Smart grid fault monitoring method and system using artificial intelligence

Abstract

The present application relates to the field of smart grid and artificial intelligence technology, and in particular provides a method and system for smart grid fault monitoring using artificial intelligence. The present application achieves comprehensive analysis and efficient use of power IoT sensor monitoring data by comprehensively using two fault discrimination decision models, grid operation status mining, status consistency analysis, and fault point location warning, providing a strong technical guarantee for the safe and stable operation of the power grid.

Description

Technical Field

[0001] The present application relates to the field of smart grid and artificial intelligence technology, and in particular to a smart grid fault monitoring method and system using artificial intelligence. Background Art

[0002] With the rapid development of power Internet of Things technology and the continuous deepening of smart grid construction, a large number of sensors and monitoring equipment are deployed in the power system to collect real-time data on the operation status of the power grid. These data are of great significance for the stable operation and fault warning of the power system. However, how to effectively use these data, accurately identify the fault status of the power grid, and issue warnings in a timely manner has always been a technical challenge faced by the power industry.

[0003] In traditional power systems, fault identification mainly relies on manual experience and simple threshold judgment. Although this method is simple, its accuracy and efficiency are low, and it cannot meet the high-precision and high-timeliness requirements of modern smart grids for fault warning. Summary of the invention

[0004] In order to improve the above problems, the present application provides a smart grid fault monitoring method and system using artificial intelligence.

[0005] The present application provides a method for monitoring faults in a smart grid using artificial intelligence, which is applied to an artificial intelligence monitoring system. The method includes:

[0006] Acquire the power Internet of Things sensor monitoring data to be processed, and determine a first fault discrimination decision model and a second fault discrimination decision model for fault identification of the power Internet of Things sensor monitoring data to be processed, wherein the second fault discrimination decision model has learned the past fault location guidance vector of the fault location sensor monitoring data;

[0007] Based on the first fault discrimination decision model, a first power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain a first power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at at least one semantic fine-grained level;

[0008] Based on the second fault discrimination decision model and in accordance with the past fault location guidance vector, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain a second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level;

[0009] Performing state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic fine-grained level to obtain state similarities and differences analysis viewpoints;

[0010] According to the state similarities and differences analysis point of view, the fault point location warning is carried out on the power Internet of Things sensor monitoring data to be processed.

[0011] In some technical solutions, the second fault discrimination decision model includes at least one semantic fine-grained level operating state mining branch, and each semantic fine-grained level operating state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the corresponding semantic fine-grained level;

[0012] Based on the second fault discrimination decision model and in accordance with the past fault location guidance vector, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level, including:

[0013] Through the operation status mining branches at each semantic fine-grained level, according to the past fault location guidance vectors at the corresponding semantic fine-grained level, the second power grid operation status mining is performed on the power Internet of Things sensor monitoring data to be processed, and the second power grid operation status vector of the power Internet of Things sensor monitoring data to be processed at at least one semantic fine-grained level is obtained.

[0014] In some technical solutions, the second fault discrimination decision model includes a first operating state mining branch at a first semantic fine-grained level, and the first operating state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level;

[0015] Through the operation state mining branches at each semantic fine-grained level, according to the past fault location guidance vector at the corresponding semantic fine-grained level, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed, and the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level is obtained, including:

[0016] Acquire potential fault search features of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level;

[0017] According to the potential fault search feature, perform a potential fault search on the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label;

[0018] Based on the first operating state mining branch, a second power grid operating state mining is performed according to the potential fault search label to obtain a second power grid operating state vector of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level.

[0019] In some technical solutions, the past fault location guidance vector at the first semantic fine-grained level includes a past fault location identification vector and a past fault location attribute vector associated with the past fault location identification vector;

[0020] According to the potential fault search feature, a potential fault search is performed on the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label, including:

[0021] Determining a feature commonality value between the potential fault search feature and the past fault location identification vector;

[0022] Determining the search confidence of the potential fault search feature according to the feature commonality value;

[0023] According to the search confidence, a feature splicing operation is performed on the past fault location attribute vector to obtain a potential fault search label that completes feature splicing.

[0024] In some technical solutions, obtaining the potential fault search features of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level includes:

[0025] Performing semantic fine-grained level update processing on the first power grid operation state vector to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel;

[0026] Performing state vector integration on the updated power grid operation state vector under the target attention channel to obtain a power grid operation state integration vector;

[0027] Based on the first operating state mining branch and according to the integrated vector of the power grid operating state, a potential fault search feature of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level is generated.

[0028] In some technical solutions, the second fault discrimination decision model includes a second operating state mining branch at a second semantic fine-grained level, and the second operating state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the second semantic fine-grained level;

[0029] Through the operation state mining branches at each semantic fine-grained level, according to the past fault location guidance vector at the corresponding semantic fine-grained level, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed, and the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level is obtained, including:

[0030] Based on the second operation state mining branch, generating a basic power grid operation state vector of the to-be-processed power Internet of Things sensor monitoring data at the second semantic fine-grained level;

[0031] Generating a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level according to the past fault location guidance vector;

[0032] According to the basic power grid operation state vector, a feature splicing operation is performed on the fault dynamic search feature to generate a second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the second semantic fine-grained level.

[0033] In some technical solutions, the fault dynamic search feature includes at least one fault dynamic search sub-vector;

[0034] According to the basic power grid operation state vector, a feature splicing operation is performed on the fault dynamic search feature, including:

[0035] Determining a characteristic commonality value between the basic power grid operation state vector and the fault dynamic search subvector;

[0036] Determining a characteristic splicing coefficient of the fault dynamic search subvector according to the characteristic commonality value;

[0037] A feature splicing operation is performed on the fault dynamic search sub-vector according to the feature splicing coefficient.

[0038] In some technical solutions, the past fault localization guidance vector at the second semantic fine-grained level includes an update variable indication corresponding to the attention channel update;

[0039] Generating a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level according to the past fault location guidance vector includes:

[0040] According to the update variable indication, the basic power grid operation state vector is updated by an attention channel to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel;

[0041] Determining a feature splicing coefficient according to an operating state relationship spectrum of the updated power grid operating state vector under the target attention channel;

[0042] According to the feature splicing coefficient, state vector splicing is performed on the basic power grid operation state vector to generate a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level.

[0043] In some technical solutions, the method further includes:

[0044] Determining fault location sensor monitoring data, a debugged first fault discrimination decision model, and a second fault discrimination decision model to be debugged, wherein the second fault discrimination decision model has learned a basic past fault location guidance vector of the fault location sensor monitoring data;

[0045] Based on the first fault discrimination decision model, the first power grid operation state mining is performed on the fault location sensor monitoring data to obtain a first power grid operation state vector of the fault location sensor monitoring data at at least one semantic fine-grained level;

[0046] According to the first power grid operation state vector, optimizing the basic past fault location guidance vector to obtain an optimized past fault location guidance vector;

[0047] Based on the second fault discrimination decision model and in accordance with the optimized past fault location guidance vector, the second power grid operation state mining is performed on the fault location sensor monitoring data to obtain a second power grid operation state vector of the fault location sensor monitoring data at the at least one semantic fine-grained level;

[0048] Performing state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic fine-grained level to obtain state similarities and differences analysis viewpoints;

[0049] According to the state similarities and differences analysis point of view, the second fault discrimination decision model is debugged to determine the debugged second fault discrimination decision model, wherein the debugged second fault discrimination decision model has learned the optimized past fault location guidance vector of the fault location sensor monitoring data.

[0050] In some technical solutions, the second fault discrimination decision model includes a first operating state mining branch at a first semantic fine-grained level, the first operating state mining branch having learned a basic past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level;

[0051] Optimizing the basic past fault location guidance vector according to the first power grid operation state vector includes:

[0052] Acquire fault location authentication features of the fault location sensor monitoring data at the first semantic fine-grained level;

[0053] According to the fault location authentication feature, a feature splicing operation is performed on the basic past fault location guidance vector to obtain a fault location interaction vector;

[0054] The basic past fault location guidance vector is optimized according to the difference between the fault location authentication feature and the fault location interaction vector.

[0055] In some technical solutions, obtaining the fault location authentication feature of the fault location sensor monitoring data at the first semantic fine-grained level includes:

[0056] Determining upstream and downstream semantic fine-grained levels of the first semantic fine-grained level;

[0057] According to the first power grid operation state vector of the fault location sensor monitoring data at the upstream and downstream semantic fine-grained level, the fault location authentication feature of the fault location sensor monitoring data at the first semantic fine-grained level is determined.

[0058] An embodiment of the present application provides an artificial intelligence monitoring system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the above method.

[0059] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. The computer program implements the above method when running.

[0060] After research and analysis, it is found that a single fault discrimination decision model may be affected by multiple factors such as data quality and model complexity, resulting in limited accuracy and stability of fault identification. In order to improve the accuracy and reliability of fault identification, the embodiment of the present application proposes a method of using two fault discrimination decision models.

[0061] For example, the embodiment of the present application first obtains the power Internet of Things sensor monitoring data to be processed, and determines two fault discrimination decision models: a first fault discrimination decision model and a second fault discrimination decision model. Among them, the second fault discrimination decision model has learned the past fault location guidance vector of the fault location sensor monitoring data, which makes the model more targeted and accurate in fault identification.

[0062] Next, the embodiment of the present application uses these two models to mine the power grid operation status of the power IoT sensor monitoring data to be processed, and obtains the first power grid operation status vector and the second power grid operation status vector at at least one semantic fine-grained level. These two vectors reflect the operation status of the power grid from different angles, providing rich information for subsequent state consistency analysis and fault point location warning.

[0063] By comparing and analyzing the consistency of these two state vectors, potential problems and risk points in power grid operation can be revealed, thereby achieving early warning and rapid location of faults. This method not only improves the accuracy and timeliness of fault identification, but also provides strong technical support for the stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A flowchart of a smart grid fault monitoring method using artificial intelligence provided in an embodiment of the present application.

[0065] Figure 2 A schematic diagram of the structure of an artificial intelligence monitoring system 200 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0067] Figure 1 A smart grid fault monitoring method using artificial intelligence is shown, which is applied to an artificial intelligence monitoring system. The method includes the following steps 110-150.

[0068] Step 110: Obtain the power Internet of Things sensor monitoring data to be processed, and determine a first fault discrimination decision model and a second fault discrimination decision model for fault identification of the power Internet of Things sensor monitoring data to be processed, wherein the second fault discrimination decision model has learned the past fault location guidance vector of the fault location sensor monitoring data.

[0069] Step 120: Perform first power grid operation state mining on the power Internet of Things sensor monitoring data to be processed based on the first fault discrimination decision model to obtain a first power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at at least one semantic fine-grained level.

