Energy management multi-agent collaborative optimization method and system based on large model agent architecture

Abstract

The invention discloses an energy management multi-agent collaborative optimization method and system based on a large model agent architecture. The system comprises a multi-agent architecture module, a model processing module, a data acquisition and management module, a system integration and interaction module and a collaborative optimization control module. The model processing module comprises a large model and a plurality of small models and is used for realizing task processing and analysis; the data acquisition and management module acquires and processes multi-source data; the system integration and interaction module realizes multi-platform integration and visual interaction; and the collaborative optimization control module ensures collaborative work and system optimization of the intelligent agents. The system has the advantages of autonomy and decentralized control, isomerism and modularization, intelligent collaboration and the like, instrument faults can be accurately diagnosed, energy management is optimized, production is intelligently controlled, enterprise energy utilization efficiency, instrument operation stability and the overall operation management level are improved, enterprises are assisted to achieve green and low-carbon transformation, and the system is suitable for popularization and application. The method is suitable for energy management and energy equipment maintenance scenes of various industrial production enterprises.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of energy management and instrument fault diagnosis in the field of industrial production, focusing on optimizing energy utilization efficiency and instrument operation status monitoring through intelligent means, and is specifically applied to energy consumption optimization, instrument maintenance, and real-time control and adjustment of energy equipment. It specifically relates to a multi-agent collaborative optimization method and system for energy management based on a large model agent architecture. Background Art

[0002] Energy is a key supporting factor for an enterprise's production and manufacturing activities, and its cost accounts for a significant proportion of the total cost of the enterprise. For high-energy-consuming enterprises, this proportion is particularly prominent.

[0003] The current industrial manufacturing industry is facing many difficulties. In terms of energy equipment management, the energy equipment management system at industrial production sites generally relies too much on manual experience and rule-based diagnostic methods. This model has many disadvantages. First, maintenance experience is difficult to be effectively accumulated, reused, and integrated into the system. Knowledge loss occurs frequently, and duplicate construction is a serious problem. Secondly, due to the lack of effective prediction methods, it is impossible to know the failure points of energy meters in advance, resulting in the inability to maintain energy equipment in a timely manner. Unexpected downtime incidents occur from time to time, which not only seriously interferes with normal production order and reduces production efficiency, but also causes a lot of unnecessary waste of energy.

[0004] In the field of energy management, the existing energy management system also has obvious shortcomings. First, in terms of real-time performance and computing efficiency, the model often has computing delays when processing complex industrial scenarios, and cannot meet the strict requirements of industrial production for millisecond-level real-time response. This seriously restricts the overall performance of the system in industrial environments with extremely high timeliness requirements. Second, there is a large room for improvement in the accuracy of energy consumption prediction, and it faces the optimization problem between real-time performance and prediction cycle, making it difficult to ensure high accuracy while ensuring the timeliness of prediction. Third, the sensitivity and specificity of energy consumption anomaly detection need to be further enhanced, making it difficult for the system to quickly and accurately identify energy consumption anomalies and take effective countermeasures in a timely manner. In addition, the degree of cross-domain knowledge integration is low, which makes it difficult for the energy management system to comprehensively consider multiple aspects of information such as instrument operation status, production plan, environmental factors, etc., and it is difficult to formulate a comprehensive, accurate and efficient energy management strategy.

[0005] In summary, the existing industrial manufacturing management system has significant defects in energy equipment maintenance and energy management. An innovative and efficient solution is urgently needed to promote industrial production towards intelligence, efficiency and sustainability. Summary of the invention

[0006] In view of the shortcomings of the current related existing technologies, the present invention provides a multi-agent collaborative optimization method and system for energy management based on a large model agent architecture; used to realize intelligent energy management optimization and energy equipment control in industrial production and energy management scenarios. The system uses an architecture that combines large models with small models, adopts task decomposition and deep learning technology, and has specific functional capabilities such as instrument fault diagnosis, energy consumption analysis (including energy consumption prediction, high energy consumption identification, energy consumption abnormality diagnosis), and production energy implementation control. The system includes multiple agents, each of which is responsible for a different workshop or energy node, and provides accurate energy management suggestions and energy equipment operation optimization control through data collection, analysis and feedback.

[0007] To achieve the above object, the present invention provides a first aspect of a multi-agent collaborative optimization method for energy management based on a large model agent architecture, characterized in that it includes the following steps:

[0008] S1 Task input and preliminary processing steps: Receive tasks input by users, perform prompt queries on the input tasks, convert them into a form suitable for large model processing, and the large model plans and decomposes the tasks, determines the subtasks to be executed, and determines the data types to be collected based on the subtasks. The instrument operation status data, energy consumption data, and production environment data are collected from the data source system including sensors, smart meters, and the enterprise's internal production database. The collected data is cleaned to remove outliers and noise, and preprocessed including normalization and feature extraction.

[0009] S2 Data collection and model analysis steps: Call the corresponding models for different subtasks. For the instrument fault diagnosis subtask, call the fault diagnosis small model group including the DBSCAN model and the AutoEncoder model. The DBSCAN model locates the fault location by clustering the instrument operation data, and the AutoEncoder model determines the fault by learning the normal operation mode of the instrument to determine the data deviation; for the energy consumption analysis subtask, call the energy analysis small model group including the ARIMA model, and select the appropriate method for energy consumption prediction, high energy consumption identification and energy consumption anomaly detection according to the characteristics of energy consumption data. If the data has a trend, the nonlinear trend prediction method is used. If there is seasonality, the seasonal autoregression model is used. If there is no trend and seasonality, the moving average method is used, and then the model analysis results are returned to the large model;

[0010] S3 result integration and control execution steps: The large model integrates the results of the small model to perform root cause analysis, explore the root causes of the problem from the perspective of the overall system, and generate optimization strategy recommendations including instrument fault handling recommendations and energy allocation strategies based on the root cause analysis. Controller instructions are generated based on the optimization strategy recommendations to achieve dynamic control of energy equipment and energy management systems. The system can dynamically adjust the control strategy based on equipment aging, production plan changes, and energy supply fluctuations.

