A method and system for monitoring and warning of partial discharge in a high-voltage circuit breaker

Abstract

The present invention provides a method and system for monitoring and warning of partial discharge in a high-voltage circuit breaker, relating to the technical field of partial discharge monitoring. The method includes: obtaining target circuit characteristic information; calling a circuit analysis model for analysis to obtain a target analysis result, including a plurality of target points; dynamically monitoring a first target point through a UHF sensor to obtain first partial discharge information; calling a signal analysis model for processing and analysis to obtain a first signal feature set; activating a fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; and when it is shown that a partial discharge fault occurs at the first target point of the target high-voltage circuit breaker, sending a first warning signal. The present invention solves the technical problems in the prior art that most of the partial discharge monitoring relies on manual experience or fixed rules, lacks intelligent analysis means, is not only prone to misjudgment or missed judgment, but also difficult to accurately locate, resulting in a reduction in the safety of high-voltage circuit breakers.

Description

Technical Field

[0001] The present invention relates to the technical field of partial discharge monitoring, and particularly to a method and system for monitoring and warning partial discharge of a high-voltage circuit breaker. Background Art

[0002] Partial discharge is a common electrical phenomenon, which usually occurs in the insulation defects or aging parts of high-voltage equipment. Partial discharge will not only affect the insulation performance of the high-voltage circuit breaker, but may ultimately lead to equipment failure, power outage or other serious accidents. Therefore, monitoring the partial discharge situation of the high-voltage circuit breaker and giving an early warning before a failure occurs is of great significance for preventing equipment damage and ensuring the operation of the power grid.

[0003] However, most traditional partial discharge monitoring relies on manual experience or fixed rules and lacks intelligent analysis means. This method is not only prone to misjudgment or missed judgment, but also difficult to cope with complex and diverse fault types. The operating environment of modern power systems is becoming increasingly complex, and the fault types are also becoming increasingly diverse. It is difficult to meet the real-time fault analysis requirements of high-voltage circuit breakers by relying solely on manual or static rules; moreover, the internal structure of high-voltage circuit breakers is complex, and it is difficult to accurately locate the position where partial discharge occurs. The traditional fault troubleshooting process often takes a lot of time to find the specific fault point, which will prolong the downtime of the equipment, affect the operation of the entire power grid, and cause huge losses to power supply and economic operation. Summary of the Invention

[0004] This application provides a method and system for monitoring and warning partial discharge of a high-voltage circuit breaker, aiming to solve the technical problems in the prior art that most partial discharge monitoring relies on manual experience or fixed rules, lacks intelligent analysis means, is not only prone to misjudgment or missed judgment, but also difficult to accurately locate, resulting in a reduction in the safety of high-voltage circuit breakers.

[0005] In the first aspect disclosed in this application, a method for monitoring and warning partial discharge of a high-voltage circuit breaker is provided. The method includes: obtaining target circuit characteristic information, where the target circuit characteristic information refers to multi-dimensional characteristic information of a target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker; calling a circuit analysis model to analyze the target circuit characteristic information to obtain a target analysis result, where the target analysis result includes multiple target points; dynamically monitoring a first target point among the multiple target points through a UHF sensor to obtain first partial discharge information; calling a signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set; activating a fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; and when the first diagnosis label indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point, sending a first warning signal.

[0006] Further, the target circuit feature information includes target circuit topology features, target circuit load features, and target historical fault features. The circuit analysis model is called to analyze the target circuit feature information to obtain a target analysis result, and the target analysis result includes multiple target points, including:

[0007] The circuit analysis model includes a first analysis channel, a second analysis channel, and a third analysis channel; wherein, through the first analysis channel, it is judged whether the number of first connection lines corresponding to the first circuit connection point is within a predetermined line threshold, and the first circuit connection point is any circuit connection point obtained based on the target circuit topology feature; if it is within, the first circuit connection point is added to the candidate key point list; wherein, through the second analysis channel, it is judged whether the first circuit load corresponding to the first circuit point satisfies a predetermined load constraint, and the first circuit load refers to the load corresponding to the first circuit point in the target circuit load; if it is satisfied, the first circuit point is added to the candidate key point list; wherein, through the third analysis channel, it is judged whether the second circuit point belongs to the historical fault point library, and the historical fault point library refers to the set of points obtained by analyzing the target historical fault features; if it belongs, the second circuit point is added to the candidate key point list; the candidate point list is de-duplicated to obtain the multiple target points.

[0008] Further, the predetermined load constraint includes a load peak constraint and a load type constraint. Judging whether the first circuit load corresponding to the first circuit point satisfies the predetermined load constraint through the second analysis channel includes:

[0009] Judging whether the first load value in the first circuit load satisfies the load peak constraint to obtain a first constraint situation; judging whether the first device type of the first load device in the first circuit load satisfies the load type constraint to obtain a second constraint situation; when the first constraint situation or the second constraint situation is satisfied, it is determined that the first circuit point satisfies the predetermined load constraint.

[0010] Further, the signal analysis model is called to process and analyze the first partial discharge information to obtain a first signal feature set, including:

[0011] Generate a first electromagnetic wave signal time series according to the first correspondence between the first monitoring time and the first electromagnetic wave signal in the first partial discharge information; collect the first time-domain features of the first electromagnetic wave signal time series through the signal analysis model to obtain the first time-domain features; collect the first frequency-domain features of the first electromagnetic wave signal spectrum obtained by converting the first electromagnetic wave signal time series through the signal analysis model to obtain the first frequency-domain features; construct the first signal feature set according to the first time-domain features and the first frequency-domain features.

[0012] Further, obtaining the first signal feature set further includes:

[0013] Activate the signal calibration model, where the signal calibration model includes a first calibrator and a second calibrator. The first calibrator is a model constructed based on the principle of a convolutional neural network, and the second calibrator is a model constructed based on the principle of a long short-term memory network; perform spatial feature analysis on the first partial discharge information through the first calibrator to obtain the first spatial features; perform temporal feature analysis on the first spatial features through the second calibrator to obtain the first temporal features; perform connection fusion on the first spatial features and the first temporal features to obtain the first comprehensive feature, and add the first comprehensive feature to the first signal feature set.

