Information generation method and device based on large model, intelligent agent and electronic equipment

Abstract

The invention provides an information generation method and device based on a large model, an intelligent agent and electronic equipment, and relates to the technical field of artificial intelligence, in particular to the technical field of large models and fault detection. According to the specific implementation scheme of the information generation method, a large model is used for analyzing operation state information and initial abnormal information of a power distribution network in the current time period, potential abnormal information associated with initial abnormal information is generated, and the initial abnormal information is obtained by detecting the operation state information through an abnormal detection model; and performing association analysis on the historical power restoration information and the potential anomaly information by using the large model, and generating target strategy information for eliminating potential anomaly.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, particularly to the fields of large models and fault detection technology, and specifically to an information generation method, device, intelligent agent, and electronic device based on large models. Background Art

[0002] With the in-depth application of artificial intelligence technology in the field of fault detection technology, not only the accurate positioning of fault points and the shortening of power outage time have been achieved, but also the collaborative scheduling of real-time load data has been realized, improving the operation efficiency of the power grid, etc. Summary of the Invention

[0003] The present disclosure provides an information generation method, device, intelligent agent, and electronic device based on large models.

[0004] According to one aspect of the present disclosure, an information generation method based on large models is provided, including: using a large model to analyze the operation state information and initial abnormal information of the distribution network in the current period to generate potential abnormal information associated with the initial abnormality, where the initial abnormal information is obtained by using an anomaly detection model to detect the operation state information; and using a large model to perform an association analysis on historical power repair information and potential abnormal information to generate target policy information for excluding potential abnormalities.

[0005] According to another aspect of the present disclosure, an information generation device based on large models is provided, including: an analysis module and a generation module.

[0006] The analysis module is configured to use a large model to analyze the operation state information and initial abnormal information of the distribution network in the current period to generate potential abnormal information associated with the initial abnormality, where the initial abnormal information is obtained by using an anomaly detection model to detect the operation state information.

[0007] The generation module is configured to use a large model to perform an association analysis on historical power repair information and potential abnormal information to generate target policy information for excluding potential abnormalities.

[0008] According to another aspect of the present disclosure, an intelligent agent for information generation is provided, including: an input module, a processing module, and an output module.

[0009] The input module is configured to receive input information. The processing module is configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and execute the information generation method described above by calling the large model to obtain output information. The output module is configured to output the output information obtained by the processing module.

[0010] According to another aspect of the present disclosure, there is provided an electronic device, including: 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 the method described above.

[0011] According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method described above.

[0012] According to another aspect of the present disclosure, there is provided a computer program product, including a computer program, where the computer program implements the method described above when executed by a processor.

[0013] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understandable through the following description. Description of the Drawings

[0014] The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them:

[0015] Figure 1 Schematically shown is an exemplary system architecture to which the information generation method and apparatus based on a large model according to an embodiment of the present disclosure can be applied;

[0016] Figure 2 Schematically shown is a flowchart of the information generation method based on a large model according to an embodiment of the present disclosure;

[0017] Figure 3 Schematically shown is a schematic diagram of the information generation method based on a large model according to an embodiment of the present disclosure;

[0018] Figure 4 Schematically shown is a schematic diagram of optimizing the large model output strategy based on feedback information according to an embodiment of the present disclosure;

[0019] Figure 5 Schematically shown is a schematic diagram of using the potential anomalies identified by the large model to retrain the anomaly detection model in reverse according to an embodiment of the present disclosure;

[0020] Figure 6 Schematically shown is a block diagram of the information generation apparatus based on a large model according to an embodiment of the present disclosure;

[0021] Figure 7 Schematically shown is a block diagram of an agent for information generation according to an embodiment of the present disclosure; and

[0022] Figure 8 Schematically shows a block diagram of an electronic device suitable for implementing a large model-based information generation method according to an embodiment of the present disclosure. Detailed implementation manners

[0023] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to assist understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0024] There are many types of distribution network faults. In related examples, a neural network model is usually trained by using known faults through methods such as machine learning and deep learning to detect distribution network faults. However, this method is limited to the fault types in the training samples, and the generalization of the model is relatively low.

[0025] A large language model (LLM) is a deep learning model trained using a large amount of text data. It can not only generate natural language text, but also deeply understand the meaning of natural language text and process various natural language tasks.

[0026] In view of this, the embodiments of the present disclosure utilize the semantic understanding ability of the large model to identify potential anomalies associated with the initial anomalies detected by the anomaly detection model from the operation state information of the distribution network. Since the potential anomalies are obtained on the basis of the large model's full analysis of the initial anomaly information and operation state information, the accuracy of the recognition result can be guaranteed while improving the generalization. Then, the large model is used to perform an association analysis on the historical power repair information and the potential anomalies to generate target policy information for eliminating the potential anomalies, realizing the early elimination of potential hazards before the faults caused by the potential anomalies occur, and further reducing the impact of the potential anomalies on the normal operation of the distribution network.

