Electric power information system inspection and resource allocation method based on knowledge base and intelligent agent

Abstract

The invention relates to the field of intelligent inspection and resource allocation, in particular to an electric power information system inspection and resource allocation method based on a knowledge base and an intelligent agent, and the method comprises the steps: collecting the data of an electric power information system, carrying out the heterogeneous data integration, and constructing the knowledge base of the electric power information system; constructing an intelligent agent oriented to the electric power information system based on a large language model, and performing RAG on the intelligent agent based on a knowledge base and open source knowledge to enhance the performance of the intelligent agent; acquiring a routing inspection demand in a natural language form, automatically generating a routing inspection script by the intelligent agent based on the routing inspection demand, and compiling the routing inspection script to generate a routing inspection task; the intelligent agent automatically executes the inspection task and compares the inspection result with the knowledge base to generate an inspection report; and the intelligent agent performs intelligent analysis on the inspection report, and performs resource allocation on abnormal items in the inspection report in combination with the knowledge base. According to the method, the operation and maintenance efficiency of the electric power information system is effectively improved, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of intelligent inspection and resource allocation, and particularly to a method for inspecting and allocating resources of a power information system based on a knowledge base and an intelligent agent. Background Art

[0002] With the advent of the information age, the power industry is facing unprecedented challenges and opportunities. Against this background, the power information system has become the key to the efficient operation and sustainable development of the power industry. The power information system is a comprehensive production management information system that connects power equipment, substations, power transmission and distribution networks, power users, and power loads of power enterprises to form power informatization. It is an integrated management platform that combines information technology and power systems, aiming to achieve intelligent monitoring, analysis, and optimization of the entire process of power production, transmission, distribution, and consumption. The power information system effectively improves the reliability, economy, and sustainability of the power grid through "data-driven".

[0003] However, as the operation years increase, problems gradually emerge in the operation and maintenance management of the power information system. On the one hand, the scale of the power system is expanding day by day, and the operation complexity of the power information system is constantly increasing, and the amount of data to be processed is growing explosively. On the other hand, with the gradual advancement of the construction of the smart grid, higher requirements are put forward for the intelligence and automation levels of the power information system.

[0004] Regularly inspecting the power information system can effectively identify potential faults and safety hazards, so as to carry out reasonable resource allocation to ensure the safety of power supply. Chinese Patent with application number 202411534543.1 discloses a method for intelligent inspection of a power information system. This method uses Python scripts to regularly analyze and diagnose the information to be diagnosed, and automatically generates feedback information according to the abnormal rating results diagnosed and the data types of the information to be diagnosed, and sends it to the mobile devices of the operation and maintenance personnel through an internal and external network isolation device, so as to achieve intelligent inspection.

[0005] However, the above method still requires a large amount of manual labor, such as the writing of Python scripts, and after the inspection is completed, the operation and maintenance personnel need to allocate resources according to the feedback information. That is to say, the automation degree of the existing methods for inspecting and allocating resources of the power information system is still insufficient, and the cost and operation and maintenance pressure are still relatively high. Summary of the Invention

[0006] In view of this, an embodiment of the present application proposes a method for inspecting and resource allocation of a power information system based on a knowledge base and an agent, aiming to build a full-stack intelligent perception system for the power information system, and use the knowledge base and the agent to automatically perform intelligent inspections according to business characteristics, and form a scientific and reasonable resource allocation solution, thereby improving resource utilization, enhancing the operation and maintenance efficiency of the power information system, and reducing the operation and maintenance costs of the power information system.

[0007] To achieve the above object, an embodiment of the present application proposes a method for inspecting and resource allocation of a power information system based on a knowledge base and an agent, which is applicable to the operation and maintenance management of the power information system, including: collecting the infrastructure data, operation data, log data, and work order data of the power information system, and performing heterogeneous data integration to build a knowledge base of the panoramic resources of the power information system; building an agent for the power information system based on a large language model, and performing RAG (Retrieval-augmented Generation) on the agent based on the knowledge base and open-source knowledge to enhance the performance of the agent; obtaining inspection requirements in the form of natural language, automatically generating an inspection script by the agent based on the inspection requirements, compiling the inspection script to generate an inspection task corresponding to the inspection requirements; automatically executing the inspection task by the agent, comparing the inspection results with the knowledge base, and generating an inspection report; intelligently parsing the inspection report by the agent, and combining the knowledge base to perform resource allocation for the abnormal items in the inspection report to solve the abnormal items.

[0008] To achieve the above object, an embodiment of the present application also proposes a system for inspecting and resource allocation of a power information system based on a knowledge base and an agent, which is applicable to the operation and maintenance management of the power information system, including: a knowledge base construction module, an agent construction module, an interaction module, and a multi-dimensional display module; the knowledge base construction module is used to collect the infrastructure data, operation data, log data, and work order data of the power information system, and perform heterogeneous data integration to build a knowledge base of the panoramic resources of the power information system; the agent construction module is used to build an agent for the power information system based on a large language model, and perform RAG on the agent based on the knowledge base and open-source knowledge to enhance the performance of the agent; the interaction module is used to obtain inspection requirements in the form of natural language; the agent is used to automatically generate an inspection script based on the inspection requirements, compile the inspection script to generate an inspection task corresponding to the inspection requirements, automatically execute the inspection task, compare the inspection results with the knowledge base, generate an inspection report, and intelligently parse the inspection report, and combine the knowledge base to perform resource allocation for the abnormal items in the inspection report to solve the abnormal items; the multi-dimensional display module is used to perform real-time multi-dimensional display on the entire workflow of the agent.

