Electricity safety monitoring method and system based on multi-agent large model ensemble learning

Abstract

The present invention relates to the technical field of electricity safety management, and in particular to an electricity safety monitoring method and system based on multi-agent large model integrated learning. The method includes preprocessing electricity data from sensors, smart meters or / and user-submitted reports; constructing an electricity safety knowledge base based on electricity safety knowledge materials in the power industry; constructing an electricity safety analysis intelligent agent model based on specific issues of electricity safety analysis; combining the preprocessed electricity data with the electricity safety knowledge base based on the electricity safety analysis intelligent agent model to analyze electricity safety hazards and obtain analysis results of electricity safety hazards; and monitoring electricity safety based on the analysis results of electricity safety hazards. The present invention significantly improves the accuracy and real-time performance of electricity safety monitoring by combining multi-agent large model integrated learning with prompt engineering and domain expert knowledge, while enhancing the system's adaptability and intelligence level.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity safety management, and in particular to an electricity safety monitoring method and system based on multi-agent large model integrated learning. Background Art

[0002] With the development of industrialization and information technology, electricity safety has become an important social issue. Traditional electricity safety monitoring methods rely on simple sensor devices and manual monitoring. Although these methods have played a certain role in the early stages, with the increase in electricity demand and the complexity of the electricity environment, these traditional methods can no longer meet the needs of modern electricity safety monitoring. Limited data processing capabilities: Most existing electricity safety monitoring systems rely on traditional data processing technologies, which limits their ability to process large-scale electricity data. With the increase in electricity-consuming devices and the explosive growth of data volume, it is difficult for these systems to effectively process and analyze data in real time, thus affecting the accuracy and timeliness of monitoring. Lack of high adaptability: Most existing systems use preset rules for monitoring and alarming. These rules lack flexibility and are difficult to adapt to complex and changing electricity environments and equipment states. Therefore, the system may not be able to accurately identify all potential safety risks, or generate a high false alarm rate. Limited intelligence level: Although some advanced monitoring systems have begun to adopt machine learning technology, most of these systems are based on a single model and lack sufficient intelligence and self-optimization capabilities. This limits the system's ability to identify and respond to unknown or rare electricity safety issues. Insufficient collaborative monitoring and decision-making capabilities: Most existing power safety monitoring systems are designed as single, centralized monitoring systems, lacking an effective multi-agent collaborative working mechanism. This limits the system's ability to handle complex scenarios and make cross-regional monitoring decisions.

[0003] The main problems faced by existing power safety monitoring technologies include limited data processing capabilities, lack of high adaptability, limited intelligence level, and insufficient collaborative monitoring and decision-making capabilities. These problems limit the effectiveness of existing technologies in ensuring power safety. Therefore, developing a power safety monitoring method and system based on multi-agent large model ensemble learning can not only improve the accuracy and real-time performance of monitoring, but also improve the adaptability and intelligence level of the system, so as to more effectively identify and prevent power safety risks. Huang L et al. pointed out in the article "A survey on hallucination in large language models": As an emerging technical means, large language models have endowed artificial intelligence technology with certain logical reasoning capabilities and natural language perception capabilities, and can achieve comparable or even superior capabilities to humans in many tasks. However, its application in the field of power safety monitoring also faces the so-called "hallucination problem". This problem mainly refers to the fact that when a large model processes specific data, it may generate predictions or suggestions that do not conform to the actual situation. This phenomenon is particularly common when the model fails to fully understand the domain expertise or the data is not representative enough. The existence of hallucination problems may cause the power safety monitoring system to make wrong judgments, thus affecting the monitoring effect of the system and the accuracy of decision-making. To solve this problem, researchers have proposed several strategies. Zuo S et al. pointed out in the article "Moebert: from bert to mixture-of-experts via importance-guided adaptation" that, first, by enhancing the representativeness and diversity of the data, it is ensured that the data for model training can fully cover various situations in the field of power safety, reducing the possibility of hallucinations caused by data bias in the model. Secondly, by introducing domain expert knowledge and combining the expert system and the large model method, the model can take into account the professional knowledge and practical experience in the field when generating predictions or suggestions, thereby improving the accuracy and reliability of the model. In addition, the model integration method is adopted to reduce the hallucination problem that may occur in a single model by combining the prediction results of multiple models, thereby improving the overall monitoring effect. When combining expert systems and integrated learning, personalized prompt templates are constructed through prompt engineering to guide large models to perform power safety detection. Furthermore, a real-time monitoring and feedback mechanism is implemented, that is, the prediction results of the model are tracked in real time in the power safety monitoring system and compared with the actual situation. Once a large deviation is found between the model prediction and the actual situation, the model is adjusted immediately or manual intervention is taken to ensure the accuracy of monitoring and decision-making. Through this dynamic adjustment mechanism, the impact of the illusion problem on the power safety monitoring effect can be effectively reduced.In short, to solve the illusion problem of large models in electricity safety monitoring, it is necessary to start from multiple aspects such as data, model design, integrated learning, prompt engineering and real-time monitoring, and comprehensively use a variety of technical means to ensure the efficiency, accuracy and reliability of the electricity safety monitoring system. Summary of the invention

[0004] To solve the above problems, the present invention aims to provide a method and system for electricity safety monitoring based on multi-agent large model integrated learning, which introduces domain expert knowledge into the large language model and integrates the prediction results of multiple models to reduce the hallucination problem that may occur in a single model. By constructing a multi-agent integrated learning electricity safety monitoring expert system, using prompt engineering to build personalized prompt templates, guiding the large model to perform electricity safety analysis according to the rules, and improving the intelligence and accuracy of user electricity safety monitoring. To achieve the above purpose, the present invention provides the following technical solutions:

[0005] In a first aspect of the present invention, a method for monitoring power safety based on multi-agent large model ensemble learning is provided, the method comprising:

[0006] Pre-processing of electricity usage data from sensors, smart meters, or / and user-submitted reports;

[0007] Build an electricity safety knowledge base based on the electricity safety knowledge materials of the power industry;

[0008] Based on the specific problems of power safety analysis, an intelligent agent model for power safety analysis is constructed;

[0009] Based on the power safety analysis intelligent agent model, the preprocessed power consumption data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained;

[0010] Based on the analysis results of the hidden dangers of electricity safety, monitoring the safety of electricity use;

[0011] The pre-processing of electricity consumption data from sensors, smart meters or / and reports submitted by users includes:

[0012] Identify missing values ​​in electricity usage data from sensors, smart meters, or / and user-submitted reports;

[0013] Correct or delete the missing values;

[0014] Based on principal component analysis and feature selection techniques, extract the data features of the electricity consumption data after correcting or deleting missing values;