[0027] Figure 1 Schematically shows an exemplary system architecture to which a large model-based information generation method and apparatus can be applied according to an embodiment of the present disclosure.

[0028] It should be noted that Figure 1The illustration below is merely an example of the system architecture to which the embodiments of the present disclosure can be applied, to assist those skilled in the art in understanding the technical content of the present disclosure. However, it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, the exemplary system architecture to which the information generation method and apparatus based on a large model can be applied may include a terminal device. However, the terminal device may implement the information generation method and apparatus based on a large model provided by the embodiments of the present disclosure without interacting with the server.

[0029] As Figure 1 shown, the system architecture 100 according to this embodiment may include an intelligent power sensing device 101, a terminal device 102, a network 103, and a server 104. The network 103 is used to provide a medium for communication links between the intelligent power sensing device 101, the terminal device 102, and the server 104. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0030] The user may use the intelligent power sensing device 101 to collect the operation status information of the distribution network and send it to the terminal device 102 through the network 103. The terminal device 102 interacts with the server 104 through the network 103 to receive or send messages, etc. Various communication client applications may be installed on the terminal device 102, such as a fault analysis application, a web browser application, a search application, an instant messaging tool, an email client, and / or a social platform software, etc.

[0031] The terminal device 102 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, and desktop computers, etc.

[0032] The server 104 may be a server providing various services, such as a background management server (merely an example) that supports the content browsed by the user using the terminal device 102. The background management server may analyze and process data such as received user requests, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0033] It should be noted that the information generation method based on a large model provided by the embodiments of the present disclosure can generally be executed by the terminal device 102. Correspondingly, the information generation apparatus based on a large model provided by the embodiments of the present disclosure may also be provided in the terminal device 102.

[0034] Alternatively, the information generation method based on a large model provided by the embodiments of the present disclosure can generally also be executed by the server 104. Correspondingly, the information generation device based on a large model provided by the embodiments of the present disclosure can generally be disposed in the server 104. The information generation method based on a large model provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 104 and capable of communicating with the terminal device 102 and / or the server 104. Correspondingly, the information generation device based on a large model provided by the embodiments of the present disclosure can also be disposed in a server or a server cluster different from the server 105 and capable of communicating with the terminal device 102 and / or the server 104.

[0035] For example, the terminal device 102 can send the operation status information of the distribution network collected by the intelligent power sensing device 101 to the server 104. The server 104 generates potential anomalies and recommended strategies by executing the information generation method based on a large model of the embodiments of the present disclosure. And feedbacks the potential anomalies and recommended strategies to the terminal device 102. The terminal device 102 displays the potential anomalies and recommended strategies to the user through a visualization page so that the user can refer to the recommended strategies to perform relevant operations for eliminating potential anomalies.

[0036] It should be understood that Figure 1 the numbers of the terminal devices, networks, and servers in

[0037] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure, and application, etc. of the user's personal information all comply with the provisions of relevant laws and regulations, adopt necessary confidentiality measures, and do not violate public order and good customs.

[0038] In the technical solution of the present disclosure, before obtaining or collecting the user's personal information, the authorization or consent of the user is obtained.

[0039] Figure 2 Schematically shows a flowchart of an information generation method based on a large model according to an embodiment of the present disclosure.

[0040] As Figure 2 shown, the information generation method 200 may include operation S210 to operation S220.

[0041] In operation S210, using a large model, analyze the operation status information and initial anomaly information of the distribution network in the current period to generate potential anomaly information associated with the initial anomaly.

[0042] In operation S220, using a large model, correlation analysis is performed on historical power restoration information and potential anomaly information to generate target policy information for excluding potential anomalies.

[0043] According to an embodiment of the present disclosure, the operating state information of the distribution network may include: information such as current, voltage, power, frequency, etc. The operating state information of the distribution network may be collected by using intelligent power sensing devices configured in the distribution network, such as: current mutual inductance devices, voltage sensing devices, etc.

[0044] According to an embodiment of the present disclosure, the initial anomaly information may be obtained by detecting the operating state information using an anomaly detection model. The anomaly detection model may be any trained lightweight deep learning network, such as: CNN (Convolutional Neural Networks), LSTM (Long Short-Term Memory), etc.

[0045] For example: Information such as the current, voltage, power, frequency, etc. of the distribution network in the current period may be input into the anomaly detection model, and the anomaly detection model outputs the initial anomaly by extracting temporal features. For example: Short circuit, arc fault or current overload, etc.

[0046] In some embodiments, a Prompt (prompt information) A may be constructed based on the operating state information and the initial anomaly information of the distribution network in the current period. The Prompt A may also include exemplary text for outputting potential anomalies associated with the reference anomaly based on the reference operating state and the reference anomaly. Then, the Prompt A is input into the large model, and the large model generates potential anomalies based on a deep analysis of the operating state information and the initial anomaly information.