[0009] To achieve the above object, an embodiment of the present application further provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable 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 execute a method for power information system inspection and resource allocation based on a knowledge base and an agent as described above.

[0010] To achieve the above object, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which when executed by a processor, can implement a method for power information system inspection and resource allocation based on a knowledge base and an agent as described above.

[0011] A method for power information system inspection and resource allocation based on a knowledge base and an agent proposed by the present application first collects data of the power information system to build a knowledge base of the panoramic resources of the power information system. Subsequently, an agent for the power information system is constructed based on the large language model, and the agent is subjected to RAG based on the knowledge base and open-source knowledge to enhance the performance of the agent. Next, the inspection requirements in the form of natural language are obtained, and the agent automatically generates an inspection script based on the inspection requirements, compiles the inspection script, generates an inspection task and automatically executes the inspection task, compares the inspection results with the knowledge base to generate an inspection report, and then intelligently analyzes the inspection report, and combines the knowledge base to allocate resources for the abnormal items in the inspection report to solve the abnormal items. This method builds a full-stack intelligent perception system for the power information system, realizes the full-automatic development of the inspection script, shortens the development time to the minute level, the agent after RAG has stronger and more stable performance, combines the power information system exclusive knowledge base to allocate resources, provides continuous enhancement guarantee for various services of the power information system, realizes the hour-level early warning of key system failures, and greatly improves the resource utilization rate, improves the operation and maintenance efficiency of the power information system, and reduces the operation and maintenance cost of the power information system, thereby providing a stable and good user experience for power generation enterprises and power users.

[0012] In some alternative embodiments, infrastructure data, operation data, historical log data, and work order data of the power information system are collected, and heterogeneous data integration is performed to build a knowledge base of the panoramic resources of the power information system, including: using NLP (Natural Language Processing) technology to extract unstructured documents storing the infrastructure data of the power information system, collecting the infrastructure data of the power information system, and storing it in Elasticsearch; using Flink to monitor the real-time business flow and log flow of the power information system, collecting the operation data and log data of the power information system, and writing them into the Kafka message queue; using Spark to process the batch-structured work orders of the power information system, collecting the work order data of the power information system, and writing them into HBase; performing heterogeneous data integration on the data stored in Elasticsearch, Kafka message queue, and HBase, and building a four-dimensional relationship model based on Neo4j to obtain a knowledge base of panoramic resources covering cloud platforms, servers, switches, network devices, storage arrays, middleware, databases, resource pools, and business applications; where the four-dimensional relationship includes the spatial relationship between entity components, the logical mapping relationship between services, the skill matrix and responsibility matrix of the operation and maintenance team, and the connection relationship between entity components and virtual components. The database established in this way can serve as the knowledge reserve support for the application technology, typical experience, and optimization practice of the power information system, providing a solid theoretical foundation for the training of intelligent agents and the intelligent agents to perform intelligent inspections and resource allocation.

[0013] In some alternative embodiments, the large language model is any one of the Guangming Power Model, the Hunyuan Model, the Spark Model, the Pangu Model, and the Tianguang Model. RAG is performed on the intelligent agent based on the knowledge base and open-source knowledge to enhance the performance of the intelligent agent, including: obtaining open-source knowledge related to the power information system, including but not limited to fault reports, accident case reviews, warning measures, rules and regulations, professional books, and expert experience; performing OCR (Optical Character Recognition) on the obtained open-source knowledge, and forming an open-source library based on the recognized data; performing RAG on the intelligent agent based on the knowledge base and the open-source library, enhancing the vertical private ability of the intelligent agent through the knowledge base, and enhancing the parallel common ability of the intelligent agent through the open-source library; where, during the process of performing RAG on the intelligent agent, it is necessary to reorder the data in the knowledge base and the open-source library in terms of vectors, bytes, paragraphs, and chapters. The application of the RAG technology introduces external open-source knowledge, which can avoid the intelligent agent from having hallucinations, and at the same time well enhances the timeliness and interpretability of the intelligent agent to perform tasks.

[0014] In some alternative embodiments, if the format of the patrol inspection requirement in natural language form is text, the patrol inspection requirement is directly input into the intelligent agent. If the format of the patrol inspection requirement in natural language form is voice, it is converted into text and then input into the intelligent agent. The intelligent agent automatically generates a patrol inspection script based on the patrol inspection requirement, compiles the patrol inspection script, and generates a patrol inspection task corresponding to the patrol inspection requirement, including: deploying the intelligent agent using a large language model deployment framework and configuring a natural language to script compiler for the deployed intelligent agent. The large language model deployment framework is any one of vLLM, Ollama, Xinference, and LMStudio, and the natural language to script compiler is any one of Cline, Continue, Roo Code, and Trae; inputting the patrol inspection requirement into the deployed intelligent agent, and the intelligent agent automatically generates a patrol inspection script using the natural language to script compiler and compiles it, determines the patrol inspection address corresponding to the patrol inspection requirement, and then uses the workflow technology to establish a workflow of the patrol inspection task corresponding to the patrol inspection requirement based on the patrol inspection address; wherein, the large language model deployment framework supports real-time display of the analysis process of the intelligent agent for the patrol inspection requirement and the workflow of the patrol inspection task established corresponding to the patrol inspection requirement. The analysis process of the intelligent agent for the patrol inspection requirement and the workflow of the patrol inspection task established corresponding to the patrol inspection requirement can be displayed in real time through the large language model deployment framework, facilitating monitoring and learning by operation and maintenance personnel.