[0014] Further, the activation of the fault diagnosis model to analyze the first signal feature set to obtain the first diagnosis label includes:

[0015] Obtain the first historical fault record according to the target historical fault features. The first historical fault record includes the first historical fault location and the first historical signal feature set of the first historical fault location when a fault occurs; obtain the first historical normal record, where the first historical normal record includes the first historical normal location and the second historical signal feature set of the first historical normal location during normal operation; perform label marking on the first historical signal feature set to obtain the first historical label, and form the first training data group; perform label marking on the first historical signal feature set to obtain the second historical label, and form the second training data group; perform supervised learning and testing on the first training data group and the second training data group to obtain the fault diagnosis model.

[0016] Further, the method for monitoring and warning partial discharge of a high-voltage circuit breaker further includes:

[0017] Activate the fault prediction model, and analyze the first signal feature set through the fault prediction model to obtain the first predicted fault type; when the first warning signal is sent, display the first predicted fault type on the remote center display; wherein, the construction process of the fault prediction model is as follows: the first historical fault record further includes the first fault type of the first historical fault point, and train the third training data set composed of the first historical signal feature set and the first fault type to obtain the fault prediction model.

[0018] In the second aspect disclosed in this application, a partial discharge monitoring and warning system for a high-voltage circuit breaker is provided. The system is used for the above-mentioned partial discharge monitoring and warning method for a high-voltage circuit breaker. The system includes: a circuit feature acquisition module, which is used to acquire target circuit feature information, and the target circuit feature information refers to multi-dimensional feature information of a target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker; a circuit feature analysis module, which is used to call a circuit analysis model to analyze the target circuit feature information to obtain a target analysis result, and the target analysis result includes multiple target points; a point dynamic monitoring module, which is used to dynamically monitor the first target point among the multiple target points through a UHF sensor to obtain the first partial discharge information; a discharge information analysis module, which is used to call a signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set; a signal feature analysis module, which is used to activate a fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; a warning signal sending module, which is used to send a first warning signal when the first diagnosis label indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point.

[0019] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0020] By collecting multi-dimensional characteristic information of the target circuit, it provides a data basis from multiple perspectives for fault analysis. This data basis can accurately reflect the operating conditions of the circuit, enabling the subsequent analysis model to have comprehensive input information; calling the circuit analysis model to comprehensively analyze the multi-dimensional data can identify the target points of multiple potential faults, narrow the scope of fault monitoring, improve the pertinence of fault detection, and avoid resource waste caused by full-network monitoring; by using a ultra-high frequency sensor to dynamically monitor the target points and capture partial discharge information, it can sensitively detect potential partial discharge problems in high-voltage circuit breakers. This non-contact monitoring method not only ensures the accuracy of monitoring but also improves the safety and real-time performance of monitoring; calling the signal analysis model to process and analyze the partial discharge signals, extracting the time-domain and frequency-domain characteristics of the signals, the obtained signal feature set not only covers the electromagnetic wave characteristics monitored in real time but also provides comprehensive signal analysis results, making the subsequent diagnosis more accurate; through the fault diagnosis model, automatically analyzing the signal feature set and generating diagnosis labels, the model identifies whether there is a fault in the circuit breaker and the specific fault type based on historical data and real-time signal characteristics. This intelligent diagnosis method reduces the need for manual intervention and improves the efficiency of fault detection and diagnosis; when a partial discharge fault is detected, an early warning signal is immediately issued. This real-time early warning can help maintenance personnel respond quickly, prevent the fault from further expanding, and reduce the risk of equipment damage and downtime. In summary, this method realizes the precise positioning of faults through real-time monitoring and intelligent fault diagnosis, reduces the blindness in traditional regular maintenance, and at the same time issues an early warning before the fault occurs, avoiding serious damage to the equipment, thereby reducing the repair time and downtime of sudden faults and directly reducing the repair and maintenance costs of the equipment.

[0021] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically illustrates the specific embodiments of this application. Brief Description of the Drawings

[0022] Figure 1 It is a schematic flow chart of a method for monitoring and early warning of partial discharge in a high-voltage circuit breaker provided by an embodiment of this application.

[0023] Figure 2 It is a schematic structural diagram of a system for monitoring and early warning of partial discharge in a high-voltage circuit breaker provided by an embodiment of this application.

[0024] Description of the reference numerals: Circuit feature acquisition module 10, circuit feature analysis module 20, point dynamic monitoring module 30, discharge information analysis module 40, signal feature analysis module 50, early warning signal sending module 60. Detailed Description of the Embodiments

[0025] By providing a method and system for monitoring and warning of partial discharge in a high-voltage circuit breaker in an embodiment of the present application, the technical problem in the prior art that most of the monitoring of partial discharge relies on manual experience or fixed rules, lacks intelligent analysis means, is not only prone to misjudgment or missed judgment, but also difficult to accurately locate, resulting in a reduction in the safety of high-voltage circuit breakers is solved.

[0026] After introducing the basic principle of the present application, various non-limiting implementation manners of the present application will be specifically introduced below in conjunction with the accompanying drawings of the specification. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0027] Embodiment 1, as Figure 1 shown, an embodiment of the present application provides a method for monitoring and warning of partial discharge in a high-voltage circuit breaker, and the method includes:

[0028] Obtain target circuit characteristic information, where the target circuit characteristic information refers to multi-dimensional characteristic information of a target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker.

[0029] The target circuit characteristic information is multi-dimensional characteristic information of the target circuit, involving data of multiple angles and types, including: target circuit topology characteristics, describing the structure and connection relationship of the circuit, including the connection methods of each node in the circuit, such as circuit breakers, transformers, loads, etc., as well as the number of lines and network layout; target circuit load characteristics, recording the load situation in the circuit, such as the type of load, the power of the load, the working state of the load, and the load types such as motors, generators, transformers, etc.; target historical fault characteristics, including the fault information that has occurred in the circuit in the past, such as the historical fault records of the circuit breaker, the types of faults that occurred, the fault points, etc.

[0030] In order to obtain the above information, the following several ways can be adopted: The real-time data of physical quantities such as current and voltage in the circuit can be obtained through a sensor monitoring system, such as current sensors, voltage sensors, etc.; The historical fault information and historical operation data can be queried through the database of the power grid management system, such as the SCADA system or other monitoring systems; The circuit topology characteristics can be extracted from the existing power system diagrams or other power grid planning data.