[0011] Furthermore, in step S1, the user input task is in the form of natural language, specifically the energy management and instrument status related requirements for a workshop or production link. The large model plans and decomposes the task based on the knowledge and capabilities acquired through incremental pre-training and instruction fine-tuning, and fully considers the logical relationship between sub-tasks during decomposition. The sensors in the data source system include temperature sensors and pressure sensors. Smart instruments are used to measure energy consumption data. The company's internal production database stores instrument operation status data and production plan data. Data cleaning uses statistical analysis methods to remove abnormal data. Normalization processing uses specific algorithms to make data comparable. Feature extraction uses principal component analysis methods.

[0012] Furthermore, in step S2, the DBSCAN model determines clustering parameters according to data density distribution during cluster analysis, the AutoEncoder model learns the normal operation mode by constructing a multi-layer neural network, and the ARIMA model determines model parameters for prediction based on the autocorrelation and seasonal characteristics of energy consumption data, and makes a comprehensive judgment based on historical energy consumption and production plan data during energy consumption analysis.

[0013] Furthermore, in step S3, during the root cause analysis process, the large model analyzes the causal relationship between the fault and abnormal energy consumption in combination with the equipment operation principle, production plan requirements and historical data experience, and comprehensively considers the enterprise's production goals, cost constraints and energy supply conditions when generating optimization strategy recommendations. The controller instructions are transmitted to the PLC through a specific communication protocol. The PLC controls the start and stop, speed and power parameters of the energy equipment according to the instructions. The system evaluates the adjustment effect through real-time monitoring data and feedback mechanism.

[0014] Another aspect of the present invention provides an energy management multi-agent collaborative optimization system based on a large model agent architecture, the system is used to implement the above method, including:

[0015] Multi-agent architecture module: It consists of multiple agents, each of which is responsible for a specific energy device or energy node. It is built based on a large model agent architecture and has independent perception, decision-making and execution capabilities. It collaborates with other agents through network communication to achieve information sharing, including sharing monitoring data, diagnostic results and optimization strategies;

[0016] Model processing module: includes large model and small model. The large model is based on Qwen2.5, and is incrementally pre-trained to incorporate professional knowledge of energy management and energy equipment maintenance. It uses instruction fine-tuning technology to understand and execute task instructions, and uses prompt engineering to achieve chain reasoning to decompose complex tasks. It also has tool calling functions to connect external tools such as PLC controllers and energy consumption analysis tools. The small model includes the DBSCAN model for low-dimensional fault diagnosis, the AutoEncoder model for high-dimensional data fault detection, and the moving average model, seasonal autoregression model, and nonlinear trend prediction model for time series data analysis in energy management, which assist the large model in fine-grained analysis;

[0017] Data collection and management module: collects instrument operation status data, energy consumption data and production environment data through sensors and intelligent instruments, cleans, preprocesses and stores the collected data. Data cleaning uses advanced algorithms to remove outliers and noise. Preprocessing includes data normalization and feature extraction. The data storage management system uses efficient structures and algorithms to classify, store and manage historical data, and processes and analyzes data with the help of big data technology and cloud computing platforms.

[0018] System integration and interaction module: realize seamless integration with one or more production data platforms such as MES, APS, MRO, SCADA and IoT, connect with other systems through standardized interfaces, develop user interfaces with visualization functions, display instrument operation status, energy consumption data, analysis results and optimization strategy information, and facilitate operators to set parameters, issue tasks and monitor the system. At the same time, it has user authority management functions to ensure data security and operational specifications.

[0019] Collaborative optimization control module: used for information sharing, collaborative cooperation and dynamic task allocation among intelligent agents, establishing a negotiation and coordination mechanism to resolve resource conflicts and task competition problems. The dynamic task allocation mechanism allocates tasks based on factors such as task priority, urgency, intelligent agent load and processing capacity, and can adjust collaborative strategies and optimization goals based on equipment aging, production plan changes and energy supply fluctuations. A performance evaluation index system is used to monitor and evaluate key indicators such as energy utilization efficiency, instrument operation stability, and task processing timeliness in real time, and collaborative strategies and optimization parameters are automatically adjusted based on the results.

[0020] Furthermore, the large model constructs a decision tree according to logical steps during chain reasoning, adjusts the reasoning path and decision-making strategy based on real-time data and feedback, and the tool calling function connects to external tools and exchanges data through secure interface technology. Incremental pre-training uses a continuous learning algorithm to update the knowledge system to adapt to environmental changes.

[0021] Furthermore, the DBSCAN model adjusts clustering parameters according to data distribution to improve positioning accuracy during low-dimensional fault diagnosis. The AutoEncoder model enhances abnormal pattern recognition capabilities through reasonable encoding and decoding structures. The moving average model dynamically adjusts the time window size according to the data. The seasonal autoregressive model optimizes parameters based on the seasonal cycle. The nonlinear trend prediction model uses a fitting algorithm to adapt to trend changes.

[0022] Furthermore, the sensors and intelligent instruments in the data acquisition and management module have high precision, high reliability and real-time transmission capabilities, the data cleaning algorithm intelligently identifies and processes abnormal data, the data storage management system supports fast retrieval, backup and recovery, and the big data technology and cloud computing platform use distributed computing architecture and parallel processing algorithms to improve efficiency, and can intelligently cache and optimize storage strategies.

[0023] Furthermore, the standardized interfaces in the system integration and interaction modules follow industry standard specifications to ensure compatibility and interoperability, the user interface visualization uses intuitive graphics and dynamic displays for easy operation, and the user authority management function is complete to ensure data security and operational specifications.