[0070] Step 130: Based on the second fault discrimination decision model and in accordance with the past fault location guidance vector, perform second power grid operation state mining on the power Internet of Things sensor monitoring data to be processed to obtain a second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at at least one semantic fine-grained level.

[0071] Step 140: Perform state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic granularity level to obtain state similarities and differences analysis viewpoints.

[0072] Step 150: Based on the state similarities and differences analysis point of view, perform fault point location warning on the power Internet of Things sensor monitoring data to be processed.

[0073] To facilitate understanding of the technical solutions described in the embodiments of the present application, an overall application scenario is first introduced below, and then each step is described in detail based on the application scenario.

[0074] In an exemplary power monitoring scenario, the artificial intelligence monitoring system obtains a large amount of pending power IoT sensor monitoring data in real time from various sensors throughout the power grid. These data reflect the real-time operating status of each node in the power grid. Based on the characteristics of the data and historical processing experience, the artificial intelligence monitoring system intelligently selects two fault discrimination decision models: the first fault discrimination decision model and the second fault discrimination decision model. In particular, the second fault discrimination decision model has mastered the past fault location guidance vectors of the fault location sensor monitoring data by learning historical fault data, which play an important role in accurately identifying abnormal conditions in the power grid.

[0075] Next, the AI ​​monitoring system uses the first fault discrimination decision model to conduct in-depth analysis of the pending power IoT sensor monitoring data. This model focuses on mining the operating status of the power grid from the data. It can generate the first power grid operating status vectors about the power grid operating status at different semantic granular levels. These vectors carefully depict the operating status of the power grid at all levels.

[0076] Then, the AI ​​monitoring system switches to the second fault identification decision model, which is characterized by the use of past fault location guidance vectors. Based on these historical experience data, it conducts a deeper mining of the power IoT sensor monitoring data to be processed, and generates a second power grid operation state vector at the same semantic granular level. These vectors not only reflect the real-time state of the power grid, but also imply the correlation with historical fault modes.

[0077] Next comes the analysis phase, where the AI ​​monitoring system will compare and analyze the state vectors generated by the two models. The AI ​​monitoring system will compare the first power grid operation state vector and the second power grid operation state vector at the same semantic granular level, analyze the state consistency between the two through a sophisticated algorithm, and draw conclusions on the state similarities and differences.

[0078] Finally, based on the above-mentioned state similarities and differences analysis point of view, the artificial intelligence monitoring system can accurately identify abnormalities or fault points in the power grid and issue early warnings in a timely manner. This early warning includes not only the location of the fault, but also the type and severity of the fault, providing valuable information for power system maintenance personnel, enabling them to respond quickly and repair the fault to ensure the stable operation of the power grid.

[0079] Through this series of intelligent analysis and early warning, the artificial intelligence monitoring system has significantly improved the power grid's fault identification and response capabilities, providing strong technical support for the safe and stable operation of the power system.

[0080] Combined with the above application scenarios, the pending power IoT sensor monitoring data refers to the raw data collected in real time from each sensor node in the power grid that has not yet been analyzed and processed. These data usually include multiple parameters such as voltage, current, power factor, temperature, humidity, etc., and are an important source of information reflecting the operating status of the power grid. These data need to be processed by professional analysis models to be converted into an accurate description of the operating status of the power grid.

[0081] The first fault discrimination decision model is an analysis module in the artificial intelligence monitoring system, which is mainly used to preliminarily analyze the pending power IoT sensor monitoring data. This model can identify the basic characteristics of the power grid operation status through preset algorithms and data mining technology, and convert them into the first power grid operation status vector. These vectors are a mathematical description of the power grid operation status, which helps to further analyze and diagnose possible problems in the power grid.

[0082] Similar to the first fault discrimination decision model, the second fault discrimination decision model is also a tool for analyzing power IoT sensor monitoring data. However, the characteristic of this model is that it incorporates past fault location guidance vectors, that is, it has learned the characteristics and patterns of faults from historical fault data. This makes the second model more sensitive and accurate in identifying and handling similar faults.

[0083] Fault location sensor monitoring data refers to sensor monitoring data that is directly related to the location of power grid faults. When a power grid fault occurs, these data will show obvious abnormalities or fluctuations, which is an important basis for fault location. By accurately capturing and analyzing these data, the artificial intelligence monitoring system can quickly locate the fault point and provide key information for subsequent maintenance work.

[0084] Past fault location guidance vectors are key information extracted from historical fault events, which are stored and represented in the form of vectors. These vectors contain important information such as data characteristics when the fault occurred, fault type, and fault location. In the subsequent fault identification process, these guidance vectors can be used as reference standards to help the artificial intelligence monitoring system more accurately determine the current operating status of the power grid and possible faults.

[0085] Furthermore, in the complex environment of power monitoring, the artificial intelligence monitoring system first performs the key steps of data acquisition and model selection, which is the content described in step 110: Step 110 begins with the artificial intelligence monitoring system receiving power IoT sensor monitoring data from key nodes of the power grid in real time through network connections and sensor interfaces. These data streams continuously flow into the artificial intelligence monitoring system, waiting to be further processed and analyzed.

[0086] After obtaining these pending power IoT sensor monitoring data, the next step of the artificial intelligence monitoring system is to determine which analysis model to use to interpret these data. In this process, the artificial intelligence monitoring system will intelligently select two key fault discrimination decision models: the first fault discrimination decision model and the second fault discrimination decision model.

[0087] The first fault discrimination decision model is a general analysis tool that can make preliminary judgments and mining on the operation status of the power grid based on the overall characteristics and trends of the data. This model does not rely on specific historical fault data, but focuses on extracting useful information from current data.

[0088] At the same time, the second fault discrimination decision model is more focused on leveraging the wisdom in historical fault data. In the previous learning process, this model has deeply analyzed a large amount of fault location sensor monitoring data and extracted past fault location guidance vectors from it. These guidance vectors are actually a kind of empirical knowledge, which reflects the data characteristics and patterns when the power grid failed in the past. By incorporating these guidance vectors into the analysis process of the second model, the artificial intelligence monitoring system can more keenly capture the fault signals that may be hidden in the current data.

[0089] It can be seen that step 110 is the starting point of the entire artificial intelligence monitoring system workflow. It ensures that the artificial intelligence monitoring system can obtain the latest and most comprehensive power Internet of Things sensor monitoring data, and lays a solid foundation for subsequent data analysis and fault warning.

[0090] Furthermore, the first power grid operation status mining is a data processing and analysis process, in which the artificial intelligence monitoring system uses the first fault discrimination decision model to conduct in-depth exploration of the power IoT sensor monitoring data to be processed. This mining process aims to identify and extract the characteristics and patterns of the current operation status of the power grid, and provide key information for subsequent status assessment and fault warning. Through the first power grid operation status mining, the artificial intelligence monitoring system can have a more comprehensive understanding of the operation status of the power grid, so as to timely discover and respond to potential problems.

[0091] In the analysis of power grid operation status, the semantic granularity level refers to the level of detail of data interpretation. Different semantic granularity levels correspond to status descriptions at different levels and levels of detail. For example, at a coarser semantic granularity level, the artificial intelligence monitoring system may only focus on the overall operation status of the power grid, such as "normal" or "abnormal"; at a finer semantic granularity level, the artificial intelligence monitoring system can further distinguish different types of abnormalities, such as "voltage abnormality" and "current overload". By analyzing at different semantic granularity levels, the artificial intelligence monitoring system can provide more accurate and targeted fault warnings and solutions.

[0092] The first grid operation state vector is a mathematical representation generated by the first fault discrimination decision model after mining the power IoT sensor monitoring data to be processed. This vector contains various characteristic values ​​that describe the grid operation state at a specific semantic granular level, such as voltage fluctuation, current stability, power factor, etc. These characteristic values ​​are combined to form a multi-dimensional vector space for quantitatively describing the operation state of the grid. By comparing the first grid operation state vectors at different time points, the artificial intelligence monitoring system can track the changing trend of the grid state and detect abnormal situations in a timely manner.

[0093] Based on this, in the key link of power monitoring, step 120 carries the important task of preliminary mining of the power grid operation status. In this step, the artificial intelligence monitoring system uses its built-in first fault discrimination decision model to conduct in-depth analysis and processing of the power Internet of Things sensor monitoring data to be processed.

[0094] For example, the AI ​​monitoring system will first input the received sensor monitoring data into the first fault discrimination decision model. This model is carefully designed and trained to identify various features and patterns contained in the power data. The model contains a series of complex algorithms and calculation processes to extract key information from the data and convert it into an intuitive description of the power grid operation status.

[0095] When mining the first grid operation status, the model focuses on information at multiple semantically granular levels. Starting from the most macroscopic level, the model determines whether the grid is in normal operation and whether there are obvious abnormalities or signs of failure. As the analysis deepens, the model will further explore more specific status details, such as which specific nodes or devices have abnormal conditions, the type and severity of these abnormalities, etc.

[0096] Finally, the first fault discrimination decision model will integrate all the mined information to generate one or more first power grid operation state vectors. These vectors not only reflect the operation state of the power grid in various aspects in terms of numerical value, but also imply the problems and potential risks that may exist in the power grid. These vectors provide an important data basis for subsequent state consistency analysis and fault warning.

[0097] Therefore, step 120 is a key link in the power monitoring process of the artificial intelligence monitoring system. It conducts in-depth mining of the power Internet of Things sensor monitoring data to be processed through the first fault identification decision model, providing strong technical support for the safe and stable operation of the power grid.

[0098] Correspondingly, the second power grid operation state mining refers to the use of the second fault discrimination decision model, combined with past fault location guidance vectors, to conduct in-depth analysis and feature extraction of the power Internet of Things sensor monitoring data to be processed. This mining process focuses on the knowledge learned from historical fault data to identify the fault modes and abnormal states that may be hidden in the current data. In this way, the second power grid operation state mining can more accurately predict and identify potential problems in the power grid and improve the accuracy and reliability of fault warning.

[0099] The second grid operation state vector is a data representation generated during the mining process of the second grid operation state. This vector contains grid state information obtained through comprehensive analysis of past fault location guidance vectors and current sensor monitoring data. Compared with the first grid operation state vector, the second grid operation state vector pays more attention to the fault characteristics and patterns learned from historical data, so it is more sensitive and accurate in identifying potential grid faults and anomalies. This vector is an important basis for subsequent state consistency analysis and fault warning.

[0100] In the power monitoring artificial intelligence monitoring system, step 130 is a crucial link, which involves using the second fault discrimination decision model and the past fault location guidance vector to conduct in-depth second power grid operation status mining on the processed power Internet of Things sensor monitoring data.

