Brain-like agent-based electric power panoramic inspection report generation method and related device

Abstract

The invention relates to the field of power and the technical field of computers, and discloses a brain-like intelligent agent-based power panoramic inspection report generation method and a related device. The power panoramic inspection report generation method comprises the following steps: acquiring a service instruction and a configuration file for generating a power panoramic inspection report; based on the service instruction and the configuration file, using a brain-like agent to generate a report, and obtaining an electric power panoramic inspection report; the brain-like agent comprises a brain-like intelligent scheduling module, an environment sensing module, a brain-like sensing module, a brain-like memory module and a brain-like world model module. According to the technical scheme provided by the invention, the labor cost can be reduced, the working efficiency can be improved, the report generation period can be shortened, the influence and interference of human factors can be remarkably reduced, and the accuracy and consistency of the electric power panoramic inspection report can be ensured.

Description

Technical Field

[0001] The present invention relates to the fields of electric power and computer technology, and in particular to a method for generating an electric power panoramic inspection report based on a brain-like intelligent agent and a related device. Background Art

[0002] In the existing practices of power inspection, operation, inspection and control, traditional methods are often confined to limited models, that is, they tend to train single scenario models with relatively narrow knowledge in a relatively isolated environment; the above-mentioned limited models have the defects of high labor costs and low work efficiency when it is necessary to obtain a comprehensive and all-round inspection, operation, inspection and control results. To explain in more detail, limited by the above-mentioned limited model, when a comprehensive and all-round inspection, operation and inspection, and regulation result is needed, each department first needs to extract key information data from its own environmental perception system, and input the extracted key information data into the established model one by one. The model will calculate, analyze and generate corresponding output results based on the input key information data; then the generated output results need to be summarized and integrated to form inspection, operation and inspection, and regulation reports for local areas, and report to the superior departments step by step; the above process is not only cumbersome, but also inefficient. After receiving the local inspection, operation and inspection, and regulation reports from various sub-departments, the superior departments in the middle stage need to spend a lot of time summarizing and integrating the reports to form higher-level inspection, operation and inspection, and regulation reports and continue to report and submit them upward. This progressive and step-by-step report summary method consumes a lot of human resources and prolongs the cycle from data collection to the final generation of power panoramic inspection, equipment operation and inspection, and power regulation reports.

[0003] In summary, a new solution for generating power inspection, operation and control reports is urgently needed to solve the technical problems of high labor cost, low work efficiency and long report generation cycle in existing technical solutions. Summary of the invention

[0004] The purpose of the present invention is to provide a method and related device for generating a power panoramic inspection report based on a brain-like intelligent agent to solve one or more of the above-mentioned technical problems. The technical solution provided by the present invention can reduce labor costs, improve work efficiency, shorten the report generation cycle, significantly reduce the influence and interference of human factors, and ensure the accuracy and consistency of the power panoramic inspection report.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for generating a power panoramic inspection report based on a brain-like intelligent agent, comprising the following steps:

[0007] Obtaining business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include a time period and a business scenario, and the configuration file at least includes a report generation style and a role definition;

[0008] Based on the business instructions and the configuration file, a brain-like intelligent agent is used to generate a report to obtain a power panoramic inspection report;

[0009] in,

[0010] The brain-like intelligent agent includes: a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module; wherein,

[0011] The brain-like intelligent scheduling module is used to understand the business instructions through the large language model after obtaining the business instructions and output local operation instructions; the local operation instructions are used to publish to other modules of the brain-like intelligent body and control the execution of other modules;

[0012] The environment perception module is used to collect and obtain environment perception information through a preset terminal based on local operation instructions;

[0013] The brain-like perception module is used to process and extract environmental perception information through multiple model prediction units based on local operation instructions to obtain scene recognition results;

[0014] The brain-like memory module is used to retrieve historical data related to the business instruction based on the local operation instruction;

[0015] The brain-like world model module is used to generate a report based on local operation instructions, the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and output a power panoramic inspection report.

[0016] A further improvement of the present invention is that

[0017] The brain-like intelligent scheduling module includes: a control unit, a first-level large language model and one or more second-level large language models; wherein,

[0018] The first-level large language model is used to receive the first-level business instructions for the locations covered by the national power grid and generate global operation instructions; the second-level large language model is used to receive the second-level business instructions for the locations covered by the provincial power grid or the global operation instructions generated by the first-level large language model and generate one or more local operation instructions;

[0019] The control unit is used to control the execution of each module in the brain-like intelligent agent based on local operation instructions.

[0020] A further improvement of the present invention is that

[0021] The business scenarios include: panoramic inspection, equipment inspection and power control;

[0022] The preset terminals in the environmental perception module include: equipment operation and maintenance data acquisition equipment, panoramic patrol data acquisition equipment and power control data acquisition equipment; wherein, the equipment operation and maintenance data acquisition equipment includes a chromatographic sensor, a voiceprint sensor, a partial discharge sensor and a ground current sensor, the panoramic patrol data acquisition equipment includes a visual data acquisition equipment and an audio data acquisition equipment, and the power control data acquisition equipment includes a meteorological sensor, a load output sensor, a data acquisition and monitoring sensor, a synchronous phasor measurement sensor and a log collection sensor.

[0023] A further improvement of the present invention is that

[0024] The brain-like perception module includes: an equipment operation and inspection model cluster, a panoramic inspection model cluster and a power regulation model cluster; wherein the equipment operation and inspection model cluster includes an equipment health status assessment model prediction unit, an equipment fault diagnosis and identification model prediction unit, an equipment fault location tracing model prediction unit and an equipment health status prediction model prediction unit; the panoramic inspection model cluster includes a target pedestrian re-identification model prediction unit, an image target detection model prediction unit, a video behavior detection model prediction unit and a target tracking model prediction unit; the power regulation model cluster includes a source-load output prediction model prediction unit, a safety boundary identification model prediction unit, a source-grid-load-storage optimization model prediction unit and a digital simulation deduction model prediction unit;

[0025] Among them, the brain-like intelligent scheduling module starts the model prediction unit that needs to be run in the brain-like perception module according to the local operation instruction, and controls the model prediction unit to stop running after the execution is completed.