[0024] Furthermore, the negotiation and coordination mechanism in the collaborative optimization control module adopts a distributed consensus algorithm to ensure decision consistency, the dynamic task allocation mechanism uses an intelligent algorithm to evaluate the task and agent status, the resource conflict and task competition resolution strategy is based on the principles of priority and fairness, and the performance evaluation index system comprehensively monitors key indicators and automatically adjusts strategies and parameters.

[0025] The present invention adopts the above technical solution, which has at least the following beneficial effects: According to the research and comparative analysis of similar system products currently on the market, the present invention has at least the following three advantages:

[0026] 1. Autonomy and decentralized control

[0027] (1) Existing technologies: There may be certain limitations in terms of autonomy and decentralized control. Although existing large models have the characteristics of cloud-edge-end collaboration, autonomous control, and data and model security, which ensure their stability and security in practical applications, they are limited in their ability to make autonomous decisions and decentralized control. In some complex energy management scenarios, more central coordination and control are required, which limits their adaptability and flexibility to a certain extent.

[0028] (2) The technology of the present invention: shows significant technical advantages in terms of autonomy and decentralized control. The model is based on an advanced multi-agent system architecture, which gives each sub-model a high degree of independence and autonomous decision-making capabilities. The sub-model can independently perform complex reasoning and planning, and select the most appropriate strategy to deal with various sub-problems. In the absence of a central control layer, the sub-model can still make effective decisions and actions, which significantly improves the robustness and flexibility of the entire system, enabling it to maintain stable operation in the face of dynamically changing environments.

[0029] 2. Heterogeneity and modularity

[0030] (1) Existing large model technology: Although it also has a modular design, it has poor flexibility in heterogeneity, which affects its adaptability in diversified energy management tasks.

[0031] (2) The technology of the present invention adopts the principle of heterogeneity, so that each intelligent agent constituting the system can have a specialized structure and function according to the nature of the task it undertakes. This design not only gives the system extremely high adaptability and flexibility, but also enables each intelligent agent to provide customized solutions for specific problems. In addition, the modular design of the model ensures the independence of each component, facilitates independent development, testing and reuse, greatly improves the efficiency of the development process, and enhances the maintainability and scalability of the system.

[0032] 3. Intelligent collaboration

[0033] (1) Existing large model technology: Although it has made some achievements in modular design, it has great limitations in dynamic asynchronous computing and self-organizing collaboration, which restricts its efficiency in processing energy management tasks that require rapid response and flexible resource scheduling.

[0034] (2) The technology of the present invention: A dynamic asynchronous computing mode is adopted in the computing architecture, which significantly improves the parallel processing capability and response speed of the system. The model can dynamically adjust resource allocation and execution order according to the urgency of the task and the computing requirements, and realizes asynchronous and efficient computing. In addition, the large energy consumption model also has excellent self-organizing ability, and forms an effective cooperation mechanism through local interaction, so that large and small models can jointly solve complex problems through consultation, coordination and cooperation, showing a strong collaborative processing capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 It is a flow chart of the multi-agent collaborative optimization method for energy management of the present invention;

[0037] Figure 2 It is a group decision flow chart of the time series prediction model of the present invention. DETAILED DESCRIPTION

[0038] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0039] Embodiment 1

[0040] See also Figure 1 and Figure 2 The first aspect of this embodiment provides a multi-agent collaborative optimization method for energy management based on a large model agent architecture, comprising the following steps:

[0041] S1 Task input and preliminary processing steps: Receive tasks input by users, perform prompt queries on the input tasks, convert them into a form suitable for large model processing, and the large model plans and decomposes the tasks, determines the subtasks to be executed, and determines the data types to be collected based on the subtasks. The instrument operation status data, energy consumption data, and production environment data are collected from the data source system including sensors, smart meters, and the enterprise's internal production database. The collected data is cleaned to remove outliers and noise, and preprocessed including normalization and feature extraction.

[0042] S2 Data collection and model analysis steps: Call the corresponding models for different subtasks. For the instrument fault diagnosis subtask, call the fault diagnosis small model group including the DBSCAN model and the AutoEncoder model. The DBSCAN model locates the fault location by clustering the instrument operation data, and the AutoEncoder model determines the fault by learning the normal operation mode of the instrument to determine the data deviation; for the energy consumption analysis subtask, call the energy analysis small model group including the ARIMA model, and select the appropriate method for energy consumption prediction, high energy consumption identification and energy consumption anomaly detection according to the characteristics of energy consumption data. If the data has a trend, the nonlinear trend prediction method is used. If there is seasonality, the seasonal autoregression model is used. If there is no trend and seasonality, the moving average method is used, and then the model analysis results are returned to the large model;

[0043] S3 result integration and control execution steps: The large model integrates the results of the small model to perform root cause analysis, explore the root causes of the problem from the perspective of the overall system, and generate optimization strategy recommendations including instrument fault handling recommendations and energy allocation strategies based on the root cause analysis. Controller instructions are generated based on the optimization strategy recommendations to achieve dynamic control of energy equipment and energy management systems. The system can dynamically adjust the control strategy based on equipment aging, production plan changes, and energy supply fluctuations.

[0044] As a preferred implementation, in step S1 of this embodiment, the user input task is in natural language form, specifically the energy management and instrument status related requirements for a workshop or production link. The large model plans and decomposes tasks based on the knowledge and capabilities acquired through incremental pre-training and instruction fine-tuning. The logical relationship between subtasks is fully considered during the decomposition. The sensors in the data source system include temperature sensors and pressure sensors. Smart instruments are used to measure energy consumption data. The company's internal production database stores instrument operating status and production plan data. Data cleaning uses statistical analysis methods to remove abnormal data. Normalization processing uses specific algorithms to make data comparable. Feature extraction uses principal component analysis methods.

[0045] As a preferred implementation, in step S2 of this embodiment, the DBSCAN model determines the clustering parameters according to the data density distribution during cluster analysis, the AutoEncoder model learns the normal operating mode by constructing a multi-layer neural network, and the ARIMA model determines the model parameters for prediction based on the autocorrelation and seasonal characteristics of the energy consumption data, and makes a comprehensive judgment based on the historical energy consumption and production plan data during energy consumption analysis.

