High-voltage Circuit Breaker and Its Disconnector Overheating Fault Prediction Method and System

Abstract

The present invention relates to the technical field of machine learning, and particularly to a high-voltage circuit breaker and a method and system for predicting the heating faults of its disconnector. The method includes the following steps: collecting multi-source data on the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and performing data preprocessing to obtain a preprocessed data set; performing an associated analysis of operation characteristics on the preprocessed data set to obtain an associated operation characteristics set; constructing a dynamic association network for the associated operation characteristics set to obtain a dynamic association network diagram; obtaining historical fault data; searching for propagation risk paths according to the historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculating the path propagation probability for the candidate risk path set to obtain a path probability data set. The present invention effectively identifies the influence of the associated high-voltage circuit breaker on the heating risk of the disconnector, thereby improving the accuracy of fault prediction and the timeliness of early warning, and providing a more reliable guarantee for the safe and stable operation of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly to a high-voltage circuit breaker and a method and system for predicting the heating fault of its disconnector. Background Art

[0002] A high-voltage circuit breaker is an important protection and control device in the power system, mainly used in high-voltage transmission lines and substations to perform on-load switching operations on the circuit and quickly cut off the fault current when a system fault occurs, protecting power equipment and limiting the expansion of accidents. The high-voltage circuit breaker is closely related to the high-voltage disconnector. The disconnector is usually used in combination with the circuit breaker to reliably isolate the live part of the circuit breaker during maintenance or repair. At the same time, frequent operation of the circuit breaker will exacerbate the wear of the disconnector, leading to heating faults of the disconnector. Therefore, in the prediction of disconnector faults, the influence of the associated high-voltage circuit breaker must be fully considered. The traditional methods for predicting the heating faults of the disconnector of the high-voltage circuit breaker mainly rely on manual inspection and regular maintenance, which have problems such as low prediction accuracy, strong hysteresis, and low efficiency. In recent years, with the rapid development of sensor technology, data analysis technology, and artificial intelligence technology, some new prediction methods have emerged, as follows:

[0003] 1. Prediction method based on temperature monitoring: By monitoring parameters such as the contact temperature and terminal temperature of the disconnector, the occurrence of faults is predicted using the temperature change trend.

[0004] 2. Prediction method based on vibration analysis: By analyzing the vibration signal of the disconnector, abnormal vibration modes are identified to predict the occurrence of faults.

[0005] 3. Prediction method based on infrared thermal imaging: Using infrared thermal imaging technology to detect the temperature distribution of the disconnector, identify abnormal hot spots, and predict the occurrence of faults.

[0006] However, most of these methods only focus on the state of the disconnector itself and lack consideration of the influence of the associated high-voltage circuit breaker, resulting in limited prediction accuracy and warning timeliness.

[0007] The disconnector of the high-voltage circuit breaker does not operate in isolation, and its faults are often closely related to one or more associated high-voltage circuit breakers. For example, frequent operation of the high-voltage circuit breaker will cause increased wear of the disconnector contacts, increasing the risk of heating faults; a fault trip of the high-voltage circuit breaker will cause the disconnector to withstand an overcurrent impact, leading to heating faults.

[0008] The main reasons for the lack of analysis of the influence of the associated high-voltage circuit breaker in traditional methods are as follows:

[0009] 1. Incomplete data collection: Traditional methods mainly collect the operating data of the disconnector itself and lack the collection and analysis of data related to the associated high-voltage circuit breaker.

[0010] 2. Difficulty in modeling association relationships: The association relationships between multiple high-voltage circuit breakers and high-voltage disconnectors are complex and diverse, making it difficult to describe them with simple mathematical models.

[0011] 3. Lack of effective analysis methods: Even if data on associated high-voltage circuit breakers is collected, there is a lack of effective analysis methods to identify fault risks. Summary of the Invention

[0012] Based on this, it is necessary to provide a high-voltage circuit breaker and its disconnector overheating fault prediction method and system to solve at least one of the above technical problems.

[0013] To achieve the above object, a method for predicting overheating faults of a disconnector of a high-voltage circuit breaker includes the following steps:

[0014] Step S1: Collect multi-source data on the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; perform running feature association analysis on the preprocessed data set to obtain an associated running feature set;

[0015] Step S2: Construct a dynamic association network for the associated running feature set to obtain a dynamic association network diagram;

[0016] Step S3: Obtain historical fault data; search for propagation risk paths based on the historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculate the path propagation probability for the candidate risk path set to obtain a path probability data set; extract the key risk propagation path based on the path probability data set and the candidate risk path set to obtain the key risk propagation path;

[0017] Step S4: Calculate the path propagation probability based on the key risk propagation path to obtain a path probability matrix; perform dynamic analysis of the fault probability based on the path probability matrix to obtain a dynamic fault probability matrix;

[0018] Step S5: Construct an overheating risk prediction model based on the associated running feature set and the dynamic fault probability matrix to obtain an overheating risk prediction model; use the overheating risk prediction model to predict the overheating risk, and calculate the comprehensive risk index to obtain the comprehensive overheating risk index;

[0019] Step S6: Obtain historical warning data; construct an adaptive warning model based on the historical warning data to obtain an adaptive warning model; perform adaptive warning of the overheating fault of the high-voltage disconnector based on the comprehensive overheating risk index and the adaptive warning model to implement the overheating fault prediction operation of the high-voltage disconnector.

[0020] By collecting multi-source data and performing preprocessing, the present invention not only ensures the data quality, but also extracts key operating characteristics related to high-voltage circuit breakers, laying a data foundation for subsequent analysis of the mutual influence between devices and construction of prediction models. By constructing a dynamic association network, the association relationship between devices and its trend of change over time are intuitively displayed, providing an intuitive visual guidance for identifying key risk propagation paths. By analyzing historical fault data and the dynamic association network, potential risk propagation paths are found, and the propagation probability of each path is evaluated, providing a basis for accurately locating key risk paths. By constructing a fault propagation model and combining it with real-time monitoring data for dynamic analysis, real-time risk assessment of key risk propagation paths is achieved, providing a dynamic probability basis for predicting the heating fault of disconnectors. By constructing a prediction model based on association characteristics and dynamic fault probability, the future heating risk of disconnectors can be predicted more accurately, and through the comprehensive risk index, a more intuitive reference basis for risk warning is provided. By constructing an adaptive warning model and dynamically adjusting the warning threshold according to the comprehensive risk index, a more accurate and flexible warning mechanism is achieved, effectively reducing the risks of false alarms and missed alarms, and improving the efficiency and accuracy of warning. Therefore, the present invention provides a method for predicting the heating fault of disconnectors of high-voltage circuit breakers. By introducing technologies such as multi-source data association analysis, dynamic association network construction, and risk propagation path identification, the drawback that existing methods lack consideration of the influence on associated high-voltage circuit breakers is effectively solved. This method can effectively identify the influence of associated high-voltage circuit breakers on the heating risk of disconnectors, thereby improving the accuracy of fault prediction and the timeliness of warning, and providing a more reliable guarantee for the safe and stable operation of the power system.

[0021] Preferably, step S1 includes the following steps:

[0022] Step S11: Collect multi-source data on high-voltage disconnectors and their associated high-voltage circuit breakers through the SCADA system to obtain an original data set; perform data preprocessing on the original data set to obtain a preprocessed data set;

[0023] Step S12: Align the data time of the preprocessed data set to obtain a time-aligned data set;

[0024] Step S13: Establish an association relationship between the high-voltage disconnector and its associated high-voltage circuit breaker to obtain association relationship data;

[0025] Step S14: Perform feature matching on the time-aligned data set according to the association relationship data within the same time window to obtain a feature-matched data set;

[0026] Step S15: Generate an associated operation feature set according to the feature-matched data set to obtain an associated operation feature set.

[0027] The present invention collects multi-source data through the SCADA system and performs preprocessing, which can eliminate noise and redundant information in the original data, improve the quality and consistency of the data, and lay a foundation for subsequent feature analysis and model construction. Data time alignment is performed on the preprocessed data set to ensure that the data of different devices are synchronized in time, eliminate the influence of time difference on correlation analysis, and ensure the accuracy of subsequent feature matching and correlation analysis. The correlation relationship data between the high-voltage disconnector and its associated high-voltage circuit breaker is established, clarifying the topological relationship between the devices, and providing a necessary information basis for subsequent analysis of the influence of the associated high-voltage circuit breaker and the risk propagation path. Feature matching is performed in the same time window according to the correlation relationship data, and the operation data of the associated high-voltage circuit breaker is correlated in the same time dimension, forming multi-dimensional correlation feature information, which provides a richer data basis for subsequent extraction of the associated operation feature set. The associated operation feature set is generated according to the feature matching data set, and the main features that can characterize the operation states of the high-voltage disconnector and the associated high-voltage circuit breaker are extracted, reducing the data dimension, reducing the computational complexity, and retaining key information at the same time, providing a more refined data representation for subsequent dynamic association network construction and fault prediction.

