Energy field corpus generation system
Through the automatic generation of corpus in the energy field through multi-agent systems, the problem of high cost and unstable quality of corpus generation in the energy field is solved, and efficient and low-cost corpus generation and model update are achieved.
Patent Information
- Application Number
- CN202510591898.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2025-08-08
AI Technical Summary
The training of large models in the energy field relies on manually labeled corpus, resulting in high labor costs and uneven quality, lacking the ability to automatically generate and update corpus, making it difficult to adapt to the rapidly developing energy field.
The collaboration mechanism of multi-agent system is adopted, including coordinating the management layer, agent layer, knowledge base and corpus, automatically generate high-quality energy field corpus, reduce manual labeling costs, improve corpus generation efficiency, and update the corpus in real time through the perception, decision-making and execution of the agent in the agent layer.
It realizes efficient and low-cost generation of high-quality energy field corpus, ensuring that the model can quickly adapt to new data and trends, reducing manual labeling costs, and improving the automation level of corpus generation.
Smart Images
Figure CN120450899A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a corpus generation system in the energy field. Background Art
[0002] The application of artificial intelligence technologies such as big models in the fields of energy and carbon emissions has gradually attracted attention. By analyzing massive amounts of data, they can provide accurate carbon emission monitoring, energy optimization strategies and policy evaluation support.
[0003] However, existing technologies for training large energy-related models rely on manually annotated corpora. Generating high-quality corpora requires extensive manual effort, resulting in high labor costs and inconsistent quality. These corpora also lack the ability to automatically generate and update corpora, making them difficult to adapt to the rapid development of the energy sector. Currently, traditional data mining techniques or simple machine learning models are increasingly being used to process corpus data, but these methods are ineffective when dealing with large-scale, complex, and dynamically changing energy data. Summary of the Invention
[0004] The present invention provides a corpus generation system in the energy field to solve the problems that corpus generation in the energy field requires high labor costs and has uneven quality, and lacks the ability to automatically generate and update corpuses.
[0005] In a first aspect, an embodiment of the present invention provides a corpus generation system in the energy field, comprising: a coordination management layer, an intelligent agent layer, a knowledge base, and a corpus;
[0006] The coordination management layer is used to allocate energy field corpus generation tasks and manage the agents in the agent layer;
[0007] The intelligent agent layer is used to perform the energy field corpus generation task;
[0008] The knowledge base is used to store knowledge related to the energy field corpus generation task;
[0009] The corpus is used to store energy field corpora generated by the intelligent agent layer, and the energy field corpora are used to train models in the energy field.
[0010] The technical solution of the embodiment of the present invention provides an energy field corpus generation system comprising: a coordination management layer, an agent layer, a knowledge base, and a corpus; wherein the coordination management layer is used to allocate energy field corpus generation tasks and manage the agents in the agent layer; the agent layer is used to execute energy field corpus generation tasks; the knowledge base is used to store knowledge related to energy field corpus generation tasks; the corpus is used to store energy field corpus generated by the agent layer, and the energy field corpus is used to train energy field models. Utilizing the collaborative mechanism of the multi-agent system, high-quality energy field corpus is automatically generated, solving the problems of high labor costs and uneven quality of energy field corpus generation, as well as the lack of the ability to automatically generate and update corpus. This reduces the cost of manual annotation, improves corpus generation efficiency, and the real-time updateable corpus ensures that the trained model can quickly adapt to new data and trends.
[0011] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.
[0013] Figure 1 A schematic diagram of the architecture of a system for generating corpus in the energy field provided by an embodiment of the present invention;
[0014] Figure 2 A schematic diagram of the architecture of a coordination management layer in an energy field corpus generation system provided by an embodiment of the present invention;
[0015] Figure 3 A schematic diagram of the architecture of the agent layer in an energy field corpus generation system provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0016] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.
[0017] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0018] Figure 1 This is a schematic diagram of the architecture of a system for generating energy-related corpus provided by an embodiment of the present invention. This embodiment is applicable to automatically generating and updating energy-related corpus and is used for model training of energy-related corpus.
[0019] like Figure 1 As shown, the energy field corpus generation system includes: a coordination management layer 10, an agent layer 20, a knowledge base 30 and a corpus 40; wherein, the coordination management layer 10 is used to assign energy field corpus generation tasks and manage the agents in the agent layer 20; the agent layer 20 is used to perform energy field corpus generation tasks; the knowledge base 30 is used to store knowledge related to energy field corpus generation tasks; the corpus 40 is used to store energy field corpus generated by the agent layer, and the energy field corpus is used to train models in the energy field.