[0015] The specific problem of power safety analysis is based on which the power safety analysis agent model is constructed, including:

[0016] Based on the specific problems of power safety analysis, a corresponding power safety analysis intelligent agent model is constructed;

[0017] According to the tendency of the power safety analysis agent model, relevant data is screened and optimized from the pre-processed power consumption data and the power safety knowledge base to generate a personalized prompt template for the power safety analysis agent model;

[0018] Based on the power safety analysis agent agent model and the personalized prompt template, construct a power safety analysis agent model;

[0019] The monitoring of electricity safety based on the analysis result of the electricity safety hazard includes:

[0020] Build a domain expert system based on knowledge and rules in the field of power safety;

[0021] Perform weighted fusion on the analysis results of different power safety hazards corresponding to different power safety hazards to obtain a comprehensive risk assessment score;

[0022] Input the comprehensive risk assessment score into the domain expert system for analysis and reasoning, verify and optimize the analysis results of the electrical safety hazards, and obtain the final assessment results;

[0023] Based on the final assessment result, generating an electricity safety assessment report;

[0024] Electricity safety is monitored based on the electricity safety assessment report.

[0025] Furthermore, the power industry's electricity safety knowledge materials include power industry standards and specifications, safe operation materials, materials on the causes and consequences of electricity safety accidents, research results and academic reports.

[0026] Furthermore, based on the electricity safety knowledge materials of the power industry, a power safety knowledge base is constructed, including:

[0027] Collect information on electricity safety knowledge in the power industry;

[0028] Screen and organize the electricity safety knowledge materials and remove outdated or erroneous information;

[0029] Classify the electricity safety knowledge materials after eliminating outdated or erroneous information and build an electricity safety knowledge base.

[0030] Furthermore, the construction of an electricity safety knowledge base based on electricity safety knowledge materials in the power industry also includes:

[0031] Regularly refresh the electricity safety knowledge base by keeping up with the latest industry trends, updating safety standards or adding recent accident cases.

[0032] Furthermore, based on the specific problems of power safety analysis, the power safety analysis agent is constructed, which also includes:

[0033] The electricity safety analysis intelligent agent model is trained.

[0034] Furthermore, based on the power safety analysis agent model, the preprocessed power data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained, including:

[0035] Based on the power safety analysis agent model, the power safety knowledge base is used to analyze and monitor the pre-processed power consumption data to identify potential safety hazards and risks;

[0036] Provide early warning and handle identified potential safety hazards and risks;

[0037] Based on the results of the early warning and processing, analysis results of the power safety hazards corresponding to the power safety hazards are obtained.

[0038] Furthermore, the potential safety hazards of electricity use include overload prediction, short circuit detection or / and equipment aging assessment;

[0039] The power safety analysis agent model includes an overload detection agent model, a short circuit prediction agent model and / or an equipment aging assessment agent model;

[0040] The analysis results of the power safety hazards include: real-time status of power equipment, abnormal event records, risk assessment and power safety rating information.

[0041] Furthermore, the electricity safety assessment report includes potential risk points, targeted safety recommendations and preventive measures.

[0042] Furthermore, the safety recommendations and measures contained in the report are intended to guide users on how to improve the electricity usage environment and enhance electricity safety, and are of high practical value and operability.

[0043] A second aspect of the present invention provides a power safety monitoring system based on multi-agent large model integrated learning, the system comprising:

[0044] A preprocessing module, for preprocessing electricity consumption data from sensors, smart meters or / and reports submitted by users;

[0045] The first building module is used to build an electricity safety knowledge base based on electricity safety knowledge materials in the power industry;

[0046] The second building module is used to build an intelligent model for power safety analysis based on specific problems of power safety analysis;

[0047] An acquisition module is used to analyze the power safety hazards by combining the preprocessed power consumption data with the power safety knowledge base based on the power safety analysis intelligent agent model, and obtain the analysis results of the power safety hazards;

[0048] A monitoring module, which monitors electricity safety based on the analysis results of the electricity safety hazards;

[0049] The steps performed by the preprocessing module include:

[0050] Identify missing values ​​in electricity usage data from sensors, smart meters, or / and user-submitted reports;

[0051] Correct or delete the missing values;

[0052] Based on principal component analysis and feature selection techniques, extract the data features of the electricity consumption data after correcting or deleting missing values;

[0053] The steps performed by the second building module include:

[0054] Based on the specific problems of power safety analysis, a corresponding power safety analysis intelligent agent model is constructed;

[0055] According to the tendency of the power safety analysis agent model, relevant data is screened and optimized from the pre-processed power consumption data and the power safety knowledge base to generate a personalized prompt template for the power safety analysis agent model;

[0056] Based on the power safety analysis agent agent model and the personalized prompt template, construct a power safety analysis agent model;

[0057] The steps performed by the monitoring module include:

[0058] Build a domain expert system based on knowledge and rules in the field of power safety;

[0059] Perform weighted fusion on the analysis results of different power safety hazards corresponding to different power safety hazards to obtain a comprehensive risk assessment score;

[0060] Input the comprehensive risk assessment score into the domain expert system for analysis and reasoning, verify and optimize the analysis results of the electrical safety hazards, and obtain the final assessment results;

[0061] Based on the final assessment result, generating an electricity safety assessment report;

[0062] Electricity safety is monitored based on the electricity safety assessment report.

[0063] Furthermore, the steps performed by the acquisition module include:

[0064] Based on the power safety analysis agent model, the power safety knowledge base is used to analyze and monitor the pre-processed power consumption data to identify potential safety hazards and risks;

[0065] Provide early warning and handle identified potential safety hazards and risks;

[0066] Based on the results of the early warning and processing, analysis results of the power safety hazards corresponding to the power safety hazards are obtained.

[0067] Furthermore, the potential safety hazards of electricity use include overload prediction, short circuit detection or / and equipment aging assessment;

[0068] The power safety analysis agent model includes an overload detection agent model, a short circuit prediction agent model and / or an equipment aging assessment agent model;

[0069] The analysis results of the power safety hazards include: real-time status of power equipment, abnormal event records, risk assessment and power safety rating information.

[0070] According to a third aspect of the present invention, an electronic device is provided, the electronic device comprising at least one processor and at least one memory, the memory being data-connected to the processor, wherein:

[0071] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can perform the above method.

[0072] According to a fourth aspect of the present invention, a computer storable medium is provided, wherein computer instructions are stored on the storable medium, and when the computer instructions are executed by a processor, the steps in the above-mentioned method are specifically executed.

[0073] According to a fifth aspect of the present invention, a computer program product is provided, comprising computer instructions, wherein when the computer instructions are executed by a processor, the steps in the above-described method are specifically executed.