[0026] A further improvement of the present invention is that

[0027] In the brain-like world model module, the steps of executing local operation instructions, generating a report according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and outputting a power panoramic inspection report are specifically as follows:

[0028] Based on the local operation instructions, the report is generated by the power graphic model according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition result, and the power panoramic inspection report is output;

[0029] The power graphic and text large model uses a large language model, with text and multimedia data addresses as input and the report text part and multimedia example addresses as output.

[0030] A further improvement of the present invention is that

[0031] The brain-like memory module is further used to package and store the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the obtained power panoramic inspection report in the brain-like memory module after obtaining the power panoramic inspection report;

[0032] in,

[0033] A key-value pair storage method is adopted when packaging and storing, with the embedded vector obtained by converting the business instruction as the key of the record, and the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the original information of the obtained power panoramic inspection report as the value of the record;

[0034] The extraction equation of stored information is expressed as,

[0035] In the formula, m * is the most suitable key obtained for the query; m is the key of the stored historical data; q represents the query; M represents the set of all memories; d rec (·) is used to measure the recency of memory, d rel (·) is used to measure the relevance of memory, d imp (·) is used to measure the importance of memory; α, β, and γ are all balance parameters.

[0036] A further improvement of the present invention is that

[0037] The brain-like intelligent agent also includes:

[0038] A self-criticism module is used to calculate the loss function based on the evaluation results, and optimize and update the brain-like world model module in the brain-like agent according to the calculation results of the loss function;

[0039] in,

[0040] The evaluation result is an evaluation result obtained by evaluating the generated power panoramic inspection report according to the local operation instruction or an evaluation result inputted externally by the user; the evaluation indicators of the evaluation result include data source traceability, data accuracy, report structure clarity, report logic, report readability, report comprehensiveness and report safety compliance;

[0041] The self-criticism module adopts a large language model;

[0042] The loss function is expressed as,

[0043] In the formula, wi is the weight of the i-th evaluation index; s i is the score of the i-th evaluation indicator.

[0044] In a second aspect, the present invention provides a system for generating a power panoramic inspection report based on a brain-like intelligent agent, comprising:

[0045] A data acquisition module, used to acquire business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include a time period and a business scenario, and the configuration file at least includes a report generation style and a role definition;

[0046] A report generation module, used to generate a report based on the business instructions and the configuration file using a brain-like intelligent agent to obtain a power panoramic inspection report;

[0047] in,

[0048] The brain-like intelligent agent includes: a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module; wherein,

[0049] The brain-like intelligent scheduling module is used to understand the business instructions through the large language model after obtaining the business instructions and output local operation instructions; the local operation instructions are used to publish to other modules of the brain-like intelligent body and control the execution of other modules;

[0050] The environment perception module is used to collect and obtain environment perception information through a preset terminal based on local operation instructions;

[0051] The brain-like perception module is used to process and extract environmental perception information through multiple model prediction units based on local operation instructions to obtain scene recognition results;

[0052] The brain-like memory module is used to retrieve historical data related to the business instruction based on the local operation instruction;

[0053] The brain-like world model module is used to generate a report based on local operation instructions, the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and output a power panoramic inspection report.

[0054] A further improvement of the present invention is that

[0055] The brain-like intelligent scheduling module includes: a control unit, a first-level large language model and one or more second-level large language models; wherein,

[0056] The first-level large language model is used to receive the first-level business instructions for the locations covered by the national power grid and generate global operation instructions; the second-level large language model is used to receive the second-level business instructions for the locations covered by the provincial power grid or the global operation instructions generated by the first-level large language model and generate one or more local operation instructions;

[0057] The control unit is used to control the execution of each module in the brain-like intelligent agent based on local operation instructions.

[0058] A further improvement of the present invention is that

[0059] The business scenarios include: panoramic inspection, equipment inspection and power control;

[0060] The preset terminals in the environmental perception module include: equipment operation and maintenance data acquisition equipment, panoramic patrol data acquisition equipment and power control data acquisition equipment; wherein, the equipment operation and maintenance data acquisition equipment includes a chromatographic sensor, a voiceprint sensor, a partial discharge sensor and a ground current sensor, the panoramic patrol data acquisition equipment includes a visual data acquisition equipment and an audio data acquisition equipment, and the power control data acquisition equipment includes a meteorological sensor, a load output sensor, a data acquisition and monitoring sensor, a synchronous phasor measurement sensor and a log collection sensor.

[0061] A further improvement of the present invention is that

[0062] The brain-like perception module includes: an equipment operation and inspection model cluster, a panoramic inspection model cluster and a power regulation model cluster; wherein the equipment operation and inspection model cluster includes an equipment health status assessment model prediction unit, an equipment fault diagnosis and identification model prediction unit, an equipment fault location tracing model prediction unit and an equipment health status prediction model prediction unit; the panoramic inspection model cluster includes a target pedestrian re-identification model prediction unit, an image target detection model prediction unit, a video behavior detection model prediction unit and a target tracking model prediction unit; the power regulation model cluster includes a source-load output prediction model prediction unit, a safety boundary identification model prediction unit, a source-grid-load-storage optimization model prediction unit and a digital simulation deduction model prediction unit;

[0063] Among them, the brain-like intelligent scheduling module starts the model prediction unit that needs to be run in the brain-like perception module according to the local operation instruction, and controls the model prediction unit to stop running after the execution is completed.

[0064] A further improvement of the present invention is that

[0065] In the brain-like world model module, the steps of executing local operation instructions, generating a report according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and outputting a power panoramic inspection report are specifically as follows:

[0066] Based on the local operation instructions, the report is generated by the power graphic model according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition result, and the power panoramic inspection report is output;

[0067] The power graphic and text large model uses a large language model, with text and multimedia data addresses as input and the report text part and multimedia example addresses as output.