[0020] The coordination management layer can be understood as a functional layer for managing intelligent agents and coordinating the allocation of energy domain corpus generation tasks. The energy domain corpus generation task can be understood as the task of generating energy domain corpus, which can be used to train energy domain models. Energy domain models include but are not limited to large models and machine learning models. The energy sector refers to the general term for technical, economic, and social activities related to the production, conversion, transmission, utilization, and management of energy. It includes traditional energy sources (such as coal, oil, and natural gas) and new energy sources (such as solar energy, wind energy, hydropower, biomass energy, and nuclear energy).
[0021] The knowledge base is used to store the knowledge and data required to generate the energy field corpus. The corpus is used to store the generated energy field corpus.
[0022] Specifically, the agent layer 20 includes multiple agents with different functions that collaborate with each other. For example, agents with different functions are designed according to system goals and task requirements, such as data collection agents, solution decision agents, and solution execution agents. Each agent has a certain degree of autonomy and professionalism, can independently complete specific tasks, and can jointly complete complex tasks by collaborating with each other, sharing information and resources. The coordination management layer 10 decomposes the energy field corpus generation task and assigns tasks according to the capabilities, status, and resource conditions of each agent in the agent layer. The agents in the agent layer 20 interact with the knowledge base 30 based on the received tasks, and collaboratively generate energy field corpus, and store the energy field corpus in the corpus 40.
[0023] As an optional embodiment of the embodiment of the present application, Figure 2 This is a schematic diagram of the architecture of the coordination management layer in the energy field corpus generation system provided by an embodiment of the present invention, such as Figure 2 As shown, the coordination management layer 10 includes: a task manager 11 and an agent manager 12;
[0024] The task manager 11 is used to decompose the energy field corpus generation task into multiple subtasks, and assign each subtask to the corresponding agent in the agent layer; the subtasks include data collection subtasks, corpus decision subtasks and corpus generation subtasks; the agent manager 12 is used to resolve competition and conflicts between the agents in the agent layer based on a preset collaboration mechanism.
[0025] The task manager 11 can be understood as a module for managing the energy field corpus generation task, and the agent manager 12 can be understood as a module for managing the relationship between multiple agents included in the agent layer.
[0026] Specifically, in the coordination management layer 10, the task manager 11 is used to perform task intent recognition and task analysis on the received energy field corpus generation task, and decompose the energy field corpus generation task into multiple subtasks, such as decomposing the energy field corpus generation task into a data collection subtask, a corpus decision subtask and a corpus generation subtask. According to the capabilities, status and resource conditions of each intelligent agent, each subtask is assigned to the corresponding intelligent agent. The task manager 11 needs to have the ability to adjust dynamically, and when the task requirements or the status of the intelligent agent change, it can reallocate tasks in time to improve the adaptability and efficiency of the system. Exemplarily, the task manager 11 can call an analytical algorithm, an analytical model or a large model to perform task analysis, and the embodiment of the present invention does not limit this.
[0027] The agent manager 12 is used to manage the collaborative relationship between multiple agents in the agent layer, and resolve competition and conflicts between agents through a preset collaborative mechanism. The preset collaborative mechanism includes: priority rules and negotiation mechanisms, etc., to ensure the stability of system operation and harmonious collaboration between agents.
[0028] This embodiment manages the energy field corpus generation task through a task manager and manages the agents through an agent manager, so that the agents can harmoniously collaborate to complete the energy field corpus generation task.
[0029] As an optional embodiment of the present application, Figure 3 This is a schematic diagram of the architecture of the agent layer in an energy field corpus generation system provided by an embodiment of the present invention, such as Figure 3 As shown, the agent layer 20 includes at least one perception agent 21, at least one decision agent 22 and at least one execution agent 23;
[0030] The perception agent 21 is used to collect and pre-process energy field data based on the data collection subtask, and store the pre-processed energy field data into the knowledge base 30;
[0031] The decision agent 22 is used to obtain knowledge information and energy field data for generating corpus from the knowledge base 30 based on the corpus decision subtask, and determine the corpus generation strategy according to the target knowledge and the energy field data;
[0032] The execution agent 23 is used to generate energy-domain corpus based on the corpus generation subtask, the corpus generation strategy, and the energy-domain data, and store the energy-domain corpus in a corpus 40 .
[0033] The perception agent can be understood as being used to perform the data collection subtask, the decision agent is used to perform the decision subtask, and the execution agent is used to perform the corpus generation subtask.