[0074] Technical effects and advantages of the present invention:

[0075] (A) Improving the accuracy and real-time performance of power safety monitoring: By combining the multi-agent large model integrated learning method, the present invention can effectively process and analyze large-scale power consumption data and realize real-time monitoring of the power environment and equipment status. Compared with the traditional power safety monitoring system, the present invention greatly reduces the illusion problem caused by data deviation or model simplification by introducing domain expert knowledge and adopting model integration methods, and improves the accuracy of monitoring. At the same time, by implementing real-time monitoring and feedback mechanisms, the model can be adjusted in time or manual intervention can be taken to ensure the accuracy of monitoring and decision-making, thereby effectively identifying and preventing power safety risks.

[0076] (B) Enhance the adaptability and intelligence level of the system: The present invention constructs a multi-agent integrated learning power safety monitoring expert system, builds a large model for prompt template generation based on prompt engineering, and dynamically generates personalized prompt templates, so that each monitoring agent focuses on a specific aspect of power safety, such as overload detection, short circuit prediction, equipment aging assessment, etc. These agents can use technologies such as large language models, combined with rules and cases in the power safety knowledge base, to identify and predict potential safety hazards. In addition, the analysis results of these agents are integrated through a weighted decision fusion mechanism, and a domain expert system is introduced. Combining the advantages of artificial intelligence and human experts, the refined design of prompt templates not only improves the accuracy and reliability of decision-making, but also greatly improves the efficiency of monitoring. This method provides an innovative solution for improving the intelligence and accuracy of power safety monitoring, which can flexibly adapt to complex and changeable power environments and equipment states, and effectively improve the adaptability and intelligence level of the system.

[0077] In summary, the present invention significantly improves the accuracy and real-time performance of power safety monitoring by combining multi-agent large model integrated learning and prompting engineering and domain expert knowledge, while enhancing the system's adaptability and intelligence level.

[0078] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A flow chart of a multi-agent large model integrated learning method for electricity safety monitoring in a specific embodiment of the present invention is shown;

[0080] Figure 2 It shows a flow chart of a multi-agent large model integrated learning power safety monitoring system in a specific embodiment of the present invention;

[0081] Figure 3The structure diagram of the multi-agent large model integrated learning method provided in a specific embodiment of the present invention is shown. DETAILED DESCRIPTION

[0082] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments 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 creative work are within the scope of protection of the present invention.

[0083] In order to solve the shortcomings of the prior art, the present invention discloses a method for monitoring power safety based on multi-agent large model ensemble learning. Figure 1 As shown, the method includes,

[0084] Step 1: Preprocessing the electricity consumption data from sensors, smart meters or / and user-submitted reports;

[0085] Step 2: Build an electricity safety knowledge base based on the electricity safety knowledge data of the power industry;

[0086] Step 3: Based on the specific problems of power safety analysis, construct a power safety analysis agent;

[0087] Step 4: Based on the power safety analysis agent, the preprocessed power data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained;

[0088] Step 5: Based on the analysis results of the potential safety hazards in electricity use, monitor electricity safety.

[0089] In a specific embodiment of the present invention, step 1: pre-processing the electricity consumption data from sensors, smart meters or / and user-submitted reports: first collect user electricity consumption data including power consumption, current, voltage, power, temperature and humidity, etc., which may come from sensors, smart meters or user reports. Then clean and standardize the data to remove noise and anomalies, ensure that the data format is suitable for model processing, and lay the foundation for ensuring the accuracy and consistency of data for subsequent electricity safety monitoring, and obtain the original sample set O.

[0090] The electricity consumption data preprocessing process in step 1 mainly consists of the following three sub-steps:

[0091] Step 101: Data collection can be done through a variety of channels, including data from different sources such as smart meters, sensors, and user reports.

[0092] Sensors are devices installed directly in the power system to monitor and record power parameters in real time. Smart meters provide detailed records of user power usage and support remote reading and management. In addition, user-submitted reports can also provide data on abnormal power usage or special circumstances.

[0093] Step 102: Identify and process missing values, erroneous values, and abnormal values ​​in the electricity usage data from sensors, smart meters, and / or user-submitted reports.

[0094] Missing values ​​and erroneous values ​​may come from equipment failures, data transmission errors, or external interference, while outliers may be caused by unusual power usage behavior. The quality of the data can be improved by applying various data cleaning techniques, such as anomaly detection algorithms and data smoothing methods;

[0095] Identify missing values, errors, and outliers in the data through the statistical analysis method Z-score and visualization technology; decide whether to correct, delete, or retain these missing values, errors, and outliers based on the situation; identify and delete duplicate records to ensure that each record in the data set is unique.

[0096] Step 103: Data conversion: Extract or construct new features from the original data according to the analysis or model requirements, such as calculating daily power consumption, peak current, etc. Perform one-hot encoding on categorical data and discretize continuous data to adapt to specific model algorithms.

[0097] Step 104: Data reduction: By applying principal component analysis (PCA) and feature selection techniques, the dimension of the data set is effectively reduced, and features with low contribution are screened out to reduce computational complexity, and the data features of the electricity consumption data after correcting or deleting missing values ​​are extracted.

[0098] Under the premise of ensuring that the data distribution remains unchanged, we reduce the amount of data through sampling to improve processing speed and efficiency. This not only simplifies the model structure, but also significantly reduces the computational burden. At the same time, on the basis of ensuring that the overall distribution of the data does not change, we adopt a carefully designed sampling strategy to reduce the amount of data, thereby optimizing processing speed and improving overall efficiency. Suppose we have an initial data set X , which contains m samples, each with n Features, namely Our goal is to simplify this dataset by dimensionality reduction and sampling. Apply PCA to reduce the dimensionality of the data. If we choose to keep k principal components, among which k<n, The goal of PCA is to find a transformation matrix , so that the transformed dataset Preserve as much variance of the original data as possible: .

[0099] In a specific embodiment of the present invention, step 2: constructing an electricity safety knowledge base based on electricity safety knowledge data of the power industry :Contains specifications, operating guidelines and cases, etc., aiming to provide theoretical support and risk prediction reference, and enhance system analysis capabilities.

[0100] In this link, the present invention is committed to creating a comprehensive knowledge base of electricity safety, which covers many aspects such as industry specifications, safety operation guidelines, and historical accident cases. By extensively collecting and carefully organizing these valuable information resources, a structured and rich knowledge system is formed. This knowledge base not only provides a solid theoretical support for the system, but also provides a key reference for identifying and predicting potential safety risks. It is an important tool for improving the system's analytical capabilities.