[0028] Preferably, step S2 includes the following steps:

[0029] Step S21: Perform correlation feature time series segmentation on the associated operation feature set to obtain a time series feature data set;

[0030] Step S22: Perform dynamic correlation relationship calculation on the time series feature data set to obtain a correlation relationship matrix;

[0031] Step S23: Perform dynamic evolution of the correlation relationship on the correlation relationship matrix to obtain a sequence of dynamic correlation matrices;

[0032] Step S24: Set the correlation strength threshold according to the sequence of dynamic correlation matrices and draw a network diagram to obtain a dynamic association network diagram.

[0033] By performing correlation feature time series segmentation on the correlation operation feature set, the present invention converts time series data into multiple time segment samples, enabling each sample to more comprehensively reflect the operating state of the device within a specific time period, and providing a more refined data basis for dynamic correlation relationship calculation. By calculating the dynamic correlation relationship of the time series feature data set, the correlation degree between different features is quantified, and a correlation relationship matrix is constructed, providing a quantitative basis for subsequent analysis of the mutual influence and risk propagation between devices. By constructing a dynamic correlation matrix sequence, the changing trend of the correlation relationship over time is captured, overcoming the limitations of traditional static analysis methods, being able to more accurately reflect the dynamic changes of the power system operating state, and providing a more reliable basis for risk warning. According to the dynamic correlation matrix sequence, a correlation intensity threshold is set, strong correlation relationships are screened out, and a dynamic correlation network diagram is drawn, presenting the complex correlation relationships in an intuitive manner, helping to more clearly understand the mutual influence between devices, and providing a direct visual guidance for identifying key risk propagation paths.

[0034] Preferably, step S3 includes the following steps:

[0035] Step S31: Perform network topology structure analysis on the dynamic correlation network diagram to obtain a network topology data set;

[0036] Step S32: Obtain historical fault data, where the historical fault data includes historical high-voltage circuit breaker fault state data, historical disconnector heating state data, and equipment operation and maintenance data;

[0037] Step S33: Search for propagation risk paths in the network topology data set according to the historical fault data to obtain a candidate risk path set;

[0038] Step S34: Calculate the path propagation probability for the candidate risk path set to obtain a path probability data set;

[0039] Step S35: Sort the candidate risk path set according to the path probability data set to obtain a sorted path probability data set;

[0040] Step S36: Use a preset sorted path threshold to screen the sorted path probability data set to obtain key risk propagation paths.

[0041] The present invention analyzes the network topology structure of the dynamic association network graph, extracts key information characterizing the network structure, such as the degree, centrality, clustering coefficient, etc. of nodes, as well as the connectivity, shortest path, etc. of the network, providing a structured data basis for subsequent search of risk propagation paths. Historical fault data is obtained, including historical high-voltage circuit breaker fault status data, historical disconnector heating status data, and equipment operation and maintenance data, providing historical experience and data support for subsequent analysis of fault propagation laws, calculation of path propagation probabilities, and identification of key risk paths. According to the historical fault data, a risk propagation path search is carried out on the network topology data set, all possible propagation paths from the fault source node to the target node are found, and a candidate risk path set is constructed, providing candidates for subsequent risk assessment and prediction. By calculating the path propagation probability of the candidate risk path set, the possibility of fault propagation for each path is quantified, providing a quantitative basis for evaluating the risk levels of different paths. According to the path probability data set, the candidate risk path set is sorted by path importance, with high-risk paths ranked first for easy focused attention, improving the efficiency of risk identification. Using a preset sorting path threshold to screen the sorted path probability data set for key paths, interference from low-risk paths is excluded, and the key risk propagation path causing disconnector heating faults is accurately located, providing a more accurate target for subsequent fault prediction and early warning.

[0042] Preferably, step S33 includes the following steps:

[0043] Step S331: Mark the fault nodes on the network topology data set according to the known high-voltage circuit breaker fault status data to obtain a fault-marked network data set;

[0044] Step S332: Mark the heating nodes on the fault-marked network data set according to the historical disconnector heating status data to obtain a heating-marked network data set;

[0045] Step S333: Screen the potential risk edges of the heating-marked network data set to obtain a risk edge data set;

[0046] Step S334: Determine the risk propagation direction of the risk edge data set to obtain a directed risk edge data set;

[0047] Step S335: Conduct a time-sequence path search on the directed risk edge data set to obtain a candidate risk path set.

[0048] The present invention marks the fault nodes of the network topology data set according to the known high-voltage circuit breaker fault status data, clarifies the starting point of fault propagation, and provides a starting point for subsequent risk path search. The heating nodes of the fault-marked network data set are marked according to the historical disconnector heating status data, clarifies the end point of risk propagation, and provides a target for subsequent risk path search. The potential risk edges of the heating-marked network data set are screened to remove the edges irrelevant to fault propagation, narrow the search scope, and improve the efficiency of path search. The risk propagation direction of the risk edge data set is determined to exclude the paths of reverse propagation, further narrow the search scope, and improve the accuracy of path search. The timing path search is performed on the directed risk edge data set, considering the timing logic of fault propagation, ensuring that the found path conforms to the actual situation, and further improving the reliability of the candidate risk path set.

[0049] Preferably, step S34 includes the following steps:

[0050] Step S341: Calculate the initial path probabilities according to the historical fault data and the candidate risk path set to obtain an initial path probability set;

[0051] Step S342: Extract the environmental factor data from the preprocessed data set and quantify the influence of environmental factors to obtain an environmental impact factor set;

[0052] Step S343: Quantify the influence of the equipment operation and maintenance data on the equipment status to obtain an equipment status factor set;

[0053] Step S344: Modify the initial path probability set according to the environmental impact factor set and the equipment status factor set to obtain a modified path probability set;

[0054] Step S345: Evaluate the confidence level of the path probabilities of the modified path probability set to obtain a path probability data set.

[0055] The present invention calculates the initial path probability based on historical fault data and the candidate risk path set, assigns a probability value of fault propagation to each candidate path, and provides a quantitative basis for subsequent risk assessment. By extracting environmental factor data from the preprocessed data set and quantifying the influence, the influence of environmental factors on the heating probability of the disconnector is taken into account, improving the comprehensiveness and accuracy of the path probability calculation. By quantifying the influence of the equipment operation and maintenance data on the equipment state, the influence of the equipment's own state on the heating probability is taken into account, further improving the accuracy and reliability of the path probability calculation. The initial path probability is corrected according to the environmental impact factor set and the equipment state factor set, comprehensively considering the influence of environmental factors and equipment state on fault propagation, making the path probability calculation result closer to the actual situation. The confidence level of the corrected path probability is evaluated, the confidence interval of the path probability is given, the uncertainty of the path probability is quantified, and the reliability and credibility of the risk assessment result are improved.

[0056] Preferably, step S4 includes the following steps:

[0057] Step S41: Quantify the critical path state of the critical risk propagation path to obtain a quantified path data set; construct a fault propagation model based on historical fault data to obtain a fault propagation model;

[0058] Step S42: Obtain the real-time monitoring data of the high-voltage circuit breaker and the disconnector; initialize the model state according to the real-time monitoring data of the high-voltage circuit breaker and the disconnector to obtain an initial state probability vector;

[0059] Step S43: Use the fault propagation model to perform state probability iterative calculation according to the initial state probability vector to obtain a state probability matrix;

[0060] Step S44: Calculate the path propagation probability according to the state probability matrix to obtain a path probability matrix;

[0061] Step S45: Perform dynamic sliding of the time window according to the path probability matrix to obtain a sequence of sliding path probability matrices;

[0062] Step S46: Integrate the probability matrix in the time dimension of the sequence of sliding path probability matrices to obtain a time-integrated probability matrix; compress the time dimension of the time-integrated probability matrix to obtain a dynamic fault probability matrix.

[0063] The present invention quantifies the node states on the critical risk propagation path, converts qualitative descriptions into quantitative data, and provides a data basis for subsequent model calculations. At the same time, a fault propagation model is constructed based on historical fault data, quantifying the fault propagation relationship between nodes and providing a model basis for dynamic fault probability assessment. Real-time monitoring data of high-voltage circuit breakers and disconnectors are obtained and converted into the initial state probability vector of the model. Combining the static model with real-time data provides real-time input for dynamic fault probability assessment. State probability iterative calculations are performed using the fault propagation model and the initial state probability vector, simulating the dynamic propagation process of faults on the critical risk propagation path and obtaining the state probabilities of each node at each time step, providing a dynamic probability basis for path propagation probability calculation. The path propagation probability is calculated based on the state probability matrix, obtaining the propagation probability of each critical risk propagation path at each time step and realizing real-time risk assessment of the fault propagation path. Time window dynamic sliding is performed according to the path probability matrix, converting discrete time steps into continuous time windows, which is closer to the actual fault development process and provides more refined time dimension information for subsequent dynamic analysis of fault probabilities. The time dimension integration and compression of the sliding path probability matrix sequence integrate the path propagation probabilities of different time windows, and finally obtain a dynamic fault probability matrix, which comprehensively combines historical data, real-time data, and time dimension information, providing a more comprehensive and accurate probability basis for subsequent risk prediction.