[0034] Specifically, the Perception Agent 21, based on the Data Collection subtask, perceives and collects data from the energy environment, preprocesses the collected data, and stores the preprocessed data in the knowledge base, providing a data basis for the subsequent generation of energy-related corpora. Based on the Corpus Decision subtask, the Decision Agent 22 retrieves the corresponding energy-related data and knowledge information from the knowledge base, determines a corpus generation strategy based on the energy-related data and knowledge information, and transmits the corpus generation strategy and energy-related data to the Execution Agent. The corpus generation strategy may include algorithms or models for generating energy-related corpora, constraints, and resource allocation methods. Based on the Corpus Generation subtask, the Execution Agent 23 generates specific execution instructions according to the Corpus Generation Strategy. The execution instructions are input into the corresponding Corpus Generation Model to obtain energy-related corpora, which are then stored in the corpus. The Corpus Generation subtask may include task objectives and output formats. Exemplarily, the executing agent 23 generates corpus generation instruction information Prompt based on the corpus generation subtask and corpus generation strategy according to the knowledge information and energy field data, inputs the corpus generation instruction information into the corpus generation model, and obtains energy field corpus; the energy field corpus can be dialogue form corpus, text form corpus, graphic form corpus or structured form corpus.
[0035] Exemplary methods for collecting data in energy environments include: using crawlers to obtain energy-related information from web pages, using text recognition to identify useful information from energy documents, and collecting field data from energy equipment through sensors deployed in energy systems. Preprocessing of the collected data can include filtering and denoising, data cleaning, data compression, and feature extraction to ensure that the preprocessed data has more concise, accurate, and significant features, facilitating subsequent analysis and decision-making.
[0036] This embodiment automatically completes the energy field corpus generation task through the respective functions and mutual cooperation of the perception agent, decision-making agent and execution agent, without the need for manual labeling, thus reducing labor costs.
[0037] As an optional embodiment of the present application, at least two decision-making agents coordinate or collaboratively optimize the algorithm to determine the corpus generation strategy based on a game theory mechanism, a voting mechanism, or a negotiation mechanism.
[0038] Specifically, the agent layer 20 may include one or more decision agents 22. If there is a decision agent, the corpus generation strategy may be determined based on a preset reasoning algorithm, which may include deductive reasoning, inductive reasoning, or fuzzy reasoning.
[0039] If there are multiple decision-making agents, they can coordinate and determine the corpus generation strategy through game theory, voting, or negotiation. Alternatively, they can use optimization algorithms (such as genetic algorithms and particle swarm optimization) for collaborative decision-making, exchanging information and conducting collaborative searches among multiple decision-making agents to find the globally optimal or near-optimal decision solution.
[0040] For example, two decision-making agents can reach a consensus through bilateral negotiation. For example, in a resource allocation problem, two decision-making agents can propose resource requirements and allocation plans to each other, and through multiple rounds of negotiation, ultimately reach a mutually acceptable resource allocation agreement. Multiple agents can also determine corpus generation strategies based on multilateral negotiation.
[0041] For example, decision-making agents can use a simple majority voting mechanism. Each agent votes individually on a given decision solution, and the solution that receives more than half the votes is adopted. This mechanism is simple and easy to implement, and is suitable for scenarios with a large number of agents and relatively clear decision solutions. Alternatively, a weighted voting mechanism can be used. The voting weights of agents are determined based on their abilities, experience, and contributions. Each agent votes on a given decision solution, and agents with higher weights have a greater influence on the decision outcome. Weighted voting can better reflect the diversity and importance of decision-making agents.
[0042] As an optional embodiment of the present application, the execution agent 23 includes: a planning module 231 and a control module 232;
[0043] The planning module 231 is used to formulate a corpus generation plan based on the corpus generation subtask and the corpus generation strategy;
[0044] The control module 232 is configured to determine a corpus generation instruction according to the corpus generation plan and the energy field data, and call a corpus generation algorithm based on the corpus generation instruction to generate energy field corpus.
[0045] The planning model is used to develop a detailed action plan, such as the execution sequence and timeline. The control module converts the plan into specific execution instructions, which then drive the corresponding objects to complete the task. The corpus generation plan can be understood as a more detailed execution plan guided by the corpus generation strategy. The corpus generation instructions control the corpus generation algorithm to produce energy-related corpora.
