intelligent algorithm development auxiliary tool for the spirit pivot plasma control system of the application includes a visualization module and an AI question and answer module.

[0021] The overall architecture of the present application takes an embedded web view (WebView) panel as the core interactive page, integrates a visualization module, an AI (artificial intelligence) question and answer module, and the WebView panel to form a collaborative intelligent development environment, as shown in Figure 1 .

[0022] The visualization module is used to parse the PCS project directory to generate an interactive view (tree diagram and organizational structure diagram) based on the WebView, support architecture parsing, file generation, code navigation (including jumping to the definition, positioning in the plug-in view), and bidirectional editing synchronization; the AI question and answer module uses a locally deployed retrieval augmented generation (RAG) system combined with a PCS knowledge vector database to provide domain-specific query and answer. The WebView panel is used as the interactive page of the user with the visualization module and the AI question and answer module, renders the tree view and organizational structure diagram of the visualization module, and displays the questions and answers of the AI question and answer module.

[0023] The WebView panel is used as the interactive page of the user with the visualization module and the AI question and answer module, renders the tree view and organizational structure diagram of the visualization module, and displays the questions and answers of the AI question and answer module. Each module closely cooperates through the WebView panel and the PCS file, and is closely linked to ensure that the development process from architecture parsing to code generation and knowledge query is efficient and consistent.

[0024] The structure and workflow of the visualization module of the present application are shown in Figure 2 , and the visualization module includes a PCS architecture tree construction module, a change content analysis module, and an automatic content generation module; the workflow of the visualization module includes:

[0025] Regarding the PCS architecture tree building module, when the intelligent algorithm development aid tool for the spiritus plasma control system of the present application is activated (plug-in activation) or the workspace spiritus plasma control system project file of the integrated development environment (VS Code) is changed (file change), the PCS directory of the workspace is scanned and the project file is parsed from the parsed spiritus plasma control system project file, the parsed project file includes Extensible Markup Language (XML) files under each directory, the hierarchical structure of control category, control algorithm, data item subset, etc. is extracted, the PCS architecture tree is built, and data basis is provided for subsequent operations. Based on this architecture tree, the PCS architecture tree building module renders a visual architecture view (including a tree view and an organizational structure diagram) through the WebView Application Programming Interface (API) provided by VS Code. Developers quickly locate nodes through keyword filtering and modify parameters (such as waveform data and control logic) using a form interface to modify the parameters of the visual architecture view (view editing changes). The changed content parsing module modifies the spiritus plasma control system project file in the workspace.

[0026] Regarding the changed content parsing module, when the visual architecture view of the WebView panel is edited and changed, the modified parameters are sent to the plug-in backend through the message passing (postMessage) mechanism. The changed content parsing module parses the corresponding PCS architecture tree node changes in the modified parameters, locks the files and changes in the workspace according to the changed PCS architecture node, and updates the Extensible Markup Language (XML) and Interface Description Language (IDL) files.

[0027] After the view in the WebView panel is modified, the spiritus plasma control system project file in the workspace is modified synchronously through the changed content parsing module. When the spiritus plasma control system project file in the workspace is modified externally, the intelligent algorithm development aid tool for the spiritus plasma control system detects the spiritus plasma control system project file change and re-synchronizes the data rendering page through the PCS architecture tree building module, realizes two-way synchronization, and ensures data consistency.

[0028] Regarding the automated content generation module, when the "Generate Code Framework" button is clicked on the page, the WebView panel will transmit the synchronized visual architecture view data directly to the automated content generation module by clicking "Generate File". The automated content generation module uses the IDL compiler to generate IDL files, and combines them with the C++ source code and configuration files generated by the template engine to generate statements for registering variables to the data engine in the Lingshu Plasma Control System project file.

[0029] The visualization module further enhances interactivity based on the Language Services Protocol (LSP) functionality. Figure 3 This is a flowchart illustrating the logical interaction between code and view in the visualization module of this invention. Code navigation (jumping to the definition and locating within the plugin view) is as follows: Figure 3 As shown in (a). Developers can jump from the variable name in the algorithm core logic (C++ file) to the variable definition in the Extensible Markup Language (XML) file, or locate the WebView node from the variable name in the algorithm core logic (C++ file) and jump directly to the corresponding node in the visual architecture view in the WebView panel.