[0047] In some embodiments, the potential anomaly associated with the initial anomaly may be the cause of the initial anomaly. For example: The initial anomaly may be a short circuit, and the potential anomaly may be a soldered joint with a virtual weld in the transmission line of the distribution network.

[0048] In some embodiments, the potential anomaly associated with the initial anomaly may also be a potential hidden danger that may be caused by the troubleshooting operation after troubleshooting the initial anomaly. For example: The initial anomaly may be an arc fault, and the potential anomaly may be partial discharge caused by the possible introduction of air gaps or impurities during troubleshooting.

[0049] According to an embodiment of the present disclosure, the historical power restoration information represents the repair operation information taken in the historical period in the case of power operation anomalies caused by historical faults.

[0050] In some embodiments, Prompt B can be constructed based on historical power restoration information and potential anomalies. The Prompt B can also include exemplary text for generating a reference strategy for excluding reference anomalies based on the reference anomalies. Then, the Prompt B is input into the large model, and the large model generates a target strategy by analyzing the association between the potential anomalies and the historical power restoration information.

[0051] For example, when the potential anomaly is partial discharge, the generated target strategy can include operations such as adding vacuum impregnation treatment to insulating components and filling silicone grease for sealing during the troubleshooting operation of arc faults. So that relevant personnel can perform the operations in the target strategy during the troubleshooting of arc faults for the potential anomalies, further reducing the occurrence probability of potential anomalies that may be introduced during the troubleshooting of arc faults.

[0052] Figure 3 Schematically shows a schematic diagram of an information generation method based on a large model according to an embodiment of the present disclosure.

[0053] As Figure 3 shown, in this Embodiment 300, first, the operating state information 301 is input into the anomaly detection model 310, and the initial anomaly 302 is output. Then, the operating state information 301 and the initial anomaly 302 are input into the large model 320, and the potential anomaly 311 is output. Finally, the historical power restoration information and the potential anomaly 311 are input into the large model 320, and the target strategy is output.

[0054] Embodiments of the present disclosure utilize the semantic understanding ability of the large model to identify potential anomalies associated with the initial anomalies detected by the anomaly detection model from the operating state information of the distribution network. Since this potential anomaly is obtained based on the large model's full analysis of the initial anomaly information and the operating state information, it is possible to ensure the accuracy of the identification results while improving the generalization ability. Then, the large model is used to perform an association analysis on the historical power restoration information and the potential anomalies, generating target strategy information for excluding the potential anomalies, realizing the early elimination of potential hazards before the occurrence of faults caused by the potential anomalies, and further reducing the impact of potential anomalies on the normal operation of the distribution network.

[0055] According to an embodiment of the present disclosure, by using a large model to analyze the operating state information and the initial anomaly information of the distribution network at the current time period, generating potential anomaly information associated with the initial anomaly can include the following operations: respectively extracting the time-domain features and frequency-domain features of the operating state information; using the large model to analyze the time-domain features and frequency-domain features to generate a time-series change trend; and using the large model to perform an association analysis on the time-series change trend and the initial anomaly information to generate potential anomaly information associated with the initial anomaly.

[0056] In some embodiments, the operation status information can be preprocessed, such as deduplication, format conversion, noise removal, outlier removal, etc., and then the time-domain features of the operation status information can be extracted. The time-domain features include, but are not limited to, the maximum value, minimum value, mean value, standard deviation, etc. of the data included in the operation status information. The frequency-domain features include, but are not limited to, the frequency distribution of the data included in the operation status information.

[0057] In some embodiments, by configuring a time window, in-depth analysis can be performed on the continuously changing time-domain features and frequency-domain features within the time window to generate a time-series change trend. Then, using a large model, correlation analysis is performed on the time-series change trend and the initial anomaly information to generate potential anomaly information associated with the initial anomaly.

[0058] For example: when the time-series change trend indicates that the current change within 10s exceeds 20% fluctuation, the initial anomaly can be overload, and the potential anomaly associated with the initial anomaly generated by the large model can be motor load anomaly.

[0059] In some embodiments, the operation of data preprocessing can be performed on the terminal device, and the preprocessed data can be transmitted to the server through the network so that the server can call the large model to perform the operations of generating potential anomalies and target policies.

[0060] During the data transmission process, the preprocessed data can be compressed before being transmitted to the server, or it can be sent to the server in a streaming transmission manner to meet the requirements of real-time concurrent processing of large-scale power data.

[0061] According to the embodiments of the present disclosure, by utilizing the deep semantic understanding ability of the large model and analyzing the time-series change trend of the time-domain features and frequency-domain features, deep and complex features in the operation status information can be mined, so that potential anomalies not recognized by the anomaly detection model can be identified, further improving the generalization of anomaly recognition.