[0068] A further improvement of the present invention is that

[0069] The brain-like memory module is further used to package and store the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the obtained power panoramic inspection report in the brain-like memory module after obtaining the power panoramic inspection report;

[0070] in,

[0071] A key-value pair storage method is adopted when packaging and storing, with the embedded vector obtained by converting the business instruction as the key of the record, and the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the original information of the obtained power panoramic inspection report as the value of the record;

[0072] The extraction equation of stored information is expressed as,

[0073] In the formula, m * is the most suitable key obtained for the query; m is the key of the stored historical data; q represents the query; M represents the set of all memories; d rec (·) is used to measure the recency of memory, d rel (·) is used to measure the relevance of memory, d imp (·) is used to measure the importance of memory; α, β, and γ are all balance parameters.

[0074] A further improvement of the present invention is that

[0075] The brain-like intelligent agent also includes:

[0076] A self-criticism module is used to calculate the loss function based on the evaluation results, and optimize and update the brain-like world model module in the brain-like agent according to the calculation results of the loss function;

[0077] in,

[0078] The evaluation result is an evaluation result obtained by evaluating the generated power panoramic inspection report according to the local operation instruction or an evaluation result inputted externally by the user; the evaluation indicators of the evaluation result include data source traceability, data accuracy, report structure clarity, report logic, report readability, report comprehensiveness and report safety compliance;

[0079] The self-criticism module adopts a large language model;

[0080] The loss function is expressed as,

[0081] In the formula, w i is the weight of the i-th evaluation index; s i is the score of the i-th evaluation indicator.

[0082] According to a third aspect of the present invention, there is provided an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a method for generating a power panoramic inspection report based on a brain-like intelligent agent as described in any one of the first aspect of the present invention.

[0083] According to a fourth aspect of the present invention, a non-transitory computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for generating a power panoramic inspection report based on a brain-like intelligent agent as described in any one of the first aspect of the present invention is implemented.

[0084] Compared with the prior art, the present invention has the following beneficial effects:

[0085] In view of the problem that the process of generating panoramic inspection reports in traditional power inspection, operation and control scenarios is cumbersome, long and labor-intensive, the present invention discloses a method for generating power panoramic inspection reports based on brain-like intelligent agents. The method is a full-process automatic panoramic inspection report generation method, which provides the architecture of brain-like intelligent agents. The use of brain-like intelligent agents to generate power panoramic inspection reports can reduce labor costs, improve work efficiency, shorten report generation cycles, and significantly reduce the impact and interference of human factors, thereby ensuring the accuracy and consistency of power panoramic inspection reports.

[0086] In the present invention, an advanced large language model is used as the understanding and parsing unit, which can directly generate operating instructions based on human language, and is flexible and convenient to use. In the present invention, the brain-like intelligent agent can autonomously perceive the environment, make decisions and perform actions without continuous human intervention. This autonomy enables the brain-like intelligent agent to efficiently complete tasks without human supervision, improving the overall efficiency and reliability of the system. In the present invention, steps and structural designs similar to those used by the human brain to process problems are adopted. The brain-like intelligent agent optimizes its own behavior through continuous learning and experience accumulation, can recognize and adapt to changes in the environment, and shows greater adaptability and flexibility when facing new tasks or complex challenges. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below; obviously, the drawings described below are some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0088] Figure 1 It is a flowchart of a method for generating a power panoramic inspection report based on a brain-like intelligent agent in an embodiment of the present invention;

[0089] Figure 2 is a schematic diagram of the architecture of a brain-like intelligent agent in an embodiment of the present invention;

[0090] Figure 3 is a schematic diagram of the framework of a brain-like intelligent scheduling module in an embodiment of the present invention;

[0091] Figure 4 is a schematic diagram of a framework of an environment perception module in an embodiment of the present invention;

[0092] Figure 5 is a schematic diagram of a framework of a brain-like perception module in an embodiment of the present invention;

[0093] Figure 6 is a structural schematic diagram of an equipment operation and inspection model cluster in an embodiment of the present invention;

[0094] Figure 7 is a structural schematic diagram of a panoramic inspection model cluster in an embodiment of the present invention;

[0095] Figure 8 is a structural schematic diagram of a power control model cluster in an embodiment of the present invention;

[0096] Fig. 9 is a schematic diagram of the framework of a brain-like world model module in an embodiment of the present invention;

[0097] Fig.10is a schematic diagram of a framework of a brain-like memory module in an embodiment of the present invention;

[0098] Fig.11 It is a schematic diagram of a power panoramic inspection report generation system based on a brain-like intelligent agent in an embodiment of the present invention. DETAILED DESCRIPTION

[0099] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention; it is obvious that the described embodiments and technical solutions are only part of the embodiments of the present invention, not all of the embodiments.

[0100] All other embodiments obtained by those of ordinary skill in the art without creative work based on the technical solutions disclosed in the embodiments of the present invention belong to the scope of protection of the present invention. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0101] See also Figure 1 and Figure 2 In an embodiment of the present invention, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is provided, comprising the following steps:

[0102] Step 1, obtaining business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include time periods and business scenarios (such as panoramic inspections, equipment inspections, power control, and other business scenarios); the configuration files at least include report generation styles and role definitions (such as business experts, technical experts, concise and clear answer styles, detailed and specific answer styles, etc.);

[0103] Step 2: Based on the business instructions and the configuration file, a brain-like intelligent agent is used to generate a report to obtain a power panoramic inspection report;

[0104] in,

[0105] The brain-like intelligent agent includes a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module;

[0106] The brain-like intelligent scheduling module is used to understand the business instructions through the large language model after obtaining the business instructions, and output local operation instructions; interpretatively, the brain-like intelligent scheduling module is used to imitate the decision-making and planning process of the decision-making unit of the human cerebral cortex on the perceived environment;

[0107] The environment perception module is used to collect and obtain environment perception information through a preset terminal based on local operation instructions; explanatory, the environment perception module is used to imitate the process of human senses (explanatory, such as vision, hearing, etc.) perceiving external environment information;

[0108] The brain-like perception module is used to process and extract environmental perception information through multiple model prediction units based on local operation instructions to obtain scene recognition results; explanatory, the brain-like perception module is used to imitate the process of human perception cerebral cortex perception units performing preliminary processing and handling of external information;

[0109] The brain-like memory module is used to retrieve historical data related to the business instruction based on the local operation instruction; explanatory, the brain-like memory module is used to imitate the process of the hippocampus of the brain storing and retrieving past experience and knowledge;

[0110] The brain-like world model module is used to generate a report based on local operation instructions, according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and output a panoramic power inspection report (exemplarily, the report can be generated through a large graphic model based on a large language model); explanatory, the brain-like world model module is used to imitate the process of information analysis, integration, high-level cognition, and action planning by the prefrontal cortex layer of the human brain.