[0101] In step 130, the artificial intelligence monitoring system first calls the second fault discrimination decision model, which is an advanced analysis model that has been integrated with historical fault data and fault location guidance vectors. The artificial intelligence monitoring system inputs the sensor monitoring data to be processed into this model. At the same time, the model refers to the past fault location guidance vectors, which are valuable experiences extracted from a large number of historical fault events. They provide the model with "memory" and "wisdom" to identify fault modes.

[0102] In the second grid operation status mining process, the model will carefully analyze the similarities between current data and historical fault data to find possible hidden fault signs. This analysis method not only focuses on the overall status of the grid, but also goes deep into each semantic granular level to comprehensively evaluate the grid status from multiple dimensions.

[0103] Finally, the second fault discrimination decision model will output one or more second power grid operation state vectors. These vectors combine real-time data and historical experience, not only reflecting the current state of the power grid, but also predicting possible development trends and potential risks. These vectors provide a more accurate and reliable basis for subsequent state consistency analysis and fault warning.

[0104] In other words, step 130 is an advanced analysis function of the artificial intelligence monitoring system in power monitoring. It combines historical fault data and real-time sensor monitoring data to conduct in-depth grid status mining, providing more powerful technical support for the safe and stable operation of the power grid.

[0105] In step 140, state consistency analysis is used to evaluate the degree of consistency between two or more data sources in describing the state of the same object or artificial intelligence monitoring system. In the field of power grid monitoring, state consistency analysis specifically refers to comparing the similarities and differences between power grid operation state vectors derived from different analysis models (such as the first fault discrimination decision model and the second fault discrimination decision model). This analysis aims to verify whether the assessment of the power grid state by different models is consistent, and to identify any possible differences or contradictions, thereby providing a more comprehensive assessment of the power grid state.

[0106] The state similarity and difference analysis viewpoint is a viewpoint or conclusion that interprets and evaluates the similarities and differences between the power grid operation state vectors based on the results of the state consistency analysis. These viewpoints may include the overall consistency of the power grid state, local differences, identification of potential fault points, and the necessity of fault warning. The state similarity and difference analysis viewpoint provides power grid operation and maintenance personnel with important information about the current state of the power grid and possible future development trends, which helps guide subsequent maintenance, inspection and fault prevention work.

[0107] In the workflow of the power monitoring artificial intelligence monitoring system, step 140 is a key analysis step, which involves performing a state consistency analysis on the first power grid operation state vector and the second power grid operation state vector, and drawing state similarities and differences analysis views therefrom.

[0108] In step 140, the artificial intelligence monitoring system first selects the first power grid operation state vector and the second power grid operation state vector of the same semantic granularity level to ensure that the two vectors have the same level of detail and focus when describing the power grid state, so that an effective comparison can be made.

[0109] Next, the AI ​​monitoring system uses algorithms and techniques to perform consistency analysis on the two vectors. This process involves calculating the similarity between the vectors, identifying common features and differences in the vectors, and evaluating the impact of these differences on the overall state of the power grid. Through this analysis, the AI ​​monitoring system can determine whether the two models are consistent in assessing the state of the power grid, and identify any potential differences or contradictions.

[0110] Finally, based on the results of the state consistency analysis, the AI ​​monitoring system will generate state similarities and differences analysis opinions. These opinions may include overall evaluation of the power grid state, identification of local abnormal areas, early warning of potential fault points, etc. These opinions provide valuable information for power grid operation and maintenance personnel, helping them to better understand the current state of the power grid and possible future development trends, so as to develop more effective maintenance strategies and fault prevention measures.

[0111] It can be seen that step 140 is a crucial analysis link in the power monitoring artificial intelligence monitoring system. It reveals the similarities and differences between the first power grid operation state vector and the second power grid operation state vector through state consistency analysis, providing strong data support and decision-making basis for the safe and stable operation of the power grid.

[0112] Based on the above, fault point location warning can be understood as the process of using advanced technical means to detect, identify and predict potential fault points in the power grid or other artificial intelligence monitoring systems, and issue warnings in a timely manner. In the power system, the fault point can be a problem point such as broken wires, equipment failure, poor contact, etc. that may cause power supply interruption or equipment damage. The positioning warning artificial intelligence monitoring system can accurately find possible fault points by collecting and analyzing various sensor data, operating status information, etc., and send warning information to the operation and maintenance personnel in a timely manner before or when the fault occurs, so as to quickly respond and repair, thereby ensuring the stable operation of the power system.

[0113] It can be understood that step 150 is a key link in the power monitoring artificial intelligence monitoring system, namely, fault point location and early warning. In this step, the artificial intelligence monitoring system uses its powerful data processing and analysis capabilities to conduct in-depth fault point location and early warning on the power IoT sensor monitoring data to be processed based on the state similarities and differences analysis viewpoints obtained in the previous steps.

[0114] Exemplarily, the artificial intelligence monitoring system first reviews and analyzes the state similarities and differences analysis viewpoints obtained after consistency analysis of the first power grid operation state vector and the second power grid operation state vector. These viewpoints reveal possible abnormalities or potential fault areas in the power grid.

[0115] Next, the AI ​​monitoring system will conduct a more detailed analysis of these potential fault areas. It will combine real-time power IoT sensor monitoring data and use advanced algorithms and models to accurately locate the fault points in these areas. This process can involve comprehensive consideration of multiple parameters such as voltage fluctuations, current anomalies, and temperature changes.

[0116] Once the AI ​​monitoring system detects that a specific point or device is at risk of failure, it will immediately trigger an early warning mechanism. This warning may be sent to maintenance personnel in the form of sound, light signals or electronic messages, ensuring that they can respond quickly and take measures to prevent or repair the failure.

[0117] In addition, the fault location warning artificial intelligence monitoring system may also have automatic or semi-automatic fault isolation functions to reduce the impact of the fault on the entire power grid artificial intelligence monitoring system. While issuing a warning, the artificial intelligence monitoring system can try to automatically adjust the operating status of the power grid to reduce the pressure on potential fault points, or isolate the fault area to prevent the fault from spreading.

[0118] It can be seen that the fault point location and early warning function of step 150 is crucial to ensure the stable operation of the power system. It can not only detect and warn potential fault points in a timely manner, but also provide accurate information and guidance to operation and maintenance personnel so that they can respond to and handle fault situations quickly and effectively.

[0119] In the power monitoring artificial intelligence monitoring system, the past fault location guidance vector, the first power grid operation status vector and the second power grid operation status vector are key data structures, which reflect the historical fault information and current operation status of the power grid in numerical form.

[0120] Past fault location guidance vectors can be generated based on historical fault data and contain a series of values ​​to guide the AI ​​monitoring system to quickly locate new faults. For example, a past fault location guidance vector can be: [0.8, 0.3, 0.1, 0.5, 0.9], where each value represents the probability or importance of a fault in the corresponding grid area. These values ​​are calculated based on multiple factors such as historical fault frequency, severity, and repair time.

[0121] The first power grid operation state vector is a vector generated by mining the current power Internet of Things sensor monitoring data through the first fault discrimination decision model. It reflects the operation state of the power grid at a specific semantic granular level. For example, a first power grid operation state vector can be: [0.7, 0.2, 0.6, 0.8, 0.5], where each value represents the operation state of different aspects of the power grid, such as voltage stability, current balance, power factor, etc. The closer these values ​​are to 1, the better the state of the corresponding aspect; otherwise, there may be potential problems.

[0122] The second grid operation state vector is similar to the first grid operation state vector, but it focuses more on the fault characteristics and patterns learned from historical data. Therefore, the generation of the second grid operation state vector will refer to the past fault location guidance vector. A typical second grid operation state vector can be: [0.6, 0.4, 0.5, 0.7, 0.4]. These values ​​also represent the operating status of different aspects of the grid, but more historical fault information is incorporated, allowing the artificial intelligence monitoring system to more accurately identify and predict potential fault points.

[0123] In practical applications, these vectors will be analyzed for state consistency to evaluate the overall operating status of the power grid and identify potential fault points. For example, by comparing the similarities and differences between the first power grid operating state vector and the second power grid operating state vector, the artificial intelligence monitoring system can identify possible problem areas in the power grid and issue early warnings in a timely manner. This vector-based numerical analysis method provides strong data support and decision-making basis for the safe and stable operation of the power grid.

[0124] By applying the embodiment of the present application, firstly, by obtaining the power IoT sensor monitoring data to be processed and determining two fault discrimination decision models, the embodiment of the present application provides a solid data foundation for subsequent power grid status analysis and fault warning. In particular, the second fault discrimination decision model has learned the past fault location guidance vector of the fault location sensor monitoring data, making the model more accurate and forward-looking in fault identification.

[0125] Secondly, the embodiment of the present application uses two models to conduct in-depth state mining on the power Internet of Things sensor monitoring data to be processed, and obtains the first power grid operation state vector and the second power grid operation state vector at at least one semantic fine-grained level. These two vectors comprehensively reflect the operation state of the power grid from different angles, providing rich information for subsequent state consistency analysis and fault point location warning.

[0126] Furthermore, the embodiment of the present application performs state consistency analysis on two power grid operation state vectors of the same semantic fine-grained level and obtains state similarities and differences analysis viewpoints. This step not only reveals potential problems and risk points in power grid operation, but also provides targeted improvement and optimization suggestions for operation and maintenance personnel.

[0127] Finally, based on the state similarities and differences analysis point of view, the embodiment of the present application can realize the fault point location warning for the processing of the power Internet of Things sensor monitoring data. This function is of great value in practical applications. It can help operation and maintenance personnel to timely discover and handle the fault points in the power grid, thereby ensuring the safe and stable operation of the power grid and reducing power outages and economic losses caused by faults.

[0128] To sum up, the embodiments of the present application realize comprehensive analysis and efficient utilization of power Internet of Things sensor monitoring data by comprehensively using two fault identification decision models, power grid operation status mining, status consistency analysis and fault point location warning and other technical means, providing strong technical guarantee for the safe and stable operation of the power grid.

[0129] In some optional embodiments, the second fault discrimination decision model includes at least one semantic fine-grained level operating status mining branch, and each of the semantic fine-grained level operating status mining branches has learned the past fault location guidance vectors of the fault location sensor monitoring data at the corresponding semantic fine-grained level; then based on the second fault discrimination decision model and according to the past fault location guidance vectors, the second power grid operating status mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operating status vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level, including: through the operating status mining branches at each semantic fine-grained level, according to the past fault location guidance vectors at the corresponding semantic fine-grained level, the second power grid operating status mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operating status vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level.

[0130] In this embodiment, the second fault discrimination decision model used by the artificial intelligence monitoring system is designed to be more sophisticated and complex. The second fault discrimination decision model is not a single structure, but contains multiple operating status mining branches for different semantic granularity levels. Such a design enables the model to more accurately capture and analyze various subtle changes in the power Internet of Things sensor monitoring data.