[0101] The step 2 specifically includes the following three sub-steps:

[0102] Step 201: Collecting electricity safety knowledge materials for the power industry: The electricity safety knowledge materials for the power industry include power industry standards and specifications, safety operation materials, materials on the causes and consequences of electricity safety accidents, research results and scientific and technological academic reports.

[0103] Collect standards and specifications for the power industry to ensure the authority and accuracy of the knowledge base content; the industry specifications include the latest national and international standards and regulations for electrical installation, maintenance and inspection.

[0104] Safety operation materials include guidance manuals and best practices for safe operation, safety operation manuals for various electrical equipment, emergency response measures, etc., providing practical safety operation knowledge; materials on the causes and consequences of power safety accidents include historical power safety accident cases, analyzing the causes, impacts, handling processes, lessons and consequences of accidents, and providing lessons for preventing future risks. Expert papers and research: including the latest research results, technical reports, etc., to provide scientific theoretical support.

[0105] Step 202: Screen, organize and structure the electricity safety knowledge materials, and remove outdated or erroneous information: Screen and organize the collected information, and remove outdated or inaccurate information. The organized information is classified according to a certain logical structure, such as different aspects of electricity safety or risk levels, to ensure that the structure of the knowledge system is clear and easy to search.

[0106] Step 203: Based on the application field or risk level of power safety, classify the power safety knowledge materials after removing outdated or erroneous information, and build a power safety knowledge base: integrate industry specifications and safety operation guidelines into the knowledge base to provide a solid theoretical basis for power safety. Utilize historical accident case analysis to identify potential safety risk points and provide reference for predicting and avoiding similar incidents.

[0107] Step 204: Application and update of knowledge base: Use the knowledge base as an analytical tool to support the identification, assessment and prediction of safety risks. Regularly update the knowledge base content by continuously following the latest industry trends, updating safety standards or new accident cases, etc. to ensure that it remains up-to-date, timely and highly accurate. Using the knowledge base as an analytical tool, it helps to discover, evaluate and predict safety risks.

[0108] In a specific embodiment of the present invention, step 3: based on the specific problems of power safety analysis, construct a power safety analysis agent: wherein, the user data in step 1 and the knowledge of the power safety knowledge base in step 2 are integrated, and according to the tendency of different models in step 3 (covering overload prediction, short circuit detection, equipment aging and other issues), a personalized prompt template is provided to prompt engineering design to ensure that the data set during each model training is optimized for its tendency, and an intelligent large language model agent for power safety is developed. During the training process, the model performance is regularly evaluated, and the training data set is adjusted as needed to optimize the accuracy and adaptability of the model. In addition, the monitoring of the model output should be strengthened to ensure that the information it provides is accurate and meets the standards and requirements for power safety.

[0109] Specific issues of the power safety analysis include overload prediction, short circuit detection and equipment aging.

[0110] Step 301: Based on the specific problems of power safety analysis, a corresponding power safety analysis agent large model is constructed;

[0111] Identify specific issues that need to be addressed in power safety analysis, such as overload prediction, short circuit detection, equipment aging assessment, etc. Decide to use advanced artificial intelligence technologies such as large language models to develop intelligent analysis agents to solve these problems.

[0112] Step 302: Based on the tendency of the power safety analysis agent large model, select and optimize relevant data from the pre-processed power data and power safety knowledge base to generate a personalized prompt template for the power safety analysis agent large model; use the prompt project to build a personalized prompt template to guide the large model to perform correct analysis. Among them, the template generation is dynamically adjusted and updated in time by the prompt generation agent large model.

[0113] Step 303: Based on the power safety analysis agent model and the personalized prompt template, construct a power safety analysis agent.

[0114] Step 304: Perform model training on the power safety analysis agent. Use the preprocessed data to train the model, and apply the large language model technology; during and after the training process, regularly evaluate the performance of the model. Use the validation set and the test set to test the performance of the model on unknown data until satisfactory accuracy and generalization ability are achieved.

[0115] Adjust the training dataset based on the results of the performance evaluation. If you find that the model performs poorly in some aspects, you can add relevant data to strengthen the model's learning in these aspects. Use resampling to sample the dataset in multiple dimensions, and use datasets for different problems to train multiple models separately as multi-agents.

[0116] In a specific embodiment of the present invention, step 4: based on the power safety analysis agent model, the pre-processed power data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained:

[0117] Model output monitoring and safety hazard identification. Develop a monitoring system to monitor the output of the model in real time. Ensure that the information provided by the model is accurate and meets the standards and requirements for power safety. Combine safety rules and cases for in-depth analysis: Use technologies such as large language models to combine safety rules and historical cases to conduct in-depth analysis of the model output. Identify potential safety hazards and risks. Achieve comprehensive safety hazard identification and early warning: Through in-depth analysis, achieve comprehensive identification and early warning of power safety hazards. Improve monitoring efficiency and accuracy, discover and handle safety issues in a timely manner, and ensure users' power safety.

[0118] In a specific embodiment of the present invention, step 5: monitoring electricity safety based on the analysis result of the electricity safety hazard, includes:

[0119] Step 501: weighted fusion of the analysis results of the power safety hazards corresponding to different power safety hazards to obtain a comprehensive risk assessment score:

[0120] Step 5011: Integrate multi-agent analysis results: Collect the output results of the aforementioned intelligent analysis agents (such as overload detection agent, short circuit prediction agent, equipment aging assessment agent, etc.). These results include real-time status of electrical equipment, abnormal event records, risk assessment, and power safety rating information.

[0121] Step 5012: Weighted fusion decision mechanism: Assign a weight to each analytical agent based on its importance and reliability. These weights can be obtained through historical data analysis or set by domain experts based on experience. Using the defined weights, the output results of different agents are weighted and fused. This step aims to comprehensively consider information from all aspects and form a comprehensive evaluation result.

[0122] Step 502: constructing a domain expert system based on knowledge and rules in the field of power safety;

[0123] Introduce domain expert system: Establish an expert system that includes knowledge and rules in the field of power safety. The system can simulate the decision-making process of human experts and provide in-depth insights. Input the weighted fusion results into the expert system, and further verify and optimize the evaluation results through the analysis and reasoning of the expert system. When necessary, submit the preliminary evaluation results of the system to human experts for review. Experts can adjust and supplement the evaluation results based on their own experience and knowledge. Integrate the feedback of human experts back into the system to continuously optimize the model and decision-making mechanism.

[0124] Step 503: input the comprehensive risk assessment score into a domain expert system for analysis and reasoning, verify and optimize the analysis result of the potential safety hazard of electricity use, and obtain the final assessment result;

[0125] Output safety assessment report: Based on the final assessment results obtained in the above steps, a detailed electricity safety assessment report is generated. The report not only lists potential risk points, but also provides targeted safety recommendations and preventive measures. The safety recommendations and measures included in the report are intended to guide users on how to improve the electricity environment and improve electricity safety, and have high practical value and operability.