[0031] Integrate the collected multi-dimensional characteristics to form target circuit characteristic information, which is the basis for subsequent analysis and helps to judge possible fault points, key monitoring positions of partial discharge, etc.

[0032] Call a circuit analysis model to analyze the target circuit characteristic information to obtain a target analysis result, and the target analysis result includes multiple target points.

[0033] The circuit analysis model consists of multiple analysis channels. Each channel evaluates different types of circuit characteristic information. Among them, the first analysis channel determines whether the number of lines at the circuit connection point is within a predetermined threshold range based on topological feature analysis; the second analysis channel determines whether the load characteristics at the circuit point satisfy the predetermined load constraints based on load feature analysis; the third analysis channel determines whether the circuit point belongs to the points in the historical fault point library based on historical fault analysis. The model performs parallel analysis on each point of the circuit through the above three analysis channels, and finally generates multiple target points, which are the circuit points with partial discharge faults.

[0034] The first target point among the multiple target points is dynamically monitored by a UHF sensor to obtain the first partial discharge information.

[0035] The UHF sensor is used to capture the electromagnetic wave signals generated by partial discharges in power equipment. Partial discharge is a tiny electrical discharge in power equipment, and this kind of discharge is often a precursor to equipment aging or failure. By monitoring partial discharges, potential faults in the equipment can be detected in advance to avoid the occurrence of large-scale faults. The advantages of the UHF sensor are high sensitivity, non-intrusive monitoring, and remote monitoring. It is suitable for high-frequency electromagnetic interference environments and can detect discharges at long distances and hidden locations.

[0036] One of the multiple target points is randomly selected as the analysis object, which is the first target point. The UHF sensor is installed at a suitable position to be able to monitor the electromagnetic wave signals of the first target point. The sensor needs to be configured according to the actual situation of the target point to ensure the effectiveness and accuracy of the monitoring. After installation, during the operation of the circuit, the UHF sensor continuously collects electromagnetic wave signals in real time. When a partial discharge occurs, electromagnetic waves within a specific frequency range will be generated, and the sensor can capture these signals and record them to obtain the first partial discharge information. The first partial discharge information includes the partial discharge state at the first target point, reflecting the intensity, frequency, timing characteristics, etc. of the partial discharge.

[0037] The signal analysis model is called to process and analyze the first partial discharge information to obtain the first signal feature set.

[0038] The signal analysis model is used to process partial discharge information. The partial discharge signal is a complex signal containing rich time-domain and frequency-domain characteristics. The main function of the signal analysis model is to extract key features from these complex signals for subsequent fault diagnosis.

[0039] Specifically, the signal analysis model first performs time-domain analysis on the first partial discharge information, including monitoring the changes of the signal at different time points, extracting time-domain features such as amplitude, duration, signal waveform, etc. The obtained time-domain features reflect the distribution of the partial discharge signal on the time axis. Then, through the signal analysis model, the partial discharge information is transformed from the time domain to the frequency domain to analyze its frequency components. Fourier transform or other spectrum analysis methods can be used to obtain the frequency-domain features representing the frequency distribution of the partial discharge signal, such as in which frequency ranges the discharge intensity is mainly concentrated. Further, the signal calibration model is used to analyze the spatial features of the partial discharge signal to reveal the distribution characteristics of the partial discharge in the physical space. Integrating the obtained time-domain, frequency-domain, and spatial features, a first signal feature set is obtained, which is a comprehensive description of the multi-dimensional features of the partial discharge signal and includes time-domain features, frequency-domain features, and spatial features.

[0040] The fault diagnosis model is activated to analyze the first signal feature set to obtain a first diagnosis label.

[0041] The fault diagnosis model is a model trained based on historical data. Specifically, it is constructed through supervised learning of historical fault records and historical normal records. Therefore, this model can identify and classify different types of partial discharge signal features. The first signal feature set is input into the fault diagnosis model. The model analyzes the input feature set to determine whether these features match the known partial discharge fault features and outputs a first diagnosis label to mark the status of this point, including whether there is a partial discharge fault and the possible fault types.

[0042] When the first diagnosis label indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point, a first warning signal is issued.

[0043] When the first diagnosis label indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point, it is confirmed that the partial discharge signal has exceeded the normal range and the circuit breaker at the target point is facing a fault risk. In this case, a first warning signal is generated and transmitted to the remote monitoring center or other relevant systems to notify the maintenance personnel or the automation system to perform corresponding fault handling to prevent the partial discharge from further expanding and causing serious accidents.

[0044] Furthermore, the target circuit feature information includes target circuit topology features, target circuit load features, and target historical fault features. The method for monitoring and warning partial discharge of a high-voltage circuit breaker includes:

[0045] The circuit analysis model includes a first analysis channel, a second analysis channel, and a third analysis channel. Among them, through the first analysis channel, it is judged whether the number of first connection lines corresponding to the first circuit connection point is within a predetermined line threshold, and the first circuit connection point is any circuit connection point obtained based on the target circuit topology feature. If it is within the threshold, the first circuit connection point is added to the candidate key point list. Among them, through the second analysis channel, it is judged whether the first circuit load corresponding to the first circuit point satisfies a predetermined load constraint, and the first circuit load refers to the load in the target circuit corresponding to the first circuit point. If it is satisfied, the first circuit point is added to the candidate key point list. Among them, through the third analysis channel, it is judged whether the second circuit point belongs to the historical fault point library, and the historical fault point library is a set of points obtained by analyzing the target historical fault feature. If it belongs, the second circuit point is added to the candidate key point list. The candidate key point list is de-duplicated to obtain the multiple target points.

[0046] The target circuit feature information is the multi-dimensional feature information of the target circuit, involving data of multiple angles and types, including: the target circuit topology feature, which describes the structure and connection relationship of the circuit, including the connection methods of each node in the circuit, such as circuit breakers, transformers, loads, etc., as well as the number of lines and network layout; the target circuit load feature, which records the load situation in the circuit, such as the type of load, the power of the load, the working state of the load, and the load types such as motors, generators, transformers, etc.; the target historical fault feature, which includes the fault information that has occurred in the circuit in the past, such as the historical fault records of circuit breakers, the types of faults that have occurred, and the fault points.

