SCADA (Supervisory Control And Data Acquisition) system based on large language model, implementation method and equipment

Abstract

The invention relates to the technical field of monitoring and data acquisition systems, and discloses an SCADA (supervisory control and data acquisition) system based on a large language model, an implementation method and equipment, and the system comprises a user interaction terminal which receives a natural language instruction input by a user and displays a dynamic user interface; the natural language processing module is used for preliminarily analyzing the natural language instruction to obtain a user intention and key information; the large language model is used for deeply analyzing the user intention and the key information and generating a task description or a task execution plan; the MCP arrangement engine is used for mapping the task description or the task execution plan to a predefined function to obtain executable structured data; and the dynamic UI generator is used for generating a dynamic user interface based on the executable structured data and a preset UI instruction. According to the method, the usability and the data analysis capability of the SCADA system are improved through natural language interaction, and meanwhile, the reliability and the safety of the system are ensured by utilizing an MCP and a predefined function mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supervisory control and data acquisition systems, in particular to a SCADA system based on a large language model, an implementation method and equipment. BACKGROUND

[0002] Supervisory Control And Data Acquisition (SCADA) systems play a core role in the field of industrial automation, enabling remote monitoring and data acquisition of industrial processes. However, existing SCADA systems have significant limitations in human-machine interaction, data analysis, and flexibility. First, the Human Machine Interface (HMI) is usually pre-configured, and if an operator wants to perform non-standardized data queries, generate temporary custom reports or monitoring charts, they often need to have professional programming or configuration skills, or rely on system integrators for secondary development, which greatly limits the flexibility and immediacy of data exploration. Second, the data analysis functions of traditional SCADA systems are relatively basic, and when faced with massive historical data and real-time logs, complex fault diagnosis, correlation analysis or predictive maintenance often require operators to perform a large amount of manual screening and comparison, which is inefficient and prone to errors.

[0003] In summary, existing SCADA systems have the problem of complex operation and poor flexibility in user-defined data analysis and display. Artificial intelligence technologies such as large language models have shown strong natural language understanding and complex task processing capabilities, and are expected to improve the usability and intelligence of SCADA systems, but the output content of the model has uncontrollable characteristics, i.e. it is difficult to directly apply to SCADA systems that require high reliability due to the "hallucination" problem. SUMMARY

[0004] Therefore, the present application provides a SCADA system based on a large language model, an implementation method and equipment to solve the problem that large language models are difficult to be directly applied to SCADA systems that require high reliability due to the "hallucination" problem.

[0005] In a first aspect, the present application provides a SCADA system based on a large language model, which comprises:

[0006] a user interaction terminal for receiving a natural language instruction input by a user and displaying a dynamic user interface;

[0007] a natural language processing module for performing preliminary analysis on the natural language instruction input by the user to obtain a user intent and key information;

[0008] A large language model is used to deeply analyze user intent and key information, and generate a task description or a task execution plan.

[0009] An MCP orchestration engine is used to map the task description or the task execution plan to a predefined function to obtain executable structured data.

[0010] A dynamic UI generator is used to generate a dynamic user interface based on the executable structured data and preset UI instructions.

[0011] The SCADA system based on the large language model provided by the application ingeniously combines the natural language understanding capability of the large language model (LLM) and the reliability requirement of the SCADA system by introducing an MCP (Model Context Protocol) orchestration engine and a predefined function, isolates the LLM from the core operation of the SCADA system through the MCP orchestration engine, makes the LLM only output a task plan instead of a direct control instruction, ensures that the system operation is strictly limited within a predefined safe range, converts the structured data into a visual interface required by the user through the dynamic UI generator, combines the natural language input and output, forms a complete interactive closed loop of instruction, analysis and feedback, improves the ease of use, greatly improves the ease of use of the SCADA system, allows the user to perform complex data query and interface customization through natural language, ensures the reliability and safety of the system, solves the influence of the LLM "illusion" problem on the SCADA data and analysis results through the design of the data interaction mechanism, improves the data analysis efficiency, and enables the operator to quickly obtain deep business insight and fault diagnosis support.

[0012] In an optional implementation, the natural language processing module is further configured to perform word segmentation processing, text filtering and text regularization data preprocessing on the input natural language instruction.

[0013] The SCADA system based on the large language model provided by the application performs word segmentation processing, text filtering and text regularization data preprocessing on the input natural language instruction through the natural language processing module, improves the subsequent processing efficiency, enhances the semantic understanding accuracy, retains the core vocabulary which is more focused on the key content of the user instruction, reduces the interference of irrelevant information on the understanding of the LLM, and makes the intent recognition and key information extraction more accurate.