[0064] Preferably, step S5 includes the following steps:

[0065] Step S51: Extract disconnector features from the associated operation feature set to obtain a disconnector feature data set; extract environmental factor features from the associated operation feature set to obtain an environmental factor feature data set;

[0066] Step S52: Perform feature set fusion on the dynamic fault probability matrix, the disconnector feature data set, and the environmental factor feature data set to obtain a fused feature data set;

[0067] Step S53: Divide the fused feature data set into a training set and a test set; use the training set to construct a prediction model for a preset neural network model to obtain a training model for predicting the heating risk; use the test set to evaluate the model performance of the training model for predicting the heating risk to obtain a prediction model for the heating risk;

[0068] Step S54: Use the prediction model for the heating risk to predict the heating risk and obtain the heating risk probability;

[0069] Step S55: Divide the heating risk probability into risk levels according to a preset risk threshold to obtain risk level data;

[0070] Step S56: Calculate the comprehensive risk index based on the risk level data to obtain the comprehensive overheating risk index.

[0071] In the present invention, by separately extracting the disconnector features and environmental factor features from the associated operation feature set, a more concise and targeted feature data set is constructed, providing more valuable input information for subsequent model training, improving the model training efficiency and prediction accuracy. By fusing the dynamic fault probability matrix, the disconnector feature data set, and the environmental factor feature data set, a fused feature data set containing more comprehensive information is constructed, which can more comprehensively depict the operation state and risk factors of the disconnector, providing data guarantee for constructing a high-precision prediction model. The fused feature data set is divided into a training set and a test set, and the preset neural network model is trained using the training set and the model is evaluated using the test set, ensuring the generalization ability and prediction accuracy of the model, and finally obtaining a good-performing overheating risk prediction model. Using the trained overheating risk prediction model to predict the overheating risk of the disconnector, a quantitative overheating risk probability is obtained, providing a direct basis for risk assessment and early warning. According to the preset risk threshold, the overheating risk probability is divided into risk levels, converting the continuous probability value into a discrete risk level, more intuitively reflecting the risk degree, facilitating the maintenance personnel to quickly judge the risk level and take corresponding measures. Calculate the comprehensive risk index according to the risk level data, assign different weights to different risk levels, and obtain an index that can comprehensively reflect various risk factors, providing a more comprehensive and refined reference basis for risk early warning and decision-making.

[0072] Preferably, step S6 includes the following steps:

[0073] Step S61: Obtain historical early warning data; perform fault early warning data annotation on the historical early warning data to obtain a historical early warning data set;

[0074] Step S62: Construct an early warning strategy library according to the historical early warning data set to obtain an early warning strategy library;

[0075] Step S63: Build a reinforcement learning environment according to the historical early warning data set and the early warning strategy library to obtain an early warning simulation environment;

[0076] Step S64: Use the early warning simulation environment to train a reinforcement learning model to obtain an adaptive early warning model;

[0077] Step S65: Generate an adaptive early warning threshold according to the comprehensive overheating risk index and the adaptive early warning model to obtain an adaptive early warning threshold;

[0078] Step S66: Use the adaptive early warning threshold to perform overheating fault early warning on the high-voltage disconnector to realize the overheating fault prediction operation of the high-voltage disconnector.

[0079] Through obtaining historical warning data and performing fault warning data annotation, the present invention constructs a historical warning data set including warning information and actual fault conditions, providing a data basis for subsequent construction of a warning strategy library and training of an adaptive warning model. According to the historical warning data set, a warning strategy library is constructed, and a variety of warning strategies are formulated in advance, covering different risk levels and countermeasures, providing a basis for strategy selection for the adaptive warning model. According to the historical warning data set and the warning strategy library, a reinforcement learning environment is built, simulating the real power system operation environment and warning decision-making process, providing a realistic experimental platform for training the adaptive warning model. The reinforcement learning model is trained using the warning simulation environment, enabling the model to continuously learn and optimize warning strategies in the interaction with the environment, and finally obtaining an adaptive warning model that can adaptively select the optimal warning strategy according to real-time risk conditions. According to the comprehensive risk index and the adaptive warning model, an adaptive warning threshold is generated, realizing the dynamic adjustment of the warning threshold according to real-time risk conditions, improving the sensitivity and accuracy of warning, and avoiding false alarms and missed alarms prone to occur in the traditional fixed-threshold warning method. Using the adaptive warning threshold to carry out heat generation fault warning of high-voltage disconnectors, converting the risk prediction result into a specific warning action, notifying the operation and maintenance personnel in time to take countermeasures, effectively preventing the occurrence of faults, and ensuring the safe and stable operation of the power system.

[0080] Preferably, the present invention also provides a heat generation fault prediction system for the disconnector of a high-voltage circuit breaker, which is used to execute the heat generation fault prediction method for the disconnector of the high-voltage circuit breaker as described above. The heat generation fault prediction system for the disconnector of the high-voltage circuit breaker includes:

[0081] A multi-source data correlation analysis module, which is used to collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; perform correlation analysis of operation characteristics on the preprocessed data set to obtain a set of correlated operation characteristics;

[0082] A dynamic correlation network construction module, which is used to construct a dynamic correlation network for the set of correlated operation characteristics to obtain a dynamic correlation network diagram;

[0083] A risk propagation path identification module, which is used to obtain historical fault data; search for propagation risk paths according to the historical fault data and the dynamic correlation network diagram to obtain a set of candidate risk paths; calculate the path propagation probability for the set of candidate risk paths to obtain a path probability data set; extract key risk propagation paths according to the path probability data set and the set of candidate risk paths to obtain key risk propagation paths;

[0084] A fault probability dynamic evaluation module, which is used to calculate the path propagation probability according to the key risk propagation path to obtain a path probability matrix; and perform dynamic analysis of the fault probability according to the path probability matrix to obtain a dynamic fault probability matrix.

[0085] A comprehensive heat risk prediction module, which is used to construct a heat risk prediction model according to the associated operation feature set and the dynamic fault probability matrix to obtain a heat risk prediction model; use the heat risk prediction model to predict the heat risk, and calculate the comprehensive risk index to obtain the comprehensive heat risk index.

[0086] An early warning adaptive adjustment module, which is used to obtain historical early warning data; construct an adaptive early warning model according to the historical early warning data to obtain an adaptive early warning model; perform adaptive early warning of the heating fault of the high-voltage disconnector according to the comprehensive heat risk index and the adaptive early warning model to realize the prediction operation of the heating fault of the high-voltage disconnector.

[0087] Preferably, a high-voltage circuit breaker includes a contact system, an arc extinguishing system, and an operating mechanism connected to both of them, and further includes a controller connected to the operating mechanism. The controller includes the disconnector heating fault prediction system of the high-voltage circuit breaker as described above.

[0088] By collecting multi-source data and performing preprocessing, the present invention not only ensures the data quality, but also extracts the key operating characteristics related to high-voltage circuit breakers, laying a data foundation for subsequent analysis of the mutual influence between devices and construction of a prediction model. By constructing a dynamic association network, the association relationship between devices and its trend of change over time are intuitively displayed, providing an intuitive visual guidance for identifying key risk propagation paths. By analyzing historical fault data and the dynamic association network, potential risk propagation paths are found, and the propagation probability of each path is evaluated, providing a basis for accurately positioning key risk paths. By constructing a fault propagation model and combining it with real-time monitoring data for dynamic analysis, real-time risk assessment of key risk propagation paths is achieved, providing a dynamic probability basis for predicting the heating fault of disconnectors. By constructing a prediction model based on association features and dynamic fault probabilities, the future heating risk of disconnectors can be predicted more accurately, and through the comprehensive risk index, a more intuitive reference basis for risk warning is provided. By constructing an adaptive warning model and dynamically adjusting the warning threshold according to the comprehensive risk index, a more accurate and flexible warning mechanism is achieved, effectively reducing the risks of false alarms and missed alarms, and improving the efficiency and accuracy of warning. Therefore, the present invention provides a method for predicting the heating fault of disconnectors of high-voltage circuit breakers. By introducing technologies such as multi-source data association analysis, dynamic association network construction, and risk propagation path identification, the drawback that the existing methods lack consideration of the influence on related high-voltage circuit breakers is effectively solved. This method can effectively identify the influence of related high-voltage circuit breakers on the heating risk of disconnectors, thereby improving the accuracy of fault prediction and the timeliness of warning, and providing a more reliable guarantee for the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a schematic diagram of the step flow of a method for predicting the heating fault of a disconnector of a high-voltage circuit breaker;

[0090] Figure 2 is Figure 1 a detailed implementation step flow diagram of step S3 in

[0091] Figure 3 is Figure 1 a detailed implementation step flow diagram of step S5 in

[0092] The implementation, functional features, and advantages of the present invention will be further described in conjunction with embodiments with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0093] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0094] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0095] It should be understood that although terms such as "first" and "second" may be used here to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0096] To achieve the above object, please refer to Figures 1 to 3 , a method for predicting the heating fault of the disconnector of a high-voltage circuit breaker, comprising the following steps:

[0097] Step S1: Collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; perform an associated analysis of operating characteristics on the preprocessed data set to obtain an associated operating characteristics set;

[0098] Step S2: Construct a dynamic association network for the associated operating characteristics set to obtain a dynamic association network diagram;

[0099] Step S3: Obtain historical fault data; search for propagation risk paths based on the historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculate the path propagation probability for the candidate risk path set to obtain a path probability data set; extract the key risk propagation path based on the path probability data set and the candidate risk path set to obtain the key risk propagation path;

[0100] Step S4: Calculate the path propagation probability according to the key risk propagation path to obtain the path probability matrix; perform dynamic analysis of the fault probability based on the path probability matrix to obtain the dynamic fault probability matrix;

[0101] Step S5: Construct a heating risk prediction model based on the associated operation feature set and the dynamic fault probability matrix to obtain the heating risk prediction model; use the heating risk prediction model to predict the heating risk and calculate the comprehensive risk index to obtain the comprehensive heating risk index;

[0102] Step S6: Obtain historical warning data; construct an adaptive warning model based on the historical warning data to obtain the adaptive warning model; perform adaptive warning of the heating fault of the high-voltage disconnector according to the comprehensive heating risk index and the adaptive warning model to realize the prediction operation of the heating fault of the high-voltage disconnector.