[0015] In some alternative embodiments, the patrol inspection tasks are divided into switch patrol inspection, cloud platform patrol inspection, and resource pool patrol inspection according to the patrol inspection direction. One patrol inspection task includes several patrol inspection items; the patrol inspection items for switch patrol inspection include checking the utilization rates of the CPU (Central Processing Unit) and memory of the switch, the voltage and current of the power module, the working status of the fan, the link aggregation status, the stacking status, the number of VLANs (Virtual Local Area Networks), network connectivity, the status of service boards, alarm records, the status of physical interfaces, temperature, and running time; the patrol inspection items for cloud platform patrol inspection include checking the utilization rates of the CPU, memory, and disks of the physical hosts in the cloud platform, the voltage and current of the power module, the working status of the fan, the storage status of OBS, the utilization rate of OBS, the utilization rates of the CPU, memory, and disks of CCE containers (Cloud Container Engine), the utilization rates of the CPU, memory, and disks of virtual machines, the instance status and disk utilization rates of RDS (Relational Database Service), DSC (DM Database), and Gauss Database on the cloud, the disk utilization rate of the MRS big data cluster, the usage rate and remaining available capacity of the storage pools in 7 AZ partitions (Available Zones), and the availability of the operation-side cloud service interface; the patrol inspection items for resource pool patrol inspection include checking the utilization rates of the CPU, memory, and disks of the physical hosts in the resource pool, the voltage and current of the power module, the working status of the fan, the status of network cards, the utilization rates of the CPU, memory, and disks of virtual machines, and the IP address, CPU model, number of CPU cores, and number of threads of the physical hosts.

[0016] In some alternative embodiments, the intelligent agent automatically executes the patrol inspection tasks, compares the patrol inspection results with the knowledge base, and generates a patrol inspection report, including: automatically executing each patrol inspection item in the patrol inspection tasks by the intelligent agent to obtain the patrol inspection results of each patrol inspection item; searching for the corresponding standard metrics of each patrol inspection item from the knowledge base, comparing the patrol inspection results of each patrol inspection item with the corresponding standard metrics in turn, determining the patrol inspection items whose patrol inspection results meet the corresponding standard metrics as normal items, and determining the patrol inspection items whose patrol inspection results do not meet the corresponding standard metrics as abnormal items; generating a patrol inspection report that supports online viewing and downloading based on the patrol inspection results and determination results of each patrol inspection item. Judging the patrol inspection results of each patrol inspection item according to the standard metrics in the knowledge base can scientifically and accurately discover the potential risks existing in the power information system.

[0017] In some alternative embodiments, the agent intelligently analyzes the inspection report and, in combination with the knowledge base, allocates resources for the abnormal items in the inspection report to resolve the abnormal items, including: the agent intelligently analyzes the inspection report, analyzes each abnormal item in turn, and determines simple abnormal items that can be automatically resolved and complex abnormal items that cannot be automatically resolved; in combination with the knowledge base, allocates resources for the abnormal items that can be automatically resolved to resolve the simple abnormal items, and the resource allocation includes elastic scaling on the cloud, automatic expansion, automatic shutdown, and automatic restart; sends the relevant information of the complex abnormal items to the operation and maintenance team to remind the operation and maintenance team to resolve the complex abnormal items. Intelligent inspection needs to balance automation and reliability. Therefore, the agent can determine the abnormal items. For simple abnormal items, the agent automatically allocates resources, and for complex abnormal items, the agent needs to promptly notify the operation and maintenance team for processing, so as to ensure that the complex abnormal items can be quickly resolved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] To more clearly illustrate the embodiments of the present application or the technical solutions in the related art, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or the related art. Obviously, the following-described drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 is a flowchart of a method for inspecting and resource allocating a power information system based on a knowledge base and an agent provided in an embodiment of the present application; Figure 2 is a schematic diagram of the workflow of an inspection task provided in an embodiment of the present application; Figure 3 is a schematic diagram of the agent analyzing inspection requirements provided in an embodiment of the present application; Figure 4 is a schematic diagram of the upper half of an inspection report provided in an embodiment of the present application; Figure 5 is a schematic diagram of the lower half of an inspection report provided in an embodiment of the present application; Figure 6 is a schematic diagram of the structure of a power information system inspection and resource allocation system based on a knowledge base and an agent provided in another embodiment of the present application; Figure 7 is a schematic diagram of the structure of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will elaborate on each embodiment of this application in conjunction with the accompanying drawings. However, those of ordinary skill in the art can understand that in each embodiment of this application, many technical details are provided to help readers better understand this application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can still be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation to the specific implementation of this application. The various embodiments can be combined and cross-referenced with each other on the premise of not being contradictory.