[0014] In an optional implementation, the SCADA system based on the large language model further includes a SCADA predefined function library and a SCADA_MCP service, and the SCADA predefined function library performs data interaction with the MCP orchestration engine through the SCADA_MCP service.

[0015] The SCADA predefined function library is used to store a plurality of standardized and verified SCADA operation functions, and the SCADA operation functions include data query operation functions, log retrieval operation functions, statistical calculation operation functions and fault code analysis operation functions.

[0016] The SCADA_MCP service is used to respond to function call requests of the MCP orchestration engine.

[0017] The SCADA system based on the large language model provided by the application fundamentally solves the contradiction between natural language interaction and industrial control system security by storing standardized and verified operation functions (data query, log retrieval, statistical calculation, fault code analysis, etc.), and the SCADA_MCP service serves as an interaction bridge between the predefined function library and the MCP orchestration engine, realizes standardized communication through a model context protocol (MCP), and the core value lies in improving interface security, compatibility and expandability.

[0018] In an optional implementation, the SCADA system based on the large language model further includes a support MCP protocol database and a database MCP service, and the support MCP protocol database is used to interact with the MCP orchestration engine through the database MCP service.

[0019] The support MCP protocol database is used to store and manage historical data, logs and configuration information of the SCADA system.

[0020] The database MCP service is used to respond to data query requests of the MCP orchestration engine.

[0021] The SCADA system based on the large language model provided by the application supports the support MCP protocol database (such as a MongoDB cluster configured with MCP) to specially store core information such as historical data, operation logs and device configurations of the SCADA system, adapt to industrial data characteristics, improve storage and query efficiency, the MCP protocol can integrate a data verification mechanism (such as request signature and data transmission encryption), ensure that the query request received by the database comes from a legal MCP orchestration engine, avoid malicious injection attacks (such as SQL injection) or illegal data theft, guarantee the integrity of industrial data, and the database MCP service serves as an interaction bridge between the support MCP protocol database and the MCP orchestration engine, and undertakes core functions such as analyzing query requests, executing data retrieval and returning structured results, the query request received by the database MCP service comes from the MCP orchestration engine, and these requests are based on key information after natural language preprocessing. The interface strictly matches these structured parameters to execute the query, ensures that the returned data is highly consistent with the user instruction, fundamentally avoids invalid queries caused by LLM "illusion" (such as fictitious device ID or time range), and guarantees the accuracy of data.

[0022] In one optional implementation, the predefined function is at least one function call sequence in the SCADA predefined function library;

[0023] The MCP orchestration engine is also used to map task descriptions or task execution plans to at least one function call sequence in the SCADA predefined function library based on preset rule sets and preset knowledge bases, to obtain executable structured data, and to initiate function call requests to the SCADA_MCP service through the MCP protocol.

[0024] This invention provides a SCADA system based on a large language model. The MCP orchestration engine, based on the mapping logic of a preset rule set and knowledge base, is the core design that balances the convenience of LLM natural language interaction with the industrial-grade security of SCADA. It eliminates risks through security filtering, ensures reliability through standardized mapping, improves controllability through transparent processes, and adapts the professional knowledge base to industrial scenarios. Ultimately, the new SCADA system enables the MCP orchestration engine to quickly adapt to new scenarios without modifying the LLM or core module logic.

[0025] In one alternative implementation, the preset UI instructions include interface type instructions in the user's intent;

[0026] The dynamic UI generator also uses executable structured data and interface type instructions in the user's intent to call preset icon libraries or preset UI components to generate a user interface, and then presents the user interface on the user's interactive terminal. The SCADA system based on a large language model provided by this invention avoids the LLM illusion problem because the input data is structured data verified by predefined functions and the MCP protocol, and the UI generator itself does not create content. This ensures that every data point and chart seen by the user originates from the database.

[0027] Secondly, this invention provides a method for implementing a SCADA system based on a large language model, the method comprising:

[0028] Receive natural language commands input by the user;

[0029] The user's input natural language commands are initially parsed to obtain user intent recognition and key information;

[0030] Perform in-depth analysis of user intent and key information to generate task descriptions or task execution plans;

[0031] Map the task description or task execution plan to a preset available function to obtain executable structured data;

[0032] A dynamic user interface is generated based on executable structured data and interface type instructions in the user intent, and then displayed.

[0033] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the SCADA system implementation method based on a large language model as described in the first aspect or any corresponding embodiment.

[0034] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the SCADA system implementation method based on a large language model as described in the first aspect or any corresponding embodiment thereof.