[0131] Each operation status mining branch is dedicated to a specific semantic granularity level, which means that they can focus on mining the grid operation status information at that level. Importantly, these branches also enrich their discriminative capabilities by learning the corresponding past fault location guidance vectors. These guidance vectors are valuable experience extracted from historical fault data, and they reflect the typical characteristics and patterns of grid failures at a specific semantic granularity level.

[0132] When the artificial intelligence monitoring system receives the pending power IoT sensor monitoring data, the second fault discrimination decision model will be activated. At this time, the operation status mining branches at each semantic fine-grained level will conduct in-depth mining of the data based on the past fault location guidance vectors that they have learned. In this process, each branch will output a second power grid operation status vector corresponding to the semantic fine-grained level that it is responsible for.

[0133] These second grid operation state vectors not only reflect the current operation state of the grid, but also imply the risk and trend of possible failures. Because each vector is generated under the guidance of past failure experience at the corresponding semantic granular level, they are highly sensitive and accurate for early identification and warning of failures.

[0134] By integrating these second grid operation state vectors at different semantic granularity levels, the AI ​​monitoring system can build a comprehensive and detailed grid status portrait. This enables the AI ​​monitoring system to detect abnormal conditions in the grid in a timely manner and respond quickly, thereby effectively preventing or reducing the occurrence of faults.

[0135] In this way, not only the accuracy and efficiency of fault identification are improved, but also the stability and safety of power grid operation are enhanced. Multiple semantically fine-grained operation status mining branches work together to enable the artificial intelligence monitoring system to capture more detailed information and detect potential fault risks at an early stage. At the same time, since each branch learns based on past fault experience, the artificial intelligence monitoring system's ability to identify various fault modes has also been significantly improved. These advantages work together to enable the artificial intelligence monitoring system to perform well in fault identification and early warning of power IoT sensor monitoring data, providing a strong guarantee for the safe and stable operation of the power grid.

[0136] In some other optional embodiments, the second fault discrimination decision model includes a first operating state mining branch at a first semantic fine-grained level, and the first operating state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level. Then, the operating state mining branches at each semantic fine-grained level perform second power grid operating state mining on the power IoT sensor monitoring data to be processed according to the past fault location guidance vector at the corresponding semantic fine-grained level, and obtain the second power grid operating state vector of the power IoT sensor monitoring data to be processed at at least one semantic fine-grained level, including: obtaining the potential fault search features of the power IoT sensor monitoring data to be processed at the first semantic fine-grained level; based on the potential fault search features, perform potential fault search in the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label; based on the first operating state mining branch, perform second power grid operating state mining according to the potential fault search label to obtain the second power grid operating state vector of the power IoT sensor monitoring data to be processed at the first semantic fine-grained level.

[0137] In other specific implementation cases, the second fault discrimination decision model adopted by the artificial intelligence monitoring system has a more detailed structure, especially for the first semantic granularity level. This model specially sets up a component called the first operating state mining branch, which has mastered the past fault location guidance vector at the first semantic granularity level through learning.

[0138] When the AI ​​monitoring system receives the pending power IoT sensor monitoring data, the first operating state mining branch of the second fault discrimination decision model begins its unique workflow. First, this branch extracts potential fault search features at the first semantic granular level from the pending power IoT sensor monitoring data. These features may be key indicators such as voltage fluctuations and current anomalies that can reflect potential problems in the power grid.

[0139] After extracting these features, the first operating state mining branch will search for potential faults in the learned past fault location guidance vectors based on these features. This process is similar to querying matching information in a database, but it is more complex and accurate because it involves multi-dimensional data analysis and pattern matching. The result of the search is one or more potential fault search tags, which identify the possible fault types and locations in the data.

[0140] With these potential fault search labels, the first operating state mining branch will further use these labels to mine the second power grid operating state. This process aims to deeply analyze the actual operating state of the power grid at the first semantic granular level, especially the part related to potential faults. Through the second power grid operating state mining, the branch will generate a second power grid operating state vector for the first semantic granular level. The second power grid operating state vector not only contains information about the current state of the power grid, but also reflects the type, severity and possible development trend of potential faults. This is extremely important for operation and maintenance personnel because it provides an accurate and efficient fault warning and location mechanism.

[0141] In this way, the AI ​​monitoring system can more accurately identify and warn of potential faults in the power grid. The first operating state mining branch combines the potential fault search features of the first semantic fine-grained level and the past fault location guidance vector to achieve in-depth analysis of the power grid state and accurate prediction of faults. This not only improves the operating efficiency and safety of the power system, but also reduces the risk of faults and maintenance costs. Overall, the implementation of this technical solution provides strong support for the intelligent management and maintenance of the power grid.

[0142] In the next step, the past fault location guidance vector at the first semantic fine-grained level includes a past fault location identification vector and a past fault location attribute vector associated with the past fault location identification vector; based on the potential fault search feature, a potential fault search is performed on the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label, including: determining a feature commonality value between the potential fault search feature and the past fault location identification vector; determining a search confidence of the potential fault search feature based on the feature commonality value; and based on the search confidence, performing a feature splicing operation on the past fault location attribute vector to obtain a potential fault search label with feature splicing completed.

[0143] In the workflow of the AI ​​monitoring system, when it comes to searching for potential faults at the first semantic fine-grained level, the AI ​​monitoring system uses a past fault location guidance vector that contains rich information. The past fault location guidance vector actually consists of two parts: a past fault location identification vector and an associated past fault location attribute vector.

[0144] The past fault location identification vector can be regarded as a "fingerprint" of a fault, which uniquely identifies each fault type. The past fault location attribute vector provides more detailed information about these faults, such as the time, location, and severity of the fault.

[0145] In the process of potential fault search, the AI ​​monitoring system will first analyze the potential fault search features in the power IoT sensor monitoring data to be processed. Potential fault search features can be sudden drops in voltage, abnormal fluctuations in current, etc., which are all signs of AI monitoring system failure.

[0146] Next, the AI ​​monitoring system calculates the feature commonality value between these potential fault search features and the past fault location identification vectors. The feature commonality value reflects the similarity between the current data and the historical fault data. If the feature commonality value is high, it means that the current potential fault may be very similar to a historical fault.

[0147] The AI ​​monitoring system will determine the search confidence of the potential fault search feature based on this feature commonality value. Search confidence is actually an indicator to measure the accuracy of the AI ​​monitoring system's judgment of the current potential fault. If the feature commonality value is high, the search confidence will also increase accordingly.

[0148] With the search confidence, the artificial intelligence monitoring system will perform feature splicing operations on the past fault location attribute vectors based on this confidence. The feature splicing operation actually combines the features of the current potential fault with the features of similar faults in the past to obtain a more complete and accurate fault description.

[0149] After completing feature splicing, the artificial intelligence monitoring system obtains a potential fault search label that completes feature splicing. The potential fault search label not only contains the information of the current potential fault, but also integrates the experience of similar faults in history, providing valuable fault warning and location basis for operation and maintenance personnel.

[0150] In this way, by combining past fault location identification vectors and past fault location attribute vectors to search for potential faults, the artificial intelligence monitoring system can more accurately identify and warn of potential faults in the power grid. This search method based on historical data not only improves the accuracy of fault identification, but also enhances the ability of the artificial intelligence monitoring system to respond to unknown faults. The feature splicing operation perfectly combines current data with historical data, providing comprehensive and accurate fault information for operation and maintenance personnel, thereby greatly improving the stability and safety of the power system.

[0151] Under some other preferred design ideas, the potential fault search features of the power Internet of Things sensor monitoring data to be processed at the first semantic fine-grained level are obtained, including: performing semantic fine-grained level update processing on the first power grid operation state vector to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel; performing state vector integration on the updated power grid operation state vector under the target attention channel to obtain a power grid operation state integration vector; based on the first operation state mining branch and according to the power grid operation state integration vector, generating the potential fault search features of the power Internet of Things sensor monitoring data to be processed at the first semantic fine-grained level.

[0152] Under some other preferred design ideas, the process of the artificial intelligence monitoring system obtaining the potential fault search features of the power Internet of Things sensor monitoring data to be processed at the first semantic fine-grained level presents more detailed and efficient characteristics.

[0153] First, the artificial intelligence monitoring system will perform preliminary processing on the pending power IoT sensor monitoring data, that is, obtain the first power grid operation state vector through the first operation state mining branch. However, the first power grid operation state vector may not be sufficient to accurately reflect the characteristics of potential faults, so further processing and treatment is required.

[0154] Next, the artificial intelligence monitoring system will perform semantic fine-grained level update processing on the first power grid operation state vector. The purpose of semantic fine-grained level update processing is to make the expression of the first power grid operation state vector more accurate and detailed, so as to better capture the subtle changes in the power grid operation state. The updated power grid operation state vector has at least one attention channel, which can help the artificial intelligence monitoring system focus more on those features that are closely related to potential faults.

[0155] After completing the update process of the first power grid operation state vector, the artificial intelligence monitoring system will perform state vector integration on the updated power grid operation state vector under the target attention channel. This step is equivalent to "refining" and "condensing" the vector, extracting the most critical and representative information from it to form a power grid operation state integration vector. The power grid operation state integration vector not only contains the core information of the original data, but also strengthens and highlights the features related to potential faults through the processing of the attention channel.

[0156] Finally, based on the first operating state mining branch and the integrated vector of the power grid operating state, the artificial intelligence monitoring system can generate potential fault search features at the first semantic fine-grained level for the power IoT sensor monitoring data to be processed. The potential fault search features are comprehensive and accurate, providing strong support for subsequent fault search and location.

[0157] By applying the above design ideas, the artificial intelligence monitoring system can more accurately and efficiently identify potential fault features in the power IoT sensor monitoring data. This not only improves the accuracy of the artificial intelligence monitoring system's fault warning, but also greatly shortens the time for fault discovery and processing. At the same time, since the artificial intelligence monitoring system can automatically process data and extract features, it also greatly reduces the workload of operation and maintenance personnel and improves the overall operation efficiency and safety of the power system.

[0158] In another possible embodiment, the second fault discrimination decision model includes a second operating state mining branch at a second semantic fine-grained level, and the second operating state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the second semantic fine-grained level. Then, the operating state mining branches at each semantic fine-grained level perform second power grid operating state mining on the power IoT sensor monitoring data to be processed according to the past fault location guidance vector at the corresponding semantic fine-grained level, and obtain the second power grid operating state vector of the power IoT sensor monitoring data to be processed at at least one semantic fine-grained level, including: based on the second operating state mining branch, generate the basic power grid operating state vector of the power IoT sensor monitoring data to be processed at the second semantic fine-grained level; based on the past fault location guidance vector, generate the fault dynamic search feature of the basic power grid operating state vector at the second semantic fine-grained level; based on the basic power grid operating state vector, perform feature splicing operation on the fault dynamic search feature to generate the second power grid operating state vector of the power IoT sensor monitoring data to be processed at the second semantic fine-grained level.