[0021] An embodiment of this application proposes a method for power information system inspection and resource allocation based on a knowledge base and an agent, which is applicable to the operation and maintenance management of power information systems and is applied to electronic devices. Among them, the electronic device can be a terminal or a server. In this embodiment and the following embodiments, the server is taken as an example for illustration. The following will specifically describe the implementation details of a method for power information system inspection and resource allocation based on a knowledge base and an agent proposed in this embodiment. The following content is only implementation details provided for convenient understanding and is not necessary for implementing this solution.

[0022] The specific process of a method for power information system inspection and resource allocation based on a knowledge base and an agent proposed in this embodiment can be as Figure 1 shown and includes: Step 101, collect the infrastructure data, operation data, log data, and work order data of the power information system, and perform heterogeneous data integration to construct a knowledge base of the panoramic resources of the power information system.

[0023] In specific implementation, the knowledge base is the theoretical foundation for the power information system. Therefore, the server needs to first construct the knowledge base of the power information system. The server needs to collect the infrastructure data, operation data, log data, and work order data of the target power information system and perform heterogeneous data integration, thereby constructing the knowledge base of the panoramic resources of the target power information system.

[0024] In one example, the infrastructure data, operation data, historical log data, and work order data of the power information system are heterogeneous and need to be collected using different technologies. The infrastructure data of the power information system is stored in the form of unstructured documents. The server needs to use NLP technology to extract the unstructured documents storing the infrastructure data of the power information system, collect the infrastructure data of the power information system, and store it in Elasticsearch. The operation data and log data of the power information system are real-time data streams. The server needs to use Flink to monitor the real-time business flow and log flow of the power information system, collect the operation data and log data of the power information system, and write them into the Kafka message queue. The work order data of the power information system is in batches and is structured. The server needs to use Spark to process the batch-structured work orders of the power information system, collect the work order data of the power information system, and write it into HBase.

[0025] In one example, after the server completes the collection of the infrastructure data, operation data, historical log data, and work order data of the power information system, it can perform heterogeneous data integration on the data stored in Elasticsearch, Kafka message queue, and HBase, and build a four-dimensional relationship model based on Neo4j to obtain a knowledge base of panoramic resources covering cloud platforms, servers, switches, network devices, storage arrays, middleware, databases, resource pools, and business applications. Among them, the four-dimensional relationships include the spatial relationships between entity components, the logical mapping relationships between services and services, the skill matrix and responsibility matrix of the operation and maintenance team, and the connection relationships between entity components and virtual components. The database established in this way can serve as the knowledge reserve support for the application technology, typical experience, and optimization practice of the power information system, providing a solid theoretical foundation for the training of intelligent agents and the execution of intelligent inspections and resource allocation by intelligent agents.

[0026] It is understandable that there are four key points in the construction of the knowledge base of the panoramic resources of the power information system. First, the integration of all factors, connecting the CMDB configuration library, monitoring platform, document knowledge base, etc., must be able to cover cloud platforms, servers (physical / virtual), switches, network devices, storage arrays, middleware, databases, resource pools and business applications. Second, dynamic link perception, through traffic monitoring and interface calls, real-time access to the operating status of the equipment, business traffic, log events and other dynamic data flows, to achieve "static ledger + real-time status" two-dimensional data fusion, timely update of the relationship. Third, knowledge relationship modeling, referring to people's social relationships. According to the comparison relationship between physical equipment, virtual resources, business systems, operation and maintenance teams, and historical events, the relationship topology of each operation and maintenance object is sorted out, and the penetrating query capability of "server, carrier system, dependent components, and responsible person" is established to present the "general outline" of information operation and maintenance. Fourth, a multi-modal interaction engine is built to support natural language query, voice command, visual map exploration and other interaction methods. The completed knowledge base supports basic queries, can achieve second-level response, and the query accuracy rate is as high as 95% or more.

[0027] Step 102: construct an intelligent agent for the power information system based on the large language model, and perform RAG on the intelligent agent based on the knowledge base and open source knowledge to enhance the performance of the intelligent agent.

[0028] In the specific implementation, after completing the construction of the knowledge base, the server needs to build an intelligent agent for the power information system based on the large language model, and perform retrieval enhancement generation on the intelligent agent based on the knowledge base and open source knowledge, thereby enhancing the performance of the intelligent agent. The application of RAG technology introduces external open source knowledge, which can prevent the intelligent agent from hallucinating, and at the same time greatly enhances the timeliness and interpretability of the intelligent agent's task execution.

[0029] In one example, the predecessor of the intelligent agent is a large language model. The server can choose any one of the Guangming Electric Power Large Model, Hunyuan Large Model, Xinghuo Large Model, Pangu Large Model, and Tiangong Large Model to build an intelligent agent for the power information system. It should be noted that the Guangming Electric Power Large Model is specially designed for the power industry, so it is best to use the Guangming Electric Power Large Model as the basis for building an intelligent agent for the power information system.