[0062] In some embodiments, when the large model performs correlation analysis on the time-series change trend and the initial anomaly, it can perform correlation analysis on the reference time-series change information included in the Prompt to generate an exemplary text of the reference anomaly as a reference example for semantic understanding. It can also use the historical time-series change trend as a reference example, enabling the large model to perform correlation analysis in combination with historical data, deepen the analysis of the operation status of the distribution network, and deeply mine the potential anomaly information associated with the initial anomaly.

[0063] In the embodiments of the present disclosure, a large model is used to perform correlation analysis on the time series change trend and the initial anomaly information to generate potential anomaly information associated with the initial anomaly, which may include the following operations: obtaining the historical operation state information of the distribution network associated with the initial anomaly; respectively extracting the historical time domain features and historical frequency domain features of the historical operation state information; using the large model to analyze the historical time domain features and historical frequency domain features to generate a historical time series change trend; and using the large model to perform correlation analysis on the time series change trend and the historical time series change trend to generate potential anomaly information associated with the initial anomaly.

[0064] According to an embodiment of the present disclosure, the historical operation state information of the distribution network associated with the initial anomaly may represent the historical operation state of the distribution network in the time periods before and after the same anomaly as the initial anomaly that has occurred in the historical period.

[0065] For example: If the initial anomaly is an arc fault, the historical operation state information in the time periods before and after the occurrence of the arc fault can be determined from the historical operation state information of the distribution network. After performing the preprocessing operations described above on the historical operation state information in this time period, the historical time domain features and historical frequency domain features are respectively extracted. The historical time domain features include, but are not limited to, the maximum value, minimum value, mean value, standard deviation, etc. of the historical data included in the historical operation state information. The historical frequency domain features include, but are not limited to, the frequency distribution of the historical data included in the historical operation state information.

[0066] In some embodiments, by configuring a time window, in-depth analysis can be performed on the continuously changing historical time domain features and historical frequency domain features within the time window to generate a historical time series change trend. Then, the large model is used to perform correlation analysis on the time series change trend and the historical time series change features to generate potential anomaly information associated with the initial anomaly.

[0067] In the process of using the large model to analyze potential anomalies, adding the historical operation state information as a reference and using the historical time series change trend as a reference example enables the large model to perform correlation analysis in combination with historical data, deepening the analysis of the operation state of the distribution network, deeply mining potential anomaly information associated with the initial anomaly, and further improving the accuracy of the potential anomalies generated by the large model.

[0068] In practical application scenarios, the potential anomaly may be an anomaly that has occurred in the historical period or an anomaly that has not occurred in the historical period. Then, when the potential anomaly has occurred in the historical period, there may be some omissions in the historical power repair information that led to the occurrence of the potential anomaly. When the potential anomaly has not occurred in the historical period, the historical power repair information may include operations for excluding potential anomalies, or operations for reducing the probability of potential anomalies occurring, etc.

[0069] Therefore, in the embodiments of the present disclosure, by using a large model, an association analysis is performed on historical power restoration information and potential anomaly information to generate policy information for excluding potential anomalies, which may include the following operations: extracting historical restoration policy information associated with the initial anomaly from the historical power restoration information; using the large model to match the historical restoration policy information with the potential anomaly information to generate target policy information.

[0070] According to an embodiment of the present disclosure, the historical restoration policy information indicates the first restoration operation performed in a historical period after an operation for excluding the initial anomaly has been executed.

[0071] In some embodiments, when a potential anomaly has occurred in the historical period, the large model can be used to generate target policy information for repairing these omissions by analyzing the omissions in the historical restoration policy information.

[0072] For example: The potential anomaly may be partial discharge associated with an arc fault. In the historical power restoration information associated with the arc fault, it may include "After excluding the arc fault, a partial discharge phenomenon occurred. For the partial discharge phenomenon, a sealing operation was performed on the insulating component." In this way, when analyzing the historical restoration policy information, the large model can determine that during the operation of repairing the arc fault, a potential anomaly of partial discharge may occur, and a sealing operation on the insulating component can be added during the operation of repairing the arc fault to reduce the occurrence probability of partial discharge. Therefore, the target policy information generated by the large model may include "Adding a sealing operation on the insulating component during the operation of repairing the arc fault".

[0073] In some embodiments, when a potential anomaly has not occurred in the historical period, the large model can be used to generate an operation for excluding the potential anomaly or an operation for reducing the occurrence probability of the potential anomaly by analyzing the specific operations in the historical restoration policy information.

[0074] For example: The potential anomaly may be partial discharge associated with an arc fault. In the historical power restoration information associated with the arc fault, it may include "After excluding the arc fault, a sealing operation was performed on the insulating component to reduce the possible introduction of air gaps or impurities during the troubleshooting." In this way, when analyzing the historical restoration policy, the large model can determine that the sealing operation performed on the insulating component reduces the possible introduction of air gaps or impurities during the troubleshooting, thereby reducing the probability of partial discharge occurrence. Thus, target policy information is generated.

[0075] By using the large model to deeply analyze the association between the restoration operations in the historical restoration policy information and the potential anomalies, the accuracy of the restoration policy is further improved.