[0047] The circuit analysis model includes three parallel analysis channels, and each channel evaluates according to different circuit features. Among them, the first analysis channel analyzes based on the topology feature, mainly analyzing the connection method of the circuit to find key nodes; the second analysis channel analyzes based on the load feature to give priority to protecting important loads and sensitive devices; the third analysis channel analyzes based on the historical fault feature, and by analyzing the past fault data, identifies high-fault areas. These channels are processed in parallel, and can independently judge different dimensional features of the circuit, so as to quickly screen out potential key points. Since they are parallel, the analysis of each channel does not depend on each other, which can significantly improve the efficiency of the entire analysis process.

[0048] The circuit topology characteristics describe the connection relationships among the nodes in the power system. In high-voltage circuits, each connection point may be connected to multiple lines, forming a complex network. Each connection point in the circuit can be identified as an intersection node of the electrical network. The goal is to analyze the number of lines at these connection points to determine whether they have high risks. Extract any circuit connection point from the target circuit topology characteristics as the analysis object and label it as the first circuit connection point.

[0049] The predetermined line threshold is a pre-set threshold used to determine whether the number of lines at a certain circuit connection point is in a high-load or critical state. By analyzing the target circuit topology characteristics, the model checks the number of the first connection lines corresponding to the first circuit connection point to determine whether it reaches the predetermined line threshold. If it reaches this threshold value, it indicates that this connection point is a critical node in the entire circuit system because the more lines there are, the greater the impact of this node on the entire system, which also means it is more likely to have problems or failures. Exemplarily, assume the threshold value is set to 4. If a certain circuit connection point is connected to 4 or more lines, then this connection point is considered a high-load or critical point. At this time, mark this point as a potential critical point and add it to the candidate critical point list.

[0050] This process is dynamic. Each connection point in the circuit will be analyzed one by one, and the qualified points will be added to the candidate list in turn. This candidate list is continuously filled in three analysis channels. It contains the critical points in the circuit that may have risks of partial discharge or other failures and serves as the key area for subsequent monitoring and analysis.

[0051] The first circuit load refers to the electrical load borne by the first circuit point. The load information usually includes current, voltage, power, load type, etc. The predetermined load constraint is a pre-set judgment condition used to determine whether the load at a certain point is normal or in a high-risk state. The predetermined load constraint includes load peak constraint and load type constraint.

[0052] Analyze the first circuit load to determine whether it meets the set load peak constraint or load type constraint. If the load exceeds the rated value or the safe operating peak of the device, it is determined that this point is in a high-load state; if the load type is a specific type, such as a device sensitive to partial discharge, it is determined that this point may have a failure risk.

[0053] When any one of the conditions is met, it is determined that this point has a potential failure risk. In this case, mark this point as a point that needs to be monitored key and add it to the candidate critical point list.

[0054] The historical fault point library is a database that stores the fault points that have occurred in the target circuit and the characteristic information related to these fault points. These fault points are extracted from the analysis of historical fault records and operation data, including the partial discharge faults that have occurred at certain specific points of the circuit breaker.

[0055] When judging the second circuit point, by comparing the second circuit point with the historical fault points in the historical fault point library, it is judged whether the point belongs to the historical fault point library. Matching can be performed based on the unique identifier of the point, such as device ID, location, connection method, etc. If the second circuit point has had a fault in the past, then this point is determined to have a high fault risk. In this case, the second circuit point is added to the candidate key point list.

[0056] The candidate key point list is generated by three parallel analysis channels. Each channel will screen out some key points during the analysis process. There may be overlapping parts among these points, that is, some points may meet the conditions of multiple channels at the same time. For example, they belong to historical fault points and meet the load constraint conditions. To avoid repeated monitoring of the same point, the candidate key point list is de-duplicated to ensure that each point only appears in the list once. After de-duplication, multiple non-repeated target points are finally generated. These points represent the positions in the circuit where partial discharge may occur and are the key objects for subsequent monitoring and fault analysis.

[0057] Furthermore, the predetermined load constraint includes a load peak constraint and a load type constraint. The method for monitoring and warning partial discharge of a high-voltage circuit breaker includes:

[0058] Judge whether the first load value in the first circuit load meets the load peak constraint to obtain the first constraint situation; judge whether the first device type of the first load device in the first circuit load meets the load type constraint to obtain the second constraint situation; when the first constraint situation or the second constraint situation is met, it is determined that the first circuit point meets the predetermined load constraint.

[0059] The predetermined load constraint is an evaluation criterion set for the load of each circuit point, used to judge whether the load devices in the circuit are in a normal or overloaded state, including a load peak constraint and a load type constraint. Among them, the load peak constraint is used to evaluate whether the load at a certain circuit point exceeds its rated capacity or the maximum load that the device can withstand. The load peak constraint can set specific thresholds according to different device types and operating conditions; the load type constraint judges whether the device has special monitoring requirements according to the type of load device in the circuit. For example, large motors, transformers, etc. are more likely to cause partial discharge during operation, so they may be subject to more stringent monitoring.

[0060] Obtain the first load value corresponding to the first circuit load at the first circuit point, such as current, voltage, power, etc. Compare the first load value with the load peak constraint corresponding to this point to determine whether it exceeds the allowable peak range. If the load value exceeds the set peak threshold, it is determined that this point meets the peak constraint, and record the judgment result as the first constraint situation, including satisfied and not satisfied.

[0061] Obtain the first device type corresponding to the first load device of the first circuit load at the first circuit point, and determine whether this device type belongs to a preset specific type. The type can be classified according to the operating characteristics and failure risks of the load device. If this device type belongs to the device type that needs to be monitored key, it is considered that it meets the load type constraint, and record the judgment result as the second constraint situation, also including satisfied and not satisfied.

[0062] Integrate the first constraint situation and the second constraint situation. When any one of the constraint conditions is satisfied, that is, the load peak exceeds the standard or the load device type belongs to a specific type, it is determined that the first circuit point meets the predetermined load constraint.