[0030] In one example, the server performs RAG on the agent based on the knowledge base and open-source knowledge. First, it is necessary to obtain open-source knowledge related to the power information system, which includes but is not limited to fault reports, accident case reviews, warning measures, regulations, professional books, expert experience, etc. Perform OCR recognition on the obtained open-source knowledge and form an open-source library based on the recognized data. Next, RAG can be performed on the agent based on the knowledge base and the open-source library, enhancing the vertical private capabilities of the agent through the knowledge base and enhancing the parallel common capabilities of the agent through the open-source library.

[0031] It should also be noted that during the process of the server performing RAG on the agent, it is necessary to reorder the data in the knowledge base and the open-source library in terms of vectors, bytes, paragraphs, and chapters to better adjust the agent.

[0032] Step 103, obtain the inspection requirements in natural language form. The agent automatically generates an inspection script based on the inspection requirements, compiles the inspection script, and generates an inspection task corresponding to the inspection requirements.

[0033] In a specific implementation, after the knowledge base and the agent are constructed, the power information system already has the conditions for intelligent inspection and resource allocation. Next, the agent will play a role. The input of the agent is the inspection requirements in natural language form, which are manually input by users, the operation and maintenance team, etc. After the agent obtains the inspection requirements, it can automatically generate an inspection script based on the inspection requirements, compile the inspection script, and generate an inspection task corresponding to the inspection requirements.

[0034] In one example, if the format of the inspection requirements in natural language form is text, the inspection requirements can be directly input into the agent. If the format of the inspection requirements in natural language form is voice, it needs to be converted into text before being input into the agent.

[0035] In one example, the constructed agent needs to be deployed using a large language model deployment framework, and a natural language to script compiler needs to be configured for the deployed agent. The large language model deployment framework selected in this embodiment can be any one of vLLM, Ollama, Xinference, and LM Studio, and the natural language to script compiler selected can be any one of Cline, Continue, Roo Code, and Trae.

[0036] In one example, after the inspection requirements are input into the deployed agent, the agent can automatically generate an inspection script using the natural language to script compiler and compile it, determine the inspection address corresponding to the inspection requirements, and then use the workflow technology to establish a workflow of the inspection task corresponding to the inspection requirements based on the inspection address.

[0037] In one example, the workflow of the patrol inspection task can be as Figure 2 shown.

[0038] In one example, the large language model deployment framework supports real-time display of the analysis process of the patrol inspection requirements by the agent, as well as the workflow of the patrol inspection task corresponding to the patrol inspection requirements. The analysis process of the patrol inspection requirements by the agent, as well as the workflow of the patrol inspection task corresponding to the patrol inspection requirements, can be displayed in real time through the large language model deployment framework, facilitating monitoring and learning by the operation and maintenance personnel.

[0039] In one example, the analysis of the patrol inspection requirements by the agent can be as Figure 3 shown.

[0040] In one example, the patrol inspection tasks can be divided into switch patrol inspection, cloud platform patrol inspection, and resource pool patrol inspection according to the patrol inspection direction. One patrol inspection task contains several patrol inspection items.

[0041] In one example, the patrol inspection items for switch patrol inspection include checking the utilization rates of the CPU and memory of the switch, the voltage and current of the power supply module, the working status of the fan, the link aggregation status, the stacking status, the number of VLANs (analyzing whether the number of VLANs changes), network connectivity (whether there are packet losses, average delay, and optical attenuation value of the switch link when pinging), the status of the service board, the alarm record, the status of the physical interface, the temperature, and the running time.

[0042] In one example, the patrol inspection items for cloud platform patrol inspection include checking the utilization rates of the CPU, memory, and disk of the physical hosts of the cloud platform, the voltage and current of the power supply module of the physical hosts of the cloud platform, the working status of the fan, the storage status of OBS, the utilization rate of OBS, the utilization rates of the CPU, memory, and disk of the CCE containers, the utilization rates of the CPU, memory, and disk of the virtual machines, the instance status and disk utilization rates of the RDS, DSC, and Gauss databases on the cloud, the disk utilization rate of the MRS big data cluster, the usage rate and remaining available capacity of the storage pools in 7 AZ partitions, and the availability of the operation-side cloud service interface.

[0043] In one example, the patrol inspection items for resource pool patrol inspection include checking the utilization rates of the CPU, memory, and disk of the physical hosts of the resource pool, the voltage and current of the power supply module, the working status of the fan, the status of the network card, the utilization rates of the CPU, memory, and disk of the virtual machines, and the IP address, CPU model, number of CPU cores, and number of threads of the physical hosts.

[0044] Step 104: The agent automatically executes the patrol inspection task, compares the inspection results with the knowledge base, and generates an inspection report.

[0045] In a specific implementation, after the intelligent agent generates an inspection task corresponding to the inspection requirements, it can automatically execute the inspection task and compare the inspection results with the knowledge base to generate an inspection report.

[0046] In one example, the intelligent agent needs to automatically execute each inspection item in the inspection task, obtain the inspection results of each inspection item, then search for the corresponding standard metrics of each inspection item in the knowledge base, compare the inspection results of each inspection item with the corresponding standard metrics in turn, determine the inspection items whose inspection results meet the corresponding standard metrics as normal items, and determine the inspection items whose inspection results do not meet the corresponding standard metrics as abnormal items. Finally, it is necessary to generate an inspection report that supports online viewing and downloading based on the inspection results and determination results of each inspection item. Judging the inspection results of each inspection item according to the standard metrics in the knowledge base can scientifically and accurately discover potential risks existing in the power information system.