[0046] As a preferred implementation, in step S3 of this embodiment, during the root cause analysis process, the large model combines the equipment operation principle, production plan requirements and historical data experience to analyze the causal relationship between the fault and abnormal energy consumption, and generates optimization strategy recommendations while comprehensively considering the company's production goals, cost constraints and energy supply conditions. The controller instructions are transmitted to the PLC through a specific communication protocol. The PLC controls the start and stop, speed and power parameters of the energy equipment according to the instructions. The system evaluates the adjustment effect through real-time monitoring data and feedback mechanism.

[0047] A second aspect of this embodiment provides an energy management multi-agent collaborative optimization system based on a large model agent architecture, the system is used to implement the above method, including:

[0048] Multi-agent architecture module: It consists of multiple agents, each of which is responsible for a specific energy device or energy node. It is built based on a large model agent architecture and has independent perception, decision-making and execution capabilities. It collaborates with other agents through network communication to achieve information sharing, including sharing monitoring data, diagnostic results and optimization strategies;

[0049] Model processing module: includes large model and small model. The large model is based on Qwen2.5, and is incrementally pre-trained to incorporate professional knowledge of energy management and energy equipment maintenance. It uses instruction fine-tuning technology to understand and execute task instructions, and uses prompt engineering to achieve chain reasoning to decompose complex tasks. It also has tool calling functions to connect external tools such as PLC controllers and energy consumption analysis tools. The small model includes the DBSCAN model for low-dimensional fault diagnosis, the AutoEncoder model for high-dimensional data fault detection, and the moving average model, seasonal autoregression model, and nonlinear trend prediction model for time series data analysis in energy management, which assist the large model in fine-grained analysis;

[0050] Data collection and management module: collects instrument operation status data, energy consumption data and production environment data through sensors and intelligent instruments, cleans, preprocesses and stores the collected data. Data cleaning uses advanced algorithms to remove outliers and noise. Preprocessing includes data normalization and feature extraction. The data storage management system uses efficient structures and algorithms to classify, store and manage historical data, and processes and analyzes data with the help of big data technology and cloud computing platforms.

[0051] System integration and interaction module: realize seamless integration with one or more production data platforms such as MES, APS, MRO, SCADA and IoT, connect with other systems through standardized interfaces, develop user interfaces with visualization functions, display instrument operation status, energy consumption data, analysis results and optimization strategy information, and facilitate operators to set parameters, issue tasks and monitor the system. At the same time, it has user authority management functions to ensure data security and operational specifications.

[0052] Collaborative optimization control module: used for information sharing, collaborative cooperation and dynamic task allocation among intelligent agents, establishing a negotiation and coordination mechanism to resolve resource conflicts and task competition problems. The dynamic task allocation mechanism allocates tasks based on factors such as task priority, urgency, intelligent agent load and processing capacity, and can adjust collaborative strategies and optimization goals based on equipment aging, production plan changes and energy supply fluctuations. A performance evaluation index system is used to monitor and evaluate key indicators such as energy utilization efficiency, instrument operation stability, and task processing timeliness in real time, and collaborative strategies and optimization parameters are automatically adjusted based on the results.

[0053] As a preferred implementation, the large model described in this embodiment constructs a decision tree according to logical steps during chain reasoning, adjusts the reasoning path and decision-making strategy based on real-time data and feedback, and the tool call function connects external tools and exchanges data through secure interface technology. Incremental pre-training uses a continuous learning algorithm to update the knowledge system to adapt to environmental changes.

[0054] As a preferred implementation, the DBSCAN model described in this embodiment adjusts clustering parameters according to data distribution to improve positioning accuracy during low-dimensional fault diagnosis. The AutoEncoder model enhances the abnormal pattern recognition capability through reasonable encoding and decoding structure. The moving average model dynamically adjusts the time window size according to the data. The seasonal autoregressive model optimizes parameters based on the seasonal cycle. The nonlinear trend prediction model uses a fitting algorithm to adapt to trend changes.

[0055] As a preferred implementation, the sensors and intelligent meters in the data acquisition and management module described in this embodiment have high precision, high reliability and real-time transmission capabilities, the data cleaning algorithm intelligently identifies and processes abnormal data, the data storage management system supports fast retrieval, backup and recovery, and the big data technology and cloud computing platform use distributed computing architecture and parallel processing algorithms to improve efficiency, and can intelligently cache and optimize storage strategies.

[0056] As a preferred implementation, the standardized interfaces in the system integration and interaction modules described in this embodiment follow industry standard specifications to ensure compatibility and interoperability, the user interface visualization uses intuitive charts and dynamic displays for easy operation, and the user authority management function is complete to ensure data security and operational specifications.

[0057] As a preferred implementation, the negotiation and coordination mechanism in the collaborative optimization control module described in this embodiment adopts a distributed consensus algorithm to ensure decision consistency, the dynamic task allocation mechanism uses an intelligent algorithm to evaluate tasks and agent status, the resource conflict and task competition resolution strategy is based on the principles of priority and fairness, and the performance evaluation index system comprehensively monitors key indicators and automatically adjusts strategies and parameters.

[0058] Embodiment 2

[0059] like Figure 1 The system flow chart analysis shown includes:

[0060] 1. Overall process architecture and logical relationship;

[0061] This flowchart clearly shows the overall operation process of the energy management multi-agent system based on the large model agent architecture. It starts with the natural language input task, goes through a series of orderly steps, and finally achieves dynamic adjustment. The steps are closely linked and the logic is clear, forming a complete closed-loop system.

[0062] The system is task-driven and uses the collaborative work of multiple modules to achieve energy management and energy equipment control optimization. Each module performs its own duties and collaborates with each other to complete complex energy management tasks.