[0159] In another embodiment of the artificial intelligence monitoring system, a second fault discrimination decision model is introduced, and the second fault discrimination decision model has a second operation state mining branch at a second semantic fine-grained level. The second operation state mining branch has learned and deeply understood the past fault location guidance vectors of the fault location sensor monitoring data at the second semantic fine-grained level.

[0160] When the artificial intelligence monitoring system receives the pending power IoT sensor monitoring data, the second operation state mining branch will start its unique workflow. First, the second operation state mining branch will generate the basic grid operation state vector of the pending power IoT sensor monitoring data at the second semantic fine-grained level based on the knowledge and model it has learned. The basic grid operation state vector is a preliminary description of the current state of the grid, which reflects the overall operation of the grid at the second semantic fine-grained level.

[0161] Next, the second operation state mining branch generates fault dynamic search features of the basic power grid operation state vector at the second semantic fine-grained level based on the past fault location guidance vector. Fault dynamic search features are the key to dynamically search for possible faults in the power grid, and they can help the artificial intelligence monitoring system quickly and accurately locate potential fault points.

[0162] Once the fault dynamic search features are generated, the artificial intelligence monitoring system will perform feature splicing operations on the fault dynamic search features based on the basic power grid operation state vector. This process actually combines the information of the basic power grid operation state vector with the information of the fault dynamic search features to generate a more comprehensive and detailed description of the power grid operation status.

[0163] Finally, through feature splicing operations, the artificial intelligence monitoring system obtains the second grid operation state vector of the power IoT sensor monitoring data to be processed at the second semantic fine-grained level. The second grid operation state vector not only contains the basic operation state information of the grid, but also integrates the dynamic search features of faults, providing strong support for subsequent fault identification and location.

[0164] It can be seen that by introducing the second operation state mining branch of the second semantic fine-grained level and combining the past fault location guidance vector to generate fault dynamic search features and feature splicing operations, the artificial intelligence monitoring system can have a deeper understanding of the operation status of the power grid and accurately and quickly locate potential fault points. This not only improves the operating efficiency and safety of the power system, but also provides more accurate and timely fault warning and location information for operation and maintenance personnel, greatly reducing the cost and time of fault detection and repair.

[0165] In some other examples, the fault dynamic search feature includes at least one fault dynamic search sub-vector. Then, according to the basic power grid operation state vector, the fault dynamic search feature is subjected to a feature splicing operation, including: determining a feature commonality value between the basic power grid operation state vector and the fault dynamic search sub-vector; determining a feature splicing coefficient of the fault dynamic search sub-vector according to the feature commonality value; and performing a feature splicing operation on the fault dynamic search sub-vector according to the feature splicing coefficient.

[0166] In other specific operation examples, the fault dynamic search feature processed by the artificial intelligence monitoring system is actually composed of multiple fault dynamic search sub-vectors. Each fault dynamic search sub-vector describes the potential power grid fault from a different angle or level.

[0167] When the AI ​​monitoring system is ready to perform feature splicing operations, it first determines the feature commonality value between the basic power grid operation state vector and each fault dynamic search sub-vector. The feature commonality value is actually a quantitative indicator that reflects the similarity or correlation between the basic power grid operation state and the specific fault dynamic characteristics. For example, if a fault dynamic search sub-vector describes a voltage fluctuation, the AI ​​monitoring system will analyze the current power grid operation state data to see if it is related to this voltage fluctuation and calculate the commonality value between them.

[0168] After calculating the feature commonality value, the artificial intelligence monitoring system will determine the feature splicing coefficient of each fault dynamic search sub-vector based on this value. The feature splicing coefficient determines the importance and influence of each fault dynamic search sub-vector in the feature splicing process. The fault dynamic search sub-vector with a high commonality value will get a larger splicing coefficient, which means that in the final feature splicing result, the information of this fault dynamic search sub-vector will be more retained and emphasized.

[0169] Finally, according to the feature splicing coefficient of each fault dynamic search sub-vector, the artificial intelligence monitoring system will perform a feature splicing operation. The feature splicing operation is similar to superimposing multiple images at different levels of transparency. The information of each fault dynamic search sub-vector will be incorporated into the final feature vector to varying degrees according to its splicing coefficient. In this way, the artificial intelligence monitoring system can generate a comprehensive and detailed state description that contains both basic power grid operation status information and multiple fault dynamic features.

[0170] It can be seen that through the feature splicing operation described above, the artificial intelligence monitoring system can more accurately capture subtle changes and potential faults in power grid operation. The advantage of this method is that it not only relies on basic power grid operation data, but also combines a variety of fault dynamic search features, thereby improving the accuracy of fault warning and diagnosis. In addition, by calculating the feature commonality value and feature splicing coefficient, the artificial intelligence monitoring system can dynamically adjust the weights of different features in the splicing process, so that the final feature vector is closer to the actual current operating status of the power grid. This not only helps to detect and handle power grid faults in a timely manner, but also provides operation and maintenance personnel with richer and more accurate operation and fault information, thereby improving the stability and reliability of the entire power system.

[0171] In some other examples, the past fault location guidance vector at the second semantic fine-grained level includes an update variable indication corresponding to the attention channel update. Based on the past fault location guidance vector, the fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level is generated, including: according to the update variable indication, the basic power grid operation state vector is updated with attention channel to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel; according to the operation state relationship spectrum of the updated power grid operation state vector under the target attention channel, the feature splicing coefficient is determined; according to the feature splicing coefficient, the state vector splicing is performed on the basic power grid operation state vector to generate the fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level.

[0172] In some specific operation examples, the AI ​​monitoring system processes the past fault location guidance vectors at the second semantic granularity level, which contain updated variable indications related to attention channel updates. This means that when the AI ​​monitoring system generates dynamic fault search features, it pays special attention to certain specific aspects of the power grid status data, which may be more likely to reveal potential fault information.

[0173] First, the AI ​​monitoring system will perform attention channel update processing on the basic power grid operation state vector according to the update variable indication. This processing process can be understood as the AI ​​monitoring system "focusing" on a specific part of the power grid state data, such as voltage fluctuations, current anomalies, etc., to obtain an updated power grid operation state vector. This updated vector has multiple attention channels, each of which focuses on different aspects of the power grid state data.

[0174] Next, the AI ​​monitoring system will determine the feature splicing coefficient based on the operating state relationship spectrum of the updated power grid operating state vector under the target attention channel. The operating state relationship spectrum can be regarded as a relationship network between various state parameters of the power grid, which reflects the overall operating state of the power grid. By analyzing the operating state relationship spectrum, the AI ​​monitoring system can determine which state parameters are more important under the current attention channel, and thus give the corresponding feature splicing coefficient.

[0175] Finally, according to the feature splicing coefficient, the artificial intelligence monitoring system will perform state vector splicing on the basic power grid operation state vector. State vector splicing is similar to fusing multiple related state parameter data together to form a more comprehensive and specific description of the power grid state. In this way, the artificial intelligence monitoring system can generate fault dynamic search features of the basic power grid operation state vector at the second semantic granular level.

[0176] In this way, the AI ​​monitoring system can more accurately locate and identify potential faults in the power grid. By introducing attention channel updating and feature splicing technology, the AI ​​monitoring system can "focus" on the key parts of the power grid status data and generate more comprehensive and specific fault dynamic search features. This not only improves the accuracy of fault detection, but also provides operation and maintenance personnel with richer fault information and positioning guidance. Therefore, this technical solution can significantly improve the intelligence level of power grid monitoring and enhance the stability and safety of power grid operation.

[0177] In some alternative embodiments, the method further includes: determining fault location sensor monitoring data, a first fault discrimination decision model after debugging, and a second fault discrimination decision model to be debugged, wherein the second fault discrimination decision model has learned the basic past fault location guidance vector of the fault location sensor monitoring data; performing first power grid operation state mining on the fault location sensor monitoring data based on the first fault discrimination decision model to obtain a first power grid operation state vector of the fault location sensor monitoring data at at least one semantic fine-grained level; optimizing the basic past fault location guidance vector based on the first power grid operation state vector to obtain an optimized past fault location guidance vector; and optimizing the basic past fault location guidance vector based on the second The fault discrimination decision model performs second power grid operation state mining on the fault location sensor monitoring data based on the optimized past fault location guidance vector to obtain a second power grid operation state vector of the fault location sensor monitoring data at at least one semantic fine-grained level; performs state consistency analysis on the first power grid operation state vector and the second power grid operation state vector at the same semantic fine-grained level to obtain state similarities and differences analysis viewpoints; and debugs the second fault discrimination decision model based on the state similarities and differences analysis viewpoints to determine a debugged second fault discrimination decision model, wherein the debugged second fault discrimination decision model has learned the optimized past fault location guidance vector of the fault location sensor monitoring data.

[0178] In this embodiment, the artificial intelligence monitoring system performs the following series of steps to optimize and improve the fault identification decision model, thereby improving the accuracy and efficiency of power grid fault detection.

[0179] First, the artificial intelligence monitoring system identified several key elements: fault location sensor monitoring data, which is the basic data for real-time monitoring of the power grid status; the first fault discrimination decision model after debugging, which is a verified and optimized model that can more accurately identify power grid faults; and the second fault discrimination decision model to be debugged. Although the second fault discrimination decision model to be debugged has learned the basic past fault location guidance vectors of the fault location sensor monitoring data, it still needs further debugging to improve its discrimination accuracy.

[0180] Next, the artificial intelligence monitoring system uses the first fault discrimination decision model to mine the first power grid operation state of the fault location sensor monitoring data. The first power grid operation state mining is to extract the operation state characteristics of the power grid at different semantic granular levels by analyzing the sensor data to form a first power grid operation state vector. The first power grid operation state vector reflects the real-time state of the power grid under various conditions.

[0181] Then, the artificial intelligence monitoring system optimizes the basic past fault location guidance vector based on the first power grid operation state vector. This is done by combining real-time state data with past fault location experience to update and improve the guidance information for fault location, thereby obtaining an optimized past fault location guidance vector.

[0182] Afterwards, the artificial intelligence monitoring system uses the second fault discrimination decision model to mine the second power grid operation state of the fault location sensor monitoring data based on the optimized past fault location guidance vector. The purpose of this step is to verify whether the optimized guidance vector can effectively improve the model's ability to identify the power grid state and obtain the second power grid operation state vector.

[0183] Subsequently, the artificial intelligence monitoring system performs state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic fine-grained level. This analysis process compares the similarities and differences between the two vectors to obtain state similarities and differences analysis views. This view reflects whether the two models have consistent judgments on the power grid state under the same conditions and what differences exist.