[0046] Specifically, within the execution agent 23, the planning module 231 generates a corpus generation plan based on the corpus generation strategy determined by the decision agent 22 and the corpus generation subtasks issued by the coordination management layer, and sends the corpus generation plan to the control module 232. Based on the corpus generation plan and the energy domain data, the control module 232 determines corpus generation instructions and, based on the corpus generation instructions, invokes a corpus generation algorithm to generate energy domain corpora that conform to the corpus generation plan.
[0047] As an optional embodiment of the embodiment of the present application, the knowledge base includes: energy data, domain knowledge, historical corpus data and logical rules;
[0048] Among them, the energy data provides data basis for the decision-making agent; the domain knowledge provides task-related background information and professional guidance for the decision-making agent; the historical corpus data provides positive or negative cases for the decision-making agent; and the logical rules provide reasoning logic and decision rules for the decision-making agent.
[0049] Specifically, the perception agent senses and collects energy data from on-site equipment in the energy sector, processes this data, and stores it in a knowledge base, providing data support for the decision-making agent. Furthermore, the knowledge base stores domain knowledge, historical corpus data, and logical rules in the energy sector, providing the decision-making agent with background knowledge in the energy sector and ensuring that the resulting corpus is more relevant to the actual energy sector. This historical corpus data can include both approved and unapproved energy sector corpus, with approved energy sector corpus serving as positive examples and unapproved energy sector corpus serving as negative examples.
[0050] As an optional embodiment of the embodiment of the present application, the system further includes an evaluation feedback module;
[0051] The evaluation feedback module is used to evaluate the energy field corpus generated by the intelligent agent layer and feed back the evaluation results to the intelligent agent layer so that each intelligent agent in the intelligent agent layer can be optimized based on the evaluation results.
[0052] Specifically, the energy-related corpus generated by the agent layer needs to be evaluated to ensure its quality. This evaluation can include accuracy, consistency, and relevance. The evaluation results are then fed back to the agent layer, enabling the decision-making and execution agents within the agent layer to optimize themselves based on these results, ensuring the quality of the energy-related corpus.
[0053] As an optional embodiment of the embodiment of the present application, multiple agents in the agent layer share data based on a preset communication topology, and each of the agents has a privacy protection mechanism;
[0054] The preset communication topology is a distributed communication topology or a centralized communication topology; and the privacy protection mechanism includes at least one of desensitization processing, encryption processing and identity authentication.
[0055] Specifically, multiple agents in the agent layer can be connected using a distributed communication topology, forming a decentralized communication network that improves the scalability and robustness of the system. Alternatively, a centralized communication topology can be used, where all agents communicate with a central node that aggregates, processes, and forwards information, facilitating centralized management and control of the agents.
[0056] Data sharing between agents based on communication protocols enables better collaboration and improves the efficiency of energy corpus generation. Information sharing can be achieved through publish-subscribe and request-response models. The communication protocols used between agents can be standardized protocols such as MQTT and AMQP to ensure compatibility and interoperability between different agents. Standardized protocols can specify communication data formats, transmission methods, connection management, and other aspects, reducing the complexity of system integration. Customized communication protocols can also be designed to meet specific system requirements and agent characteristics. Customized protocols can more flexibly meet the system's unique communication requirements, such as efficiently transmitting specific data types and implementing specialized communication security mechanisms.
[0057] Furthermore, while agents are sharing information, they can also implement privacy protection mechanisms as needed. For sensitive information, privacy protection measures such as encryption and anonymization are required to prevent leakage and misuse. Encrypted communication technology is used to ensure the security of data transmitted between agents, preventing data from being intercepted, tampered with, or forged. Identity authentication is performed on both communicating parties to ensure the legitimacy and credibility of the communication.
[0058] As an optional embodiment of the embodiment of the present application, the system also includes: a virtual environment simulator, which is used to simulate real task scenarios and provide a training and testing environment for the agents in the agent layer.
[0059] Specifically, by testing and optimizing the intelligent agent layer through a virtual environment simulator, the physical laws, environmental changes, and task scenarios of a real energy environment can be simulated, providing a safe and controllable training and testing environment for the intelligent agent, helping developers evaluate the performance of the system and the synergistic effect of the intelligent agent.
[0060] As an optional embodiment of the embodiment of the present application, the system also includes: a real environment interface, which is used to realize the interaction between the intelligent agent in the intelligent agent layer and the external environment, and to perform optimized learning of the intelligent agent based on the perceived external environment data.