[0063] Furthermore, the method for monitoring and warning partial discharge of a high-voltage circuit breaker includes:

[0064] According to the first correspondence relationship between the first monitoring time and the first electromagnetic wave signal in the first partial discharge information, generate the first electromagnetic wave signal time sequence; collect the time-domain characteristics of the first electromagnetic wave signal time sequence through the signal analysis model to obtain the first time-domain characteristics; collect the frequency-domain characteristics of the first electromagnetic wave signal spectrum obtained by converting the first electromagnetic wave signal time sequence through the signal analysis model to obtain the first frequency-domain characteristics; construct the first signal feature set according to the first time-domain characteristics and the first frequency-domain characteristics.

[0065] The first partial discharge information contains multiple monitoring time points and corresponding electromagnetic wave signal data. These time points are matched with the corresponding signal values to generate the first electromagnetic wave signal time sequence. Specifically, the monitoring time includes the moment when the partial discharge event occurs, and the electromagnetic wave signal is the electromagnetic wave amplitude and frequency information corresponding to these moments. These data points are arranged in chronological order to form a time series for subsequent analysis.

[0066] Extract the time-domain characteristics of the first electromagnetic wave signal time sequence. The time-domain characteristics include the amplitude of the signal (maximum value, minimum value, mean value, etc.), the waveform characteristics of the signal (such as zero-crossing points, waveform symmetry), and the time-domain statistical characteristics of the signal (such as variance, standard deviation). These characteristics can reflect the change characteristics of the electromagnetic wave signal in time. After time-domain analysis, a series of time-domain characteristics are extracted and recorded as the first time-domain characteristics.

[0067] Perform a Fourier transform on the timing sequence of the first electromagnetic wave signal to convert the time-domain signal into a frequency-domain signal. The obtained spectrum of the first electromagnetic wave signal provides the intensity information of the signal at different frequencies, which can reveal the frequency components in the signal. Collect the frequency-domain characteristics of the spectrum of the first electromagnetic wave signal. The frequency-domain characteristics include the main frequency components (such as the main frequency and its amplitude), the bandwidth of the spectrum (the range of signal energy distribution), and the frequency-domain statistical characteristics (such as the frequency mean, frequency variance, etc.). The collected frequency-domain characteristics form the first frequency-domain characteristics.

[0068] Integrate the first time-domain characteristics and the first frequency-domain characteristics to obtain the first signal feature set. This signal feature set integrates the information in the time domain and the frequency domain, provides the comprehensive characteristics of the electromagnetic wave signal, and can effectively support subsequent fault diagnosis.

[0069] Furthermore, the method for monitoring and warning partial discharge of a high-voltage circuit breaker further includes:

[0070] Activate the signal calibration model. The signal calibration model includes a first calibrator and a second calibrator. The first calibrator is a model constructed based on the principle of convolutional neural network, and the second calibrator is a model constructed based on the principle of long short-term memory network. Analyze the spatial characteristics of the first partial discharge information through the first calibrator to obtain the first spatial characteristics. Analyze the timing characteristics of the first spatial characteristics through the second calibrator to obtain the first time characteristics. Connect and fuse the first spatial characteristics and the first time characteristics to obtain the first comprehensive characteristic, and add the first comprehensive characteristic to the first signal feature set.

[0071] The signal calibration model includes a first calibrator and a second calibrator. Among them, the first calibrator is constructed based on a convolutional neural network (CNN) and is used for spatial feature analysis. The convolutional neural network is good at processing signals or data with spatial structures, such as images, signal distributions, etc., and can identify spatial relationships and patterns by extracting local features. The second calibrator is constructed based on a long short-term memory network (LSTM) and is used for timing feature analysis. The long short-term memory network can process time series data and extract the dynamic features of the signal changing over time.

[0072] The convolutional neural network extracts the spatial features of data layer by layer through multiple convolutional layers and pooling layers. Each convolutional layer extracts the features of a local area by applying filters or convolutional kernels, and the pooling layer is used for dimensionality reduction and extracting more abstract features. The partial discharge signal can be manifested as the signal intensity of spatial distribution, such as the signal amplitude and frequency detected by multiple sensors at different spatial positions. The convolutional neural network processes these spatial data to extract specific spatial patterns related to partial discharge, and extracts the spatial features of the partial discharge signal, such as the signal intensity difference at specific sensor positions, the spatial distribution of the partial discharge source position, and the spatial diffusion pattern of the electromagnetic wave signal, to obtain the first spatial feature.

[0073] The long short-term memory network performs excellently in processing sequence data with time correlation because it can remember the dependencies over a long period of time. Starting from the first spatial feature, it further analyzes the variation laws of these spatial features over time, such as the variation of the signal amplitude over time, the periodicity or randomness of the discharge, and the fluctuation characteristics of the partial discharge signal, such as whether there is a trend of gradual increase or decrease. Through the long short-term memory network model, these temporal characteristics are identified and the temporal features are extracted to facilitate understanding the temporal evolution of the partial discharge signal. The obtained first temporal feature can reflect the changes of the partial discharge signal on the time axis.

[0074] The first spatial feature and the first temporal feature are concatenated and fused. Concatenating and fusing means combining features of different dimensions into a more representative feature set to capture the comprehensive characteristics of the partial discharge signal. The fused feature set can simultaneously reflect the spatial distribution of the partial discharge signal and the changes of the signal at different time points. Through concatenating and fusing, the first comprehensive feature is finally generated. This comprehensive feature integrates the information of both the spatial and temporal dimensions and can describe the partial discharge phenomenon more comprehensively. The first comprehensive feature is added to the first signal feature set.

[0075] Furthermore, the method for monitoring and warning partial discharge of a high-voltage circuit breaker includes:

[0076] Obtain the first historical fault record according to the target historical fault feature. The first historical fault record includes the first historical fault point and the first historical signal feature set of the first historical fault point when a fault occurs; obtain the first historical normal record. The first historical normal record includes the first historical normal point and the second historical signal feature set of the first historical normal point during normal operation; perform label marking on the first historical signal feature set to obtain the first historical label and form the first training data group; perform label marking on the second historical signal feature set to obtain the second historical label and form the second training data group; perform supervised learning and testing on the first training data group and the second training data group to obtain the fault diagnosis model.

[0077] Target historical fault features include the fault information that has occurred in the circuit in the past, such as the historical fault records of circuit breakers, the types of faults that occurred, the fault points, etc. The first historical fault record is the historical data extracted from the target historical fault features. Among them, the first historical fault point refers to the specific location or equipment where the fault occurred during a certain historical fault; the first historical signal feature set contains the signal feature data collected when a fault occurred at the first historical fault point, including information such as the time domain, frequency domain, and spatial features during the fault.