[0047] Taking the inspection item of checking the CPU and memory utilization rates of the switch as an example, the standard metric for this inspection item is below 70%. If the inspection result of this inspection item is 85%, then this inspection item needs to be determined as an abnormal item.

[0048] In one example, the inspection report automatically generated by the intelligent agent can be as Figure 4 、 Figure 5 shown, Figure 4 showing the upper part of the inspection report, Figure 5 showing the lower part of the inspection report.

[0049] Step 105, the intelligent agent intelligently analyzes the inspection report and combines the knowledge base to allocate resources for the abnormal items in the inspection report to resolve the abnormal items.

[0050] In a specific implementation, after the intelligent agent generates the inspection report, it also needs to intelligently analyze the inspection report and combine the knowledge base to allocate resources for the abnormal items in the inspection report to resolve the abnormal items.

[0051] In one example, the intelligent agent intelligently analyzes the inspection report, analyzes each abnormal item in turn, and determines the simple abnormal items that can be automatically resolved and the complex abnormal items that cannot be automatically resolved. Subsequently, in combination with the knowledge base, resources are allocated for the abnormal items that can be automatically resolved to resolve the simple abnormal items. The resource allocation performed includes elastic scaling in the cloud, automatic expansion, automatic shutdown, automatic restart, etc. For the complex abnormal items that the intelligent agent cannot automatically resolve, relevant information needs to be sent to the operation and maintenance team to remind the operation and maintenance team to resolve the complex abnormal items.

[0052] It is understandable that intelligent inspection needs to balance automation and reliability. Therefore, the intelligent agent can determine abnormal items. For simple abnormal items, the intelligent agent automatically allocates resources. For complex abnormal items, the intelligent agent needs to promptly notify the operation and maintenance team for handling, so as to ensure that complex abnormal items can be quickly resolved.

[0053] A power information system inspection and resource allocation method based on a knowledge base and an intelligent agent proposed in this embodiment first collects data of the power information system to construct a knowledge base of the panoramic resources of the power information system. Subsequently, an intelligent agent for the power information system is constructed based on a large language model, and the intelligent agent is RAGed based on the knowledge base and open-source knowledge to enhance the performance of the intelligent agent. Next, inspection requirements in natural language form are obtained, and the intelligent agent automatically generates an inspection script based on the inspection requirements, compiles the inspection script, generates an inspection task and automatically executes the inspection task, compares the inspection results with the knowledge base to generate an inspection report, and then intelligently analyzes the inspection report, and allocates resources for abnormal items in the inspection report in combination with the knowledge base to resolve the abnormal items. This method builds a full-stack intelligent perception system for the power information system, realizes the full automation of the development of inspection scripts, shortens the development time to the minute level, the intelligent agent with RAG has stronger and more stable performance, allocates resources in combination with the power information system exclusive knowledge base, provides continuous enhancement guarantee for various services of the power information system, realizes the hour-level early warning of key system failures, and greatly improves the resource utilization rate, improves the operation and maintenance efficiency of the power information system, and reduces the operation and maintenance cost of the power information system, thereby providing a stable and good user experience for power generation enterprises and electricity users.

[0054] The step division of the above various methods is only for clear description. When implemented, they can be combined into one step, or some steps can be split into multiple steps. As long as the same logical relationship is included, it is within the protection scope of this application; adding insignificant modifications or introducing insignificant designs to the algorithm or process, but not changing the core design of its algorithm and process are all within the protection scope of this application.

[0055] Another embodiment of this application proposes a power information system inspection and resource allocation system based on a knowledge base and an intelligent agent, which is applicable to the operation and maintenance management of the power information system. The details of the power information system inspection and resource allocation system based on a knowledge base and an intelligent agent proposed in this embodiment will be specifically described below. The following content is only implementation details provided for convenient understanding and is not necessary for implementing this example.

[0056] Figure 6It is a schematic diagram of a power information system inspection and resource allocation system based on a knowledge base and an agent proposed in this embodiment, including: a knowledge base construction module 21, a knowledge base 22, an agent construction module 23, an agent 24, an interaction module 25, and a multi-dimensional display module 26.

[0057] The knowledge base construction module 21 is used to collect infrastructure data, operation data, log data, and work order data of the power information system 30, perform heterogeneous data integration, and construct a knowledge base 22 of the panoramic resources of the power information system 30.

[0058] The agent construction module 23 is used to construct an agent 24 for the power information system 30 based on a large language model, and perform RAG on the agent 24 based on the knowledge base 22 and open-source knowledge to enhance the performance of the agent 24.

[0059] The interaction module 25 is used to obtain inspection requirements in the form of natural language.

[0060] The agent 24 is used to automatically generate an inspection script based on the inspection requirements, compile the inspection script, generate an inspection task corresponding to the inspection requirements, automatically execute the inspection task, compare the inspection results with the knowledge base 22, generate an inspection report, perform intelligent analysis on the inspection report, and perform resource allocation for the abnormal items in the inspection report in combination with the knowledge base to solve the abnormal items.

[0061] The multi-dimensional display module 26 is used to perform real-time multi-dimensional display on the entire workflow of the agent 24.