[0111] In the technical solution provided by the embodiment of the present invention, a brain-like intelligent agent is used to generate a power panoramic inspection report according to business instructions and configuration files, which solves the technical problems of the existing technical solutions that the power panoramic inspection report generation process is cumbersome, has a long cycle, and consumes a lot of manpower.

[0112] To further explain, in order to overcome the drawbacks of the existing technical solutions, the embodiments of the present invention specifically disclose a method for generating a panoramic power inspection report based on a brain-like intelligent agent, which fully utilizes the excellent language understanding ability and efficient generation ability of the large language model, and integrates the originally scattered and cumbersome report generation process (such as inspection, operation and inspection, and control report generation process) into an end-to-end automated process, which can realize real-time data collection, rapid analysis and intelligent report generation, greatly improves work efficiency, reduces labor costs, and shortens the report generation cycle.

[0113] Principle explanation, brain-like intelligent agent shows excellent autonomy and intelligent characteristics. It can be seamlessly integrated into various environmental perception systems of power inspection, operation and control, and can capture and extract key information in real time and automatically. With the help of advanced large language model technology, these massive data will be deeply analyzed and finely mined to reveal potential hidden dangers and abnormal situations. By using intelligent algorithms and efficient computing power, brain-like intelligent agents can not only quickly analyze complex data patterns, but also accurately judge the operating status of power facilities, and then generate comprehensive and accurate reports. These reports can not only cover the details of each inspection, operation and control area, but also provide targeted suggestions and solutions, which can provide valuable data support for decision-makers. Therefore, the technical solution disclosed in the embodiment of the present invention can not only greatly improve work efficiency, but also significantly reduce the influence and interference of human factors, and ensure the accuracy and consistency of reports. For the power industry, this efficient and accurate report generation method can provide strong support and guarantee for safe operation, help power companies to discover and solve problems in a timely manner, reduce potential safety risks, and ensure the stable operation of the power system.

[0114] In one embodiment of the present invention, after step 2 is completed, the method further includes:

[0115] Step 3: Package and store the configuration file, the business instruction, the environmental perception information, the scene recognition result and the obtained power panoramic inspection report in the brain-like memory module. Explanatoryally, by storing the generated report, the historical data of the brain-like memory module can be expanded to facilitate the generation of subsequent reports; further explanatoryly, the data packaged and stored in step 3 is the historical data that can be called in the subsequent report generation process.

[0116] See also Figure 3 In one embodiment of the present invention, the brain-like intelligent scheduling module includes a first-level large language model, one or more second-level large language models and a control unit; explanatory, the first-level large language model is also called the headquarters-side large language model, and the second-level large language model is also called the provincial-side large language model;

[0117] The first-level large language model is used to receive first-level business instructions (interpretatively, business instructions given by the headquarters, such as business instructions for locations covered by the national power grid) and generate global operation instructions; the second-level large language model is used to receive second-level business instructions (interpretatively, business instructions given by the provincial side, such as business instructions for locations covered by the power grid within the province) or global operation instructions and generate one or more local operation instructions; explanatoryly, local operation instructions are used to instruct the execution process of the panoramic power inspection.

[0118] The control unit is used to control the execution of each module in the brain-like intelligent agent based on local operation instructions; specifically, when the input is a first-level business instruction, the first-level large language model is controlled to receive the first-level business instruction and output the global operation instruction, and the second-level large language model is controlled to receive the global operation instruction and output the local operation instruction; when the input is a second-level business instruction, the second-level large language model is controlled to receive the second-level business instruction and output the local operation instruction; finally, the execution of each module in the brain-like intelligent agent is controlled based on the local operation instruction.

[0119] In the technical solution provided by the embodiment of the present invention, when faced with complex tasks, humans tend to break them down into simpler sub-tasks and solve them individually; the brain-like intelligent scheduling module is intended to give brain-like intelligent agents this human ability, thereby making the behavior of brain-like intelligent agents more reasonable, powerful and reliable.

[0120] See also Figure 4 In one embodiment of the present invention, the environmental perception module specifically includes a series of data acquisition devices and a data collection device; among the series of data acquisition devices, the equipment operation and inspection data acquisition device includes a chromatographic sensor, a voiceprint sensor, a partial discharge sensor, a ground current sensor, etc.; the panoramic inspection data acquisition device includes a visual data acquisition device (such as a camera, an inspection robot, an inspection drone, etc.) and an audio data acquisition device (such as an audio monitoring device, etc.); the power control data acquisition device includes a meteorological sensor, a load output sensor, a synchronous phasor measurement sensor, a data acquisition and monitoring sensor, and a log collection sensor provided by the equipment.

[0121] In the environmental perception module of the embodiment of the present invention, after the environmental data is collected, the collected environmental data is transmitted to the data collection device through the network for aggregation; the data collection device receives the collected data and records the source of the data, and can continuously send the data outward from different data interfaces according to the data type and device type.