[0103] In the embodiment of the present invention, refer to Figure 1 As described, it is a schematic diagram of the step flow of the method for predicting the heating fault of the disconnector of the high-voltage circuit breaker of the present invention. In this example, the method for predicting the heating fault of the disconnector of the high-voltage circuit breaker includes the following steps:

[0104] Step S1: Collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; perform associated operation feature analysis on the preprocessed data set to obtain an associated operation feature set;

[0105] In the embodiment of the present invention, multi-source data such as temperature, current, voltage, opening and closing states, etc. of the target high-voltage disconnector and its associated high-voltage circuit breaker are collected through the SCADA system within a period of time to form an original data set. The Python script is used to clean the original data set, remove missing values and outliers, and perform time alignment to obtain the preprocessed data set. Then, the principal component analysis method (PCA) is used to extract the associated operation features of the preprocessed data set to obtain an associated operation feature set including the main operation features of the disconnector and the associated high-voltage circuit breaker.

[0106] Step S2: Construct a dynamic association network for the associated operation feature set to obtain a dynamic association network diagram;

[0107] In the embodiments of the present invention, the associated operation feature set is segmented into multiple time segments with a 24-hour period in chronological order, and each segment forms a sample to form a time series feature data set. The mutual information method is used to calculate the association relationship between features within each sample, obtaining a daily association relationship matrix, and all matrices are arranged in chronological order to form an association relationship matrix sequence. An association strength threshold is set to screen out strong association relationships, and the Gephi software is used to draw a dynamic association network diagram to display the association relationship and evolution trend between devices.

[0108] Step S3: Obtain historical fault data; search for propagation risk paths based on the historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculate the path propagation probability for the candidate risk path set to obtain a path probability data set; extract the key risk propagation paths based on the path probability data set and the candidate risk path set to obtain the key risk propagation paths;

[0109] In the embodiments of the present invention, the NetworkX software is used to perform topological structure analysis on the dynamic association network diagram to obtain a network topology data set containing node attributes and network structure information. The historical high-voltage circuit breaker fault status data, historical disconnector heating status data, and equipment operation and maintenance data are obtained from the database. According to the historical fault data, the high-voltage circuit breaker nodes and heating disconnector nodes that have failed in the network topology data set are marked, and the potential risk edges connecting the fault nodes and heating nodes are screened. The risk propagation direction is determined based on the fault time and heating time to obtain a directed risk edge data set. A time series path search is performed in the directed risk edge data set to obtain a candidate risk path set.

[0110] Step S4: Calculate the path propagation probability based on the key risk propagation paths to obtain a path probability matrix; perform dynamic analysis of the fault probability based on the path probability matrix to obtain a dynamic fault probability matrix;

[0111] In the embodiments of the present invention, for each candidate risk path, the frequency of simultaneous occurrence of each edge on the path in the historical data is calculated as the propagation probability, and the propagation probabilities of all edges are multiplied to obtain the initial path propagation probability. The environmental factor data is extracted from the preprocessed data set, and the linear regression model is used to quantify the impact of environmental factors on the heating probability of the disconnector to obtain an environmental impact factor set. The device status information is extracted from the equipment operation and maintenance data, and the decision tree model is used to quantify the impact of the device status on the heating probability of the disconnector to obtain a device status factor set. The initial path probability is corrected according to the environmental impact factor and the device status factor, and the Bootstrap method is used for confidence evaluation to finally obtain a path probability data set containing the path propagation probability and its confidence interval. According to a preset sorting path threshold, such as 0.05, the paths with a propagation probability greater than or equal to this threshold are selected as the key risk propagation paths.

[0112] Step S5: Construct a heating risk prediction model based on the associated operation feature set and the dynamic fault probability matrix to obtain the heating risk prediction model; use the heating risk prediction model to predict the heating risk and calculate the comprehensive risk index to obtain the comprehensive heating risk index;

[0113] In the embodiment of the present invention, the node states on the key risk propagation path are quantified, and a fault propagation model is constructed using a Bayesian network. Real-time monitoring data is obtained, and the model state probability vector is initialized. The state probability matrix is obtained by performing state probability iterative calculation using the fault propagation model, and the path probability matrix is calculated based on the state probability matrix. The path probability matrix is subjected to time window sliding and time dimension integration to finally obtain the dynamic fault probability matrix. A heating risk prediction model is trained using the associated operation feature set and the dynamic fault probability matrix. The heating risk prediction is performed using this model to obtain the heating risk probability, and the risk level is divided according to the preset risk threshold, and finally the comprehensive risk index is calculated.

[0114] Step S6: Obtain historical warning data; construct an adaptive warning model based on the historical warning data to obtain the adaptive warning model; perform adaptive warning of the heating fault of the high-voltage disconnector according to the comprehensive heating risk index and the adaptive warning model to realize the heating fault prediction operation of the high-voltage disconnector;

[0115] In the embodiment of the present invention, historical warning data is obtained, and it is marked whether each warning data corresponds to a real fault event. An early warning strategy library containing multiple early warning strategies is constructed based on the historical warning data. The reinforcement learning tool OpenAI Gym is used to construct an early warning simulation environment, and the DQN algorithm is used to train the adaptive early warning model. The optimal early warning strategy is selected according to the comprehensive risk index obtained in S5 and the adaptive early warning model, and the adaptive early warning threshold is determined. Continuously monitor the operation state of the disconnector. When the comprehensive risk index exceeds the adaptive early warning threshold, execute the early warning strategy to notify the operation and maintenance personnel to take measures.

[0116] Preferably, step S1 includes the following steps:

[0117] Step S11: Collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system to obtain the original data set; perform data preprocessing on the original data set to obtain the preprocessed data set;

[0118] Step S12: Align the data time of the preprocessed data set to obtain the time-aligned data set;

[0119] Step S13: Establish an association relationship between the high-voltage disconnector and its associated high-voltage circuit breaker to obtain the association relationship data;

[0120] Step S14: Perform feature matching for the time-aligned data set within the same time window according to the association relationship data to obtain a feature matching data set;

[0121] Step S15: Generate an associated operation feature set according to the feature matching data set to obtain an associated operation feature set.

[0122] In the embodiment of the present invention, an original data set of the target high-voltage disconnector and its associated high-voltage circuit breaker is collected through the SCADA system during the period from January 1, 2023 to December 31, 2023. The original data set includes data such as the temperature, current, and voltage of the disconnector recorded every 1 minute, as well as the opening and closing states, current, and voltage of the associated high-voltage circuit breaker. A data cleaning script is written in the Python language to remove missing values and outliers from the original data set. For missing values, linear interpolation is used for filling; for outliers, the 3σ criterion is used for identification and elimination. Finally, a preprocessed data set is obtained. The data in the preprocessed data set is aligned in time using the Pandas library in Python. First, the collection time of all data is uniformly converted into the standard time format (e.g., YYYY-MM-DD HH:MM:SS). Then, in seconds, the data of different devices is aligned according to the time stamp to ensure that the data of all devices exists completely at the same time point. If there is a situation where the time stamps are inconsistent, the data of other devices is linearly interpolated and filled based on the data of the device with the earliest occurrence of the time stamp, and finally a time-aligned data set is obtained. According to the primary wiring diagram of the power system and the equipment operation and maintenance records, the association relationship data between the high-voltage disconnector and its associated high-voltage circuit breaker is established. For example, the number of the high-voltage circuit breaker connected to each disconnector is recorded, as well as the connection method between the two (e.g., series or parallel). These association relationship data are stored in the database for subsequent steps to call. The time window length is set to 1 hour. According to the association relationship data established in step S13, the high-voltage disconnector data and the associated high-voltage circuit breaker data within the same time window in the time-aligned data set are feature-matched. For example, the temperature, current, and voltage data of a certain disconnector within a certain hour are matched with the opening and closing states, current, and voltage data of its associated high-voltage circuit breaker within the same hour to form a feature vector. Finally, a feature matching data set containing the feature matching results of all time windows is obtained. The associated operation feature set is generated using the feature matching data set. The principal component analysis method (PCA) is used to reduce the dimension of the feature vectors within each time window in the feature matching data set, extract the main features that can characterize the operation states of the high-voltage disconnector and the associated high-voltage circuit breaker within that time window, and combine these main features to form the associated operation feature set. For example, the associated operation feature set within a certain time window includes features such as the average temperature of the disconnector, the number of opening and closing operations of the associated high-voltage circuit breaker, and the maximum current of the associated high-voltage circuit breaker.