[0078] The first historical normal record refers to the historical data of the circuit under normal operating conditions. The normal record is opposite to the fault record and can provide a comparison sample for the model to help the model distinguish normal and abnormal signal patterns. Among them, the first historical normal point refers to the point where the equipment was operating normally in history. Different from the fault point, the record of the normal point represents the operating state without partial discharge or other abnormal events; the second historical signal feature set is the signal feature data of the equipment in the normal state, including signal data such as the current, voltage waveform, frequency domain characteristics, and load conditions during the healthy operation of the equipment. The normal record and the fault record together constitute the positive and negative sample sets required for model training.

[0079] Labeling is to assign a corresponding fault label to each historical signal sample. The label represents the actual state of the signal. The first historical signal feature set corresponds to the fault state, so a fault label will be assigned. For example, if a certain signal feature set indicates that the equipment has experienced a partial discharge fault in history, then the label "fault" or a more specific label such as "partial discharge fault" can be assigned to this signal. The obtained first historical label is the annotation result corresponding to the first historical signal feature set, indicating that these signal feature data are derived from the fault state. The label provides a clear reference for the model, enabling the model to identify the association between similar features and faults. The first training data group is a data set composed of the first historical signal feature set and its corresponding fault label. This data set is used to train the machine learning model to help the model learn the signal performance of the equipment when a fault occurs.

[0080] Similarly, label the second historical signal feature set, these normal signal data. Since these data come from the normal state of the equipment, a label of the normal state will be assigned, such as normal operation or no fault. The second historical label corresponds to the label in the normal operation state, indicating that these signal data come from the normal operation process of the equipment. The label provides reference data for the model on the normal operation of the equipment, helping the model distinguish the differences between faults and normal states. The second training data group is composed of the second historical signal feature set and the corresponding normal operation label, helping the model learn the signal performance of the equipment in the healthy state.

[0081] Supervised learning is a method of machine learning in which the training dataset consists of input features and output labels. By learning the correspondence between the input data and the output labels, the model can predict the corresponding labels based on the newly input data.

[0082] During the supervised learning process, the first training data set and the second training data set are input into a machine learning algorithm such as a neural network, a support vector machine, etc. The model learns the differences between the signal features in the normal and faulty states through these data learning devices. Through the training algorithm, the model gradually learns how to predict the correct output labels based on the input features.

[0083] Specifically, by inputting a large amount of historical signal features, the model will attempt to fit these data and find the mapping relationship between the signal features and the faulty / normal states. The goal of training is to enable the model to identify and distinguish between normal and faulty signals. The goal of model training is to minimize the prediction error of the model, that is, to minimize the gap between the labels predicted by the model and the true labels. After training is completed, an independent test set is used to test the actual performance of the model. The test set contains data that the model has not seen during the training process, which can evaluate the performance of the model in a real scenario. After training and verification, the model finally learns how to determine whether the device is in a normal state or a faulty state based on the input signal features. This model is the fault diagnosis model.

[0084] Furthermore, the method for monitoring and warning of partial discharge of a high-voltage circuit breaker further includes:

[0085] Activate the fault prediction model, and analyze the first signal feature set through the fault prediction model to obtain the first predicted fault type; when the first warning signal is issued, display the first predicted fault type on the remote center display; wherein, the construction process of the fault prediction model is as follows: the first historical fault record further includes the first fault type of the first historical fault point. Train the third training data set composed of the first historical signal feature set and the first fault type to obtain the fault prediction model.

[0086] Activate the pre-constructed fault prediction model. The model analyzes the input first signal feature set and predicts whether there is a potential fault risk at the first target point by identifying various signal features in the feature set. After the analysis is completed, output the first predicted fault type, that is, the fault type that the model determines may currently occur at the first target point. Specific fault types such as partial discharge faults, short circuit faults, overload faults, etc.

[0087] When the first warning signal is triggered, the first predicted fault type is displayed on the monitor of the remote monitoring center, including the fault location and the fault type. By displaying the predicted fault type in the remote center, the monitoring personnel can grasp the health status of the equipment in real time and make timely responses.

[0088] The construction process of the fault prediction model is as follows:

[0089] The first historical fault record also includes the first fault type of the first historical fault point. The fault type refers to the specific form or category of the equipment fault, such as partial discharge fault, short circuit fault, overload fault. The first historical fault point is associated with the corresponding specific fault type, and these fault types provide labeled data for subsequent model training.

[0090] The first historical signal feature set contains the signal features collected at the first historical fault point when a fault occurs. These features record the operating conditions of the equipment during the fault. The first historical signal feature set is associated with the specific first fault type to form a complete sample, that is, the third training data set. Among them, the input of each sample is the historical fault signal feature, which describes the electrical parameters of the equipment when a fault occurs, and the output is the fault type label, which indicates which specific fault type these signal features correspond to.

[0091] The fault prediction model is a multi-class classification model. Its goal is to predict the fault type based on the input signal features. Therefore, the model needs to be trained on the third training data set so that it can classify faults according to the signal features. Specifically, the model receives the historical signal feature set as input, learns the relationship between them and the fault type according to the input features, and the output is the predicted fault type. During the model training process, the goal is to minimize the error between the prediction result and the actual fault type. By continuously adjusting the parameters, the model finally learns to extract the association between the signal features and the fault type from the historical data, and finally obtains a fault prediction model that meets the accuracy requirement after training.