[0122] See also Figures 5 to 8 In one embodiment of the present invention, the brain-like perception module is composed of a plurality of model clusters of preset scenarios, including an equipment operation and inspection model cluster, a panoramic inspection model cluster, and a power control model cluster; wherein,

[0123] The equipment operation and inspection model cluster includes an equipment health status assessment model prediction unit, an equipment fault diagnosis and identification model prediction unit, an equipment fault location tracing model prediction unit, and an equipment health status prediction model prediction unit; wherein each model prediction unit is further divided into a data receiving step, a data preprocessing step, a model prediction step, and a result post-processing step; illustratively, as Figure 6As shown; further exemplarily, the equipment health status assessment model prediction unit and the equipment fault diagnosis and identification model prediction unit can adopt the MAE model (masked autoencoder model) model, and the equipment fault location tracing model prediction unit and the equipment health status prediction model prediction unit can adopt the SwinTransformer model;

[0124] The panoramic inspection model cluster includes a target pedestrian re-identification (ReID) model prediction unit, an image target detection model prediction unit, a video behavior detection model prediction unit, a target tracking model prediction unit, etc.; wherein each model prediction unit is further divided into a data receiving step, a data preprocessing step, a model prediction step, and a result post-processing step; illustratively, as Figure 7 As shown; further exemplarily, the target pedestrian re-identification model prediction unit and the image target detection model prediction unit may adopt the Yolov8 model, and the video behavior detection model prediction unit and the target tracking model prediction unit may adopt the Yolov5 model;

[0125] The power control model cluster includes a source-load output prediction model prediction unit, a safety boundary identification model prediction unit, a source-grid-load-storage optimization model prediction unit, a digital simulation deduction model prediction unit, etc.; wherein each model prediction unit is further divided into a data receiving step, a data preprocessing step, a model prediction step, and a result post-processing step; illustratively, as Figure 8 As shown; further exemplarily, the source-load output prediction model prediction unit, the safety boundary identification model prediction unit, the source-grid-load-storage optimization model prediction unit, and the digital simulation deduction model prediction unit can all adopt the Spikformer model.

[0126] To further explain, each model prediction unit provides a callable interface to the outside world. The brain-like intelligent scheduling module starts the required model prediction unit according to the local operation instructions output by the secondary large language model, and controls the corresponding model prediction unit to stop running when the execution ends.

[0127] See also Fig. 9 In one embodiment of the present invention, the brain-like world model is composed of a large power graph and text model. The large power graph and text model generates a power panoramic inspection report and predicts the next action to be taken based on user needs and the results output by the brain-like perception module; wherein,

[0128] The control unit in the brain-like intelligent scheduling module first extracts historical data from the brain-like memory module according to the business instructions, including: converting the business instructions into embedding vectors using the Word2Vec (a model used to generate word vectors) method, and using the obtained embedding vectors to retrieve relevant information from the brain-like memory module; wherein, the historical information in the brain-like memory module is utilized in the form of a mixture of short-term memory and long-term memory. This method explicitly simulates human short-term and long-term memory. Short-term memory temporarily caches recent perceptions, while long-term memory consolidates important information over time; the information provided in the prompt is regarded as short-term memory, and the information in the brain-like memory module is regarded as long-term memory. By integrating short-term and long-term memory, the long-term reasoning ability and the ability to accumulate valuable experience of the intelligent body can be enhanced, which is crucial to completing tasks in complex environments; then the business instructions, global operation instructions, configuration files, local operation instructions, historical data, and brain-like perception results are input into the large electric power graphic model together to obtain the generated panoramic inspection, equipment operation and inspection, power regulation report, and the actions that may be taken next.

[0129] In the power graphic model of the embodiment of the present invention, the input form is text and multimedia data (picture / audio) address, and the output form is the report text part and multimedia (picture / audio) example address; the power graphic model learns the complex relationship between different modal data through deep learning technology, so as to have stronger cross-modal understanding ability. Exemplarily, using a large language model, the multimedia data storage location is used instead of the multimedia data as input, and the future goal is to replace it with input and output that can directly support multiple modalities such as text, image and audio.

[0130] Specifically, a business instruction is a command entered by a user, such as "Please help me generate a drone transmission channel inspection report from 2024 to date"; a global operation instruction is an operation instruction generated by a first-level large language model based on the business instruction, such as "According to the above command: Please help me generate [between January 1, 2024 and today] [Beijing, Tianjin, Hebei Province, northern Hebei, Shanxi Province, Shandong Province, Shanghai, Jiangsu Province, Zhejiang Province, Anhui Province, Fujian Province, Hubei Province, Hunan Province, Henan Province, Jiangxi Province, Liaoning Province, Jilin Province, Heilongjiang Province, eastern Inner Mongolia, Shaanxi Province, Gansu Province, Qinghai Province, Ningxia Hui Autonomous Region, Xinjiang, Sichuan Province, Chongqing City, Tibet] [UAV Transmission Channel Inspection Report]"; the configuration file is the role played by the user through language commands before the large language model is executed, such as "business expert", "technical expert", "concise and clear answer style", "detailed and specific answer style", etc. The large language model creates the corresponding configuration file according to the setting; the local operation instruction is the instruction obtained by parsing the global operation instruction or the secondary business instruction, for example, the local operation instruction generated according to the above global operation instruction is shown in Table 1.

[0131] Table 1. Local operation instructions

[0132]

[0133] See also Fig.10 In one embodiment of the present invention, the brain-like intelligent agent combines the principles and mechanisms of cognitive science to study the human memory process. Human memory follows an overall process, from sensory memory that records sensory input, to short-term memory that briefly maintains information, to long-term memory that consolidates information over a long period of time; in the technical solution of the embodiment of the present invention, when designing a brain-like memory module, short-term memory only includes input information within the context window, and long-term memory includes external vector storage. The brain-like memory module is used to store the historical data of the entire brain-like intelligent agent. The brain-like memory module plays a very important role in the design of the entire brain-like intelligent agent. It stores environmental perception information perceived from the environment and key information in the execution process, and uses recorded memory to promote future actions. The brain-like memory module can help the intelligent agent accumulate experience, evolve itself, and act in a more consistent, reasonable, and effective manner.

[0134] The brain-like memory module includes a brain-like memory data repository, a memory writing unit, a memory reading unit, and a memory reflection unit, and their execution is controlled by the control unit of the brain-like intelligent scheduling module. Among them, the brain-like memory data repository is used to store data generated during the operation of the panoramic inspection, equipment operation and inspection, and power control system; the purpose of the memory writing unit is to store information about the perceived environment, and storing valuable information in the memory provides a basis for future retrieval of the information memory, so that the agent can act more effectively and reasonably. Through the memory writing unit, the data generated during the operation of the system in the panoramic inspection, equipment operation and inspection, and power control scenarios are organized into a record, and each record is a key-value pair, which is convenient for efficient storage and retrieval; when storing, the business instructions are converted into an embedded vector using the Word2Vec method as the key of the record, and the original information such as business instructions, global operation instructions, configuration files, local operation instructions, brain-like perception results, and power (panoramic inspection, equipment operation and inspection, power control) reports are used as the value of the record. There are several advantages to using the original information, including: first, the memory information can be expressed in a flexible and understandable way; in addition, it retains rich semantic information and can provide comprehensive signals to guide subsequent behaviors.