[0126] Step 504: Generate an electricity safety assessment report based on the final assessment result;

[0127] Step 505: Monitor electricity safety based on the electricity safety assessment report.

[0128] In summary, the method of the present invention starts with data preprocessing, including collecting and cleaning electricity data (such as electricity consumption, current, voltage, power, etc.) to ensure the accuracy and consistency of the data. Next, a knowledge base containing electricity safety rules, standards and historical cases is established to provide theoretical support for safety analysis. A large model of prompt generation agents is constructed through prompt engineering, personalized prompt templates are dynamically generated, and multiple large language model agents focusing on different aspects of electricity safety are developed (such as overload detection, short circuit prediction, equipment aging assessment, etc.), and these agents are used to analyze data and identify potential risks in different fields. Finally, by integrating the analysis results of these agents and combining the judgment of the domain expert system, a final electricity safety assessment report is formed, which includes potential risks, safety recommendations and preventive measures, aiming to provide users with practical electricity safety guidance.

[0129] The present invention also provides a power safety monitoring system based on multi-agent large model integrated learning, the system comprising:

[0130] A preprocessing module, for preprocessing electricity consumption data from sensors, smart meters or / and reports submitted by users;

[0131] The first building module is used to build an electricity safety knowledge base based on electricity safety knowledge materials in the power industry;

[0132] The second building module is used to build an intelligent power safety analysis agent based on the specific problems of power safety analysis;

[0133] An acquisition module is used to analyze the power safety hazards by combining the preprocessed power consumption data with the power safety knowledge base based on the power safety analysis intelligent agent model, and obtain the analysis results of the power safety hazards;

[0134] The monitoring module monitors the safety of electricity use based on the analysis results of the potential safety hazards of electricity use.

[0135] In a specific embodiment of the present invention, the second building block includes:

[0136] The first construction submodule is used to construct a corresponding power safety analysis intelligent agent model based on power safety hazards;

[0137] The first production submodule is used to screen and optimize relevant data from the preprocessed power consumption data and the power safety knowledge base according to the tendency of the power safety analysis intelligent agent model, and generate a personalized prompt template for the power safety analysis intelligent agent large model;

[0138] The second construction submodule is used to construct a power safety analysis agent model based on the power safety analysis agent large model and the personalized prompt template;

[0139] A training submodule, used for training the intelligent agent model;

[0140] The first acquisition submodule is used to analyze and monitor the pre-processed electricity usage data based on the trained intelligent agent model and the electricity safety knowledge base to obtain the analysis results of the electricity safety hazards corresponding to the electricity safety hazards.

[0141] Regarding the system device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0142] An electronic device, comprising at least one processor and at least one memory, wherein the memory is data-connected to the processor,

[0143] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can perform the above method.

[0144] A computer storable medium, characterized in that computer instructions are stored on the storable medium, and when the computer instructions are executed by a processor, the steps in the above method are specifically executed.

[0145] A computer program product includes computer instructions, wherein when the computer instructions are executed by a processor, the steps in the above method are specifically executed.

[0146] The following will provide further explanation in conjunction with specific embodiments.

[0147] Figure 2This is a method and system flow chart for monitoring power safety based on multi-agent large model integrated learning. In building an efficient power safety analysis system, the power information from multiple data sources is firstly preprocessed in a refined manner to ensure the quality and consistency of the data. This includes all kinds of data from smart meters to sensors and user reports. By eliminating irregular items, a clear and reliable data foundation is provided for the model. Next, a comprehensive power safety knowledge base is constructed, covering industry specifications, safety operation guidelines and historical accident cases, etc., to provide a solid theoretical support for the system and a key reference for safety risk prediction. Based on this, intelligent analysis agents for power safety issues, such as overload prediction, short circuit detection and equipment aging assessment, are developed, and multi-dimensional and in-depth risk analysis is performed using large language models and prompt engineering and other technologies. In addition, the present invention also develops a prompt generation agent model that can dynamically generate prompt templates required by the analysis agent, which can generate personalized prompt templates for each agent in a timely manner in combination with the latest power safety standards. Finally, through a set of refined decision-making mechanisms, the analysis results of each intelligent agent are integrated, combined with the deep insights of the domain expert system, to form a final safety assessment report. The report not only clearly points out potential risk points, but also provides targeted safety recommendations and preventive measures, providing users with a practical and efficient electricity safety guidance program.

[0148] Figure 3 This is a structural diagram of the multi-agent large model ensemble learning method. Select suitable pre-trained large models according to the task requirements. These models can be different versions of the same type or models with different architectures to cover a wider feature space. Evaluate the performance of each model on the target task individually. This step is crucial for the subsequent weight assignment. Assign weights to each model based on the performance of a single model. Weights can be determined by heuristic methods or automatically adjusted by optimization algorithms (such as genetic algorithms, gradient descent, etc.). Multiply the prediction results of each model by its corresponding weight, and then combine these weighted results. Finally, integrate the analysis results of each intelligent agent and combine the deep insights of the domain expert system to form a final safety assessment report.

[0149] The method provided in a specific embodiment of the present invention comprises the following steps:

[0150] Step 1: Preprocess the electricity consumption data from sensors, smart meters, or / and user-submitted reports: First, collect user electricity consumption data including power consumption, current, voltage, power, temperature, and humidity, which may come from sensors, smart meters, or user reports. Then clean and standardize the data to remove noise and anomalies, ensure that the data format is suitable for model processing, and lay the foundation for ensuring the accuracy and consistency of data for subsequent electricity safety monitoring, and obtain the original sample set O;

[0151] The data preprocessing process in the present invention mainly consists of the following three sub-steps:

[0152] Step 101: Data collection can be done through a variety of channels. Sensors are devices installed directly in the power system that monitor and record power parameters in real time. Smart meters provide detailed records of user power usage and support remote reading and management. In addition, user-submitted reports can also provide data on abnormal or special circumstances.

[0153] Step 102: This step aims to identify and deal with errors and outliers in the data. Noise may come from equipment failure, data transmission errors, or external interference, while outliers may be caused by unusual power usage behavior. The quality of the data can be improved by applying various data cleaning techniques, such as anomaly detection algorithms and data smoothing methods.