[0076] Since there are many types of faults in the distribution network and there are also many reasons for the same fault, historical operating environments, such as weather and equipment aging degree, may all lead to the occurrence of certain faults. Therefore, in order to further reduce the redundant and interfering information in the historical repair strategy information, the historical operating environment of the distribution network can be extracted from the historical power repair information; using a large model, the second repair operations associated with the historical operating environment can be deleted from the historical repair strategy information to obtain the target historical repair strategy information; and using a large model, the target historical repair strategy information can be matched with potential abnormal information to generate the target strategy information.

[0077] For example, for a short - circuit fault, strong wind weather may cause damage to insulation supports, and wires may collide and rub against each other, resulting in damage to wire insulation and thus triggering a short - circuit. In addition, lightning strikes may also cause flashover or lightning arrester operation, thereby triggering a short - circuit fault. Extreme weather such as heavy rain and heavy snow may also cause electrical equipment to get waterlogged or iced, increasing the risk of short - circuit.

[0078] If there is no such special weather in the current period, the large model can be used to delete the repair operations associated with the historical operating environment from the historical repair strategy information, further reducing the interference of historical redundant information on the output strategy of the large model.

[0079] In order to further improve the accuracy of the strategy output by the large model, in addition to pre - training the large model using sample data in the field of power detection technology, the model parameters of the large model can also be reversely optimized based on the feedback information of relevant personnel on the target strategy, so as to use the self - correction ability of the large model to improve the matching degree between the target strategy and user needs, and further improve the user experience.

[0080] In the embodiments of the present disclosure, the above - mentioned method may further include the following operations: obtaining feedback information on the target strategy information; using the large model to optimize the target strategy information based on the feedback information to generate optimized strategy information.

[0081] Figure 4 Schematically shows a schematic diagram of optimizing the large - model output strategy based on feedback information according to an embodiment of the present disclosure.

[0082] As Figure 4 shown, in this embodiment 400, the historical power repair information 321 and the potential line 311 can be input into the large model 320 to output the target strategy 322. Then, feedback information of relevant personnel on the target strategy can be obtained, and the feedback information includes but is not limited to: the effectiveness degree of the strategy, the rationality degree of the strategy, etc. Then, the model parameters of the large model 320 can be adjusted based on the feedback information, and the large model can be used for strategy optimization to generate the optimized strategy 402.

[0083] In some embodiments, the feedback information indicates that the target policy information includes redundant policies. Using a large model, the target policy information is optimized based on the feedback information to generate optimized policy information, which may include the following operations: In response to determining that the feedback information indicates that the target policy information includes redundant policies, using the large model, redundant operations are deleted from the target policy information based on the feedback information.

[0084] For example: The feedback information may include redundant policies in the target policy information that are irrelevant to both the initial anomaly and potential anomalies. While deleting the redundant operations, the weights associated with the redundant policies can also be reduced, which can reduce the generation of such redundant policies when the large model performs policy optimization.

[0085] In some embodiments, the feedback information indicates that the execution order among the third repair operations in the target policy information is incorrect. Using a large model, the target policy information is optimized based on the feedback information to generate optimized policy information, which may include the following operations: In response to determining that the feedback information indicates that the execution order among the third repair operations in the target policy information is incorrect, using the large model, the execution order among the third repair operations in the target policy information is corrected based on the feedback information.

[0086] For example: The target policy information may be: Before excluding the arc fault, perform the sealing operation of the insulating component. The feedback information may include that the sealing operation should be performed after excluding the arc fault, and it can be determined that the execution order between the operation of excluding the arc fault and the sealing operation of the insulating component in the target policy is incorrect. The execution order between the exclusion operation and the sealing operation of the insulating component can be corrected using the large model to generate optimized policy information. At the same time, the execution order weight can also be increased so that when the large model generates policies subsequently, it pays more attention to the execution order among the repair operations.

[0087] According to the embodiments of the present disclosure, based on the feedback information of relevant personnel on the target policy, the model parameters of the large model are reversely optimized to utilize the self-correction ability of the large model to improve the matching degree between the target policy and user requirements, and further improve the user experience.

[0088] Since the anomaly detection model trained based on sample data is limited by the sample data, resulting in low generalization. When the large model identifies relatively accurate potential anomalies, they can be used as sample data to reversely train the anomaly detection model, thereby further improving the detection accuracy of the anomaly detection model.

[0089] Figure 5 A schematic diagram showing the reverse training of the anomaly detection model using the potential anomalies identified by the large model according to the embodiments of the present disclosure is schematically shown.

[0090] Such as Figure 5As shown, based on Embodiment 300, in this Embodiment 500, sample data is constructed using potential anomalies 311 and operating status information 301 for training the anomaly detection model 310.

[0091] In some embodiments, the above method may further include the following operations: generating target operating status information by deleting noise information and anomaly information from the operating status information; and generating sample data for training the anomaly detection model based on the target operating status information and potential anomaly information.