[0062] It is worth mentioning that each module involved in this embodiment is a logical module. In actual application, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, units that are not closely related to solving the technical problems proposed in this application are not introduced in this embodiment, but this does not mean that there are no other units in this embodiment.

[0063] It is not difficult to find that this embodiment is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in cooperation with the above method embodiment. The relevant technical details and technical effects mentioned in the above method embodiment are still valid in this embodiment. To avoid repetition, they are not elaborated here. Correspondingly, the relevant technical details mentioned in this embodiment can also be applied to the above method embodiment.

[0064] Another embodiment of this application proposes an electronic device, and its specific structure can be as Figure 7As shown in the figure, it includes: at least one processor 401; and a memory 402 communicatively connected to the at least one processor 401; wherein, the memory 402 stores instructions executable by the at least one processor 401, and the instructions are executed by the at least one processor 401 to enable the at least one processor 401 to execute a power information system inspection and resource allocation method based on a knowledge base and an agent as described in each of the above method embodiments.

[0065] Among them, the memory and the processor are connected by a bus. The bus can include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and the memory together. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art, so they will not be further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be an element or multiple elements, such as multiple receivers and transmitters, and provides a unit for communicating with various other devices on the transmission medium.

[0066] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can be used to store data used by the processor when executing operations.

[0067] Another embodiment of the present application proposes a computer-readable storage medium storing a computer program, which when executed by a processor, can implement a power information system inspection and resource allocation method based on a knowledge base and an agent as described in the above method embodiments.

[0068] That is, those skilled in the art can understand that all or part of the steps in the above method embodiments can be completed by instructing relevant hardware through a program. The program is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: USB flash drives, mobile hard disks, ROM (Read-Only Memory), RAM (Random Access Memory), magnetic disks, or optical discs and other various media that can store program codes.

[0069] Those of ordinary skill in the art can understand that the above embodiments are specific embodiments for implementing the present application, and in practical applications, various changes can be made in form and details without departing from the spirit and scope of the present application.

Claims

1. A method for inspection and resource allocation of power information system based on knowledge base and intelligent agent, suitable for operation and maintenance management of power information system, characterized in that: The method comprises: Collect the infrastructure data, operation data, log data, and work order data of the power information system, and integrate heterogeneous data to build a knowledge base of panoramic resources of the power information system; Based on the large language model, an intelligent agent for the power information system is constructed, and RAG is performed on the intelligent agent based on the knowledge base and open source knowledge to enhance the performance of the intelligent agent; Obtain inspection requirements in natural language form, and the intelligent agent automatically generates inspection scripts based on the inspection requirements, compiles the inspection scripts, and generates inspection tasks corresponding to the inspection requirements; The intelligent agent automatically performs inspection tasks, compares the inspection results with the knowledge base, and generates inspection reports; The intelligent agent performs intelligent analysis on the inspection report and allocates resources based on the knowledge base to resolve the abnormal items in the inspection report.

2. According to claim 1, a method for power information system inspection and resource allocation based on knowledge base and intelligent agent, characterized in that: Collect the infrastructure data, operation data, historical log data, and work order data of the power information system, and integrate heterogeneous data to build a knowledge base of panoramic resources of the power information system, including: Use NLP technology to extract unstructured documents that store the infrastructure data of the power information system, collect the infrastructure data of the power information system, and store it in Elasticsearch; Use Flink to monitor the real-time business flow and log flow of the power information system, collect the operation data and log data of the power information system, and write them into the Kafka message queue; Use Spark to process the batch structured work orders of the power information system, collect the work order data of the power information system, and write it into HBase; Heterogeneous data integration is performed on the data stored in Elasticsearch, Kafka message queues, and HBase, and a four-dimensional relationship model is built based on Neo4j to obtain a knowledge base of panoramic resources covering cloud platforms, servers, switches, network devices, storage arrays, middleware, databases, resource pools, and business applications. The four-dimensional relationship includes the spatial relationship between physical components, the logical mapping relationship between businesses and services, the skill matrix and responsibility matrix of the operation and maintenance team, and the connection relationship between physical components and virtual components.

3. The method for power information system inspection and resource allocation based on knowledge base and intelligent agent according to claim 2 is characterized in that: The large language model is any one of the Guangming Electric Power Large Model, Hunyuan Large Model, Xinghuo Large Model, Pangu Large Model, and Tiangong Large Model. Based on the knowledge base and open source knowledge, RAG is performed on the intelligent agent to enhance the performance of the intelligent agent, including: Acquire open source knowledge related to power information systems, including but not limited to fault reports, accident case reviews, early warning measures, rules and regulations, professional books, and expert experience; Perform OCR recognition on the acquired open source knowledge and form an open source library based on the recognized data; Conduct RAG on the intelligent agent based on the knowledge base and open source library, enhance the vertical private capabilities of the intelligent agent through the knowledge base, and enhance the parallel shared capabilities of the intelligent agent through the open source library; In the process of performing RAG on the intelligent agent, it is necessary to reorder the data in the knowledge base and open source library in terms of vectors, bytes, paragraphs, and chapters.