[0030] For the implementation of the jump definition in the visualization module, such as Figure 3 As shown in (b), the plugin parses the registration statement in the algorithm header file, establishes a mapping relationship between variable identifiers and registration statements, and obtains the unique identifier (varname) of the variable. Using varname, the system searches for the configuration file, i.e., the XML file in the directory structure, locates the target location of the (variable), and jumps to the target location of the variable.

[0031] For the implementation of locating the visualization module in the plugin view, such as Figure 3 As shown in (c), by parsing the variable at the current cursor position, the algorithm header file registration statement is also parsed to obtain the variable's unique identifier (varname). During the visualization architecture parsing process, the system establishes a mapping relationship between varname and architecture nodes, i.e., a hash mapping table. Therefore, by looking up the hash mapping table using varname, the target position can be directly located on the page, opening the WebView panel and highlighting the corresponding node in the visualization architecture view. The architecture tree and parameter editing results of the visualization module provide input for automated generation, and the generated files, in turn, support LSP navigation, forming a closed-loop process that significantly reduces error rates and repetitive workload.

[0032] like Figure 4 As shown, the AI ​​question-answering module utilizes a local search-enhanced generation (RAG) system to display domain-specific answers through an embedded WebView panel, sharing an interface with the visualization module. Developers can access the technical guidance provided by the AI ​​question-answering module in real time while editing parameters.

[0033] The whole process starts from building a knowledge database. The AI Q&A module uses LanceDB (an efficient vector database) to store documents related to the plasma control system (PCS) such as technical manuals, experimental reports, and algorithm specifications. After data cleaning and structuring (knowledge processing), these documents are converted into high-dimensional embedding vectors using bge-large-zh (a Chinese embedding model). Deep learning is used to capture semantic information and store it back into LanceDB, forming a structured knowledge database: data block knowledge base, providing reliable data sources for subsequent queries. At the same time, local deployment ensures the confidentiality and integrity of sensitive information in the LingShu PCS.

[0034] Based on this database, developers input queries (such as "how to configure plasma current parameters") into the WebView panel. DeepSeek-R1 Large Language Model (LLM) rewrites the query based on the input to optimize keywords and semantics, enhancing retrieval accuracy. The RAG system vectorizes the rewritten query using bge-large-zh. Multi-path recall retrieval (vector retrieval and keyword retrieval) is used to feedback the most relevant results (questions + relevant knowledge). Vector retrieval matches similarity based on embedding vectors, while keyword retrieval matches based on key terms in the query. The most relevant document fragments are retrieved. The query content and the retrieved most relevant document fragments (questions and relevant knowledge) are input into the DeepSeek-R1 prediction model as context to generate accurate and structured answers. The output answers are displayed in text form on the WebView panel.

[0035] The local deployment of the large language model in the AI Q&A module ensures data security, and the answer content can guide developers to adjust parameters or optimize algorithms during the development of the PCS control algorithm. Developers can edit and verify the guidance content simultaneously, enhancing development efficiency and learning experience, and providing real-time support for the LingShu PCS algorithm development.

[0036] The present invention proposes an intelligent algorithm development tool for the LingShu plasma control system. By integrating a visualization module (page display, automatic file generation, and language service agreement (LSP) function), and an RAG (retrieval enhancement generation) driven AI Q&A module, it solves the problems of complex parameter configuration, lack of visualization, insufficient automation, and lack of domain intelligence support in traditional PCS algorithm development. The invention provides an intuitive WebView panel to realize the visualization of PCS architecture, ensures real-time consistency between parameter editing and files through a bidirectional synchronization mechanism, automatically generates code and configuration files to reduce repetitive work, and provides domain-specific technical guidance through a local RAG system, significantly improving development efficiency, accuracy, and engineering level, providing a new development solution for nuclear fusion research.

[0037] In addition, alternative solutions that can achieve the purpose of the invention include:

[0038] (1) Independent desktop application instead of VS Code plugin:

[0039] An independent desktop application with a custom graphical user interface (GUI) can be developed to replace the VS Code plugin (no longer based on the environment of the VS Code editor), retaining the parameter editing of the visualization module and the functions of the AI question and answer module. This solution increases development costs and loses the integration advantages of the VS Code IDE, and requires redesign of related synchronization and interaction logic to adapt to the new GUI framework.