[0092] For example: The Kalman filter or median filter method can be used to remove noise information from the operating status information. Anomaly data detection methods such as the interquartile range method can be used to delete severely abnormal data from the operating status information. Then, the target operating status information and potential anomaly information can be used as sample data to train the anomaly detection model.

[0093] In some embodiments, generating sample data for training the fault detection model based on the target operating status information and potential anomaly information may include the following operations: respectively extracting the time-domain features and frequency-domain features of the target operating status information; performing time-series detection on the time-domain features and frequency-domain features to generate time-series change features; and generating sample data based on the time-domain features, frequency-domain features, time-series change features, and potential anomaly information.

[0094] For example: Based on the methods for extracting time-domain features, frequency-domain features, and time-series change features from the operating status information described above, the target operating status information can be processed, which will not be elaborated here. Then, sample data can be generated based on the time-domain features, frequency-domain features, time-series change features, and potential anomaly information. Since the number of abnormal sample data is small, data augmentation can be further used to increase the number of sample data and further improve the quality of the sample data.

[0095] When potential anomalies identified by the large model are used as sample data to inversely train the anomaly detection model, the detection accuracy of the anomaly detection model can be further improved.

[0096] Figure 6 The block diagram of an information generation device based on a large model according to an embodiment of the present disclosure is schematically shown.

[0097] As Figure 6 shown, the information generation device 600 may include an analysis module 610 and a generation module 620.

[0098] The analysis module 610 is configured to use the large model to analyze the operating status information and initial anomaly information of the distribution network in the current period, and generate potential anomaly information associated with the initial anomaly, where the initial anomaly information is obtained by detecting the operating status information using the anomaly detection model.

[0099] A generation module 620, configured to use a large model to perform correlation analysis on historical power restoration information and potential anomaly information, and generate target policy information for excluding potential anomalies.

[0100] According to an embodiment of the present disclosure, the analysis module includes: a first extraction sub-module, a first analysis sub-module, and a first generation sub-module.

[0101] The first extraction sub-module is configured to extract the time-domain features and frequency-domain features of the operating state information respectively.

[0102] The first analysis sub-module is configured to use a large model to analyze the time-domain features and frequency-domain features, and generate a time-series change trend.

[0103] The first generation sub-module is configured to use a large model to perform correlation analysis on the time-series change trend and the initial anomaly information, and generate potential anomaly information associated with the initial anomaly.

[0104] According to an embodiment of the present disclosure, the first generation sub-module includes: an acquisition unit, an extraction unit, an analysis unit, and a generation unit.

[0105] The acquisition unit is configured to acquire the historical operating state information of the distribution network associated with the initial anomaly.

[0106] The extraction unit is configured to extract the historical time-domain features and historical frequency-domain features of the historical operating state information respectively.

[0107] The analysis unit is configured to use a large model to analyze the historical time-domain features and historical frequency-domain features, and generate a historical time-series change trend.

[0108] The generation unit is configured to use a large model to perform correlation analysis on the time-series change trend and the historical time-series change trend, and generate potential anomaly information associated with the initial anomaly.

[0109] According to an embodiment of the present disclosure, the generation module includes: a second extraction sub-module, a second generation sub-module.

[0110] The second extraction sub-module is configured to extract historical repair policy information associated with the initial anomaly from the historical power restoration information; the historical repair policy information indicates the first repair operation performed after the initial anomaly has been excluded during a historical period.

[0111] The second generation sub-module is configured to use a large model to match the historical repair policy information with the potential anomaly information, and generate target policy information.

[0112] According to an embodiment of the present disclosure, the above-mentioned generation module further includes: a third extraction sub-module, a first deletion sub-module, and a generation sub-module.

[0113] The third extraction sub-module is used to extract the historical operating environment of the distribution network from the historical power repair information.

[0114] The first deletion sub-module is used to use the large model to delete the second repair operations associated with the historical operating environment from the historical repair strategy information to obtain the target historical repair strategy information.

[0115] The generation sub-module is used to use the large model to match the target historical repair strategy information with the potential abnormal information to generate the target strategy information.

[0116] According to an embodiment of the present disclosure, the above information generation device further includes: an acquisition module and an optimization module.

[0117] The acquisition module is used to acquire feedback information for the target strategy information.

[0118] The optimization module is used to use the large model to optimize the target strategy information based on the feedback information to generate the optimized strategy information.

[0119] According to an embodiment of the present disclosure, the optimization module includes: a second deletion sub-module and a correction sub-module.

[0120] The second deletion sub-module is used to respond to determining that the feedback information indicates that the target strategy information includes redundant strategies, and use the large model to delete the redundant operations from the target strategy information based on the feedback information.

[0121] The correction sub-module is used to respond to determining that the feedback information indicates that the execution order between the third repair operations in the target strategy information is incorrect, and use the large model to correct the execution order between the third repair operations in the target strategy information based on the feedback information.