[0063] 2. Functions and data flows of each process node;

[0064] Natural language input task (prompt query): As the starting point of the system, the user inputs the task in natural language, which reflects the user-friendly interaction of the system and facilitates operators to issue various energy management and energy equipment-related instructions, such as instrument fault diagnosis requests, energy consumption optimization requirements, etc. The input tasks will enter the subsequent large model processing link.

[0065] Large model processing (LLM large model): The large model is the core decision-making unit of the system. After receiving the user input task, it first deeply analyzes the user's intention through deep learning and context understanding technology. This step is the basis for subsequent precise operations. Then, the large model breaks down complex tasks into specific subtasks, such as breaking down instrument fault diagnosis tasks into subtasks such as data collection, fault location, and root cause analysis, in preparation for the subsequent targeted calling of different models and data sources.

[0066] Data collection and processing: According to the subtask requirements disassembled from the large model, relevant data is collected from the data source system. The data source covers a wide range, including sensors, smart meters, and internal production databases of enterprises, etc., to ensure the acquisition of comprehensive and accurate instrument operation status data, energy consumption data, production environment data and other information. The collected data needs to be cleaned and preprocessed. The cleaning process removes outliers and noise interference. The preprocessing includes normalization processing to make the data comparable, and feature extraction for subsequent model analysis. The processed data will be provided to the corresponding small model for analysis.

[0067] Small model analysis: For different types of tasks, the system calls different small models for professional analysis. In terms of fault diagnosis, models such as DBSCAN (density-based clustering algorithm) and AutoEncoder (autoencoder) are used. The DBSCAN model can accurately locate the fault location of the instrument through cluster analysis of the instrument operation data, and the AutoEncoder model learns the normal operation mode of the instrument and compares the real-time data deviation to judge the fault situation; in terms of energy consumption analysis, models such as ARIMA (time series prediction model) are used for energy consumption prediction, high energy consumption identification and energy consumption anomaly detection. After the small model analysis is completed, the results are returned to the large model.

[0068] Large model integration and decision-making (root cause analysis, optimization strategy recommendations): After receiving the analysis results of the small model, the large model conducts a comprehensive root cause analysis to deeply explore the root cause of the problem from the perspective of the entire system, not just limited to the surface fault phenomenon. Based on the root cause analysis results, the large model further generates comprehensive optimization strategy recommendations, covering instrument fault handling recommendations, energy equipment operating parameter adjustment strategies, production process optimization solutions, and energy allocation strategies, etc., providing detailed guidance for the optimization control of the system.

[0069] Control execution (generating controller instructions, dynamic adjustment): Finally, the controller instructions are generated according to the optimization strategy recommendations, and the instructions are transmitted to the energy equipment and energy management system through the connection with the PLC (programmable logic controller), so as to realize the real-time and dynamic adjustment of the operation of energy equipment and energy use. In addition, the system can continuously adjust the control strategy according to the actual situation such as equipment aging, production plan changes and energy supply fluctuations to ensure that the system always maintains the optimal operating state.

[0070] like Figure 2 The analysis of the group decision flow chart of the time series forecasting model includes:

[0071] 1. Time series data processing logic;

[0072] This flowchart focuses on the processing of time series data. According to the characteristics of time series data such as energy consumption data in energy management, a scientific and reasonable prediction model decision process is designed. The core logic is to classify and process the data according to its trend and periodic characteristics to select the most suitable prediction method and improve the accuracy and effectiveness of the prediction.

[0073] It reflects the system's ability to fine-tune management and targeted processing of different types of time series data, ensuring accurate energy consumption prediction and anomaly detection under various complex data patterns, and providing a reliable decision-making basis for energy management.

[0074] 2. Processing procedures and model selection basis for different situations;

[0075] Trend judgment and model selection: The system first judges the trend of time series data. If the data shows an obvious trend, such as energy consumption gradually increasing or decreasing over time, the nonlinear trend prediction method is selected for prediction. This is because the nonlinear trend prediction method can better fit the nonlinear change trend of the data, capture the complex law of energy consumption changing over time, and thus provide more accurate prediction results.

[0076] Periodicity judgment and model selection (in the absence of trend): When the data has no trend, it enters the periodicity judgment stage. If the data has obvious periodicity, such as energy consumption data that fluctuates according to daily, weekly, monthly and other periodic rules, the system uses the seasonal autoregression model for prediction. The seasonal autoregression model can fully consider the seasonal periodic characteristics of the data, use the periodic rules in historical data to predict future data, and effectively deal with the impact of periodic changes.

[0077] Processing of non-trend and non-periodic situations: If the data has neither trend nor periodicity, the moving average method is used for prediction. The moving average method smoothes data fluctuations by calculating the average value of data within a certain time window. It has a good prediction effect on stable time series data and is suitable for situations where energy consumption data is relatively stable and has no obvious regular changes. Through this case-by-case processing method, the system can select the most appropriate prediction model based on the actual characteristics of the data, maximize the prediction accuracy, and provide strong support for energy management decisions.

[0078] Embodiment 3

[0079] See also Figure 1 and Figure 2 This embodiment provides an implementation method of a multi-agent collaborative optimization method for energy management based on a large model agent architecture, comprising the following steps:

[0080] 1. System initialization and preparation

[0081] Build an energy management multi-agent system based on a large model agent architecture, and ensure that the system includes major components such as scenario function module, large language model module, production data docking and control system integration module, and industry standard docking module. Establish stable communication connections between modules to achieve smooth transmission of data and instructions.

[0082] Initialize the large model (based on ChatGLM3-6B or Qwen2.5), including completing incremental pre-training, integrating professional knowledge in specific fields such as energy consumption management and energy equipment maintenance into the model parameters, so that it has the basic ability to understand and handle industrial scenario tasks. At the same time, for different task scenarios, use instruction fine-tuning technology to optimize the large model to ensure that it can accurately understand and execute complex instructions such as "fault analysis" and "energy consumption optimization". Set prompt engineering related parameters so that the large model can perform chain reasoning and achieve effective decomposition and step-by-step resolution of complex tasks. Configure the tool call function of the large model, and establish a connection interface with external tools (such as PLC controllers, energy consumption analysis tools, etc.) through AIAgent technology to ensure that the operation of external tools can be realized.