[0123] Preferably, step S2 includes the following steps:

[0124] Step S21: Perform time series segmentation on the associated operation feature set to obtain a time series feature data set;

[0125] Step S22: Calculate the dynamic correlation relationship of the time series feature dataset to obtain a correlation relationship matrix;

[0126] Step S23: Perform dynamic evolution of the correlation relationship on the correlation relationship matrix to obtain a sequence of dynamic correlation matrices;

[0127] Step S24: Set the correlation strength threshold according to the sequence of dynamic correlation matrices and draw a network diagram to obtain a dynamic correlation network diagram.

[0128] In the embodiment of the present invention, the correlation operation feature set obtained in step S1 is arranged in chronological order. Considering the periodic change of the operation state of the power system, the time series is segmented into multiple time segments with a period of 24 hours, and each time segment corresponds to one day. The correlation operation feature data within each time segment is used as an independent sample to form a time series feature dataset. For example, the correlation operation feature set during the period from 00:00:00 on January 1, 2023 to 23:59:59 on December 31, 2023 is segmented into 365 samples, and each sample contains the correlation operation feature data of all time windows within that day. The mutual information method is used to calculate the correlation relationship between every two features in the time series feature dataset. For the sample data of each day, calculate the mutual information value between every two features within the sample and store the results in a matrix, which is the correlation relationship matrix. The rows and columns of the matrix represent different features respectively, and the value of the matrix element represents the mutual information value between the corresponding features, which is used to quantify the correlation degree between the features. In order to capture the changing trend of the correlation relationship over time, the daily correlation relationship matrices calculated in step S22 are arranged in chronological order to form a sequence of correlation relationship matrices. For example, the correlation relationship matrices calculated daily within 365 days are arranged in date order to form a matrix sequence with a length of 365, which is used to characterize the dynamic evolution process of the correlation relationship within one year. Analyze the changing trend of each element value in the sequence of correlation relationship matrices and set the correlation strength threshold in combination with expert experience. Compare all element values in the matrix sequence with this threshold. Elements higher than the threshold represent strong correlation relationships between the corresponding features, while elements lower than the threshold represent weak correlation relationships. According to the filtered strong correlation relationships, use the network diagram drawing tool Gephi to use features as nodes and strong correlation relationships as edges to draw a dynamic correlation network diagram. In the network diagram, the size of the node can represent the importance of the feature, the thickness of the edge can represent the correlation strength, and the depth of the color can represent the changing trend of the correlation relationship over time.

[0129] Preferably, step S3 includes the following steps:

[0130] Step S31: Analyze the network topology structure of the dynamic correlation network diagram to obtain a network topology dataset;

[0131] Step S32: Obtain historical fault data, where the historical fault data includes historical high-voltage circuit breaker fault status data, historical disconnector heating status data, and equipment operation and maintenance data;

[0132] Step S33: Search for propagation risk paths in the network topology dataset based on the historical fault data to obtain a candidate risk path set;

[0133] Step S34: Calculate the path propagation probability for the candidate risk path set to obtain a path probability dataset;

[0134] Step S35: Sort the candidate risk path set according to the path importance based on the path probability dataset to obtain a sorted path probability dataset;

[0135] Step S36: Use a preset sorted path threshold to screen the sorted path probability dataset to obtain critical risk propagation paths.

[0136] As an example of the present invention, referring to Figure 2 as shown, in this example, step S3 includes:

[0137] Step S31: Perform network topology structure analysis on the dynamic association network diagram to obtain a network topology dataset;

[0138] In the embodiment of the present invention, the NetworkX network analysis library is used to perform network topology structure analysis on the dynamic association network diagram obtained in step S2. All nodes (representing features) and the connection relationships between nodes (representing association relationships) in the network diagram are extracted to construct a network topology dataset. This dataset contains topological attribute information such as the degree, centrality, and clustering coefficient of the nodes, as well as network structure information such as the shortest path length and connectivity between nodes.

[0139] Step S32: Obtain historical fault data, where the historical fault data includes historical high-voltage circuit breaker fault status data, historical disconnector heating status data, and equipment operation and maintenance data;

[0140] In the embodiment of the present invention, historical fault data from January 1, 2022 to December 31, 2022 is obtained from the power system operation and maintenance database, including historical high-voltage circuit breaker fault status data, historical disconnector heating status data, and equipment operation and maintenance data. Among them, the historical high-voltage circuit breaker fault status data contains information such as the fault occurrence time, fault type, and faulty equipment; the historical disconnector heating status data contains information such as the heating time, heating degree, and heating equipment; the equipment operation and maintenance data contains information such as the equipment operation time, maintenance records, and environmental temperature.

[0141] Step S33: Search for propagation risk paths in the network topology dataset based on historical fault data to obtain a candidate risk path set;

[0142] In the embodiment of the present invention, in combination with the historical fault data obtained in step S32, a propagation risk path search is performed on the network topology dataset constructed in step S31. First, according to the historical fault state data of high-voltage circuit breakers and the historical heating state data of disconnectors, the high-voltage circuit breaker nodes with faults in the network topology dataset are marked as fault source nodes, and the disconnector nodes with heating are marked as target nodes. Then, the Dijkstra algorithm is used to search for all paths from the fault source nodes to the target nodes in the network topology dataset, and these paths are used as the candidate risk path set.

[0143] Step S34: Calculate the path propagation probability for the candidate risk path set to obtain a path probability dataset;

[0144] In the embodiment of the present invention, according to the historical fault data, the path propagation probability of each path in the candidate risk path set is calculated. The calculation method is as follows: For each edge (representing an association relationship) on the path, count the frequency of occurrence of this association relationship in the historical fault data as the propagation probability of this edge. For example, if the high-voltage circuit breaker and the disconnector connected by a certain edge have failed simultaneously 10 times in the past year, then the propagation probability of this edge is 10 / 365. Multiply the propagation probabilities of all edges on the path to obtain the propagation probability of this path.

[0145] Step S35: Sort the candidate risk path set according to the path probability dataset to obtain a sorted path probability dataset;

[0146] In the embodiment of the present invention, according to the path propagation probability calculated in step S34, the path importance of the candidate risk path set is sorted. Sort the path propagation probabilities from high to low. The higher the propagation probability of a path, the higher the risk of its occurrence of a fault and the higher its importance. Finally, a sorted path probability dataset is obtained.

[0147] Step S36: Use a preset sorted path threshold to screen the key paths from the sorted path probability dataset to obtain the key risk propagation paths;

[0148] In the embodiment of the present invention, the preset sorted path threshold is 0.05, and the key paths are screened from the sorted path probability dataset. Select the paths with a propagation probability greater than or equal to 0.05 as the key risk propagation paths. These paths represent the high-risk propagation paths leading to the heating faults of disconnectors and need to be focused on.

[0149] Preferably, step S33 includes the following steps:

[0150] Step S331: Mark the fault nodes in the network topology data set according to the known high-voltage circuit breaker fault status data to obtain a fault-marked network data set;

[0151] Step S332: Mark the overheated nodes in the fault-marked network data set according to the historical disconnector overheating status data to obtain an overheat-marked network data set;

[0152] Step S333: Screen the potential risk edges in the overheat-marked network data set to obtain a risk edge data set;

[0153] Step S334: Determine the risk propagation direction for the risk edge data set to obtain a directed risk edge data set;

[0154] Step S335: Search for the time-sequence paths in the directed risk edge data set to obtain a candidate risk path set.

[0155] In the embodiment of the present invention, the historical high-voltage circuit breaker fault status data obtained in the reading step S32 is read, and the high-voltage circuit breaker number where the fault occurs and the corresponding fault time are extracted. The node corresponding to the fault high-voltage circuit breaker number in the network topology dataset is marked as a fault node, and the fault occurrence time is recorded. For example, assuming that the historical data records that the high-voltage circuit breaker A fails at 10:00:00 on March 15, 2022, the node representing the high-voltage circuit breaker A in the network topology dataset is marked as a fault node, and the fault time is recorded as 10:00:00 on March 15, 2022. After all the fault nodes are marked, a fault-marked network dataset is obtained. The historical disconnector heating status data obtained in the reading step S32 is read, and the heating disconnector number and the corresponding heating time are extracted. The node corresponding to the heating disconnector number in the fault-marked network dataset processed in step S331 is marked as a heating node, and the heating time is recorded. For example, assuming that the historical data records that the disconnector B has heating at 10:30:00 on March 15, 2022, the node representing the disconnector B in the fault-marked network dataset is marked as a heating node, and the heating time is recorded as 10:30:00 on March 15, 2022. After all the heating nodes are marked, a heating-marked network dataset is obtained. In the heating-marked network dataset, all the edges connecting the fault nodes and the heating nodes are screened out as potential risk edges. For example, assuming that in the heating-marked network dataset, there is an edge connecting the fault node A and the heating node B, then this edge is regarded as a potential risk edge because the fault of the high-voltage circuit breaker A may cause the heating of the disconnector B. All the edges meeting the conditions are stored in the risk edge dataset. According to the fault time and the heating time recorded in steps S331 and S332, the risk propagation direction in the risk edge dataset is determined. If the fault time of a fault node of an edge is earlier than the heating time of its connected heating node, it is considered that the risk propagation direction of this edge is from the fault node to the heating node, and this edge is set as a directed edge with the direction from the fault node to the heating node; otherwise, it is considered that this edge does not form a risk propagation path and is removed from the risk edge dataset. After the direction is determined, a directed risk edge dataset is obtained. According to the directed risk edge dataset obtained in step S334, a temporal path search is performed in the heating-marked network dataset. Starting from each fault node, a depth-first search is performed along the direction of the directed edge to find all the paths that can reach the heating node. During the search process, it is required that the fault time or the heating time of each node on the path increases in chronological order to ensure that the path conforms to the temporal logic of risk propagation. All the found paths are used as the candidate risk path set.