[0040] (2) Cloud deployment instead of local RAG system:

[0041] The RAG system can be adjusted from local deployment to cloud deployment, using a third-party large language model (LLM), integrated with the VS Code plugin through a secure API, retaining query processing functions, to improve system processing capacity and data storage scalability. However, this alternative sacrifices local data security and may introduce data leakage risks.

[0042] (3) Other models instead of DeepSeek-R1 large language model (LLM):

[0043] The invention uses DeepSeek-R1 large language model as the default choice, as it performs well in general semantic understanding and response efficiency. DeepSeek-R1 large language model (LLM) can be replaced by other adaptive models, such as LLaMA (Large Language Model, a highly efficient open-source language model) or GPT-Neo (Generative Pretrained Transformer, another open-source language model), combined with local LanceDB (efficient vector database) for semantic retrieval and knowledge matching, maintaining the integration of RAG pipeline and WebView panel. Other plugin settings remain unchanged.

[0044] Those skilled in the art will appreciate that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented in various computer languages.

[0045] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart

[0046] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart

[0047] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheet block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 Figure 1 means for functionally implementing the steps listed in the flowchart

[0048] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments described and illustrated herein, without departing from the spirit and scope of the application. Accordingly, it is intended that all subject matter contained in the above description be interpreted as illustrative only and basis for claims.

[0049] It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application cover the modifications and variations of this application provided they come within the scope of the appended claims and their equivalents.

[0050] The embodiments described above are only examples of the present application, and the scope of the present application is not limited thereto. Any equivalent structure or equivalent flowchart variation of the present application, or direct or indirect application thereof to other related fields, are also included in the scope of the present application.

[0051] The description herein of any aspects of the application encompassing present technology that is not specifically described in the specification is considered to be prior art to the application.

Claims

1. An intelligent algorithm development aid for a wick plasma control system, characterized by, Comprise: Visual module, AI question and answer module, WebView panel; The visual module is used to parse the PCS project directory, generate an interactive view based on WebView, support architecture parsing, generate files, code navigation, and two-way editing synchronization; The visual module comprises: a PCS architecture tree construction module, a change content analysis module, and an automatic generation content module; The AI question and answer module adopts a locally deployed search enhancement generation system, combines a PCS knowledge vector database, and provides query and answer in the PCS field of plasma control system; The WebView panel is used as an interactive page of the user with the visual module and the AI question and answer module, renders the tree view and organization architecture diagram of the visual module, and displays the question and answer of the AI question and answer module; Regarding the change content analysis module of the visual module, when the visual architecture view of the WebView panel is edited and changed, the modified parameters are sent to the plug-in backend through the message passing mechanism, the change content analysis module analyzes the changes of the corresponding PCS architecture tree nodes in the modified parameters, analyzes the corresponding files and change contents that need to be changed in the PCS architecture node according to the changed PCS architecture node, and updates the extensible markup language and interface description language files; After the view in the WebView panel is modified, the work area of the plasma control system project file is modified synchronously through the change content analysis module, and when the work area of the plasma control system project file is externally modified, the intelligent algorithm development tool for plasma control system detects the change of the plasma control system project file and re-synchronizes the data rendering page through the PCS architecture tree construction module.

2. The intelligent algorithm development aid for a wits-on-plasma control system of claim 1, wherein, Regarding the PCS architecture tree construction module of the visual module, when the intelligent algorithm development tool for plasma control system is activated or the work area plasma control system project file of the integrated development environment is changed, the PCS directory of the work area is scanned and the project file is parsed from the parsed plasma control system project file, the parsed project file includes extensible markup language files under each directory, control directories, control algorithms, and data item subsets are extracted, and a PCS architecture tree is constructed; Based on the PCS architecture tree, the PCS architecture tree construction module renders the visual architecture view through the WebView application programming interface provided by VS Code, the developer quickly locates the node through keyword filtering, modifies the parameters of the visual architecture view using the form interface, and modifies the work area of the plasma control system project file through the change content analysis module.