[0035] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the SCADA system implementation method based on a large language model as described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0036] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the structure of a SCADA system based on a large language model according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of another SCADA system based on a large language model according to an embodiment of the present invention;

[0039] Figure 3 This is a flowchart illustrating the implementation method of a SCADA system based on a large language model according to an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0041] In order to make the purposes, 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 with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.

[0042] In recent years, large language models have been considered to have the potential to revolutionize the human-machine interaction mode of SCADA due to their powerful natural language processing and logical reasoning capabilities. Although artificial intelligence technologies such as large language models (LLM) exhibit strong natural language understanding and complex task processing capabilities, they have the potential to improve the ease of use and intelligence of SCADA systems. However, the "hallucination" problem of LLM, which can generate seemingly reasonable but actually incorrect or ex nihilo information, is unacceptable in the field of industrial control that requires reliability and safety. It can also be said that it can generate data or logic that does not conform to reality, making it face great challenges in the application of industrial SCADA systems that have extremely high requirements for safety and reliability. If LLM is directly applied to SCADA systems, especially in the data display and analysis link, there is a significant risk of introducing incorrect information and misleading operators' decision-making, which hinders its deep application in the SCADA field. How to safely and reliably use LLM to improve the ease of use and intelligence of SCADA systems is a technical problem that needs to be solved urgently.

[0043] The embodiments of the present application provide a SCADA system based on a large language model language. By introducing an MCP (Model Context Protocol, MCP for short) orchestration engine and a predefined function, the natural language understanding capability of the large language model (Large Language Model, LLM for short) is ingeniously combined with the reliability requirements of the SCADA system. The LLM is isolated from the core operation of the SCADA system through the MCP orchestration engine, so that the LLM only outputs task planning rather than direct control instructions, ensuring that system operation is strictly limited within a predefined safe range. A dynamic UI (User Interface, UI for short) generator is used to convert structured data into a visual interface required by the user, and natural language input and output are combined to form a complete interactive closed loop of instructions, analysis and feedback, thereby improving the ease of use and greatly improving the ease of use of the SCADA system. The user can perform complex data queries and interface customization through natural language, ensuring the reliability and safety of the system, and the influence of the LLM "hallucination" problem on SCADA data and analysis results is solved through data interaction mechanism design.

[0044] The present application aims to solve the problems of existing SCADA systems in user-defined data analysis and display, such as complex operation, poor flexibility, and the problem of large language models (LLM models) being difficult to directly apply to SCADA systems requiring high reliability due to the "hallucination" problem. The embodiments of the present application provide a SCADA system based on a large language model language, which can realize natural language interaction using LLM to improve the usability and data analysis capability of the SCADA system while ensuring system reliability and security.

[0045] According to the embodiments of the present application, a SCADA system based on a large language model is provided, Figure 1 is a structural schematic diagram of the SCADA system based on a large language model according to the embodiments of the present application, as Figure 1 shown, the system comprises:

[0046] User interaction terminal 1 for receiving user input natural language instructions; natural language processing module 2 for preliminary analysis of user input natural language instructions to obtain user intent and key information; large language model 3 for deep analysis of user intent and key information to generate task description or task execution plan; MCP orchestration engine 4 for mapping task description or task execution plan to predefined functions to obtain executable structured data; dynamic UI generator 9 for generating user interface based on executable structured data and preset UI instructions.

[0047] Specifically, as Figure 2 shown, the system comprises:

[0048] User interaction terminal 1 (User Interaction Terminal): receives user input natural language instructions and displays dynamic user interface generated by the system.

[0049] Natural language processing module 2 (NLP Module): pre-processes, identifies intent, and extracts key information from input natural language. The natural language processing module 2 is also used for word segmentation processing, text filtering, and text normalization data preprocessing of input natural language instructions.

[0050] Large language model 3 (Large Language Model, LLM for short): understand user intent, generate task description or task execution plan.

[0051] MCP orchestration engine 4: a core component responsible for mapping the output of LLM to predefined functions and coordinating interactions with backend systems through the MCP protocol. The MCP protocol (Model Context Protocol) aims to establish a unified communication framework for large language models, MCP orchestration engines, data sources, and services, enabling efficient cross-platform data interaction.

[0052] The system of the embodiment introduces an MCP orchestration engine 4 as a key intermediate layer between LLM and the core functions of the SCADA system. The user's natural language request is first analyzed by the natural language processing module 2 to extract the user's intent and key information, and then submitted to the large language model 3 for deep understanding and task planning. However, the LLM model does not directly access the SCADA system or database, but generates a task description or task execution plan that conforms to the MCP specification.