[0092] In summary, a method for monitoring and warning partial discharge of a high-voltage circuit breaker provided by an embodiment of the present application has the following technical effects:

[0093] By collecting multi-dimensional characteristic information of the target circuit, it provides a data basis from multiple perspectives for fault analysis. This data basis can accurately reflect the operating conditions of the circuit, enabling the subsequent analysis model to have comprehensive input information; calling the circuit analysis model to comprehensively analyze the multi-dimensional data can identify the target points of multiple potential faults, narrow the scope of fault monitoring, improve the pertinence of fault detection, and avoid resource waste caused by full-network monitoring; by using ultra-high frequency sensors to dynamically monitor the target points and capture partial discharge information, it can sensitively detect potential partial discharge problems in high-voltage circuit breakers. This non-contact monitoring method not only ensures the accuracy of monitoring but also improves the safety and real-time performance of monitoring; calling the signal analysis model to process and analyze the partial discharge signals, extracting the time-domain and frequency-domain characteristics of the signals, the obtained signal feature set not only covers the electromagnetic wave characteristics monitored in real time but also provides comprehensive signal analysis results, making the subsequent diagnosis more accurate; through the fault diagnosis model, automatically analyzing the signal feature set and generating diagnostic labels, the model can identify whether there is a fault in the circuit breaker and the specific fault type based on historical data and real-time signal characteristics. This intelligent diagnosis method reduces the need for manual intervention and improves the efficiency of fault detection and diagnosis; when a partial discharge fault is detected, an early warning signal is immediately issued. This real-time early warning can help maintenance personnel respond quickly, prevent the further expansion of the fault, and reduce the risk of equipment damage and downtime. In summary, through real-time monitoring and intelligent fault diagnosis, this method realizes the accurate positioning of faults, reduces the blindness in traditional regular maintenance. At the same time, it issues an early warning before the fault occurs, avoids serious damage to the equipment, thereby reducing the repair time and downtime of sudden faults, and directly reducing the repair and maintenance costs of the equipment.

[0094] Embodiment 2, based on the same inventive concept as the method for monitoring and early warning of partial discharge in a high-voltage circuit breaker in the foregoing embodiment, as Figure 2 shown, the embodiment of the present application provides a system for monitoring and early warning of partial discharge in a high-voltage circuit breaker, and the system includes:

[0095] Circuit feature acquisition module 10, the circuit feature acquisition module 10 is used to acquire target circuit feature information, the target circuit feature information refers to multi-dimensional feature information of a target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker; Circuit feature analysis module 20, the circuit feature analysis module 20 is used to call a circuit analysis model to analyze the target circuit feature information to obtain a target analysis result, the target analysis result includes multiple target points; Point position dynamic monitoring module 30, the point position dynamic monitoring module 30 is used to dynamically monitor a first target point among the multiple target points through a UHF sensor to obtain first partial discharge information; Discharge information analysis module 40, the discharge information analysis module 40 is used to call a signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set; Signal feature analysis module 50, the signal feature analysis module 50 is used to activate a fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; Early warning signal sending module 60, the early warning signal sending module 60 is used to send a first early warning signal when the first diagnosis label indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point.

[0096] Furthermore, the target circuit feature information includes target circuit topology features, target circuit load features, and target historical fault features. The system further includes multiple target point acquisition modules to perform the following operation steps:

[0097] The circuit analysis model includes a first analysis channel, a second analysis channel, and a third analysis channel; among them, through the first analysis channel, it is judged whether the number of first connection lines corresponding to a first circuit connection point is within a predetermined line threshold, and the first circuit connection point is any circuit connection point obtained based on the target circuit topology feature; if it is within, add the first circuit connection point to the candidate key point list; among them, through the second analysis channel, it is judged whether the first circuit load corresponding to a first circuit point satisfies a predetermined load constraint, and the first circuit load refers to the load corresponding to the first circuit point in the target circuit load; if it is satisfied, add the first circuit point to the candidate key point list; among them, through the third analysis channel, it is judged whether a second circuit point belongs to a historical fault point library, and the historical fault point library refers to a set of points obtained by analyzing the target historical fault features; if it belongs, add the second circuit point to the candidate key point list; perform duplicate removal processing on the candidate key point list to obtain the multiple target points.

[0098] Furthermore, the predetermined load constraint includes a load peak constraint and a load type constraint. The system further includes a first circuit point determination module to perform the following operation steps:

[0099] Determine whether the first load value in the first circuit load satisfies the load peak constraint to obtain a first constraint situation; determine whether the first device type of the first load device in the first circuit load satisfies the load type constraint to obtain a second constraint situation; when the first constraint situation or the second constraint situation is satisfied, it is determined that the first circuit point satisfies the predetermined load constraint.

[0100] Furthermore, the system further includes a first signal feature set acquisition module to perform the following operation steps:

[0101] Generate a first electromagnetic wave signal time series according to the first correspondence between the first monitoring time and the first electromagnetic wave signal in the first partial discharge information; collect time domain features of the first electromagnetic wave signal time series through the signal analysis model to obtain first time domain features; collect frequency domain features of the first electromagnetic wave signal spectrum obtained by converting the first electromagnetic wave signal time series through the signal analysis model to obtain first frequency domain features; construct the first signal feature set according to the first time domain features and the first frequency domain features.

[0102] Furthermore, the system further includes a first comprehensive feature addition module to perform the following operation steps:

[0103] Activate the signal calibration model, where the signal calibration model includes a first calibrator and a second calibrator, the first calibrator is a model constructed based on the principle of convolutional neural network, and the second calibrator is a model constructed based on the principle of long short-term memory network; perform spatial feature analysis on the first partial discharge information through the first calibrator to obtain first spatial features; perform time series feature analysis on the first spatial features through the second calibrator to obtain first time features; perform connection fusion on the first spatial features and the first time features to obtain a first comprehensive feature, and add the first comprehensive feature to the first signal feature set.

[0104] Furthermore, the system further includes a fault diagnosis model acquisition module to perform the following operation steps:

[0105] Obtain the first historical fault record according to the target historical fault characteristics, where the first historical fault record includes the first historical fault location and the first historical signal feature set of the first historical fault location when a fault occurs; obtain the first historical normal record, where the first historical normal record includes the first historical normal location and the second historical signal feature set of the first historical normal location during normal operation; perform label marking on the first historical signal feature set to obtain the first historical label, and form the first training data group; perform label marking on the second historical signal feature set to obtain the second historical label, and form the second training data group; perform supervised learning and verification on the first training data group and the second training data group to obtain the fault diagnosis model.

[0106] Furthermore, the system further includes a fault type display module to perform the following operating steps:

[0107] Activate the fault prediction model, and analyze the first signal feature set through the fault prediction model to obtain the first predicted fault type; when the first warning signal is issued, display the first predicted fault type on the remote central display; where the construction process of the fault prediction model is as follows: the first historical fault record further includes the first fault type of the first historical fault location, and train the third training data group formed by the first historical signal feature set and the first fault type to obtain the fault prediction model.