[0156] Preferably, step S34 includes the following steps:

[0157] Step S341: Calculate the initial path probabilities based on historical failure data and the candidate risk path set to obtain the initial path probability set;

[0158] Step S342: Extract environmental factor data from the preprocessed data set and quantify the impact of environmental factors to obtain the environmental impact factor set;

[0159] Step S343: Quantify the impact on the device status for the device operation and maintenance data to obtain the device status factor set;

[0160] Step S344: Modify the initial path probability set according to the environmental impact factor set and the device status factor set to obtain the modified path probability set;

[0161] Step S345: Evaluate the confidence level of the path probabilities for the modified path probability set to obtain the path probability data set.

[0162] In the embodiments of the present invention, each path in the candidate risk path set obtained by traversing step S33 is traversed. For each path, the propagation probability of each edge on the path is calculated. The specific calculation method is as follows: The frequency of simultaneous occurrence of the two nodes (corresponding to an association relationship) connected by the edge is counted in the historical failure data. For example, if the number of times the high-voltage circuit breaker and the disconnecting switch connected by a certain edge have failed simultaneously in the past year is 10 times, then the propagation probability of this edge is 10 / 365. Multiply the propagation probabilities of all the edges on the path to obtain the initial propagation probability of this path. Store the initial propagation probabilities of all paths in the initial path probability set. Extract environmental factor data from the preprocessed data set obtained in step S1, including environmental temperature, humidity, wind speed, etc. Use a linear regression model to establish a relationship model between environmental factors and the heating state of the disconnecting switch respectively. According to the model coefficients, quantify the influence degree of environmental factors on the heating probability of the disconnecting switch, and store the quantification results in the environmental impact factor set. For example, if the heating probability of the disconnecting switch increases by 0.01 for every 1 degree Celsius increase in environmental temperature, then the impact factor of environmental temperature is set to 0.01. Extract device status information from the device operation and maintenance data obtained in step S32, including device operation time, maintenance times, failure times, etc. Use a decision tree model to establish a relationship model between device status and the heating state of the disconnecting switch. According to the importance of each feature in the model, quantify the influence degree of device status on the heating probability of the disconnecting switch, and store the quantification results in the device status factor set. For example, if the heating probability of the disconnecting switch increases by 0.05 for every 1000 hours increase in device operation time, then the impact factor of device operation time is set to 0.0005. According to the environmental impact factor set and the device status factor set obtained in steps S342 and S343, perform path probability correction on the initial path probability set obtained in step S341. The specific correction method is as follows: For each path, calculate the weighted sum of the environmental impact factor and the device status factor according to the environmental factors and device status corresponding to the nodes on the path, and use this weighted sum as the correction coefficient. Multiply the initial propagation probability of the path by the correction coefficient to obtain the corrected path propagation probability, and store all the corrected path propagation probabilities in the corrected path probability set. Use the Bootstrap method to evaluate the confidence level of each path probability in the corrected path probability set. The specific method is as follows: Perform multiple sampling with replacement on the original data set, and calculate the path propagation probability according to the methods in steps S341 to S344 after each sampling. Finally, obtain the distribution of each path propagation probability, and calculate the 95% confidence interval according to the distribution. Store the propagation probabilities and their confidence intervals of all paths in the path probability data set.

[0163] Preferably, step S4 includes the following steps:

[0164] Step S41: Quantify the critical path status of the critical risk propagation path to obtain a quantified path data set; construct a fault propagation model based on historical fault data to obtain a fault propagation model;

[0165] Step S42: Obtain the real-time monitoring data of the high-voltage circuit breaker and the disconnector; initialize the model state according to the real-time monitoring data of the high-voltage circuit breaker and the disconnector to obtain an initial state probability vector;

[0166] Step S43: Use the fault propagation model to perform state probability iterative calculation according to the initial state probability vector to obtain a state probability matrix;

[0167] Step S44: Calculate the path propagation probability according to the state probability matrix to obtain a path probability matrix;

[0168] Step S45: Perform dynamic sliding of the time window according to the path probability matrix to obtain a sequence of sliding path probability matrices;

[0169] Step S46: Integrate the time dimension of the sliding path probability matrix sequence to obtain a time-integrated probability matrix; compress the time dimension of the time-integrated probability matrix to obtain a dynamic fault probability matrix.

[0170] In the embodiment of the present invention, for each critical risk propagation path screened in step S3, the states of each node on the path are quantified to construct a quantified path data set. The quantification method is as follows:

[0171] For the high-voltage circuit breaker node, according to its real-time monitoring data, its state is quantified as 0 or 1, where 0 represents the normal state and 1 represents the fault state. For example, if the real-time current of the high-voltage circuit breaker exceeds the preset threshold, it is considered that the high-voltage circuit breaker is in the fault state and its state is quantified as 1, otherwise it is quantified as 0.

[0172] For the disconnector node, according to its real-time monitoring data, its state is quantified as a continuous value between 0 and 1, indicating its heating degree. For example, 0 represents no heating and 1 represents severe heating. Fuzzy logic and other methods can be used to quantify the heating degree of the disconnector according to indicators such as temperature and current of the disconnector.

[0173] According to historical fault data, use a Bayesian network to construct a fault propagation model. This model describes the probabilistic dependence relationship between the states of each node on the critical risk propagation path. For example, the model can be expressed as: P(disconnector heating|high-voltage circuit breaker fault, ambient temperature), that is, the probability of the disconnector heating under the conditions of high-voltage circuit breaker fault and ambient temperature.

[0174] Obtain the real-time monitoring data of high-voltage circuit breakers and disconnectors through the SCADA system, including current, voltage, temperature, etc. According to the state quantization method defined in step S41, convert the real-time monitoring data into the initial state probability of each node in the model. For example, assume that a certain critical risk propagation path includes high-voltage circuit breaker A and disconnector B. According to the real-time monitoring data, the current of high-voltage circuit breaker A does not exceed the threshold, so its initial state probability is P(A = 0) = 0.9, P(A = 1) = 0.1; the temperature of disconnector B is 40 degrees Celsius, and its heating degree is 0.2 according to fuzzy logic quantization, so its initial state probability is P(B = 0.2) = 1. Combine the initial state probabilities of all nodes into an initial state probability vector. Use the fault propagation model constructed in step S41 and the initial state probability vector obtained in step S42 to perform state probability iterative calculations. During the iteration process, according to the inference mechanism of the Bayesian network, calculate the state probability of each node at each time step. For example, according to the fault propagation model P(B = 1|A = 1) = 0.8 and the initial state probability P(A = 1) = 0.1, the probability that disconnector B is in a fault state at the next time step can be calculated as 0.8 * 0.1 = 0.08. Store the state probabilities of all nodes at each time step in a matrix to obtain a state probability matrix. According to the state probability matrix, calculate the propagation probability of each critical risk propagation path at each time step. The specific calculation method is: multiply the probabilities that all nodes on the path are in a fault state at this time step. For example, assume that a certain path includes node A and node B. At a certain time step, P(A = 1) = 0.2, P(B = 1) = 0.1, then the propagation probability of this path at this time step is 0.2 * 0.1 = 0.02. Store the propagation probabilities of all paths at all time steps in a matrix to obtain a path probability matrix. Set the time window length to 1 hour and slide the path probability matrix dynamically according to the time window. For example, assume that the path probability matrix contains 24 time steps, then it can be slid into 23 sliding windows each containing 2 time steps. Each sliding window corresponds to a path probability matrix, forming a sequence of sliding path probability matrices. Concatenate all the matrices in the sequence of sliding path probability matrices in chronological order to obtain a time-integrated probability matrix. For example, concatenate the path probability matrices corresponding to 23 sliding windows in chronological order into a larger matrix. Then, perform time dimension compression on the time-integrated probability matrix, and average the propagation probabilities within each time window to obtain the final dynamic fault probability matrix.

[0175] Preferably, step S5 includes the following steps:

[0176] Step S51: Extract disconnector features from the associated operation feature set to obtain a disconnector feature data set; extract environmental factor features from the associated operation feature set to obtain an environmental factor feature data set;

[0177] Step S52: Perform feature set fusion on the dynamic fault probability matrix, the disconnector feature data set, and the environmental factor feature data set to obtain a fused feature data set;

[0178] Step S53: Divide the fused feature data set to obtain a training set and a test set; use the training set to construct a prediction model for a preset neural network model to obtain a fever risk prediction training model; use the test set to evaluate the performance of the fever risk prediction training model to obtain a fever risk prediction model;

[0179] Step S54: Use the fever risk prediction model to predict the fever risk to obtain a fever risk probability;

[0180] Step S55: Perform risk level division on the fever risk probability according to a preset risk threshold to obtain risk level data;

[0181] Step S56: Calculate a comprehensive risk index based on the risk level data to obtain a comprehensive fever risk index.