[0083] Deploy small models for specific tasks, such as the DBSCAN model for low-dimensional fault diagnosis, the AutoEncoder model for high-dimensional data fault detection, and the moving average model, seasonal autoregression model, and nonlinear trend prediction model for time series data analysis in energy management. Ensure that an effective data interaction channel is established between the small model and the large model so that it can receive instructions from the large model and return analysis results.

[0084] Connect data source systems, including various sensors (such as temperature sensors, pressure sensors, etc.), smart meters (for measuring energy consumption data, etc.), and internal production databases of enterprises to ensure that the system can collect information such as instrument operation status data, energy consumption data, and production environment data in real time. Calibrate and test data acquisition equipment to ensure data accuracy and reliability. Set up a data storage management system and use efficient data structures and algorithms to classify, store, and manage the collected historical data to facilitate subsequent data query, analysis, and mining.

[0085] 2. Task Input and Processing

[0086] The user inputs tasks (prompt queries) in natural language, such as "analyze the reasons for excessive energy consumption in a production workshop and propose optimization solutions." The input tasks enter the LLM model through the system interface.

[0087] After receiving the task, the LLM model uses deep learning and context understanding technology to conduct an in-depth analysis of the user's intention. Through the semantic understanding and grammatical parsing of natural language texts and matching with pre-trained knowledge, the core content of the user's needs can be accurately grasped. For example, key information such as the instrument range involved in the task, energy consumption-related issues, and expected optimization goals can be identified.

[0088] According to the analysis results, the big model breaks down complex tasks into specific subtasks. For example, in response to the above-mentioned problem of excessive energy consumption, the big model breaks down the data collection subtask (including collecting the real-time energy consumption data, operating parameters, ambient temperature and other related data of the production workshop), the fault diagnosis subtask (determining whether there is a fault in the instrument that causes abnormal energy consumption), and the energy consumption analysis subtask (analyzing the trend and periodic characteristics of energy consumption data, and identifying high energy consumption periods and links). At the same time, the data type and data source required for each subtask are determined to prepare for subsequent data collection and model calls.

[0089] 3. Data Collection and Preprocessing

[0090] According to the subtask requirements disassembled from the large model, the system collects corresponding data from the data source system. The sensor obtains the instrument operation status data (such as temperature, pressure, flow, etc.) in real time, and the smart instrument accurately measures the energy consumption data. At the same time, the historical operation data and production plan data related to the workshop are extracted from the internal production database of the enterprise.

[0091] The collected data enters the data processing stage. First, data cleaning is performed, and advanced algorithms (such as outlier detection algorithms based on statistical analysis) are used to identify and remove outliers and noise interference in the data to ensure the accuracy and reliability of the data. For example, data points that are obviously deviated from the normal energy consumption range are corrected or deleted.

[0092] Preprocess the cleaned data. Use normalization algorithms (such as minimum-maximum normalization or mean-standard deviation normalization) to convert data of different types and ranges into a unified numerical range to make the data comparable and facilitate subsequent model analysis. At the same time, use feature extraction techniques (such as principal component analysis PCA and other methods) to extract key features from the original data, reduce data dimensions, improve model calculation efficiency, and retain the main information of the data. The preprocessed data will be provided to the corresponding small models for analysis according to the subtask requirements.

[0093] 4. Model Analysis and Diagnosis

[0094] Fault diagnosis small model group

[0095] For the instrument fault diagnosis subtask, the system calls the DBSCAN model and AutoEncoder model in the fault diagnosis small model group.

[0096] The DBSCAN model receives instrument operation data, determines clustering parameters based on the data density distribution, and performs cluster analysis on the data. By identifying dense and sparse areas of data points, the instrument operation status is divided into normal status and possible fault status. For example, if the operation data points of an instrument form a dense cluster under normal operation, but deviate from the cluster when an abnormality occurs, the DBSCAN model can accurately detect this change and locate the instrument fault. For example, if the temperature data clustering of a motor is abnormal, it may indicate that there is a fault in the motor cooling system.

[0097] The AutoEncoder model builds a multi-layer neural network to learn the data feature representation of the instrument under normal operating mode. In real-time monitoring, the collected data is input into the model to calculate the deviation between the output and input data. If the deviation exceeds the set threshold, the instrument is judged to be faulty. For example, for the attribute data of a certain instrument, after the AutoEncoder model learns the normal operating mode, when the instrument has faults such as inaccurate data collection and network offline, the deviation of the attribute data will trigger a fault alarm and provide assessment information on the degree of fault.

[0098] Using energy to analyze small model groups

[0099] In the energy consumption analysis subtask, the system calls the ARIMA model and other related algorithm models in the energy consumption analysis small model group.

[0100] First, the trend and periodicity of energy consumption data are judged. If the data shows an obvious trend (such as energy consumption gradually increasing or decreasing over time), the nonlinear trend prediction method is used for prediction. This method fits the changing trend of energy consumption data by constructing a suitable nonlinear function model (such as polynomial function, exponential function, etc.), predicts future energy consumption trends based on historical data, and provides a basis for enterprises to plan energy supply in advance.

[0101] If the energy consumption data has no trend but has obvious periodicity (such as fluctuations in daily, weekly, monthly, etc.), the seasonal autoregressive model is applied. This model considers the seasonal periodic characteristics in historical data, establishes an autoregressive equation, and uses the data of past cycles to predict the energy consumption of the same cycle in the future, helping enterprises to arrange production plans reasonably and reduce energy consumption during peak hours.