[0154] Step 103: By applying principal component analysis (PCA) and feature selection techniques, we can effectively reduce the dimensionality of the dataset and filter out those features with low contribution. This not only simplifies the model structure, but also significantly reduces the computational burden. At the same time, while ensuring that the overall distribution of the data is not changed, a carefully designed sampling strategy is adopted to reduce the amount of data, thereby optimizing the processing speed and improving the overall efficiency. Suppose we have an initial dataset , which contains samples, each with Features, namely Our goal is to simplify this dataset by dimensionality reduction and sampling. Apply PCA to reduce the dimensionality of the data. If we choose to keep principal components, among which , the goal of PCA is to find a transformation matrix , so that the transformed dataset Preserve as much variance of the original data as possible: .

[0155] Step 2: Build a knowledge base on electricity safety based on the electricity safety knowledge of the power industry. Build a comprehensive knowledge base on electricity safety , including specifications, operating guidelines and cases, aims to provide theoretical support and risk prediction references and enhance system analysis capabilities.

[0156] The construction of the electricity safety database in the present invention mainly consists of the following three sub-steps:

[0157] Step 201: Content collection. Including industry specifications: collect the latest national and international standards and regulations for electrical installation, maintenance and inspection. Safety operation guide: organize safety operation manuals, emergency response measures, etc. for various electrical equipment. Historical accident cases: collect historical electrical accident cases, including accident causes, impacts, handling processes and lessons. Expert papers and research: including the latest research results, technical reports, etc., to provide scientific theoretical support.

[0158] Step 202: Organize and structure the electrical safety knowledge in step 201. Carefully screen and organize the collected information to eliminate outdated or erroneous information. Then, classify the refined information according to a clear logical structure, such as various fields of electrical safety or different risk levels, to ensure that the knowledge system is clearly structured and easy to find.

[0159] Step 203: Use the knowledge base as an analysis tool to help discover, evaluate and predict safety risks. Regularly refresh the knowledge base content by continuously following the latest industry trends, updating safety standards or adding recent accident cases to ensure that it remains up-to-date and highly accurate.

[0160] Step 3: Based on the specific problems of power safety analysis, build a power safety analysis agent. Combining the user data in step 1 and the knowledge base of power safety in step 2, according to the tendency of different models (covering overload prediction, short circuit detection, equipment aging and other issues), a personalized prompt template is provided by prompting engineering design to ensure that the data set for each model training is optimized for its tendency, and an intelligent large language model agent for power safety is developed. During the training process, the model performance is evaluated regularly, and the training data set is adjusted as needed to optimize the accuracy and adaptability of the model. In addition, the monitoring of the model output should be strengthened to ensure that the information it provides is accurate and meets the standards and requirements of power safety. Using technologies such as large language models, combined with safety rules and case in-depth analysis, comprehensive safety hazard identification and early warning can be achieved to improve monitoring efficiency and accuracy.

[0161] The large language model agent for power safety analysis in the present invention is mainly composed of the following three sub-steps:

[0162] Step 301: Dataset optimization and model tendency analysis. According to different power safety issues (such as overload prediction, short circuit detection, equipment aging, etc.), determine the type of model that needs to be developed. Each type of model has its specific data requirements and tendencies. According to the tendency of each model, filter and optimize relevant data from user data and power safety knowledge base, and generate "intelligent agent big model" through prompts. It is an independent intelligent agent big model used to generate personalized prompt templates. Ensure that the data set can cover various situations that the model needs to predict or detect, so that the data set is more focused on the specific needs of the model.

[0163] Specifically, the K-Mean nearest neighbor clustering method is used to classify the data set:

[0164]

[0165] in, Represents data points With cluster center The distance between Indicates the index of the cluster center that minimizes the distance , so that the data points Assigned to the subset represented by the closest cluster center.

[0166] The design of the prompt template is shown in Table 1-4. Generally speaking, this refined prompt template is refined by domain experts, or it can be assisted by a more advanced large model. In the present invention, this process combines the latest knowledge base of power safety standards and the auxiliary prompts of artificial domain experts to dynamically generate the design of the power safety analysis prompt template required by each large model, which can help the large language model to identify safety monitoring tasks and guide different agents to focus on the analysis of specific safety issues.

[0167] For example, a computer-generated overload detection analysis prompt template is shown in Table 1:

[0168] Table 1: Overload detection analysis prompt template

[0169]

[0170] The computer generated short circuit prediction analysis prompt template is shown in Table 2:

[0171] Table 2: Short circuit prediction analysis prompt template

[0172]

[0173] The computer-generated equipment aging assessment analysis prompt template is shown in Table 3:

[0174] Table 3: Equipment aging assessment analysis prompt template

[0175]

[0176] The computer-generated comprehensive electricity safety detection analysis prompt template is shown in Table 4:

[0177] Table 4: Comprehensive electricity safety detection and analysis prompt template

[0178]

[0179] Step 302: Model training and performance evaluation. Use the optimized dataset to train each model. Apply large language model technology. Regularly evaluate the performance of the model during and after training. Use the validation set and test set to verify the performance of the model on unknown data to ensure the accuracy and generalization ability of the model. Adjust the training dataset based on the results of the performance evaluation. If it is found that the model performs poorly in some aspects, relevant data can be added to strengthen the model's learning in these aspects.

[0180] Step 4: Based on the electricity safety analysis agent, the pre-processed electricity data is combined with the electricity safety knowledge base to analyze the electricity safety hazards and obtain the analysis results of the electricity safety hazards. Develop a monitoring system to monitor the output of the model in real time. Ensure that the information provided by the model is accurate and meets the standards and requirements for electricity safety. Conduct in-depth analysis in combination with safety rules and cases: Use technologies such as large language models, combined with safety rules and historical cases, to conduct in-depth analysis of the model output. Identify potential safety hazards and risks. Achieve comprehensive identification and early warning of safety hazards: Through in-depth analysis, comprehensive identification and early warning of electricity safety hazards can be achieved. Improve monitoring efficiency and accuracy, promptly discover and deal with safety issues, and ensure users' electricity safety.

[0181] Step 5: Based on the analysis results of the power safety hazards, monitor power safety: The power safety decision-making expert model integrates the results of all large language model agent analysis, uses a precise weighted fusion strategy for comprehensive evaluation, and outputs the ultimate safety assessment report. This model incorporates the domain expert system and combines the deep insights of human experts to significantly improve the accuracy and reliability of decision-making. The report not only thoroughly marks various potential risks, but also puts forward specific safety recommendations and prevention strategies, providing users with practical and efficient power safety solutions.

[0182] The power safety decision-making expert model in the present invention is mainly composed of the following three sub-steps:

[0183] Step 501: Integrate the results of multiple agents to improve the quality of decision making. This integration is not a simple merger, but an accurate weighted fusion strategy is adopted, which means that the outputs of different agents are given different weights according to their reliability and importance. This method can ensure that the final safety assessment report integrates all available information and reflects the most in-depth analysis.

