Method and system for monitoring faults of connector in real time

Abstract

The invention relates to the technical field of connector fault monitoring, in particular to a method and a system for monitoring a connector fault in real time. The method comprises the following steps of performing operation state data acquisition on a connector, and performing timestamp marking to obtain operation state characteristic time sequence marking data; performing abnormal state identification on the operation state characteristic time sequence mark data to obtain current / voltage abnormal fluctuation data; according to the current / voltage abnormal fluctuation data, carrying out fault triggering pre-risk factor identification to obtain a fault triggering pre-risk factor; performing fault lag severity level prediction according to the fault triggering pre-risk factor to obtain fault lag severity level prediction data; and performing connector fault identification model construction on the fault lagging severity level prediction data and the fault triggering pre-risk factors to obtain a connector fault identification model. According to the invention, the connector fault monitoring technology is optimized, so that the connector fault monitoring technology is more perfect.

Description

Technical Field

[0001] The present invention relates to the technical field of connector fault monitoring, and in particular to a method and system for real-time monitoring of connector faults. Background Art

[0002] As a vital component, connectors are widely used in industries such as electricity, communications, transportation, and aerospace. Connector failure will not only lead to system performance degradation, but may even cause serious equipment failure, thereby affecting the safety and stability of the entire system. Therefore, real-time monitoring of connector failures is of great significance for improving equipment reliability and maintenance efficiency. Connector failures usually manifest themselves in the form of poor contact, aging, corrosion, overload, or physical damage, which are often not easily detected directly by external monitoring tools or ordinary sensors. Therefore, developing a method that can monitor connectors in real time and accurately has become a difficult problem that needs to be solved urgently in academia and industry. Especially in some high-risk and high-demand fields, such as aerospace, military equipment, and new energy vehicles, connector fault monitoring is not only related to the operating efficiency of the equipment itself, but also involves personal safety and the protection of important facilities. However, a traditional method for real-time monitoring of connector failures has the problems of inaccurate analysis of performance release imbalance and inaccurate identification of pre-failure risk factors. Summary of the invention

[0003] Based on this, it is necessary to provide a method and system for real-time monitoring of connector faults to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for real-time monitoring of connector faults is provided, the method comprising the following steps:

[0005] Step S1: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector; timestamping the operation status data of the connector to obtain the operation status characteristic time series marking data;

[0006] Step S2: using a preset abnormal risk identification model to identify abnormal states of the operating state feature time series marking data, and obtain current / voltage abnormal fluctuation data; performing performance release imbalance analysis based on the current / voltage abnormal fluctuation data, and obtaining abnormal current / voltage performance imbalance impact data; performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data, and obtaining the fault triggering pre-risk factor;

[0007] Step S3: performing time lag analysis on the current / voltage abnormal fluctuation data according to the pre-risk factor for fault triggering to obtain pre-risk current / voltage lag data; performing fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data;

[0008] Step S4: Based on the policy gradient algorithm, a connector fault identification model is constructed for the fault lag severity level prediction data and the fault triggering pre-risk factors to obtain the connector fault identification model, and the connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