[0182] As an example of the present invention, referring to Figure 3 as shown, in this example, step S5 includes:

[0183] Step S51: Extract disconnector features from the associated operation feature set to obtain a disconnector feature data set; extract environmental factor features from the associated operation feature set to obtain an environmental factor feature data set;

[0184] In the embodiment of the present invention, features related to the target disconnector are extracted from the associated operation feature set obtained in step S1, including the temperature, current, voltage, etc. of the disconnector, to form a disconnector feature data set. At the same time, features related to environmental factors, such as environmental temperature, humidity, wind speed, etc., are extracted to form an environmental factor feature data set.

[0185] Step S52: Perform feature set fusion on the dynamic fault probability matrix, the disconnector feature data set, and the environmental factor feature data set to obtain a fused feature data set;

[0186] In the embodiments of the present invention, the dynamic fault probability matrix obtained in step S4, the disconnector feature dataset extracted in step S51, and the environmental factor feature dataset are subjected to feature set fusion to form a fused feature dataset. The fusion method is as follows: at each time step, the corresponding dynamic fault probability, disconnector features, and environmental factor features are concatenated into a feature vector, and the feature vectors of all time steps form the fused feature dataset.

[0187] Step S53: Divide the fused feature dataset to obtain a training set and a test set; use the training set to construct a prediction model for a preset neural network model to obtain a training model for predicting the heating risk; use the test set to evaluate the performance of the training model for predicting the heating risk to obtain a prediction model for the heating risk;

[0188] In the embodiments of the present invention, the fused feature dataset is divided into a training set and a test set according to a ratio of 7:3. The preset long short-term memory neural network (LSTM) model is trained using the training set to construct a training model for predicting the heating risk. The input of the LSTM model is the fused feature vector, and the output is the heating probability of the disconnector within a future period of time (such as the next 1 hour). The performance of the trained model is evaluated using the test set, and the evaluation metrics include accuracy, precision, recall, etc. Finally, a prediction model for the heating risk is obtained.

[0189] Step S54: Use the prediction model for the heating risk to predict the heating risk to obtain the heating risk probability;

[0190] In the embodiments of the present invention, the prediction model for the heating risk trained in step S53 is used to predict the heating risk of the disconnector at the current moment. The input of the model is the fused feature vector at the current moment, and the output is the heating risk probability of the disconnector within a future period of time (such as the next 1 hour).

[0191] Step S55: Divide the heating risk probability according to a preset risk threshold to obtain risk level data;

[0192] In the embodiments of the present invention, according to the preset risk thresholds, such as 0.1, 0.5, and 0.9, the heating risk probability obtained in step S54 is divided into different risk levels. For example, the case where the heating risk probability is lower than 0.1 is classified as a low risk level, the case where the heating risk probability is between 0.1 and 0.5 is classified as a medium risk level, and the case where the heating risk probability is higher than 0.9 is classified as a high risk level.

[0193] Step S56: Calculate the comprehensive risk index based on the risk level data to obtain the comprehensive heating risk index;

[0194] In the embodiment of the present invention, according to the risk level data divided in step S55, a comprehensive risk index is calculated. The calculation method is as follows: different weights are assigned to each risk level. For example, the weight of the low risk level is 1, the weight of the medium risk level is 3, and the weight of the high risk level is 5. The weight corresponding to the risk level at the current moment is used as the comprehensive risk index. For example, if the fever risk probability at the current moment is 0.6 and it belongs to the medium risk level, the comprehensive risk index is 3.

[0195] Preferably, step S6 includes the following steps:

[0196] Step S61: Obtain historical warning data; perform fault warning data annotation on the historical warning data to obtain a historical warning data set;

[0197] Step S62: Construct a warning strategy library according to the historical warning data set to obtain a warning strategy library;

[0198] Step S63: Build a reinforcement learning environment according to the historical warning data set and the warning strategy library to obtain a warning simulation environment;

[0199] Step S64: Use the warning simulation environment to train a reinforcement learning model to obtain an adaptive warning model;

[0200] Step S65: Generate an adaptive warning threshold according to the comprehensive fever risk index and the adaptive warning model to obtain an adaptive warning threshold;

[0201] Step S66: Use the adaptive warning threshold to perform a fever fault warning on the high-voltage disconnector to realize the fever fault prediction operation of the high-voltage disconnector.

[0202] In the embodiment of the present invention, historical warning data is obtained from the power system operation and maintenance database, including information such as warning time, warning level, warning reason, and actual fault situation. The historical warning data is annotated to mark whether each warning data corresponds to a real fault event. For example, if a certain warning data actually causes a fever fault of the disconnector within a certain period of time (for example, within 24 hours) after that, it is marked as "true warning", otherwise it is marked as "false alarm". Finally, a historical warning data set containing warning data and annotation information is obtained.

[0203] According to the true warning ratio and false alarm ratio corresponding to different warning levels in the historical warning data set, a warning strategy library is constructed. The warning strategy library contains a variety of warning strategies, and each strategy corresponds to different warning thresholds and warning methods. For example, three warning strategies can be set:

[0204] Strategy 1: When the comprehensive risk index is greater than or equal to 2, send a text message to notify the maintenance personnel to pay attention;

[0205] Strategy 2: When the comprehensive risk index is greater than or equal to 4, notify the operation and maintenance personnel by phone to take measures;

[0206] Strategy 3: When the comprehensive risk index is equal to 5, immediately start the emergency plan.

[0207] Use the reinforcement learning tool OpenAI Gym to build an early warning simulation environment. This environment simulates the operating state of the power system and calculates the comprehensive risk index according to the methods in steps S1 to S5. The state space of the environment is the value range of the comprehensive risk index, the action space is the numbers of all early warning strategies in the early warning strategy library, and the reward function is designed as follows: when the adopted early warning strategy can be timely and effective when a fault occurs, a positive reward is given; when there is a missed report or false alarm, a negative reward is given. Use the Deep Q-Network (DQN) algorithm to train the reinforcement learning model in the early warning simulation environment constructed in step S63. During the training process, the agent selects an early warning strategy from the early warning strategy library according to the current comprehensive risk index, and adjusts the strategy selection according to the feedback of the environment, and finally learns an adaptive early warning model that can adaptively select the optimal early warning strategy according to the comprehensive risk index. According to the comprehensive risk index calculated in step S5 and the adaptive early warning model trained in step S64, select the optimal early warning strategy. For example, if the current comprehensive risk index is 3 and the adaptive early warning model selects early warning strategy 1, then set the early warning threshold to 2 and send a text message to notify the operation and maintenance personnel to pay attention. Continuously monitor the operating state of the disconnector and update the comprehensive risk index in real time according to the methods in steps S1 to S5. When the comprehensive risk index exceeds the adaptive early warning threshold determined in step S65, immediately execute the corresponding early warning strategy, notify the operation and maintenance personnel to take corresponding measures to prevent the occurrence of the disconnector heating fault and ensure the safe and stable operation of the power system.

[0208] Preferably, the present invention also provides a disconnector heating fault prediction system for a high-voltage circuit breaker, which is used to execute the disconnector heating fault prediction method for the high-voltage circuit breaker as described above. The disconnector heating fault prediction system for the high-voltage circuit breaker includes:

[0209] A multi-source data correlation analysis module, which is used to collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, perform data preprocessing to obtain a preprocessed data set; perform correlation analysis of operation characteristics on the preprocessed data set to obtain a set of correlated operation characteristics;

[0210] A dynamic correlation network construction module, which is used to construct a dynamic correlation network for the set of correlated operation characteristics to obtain a dynamic correlation network diagram;

[0211] A risk propagation path identification module, which is used to obtain historical fault data; search for risk propagation paths according to the historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculate the path propagation probability for the candidate risk path set to obtain a path probability data set; extract key risk propagation paths according to the path probability data set and the candidate risk path set to obtain key risk propagation paths.

[0212] A fault probability dynamic assessment module, which is used to calculate the path propagation probability according to the key risk propagation paths to obtain a path probability matrix; perform dynamic analysis of the fault probability according to the path probability matrix to obtain a dynamic fault probability matrix.

[0213] A heating risk comprehensive prediction module, which is used to construct a heating risk prediction model according to the associated operation feature set and the dynamic fault probability matrix to obtain a heating risk prediction model; use the heating risk prediction model to predict the heating risk and calculate the comprehensive risk index to obtain the comprehensive heating risk index.

[0214] An early warning adaptive adjustment module, which is used to obtain historical early warning data; construct an adaptive early warning model according to the historical early warning data to obtain an adaptive early warning model; perform adaptive early warning of the heating fault of the high-voltage disconnector according to the comprehensive heating risk index and the adaptive early warning model to realize the prediction operation of the heating fault of the high-voltage disconnector.

[0215] Preferably, a high-voltage circuit breaker includes a contact system, an arc extinguishing system, and an operating mechanism connected to both of them, and further includes a controller connected to the operating mechanism. The controller includes the disconnector heating fault prediction system of the high-voltage circuit breaker as described above.

[0216] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application document are intended to be included in the present invention.