[0053] The MCP-compliant task description or plan refers to a series of intermediate instruction sets generated by the LLM model based on the MCP specification and predefined function library and generation rules, describing the process and content required to complete the user's request.

[0054] Dynamic UI generator 9: generates a dynamic user interface based on executable structured data and UI instructions (including interface type instructions in the user's intent) passed by the MCP orchestration engine. The dynamic UI generator 9 receives executable structured and trusted data returned by the predefined function execution from the MCP orchestration engine, and dynamically combines and renders the user interface according to the interface type requirements (such as tables, trend charts, dashboards, etc.) implied in the user's original request (i.e., UI instructions). This ensures that the interface and data the user ultimately sees are reliable SCADA functions and databases, rather than LLM-generated free-form, thus eliminating the impact of the "hallucination" problem on the final results.

[0055] The SCADA system based on large language models provided by the embodiment allows the capabilities of the LLM model to be used to understand user intent and dynamically orchestrate task flows, while strictly limiting its actual operations within predefined, secure functions and performing reliable execution and data acquisition through the MCP protocol, ultimately presenting trusted results through the dynamic UI generator. The "hallucination" problem of LLM affects SCADA data and analysis results, improving data analysis efficiency and enabling operators to quickly obtain deep business insights and fault diagnosis support.

[0056] In an optional embodiment, the large language model-based SCADA system further comprises a SCADA predefined function library 5 and a SCADA_MCP service 6, the SCADA predefined function library 5 interacts with the MCP orchestration engine 4 through the SCADA_MCP service 6; the SCADA predefined function library is used to store a plurality of standardized and verified SCADA operation functions, the SCADA operation functions include data query operation functions, log retrieval operation functions, statistical calculation operation functions and fault code analysis operation functions; the SCADA_MCP service (SCADA MCP Server) is used to respond to the function call request of the MCP orchestration engine.

[0057] Wherein, the predefined function is at least one function call sequence in the SCADA predefined function library. The MCP orchestration engine is further configured to map the task description or the task execution plan to at least one function call sequence in the SCADA predefined function library based on a preset rule set and a preset knowledge base, to obtain an executable structured data, and to initiate a function call request to the SCADA_MCP service through the MCP protocol.

[0058] It can be understood that the large language model-based SCADA system provided in the embodiment has a SCADA predefined function library built-in. The function library contains a series of atomic operation functions that have been strictly tested, verified and have clear functions, these functions cover the commonly used functions of data query, log retrieval, statistical calculation, fault code analysis and the like of the SCADA system, but do not contain any control instructions. These functions are encapsulated and exposed to the outside in the form of "Tools" through a SCADA_MCP service.

[0059] In an optional embodiment, the large language model-based SCADA system further comprises a support MCP protocol database 7 and a database MCP service 8, the support MCP protocol database interacts with the MCP orchestration engine through the database MCP service; the support MCP protocol database is used to store and manage the historical data, logs and configuration information of the SCADA system; the database MCP service is used to respond to the data query request of the MCP orchestration engine.

[0060] It can be understood that after the MCP orchestration engine receives the task description generated by the LLM model, it does not directly execute the instructions of the LLM model, but strictly maps it to one or more function call sequences in the predefined function library. This mapping process is based on a preset rule set and knowledge base, ensuring that the function call request of the LLM model is ultimately converted into a series of safe, controllable, and deterministic function executions. The MCP orchestration engine then initiates a function call request to the SCADA_MCP service through the MCP protocol. The SCADA system based on a large language model provided in this embodiment utilizes a database supporting the MCP protocol (such as MongoDB and its database MCP service) to store and manage historical data, logs, configuration information, etc. of the SCADA. The MCP orchestration engine also interacts with the database_MCP service through the MCP protocol to safely and efficiently obtain the required data. The use of the MCP protocol ensures the standardization, structuring, and security of communication between the LLM, the MCP orchestration engine, the SCADA predefined function library, and the database supporting the MCP protocol.

[0061] The embodiment of the application also provides a method for implementing a SCADA system based on a large language model, and a flowchart of the method is as shown in Figure 3 The method comprises the following steps:

[0062] Step S301: receiving a natural language instruction input by a user.

[0063] Specifically, the natural language instruction input by the user is received through the user interaction terminal 1, and the natural language instruction is transmitted to the natural language processing module 2.

[0064] Step S302: performing preliminary analysis on the natural language instruction input by the user to obtain user intent recognition and key information.

[0065] Specifically, the natural language processing module 2 performs user intent recognition and key information extraction after performing data preprocessing such as word segmentation processing, text filtering, and text regularization on the input natural language instruction, obtains user intent and key information, and transmits the user intent and key information to the large language model 3 (i.e. LLM model).