[0009] The present invention uses a distributed sensor system to monitor and collect data on the connector in real time, and comprehensively obtains the operating status data of the connector, including multi-dimensional information such as current, voltage, and temperature, to ensure that the collected status data has high representativeness and integrity. By timestamping these data, each data point can be accurately associated with the corresponding time to form labeled data with time series characteristics. The time series labeled data is not only convenient for subsequent analysis, but also can reveal the changes in the operating status of the connector at different time points, providing a solid foundation for abnormal detection and fault prediction. First, the preset abnormal risk identification model is used to analyze the operating status characteristic time series labeled data obtained in the previous step, and abnormal fluctuations of current or voltage are automatically identified. Through in-depth analysis of abnormal fluctuations, performance release imbalance analysis is further performed to identify performance imbalance caused by current / voltage instability. These analysis results can accurately reflect the potential performance degradation or failure risk of the connector. Then, based on the performance imbalance data of abnormal current / voltage, the fault trigger pre-risk factor identification is further carried out to discover potential fault risk factors in advance and predict the occurrence time of the fault. Through this series of analysis, early warning can be achieved, the probability of equipment failure can be reduced, and the stability and reliability of the system can be improved. By analyzing the time lag relationship between the fault triggering pre-risk factor and the current / voltage abnormal fluctuation data, the time delay characteristics of the current / voltage fluctuation before the fault occurs are identified. Through hysteresis analysis, the time difference between the abnormal fluctuation and the potential fault can be tracked, and the pre-risk current / voltage hysteresis data can be further obtained. These hysteresis data provide an accurate time window for subsequent fault prediction. Based on these hysteresis data, machine learning or statistical methods are used to predict the severity of the fault hysteresis, evaluate the severity of the fault and its impact on the performance of the connector. This prediction can provide an important basis for maintenance decisions, identify serious faults in advance, and prepare for emergency treatment. Using the policy gradient algorithm, a reinforcement learning method, a connector fault identification model is constructed based on the comprehensive analysis of the fault hysteresis severity level prediction data and the pre-risk factor. The policy gradient algorithm can dynamically adjust the identification strategy by self-optimizing the model to improve the accuracy and flexibility of fault detection. The model can not only identify the current fault risk, but also adjust the fault identification strategy in time according to historical data and prediction results, making fault monitoring more accurate. The built fault identification model will be uploaded to the cloud platform, using the powerful computing power of cloud computing for real-time monitoring and data analysis. The cloud platform can efficiently process large amounts of data from various sensors, detect connector faults in real time, and provide timely fault alerts and suggestions to operation and maintenance personnel, greatly improving fault diagnosis and response speed.Therefore, the present invention is an optimization processing of a traditional method for real-time monitoring of connector faults, which solves the problems of inaccurate performance release imbalance analysis and inaccurate identification of pre-fault risk factors in the traditional method for real-time monitoring of connector faults, improves the accuracy of performance release imbalance analysis and improves the accuracy of identification of pre-fault risk factors.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector;

[0012] Step S12: Cleaning the connector operation status data to obtain connector operation status cleansing data;

[0013] Step S13: extracting features from the connector operation status cleaning data to obtain connector operation status feature data;

[0014] Step S14: Timestamp the connector operation status characteristic data to obtain operation status characteristic timing mark data.

[0015] The present invention deploys distributed sensors to conduct comprehensive real-time monitoring of the connector, and collects multi-dimensional operating status data including current, voltage, temperature, vibration, etc. Since the connector involves multiple physical parameters during operation, the distributed sensors can monitor the performance of the device from different positions and angles at the same time, ensuring the comprehensiveness and high accuracy of data collection. The data collection at this stage provides the original data basis for subsequent fault detection and performance analysis, ensuring the accuracy and reliability of the data source, and is the prerequisite for subsequent data processing. The collected original data often contains noise, missing values ​​or inconsistencies. In order to ensure the data quality and the accuracy of subsequent analysis, these data need to be cleaned. The data cleaning process includes operations such as removing incomplete or erroneous data points, filling missing values, correcting outliers, and removing environmental noise interference. Through data cleaning, the reliability and consistency of the data can be significantly improved, ensuring that the subsequent feature extraction and fault identification processes can be based on high-quality data, thereby improving the diagnostic accuracy and efficiency of the entire system. Feature extraction is performed on the cleaned connector operating status data. The core of this step is to extract meaningful features from a large amount of time series data, such as statistical indicators such as mean, variance, peak, frequency, and advanced features indicating faults or performance changes. These features can compress the original data and extract the important information that best reflects the health status of the connector without losing key information. These feature data provide the basis for subsequent anomaly detection, fault prediction and performance evaluation, reduce the dimension of the data, and improve the training effect and prediction accuracy of the model. The connector operation status data after feature extraction is timestamped to accurately associate each feature data point with the corresponding time. The purpose of timestamp marking is to provide time series attributes for the data, thereby laying the foundation for subsequent timing analysis and model prediction. This step not only ensures the time correlation of the data, so that the change of each feature value can correspond to the time dimension, but also reveals the performance fluctuation trend of the connector at different time points. Through timestamp-marked data, early prediction of connector failures can be achieved, and detailed time dimension information can be provided for fault trend analysis.