4. A method for power information system inspection and resource allocation based on knowledge base and intelligent agent according to claim 3, characterized in that: If the inspection requirement in natural language is in text format, the inspection requirement is directly input into the intelligent agent. If the inspection requirement in natural language is in voice format, it is converted into text and input into the intelligent agent. The intelligent agent automatically generates an inspection script based on the inspection requirement, compiles the inspection script, and generates an inspection task corresponding to the inspection requirement, including: Use the large language model deployment framework to deploy the agent and configure the natural language to script compiler for the deployed agent. The large language model deployment framework is any one of vLLM, Ollama, Xinference, and LM Studio, and the natural language to script compiler is any one of Cline, Continue, Roo Code, and Trae. The inspection requirements are input into the deployed intelligent agent, which automatically generates and compiles the inspection script using the natural language to script compiler, determines the inspection address corresponding to the inspection requirements, and then uses the workflow technology to establish the inspection task workflow corresponding to the inspection requirements based on the inspection address. Among them, the large language model deployment framework supports the real-time display of the intelligent agent's analysis process of inspection needs, as well as the establishment of a workflow of inspection tasks corresponding to the inspection needs.

5. A method for power information system inspection and resource allocation based on knowledge base and intelligent agent according to any one of claims 1 to 4, characterized in that: Inspection tasks are divided into switch inspection, cloud platform inspection, and resource pool inspection according to the inspection direction. One inspection task contains several inspection items. The switch inspection items include checking the CPU and memory utilization of the switch, the voltage and current of the power module, the working status of the fan, the link aggregation status, the stacking status, the number of VLANs, the network connectivity, the service card status, the alarm record, the physical interface status, the temperature, and the running time; The cloud platform inspection includes checking the CPU, memory, and disk utilization of the cloud platform's physical hosts, the voltage and current of the power modules, the working status of the fans, the storage status of OBS, the utilization of OBS, the CPU, memory, and disk utilization of CCE containers, the CPU, memory, and disk utilization of virtual machines, the instance status and disk utilization of the RDS, DSC, and Gauss databases on the cloud, the disk utilization of the MRS big data cluster, the storage pool utilization and remaining available capacity of the seven AZ partitions, and the availability of the cloud service interface on the operation side. The inspection items of resource pool inspection include checking the CPU, memory and disk utilization of the physical host of the resource pool, the voltage and current of the power module, the working status of the fan, the status of the network card, the CPU, memory and disk utilization of the virtual machine, and the IP address, CPU model, number of CPU cores and number of threads of the physical host.

6. A method for power information system inspection and resource allocation based on knowledge base and intelligent agent according to claim 5, characterized in that: The intelligent agent automatically performs inspection tasks, compares the inspection results with the knowledge base, and generates inspection reports, including: The intelligent agent automatically executes each inspection item in the inspection task and obtains the inspection results of each inspection item; Search the knowledge base for the standard indicators corresponding to each inspection item, compare the inspection results of each inspection item with the corresponding standard indicators in turn, determine the inspection items whose inspection results meet the corresponding standard indicators as normal items, and determine the inspection items whose inspection results do not meet the corresponding standard indicators as abnormal items; Based on the inspection results and judgment results of each inspection item, an inspection report is generated that supports online viewing and downloading.

7. A method for power information system inspection and resource allocation based on knowledge base and intelligent agent according to claim 6, characterized in that: The intelligent agent performs intelligent analysis on the inspection report and allocates resources based on the knowledge base to solve the abnormal items in the inspection report, including: The intelligent agent performs intelligent analysis on the inspection report, analyzes each abnormal item in turn, and determines the simple abnormal items that can be solved automatically and the complex abnormal items that cannot be solved automatically; Combined with the knowledge base, resources are allocated for abnormal items that can be automatically resolved to resolve simple abnormal items. Resource allocation includes cloud elastic scaling, automatic capacity expansion, automatic shutdown, and automatic restart. Send relevant information of complex exceptions to the operation and maintenance team to remind the operation and maintenance team to solve the complex exceptions.

8. A power information system inspection and resource allocation system based on knowledge base and intelligent agent, suitable for operation and maintenance management of power information system, characterized in that: The system comprises: The knowledge base construction module is used to collect the infrastructure data, operation data, log data, and work order data of the power information system, and integrate heterogeneous data to build a knowledge base of panoramic resources of the power information system; The intelligent agent building module is used to build an intelligent agent for the power information system based on a large language model, and to perform RAG on the intelligent agent based on the knowledge base and open source knowledge to enhance the performance of the intelligent agent; Interaction module, used to obtain inspection requirements in natural language form; Intelligent agent, used to automatically generate inspection scripts based on inspection requirements, compile inspection scripts, generate inspection tasks corresponding to inspection requirements, automatically execute inspection tasks, compare inspection results with knowledge base, generate inspection reports, and intelligently analyze inspection reports, and allocate resources for abnormal items in inspection reports in combination with knowledge base to resolve abnormal items; The multi-dimensional display module is used to display the entire workflow of the intelligent agent in real time and in multiple dimensions.

9. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, 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 so that the at least one processor can execute a power information system inspection and resource allocation method based on a knowledge base and an intelligent agent as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement a power information system inspection and resource allocation method based on a knowledge base and an intelligent agent as described in any one of claims 1 to 7.

Images