[0066] Step S303: performing deep analysis on the user intent and key information to generate a task description or a task execution plan.

[0067] Specifically, the LLM model uses its powerful understanding ability to analyze user requirements and generate a preliminary task execution plan or task description. Importantly, this plan is not a direct system operation instruction, but a high-level description. The LLM model transmits the task execution plan or task description to the MCP orchestration engine 4.

[0068] Step S304, mapping the task description or task execution plan to the preset available functions to obtain executable structured data.

[0069] Specifically, after receiving the task description generated by the LLM model, the MCP orchestration engine 4 does not directly execute the instructions of the LLM model, but strictly maps it to one or more function call sequences in the predefined function library. This mapping process is based on a preset rule set and knowledge base, ensuring that the function call request of the LLM model is ultimately converted into a series of safe, controllable, and deterministic function executions. The MCP orchestration engine then initiates a function call request to the SCADA_MCP service through the MCP protocol. The SCADA system based on large language models provided in this embodiment utilizes a database supporting the MCP protocol (such as MongoDB and its database MCP service) to store and manage historical data, logs, configuration information, etc. of the SCADA. The MCP orchestration engine also interacts with the database_MCP service through the MCP protocol to safely and efficiently obtain the required data. The use of the MCP protocol ensures the standardization, structuring, and security of communication between the LLM, the MCP orchestration engine, the SCADA predefined function library, and the database supporting the MCP protocol.

[0070] Step S305, generating a dynamic user interface based on the executable structured data and the interface type instruction in the user intent, and displaying the dynamic user interface.

[0071] Specifically, the dynamic UI generator 9 receives structured and trusted data returned by the predefined function execution from the MCP orchestration engine, and dynamically combines and renders the user interface according to the interface type (such as table, trend chart, dashboard, etc.) requirement implied in the user's original request, and displays the dynamic user interface through the user interaction terminal 1. This ensures that the interface and data ultimately seen by the user are from the reliable SCADA predefined function library 5 and the database supporting the MCP protocol 7, rather than the free generation of the LLM, thereby eliminating the influence of the "illusion" problem on the final result.

[0072] As one or more specific application embodiments of the embodiments of the present application, in combination with Figure 2 The SCADA system based on large language models provided by the present application is further described in detail as follows: Figure 2 As shown in the figure, the system comprises:

[0073] User interaction terminal 1 (User Interaction Terminal): receives natural language instructions input by the user, and displays the dynamic user interface generated by the system.

[0074] User Interaction Terminal (User Interaction Terminal): receives natural language instructions input by the user, and displays dynamic user interfaces generated by the system.

[0075] Natural Language Processing Module 2 (NLP Module): preprocesses, identifies intent and extracts key information from input natural language. The natural language processing module 2 is also used for word segmentation processing, text filtering and text normalization data preprocessing of input natural language instructions.

[0076] Large Language Model 3 (Large Language Model, LLM for short): understand user intent, generate task description or task execution plan.

[0077] MCP orchestration engine 4 (MCP Orchestration Engine): the core component, responsible for mapping the output of the LLM to the pre-defined function, and coordinating the interaction with the back-end system through the MCP protocol. Among them, MCP protocol (Model Context Protocol, MCP for short) aims to establish a unified communication framework for large language models (LLM model), MCP orchestration engine, data source and service, to realize cross-platform and efficient data interaction.

[0078] SCADA pre-defined function library 5 (SCADA Pre-defined Function Library): contains a series of safe, read-only, verified SCADA operation functions.

[0079] SCADA_MCP service 6 (SCADA MCP Server): as the MCP interface of the function library, it responds to the function call request of the orchestration engine.

[0080] MCP-enabled database 7 (MCP-enabled Database): stores SCADA historical data (such as historical operation data), logs, configuration information, etc., such as MongoDB.

[0081] Database MCP service 8 (Database MCP Server): as the MCP interface of the database, it responds to the data query request of the orchestration engine.

[0082] Dynamic UI generator 9 (Dynamic UI Generator): generates user interfaces according to the structured data and UI instructions passed by the orchestration engine.

[0083] The workflow of the SCADA system based on large language model provided by the present application is as follows:

[0084] The user inputs a natural language instruction through the user interaction terminal 1, for example, "Please generate a trend chart of the main hoist motor temperature and vibration frequency in the past 8 hours, and highlight all alarm points where the temperature exceeds 80 degrees Celsius."

[0085] The instruction is sent to the natural language processing module 2, which performs word segmentation, removes stop words, identifies user intent (generating trend chart, highlighting alarms), and extracts key entities (time range: past 8 hours; object: A production line No. 1 motor; parameters: temperature, vibration frequency; conditions: temperature > 80 degrees Celsius).