[0016] Preferably, step S2 comprises the following steps:

[0017] Step S21: using a preset abnormal risk identification model to identify the abnormal state of the operating state feature time series mark data, and obtaining operating state abnormal feature identification data;

[0018] Step S22: extracting abnormal current / voltage fluctuations from the abnormal operating state feature recognition data to obtain abnormal current / voltage fluctuation data;

[0019] Step S23: performing performance release imbalance analysis on the connector operation status data according to the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data;

[0020] Step S24: performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data to obtain the fault triggering pre-risk factor.

[0021] The present invention analyzes the time series data of the operating state characteristics marked by the timestamp through a preset abnormal risk identification model to identify the abnormal state therein. The abnormal risk identification model is usually based on machine learning or statistical analysis methods, and can accurately find the difference between normal operation and abnormal behavior. The core of this step is to screen the real-time monitoring data through a model trained with historical data, automatically mark potential abnormal data points, and thus provide accurate abnormal signals for subsequent fault analysis. Through this step, the abnormal state of the connector can be detected in real time, which provides an important basis for taking measures in advance and avoiding further faults. For the abnormal state characteristic data identified in step S21, the abnormal fluctuation information of current and voltage is further extracted. The current and voltage fluctuations are direct indications of the abnormality of the connector and its electrical system, and can reflect whether the electrical performance of the equipment is in normal working condition. By extracting the current / voltage abnormal fluctuation data, the unstable factors in the electrical system can be identified and the problem areas can be located. This step provides key data support for the subsequent performance imbalance analysis, which can help to more accurately reveal the abnormal fluctuations of current or voltage and their potential impact on the system, and further optimize the fault warning and diagnosis process. By using the information extracted from the abnormal current / voltage fluctuation data, the overall performance of the connector is analyzed to identify the performance release imbalance caused by abnormal current or voltage fluctuations. This analysis mainly explores the impact of current / voltage fluctuations on the operating performance of the equipment by modeling the relationship between electrical characteristics and connector performance. The imbalance of current and voltage causes problems such as overheating, reduced efficiency, or long-term system instability in the equipment. Through this step, the specific impact of current / voltage fluctuations on connector performance can be identified, providing a basis for fault prediction and helping to take targeted preventive measures. By analyzing the impact data of abnormal current / voltage performance imbalance, the pre-risk factors that trigger faults are identified. These pre-risk factors are key indicators closely related to problems such as unstable electrical performance, excessive load, and excessive temperature. By identifying these factors, the occurrence and development trend of potential faults can be predicted. Through this analysis step, early signs of faults can be discovered in time and intervention can be carried out in advance to avoid the deterioration of equipment faults. The identification of risk factors provides valuable information for subsequent fault diagnosis and decision-making, improving the intelligence level of the system and the response speed of operation and maintenance.

[0022] Preferably, step S23 includes the following steps:

[0023] Step S231: performing current abnormal fluctuation spectrum analysis on the current / voltage abnormal fluctuation data to obtain a current abnormal fluctuation spectrum diagram;

[0024] Step S232: Calculate the frequency mutation point mean difference of the abnormal current fluctuation spectrum to obtain abnormal current frequency mutation point mean difference data;

[0025] Step S233: performing voltage over-limit abnormality identification on the current / voltage abnormal fluctuation data to obtain voltage over-limit abnormality data;

[0026] Step S234: Perform performance release imbalance analysis based on the abnormal current frequency mutation point average difference data and the voltage over-limit abnormal data to obtain abnormal current / voltage performance imbalance impact data.