[0122] According to an embodiment of the present disclosure, the above information generation device further includes: a deletion module and a sample generation module.

[0123] The deletion module is used to generate the target operating state information by deleting the noise information and abnormal information in the operating state information.

[0124] The sample generation module is used to generate sample data for training the anomaly detection model according to the target operating state information and the potential abnormal information.

[0125] According to an embodiment of the present disclosure, the sample generation module includes: a fourth extraction sub-module, a detection sub-module, and a sample generation sub-module.

[0126] The fourth extraction sub-module is used to extract the time domain features and frequency domain features of the target operating state information respectively.

[0127] The detection sub-module is used to perform sequential detection on time-domain features and frequency-domain features to generate sequential change features.

[0128] The sample generation sub-module is used to generate sample data according to time-domain features, frequency-domain features, sequential change features, and potential anomaly information.

[0129] Figure 7 Schematically shows a block diagram of an agent for information generation according to an embodiment of the present disclosure.

[0130] In an embodiment of the present disclosure, inspired by the von Neumann architecture in modern computer theory, as Figure 7 shown, the AI agent 700 may include three core modules: an input module 710, an output module 720, and a processing module 730. The processing module 730 may include a control unit 731, a storage unit 732, and an arithmetic unit 733.

[0131] The input module 710 is responsible for receiving or sensing information such as queries, requests, instructions, signals, or data from the outside world (such as users or the external environment), and converting it into a format that the AI agent 700 can understand and process. The input module 710 is the primary link for the AI agent 700 to interact with the outside world. It enables the AI agent 700 to efficiently and accurately obtain necessary "sensory" information from the outside world and respond to this information.

[0132] In an example, the input information received by using the input module 710 may include the initial anomaly information and the operating state information described above.

[0133] In an example, the processing module 730 is the core support for the AI agent 700 to process complex tasks. The processing module 730 may determine a target task based on the input information received by the input module 1010, determine a large model based on the target task, and output a target policy by invoking the large model to execute the information generation method based on the large model described above.

[0134] In an example, the control unit 731 in the processing module 730 will continuously interact with the storage unit 732, the arithmetic unit 733, and / or the output module 720 during operation. However, it should be noted that in the embodiment of the present disclosure, the control unit 731 acts as a single initiator to initiate communication with the storage unit 732, the arithmetic unit 733, and / or the output module 720, and there is no communication coupling between the storage unit 732, the arithmetic unit 733, and the output module 720.

[0135] In the example, the performance of the control unit 731 can be closely related to the large model on which the AI agent 700 is based. To fully utilize the capabilities of the large language model, the internal structure of the control unit 731 can be designed to be highly configurable and extensible to handle various different types of tasks and requirements in real-world scenarios.

[0136] The storage unit 732 can be responsible for memorizing information such as historical interactions and event streams. The text generated in each round as described above can be included in the storage unit 732.

[0137] In the example, after the AI agent 700 obtains an information optimization request, the AI agent 700 can call the large model to execute the task corresponding to the input information and output the corresponding text. The corresponding text can be stored in the storage unit 732. The AI agent 700 can retrieve relevant data resources from the storage unit 732 and feedback them to the control unit 731. Then, the control unit 731 can utilize the feedback data resources to generate a target strategy. It can also retrieve relevant data resources from the storage unit 732 and feedback them to the control unit 731. Then, the control unit 731 can utilize the returned data resources to generate a target strategy. And the target strategy is passed to the output module 720.

[0138] The arithmetic unit 733 can be regarded as a predefined tool library. Renderers and display controls as described above can be included in the arithmetic unit 733.

[0139] In the example, when the AI agent 700 needs to render multiple output data, relevant renderers and display controls can be called from the arithmetic unit 733 and feedback them to the control unit 732. Then, the control unit 732 can utilize the feedback renderers and display controls to render the first search result and pass the first search result to the output module 720. It can be understood that although the large language model has excellent language understanding and generation capabilities, like humans, the tasks it can solve without any tools are very limited. When the AI agent 700 is given the ability to call tools, it can achieve tasks such as performing mathematical operations with the help of a calculator, performing data analysis with the help of Python, and performing prediction tasks with the help of a search engine.

[0140] In the example, the output module 720 can output the target strategy described above.

[0141] The AI agent 700 according to the embodiments of the present disclosure can simply and effectively improve the degree of intelligence and enhance flexibility and versatility.

[0142] According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0143] According to an embodiment of the present disclosure, an electronic device 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 the method as described above.

[0144] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the method as described above.

[0145] According to an embodiment of the present disclosure, a computer program product includes a computer program which, when executed by a processor, implements the method as described above.