[0086] The processed information is sent to the large language model 3 (LLM model), which uses its powerful understanding ability to analyze user needs and generate a preliminary task execution plan. Importantly, this plan is not a direct system operation instruction, but a high-level description, such as: "Need to obtain main hoist motor temperature data (8 hours), obtain main hoist motor vibration data (8 hours), obtain main hoist motor temperature alarm data (8 hours, > 80 degrees), and then display these data on a trend chart."

[0087] The task execution plan is passed to the core MCP orchestration engine 4, which is the core component of the system. The MCP orchestration engine 4 does not blindly trust the LLM plan, but takes it as input and starts a strict mapping and execution process. It will query the internal knowledge base and rule set to accurately map the LLM plan to the functions available in the SCADA predefined function library 5. For the above example, it may be mapped to the following function call sequence:

[0088] Func_GetData('Motor_A1_Temp','Now-8H','Now');

[0089] Func_GetData('Motor_A1_Vib','Now-8H','Now');

[0090] Func_GetAlarms('Motor_A1_Temp','Now-8H','Now','>80')。

[0091] The SCADA predefined function library 5 guarantees the security of the system. It only contains authorized and well-tested functions. These functions are designed to be read-only or perform secure calculation analysis, such as Func_GetData for getting historical data, Func_GetAlarms for querying alarm logs, Func_Calculate_Avg for calculating average, Func_Correlate_Events for analyzing event correlation, etc. Any function that contains modifying device status or issuing control instructions is rejected. These functions are exposed through the SCADA_MCP service 6. The MCP protocol ensures the standardization and security of the interface, and only legitimate requests can call these functions.

[0092] At the same time, the massive historical data, logs, etc. of the SCADA system are stored in the MCP-supported database 7, such as a MongoDB cluster configured with a database MCP service 8. If the MCP orchestration engine 4 needs to query the database directly (or query inside the predefined function), it also interacts with the database MCP service 8 through the MCP protocol to obtain executable structured data.

[0093] The MCP orchestration engine 4 sends a call request to the SCADA_MCP service 6 and / or the database MCP service 8 through the MCP protocol. The server executes the corresponding function or query and returns the result to the MCP orchestration engine 4 in a structured data format (such as JSON format) through the MCP protocol.

[0094] After the MCP orchestration engine 4 collects all the required data, it combines the interface rendering requirements parsed from the user's original request (such as "trend chart"), and sends these trusted, structured data and UI instructions (such as {'type':'line_chart','data':[...],'highlight':[...]} ) to the dynamic UI generator 9.

[0095] The dynamic UI generator 9 is responsible for the final rendering of the interface. It calls the corresponding chart library or UI component according to the received instructions and data, generates the interface requested by the user, and presents it on the user interaction terminal 1. Since the input data is structured data verified by predefined functions and the MCP protocol, and the dynamic UI generator 9 itself does not create content, the LLM model hallucination problem is avoided, ensuring that every data point and chart the user sees is from the SCADA predefined function library and the MCP protocol-supported database.

[0096] Taking the daily fault comprehensive analysis of the shore-based container crane (shore crane) as an example, the process includes:

[0097] User input: "Find all major faults of QC_07 crane in the past 24 hours and analyze possible causes."

[0098] Natural Language Processing module 2 (NLP) extracts the intent (find all major faults, analyze possible causes) and key entities (i.e., critical information) (object: QC_07 crane; time: past 24 hours).

[0099] Large Language Model 3 understands that the user needs a comprehensive, cross-subsystem analysis and generates a multi-step plan, including:

[0100] 1. Retrieve all major fault records of QC_07 crane in the past 24 hours (e.g., motor trip, spreader failure, PLC communication error, sensor failure, etc.).

[0101] 2. Retrieve all relevant alarm logs during the same period (e.g., overload, overspeed, limit, emergency stop, etc.).

[0102] 3. Retrieve all operator manual interventions or critical operation records during the same period.

[0103] 4. Perform time series sorting and correlation analysis on all events to find event clusters or causal chains.

[0104] 5. Based on the discovered correlation patterns, query the crane fault diagnosis knowledge base to propose possible root causes.

[0105] 6. Generate a comprehensive analysis report.

[0106] MCP orchestration engine 4 receives the plan output from the LLM model and accurately maps it to a more specific sequence of function calls available in the SCADA pre-defined function library 5. This may involve multiple functions:

[0107] Func_GetMajorFaults('QC_07','Now-24H','Now')-> calls the SCADA_MCP service to return the list of major faults.