[0027] The present invention generates a current abnormal fluctuation spectrum diagram by performing spectrum analysis on current / voltage abnormal fluctuation data. Current spectrum analysis can reveal the frequency characteristics of current fluctuations and identify abnormal activities within the frequency range. In this way, normal fluctuations and potential fault signals can be effectively distinguished. For example, abnormal fluctuations at certain frequencies indicate that certain components (such as connectors) inside the device have poor contact or other electrical faults. The current spectrum diagram provides detailed frequency information for subsequent analysis, helping engineers to deeply understand the source of the fluctuations and further guide the fault location and diagnosis of the equipment. By calculating the mean difference of the frequency mutation points of the current abnormal fluctuation spectrum diagram, the mean difference data of the abnormal current frequency mutation points is obtained. The mean difference of the frequency mutation points is an indicator for measuring sudden changes in the current spectrum and can effectively capture the abnormal bands of current fluctuations. This calculation method helps detect rapid fluctuations or sharp changes in the current by quantifying the mutations in frequency, which is usually a precursor to system failure or abnormality. For example, if a frequency mutation occurs in a certain frequency band of the current, it indicates that the electrical connection or load in the device is abnormal. This step helps to identify the details of the current abnormality and provides a basis for further fault analysis. By analyzing the current and voltage abnormal fluctuation data, the abnormal situation of voltage exceeding the limit is identified. Voltage over-limit anomaly usually refers to the situation where the voltage exceeds the design threshold during the operation of the equipment, which can cause damage to electrical components, performance degradation or system instability. This step automatically identifies the period or event of voltage abnormality by setting the voltage threshold and comparing it with the collected voltage data. Voltage over-limit anomaly identification helps to discover the potential risks of voltage fluctuations to the system as early as possible, helps engineers monitor the voltage status of the equipment in real time, and prevents equipment failures caused by overvoltage or undervoltage. According to the mean difference data of the abnormal current frequency mutation points obtained in step S232 and the voltage over-limit abnormal data identified in step S233, a performance release imbalance analysis is performed to finally obtain the abnormal current / voltage performance imbalance impact data. By jointly analyzing the frequency mutation of the current and the over-limit anomaly of the voltage, the impact of current and voltage fluctuations on the overall performance of the connector and the equipment can be comprehensively evaluated. Performance release imbalance analysis helps identify the current and voltage imbalance problems that cause faults in the electrical system and reveals the failure mode of the equipment under electrical instability. Through this step, the performance imbalance of the system can be effectively diagnosed, equipment failures can be predicted in advance, and reasonable preventive maintenance measures can be proposed.

[0028] Preferably, step S232 includes the following steps:

[0029] Perform frequency amplitude change analysis on the abnormal current fluctuation spectrum to obtain frequency amplitude change data;

[0030] The frequency negative sequence component is calculated for the abnormal current fluctuation spectrum according to the frequency amplitude change data to obtain the frequency negative sequence component data;

[0031] Based on the frequency negative sequence component data, the frequency amplitude change data is marked with frequency mutation point distribution to obtain frequency mutation point distribution data;

[0032] The phase difference of the mutation point is calculated for the frequency mutation point distribution data to obtain the phase difference data of the mutation point;

[0033] The frequency mutation point mean difference data is calculated based on the mutation point phase difference data and the frequency mutation point distribution data to obtain the abnormal current frequency mutation point mean difference data.