[0217] The above are only specific embodiments of the present invention, which enable those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for predicting thermal failure of an isolating switch of a high-voltage circuit breaker, characterized in that: The following steps are involved: Step S1: collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; Performing operation feature correlation analysis on the preprocessed data set to obtain a correlation operation feature set; Step S2: constructing a dynamic association network for the associated operation feature set to obtain a dynamic association network graph; Step S3: Acquire historical fault data; search for propagation risk paths based on historical fault data and the dynamic association network diagram to obtain a candidate risk path set; calculate the path propagation probability of the candidate risk path set to obtain a path probability data set; extract key risk propagation paths based on the path probability data set and the candidate risk path set to obtain key risk propagation paths; Step S4: Calculate the path propagation probability according to the key risk propagation path to obtain a path probability matrix; perform a dynamic analysis of the fault probability according to the path probability matrix to obtain a dynamic fault probability matrix; Step S5: construct a fever risk prediction model according to the associated operation feature set and the dynamic fault probability matrix to obtain a fever risk prediction model; use the fever risk prediction model to predict the fever risk, and calculate the comprehensive risk index to obtain a comprehensive fever risk index, wherein step S5 is specifically as follows: Step S51: extracting the isolating switch features from the associated operation feature set to obtain an isolating switch feature data set; Extracting environmental factor features from the associated operation feature set to obtain an environmental factor feature data set; Step S52: performing feature set fusion on the dynamic fault probability matrix, the disconnector feature data set, and the environmental factor feature data set to obtain a fused feature data set; Step S53: dividing the fused feature data set into a data set to obtain a training set and a test set; using the training set to construct a prediction model for the preset neural network model to obtain a fever risk prediction training model; using the test set to evaluate the model performance of the fever risk prediction training model to obtain a fever risk prediction model; Step S54: using the fever risk prediction model to predict the fever risk and obtain the fever risk probability; Step S55: classifying the fever risk probability into risk levels according to a preset risk threshold to obtain risk level data; Step S56: Calculate the comprehensive risk index according to the risk level data to obtain a comprehensive fever risk index, wherein the comprehensive risk index calculation is specifically as follows: Assign different weights to each risk level in the risk level data, where the low risk level has a weight of 1, the medium risk level has a weight of 3, and the high risk level has a weight of 5. The weights corresponding to the risk levels are used as the comprehensive risk index; Step S6: Obtain historical warning data; An adaptive early warning model is constructed according to historical early warning data to obtain an adaptive early warning model; Based on the comprehensive heating risk index and the adaptive warning model, adaptive warning of heating failure of high-voltage disconnectors is performed to realize the prediction of heating failure of high-voltage disconnectors.

2. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: performing multi-source data acquisition on the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system to obtain an original data set; performing data preprocessing on the original data set to obtain a preprocessed data set; Step S12: performing data time alignment on the preprocessed data set to obtain a time-aligned data set; Step S13: establishing an association relationship between the high-voltage disconnector and its associated high-voltage circuit breaker to obtain association relationship data; Step S14: performing feature matching of the same time window on the time-aligned data set according to the association relationship data to obtain a feature matching data set; Step S15: generating an associated operation feature set according to the feature matching data set to obtain an associated operation feature set; Step S2 includes the following steps: Step S21: performing time series segmentation of the associated operation feature set to obtain a time series feature data set; Step S22: Calculate the dynamic correlation relationship of the time series feature data set to obtain a correlation relationship matrix; Step S23: dynamically evolving the correlation relationship matrix to obtain a dynamic correlation matrix sequence; Step S24: setting the association strength threshold according to the dynamic association matrix sequence, and drawing a network diagram to obtain a dynamic association network diagram.

3. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing network topology structure analysis on the dynamic association network graph to obtain a network topology data set; Step S32: Acquire historical fault data, wherein the historical fault data includes historical high-voltage circuit breaker fault state data, historical disconnector heating state data, and equipment operation and maintenance data; Step S33: searching for a propagation risk path on the network topology data set according to the historical fault data to obtain a candidate risk path set; Step S34: Calculate the path propagation probability of the candidate risk path set to obtain a path probability data set; Step S35: sorting the candidate risk path set by path importance according to the path probability data set to obtain a sorted path probability data set; Step S36: Use the preset sorting path threshold to perform key path screening on the sorting path probability data set to obtain the key risk propagation path.

4. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 3, characterized in that: Step S33 includes the following steps: Step S331: marking fault nodes of the network topology data set according to known high-voltage circuit breaker fault status data to obtain a fault-marked network data set; Step S332: marking heating nodes of the fault marking network data set according to the historical isolating switch heating state data to obtain the heating marking network data set; Step S333: Screening the fever mark network data set for potential risk edges to obtain a risk edge data set; Step S334: determining the risk propagation direction of the risk edge data set to obtain a directed risk edge data set; Step S335: Perform a temporal path search on the directed risk edge data set to obtain a candidate risk path set.

5. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 3, characterized in that: Step S34 includes the following steps: Step S341: Calculate the initial path probability based on the historical fault data and the candidate risk path set to obtain an initial path probability set; Step S342: extracting environmental factor data from the preprocessed data set, and quantifying the impact of environmental factors to obtain an environmental impact factor set; Step S343: quantifying the impact of equipment status on equipment operation and maintenance data to obtain an equipment status factor set; Step S344: modifying the initial path probability set according to the environmental impact factor set and the device status factor set to obtain a modified path probability set; Step S345: performing path probability confidence evaluation on the modified path probability set to obtain a path probability data set.

6. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: quantify the critical path state of the critical risk propagation path to obtain a quantified path data set; construct a fault propagation model based on historical fault data to obtain a fault propagation model; Step S42: acquiring real-time monitoring data of the high-voltage circuit breaker and the disconnector; initializing the model state according to the real-time monitoring data of the high-voltage circuit breaker and the disconnector to obtain an initial state probability vector; Step S43: using the fault propagation model to perform iterative state probability calculation according to the initial state probability vector to obtain a state probability matrix; Step S44: Calculate the path propagation probability according to the state probability matrix to obtain a path probability matrix; Step S45: dynamically sliding the time window according to the path probability matrix to obtain a sliding path probability matrix sequence; Step S46: integrating the probability matrix time dimension of the sliding path probability matrix sequence to obtain a time-integrated probability matrix; and compressing the time-integrated probability matrix in the time dimension to obtain a dynamic fault probability matrix.

7. The method for predicting thermal failure of the isolating switch of a high-voltage circuit breaker according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: Acquire historical warning data; perform fault warning data labeling on the historical warning data to obtain a historical warning data set; Step S62: constructing a warning strategy library according to the historical warning data set to obtain a warning strategy library; Step S63: Building a reinforcement learning environment based on the historical warning data set and the warning strategy library to obtain a warning simulation environment; Step S64: using the early warning simulation environment to perform reinforcement learning model training to obtain an adaptive early warning model; Step S65: generating an adaptive warning threshold according to the comprehensive fever risk index and the adaptive warning model to obtain an adaptive warning threshold; Step S66: Use the adaptive warning threshold to perform a high-voltage disconnector thermal fault warning to achieve a high-voltage disconnector thermal fault prediction operation.

8. A thermal fault prediction system for a disconnector of a high-voltage circuit breaker, characterized in that: Used to execute the method for predicting thermal failure of the isolating switch of the high-voltage circuit breaker according to claim 1, the thermal failure prediction system of the isolating switch of the high-voltage circuit breaker comprises: The multi-source data association analysis module is used to collect multi-source data of the high-voltage disconnector and its associated high-voltage circuit breaker through the SCADA system, and perform data preprocessing to obtain a preprocessed data set; perform operation feature association analysis on the preprocessed data set to obtain an associated operation feature set; A dynamic association network construction module is used to construct a dynamic association network for the association operation feature set to obtain a dynamic association network graph; The risk propagation path identification module is used to obtain historical fault data; search for propagation risk paths based on historical fault data and dynamic association network diagrams to obtain a candidate risk path set; calculate the path propagation probability of the candidate risk path set to obtain a path probability data set; extract key risk propagation paths based on the path probability data set and the candidate risk path set to obtain key risk propagation paths; The fault probability dynamic assessment module is used to calculate the path propagation probability according to the key risk propagation path to obtain the path probability matrix; perform a dynamic analysis of the fault probability according to the path probability matrix to obtain a dynamic fault probability matrix; The fever risk comprehensive prediction module is used to construct a fever risk prediction model according to the associated operation feature set and the dynamic fault probability matrix to obtain the fever risk prediction model; the fever risk prediction model is used to predict the fever risk and calculate the comprehensive risk index to obtain the comprehensive fever risk index; The early warning adaptive adjustment module is used to obtain historical early warning data; construct an adaptive early warning model based on the historical early warning data to obtain an adaptive early warning model; and perform an adaptive early warning of high-voltage disconnector thermal faults based on a comprehensive thermal risk index and an adaptive early warning model to realize the high-voltage disconnector thermal fault prediction operation.

9. A high voltage circuit breaker, characterized in that: It comprises a contact system, an arc extinguishing system and an operating mechanism connected to the two, and also comprises a controller connected to the operating mechanism, wherein the controller comprises the isolating switch thermal fault prediction system of the high-voltage circuit breaker as claimed in claim 8.

Images