[0108] Through the foregoing detailed description of a method for monitoring and warning partial discharge of a high-voltage circuit breaker in this specification, those skilled in the art can clearly know a system for monitoring and warning partial discharge of a high-voltage circuit breaker in this embodiment. Since it corresponds to the method disclosed in the embodiment, it is described relatively simply, and the relevant parts can be referred to the description of the method part.

[0109] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker, characterized in that: The method comprises: Acquire target circuit characteristic information, wherein the target circuit characteristic information refers to multi-dimensional characteristic information of the target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker; Calling a circuit analysis model to analyze the target circuit characteristic information to obtain a target analysis result, wherein the target analysis result includes a plurality of target points; Dynamically monitoring a first target point among the multiple target points by using a UHF sensor to obtain first partial discharge information; Calling a signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set; activating the fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; When the first diagnostic tag indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point, a first warning signal is issued; The target circuit characteristic information includes target circuit topology characteristics, target circuit load characteristics and target historical fault characteristics. The calling circuit analysis model analyzes the target circuit characteristic information to obtain a target analysis result. The target analysis result includes multiple target points, including: The circuit analysis model includes a first analysis channel, a second analysis channel and a third analysis channel; Wherein, determining whether the number of first connection lines corresponding to a first circuit connection point is within a predetermined line threshold through the first analysis channel, the first circuit connection point being any circuit connection point obtained based on the target circuit topology feature; If yes, add the first circuit connection point to the candidate key point list; Wherein, judging whether a first circuit load corresponding to a first circuit point satisfies a predetermined load constraint through the second analysis channel, the first circuit load refers to a load corresponding to the first circuit point in the target circuit load characteristic; If satisfied, adding the first circuit point to the candidate key point list; Wherein, judging whether the second circuit point belongs to a historical fault point library through the third analysis channel, the historical fault point library refers to a point set obtained by analyzing the target historical fault characteristics; If yes, adding the second circuit point to the candidate key point list; The candidate key point list is deduplicated to obtain the multiple target points.

2. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to claim 1, characterized in that: The predetermined load constraint includes a load peak constraint and a load type constraint, and judging whether the first circuit load corresponding to the first circuit point satisfies the predetermined load constraint through the second analysis channel includes: Determine whether a first load value in the first circuit load satisfies the load peak constraint to obtain a first constraint condition; Determine whether a first device type of a first load device in the first circuit load satisfies the load type constraint, and obtain a second constraint condition; When the first constraint condition or the second constraint condition is satisfied, it is determined that the first circuit point satisfies the predetermined load constraint.

3. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to claim 1, characterized in that: The calling of the signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set includes: Generate a first electromagnetic wave signal time sequence according to a first corresponding relationship between a first monitoring time and a first electromagnetic wave signal in the first partial discharge information; Collecting time domain features of the first electromagnetic wave signal time series through the signal analysis model to obtain a first time domain feature; Collect frequency domain features of the first electromagnetic wave signal spectrum obtained by converting the first electromagnetic wave signal time series through the signal analysis model to obtain first frequency domain features; The first signal feature set is established according to the first time domain feature and the first frequency domain feature.

4. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to claim 3, characterized in that: The obtaining of the first signal feature set further includes: Activate a signal calibration model, wherein the signal calibration model includes a first calibrator and a second calibrator, wherein the first calibrator is a model constructed based on the principle of a convolutional neural network, and the second calibrator is a model constructed based on the principle of a long short-term memory network; Performing spatial feature analysis on the first partial discharge information by the first calibrator to obtain a first spatial feature; Performing a time series feature analysis on the first spatial feature by the second calibrator to obtain a first time feature; The first spatial feature and the first temporal feature are connected and fused to obtain a first comprehensive feature, and the first comprehensive feature is added to the first signal feature set.

5. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to claim 3, characterized in that: The activating fault diagnosis model analyzes the first signal feature set to obtain a first diagnosis label, including: Acquire a first historical fault record according to the target historical fault feature, the first historical fault record including a first historical fault point and a first historical signal feature set when the first historical fault point fails; Acquire a first historical normal record, where the first historical normal record includes a first historical normal point and a second historical signal feature set of the first historical normal point during normal operation; Labeling the first historical signal feature set to obtain a first historical label and forming a first training data group; Labeling the second historical signal feature set to obtain a second historical label and forming a second training data group; Supervised learning and verification are performed on the first training data group and the second training data group to obtain the fault diagnosis model.

6. A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to claim 5, characterized in that: The high-voltage circuit breaker partial discharge monitoring and early warning method further comprises: activating a fault prediction model, and analyzing the first signal feature set through the fault prediction model to obtain a first predicted fault type; When the first warning signal is issued, the first predicted fault type is displayed on a remote center display; The construction process of the fault prediction model is as follows: The first historical fault record also includes a first fault type of the first historical fault point, and a third training data group consisting of the first historical signal feature set and the first fault type is trained to obtain the fault prediction model.

7. A high voltage circuit breaker partial discharge monitoring and early warning system, characterized in that: A method for monitoring and early warning of partial discharge of a high-voltage circuit breaker according to any one of claims 1 to 6, wherein the method comprises: A circuit feature acquisition module, the circuit feature acquisition module is used to acquire target circuit feature information, the target circuit feature information refers to multi-dimensional feature information of the target circuit, and the target circuit refers to a circuit including a target high-voltage circuit breaker; A circuit characteristic analysis module, wherein the circuit characteristic analysis module is used to call a circuit analysis model to analyze the target circuit characteristic information to obtain a target analysis result, wherein the target analysis result includes a plurality of target points; A point dynamic monitoring module, the point dynamic monitoring module is used to dynamically monitor a first target point among the multiple target points through a UHF sensor to obtain first partial discharge information; A discharge information analysis module, the discharge information analysis module is used to call a signal analysis model to process and analyze the first partial discharge information to obtain a first signal feature set; A signal feature analysis module, the signal feature analysis module is used to activate the fault diagnosis model to analyze the first signal feature set to obtain a first diagnosis label; A warning signal issuing module is used to issue a first warning signal when the first diagnostic tag indicates that the target high-voltage circuit breaker has a partial discharge fault at the first target point.

Images