[0034] The present invention can capture the amplitude changes of current fluctuations in different frequency bands by analyzing the frequency amplitude changes of the current abnormal fluctuation spectrum. This analysis can reveal the intensity changes of current fluctuations, especially the aggravation or weakening of abnormal fluctuations, and help find fault signs. For example, abnormal changes in current amplitude indicate unstable conditions in the connector, load or electrical system, such as overload, short circuit or poor contact. The frequency amplitude change data provides basic data for the subsequent calculation of frequency negative sequence components, and can provide early clues for equipment fault diagnosis from the amplitude change level. The frequency amplitude change data is used to calculate the frequency negative sequence component of the current abnormal fluctuation spectrum. The frequency negative sequence component is a reverse component in the current signal, which is usually related to system imbalance or fault (such as asymmetric load, poor contact, etc.). Calculating the frequency negative sequence component can help distinguish between normal and abnormal current waveforms, thereby more accurately identifying the fault components in the current fluctuation. The appearance of the negative sequence component indicates that there is a potential imbalance in the electrical system of the equipment, which is usually related to equipment failure or abnormal operating status. This step provides a deeper current signal analysis result for further diagnosis and positioning of the fault. Based on the frequency negative sequence component data obtained in the previous step, the frequency amplitude change data is marked with the distribution of frequency mutation points. This analysis helps identify the moments of sharp changes in the system by marking the frequency points of mutations in the current waveform. These mutation points are caused by rapid changes or unstable behaviors caused by faults. Mutation points usually indicate sudden changes inside the equipment, such as abnormal events such as short circuits, disconnections, or poor electrical connections. By marking the distribution of mutation points, abnormal areas in current fluctuations can be located more accurately and provide important clues for subsequent fault location. By calculating the phase difference of mutation points for the distribution data of frequency mutation points, the phase changes in current fluctuations can be further analyzed. Phase difference refers to the amplitude of the change in the phase of the current signal between different time points. The increase in phase difference is usually related to the asymmetry of the system or electrical problems inside the equipment, such as poor contact or unbalanced load. By calculating the phase difference of the mutation point, the specific moment of abnormal current fluctuation and its impact range can be revealed, thereby providing key timing information for fault diagnosis and helping to determine the exact location and nature of the fault. Combined with the mutation point phase drop data and frequency mutation point distribution data, the frequency amplitude change data is used to calculate the frequency mutation point mean difference. This process further quantifies the degree of frequency mutation by comprehensively analyzing the phase change and frequency distribution of the mutation point, and obtains the frequency mutation point mean difference data. The mutation point mean difference reflects the regularity and stability of the current fluctuation mutation, and can effectively reveal the pattern and impact of current anomalies in electrical systems. Through this step, abnormal components in current fluctuations can be more accurately identified, especially in complex electrical environments, helping the system to accurately determine which frequency mutations have a higher risk of failure, providing support for preventive maintenance and fault prediction of equipment.

[0035] Preferably, step S233 includes the following steps:

[0036] Perform voltage amplitude outlier detection on current / voltage abnormal fluctuation data to obtain voltage amplitude outlier data;

[0037] Performing time series concentration analysis on voltage amplitude outlier data to obtain amplitude outlier time series concentration data;

[0038] Performing spatial interpolation processing on the voltage amplitude outlier data according to the amplitude outlier time series concentrated data to obtain the voltage outlier spatial interpolation data;

[0039] Voltage over-limit anomaly identification is performed based on the voltage outlier spatial interpolation data to obtain voltage over-limit anomaly data.

[0040] The present invention first performs outlier detection of voltage amplitude in current / voltage abnormal fluctuation data. This process can effectively locate potential abnormal sources by identifying and screening data points that deviate from the normal fluctuation range, thereby providing high-quality abnormal data for subsequent analysis. This operation ensures the accurate screening of abnormal data and the pertinence of subsequent processing, and reduces unnecessary data noise interference. The time series concentration analysis of voltage amplitude outlier data is intended to study the time distribution law of outlier occurrence. By analyzing its time concentration, it can be identified whether the abnormal fluctuation has periodic, regular or sudden characteristics, thereby providing important clues for further abnormal cause analysis. This step can help study the time series characteristics of abnormal voltage fluctuations and identify external factors or equipment failures. Based on the amplitude outlier time series concentration data, spatial interpolation processing is performed. This step uses a spatial interpolation method to compensate and optimize the voltage amplitude outlier data for position, thereby filling the gaps caused by abnormal data points. Spatial interpolation can improve the spatial consistency of data, making data smoother and more accurate between different geographical locations or device nodes, and avoiding misjudgment caused by missing or incomplete data. Based on the voltage outlier data after spatial interpolation, voltage over-limit anomaly identification is performed. This step quickly identifies abnormal situations where the voltage exceeds the set standard range by comparing the voltage data with the normal threshold range. In this way, voltage over-limit problems can be located on a more accurate basis, and alarms can be issued in time and corresponding measures can be taken to ensure the safe operation of the power system and prevent equipment from being damaged due to voltage instability.