[0184] Finally, the artificial intelligence monitoring system debugs the second fault discrimination decision model based on the state similarities and differences analysis point of view. The purpose of debugging is to reduce the differences and errors in model judgments and make it more consistent with the actual power grid state. After debugging, the second fault discrimination decision model has learned the optimized past fault location guidance vector, so its fault discrimination accuracy and efficiency have been significantly improved.

[0185] In this way, not only the accuracy of the fault identification decision model is improved, but also the adaptability and robustness of the model are enhanced. The optimized model can better combine real-time grid status data and past fault location experience, so as to more accurately identify and predict grid faults. The implementation of this technical solution not only improves the reliability and safety of grid operation, but also provides more accurate fault location and prediction information for operation and maintenance personnel, which helps to reduce the time and economic costs of fault detection and repair.

[0186] In some other alternative embodiments, the second fault discrimination decision model includes a first operating state mining branch at a first semantic fine-grained level, and the first operating state mining branch has learned the basic past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level; based on the first power grid operating state vector, the basic past fault location guidance vector is optimized, including: obtaining the fault location authentication feature of the fault location sensor monitoring data at the first semantic fine-grained level; based on the fault location authentication feature, performing a feature splicing operation on the basic past fault location guidance vector to obtain a fault location interaction vector; based on the difference between the fault location authentication feature and the fault location interaction vector, the basic past fault location guidance vector is optimized.

[0187] For another example, the second fault discrimination decision model relied on by the artificial intelligence monitoring system has a more detailed structure, especially the second fault discrimination decision model includes a first operation state mining branch for the first semantic fine-grained level. The first operation state mining branch has mastered the past fault location guidance vector based on the fault location sensor monitoring data at the first semantic fine-grained level through learning.

[0188] When the AI ​​monitoring system needs to optimize the basic past fault location guidance vector, it will first obtain the fault location authentication features of the fault location sensor monitoring data at the first semantic granular level. The fault location authentication features are extracted after in-depth analysis of the sensor monitoring data, and can accurately reflect the real-time status of the power grid and possible fault information at the first semantic granular level.

[0189] Next, the AI ​​monitoring system will use the fault location authentication features to perform feature splicing operations on the basic past fault location guidance vectors. The feature splicing operation is equivalent to integrating the real-time grid status information with the historical fault location experience to form a more comprehensive and detailed fault location interaction vector. The fault location interaction vector not only contains the guidance information of historical faults, but also integrates the current grid status characteristics.

[0190] Then, the AI ​​monitoring system will optimize the basic past fault location guidance vector based on the difference between the fault location authentication feature and the fault location interaction vector. The process of optimizing the basic past fault location guidance vector is actually a dynamic adjustment process. The AI ​​monitoring system will continuously update and improve the guidance vector based on the difference between the real-time power grid status and the historical fault location experience, making it more suitable for the current power grid status.

[0191] Through the above optimization process, the first operating state mining branch of the second fault discrimination decision model can more accurately identify and locate faults in the power grid. At the same time, since the optimization process is based on real-time power grid state data, the model can adapt to changes in power grid state more quickly and improve the real-time and accuracy of fault detection.

[0192] With this design, the artificial intelligence monitoring system not only improves the accuracy of the fault identification decision model through continuous learning and optimization, but also makes it more adaptive. Especially at the first semantic fine-grained level, the fault identification decision model can make full use of the real-time fault location authentication features to optimize the past fault location guidance vectors, so as to more accurately identify and locate faults in the power grid. The implementation of this technical solution has significantly improved the efficiency and accuracy of power grid fault detection, and provided a strong technical guarantee for the safe and stable operation of the power grid.

[0193] In some other replaceable embodiments, obtaining the fault location authentication features of the fault location sensor monitoring data at the first semantic fine-grained level includes: determining the upstream and downstream semantic fine-grained levels of the first semantic fine-grained level; and determining the fault location authentication features of the fault location sensor monitoring data at the first semantic fine-grained level based on the first power grid operation state vectors of the fault location sensor monitoring data at the upstream and downstream semantic fine-grained levels.

[0194] Furthermore, the artificial intelligence monitoring system adopts a more sophisticated and comprehensive approach when acquiring fault location authentication features of fault location sensor monitoring data at the first semantic granular level.

[0195] First, the artificial intelligence monitoring system will determine the upstream and downstream semantic granularity levels of the first semantic granularity level. The semantic granularity level can be understood as the level of detail or level of data analysis. Different semantic granularity levels focus on different aspects or details of the data. The upstream and downstream semantic granularity levels refer to the more macroscopic or microscopic levels associated with the first semantic granularity level.

[0196] Next, the artificial intelligence monitoring system will determine the fault location authentication features at the first semantic granular level based on the first grid operation state vector of the fault location sensor monitoring data at these upstream and downstream semantic granular levels. In other words, the artificial intelligence monitoring system not only considers the data features at the current semantic granular level, but also comprehensively considers the broader or more specific data states associated with it. In this way, the artificial intelligence monitoring system can capture the real-time status of the power grid more comprehensively, thereby more accurately identifying and locating faults.

[0197] For example, the AI ​​monitoring system can first analyze the power grid operation state vectors at the upstream and downstream semantic granular levels to find out the features or patterns related to the first semantic granular level. Then, these features and patterns, as well as the sensor monitoring data of the first semantic granular level itself, are combined to comprehensively determine the fault location authentication features. This process can involve multiple complex steps such as data fusion, feature extraction, and pattern recognition.

[0198] Ultimately, by comprehensively considering information at multiple semantically fine-grained levels, the AI ​​monitoring system is able to generate more comprehensive and accurate fault location authentication features, thereby improving the accuracy of its fault identification and location.

[0199] This design can more comprehensively capture and analyze the real-time status data of the power grid, thereby generating more accurate and targeted fault location authentication features. This not only improves the accuracy of fault detection of the artificial intelligence monitoring system, but also enhances the artificial intelligence monitoring system's ability to perceive and understand complex power grid conditions. This cross-semantic fine-grained data analysis method provides new ideas and technical means for power grid fault detection, which helps to improve the safety and stability of power grid operation.

[0200] In an independent embodiment, the method of performing fault point location warning on the power Internet of Things sensor monitoring data to be processed based on the state similarities and differences analysis point of view includes: obtaining a first state jump description variable and a first state offset distribution of the power Internet of Things sensor monitoring data to be processed based on the state similarities and differences analysis point of view, and obtaining a second state jump description variable and a second state offset distribution of the fault location sensor monitoring data; determining whether the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data based on the first state jump description variable and the second state jump description variable, and According to the first state offset distribution and the second state offset distribution, determine whether the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed and the abnormal sensor monitoring node of the fault location sensor monitoring data match; if the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data and the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data, then determine the IOT fault point of the power Internet of Things sensor monitoring data to be processed according to the abnormal sensor monitoring node of the fault location sensor monitoring data.

[0201] For example, the artificial intelligence monitoring system can use the state difference analysis point of view to locate the fault point and warn the processing of power Internet of Things sensor monitoring data.

[0202] First, the artificial intelligence monitoring system obtains the state jump description variables and state offset distribution of the power IoT sensor monitoring data to be processed and the fault location sensor monitoring data based on the state similarity and difference analysis point of view. The state jump description variable can be understood as a quantitative indicator describing the sudden change of the data state, while the state offset distribution reflects the deviation of the data state from the normal range.

[0203] For example, the artificial intelligence monitoring system extracts the first state transition description variable and the first state offset distribution of the power IoT sensor monitoring data to be processed. The first state transition description variable and the first state offset distribution respectively describe the state transition degree of the data and the offset of the node data. At the same time, the artificial intelligence monitoring system also obtains the second state transition description variable and the second state offset distribution of the fault location sensor monitoring data for comparison and reference.

[0204] Next, the AI ​​monitoring system compares the state transition description variables of the two sets of data to determine whether the state transition trend of the power IoT sensor monitoring data to be processed matches the state transition trend of the fault location data. At the same time, the AI ​​monitoring system also compares the state offset distribution of the two sets of data to determine whether the abnormal sensor monitoring nodes in the power IoT sensor monitoring data to be processed match the abnormal nodes in the fault location data.

[0205] If the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data, and the abnormal sensor monitoring nodes of the two are also consistent, then the artificial intelligence monitoring system can determine the Internet of Things (IOT) fault point in the power Internet of Things sensor monitoring data to be processed based on the abnormal sensor monitoring nodes in the fault location sensor monitoring data.

[0206] The above process is highly automated and can complete the analysis of a large amount of data in a short time and accurately locate possible fault points, thereby greatly improving the efficiency and accuracy of fault detection. In this way, the artificial intelligence monitoring system can efficiently analyze the pending power Internet of Things sensor monitoring data, and accurately locate potential IoT fault points by comparing it with the fault location sensor monitoring data. The application of this method not only improves the automation level of fault detection, but also significantly shortens the time for fault discovery and processing, providing a strong guarantee for the stable operation of the power system. At the same time, through the comprehensive analysis of state jumps and state offsets, the artificial intelligence monitoring system can more comprehensively evaluate the health status of the power Internet of Things sensor network, providing important data support for preventive and predictive maintenance.

[0207] In an independent embodiment, determining whether the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data based on the first state jump description variable and the second state jump description variable includes: determining a first matching weight between the first state jump description variable and the second state jump description variable; if the first matching weight is greater than a first set threshold, determining that the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data; determining whether the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data based on the first state offset distribution and the second state offset distribution includes: determining a second matching weight between the first state offset distribution and the second state offset distribution; if the second matching weight is greater than a second set threshold, determining that the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data.

[0208] In an independent embodiment, if the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data and the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data, then the IOT fault point of the power Internet of Things sensor monitoring data to be processed is determined according to the abnormal sensor monitoring node of the fault location sensor monitoring data, including: determining the third matching weight between the power Internet of Things sensor monitoring data to be processed and the fault location sensor monitoring data according to the first matching weight and the second matching weight; if the third matching weight is greater than the third set threshold, determining the IOT fault point of the power Internet of Things sensor monitoring data to be processed according to the abnormal sensor monitoring node of the fault location sensor monitoring data.