3. The intelligent algorithm development aid for a wits-and-crafts plasma control system of claim 2, wherein, Regarding the automatic generation content module of the visual module, when the code framework generation box in the page is clicked, the synchronized visual architecture view data of the WebView panel is directly transmitted to the automatic generation content module through the click generation file, the automatic generation content module generates an IDL file using an IDL compiler, combines C++ source code and configuration files generated by a template engine, and generates a statement for registering variables to a data engine in the plasma control system project file.

4. The intelligent algorithm development aid for a wits-and-crafts plasma control system of claim 3, wherein, Regarding the automatic content generation module, when the code generation framework is clicked in the page, the WebView panel directly transmits the synchronized visual architecture view data to the automatic content generation module through the click generation file, the automatic content generation module generates an IDL file by using an IDL compiler, and generates a variable registration statement in a project file of a plasma control system in combination with C++ source code and a configuration file generated by a template engine.

5. The intelligent algorithm development aid for a wits-in-the-box plasma control system of claim 1 wherein, The code navigation includes: jumping to a definition, locating in a plugin view; For the implementation of jumping to a definition, a plugin parses a header file registration statement to establish a mapping relationship between a variable identifier and the registration statement, and obtain a unique identifier of the variable; the system finds a configuration file through the unique identifier of the variable, locates a target position and jumps to the target position; For the implementation of locating in a plugin view, a variable at a current cursor position is parsed, and a header file registration statement is also parsed to obtain a unique identifier of the variable; in a visual architecture parsing process, the system establishes a mapping relationship between the unique identifier of the variable and an architecture node, that is, a hash mapping table, and directly locates a target position in a page through the unique identifier of the variable to open a WebView panel and highlight a corresponding node in a visual architecture view; an architecture tree and a parameter editing result of a visual module provide input for automatic generation, and a generated file supports LSP navigation in turn to form a closed-loop process.

6. The intelligent algorithm development aid for a wits-in-the-box plasma control system of claim 1 wherein, The AI question and answer module utilizes a local retrieval enhancement generation system, displays a plasma control system field answer through an embedded WebView panel, and shares an interface with a visual module; a developer can query technical guidance given by the AI question and answer module in real time when editing parameters; Starting from constructing a knowledge database, the AI question and answer module stores plasma control system related documents in a high-efficiency vector database, generates high-dimensional embedded vectors from the documents after data cleaning and structuring, captures semantic information through deep learning, and stores the semantic information back to the high-efficiency vector database to form a structured knowledge database: a data block knowledge base; at the same time, through local deployment, the security and integrity of sensitive information of the plasma control system are ensured; Based on the data block knowledge base, a developer inputs a query in a WebView panel, a DeepSeek-R1 large language model rewrites the query according to the input to optimize keywords and semantics, and enhances the retrieval accuracy; the retrieval enhancement generation system vectorizes the rewritten query based on bge-large-zh; the most relevant result is fed back through multi-path recall retrieval, vector retrieval is matched according to embedded vectors, and keyword retrieval is matched according to key terms in the query, and the most relevant document fragment is retrieved; the query content and the most relevant document fragment are input as context into a DeepSeek-R1 prediction model to infer and analyze to generate an accurate and structured answer, and the output answer is displayed in a text form in the WebView panel.

7. The intelligent algorithm development aid for a wits-in-the-box plasma control system of any of claims 2 to 4, wherein, Develop a standalone desktop application with a custom graphical user interface to replace the VS Code plugin, retaining the parameter editing of the visualization module and the functionality of the AI question-answering module.

8. The intelligent algorithm development aid for a wits-in-the-box plasma control system of any of claims 1 to 6, wherein, Adjust the retrieval augmentation generation system from local deployment to cloud deployment, utilize third-party large language models, integrate with the VS Code plugin through secure APIs, and retain query processing functionality.

9. The intelligent algorithm development aid for a wits-in-the-box plasma control system of any of claims 1 to 6, wherein, Replace the DeepSeek-R1 large language model with LLaMA, GPT-Neo, or other adapted models, and combine it with a local efficient vector database for semantic retrieval and knowledge matching, maintaining the integration of the retrieval augmentation generation pipeline and the WebView panel.

Images