[0041] Preferably, the present invention further provides a real-time monitoring system for connector faults, which is used to execute the real-time monitoring method for connector faults as described above. The real-time monitoring system for connector faults comprises:

[0042] The connector operation status acquisition module is used to collect the operation status data of the connector through distributed sensors to obtain the connector operation status data; timestamp the connector operation status data to obtain the operation status characteristic time series mark data;

[0043] The pre-factor identification module is used to identify the abnormal state of the operating state characteristic time series mark data using a preset abnormal risk identification model to obtain current / voltage abnormal fluctuation data; perform performance release imbalance analysis based on the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data; identify the fault trigger pre-risk factor of the abnormal current / voltage performance imbalance impact data to obtain the fault trigger pre-risk factor;

[0044] The fault lag severity level prediction module is used to perform time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk factor to obtain pre-risk current / voltage lag data; perform fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data;

[0045] The fault identification model is constructed to construct a connector fault identification model based on the fault lag severity prediction data and the fault triggering pre-risk factors based on the policy gradient algorithm, and the connector fault identification model is obtained. The connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

[0046] The beneficial effect of the present invention is that a distributed sensor system is used to monitor and collect data of the connector in real time, and the running status data of the connector is fully obtained, including multi-dimensional information such as current, voltage, and temperature, so as to ensure that the collected status data has high representativeness and integrity. By timestamping these data, each data point can be accurately associated with the corresponding time to form labeled data with time series characteristics. The time series labeled data is not only convenient for subsequent analysis, but also can reveal the changes in the running status of the connector at different time points, providing a solid foundation for abnormal detection and fault prediction. First, the preset abnormal risk identification model is used to analyze the running status characteristic time series labeled data obtained in the previous step, and the abnormal fluctuation of current or voltage is automatically identified. Through in-depth analysis of abnormal fluctuations, performance release imbalance analysis is further performed to identify the performance imbalance phenomenon caused by current / voltage instability. These analysis results can accurately reflect the potential performance degradation or failure risk of the connector. Then, based on the performance imbalance data of abnormal current / voltage, the fault triggering pre-risk factor identification is further carried out to discover potential fault risk factors in advance and predict the occurrence time of the fault. Through this series of analysis, early warning can be achieved, the probability of equipment failure can be reduced, and the stability and reliability of the system can be improved. By analyzing the time lag relationship between the fault triggering pre-risk factor and the current / voltage abnormal fluctuation data, the time delay characteristics of the current / voltage fluctuation before the fault occurs are identified. Through hysteresis analysis, the time difference between the abnormal fluctuation and the potential fault can be tracked, and the pre-risk current / voltage hysteresis data can be further obtained. These hysteresis data provide an accurate time window for subsequent fault prediction. Based on these hysteresis data, machine learning or statistical methods are used to predict the severity of the fault hysteresis, evaluate the severity of the fault and its impact on the performance of the connector. This prediction can provide an important basis for maintenance decisions, identify serious faults in advance, and prepare for emergency treatment. Using the policy gradient algorithm, a reinforcement learning method, a connector fault identification model is constructed based on the comprehensive analysis of the fault hysteresis severity level prediction data and the pre-risk factor. The policy gradient algorithm can dynamically adjust the identification strategy by self-optimizing the model to improve the accuracy and flexibility of fault detection. The model can not only identify the current fault risk, but also adjust the fault identification strategy in time according to historical data and prediction results, making fault monitoring more accurate. The built fault identification model will be uploaded to the cloud platform, using the powerful computing power of cloud computing for real-time monitoring and data analysis. The cloud platform can efficiently process large amounts of data from various sensors, detect connector faults in real time, and provide timely fault alerts and suggestions to operation and maintenance personnel, greatly improving fault diagnosis and response speed.Therefore, the present invention is an optimization processing of a traditional method for real-time monitoring of connector faults, which solves the problems of inaccurate performance release imbalance analysis and inaccurate identification of pre-fault risk factors in the traditional method for real-time monitoring of connector faults, improves the accuracy of performance release imbalance analysis and improves the accuracy of identification of pre-fault risk factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a process flow of a method for real-time monitoring of connector faults;