[0135] To explain by way of specific example, during the memory writing process, the key (embedded representation) of the current record c is firstly similarly calculated to the key of the existing data in the brain-like memory data repository, and the current record and the records with higher similarity are grouped into a group; if there are no records with higher similarity, a separate group is created for the current record, and the current record is directly saved in it; if there is a record s with higher similarity, the number of records n in the group where s is located is determined, and if n is less than the upper limit of the group size N (exemplarily, N=5), the current record c is directly saved in the group where s is located; if n is equal to the upper limit of the group size N (N=5), a large language model is used to condense all sequences in the group into one sequence, the original sequence in the group is replaced by the newly generated record, and the current record is stored in the group. During the storage process, if a memory overflow problem is encountered, the oldest entry is overwritten in a first-in-first-out (FIFO) manner.

[0136] The purpose of memory reading is to extract meaningful information from memory to enhance the effect of brain-like world models. The key to the memory reading unit is how to extract valuable information from historical behaviors. Three criteria are used to measure, namely recency, relevance, and importance, so that the most suitable memory is more likely to be retrieved. The following equation for memory information extraction is used:

[0137]

[0138] In the formula, m *is the most suitable key obtained for the query; m is the key of the stored historical data; q represents the query, that is, the task being processed by the agent; M is the set of all memories; d rec (·) is used to measure the recency of memory, d rel (·) is used to measure the relevance of memory, d imp (·) is used to measure the importance of memory; α, β, γ are balance parameters;

[0139] These scoring functions can be implemented using various methods, based on the HNSW (Hierarchical-Navigable-Small-World approach);

[0140] The memory reflection unit mimics the human ability to witness and evaluate their own cognitive, emotional, and behavioral processes. The goal of the memory reflection unit is to provide the brain-like world model module with the ability to independently summarize and infer more abstract, complex, and high-level information, summarizing past experiences stored in memory into broader and more abstract insights. First, the power graph model generates n key questions based on its recent memory; then, these questions are used to query the memory to obtain relevant information. Based on the information obtained, the agent generates n insights that reflect the high-level ideas of the brain-like world model module.

[0141] In one embodiment of the present invention, in many real-world scenarios, it is extremely difficult for humans to generate a perfect plan directly from the beginning when solving complex tasks, because it requires considering various complex prerequisites. When studying how humans handle complex tasks, we found that individuals may repeatedly formulate and modify their plans based on self-reflection. In order to simulate this human ability, the brain-like intelligent agent in the method of the embodiment of the present invention performs self-evaluation after taking action.

[0142] See also Figure 2 In one embodiment of the present invention, the brain-like intelligent agent further comprises:

[0143] A self-criticism module is used to imitate the process of human self-criticism and self-reflection, calculate the loss function based on the evaluation results of the power panoramic inspection report generated according to the local operation instructions or the evaluation results input by the user, and optimize and update the brain-like world model module in the brain-like intelligent agent according to the calculation results of the loss function;

[0144] Among them, the self-criticism module adopts a large model of pictures and texts;

[0145] The evaluation indicators of the evaluation results include data source traceability, data accuracy, report structure clarity, report logic, report readability, report comprehensiveness and report security compliance; each evaluation indicator has a score of [0,1], 0 means very poor and 1 means very good; in the loss function, the score of each evaluation indicator and the full score of 1 are used to calculate the loss, and the sum is used as the final loss. The loss function is expressed as:

[0146]

[0147] In the formula, w i is the weight of the i-th evaluation index; s i is the score of the i-th evaluation indicator;

[0148] Among them, automatic updates are performed according to the loss function, and the model can continuously improve itself through self-evaluation.

[0149] Explanativily, the evaluation results used to calculate the loss function can be the evaluation results of external input by the user. The planning ability of the brain-like agent is enhanced through direct interaction with humans. External feedback is a subjective signal, which can effectively align the intelligent agent with human values ​​and preferences, and also help alleviate hallucinations. Among them, based on the power panoramic inspection report generated by the brain-like agent, the loss function is calculated after the user scores and obtains the evaluation results, and the brain-like world model module is updated to update and optimize the brain-like agent.

[0150] The technical solution disclosed in the embodiment of the present invention uses a brain-like intelligent agent to imitate the process of brain processing information, and completes the complete power panoramic inspection process from perception to planning and then to action; wherein, a large language model is used as the core unit of the brain-like intelligent scheduling module, so that users can use the entire brain-like intelligent agent in the form of conventional dialogue, and the operation is convenient and flexible; in the preferred scheme of the embodiment of the present invention, it can be continuously adjusted and updated according to the results of self-evaluation or user feedback, and can continuously adapt to user usage habits; the technical solution of the embodiment of the present invention adopts a modular design, which has the advantages of flexibility and convenience when adding modules or units in modules.