[0146] Figure 8 FIG. shows a schematic block diagram of an exemplary electronic device 800 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, for example, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as, for example, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0147] As Figure 8 shown, the device 800 includes a computing unit 801 which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0148] A plurality of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0149] The computing unit 801 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 executes the various methods and processes described above, such as the large model-based information generation method. For example, in some embodiments, the large model-based information generation method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the large model-based information generation method described above can be executed. Alternatively, in other embodiments, the computing unit 801 can be configured to execute the large model-based information generation method in any other suitable manner (e.g., by means of firmware).

[0150] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), system-on-a-chip systems (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general-purpose programmable processor, that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0151] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0152] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0153] To provide for interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic, speech, or tactile input).

[0154] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or in a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0155] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0156] It should be understood that the various forms of processes shown above can be used, with steps reordered, added or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution disclosed in this disclosure can be achieved, and no limitations are imposed herein.

[0157] The above specific embodiments do not constitute a limitation on the protection scope of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. A method for generating information based on a large model, comprising: Using the large model, analyzing the operation status information and initial abnormal information of the distribution network in the current period, and generating potential abnormal information associated with the initial abnormality, wherein the initial abnormal information is obtained by detecting the operation status information using the abnormality detection model; and The large model is used to perform correlation analysis on the historical power restoration information and the potential abnormality information, and generate target strategy information for eliminating potential abnormalities.

2. The method according to claim 1, wherein: The large model is used to analyze the operation status information and initial abnormal information of the distribution network in the current period to generate potential abnormal information associated with the initial abnormality, including: Respectively extracting time domain features and frequency domain features of the operating status information; Analyzing the time domain features and the frequency domain features using the large model to generate a time series variation trend; and The large model is used to perform correlation analysis on the time series variation trend and the initial abnormal information to generate potential abnormal information associated with the initial abnormality.

3. The method according to claim 2, wherein: The large model is used to perform correlation analysis on the time series variation trend and the initial abnormal information to generate potential abnormal information associated with the initial abnormality, including: Acquiring historical operating status information of the distribution network associated with the initial anomaly; Respectively extracting historical time domain features and historical frequency domain features of the historical operating status information; Analyzing the historical time domain features and the historical frequency domain features using the large model to generate a historical time series change trend; and The large model is used to perform correlation analysis on the time series change trend and the historical time series change trend to generate potential abnormality information associated with the initial abnormality.

4. The method according to claim 1, wherein: The method of using the large model to correlate historical power restoration information with the potential abnormality information and generate strategy information for eliminating potential abnormalities includes: Extracting historical repair strategy information associated with the initial abnormality from historical power repair information; the historical repair strategy information indicates a first repair operation performed after an operation for eliminating the initial abnormality has been performed during a historical period; and The historical repair strategy information is matched with the potential abnormality information by using the large model to generate the target strategy information.

5. The method according to claim 4, further comprising: Extract the historical operating environment of the distribution network from the historical power repair information; Using the large model, deleting the second repair operation associated with the historical operating environment from the historical repair strategy information to obtain target historical repair strategy information; as well as The target historical repair strategy information is matched with the potential abnormality information by using the large model to generate the target strategy information.

6. The method according to any one of claims 1 to 5, further comprising: Obtaining feedback information for the target strategy information; The target strategy information is optimized based on the feedback information using the large model to generate optimized strategy information.

7. The method according to claim 6, wherein: The step of optimizing the target strategy information based on the feedback information by using the large model to generate optimized strategy information includes: In response to determining that the feedback information indicates that the target policy information includes a redundant policy, utilizing the large model to delete the redundant operation from the target policy information based on the feedback information; and In response to determining that the feedback information indicates that the execution order between the third repair operations in the target strategy information is wrong, using the large model, correcting the execution order between the third repair operations in the target strategy information based on the feedback information.

8. The method according to any one of claims 1 to 7, further comprising: Generate target operating state information by deleting noise information and abnormal information in the operating state information; as well as According to the target operating state information and the potential abnormality information, sample data for training the abnormality detection model is generated.

9. The method according to claim 8, wherein: The generating, according to the target operating state information and the potential abnormality information, sample data for training the abnormality detection model comprises: Respectively extracting time domain features and frequency domain features of the target operating status information; Performing time series detection on the time domain features and the frequency domain features to generate time series change features; and The sample data is generated according to the time domain features, the frequency domain features, the time series change features and the potential abnormality information.

10. An information generation device based on a large model, comprising: An analysis module, configured to analyze the operation status information and initial abnormality information of the distribution network in the current period by using a large model, and generate potential abnormality information associated with the initial abnormality, wherein the initial abnormality information is obtained by detecting the operation status information by using an abnormality detection model; and A generation module is used to use the large model to perform correlation analysis on the historical power restoration information and the potential abnormality information, and generate target strategy information for eliminating potential abnormalities.

11. An intelligent agent for information generation, comprising: An input module, used for receiving input information; a processing module, configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and obtain output information by executing the method of any one of claims 1 to 9 by calling the large model; as well as An output module is used to output the output information obtained by the processing module.

12. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

14. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images