[0184] Assume that The evaluation result of each intelligent agent for a specific power safety risk is ,in The evaluation result of each agent is assigned a weight according to its reliability and importance. The final comprehensive risk assessment score ( ) can be calculated by weighted average:

[0185] .

[0186] Step 502: Risk level assessment. The power safety assessment reports analyzed by each intelligent agent are summarized, and the expert model is further integrated to generate a complete assessment report. The overall power safety risk assessment level is obtained by referring to the comprehensive risk score obtained in step 4.

[0187] Step 503: Human expert enhancement. The model not only relies on algorithms and automated analysis, but also incorporates the knowledge of domain expert systems and the insights of human experts. This combination of artificial intelligence and human expert knowledge greatly enhances the accuracy and reliability of the model in power safety assessment. In this way, the model can identify and thoroughly mark various potential power risks, while providing specific safety recommendations and preventive measures.

[0188] After the above five steps, a power safety monitoring method and system based on multi-agent large model integrated learning is formed, so that the large language model can accurately generate power safety analysis and evaluation reports, minimizing possible hallucination problems.

[0189] In general, this technology significantly improves the effect of power safety monitoring through advanced data preprocessing and intelligent analysis modules, which is specifically reflected in the following aspects: Building a comprehensive power safety knowledge base: This technology deeply explores industry standards, safety specifications and historical power accident cases to establish a structured knowledge base. This knowledge base provides a rich theoretical basis and reference basis for the intelligent agent, greatly enhancing the system's ability to identify and predict potential safety risks. Realizing multi-agent integrated learning: By developing multiple intelligent agents focusing on different aspects of power safety (such as overload detection, short circuit prediction, equipment aging assessment, etc.), and using advanced technologies such as large language models for data analysis, and using prompt engineering to build a prompt generation model to dynamically generate personalized prompt templates, this technology can comprehensively consider various safety factors and achieve accurate identification and prediction of potential safety hazards. Improve the accuracy of monitoring: Through the dual advantages of integrated learning and domain expert systems, this technology significantly reduces the problem of misjudgment caused by data deviation or model simplification, and implements real-time monitoring and feedback mechanisms to ensure the high accuracy of monitoring and decision-making, and effectively prevent power safety risks. Enhanced system adaptability and intelligence: The multi-agent integrated learning framework of this technology, combined with artificial intelligence and human expert judgment, not only improves the accuracy and reliability of decision-making, but also improves monitoring efficiency. This innovative approach enables the system to flexibly adapt to complex and changing power consumption environments and equipment states, improving the system's adaptability and intelligence.

[0190] Example 2

[0191] According to the power consumption data and equipment parameters shown in Table 5, the corresponding data is input according to the actual use example input information diagram.

[0192] Table 5: Input information table for actual use example

[0193]

[0194] The comprehensive power grid security monitoring and analysis report actually output after monitoring according to the method of the present invention is as follows:

[0195] 1. Safety analysis report: Based on the power consumption data and equipment parameters provided, the current power consumption situation has the following potential safety hazards:

[0196] Overload risk: Power consumption peaks approached or exceeded the equipment rated power many times, especially at the 600kW peak.

[0197] Short circuit risk: The insulation resistance of the equipment is low (10MΩ), and short circuit events have occurred in the past.

[0198] Equipment aging risk: The equipment has been in use for 6 years and shows signs of aging.

[0199] The overall risk rating is: 8 (high risk)

[0200] 2. Overload detection results: The current power load situation shows that the equipment is close to or exceeds the rated power many times, and there is a risk of overload. High temperature (30°C) and high humidity (60%) further increase the load on the equipment.

[0201] Overload precautions:

[0202] 1) Equipment running time, avoid long-term high-load operation.

[0203] 2) Increase the heat dissipation measures of the equipment to reduce the temperature of the equipment.

[0204] 3) Check equipment cables and connections regularly to ensure they are functioning properly.

[0205] 3. Short circuit prediction results:

[0206] The insulation resistance of the equipment is low (10MΩ), and short-circuit incidents have occurred in the past, indicating that the equipment has a certain short-circuit risk.

[0207] Short circuit prevention measures:

[0208] 1) Check the insulation condition of the equipment regularly and replace aging or damaged insulation materials in time.

[0209] 2) Ensure that the environment around the equipment is dry to reduce the impact of humidity on equipment insulation.

[0210] 3) Install a short-circuit protection device to cut off the power supply in time to prevent greater damage caused by short circuit.

[0211] 4. Equipment aging assessment results: The equipment has been in use for 6 years. Although it is regularly maintained every year, the aging speed of the equipment may be accelerated under high load, high temperature and high humidity environment.

[0212] Equipment maintenance suggestions:

[0213] 1) Strengthen daily maintenance of equipment, especially inspection and maintenance after high-load operation.

[0214] 2) Consider replacing some key components, such as cables and insulation materials, to extend the service life of the equipment.

[0215] 3) Develop a more frequent maintenance plan based on the actual usage of the equipment.

[0216] 5. Comprehensive preventive measures are recommended:

[0217] 1) Optimize power load distribution according to the equipment's operating status and environmental conditions to avoid long-term high-load operation.

[0218] 2) Conduct regular and comprehensive equipment inspections and maintenance to promptly identify and address potential safety hazards.

[0219] 3) Strengthen the heat dissipation and moisture-proof measures of the equipment to ensure a good operating environment for the equipment.

[0220] 4) Improve operators’ safety awareness and skills to ensure rapid response in case of emergency.

[0221] 6. Emergency treatment plan:

[0222] 1) When overload or short circuit occurs, cut off the power supply immediately to prevent equipment damage.

[0223] 2) Contact professional maintenance personnel for inspection and repair.

[0224] 3) Emergency contact information: XXXXXX.

[0225] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for monitoring power safety based on multi-agent large model ensemble learning, characterized in that: The method comprises, Pre-processing of electricity usage data from sensors, smart meters, or / and user-submitted reports; Build an electricity safety knowledge base based on the electricity safety knowledge materials of the power industry; Based on the specific problems of power safety analysis, an intelligent agent model for power safety analysis is constructed; Based on the power safety analysis intelligent agent model, the preprocessed power consumption data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained; Based on the analysis results of the hidden dangers of electricity safety, monitoring the safety of electricity use; The pre-processing of electricity consumption data from sensors, smart meters or / and reports submitted by users includes: Identify missing values ​​in electricity usage data from sensors, smart meters, or / and user-submitted reports; Correct or delete the missing values; Based on principal component analysis and feature selection techniques, extract the data features of the electricity consumption data after correcting or deleting missing values; The specific problem of power safety analysis is based on which the power safety analysis agent model is constructed, including: Based on the specific problems of power safety analysis, a corresponding power safety analysis intelligent agent model is constructed; According to the tendency of the power safety analysis agent model, relevant data is screened and optimized from the pre-processed power consumption data and the power safety knowledge base to generate a personalized prompt template for the power safety analysis agent model; Based on the power safety analysis agent agent model and the personalized prompt template, construct a power safety analysis agent model; The monitoring of electricity safety based on the analysis result of the electricity safety hazard includes: Build a domain expert system based on knowledge and rules in the field of power safety; Perform weighted fusion on the analysis results of different power safety hazards corresponding to different power safety hazards to obtain a comprehensive risk assessment score; Input the comprehensive risk assessment score into the domain expert system for analysis and reasoning, verify and optimize the analysis results of the electrical safety hazards, and obtain the final assessment results; Based on the final assessment result, generating an electricity safety assessment report; Electricity safety is monitored based on the electricity safety assessment report.