[0151] The following are device embodiments of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0152] See also Fig.11 In an embodiment of the present invention, a system for generating a power panoramic inspection report based on a brain-like intelligent agent is provided, comprising:

[0153] A data acquisition module, used to acquire business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include a time period and a business scenario, and the configuration file at least includes a report generation style and a role definition;

[0154] A report generation module, used to generate a report based on the business instructions and the configuration file using a brain-like intelligent agent to obtain a power panoramic inspection report;

[0155] in,

[0156] The brain-like intelligent agent includes: a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module; wherein,

[0157] The brain-like intelligent scheduling module is used to imitate the decision-making and planning process of the human cerebral cortex decision unit on the perceived environment. After obtaining the business instructions, the business instructions are understood through the large language model and local operation instructions are output; the local operation instructions are used to publish to other modules of the brain-like intelligent body and control the execution of other modules;

[0158] The environment perception module is used to imitate the process of human senses perceiving external environment information, and based on local operation instructions, collects and obtains environment perception information through a preset terminal;

[0159] The brain-like perception module is used to imitate the process of human cerebral cortex perception units performing preliminary processing and handling of external information, and based on local operation instructions, processes and extracts environmental perception information through multiple model prediction units to obtain scene recognition results;

[0160] The brain-like memory module is used to imitate the process of the human brain hippocampus storing and retrieving past experience and knowledge, and retrieve historical data related to the business instruction based on the local operation instruction;

[0161] The brain-like world model module is used to imitate the process of information analysis, integration, high-level cognition, and action planning by the prefrontal cortex layer of the human brain. Based on local operation instructions, it generates a report according to the configuration file, the business instructions, the historical data, the environmental perception information, and the scene recognition results, and outputs a power panoramic inspection report.

[0162] In one embodiment of the present invention, a computer device is provided, the computer device comprising a processor and a memory, the memory being used to store a computer program, the computer program comprising program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used to perform the operation of the method for generating a power panoramic inspection report based on a brain-like intelligent agent.

[0163] In one embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM (Random Access Memory) memory, or a non-volatile memory (non-volatile memory), such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the method for generating a power panoramic patrol report based on a brain-like intelligent agent in the above embodiment.

[0164] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) that contain computer-usable program code.

[0165] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0166] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0167] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for generating a power panoramic inspection report based on a brain-like intelligent agent, characterized in that: The following steps are involved: Obtaining business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include a time period and a business scenario, and the configuration file at least includes a report generation style and a role definition; Based on the business instructions and the configuration file, a brain-like intelligent agent is used to generate a report to obtain a power panoramic inspection report; in, The brain-like intelligent agent includes: a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module; wherein, The brain-like intelligent scheduling module is used to understand the business instructions through a large language model and output local operation instructions after obtaining the business instructions; The environment perception module is used to collect and obtain environment perception information through a preset terminal based on local operation instructions; The brain-like perception module is used to process and extract environmental perception information through multiple model prediction units based on local operation instructions to obtain scene recognition results; The brain-like memory module is used to retrieve historical data related to the business instruction based on the local operation instruction; The brain-like world model module is used to generate a report based on local operation instructions, the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and output a power panoramic inspection report.

2. According to claim 1, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: The brain-like intelligent scheduling module includes: a control unit, a first-level large language model and one or more second-level large language models; wherein, The first-level large language model is used to receive the first-level business instructions for the locations covered by the national power grid and generate global operation instructions; the second-level large language model is used to receive the second-level business instructions for the locations covered by the provincial power grid or the global operation instructions generated by the first-level large language model and generate one or more local operation instructions; The control unit is used to control the execution of each module in the brain-like intelligent agent based on local operation instructions.

3. According to claim 1, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: The business scenarios include: panoramic inspection, equipment inspection and power control; The preset terminals in the environmental perception module include: equipment operation and maintenance data acquisition equipment, panoramic patrol data acquisition equipment and power control data acquisition equipment; wherein, the equipment operation and maintenance data acquisition equipment includes a chromatographic sensor, a voiceprint sensor, a partial discharge sensor and a ground current sensor, the panoramic patrol data acquisition equipment includes a visual data acquisition equipment and an audio data acquisition equipment, and the power control data acquisition equipment includes a meteorological sensor, a load output sensor, a data acquisition and monitoring sensor, a synchronous phasor measurement sensor and a log collection sensor.

4. According to claim 3, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: The brain-like perception module includes: an equipment operation and inspection model cluster, a panoramic inspection model cluster and a power regulation model cluster; wherein the equipment operation and inspection model cluster includes an equipment health status assessment model prediction unit, an equipment fault diagnosis and identification model prediction unit, an equipment fault location tracing model prediction unit and an equipment health status prediction model prediction unit; the panoramic inspection model cluster includes a target pedestrian re-identification model prediction unit, an image target detection model prediction unit, a video behavior detection model prediction unit and a target tracking model prediction unit; the power regulation model cluster includes a source-load output prediction model prediction unit, a safety boundary identification model prediction unit, a source-grid-load-storage optimization model prediction unit and a digital simulation deduction model prediction unit; Among them, the brain-like intelligent scheduling module starts the model prediction unit that needs to be run in the brain-like perception module according to the local operation instruction, and controls the model prediction unit to stop running after the execution is completed.

5. According to claim 1, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: In the brain-like world model module, the steps of executing local operation instructions, generating a report according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and outputting a power panoramic inspection report are specifically as follows: Based on the local operation instructions, the report is generated by the power graphic model according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition result, and the power panoramic inspection report is output; The power graphic and text large model uses a large language model, with text and multimedia data addresses as input and the report text part and multimedia example addresses as output.

6. According to claim 1, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: The brain-like memory module is further used to package and store the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the obtained power panoramic inspection report in the brain-like memory module after obtaining the power panoramic inspection report; in, A key-value pair storage method is adopted when packaging and storing, with the embedded vector obtained by converting the business instruction as the key of the record, and the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the original information of the obtained power panoramic inspection report as the value of the record; The extraction equation of stored information is expressed as, In the formula, m * is the most suitable key obtained for the query; m is the key of the stored historical data; q represents the query; M represents the set of all memories; d rec (·) is used to measure the recency of memory, d rel (·) is used to measure the relevance of memory, d imp (·) is used to measure the importance of memory; α, β, and γ are all balance parameters.

7. According to claim 1, a method for generating a power panoramic inspection report based on a brain-like intelligent agent is characterized in that: The brain-like intelligent agent also includes: A self-criticism module is used to calculate the loss function based on the evaluation results, and optimize and update the brain-like world model module in the brain-like agent according to the calculation results of the loss function; in, The evaluation result is an evaluation result obtained by evaluating the generated power panoramic inspection report according to the local operation instruction or an evaluation result inputted externally by the user; the evaluation indicators of the evaluation result include data source traceability, data accuracy, report structure clarity, report logic, report readability, report comprehensiveness and report safety compliance; The self-criticism module adopts a large language model; The loss function is expressed as, In the formula, w i is the weight of the i-th evaluation index; s i is the score of the i-th evaluation indicator.