[0209] In detail, the artificial intelligence monitoring system can also comprehensively judge whether the state jump trend between the power Internet of Things sensor monitoring data to be processed and the fault location sensor monitoring data is consistent based on two key state jump description variables, namely the first state jump description variable and the second state jump description variable. The implementation of this process is first to determine the first matching weight between the two state jump description variables through a complex algorithm. The first matching weight can be understood as the similarity or consistency of the state jump trends of the two. If the calculated first matching weight is greater than the preset first set threshold, the artificial intelligence monitoring system will determine that the state jump trend of the power Internet of Things sensor monitoring data to be processed is consistent with the state jump trend of the fault location sensor monitoring data, that is, the change trends of the two are matched. Then, the artificial intelligence monitoring system will further analyze whether the abnormal sensor monitoring nodes in the two data match. This is achieved by analyzing the first state offset distribution and the second state offset distribution. Similarly, the artificial intelligence monitoring system will calculate a second matching weight to quantify the consistency between the two state offset distributions. When the second matching weight is greater than the preset second set threshold, the artificial intelligence monitoring system will determine that the abnormal sensor monitoring node in the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node in the fault location sensor monitoring data.

[0210] After confirming that both the state jump trend and the abnormal sensor monitoring node match, the artificial intelligence monitoring system will proceed to the next step, which is to determine the Internet of Things (IOT) fault point in the power Internet of Things sensor monitoring data to be processed based on these matching information. This step is to calculate the third matching weight between the power Internet of Things sensor monitoring data to be processed and the fault location sensor monitoring data by comprehensively considering the first matching weight and the second matching weight. If this third matching weight is greater than the preset third setting threshold, the artificial intelligence monitoring system will accurately locate the Internet of Things fault point in the power Internet of Things sensor monitoring data to be processed based on the abnormal sensor monitoring node identified in the fault location sensor monitoring data.

[0211] Through the above detailed and precise analysis process, the artificial intelligence monitoring system can not only effectively identify abnormal conditions in the power IoT sensor network, but also accurately locate the specific fault point, greatly improving the efficiency of fault detection and processing. This is of great significance for ensuring the stable operation of the power system, improving service quality, and reducing economic losses caused by faults.

[0212] In an independent embodiment, the first state jump description variable includes a third state jump description variable of each first sensor monitoring field in the power Internet of Things sensor monitoring data to be processed, and the second state jump description variable includes a fourth state jump description variable of each second sensor monitoring field in the fault location sensor monitoring data, and each first sensor monitoring field corresponds to a second sensor monitoring field; the first state jump description variable of the power Internet of Things sensor monitoring data to be processed and the second state jump description variable of the fault location sensor monitoring data are obtained, including: respectively mapping the power Internet of Things sensor monitoring data to be processed and the fault location sensor monitoring data to obtain the first sensor monitoring text and the second sensor monitoring text; respectively decomposing the first sensor monitoring text and the second sensor monitoring text into multiple first sensor monitoring fields and multiple second sensor monitoring fields; respectively obtaining the third state jump description variable of each first sensor monitoring field and the fourth state jump description variable of the second sensor monitoring field corresponding to each first sensor monitoring field.

[0213] In this embodiment, the data processed by the artificial intelligence monitoring system mainly includes the to-be-processed power IoT sensor monitoring data and the fault location sensor monitoring data. In order to deeply understand these two types of data, the artificial intelligence monitoring system will first focus on their state transition description variables.

[0214] For example, the power IoT sensor monitoring data to be processed includes a series of first sensor monitoring fields, each of which has a corresponding third state transition description variable. Similarly, the fault location sensor monitoring data also consists of a series of second sensor monitoring fields, each of which is accompanied by a fourth state transition description variable. Importantly, each first sensor monitoring field here corresponds to a second sensor monitoring field.

[0215] When processing the above data, the artificial intelligence monitoring system will first perform field mapping, which is a process of converting data into a text format understandable to the artificial intelligence monitoring system. Through field mapping, the original to-be-processed power IoT sensor monitoring data and fault location sensor monitoring data are converted into the first sensor monitoring text and the second sensor monitoring text.

[0216] Next, the artificial intelligence monitoring system will disassemble the first sensor monitoring text and the second sensor monitoring text into multiple first sensor monitoring fields and multiple second sensor monitoring fields. This is done to be able to analyze the state transition of each field in more detail.

[0217] After the disassembly is completed, the artificial intelligence monitoring system will obtain the third state jump description variable of each first sensor monitoring field and the fourth state jump description variable of the second sensor monitoring field corresponding to each first sensor monitoring field. The third state jump description variable and the fourth state jump description variable reflect the change characteristics of the data in the time series, which is an important basis for the subsequent analysis of faults and anomalies.

[0218] Through the above steps, the artificial intelligence monitoring system can conduct in-depth analysis of the pending power Internet of Things sensor monitoring data and fault location sensor monitoring data, thereby providing accurate data support for subsequent fault location and preventive measures. In this way, not only the accuracy of fault detection is improved, but also the artificial intelligence monitoring system can respond to and solve potential problems more quickly, thereby effectively improving the stability and safety of the power system. At the same time, through the detailed analysis of each sensor monitoring field, the artificial intelligence monitoring system can more accurately identify abnormal situations, reduce the possibility of false alarms and missed alarms, and further improve the operating efficiency and service quality of the power system.

[0219] In an independent embodiment, determining the first matching weight between the first state jump description variable and the second state jump description variable includes: respectively determining the matching weights between the third state jump description variable of each first sensor monitoring field and the fourth state jump description variable of the second sensor monitoring field corresponding to each first sensor monitoring field, to obtain multiple fourth matching weights; and determining the global matching weight of the multiple fourth matching weights as the first matching weight between the first state jump description variable and the second state jump description variable.

[0220] In an independent embodiment, the first state offset distribution of the power Internet of Things sensor monitoring data to be processed includes the third state offset distribution of each sensor monitoring data unit to be processed in the power Internet of Things sensor monitoring data to be processed, and the second state offset distribution of the fault location sensor monitoring data includes the fourth state offset distribution of each fault location sensor monitoring data unit in the fault location sensor monitoring data, and each of the sensor monitoring data units to be processed corresponds to a fault location sensor monitoring data unit; the determination of the second matching weight between the first state offset distribution and the second state offset distribution includes: respectively determining the matching weight between the third state offset distribution of each sensor monitoring data unit to be processed and the fourth state offset distribution of the fault location sensor monitoring data unit corresponding to each sensor monitoring data unit to be processed, to obtain multiple fifth matching weights; and determining the global matching weight of the multiple fifth matching weights as the second matching weight between the first state offset distribution and the second state offset distribution.

[0221] In the above two embodiments, the artificial intelligence monitoring system takes a meticulous approach when determining the first matching weight between the first state transition description variable and the second state transition description variable. First, the artificial intelligence monitoring system pays attention to each first sensor monitoring field in the power IoT sensor monitoring data to be processed, and the third state transition description variable corresponding to the first sensor monitoring field. At the same time, the artificial intelligence monitoring system also checks the second sensor monitoring field corresponding to each first sensor monitoring field in the fault location sensor monitoring data, and the fourth state transition description variable of the second sensor monitoring field.

[0222] Next, the AI ​​monitoring system will determine the matching weights between the third state transition description variable of each first sensor monitoring field and the fourth state transition description variable of its corresponding second sensor monitoring field. This process is achieved through complex algorithms that comprehensively consider multiple factors such as the amplitude, frequency and trend of the state transition to quantify the similarity or difference between the two. In this way, the AI ​​monitoring system will obtain multiple fourth matching weights, each of which reflects the degree of matching between the state transition description variables of the corresponding field.

[0223] Finally, in order to obtain a global evaluation, the AI ​​monitoring system determines the global matching weight of these fourth matching weights. The global matching weight is based on the comprehensive consideration of all fourth matching weights, which represents the overall matching degree between the first state jump description variable and the second state jump description variable, that is, the first matching weight.

[0224] Similarly, when determining the second matching weight between the first state offset distribution and the second state offset distribution, the artificial intelligence monitoring system also adopts a similar method. The artificial intelligence monitoring system will respectively examine the third state offset distribution of each sensor monitoring data unit to be processed in the power Internet of Things sensor monitoring data to be processed, and the fourth state offset distribution of the fault location sensor monitoring data unit corresponding to the sensor monitoring data unit to be processed.

[0225] By calculating the matching weight between the third state offset distribution of each sensor monitoring data unit to be processed and the fourth state offset distribution of its corresponding fault location sensor monitoring data unit, the artificial intelligence monitoring system obtains multiple fifth matching weights. The fifth matching weights reflect the similarity or difference of each data unit in terms of state offset.

[0226] Finally, the AI ​​monitoring system determines the global matching weight of these fifth matching weights as the overall matching degree between the first state offset distribution and the second state offset distribution, i.e., the second matching weight. The global matching weight is derived based on the comprehensive evaluation of all fifth matching weights, and it represents the overall similarity or difference between the two data in terms of state offset distribution.

[0227] In this way, the matching degree between the to-be-processed power IoT sensor monitoring data and the fault location sensor monitoring data in terms of state transition and state offset can be accurately evaluated. This not only helps the artificial intelligence monitoring system to more accurately identify potential fault points, but also provides strong data support for subsequent fault prevention and repair work. Therefore, this meticulous data analysis method is of great significance for improving the stability and safety of the power system.

[0228] In an independent embodiment, the first state jump description variable includes a first peak jump number and a first valley jump number in the power Internet of Things sensor monitoring data to be processed, and the second state jump description variable includes a second peak jump number and a second valley jump number in the fault location sensor monitoring data; determining whether the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data according to the first state jump description variable and the second state jump description variable, including: determining the power Internet of Things to be processed according to the first peak jump number and the second peak jump number. The peak jump matching weight between the sensor monitoring data and the fault location sensor monitoring data; determining the valley jump matching weight between the power Internet of Things sensor monitoring data to be processed and the fault location sensor monitoring data according to the first valley jump number and the second valley jump number; determining the first matching weight between the first state jump description variable and the second state jump description variable according to the peak jump matching weight and the valley jump matching weight; if the first matching weight is greater than the first set threshold, determining that the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data.

[0229] Furthermore, the AI ​​monitoring system will conduct an in-depth analysis of the state transition trends of the power IoT sensor monitoring data and fault location sensor monitoring data to be processed. In order to more accurately describe the state transitions of these two types of data, the AI ​​monitoring system will pay special attention to the peak transitions and valley transitions in the data.

[0230] First, the artificial intelligence monitoring system will calculate the number of first peak jumps and the number of first valley jumps in the power IoT sensor monitoring data to be processed. The number of first peak jumps and the number of first valley jumps reflect the jumps of the maximum and minimum values ​​in the data. Similarly, the artificial intelligence monitoring system will also calculate the number of second peak jumps and the number of second valley jumps in the fault location sensor monitoring data.

[0231] Next, the AI ​​monitoring system will use the number of peak jumps and the number of valley jumps to determine the degree of match between the two data. For example, the AI ​​monitoring system will calculate the peak jump matching weight between the to-be-processed power IoT sensor monitoring data and the fault location sensor monitoring data through a specific algorithm based on the first peak jump number and the second peak jump number. This weight reflects the similarity of the two data in terms of peak jumps.