2. The method for power safety monitoring based on multi-agent large model ensemble learning according to claim 1 is characterized in that: The electricity safety knowledge materials in the power industry include power industry standards and specifications, safe operation materials, materials on the causes and consequences of electricity safety accidents, research results and academic reports.

3. The method for power safety monitoring based on multi-agent large model ensemble learning according to claim 2 is characterized in that: Based on the electricity safety knowledge materials of the power industry, a power safety knowledge base is constructed, including: Collect information on electricity safety knowledge in the power industry; Screen and organize the electricity safety knowledge materials and remove outdated or erroneous information; Classify the electricity safety knowledge materials after eliminating outdated or erroneous information and build an electricity safety knowledge base.

4. The method for monitoring power safety based on multi-agent large model integrated learning according to any one of claims 1 to 3, characterized in that: The construction of the electricity safety knowledge base based on the electricity safety knowledge data of the power industry also includes: Regularly refresh the electricity safety knowledge base by keeping up with the latest industry trends, updating safety standards or adding recent accident cases.

5. The method for power safety monitoring based on multi-agent large model ensemble learning according to claim 1 is characterized in that: Based on the specific problems of power safety analysis, the power safety analysis agent is constructed, which also includes: The electricity safety analysis intelligent agent model is trained.

6. The method for power safety monitoring based on multi-agent large model ensemble learning according to claim 1 is characterized in that: Based on the power safety analysis intelligent agent model, the preprocessed power data is combined with the power safety knowledge base to analyze the power safety hazards, and the analysis results of the power safety hazards are obtained, including: Based on the power safety analysis agent model, the power safety knowledge base is used to analyze and monitor the pre-processed power consumption data to identify potential safety hazards and risks; Provide early warning and handle identified potential safety hazards and risks; Based on the results of the early warning and processing, analysis results of the power safety hazards corresponding to the power safety hazards are obtained.

7. The method for power safety monitoring based on multi-agent large model integrated learning according to claim 6 is characterized in that: Electrical safety hazards include overload prediction, short circuit detection and / or equipment aging assessment; The power safety analysis agent model includes an overload detection agent model, a short circuit prediction agent model and / or an equipment aging assessment agent model; The analysis results of the power safety hazards include: real-time status of power equipment, abnormal event records, risk assessment and power safety rating information.

8. The method for power safety monitoring based on multi-agent large model ensemble learning according to claim 1 is characterized in that: The electricity safety assessment report includes potential risk points, targeted safety recommendations and preventive measures.

9. The method for electricity safety monitoring based on multi-agent large model ensemble learning according to claim 8 is characterized in that: The safety recommendations and measures included in the report are intended to guide users on how to improve the electricity usage environment and enhance electricity safety, and are of high practical value and operability.

10. An electricity safety monitoring system based on multi-agent large model ensemble learning, characterized in that: The system comprises, A preprocessing module, for preprocessing electricity consumption data from sensors, smart meters or / and reports submitted by users; The first building module is used to build an electricity safety knowledge base based on electricity safety knowledge materials in the power industry; The second building module is used to build an intelligent model for power safety analysis based on specific problems of power safety analysis; An acquisition module is used to analyze the power safety hazards by combining the preprocessed power consumption data with the power safety knowledge base based on the power safety analysis intelligent agent model, and obtain the analysis results of the power safety hazards; A monitoring module, which monitors electricity safety based on the analysis results of the electricity safety hazards; The steps performed by the preprocessing module include: Identify missing values ​​in electricity usage data from sensors, smart meters, or / and user-submitted reports; Correct or delete the missing values; Based on principal component analysis and feature selection techniques, extract the data features of the electricity consumption data after correcting or deleting missing values; The steps performed by the second building module include: Based on the specific problems of power safety analysis, a corresponding power safety analysis intelligent agent model is constructed; According to the tendency of the power safety analysis agent model, relevant data is screened and optimized from the pre-processed power consumption data and the power safety knowledge base to generate a personalized prompt template for the power safety analysis agent model; Based on the power safety analysis agent agent model and the personalized prompt template, construct a power safety analysis agent model; The steps performed by the monitoring module include: Build a domain expert system based on knowledge and rules in the field of power safety; Perform weighted fusion on the analysis results of different power safety hazards corresponding to different power safety hazards to obtain a comprehensive risk assessment score; Input the comprehensive risk assessment score into the domain expert system for analysis and reasoning, verify and optimize the analysis results of the electrical safety hazards, and obtain the final assessment results; Based on the final assessment result, generating an electricity safety assessment report; Electricity safety is monitored based on the electricity safety assessment report.

11. The power safety monitoring system based on multi-agent large model integrated learning according to claim 10 is characterized in that: The steps performed by the acquisition module include: Based on the power safety analysis agent model, the power safety knowledge base is used to analyze and monitor the pre-processed power consumption data to identify potential safety hazards and risks; Provide early warning and handle identified potential safety hazards and risks; Based on the results of the early warning and processing, analysis results of the power safety hazards corresponding to the power safety hazards are obtained.

12. The power safety monitoring system based on multi-agent large model integrated learning according to claim 10 is characterized in that: Electrical safety hazards include overload prediction, short circuit detection and / or equipment aging assessment; The power safety analysis agent model includes an overload detection agent model, a short circuit prediction agent model and / or an equipment aging assessment agent model; The analysis results of the power safety hazards include: real-time status of power equipment, abnormal event records, risk assessment and power safety rating information.

13. An electronic device, comprising at least one processor and at least one memory, wherein the memory is data-connected to the processor, wherein: The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

14. A computer storable medium, characterized in that: The storable medium stores computer instructions, and when the computer instructions are executed by the processor, the steps in the method according to any one of claims 1 to 9 are specifically performed.

15. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the steps in the method according to any one of claims 1 to 9 are specifically performed.

Images