[0048] Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG.

[0049] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0050] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying 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 implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0052] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0053] To achieve this, please refer to Figure 1 to Figure 2 , a method for real-time monitoring of connector faults, the method comprising the following steps:

[0054] Step S1: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector; timestamping the operation status data of the connector to obtain the operation status characteristic time series marking data;

[0055] Step S2: using a preset abnormal risk identification model to identify abnormal states of the operating state feature time series marking data, and obtain current / voltage abnormal fluctuation data; performing performance release imbalance analysis based on the current / voltage abnormal fluctuation data, and obtaining abnormal current / voltage performance imbalance impact data; performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data, and obtaining the fault triggering pre-risk factor;

[0056] Step S3: performing time lag analysis on the current / voltage abnormal fluctuation data according to the pre-risk factor for fault triggering to obtain pre-risk current / voltage lag data; performing fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data;

[0057] Step S4: Based on the policy gradient algorithm, a connector fault identification model is constructed for the fault lag severity level prediction data and the fault triggering pre-risk factors to obtain the connector fault identification model, and the connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

[0058] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a flow chart of a method for real-time monitoring of connector faults according to the present invention. In this example, the method for real-time monitoring of connector faults includes the following steps:

[0059] Step S1: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector; timestamping the operation status data of the connector to obtain the operation status characteristic time series marking data;

[0060] In an embodiment of the present invention, the operation status data of the connector is collected through a distributed sensor system. Specifically, the distributed sensors will be installed at key parts of the connector, such as current, voltage and temperature detection points. Each sensor is responsible for collecting operation data in a specified area, such as current value, voltage value, connector surface temperature and the temperature of its surrounding environment. These sensors will collect data regularly or when a specific event occurs, and transmit the data to a local data aggregation device through a wireless communication module. During the data collection process, time synchronization technology is used to ensure that the data collected by each sensor has a consistent time reference and ensure the timing between the data. This is achieved through hardware clock synchronization and a timestamp-based correction mechanism. After data collection, timestamp marking is first performed, that is, a time stamp is added to each collected data. The time stamp accuracy must reach the millisecond level to ensure accurate tracking of changes in the operation status. The "operation status feature timing mark data" obtained through this step will provide a time-based, dynamically changing connector operation status data for subsequent steps.

[0061] Step S2: using a preset abnormal risk identification model to identify abnormal states of the operating state feature time series marking data, and obtain current / voltage abnormal fluctuation data; performing performance release imbalance analysis based on the current / voltage abnormal fluctuation data, and obtaining abnormal current / voltage performance imbalance impact data; performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data, and obtaining the fault triggering pre-risk factor;