[0108] Func_GetAlarmLogs('QC_07','Now-24H','Now','Severity>=High')-> calls the SCADA_MCP service to return high-priority alarms.

[0109] Func_GetOperatorLogs('QC_07','Now-24H','Now','Critical_Only')-> calls the SCADA_MCP service to return critical operation records.

[0110] Func_MultiCorrelateEvents(fault_data,alarm_data,operator_data,'15Min_Window') -> invokes the SCADA_MCP service, which is a core analysis function responsible for building event timelines and identifying potential correlations.

[0111] Func_RootCauseAnalysis(correlation_results,'QC_07_KB') -> invokes the SCADA_MCP service to perform root cause inference based on correlation results and the knowledge base.

[0112] The SCADA pre-defined function library executes, they are invoked through the SCADA_MCP service 6. These functions retrieve the required log data of various types from the supporting MCP protocol database 7, and execute complex analysis logic. For example, Func_MultiCorrelateEvents might find that after a PLC communication error, multiple subsystems' sensor readings are abnormal and eventually a spreader operation failure occurs. Func_RootCauseAnalysis would then find the rule "PLC communication interruption can cause multiple system data unreliable and trigger chain failure" in the knowledge base based on this pattern, and propose it as the main possibility.

[0113] The dynamic UI generator 9 receives highly structured analysis results from the MCP orchestration engine 4. These results are not just raw data, but contain comprehensive information of time correlations and cause inferences. The dynamic UI generator renders them into a multi-dimensional analysis dashboard, which can include:

[0114] An event timeline: clearly shows the sequence and time interval of all key faults, alarms, and operations in the past 24 hours.

[0115] A fault clustering graph: visually displays which events tend to occur together.

[0116] A root cause analysis report: in text form, clearly list the most likely root cause (or several alternative causes), and provide evidence of the event chain supporting this inference (for example, "main possibility: 'PLC network interruption' failure at 14:35, followed by 'crane position sensor signal loss' and 'hoist motor controller communication timeout' occurring consecutively within 5 minutes, eventually leading to 'work cycle interruption' at 14:42. It is recommended to prioritize checking the PLC network switch and main controller status. "). This dynamically generated interface is presented to the user through the user interaction terminal 1, enabling them to quickly grasp the overall fault situation in the past 24 hours and obtain reliable, data-driven diagnostic directions. The dynamic UI generator 9 receives the correlation analysis results and fault prompt texts (which can come from a predefined knowledge base, and the LLM can participate in sorting but cannot generate out of thin air), and generates an interface containing a list of related events, statistical charts, and comprehensive judgment prompts.

[0117] In this way, the SCADA system based on a large language model provided by the embodiment safely and reliably improves the data analysis and user interaction experience of the SCADA system by using the LLM without issuing control instructions and relying on real-time. Compared with the traditional SCADA system, the time for generating a new custom monitoring chart or a complex analysis report can be reduced by more than 50%, improving efficiency. Through the constraint mechanism of MCP and predefined functions, the influence of the LLM model "hallucination" problem on SCADA data display and analysis results is solved, ensuring the data reliability of the human-computer interaction interface. Only text is used for human-computer interaction, and the training time of the operator to master advanced data analysis functions can be shortened by 50%.

[0118] The SCADA system based on a large language model in the embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory executing one or more software or fixed programs, and / or other devices that can provide the above functions.

[0119] The embodiment of the application also provides a computer device having the SCADA system based on a large language model as shown in Figure 1 and Figure 2 .

[0120] Please refer to Figure 4 , Figure 4 is a structural schematic diagram of a computer device provided by an optional embodiment of the application, as Figure 4As shown, the computer device includes one or more processors 10, memory 20, and interfaces 30 for external devices such as a keyboard and a mouse and a disk drive. One or more of the interfaces 30 enable a user to interact with the computer device. In some embodiments, the interface 30 also includes an input device, such as a microphone, or output device, such as a speaker. Figure 4 The processor 10 is used in the description as an example.

[0121] The processor 10 can be a central processing unit, a network processor, or a combination thereof. The processor 10 can further include a hardware chip. The hardware chip can be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device can be a complex programmable logic device, a field programmable logic device, a general array logic, or any combination thereof.

[0122] The memory 20 stores instructions that can be executed by the at least one processor 10 to cause the at least one processor 10 to perform the methods described in the above embodiments.

[0123] The memory 20 can include a program region and a data region. The program region can store an operating system, application programs required by at least one function, and the like. The data region can store data created according to the use of the computer device, and the like. In addition, the memory 20 can include a high-speed random access memory, and can further include a non-transitory memory such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid state storage device. In some alternative embodiments, the memory 20 can optionally include a memory disposed remotely from the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0124] The memory 20 can include a volatile memory, such as a random access memory, and can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid state disk. The memory 20 can further include a combination of the above-mentioned types of memories.