[0232] Similarly, the artificial intelligence monitoring system will also calculate the valley jump matching weight based on the number of the first valley jump and the number of the second valley jump. The valley jump matching weight represents the similarity of the two data in terms of valley jump.

[0233] After obtaining the peak jump matching weight and the valley jump matching weight, the artificial intelligence monitoring system will comprehensively consider the peak jump matching weight and the valley jump matching weight, and determine the first matching weight between the first state jump description variable and the second state jump description variable through a certain algorithm. The first matching weight is a value that combines the similarity of the peak and valley jumps, and can fully reflect the matching degree of the state jump trend between the to-be-processed power Internet of Things sensor monitoring data and the fault location sensor monitoring data.

[0234] Finally, the artificial intelligence monitoring system will determine whether the first matching weight is greater than the preset first set threshold. If it is greater, it means that the state jump trend of the power IoT sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data, which further indicates that the two data show similar rules in state change, which is very important for subsequent fault location and analysis.

[0235] With this design, the AI ​​monitoring system can more accurately grasp the dynamic characteristics of the power IoT sensor data, thereby improving the sensitivity and accuracy of fault detection. This not only helps to detect potential power failures in a timely manner, but also provides strong data support for the stable operation of the power system.

[0236] Further, Figure 2 This is a schematic diagram of the structure of an artificial intelligence monitoring system 200 provided in an embodiment of the present application. Figure 2The artificial intelligence monitoring system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0237] Alternatively, if Figure 2 As shown, the artificial intelligence monitoring system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0238] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0239] Alternatively, if Figure 2 As shown, the artificial intelligence monitoring system 200 may also include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0240] Optionally, the artificial intelligence monitoring system 200 can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0241] It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiment can be completed by the hardware integrated logic circuit or software instructions in the processor. The above processor can be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined and performed. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, and other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0242] It can be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0243] It should be understood that the above-mentioned memory is exemplary but not restrictive. For example, the memory in the embodiments of the present application may also be static random access memory (static RAM, SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (synch link DRAM, SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DRRAM), etc. That is to say, the memory in the embodiments of the present application is intended to include but not limited to these and any other suitable types of memory.

[0244] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0245] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations.

Claims

1. A smart grid fault monitoring method using artificial intelligence, characterized in that: Applied to an artificial intelligence monitoring system, the method comprises: Acquire the power Internet of Things sensor monitoring data to be processed, and determine a first fault discrimination decision model and a second fault discrimination decision model for fault identification of the power Internet of Things sensor monitoring data to be processed, wherein the second fault discrimination decision model has learned the past fault location guidance vector of the fault location sensor monitoring data; Based on the first fault discrimination decision model, a first power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain a first power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at at least one semantic fine-grained level; Based on the second fault discrimination decision model and according to the past fault location guidance vector, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level; wherein the second fault discrimination decision model includes at least one operation state mining branch at the semantic fine-grained level, and each operation state mining branch at the semantic fine-grained level has learned the past fault location guidance vector of the fault location sensor monitoring data at the corresponding semantic fine-grained level; based on the second fault discrimination decision model and according to the past fault location guidance vector, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level, including: through the operation state mining branches at each semantic fine-grained level, according to the past fault location guidance vector at the corresponding semantic fine-grained level, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed to obtain the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level; Performing state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic fine-grained level to obtain state similarities and differences analysis viewpoints; Based on the state similarities and differences analysis point of view, the fault point location warning is performed on the power Internet of Things sensor monitoring data to be processed; Among them, the fault point location warning is performed on the power Internet of Things sensor monitoring data to be processed based on the state similarities and differences analysis point of view, including: according to the state similarities and differences analysis point of view, obtaining the first state jump description variable and the first state offset distribution of the power Internet of Things sensor monitoring data to be processed, and obtaining the second state jump description variable and the second state offset distribution of the fault location sensor monitoring data; determining whether the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data based on the first state jump description variable and the second state jump description variable, and The first state offset distribution and the second state offset distribution are used to determine whether the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data; if the state jump trend of the power Internet of Things sensor monitoring data to be processed matches the state jump trend of the fault location sensor monitoring data and the abnormal sensor monitoring node of the power Internet of Things sensor monitoring data to be processed matches the abnormal sensor monitoring node of the fault location sensor monitoring data, then the IOT fault point of the power Internet of Things sensor monitoring data to be processed is determined according to the abnormal sensor monitoring node of the fault location sensor monitoring data; Among them, semantic granularity refers to the level of detail of data interpretation; different semantic granularity levels correspond to different levels and levels of detailed state description; The second fault discrimination decision model includes a first operation state mining branch at a first semantic fine-grained level, wherein the first operation state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level; Through the operation state mining branches at each semantic fine-grained level, according to the past fault location guidance vector at the corresponding semantic fine-grained level, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed, and the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level is obtained, including: Acquire potential fault search features of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level; According to the potential fault search feature, perform a potential fault search on the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label; Based on the first operating state mining branch, a second power grid operating state mining is performed according to the potential fault search label to obtain a second power grid operating state vector of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level.

2. The method for monitoring faults in a smart grid using artificial intelligence according to claim 1, characterized in that: The past fault location guidance vector at the first semantic fine-grained level includes a past fault location identification vector and a past fault location attribute vector associated with the past fault location identification vector; According to the potential fault search feature, a potential fault search is performed on the past fault location guidance vector at the first semantic fine-grained level to obtain a potential fault search label, including: Determining a feature commonality value between the potential fault search feature and the past fault location identification vector; Determining the search confidence of the potential fault search feature according to the feature commonality value; According to the search confidence, a feature splicing operation is performed on the past fault location attribute vector to obtain a potential fault search label that completes feature splicing.

3. The method for monitoring faults in a smart grid using artificial intelligence according to claim 1, wherein: Acquiring potential fault search features of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level, including: Performing semantic fine-grained level update processing on the first power grid operation state vector to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel; Performing state vector integration on the updated power grid operation state vector under the target attention channel to obtain a power grid operation state integration vector; Based on the first operating state mining branch and according to the integrated vector of the power grid operating state, a potential fault search feature of the to-be-processed power Internet of Things sensor monitoring data at the first semantic fine-grained level is generated.

4. The method for monitoring faults in a smart grid using artificial intelligence according to claim 1, wherein: The second fault discrimination decision model includes a second operation state mining branch at a second semantic fine-grained level, and the second operation state mining branch has learned the past fault location guidance vector of the fault location sensor monitoring data at the second semantic fine-grained level; Through the operation state mining branches at each semantic fine-grained level, according to the past fault location guidance vector at the corresponding semantic fine-grained level, the second power grid operation state mining is performed on the power Internet of Things sensor monitoring data to be processed, and the second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the at least one semantic fine-grained level is obtained, including: Based on the second operation state mining branch, generating a basic power grid operation state vector of the to-be-processed power Internet of Things sensor monitoring data at the second semantic fine-grained level; Generating a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level according to the past fault location guidance vector; According to the basic power grid operation state vector, a feature splicing operation is performed on the fault dynamic search feature to generate a second power grid operation state vector of the power Internet of Things sensor monitoring data to be processed at the second semantic fine-grained level.

5. The method for monitoring faults in a smart grid using artificial intelligence as claimed in claim 4, characterized in that: The fault dynamic search feature includes at least one fault dynamic search sub-vector; According to the basic power grid operation state vector, a feature splicing operation is performed on the fault dynamic search feature, including: Determining a characteristic commonality value between the basic power grid operation state vector and the fault dynamic search subvector; Determining a characteristic splicing coefficient of the fault dynamic search subvector according to the characteristic commonality value; A feature splicing operation is performed on the fault dynamic search sub-vector according to the feature splicing coefficient.

6. The method for monitoring faults in a smart grid using artificial intelligence according to claim 4, characterized in that: The past fault location guidance vector at the second semantic fine-grained level includes an update variable indication corresponding to the attention channel update; Generating a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level according to the past fault location guidance vector includes: According to the update variable indication, the basic power grid operation state vector is updated by an attention channel to obtain an updated power grid operation state vector, wherein the updated power grid operation state vector has at least one attention channel; Determining a feature splicing coefficient according to an operating state relationship spectrum of the updated power grid operating state vector under the target attention channel; According to the feature splicing coefficient, state vector splicing is performed on the basic power grid operation state vector to generate a fault dynamic search feature of the basic power grid operation state vector at the second semantic fine-grained level.

7. The method for monitoring faults in a smart grid using artificial intelligence according to claim 1, characterized in that: The method further comprises: Determining fault location sensor monitoring data, a debugged first fault discrimination decision model, and a second fault discrimination decision model to be debugged, wherein the second fault discrimination decision model has learned a basic past fault location guidance vector of the fault location sensor monitoring data; Based on the first fault discrimination decision model, the first power grid operation state mining is performed on the fault location sensor monitoring data to obtain a first power grid operation state vector of the fault location sensor monitoring data at at least one semantic fine-grained level; According to the first power grid operation state vector, optimizing the basic past fault location guidance vector to obtain an optimized past fault location guidance vector; Based on the second fault discrimination decision model and in accordance with the optimized past fault location guidance vector, the second power grid operation state mining is performed on the fault location sensor monitoring data to obtain a second power grid operation state vector of the fault location sensor monitoring data at the at least one semantic fine-grained level; Performing state consistency analysis on the first power grid operation state vector and the second power grid operation state vector of the same semantic fine-grained level to obtain state similarities and differences analysis viewpoints; According to the state similarities and differences analysis viewpoint, the second fault discrimination decision model is debugged to determine a debugged second fault discrimination decision model, wherein the debugged second fault discrimination decision model has learned the optimized past fault location guidance vector of the fault location sensor monitoring data; Among them, the second fault discrimination decision model includes a first operating state mining branch at a first semantic fine-grained level, and the first operating state mining branch has learned the basic past fault location guidance vector of the fault location sensor monitoring data at the first semantic fine-grained level; according to the first power grid operation state vector, the basic past fault location guidance vector is optimized, including: obtaining the fault location authentication feature of the fault location sensor monitoring data at the first semantic fine-grained level; according to the fault location authentication feature, performing a feature splicing operation on the basic past fault location guidance vector to obtain a fault location interaction vector; according to the difference between the fault location authentication feature and the fault location interaction vector, the basic past fault location guidance vector is optimized; Among them, obtaining the fault location authentication features of the fault location sensor monitoring data at the first semantic fine-grained level includes: determining the upstream and downstream semantic fine-grained levels of the first semantic fine-grained level; and determining the fault location authentication features of the fault location sensor monitoring data at the first semantic fine-grained level based on the first power grid operation state vector of the fault location sensor monitoring data at the upstream and downstream semantic fine-grained levels.

8. An artificial intelligence monitoring system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 7.

Images