8. A power panoramic inspection report generation system based on brain-like intelligent agent, characterized in that: include: A data acquisition module, used to acquire business instructions and configuration files for generating a power panoramic inspection report; wherein the business instructions at least include a time period and a business scenario, and the configuration file at least includes a report generation style and a role definition; A report generation module, used to generate a report based on the business instructions and the configuration file using a brain-like intelligent agent to obtain a power panoramic inspection report; in, The brain-like intelligent agent includes: a brain-like intelligent scheduling module, an environment perception module, a brain-like perception module, a brain-like memory module and a brain-like world model module; wherein, The brain-like intelligent scheduling module is used to understand the business instructions through the large language model after obtaining the business instructions and output local operation instructions; the local operation instructions are used to publish to other modules of the brain-like intelligent body and control the execution of other modules; The environment perception module is used to collect and obtain environment perception information through a preset terminal based on local operation instructions; The brain-like perception module is used to process and extract environmental perception information through multiple model prediction units based on local operation instructions to obtain scene recognition results; The brain-like memory module is used to retrieve historical data related to the business instruction based on the local operation instruction; The brain-like world model module is used to generate a report based on local operation instructions, the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and output a power panoramic inspection report.

9. According to claim 8, a power panoramic inspection report generation system based on brain-like intelligent agent is characterized in that: The brain-like intelligent scheduling module includes: a control unit, a first-level large language model and one or more second-level large language models; wherein, The first-level large language model is used to receive the first-level business instructions for the locations covered by the national power grid and generate global operation instructions; the second-level large language model is used to receive the second-level business instructions for the locations covered by the provincial power grid or the global operation instructions generated by the first-level large language model and generate one or more local operation instructions; The control unit is used to control the execution of each module in the brain-like intelligent agent based on local operation instructions.

10. The power panoramic inspection report generation system based on brain-like intelligent agent according to claim 8 is characterized in that: The business scenarios include: panoramic inspection, equipment inspection and power control; The preset terminals in the environmental perception module include: equipment operation and maintenance data acquisition equipment, panoramic patrol data acquisition equipment and power control data acquisition equipment; wherein, the equipment operation and maintenance data acquisition equipment includes a chromatographic sensor, a voiceprint sensor, a partial discharge sensor and a ground current sensor, the panoramic patrol data acquisition equipment includes a visual data acquisition equipment and an audio data acquisition equipment, and the power control data acquisition equipment includes a meteorological sensor, a load output sensor, a data acquisition and monitoring sensor, a synchronous phasor measurement sensor and a log collection sensor.

11. The power panoramic inspection report generation system based on brain-like intelligent agent according to claim 10 is characterized in that: The brain-like perception module includes: an equipment operation and inspection model cluster, a panoramic inspection model cluster and a power regulation model cluster; wherein the equipment operation and inspection model cluster includes an equipment health status assessment model prediction unit, an equipment fault diagnosis and identification model prediction unit, an equipment fault location tracing model prediction unit and an equipment health status prediction model prediction unit; the panoramic inspection model cluster includes a target pedestrian re-identification model prediction unit, an image target detection model prediction unit, a video behavior detection model prediction unit and a target tracking model prediction unit; the power regulation model cluster includes a source-load output prediction model prediction unit, a safety boundary identification model prediction unit, a source-grid-load-storage optimization model prediction unit and a digital simulation deduction model prediction unit; Among them, the brain-like intelligent scheduling module starts the model prediction unit that needs to be run in the brain-like perception module according to the local operation instruction, and controls the model prediction unit to stop running after the execution is completed.

12. According to claim 8, a power panoramic inspection report generation system based on brain-like intelligent agent is characterized in that: In the brain-like world model module, the steps of executing local operation instructions, generating a report according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition results, and outputting a power panoramic inspection report are specifically as follows: Based on the local operation instructions, the report is generated by the power graphic model according to the configuration file, the business instructions, the historical data, the environmental perception information and the scene recognition result, and the power panoramic inspection report is output; The power graphic and text large model uses a large language model, with text and multimedia data addresses as input and the report text part and multimedia example addresses as output.

13. According to claim 8, a power panoramic inspection report generation system based on brain-like intelligent agent is characterized in that: The brain-like memory module is further used to package and store the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the obtained power panoramic inspection report in the brain-like memory module after obtaining the power panoramic inspection report; in, A key-value pair storage method is adopted when packaging and storing, with the embedded vector obtained by converting the business instruction as the key of the record, and the configuration file, the business instruction, the global operation instruction, the local operation instruction, the environmental perception information, the scene recognition result and the original information of the obtained power panoramic inspection report as the value of the record; The extraction equation of stored information is expressed as, In the formula, m * is the most suitable key obtained for the query; m is the key of the stored historical data; q represents the query; M represents the set of all memories; d rec (·) is used to measure the recency of memory, d rel (·) is used to measure the relevance of memory, d imp (·) is used to measure the importance of memory; α, β, and γ are all balance parameters.

14. The power panoramic inspection report generation system based on brain-like intelligent agent according to claim 8 is characterized in that: The brain-like intelligent agent also includes: A self-criticism module is used to calculate the loss function based on the evaluation results, and optimize and update the brain-like world model module in the brain-like agent according to the calculation results of the loss function; in, The evaluation result is an evaluation result obtained by evaluating the generated power panoramic inspection report according to the local operation instruction or an evaluation result inputted externally by the user; the evaluation indicators of the evaluation result include data source traceability, data accuracy, report structure clarity, report logic, report readability, report comprehensiveness and report safety compliance; The self-criticism module adopts a large language model; The loss function is expressed as, In the formula, w i is the weight of the i-th evaluation index; s i is the score of the i-th evaluation indicator.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the method for generating a power panoramic inspection report based on a brain-like intelligent agent as described in any one of claims 1 to 7.

16. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the method for generating a power panoramic inspection report based on a brain-like intelligent agent as described in any one of claims 1 to 7.

Images