[0102] If the energy consumption data has neither trend nor periodicity, the moving average method is used for prediction. By calculating the average value of energy consumption data within a certain time window, smoothing data fluctuations, and making short-term predictions for relatively stable energy consumption data, the system provides a reference for enterprises to adjust their energy use strategies in real time. At the same time, based on these analysis models, the system can also identify high energy consumption and detect energy anomalies, timely discover unreasonable energy consumption, and provide support for energy conservation and consumption reduction.

[0103] V. Results Integration and Decision-making

[0104] After the fault diagnosis and energy analysis small model completes the analysis, it returns the results to the LLM large model. The large model receives the analysis results from different small models and conducts a comprehensive root cause analysis. Combining the equipment operation principle, production plan requirements and historical data experience, the root cause of the problem is deeply explored from the perspective of the overall system. For example, if the instrument fault diagnosis small model detects that the energy consumption of a certain instrument is too high and there are abnormal operating parameters, the large model will further analyze the causal relationship between the abnormal parameters and the excessive energy consumption, considering a variety of possible factors such as equipment aging, environmental factors, improper operation, etc., and determine that the root cause of the excessive energy consumption is the wear of a key component of the instrument, resulting in a decrease in efficiency, rather than simply the influence of the external environment.

[0105] Based on the results of the root cause analysis, the large model generates comprehensive optimization strategy recommendations. The optimization strategy content covers instrument fault handling recommendations (such as specific plans for repairing or replacing worn parts), energy equipment operating parameter adjustment strategies (such as optimizing energy equipment operating speed, temperature setting and other parameters) and energy allocation strategies (such as reasonably allocating energy supply to different workshops, giving priority to ensuring energy demand in key links, etc.). These optimization strategies are designed to improve the operating efficiency of energy equipment, reduce energy consumption, and improve the overall production benefits, while ensuring that the system operates in a stable and reliable state.

[0106] 6. Control Execution and Dynamic Adjustment

[0107] According to the optimization strategy recommendations, the LLM large model generates corresponding controller instructions. The instructions are transmitted to the PLC (programmable logic controller) through a specific communication protocol. The PLC controls the operating parameters of the energy equipment and energy management system according to the received instructions. For example, the start and stop time, operating speed, power output, etc. of the energy equipment are adjusted to achieve precise control of energy use. At the same time, the system can monitor the changes in relevant data in real time and dynamically adjust the control strategy according to the actual situation such as equipment aging, production plan changes and energy supply fluctuations. For example, when the performance of energy equipment decreases due to aging, the system automatically reduces the operating load of the energy equipment, optimizes energy distribution, ensures that the energy equipment operates in a safe and reliable state, and avoids energy waste; if the production plan changes, the system adjusts the operating parameters of the energy equipment and the energy supply strategy in time according to the requirements of the new production plan to ensure the smooth progress of production; when the energy supply is tight, the energy demand of key production links is prioritized, and the efficient use of energy is achieved by reasonably adjusting the operation of non-critical equipment, ensuring that the system always maintains the optimal operating state and constantly adapts to various complex and changing industrial production environments.

[0108] During the entire system operation process, the system performance is continuously monitored, instrument operation data, energy consumption data and user feedback information are collected, and the model is evaluated and optimized regularly. Based on new data and actual operation conditions, the large model is incrementally trained to update its knowledge and strategies, improve the system's adaptability and processing capabilities for different tasks and scenarios, ensure that the system always maintains an efficient, intelligent and stable operation state, and achieve continuous optimization of energy management and energy equipment control.

[0109] Although the embodiments of the present invention have been shown and described above, it will be appreciated that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, substitute and vary the above embodiments within the scope of the present invention.

Claims

1. A multi-agent collaborative optimization method for energy management based on a large model agent architecture, characterized by: The following steps are involved: S1 Task input and preliminary processing steps: Receive tasks input by users, perform prompt queries on the input tasks, convert them into a form suitable for large model processing, and the large model plans and decomposes the tasks, determines the subtasks to be executed, and determines the data types to be collected based on the subtasks. The instrument operation status data, energy consumption data, and production environment data are collected from the data source system including sensors, smart meters, and the enterprise's internal production database. The collected data is cleaned to remove outliers and noise, and preprocessed including normalization and feature extraction. S2 Data collection and model analysis steps: Call the corresponding models for different subtasks. For the instrument fault diagnosis subtask, call the fault diagnosis small model group including the DBSCAN model and the AutoEncoder model. The DBSCAN model locates the fault location by clustering the instrument operation data, and the AutoEncoder model determines the fault by learning the normal operation mode of the instrument to determine the data deviation; for the energy consumption analysis subtask, call the energy analysis small model group including the ARIMA model, and select the appropriate method for energy consumption prediction, high energy consumption identification and energy consumption anomaly detection according to the characteristics of energy consumption data. If the data has a trend, the nonlinear trend prediction method is used. If there is seasonality, the seasonal autoregression model is used. If there is no trend and seasonality, the moving average method is used, and then the model analysis results are returned to the large model; S3 result integration and control execution steps: The large model integrates the results of the small model to perform root cause analysis, explore the root causes of the problem from the perspective of the overall system, and generate optimization strategy recommendations including instrument fault handling recommendations and energy allocation strategies based on the root cause analysis. Controller instructions are generated based on the optimization strategy recommendations to achieve dynamic control of energy equipment and energy management systems. The system can dynamically adjust the control strategy based on equipment aging, production plan changes, and energy supply fluctuations.

2. The energy management multi-agent collaborative optimization method based on a large model agent architecture according to claim 1 is characterized by: In step S1, the user inputs the task in natural language form, specifically the energy management and instrument status related requirements for a workshop or production link. The large model plans and decomposes the task based on the knowledge and capabilities acquired through incremental pre-training and instruction fine-tuning. The logical relationship between sub-tasks is fully considered during the decomposition. The sensors in the data source system include temperature sensors and pressure sensors. Smart instruments are used to measure energy consumption data. The company's internal production database stores instrument operation status data and production plan data. Data cleaning uses statistical analysis methods to remove abnormal data. Normalization processing uses specific algorithms to make data comparable. Feature extraction uses principal component analysis methods.