[0125] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 can be connected by a bus or other means, Figure 4 The bus connection is taken as an example.

[0126] The input device 30 can receive inputted digital or character information, and generate key signal input related to user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a trackpad, a touchpad, a pointing stick, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (e.g., an LED), a tactile feedback device (e.g., a vibration motor), etc. The display device includes but is not limited to a liquid crystal display, a light-emitting diode, a display, and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0127] The embodiments of the present application also provide a computer readable storage medium, and the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or stored in a remote storage medium or a non-transitory machine readable storage medium downloaded through a network and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk or a solid state disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that the computer, processor, microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code, which, when accessed and executed by the computer, processor or hardware, implements the method shown in the above embodiments.

[0128] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, the method and / or technical solutions according to the present application can be called or provided. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source file, executable file, installation package file, etc., and accordingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer executes the corresponding compiled program after compiling the instructions, or the computer reads and executes the instructions, or the computer executes the corresponding installed program after reading and installing the instructions. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0129] While embodiments of the application have been described in connection with the preferred embodiments of the various figures, those of ordinary skill in the art will appreciate that various modifications and variations of the preferred embodiments can be employed without departing from the spirit and scope of the application.

Claims

1. A large language model-based SCADA system, characterized by, The system comprises: a user interaction terminal configured to receive a natural language instruction input by a user and display a dynamic user interface; a natural language processing module configured to perform preliminary analysis on the natural language instruction input by the user to obtain a user intent and key information; a large language model configured to perform deep analysis on the user intent and the key information to generate a task description or a task execution plan; an MCP orchestration engine configured to map the task description or the task execution plan to a predefined function to obtain executable structured data; a dynamic UI generator configured to generate a dynamic user interface based on the executable structured data and preset UI instructions.

2. The system of claim 1, wherein, The natural language processing module is further configured to perform word segmentation processing, text filtering, and text normalization data preprocessing on the input natural language instruction.

3. The system of claim 1, wherein, The system further comprises a SCADA predefined function library and a SCADA_MCP service, wherein the SCADA predefined function library is configured to interact with the MCP orchestration engine through the SCADA_MCP service; the SCADA predefined function library is configured to store a plurality of standardized and verified SCADA operation functions, wherein the SCADA operation functions include data query operation functions, log retrieval operation functions, statistical calculation operation functions, and fault code analysis operation functions; the SCADA_MCP service is configured to respond to a function call request of the MCP orchestration engine.

4. The system of claim 1, wherein, The system further comprises a support MCP protocol database and a database MCP service, wherein the support MCP protocol database is configured to interact with the MCP orchestration engine through the database MCP service; the support MCP protocol database is configured to store and manage historical data, logs, and configuration information of a SCADA system; the database MCP service is configured to respond to a data query request of the MCP orchestration engine.

5. The system of claim 3, wherein, The predefined function is at least one function call sequence in the SCADA predefined function library; the MCP orchestration engine is further configured to map the task description or the task execution plan to at least one function call sequence in the SCADA predefined function library based on a preset rule set and a preset knowledge base to obtain executable structured data, and initiate a function call request to the SCADA_MCP service through an MCP protocol.

6. The system of claim 1, wherein, The preset UI instructions include an interface type instruction in the user intent; the dynamic UI generator is further configured to generate a user interface based on the executable structured data and the interface type instruction in the user intent by calling a preset icon library or a preset UI component, and present the user interface on the user interaction terminal. 7.A method for implementing a SCADA system based on a large language model, characterized in that, The method comprises: receiving a natural language instruction input by a user; performing preliminary analysis on the natural language instruction input by the user to obtain a user intent and key information; performing deep analysis on the user intent and the key information to generate a task description or a task execution plan; mapping the task description or the task execution plan to a predefined function to obtain executable structured data; Generating a dynamic user interface based on executable structured data and interface type instructions in a user intent, and presenting the dynamic user interface.

8. A computer device, comprising: Comprising: A memory and a processor, which are connected in communication with each other, the memory has computer instructions stored therein, and the processor executes the computer instructions to perform the method for implementing a SCADA system based on a large language model according to claim 7.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium has computer instructions stored thereon, and the computer instructions are used to make a computer execute the method for implementing a SCADA system based on a large language model according to claim 7.

10. A computer program product, characterised in that, The computer instructions are used to make a computer execute the method for implementing a SCADA system based on a large language model according to claim 7.

Images