[0061] Specifically, when the system is deployed in a real energy environment, the real-world interface is responsible for the interaction between the intelligent agent and the external world, including the collection of sensor data and the sending of control instructions. It needs to be highly reliable and real-time to ensure that the intelligent agent can accurately perceive the environment and respond promptly.
[0062] Through machine learning and reinforcement learning, agents can learn from their interactions with the environment, continuously optimizing their decision-making models and behavioral strategies, and improving their ability to survive and complete tasks in complex environments. The learning process can be either online or offline.
[0063] In another alternative embodiment, redundancy is introduced between key modules and agents. When a module or agent fails, a backup module or agent can immediately take over its functions, ensuring normal system operation. A detection module is incorporated into the system to monitor the system's operating status in real time, detect anomalies promptly, and take appropriate action. Exception handling includes fault isolation, system restart, and emergency risk avoidance to prevent the anomaly from spreading and causing further losses.
[0064] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.
[0065] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.
Claims
1. A corpus generation system in the energy field, characterized by: include: Coordinate the management layer, agent layer, knowledge base and corpus; The coordination management layer is used to allocate energy field corpus generation tasks and manage the agents in the agent layer; The intelligent agent layer is used to perform the energy field corpus generation task; The knowledge base is used to store knowledge related to the energy field corpus generation task; The corpus is used to store energy field corpora generated by the intelligent agent layer, and the energy field corpora are used to train models in the energy field.
2. The energy field corpus generation system according to claim 1, characterized in that: The coordination management layer includes: a task manager and an agent manager; The task manager is used to decompose the energy field corpus generation task into multiple subtasks and assign each subtask to a corresponding agent in the agent layer; the subtasks include a data collection subtask, a corpus decision subtask, and a corpus generation subtask; The agent manager is used to resolve competition and conflicts between agents in the agent layer based on a preset collaboration mechanism.
3. The energy field corpus generation system according to claim 2, characterized in that: The agent layer includes at least one perception agent, at least one decision agent and at least one execution agent; The perception agent is used to collect and pre-process energy field data based on the data collection subtask, and store the pre-processed energy field data in the knowledge base; The decision agent is used to obtain knowledge information and energy field data for generating corpus from a knowledge base based on the corpus decision subtask, and determine a corpus generation strategy based on the knowledge information and the energy field data; The execution agent is used to generate energy field corpus based on the corpus generation subtask, the corpus generation strategy and the energy field data, and store the energy field corpus in a corpus.
4. The energy field corpus generation system according to claim 3, characterized in that: At least two decision-making agents coordinate and determine the corpus generation strategy based on a game theory mechanism, a voting mechanism, or a negotiation mechanism.
5. The energy field corpus generation system according to claim 3, characterized in that: The execution agent includes: a planning module and a control module; The planning module is used to formulate a corpus generation plan according to the corpus generation subtask and the corpus generation strategy; The control module is used to determine a corpus generation instruction according to the corpus generation plan and the energy field data, and to call a corpus generation algorithm based on the corpus generation instruction to generate energy field corpus.
6. The energy field corpus generation system according to any one of claims 3 to 5, characterized in that: The knowledge base includes: energy data, domain knowledge, historical corpus data and logical rules; Among them, the energy data provides data basis for the decision-making agent; the domain knowledge provides task-related background information and professional guidance for the decision-making agent; the historical corpus data provides positive or negative cases for the decision-making agent; and the logical rules provide reasoning logic and decision rules for the decision-making agent.
7. The energy field corpus generation system according to any one of claims 1 to 5, characterized in that: The system also includes an evaluation feedback module; The evaluation feedback module is used to evaluate the energy field corpus generated by the intelligent agent layer and feed back the evaluation results to the intelligent agent layer so that each intelligent agent in the intelligent agent layer can be optimized based on the evaluation results.
8. The energy field corpus generation system according to any one of claims 1 to 5, characterized in that: The multiple agents in the agent layer share data based on a preset communication topology, and each agent has a privacy protection mechanism; The preset communication topology is a distributed communication topology or a centralized communication topology; and the privacy protection mechanism includes at least one of desensitization processing, encryption processing and identity authentication.
9. The energy field corpus generation system according to any one of claims 1 to 5, characterized in that: The system also includes: a virtual environment simulator, which is used to simulate real task scenarios and provide a training and testing environment for the agents in the agent layer.
10. The energy field corpus generation system according to any one of claims 1 to 5, characterized in that: The system further includes: a real environment interface for realizing the interaction between the intelligent agent in the intelligent agent layer and the external environment, and performing optimized learning of the intelligent agent based on the perceived external environment data.