3. The energy management multi-agent collaborative optimization method based on a large model agent architecture according to claim 1 is characterized by: In step S2, the DBSCAN model determines clustering parameters according to data density distribution during cluster analysis, the AutoEncoder model learns the normal operation mode by building a multi-layer neural network, and the ARIMA model determines model parameters for prediction based on the autocorrelation and seasonal characteristics of energy consumption data. When analyzing energy consumption, a comprehensive judgment is made based on historical energy consumption and production plan data.

4. The energy management multi-agent collaborative optimization method based on a large model agent architecture according to claim 1 is characterized by: In step S3, during the root cause analysis process, the large model combines the equipment operation principle, production plan requirements and historical data experience to analyze the causal relationship between the fault and abnormal energy consumption, and generates optimization strategy recommendations by comprehensively considering the company's production goals, cost constraints and energy supply conditions. The controller instructions are transmitted to the PLC through a specific communication protocol. The PLC controls the start and stop, speed and power parameters of the energy equipment according to the instructions. The system evaluates the adjustment effect through real-time monitoring data and feedback mechanism.

5. An energy management multi-agent collaborative optimization system based on a large model agent architecture, characterized by: The system is used to implement the method described in any one of claims 1 to 3, including: Multi-agent architecture module: It consists of multiple agents, each of which is responsible for a specific energy device or energy node. It is built based on a large model agent architecture and has independent perception, decision-making and execution capabilities. It collaborates with other agents through network communication to achieve information sharing, including sharing monitoring data, diagnostic results and optimization strategies; Model processing module: It includes large models and small models. The large model is based on Qwen2.5, and is incrementally pre-trained to incorporate professional knowledge of energy management and instrument maintenance. It uses instruction fine-tuning technology to understand and execute task instructions, and uses prompt engineering to achieve chain reasoning to decompose complex tasks. It also has tool calling functions to connect external tools such as PLC controllers and energy consumption analysis tools. The small model includes the DBSCAN model for low-dimensional fault diagnosis, the AutoEncoder model for high-dimensional data fault detection, and the moving average model, seasonal autoregression model, and nonlinear trend prediction model for time series data analysis in energy management, which assist the large model in fine-grained analysis; Data collection and management module: collects instrument operation status data, energy consumption data and production environment data through sensors and intelligent instruments, cleans, preprocesses and stores the collected data. Data cleaning uses advanced algorithms to remove outliers and noise. Preprocessing includes data normalization and feature extraction. The data storage management system uses efficient structures and algorithms to classify, store and manage historical data, and processes and analyzes data with the help of big data technology and cloud computing platforms. System integration and interaction module: realize seamless integration with one or more production data platforms such as MES, APS, MRO, SCADA and IoT, connect with other systems through standardized interfaces, develop user interfaces with visualization functions, display instrument operation status, energy consumption data, analysis results and optimization strategy information, and facilitate operators to set parameters, issue tasks and monitor the system. At the same time, it has user authority management functions to ensure data security and operation specifications; Collaborative optimization control module: used for information sharing, collaborative cooperation and dynamic task allocation among intelligent agents, establishing a negotiation and coordination mechanism to resolve resource conflicts and task competition problems. The dynamic task allocation mechanism allocates tasks based on task priority, urgency, intelligent agent load and processing capacity factors, and can adjust collaborative strategies and optimization goals according to equipment aging, production plan changes and energy supply fluctuations. A performance evaluation index system is used to monitor and evaluate key indicators of energy utilization efficiency, instrument operation stability, and task processing timeliness in real time, and collaborative strategies and optimization parameters are automatically adjusted based on the results.

6. The energy management multi-agent collaborative optimization system based on a large model agent architecture according to claim 5 is characterized by: The large model constructs a decision tree according to logical steps during chain reasoning, adjusts the reasoning path and decision-making strategy based on real-time data and feedback, and the tool call function connects to external tools and exchanges data through secure interface technology. Incremental pre-training uses a continuous learning algorithm to update the knowledge system to adapt to environmental changes.

7. The energy management multi-agent collaborative optimization system based on a large model agent architecture according to claim 5 is characterized by: The DBSCAN model adjusts clustering parameters according to data distribution to improve positioning accuracy during low-dimensional fault diagnosis. The AutoEncoder model enhances abnormal pattern recognition capabilities through reasonable encoding and decoding structures. The moving average model dynamically adjusts the time window size according to data. The seasonal autoregressive model optimizes parameters based on seasonal cycles. The nonlinear trend prediction model uses a fitting algorithm to adapt to trend changes.

8. The energy management multi-agent collaborative optimization system based on a large model agent architecture according to claim 5 is characterized by: The sensors and intelligent instruments in the data acquisition and management module have high precision, high reliability and real-time transmission capabilities. The data cleaning algorithm intelligently identifies and processes abnormal data. The data storage management system supports fast retrieval, backup and recovery. The big data technology and cloud computing platform use distributed computing architecture and parallel processing algorithms to improve efficiency, and can intelligently cache and optimize storage strategies.

9. The energy management multi-agent collaborative optimization system based on a large model agent architecture according to claim 5 is characterized by: The standardized interfaces in the system integration and interaction modules comply with industry standard specifications to ensure compatibility and interoperability. The user interface visualization uses intuitive charts and dynamic displays for easy operation. The user authority management function is complete to ensure data security and operational specifications.

10. The energy management multi-agent collaborative optimization system based on large model agent architecture according to claim 5 is characterized in that: The negotiation and coordination mechanism in the collaborative optimization control module adopts a distributed consensus algorithm to ensure decision consistency. The dynamic task allocation mechanism uses an intelligent algorithm to evaluate the task and agent status. The resource conflict and task competition resolution strategy is based on the principles of priority and fairness. The performance evaluation index system comprehensively monitors key indicators and automatically adjusts strategies and parameters.

Images