[0062] In the embodiment of the present invention, the "operating state characteristic timing mark data" is firstly identified as abnormal by using a preset abnormal risk identification algorithm. The algorithm used at this time is based on a preset rule set and threshold value, and these rules can be empirical formulas or statistical models based on historical data. For current and voltage fluctuations, a normal fluctuation range is set, and any fluctuation beyond the range is identified as abnormal. Specifically, if the numerical changes of current and voltage exceed the set ±10% fluctuation range, they are regarded as abnormal data. The identification of abnormal state is carried out by comparing the difference between the data value collected at each moment and the range. After identifying the abnormal fluctuation, the current / voltage abnormal fluctuation extraction is performed to extract the time point and specific fluctuation amplitude of these abnormal values. Based on these extracted data, the next step is to perform performance release imbalance analysis. The specific method is to calculate the relative change between current and voltage to evaluate whether the performance release of the connector under abnormal fluctuations is balanced. If the current or voltage fluctuations are large and frequent, the energy release and absorption of the connector will be uneven, thereby affecting the normal operation of the equipment. According to the degree and duration of abnormal current / voltage fluctuations, the calculated "abnormal current / voltage performance imbalance impact data" will provide a basis for the subsequent identification of fault risk factors. The identification of fault-triggering pre-risk factors is completed by comparing the performance imbalance data with the historical fault records of the connector, and the pre-risk factors with a high correlation with the fault are extracted. These factors can be used to predict the occurrence of faults.

[0063] Step S3: performing time lag analysis on the current / voltage abnormal fluctuation data according to the pre-risk factor for fault triggering to obtain pre-risk current / voltage lag data; performing fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data;

[0064] In an embodiment of the present invention, the "fault-triggering pre-risk factor" is first normalized. Normalization is to eliminate the influence of different sensor data ranges and dimensions so that the data can be analyzed on a unified scale. The specific method of normalization is to use the minimum-maximum normalization algorithm to map all risk factor data to between 0 and 1. The normalized data is called the "fault-triggering pre-risk normalization factor". Next, the normalized data is used to perform a time lag analysis on the "current / voltage abnormal fluctuation data". Specifically, this step determines the lag of data changes by calculating the difference between the current / voltage data at the current moment and the data at the previous moment. The autocorrelation function (ACF) is used to quantify the relationship between time series data, and further determine the time delay relationship between current and voltage fluctuations and faults. This analysis helps identify the delay phenomenon caused by the fault and obtains the "pre-risk current / voltage lag data".

[0065] Step S4: Based on the policy gradient algorithm, a connector fault identification model is constructed for the fault lag severity level prediction data and the fault triggering pre-risk factors to obtain the connector fault identification model, and the connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

[0066] In an embodiment of the present invention, based on the "fault hysteresis severity level prediction data" and "fault triggering pre-risk factor" data obtained in step S3, a "connector fault identification model" is constructed by a policy gradient algorithm. The policy gradient algorithm is used here to optimize the fault identification process. The algorithm models the identification process of the system as a decision problem, in which each decision (i.e., fault identification) is based on the current input data (such as current / voltage abnormal data and hysteresis data) to generate an output result (fault type and severity). By adjusting the parameters in the decision-making process, the accuracy of the identification result is maximized. During the training process, the model will optimize its performance based on the continuously updated input data (such as newly collected current and voltage abnormal fluctuation data) by continuously adjusting the identification strategy. Through multiple rounds of feedback mechanism, the identification process is continuously optimized to improve the accuracy of fault prediction and identification. The "connector fault identification model" after construction will be sent to the cloud platform, which will monitor the operating status of the connector in real time. When the operating status data of the connector is uploaded through the cloud platform, the system will immediately perform fault monitoring through the identification model, so as to detect potential faults in time and trigger the early warning mechanism to respond.

[0067] Preferably, step S1 comprises the following steps:

[0068] Step S11: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector;

[0069] Step S12: Cleaning the connector operation status data to obtain connector operation status cleansing data;

[0070] Step S13: extracting features from the connector operation status cleaning data to obtain connector operation status feature data;

[0071] Step S14: Timestamp the connector operation status characteristic data to obtain operation status characteristic timing mark data.

[0072] In an embodiment of the present invention, the operation status data of the connector is collected by distributed sensors. The distributed sensor network covers all working parts of the connector, including real-time collection of key parameters such as current, voltage, temperature, and humidity. The layout of each sensor should be reasonably arranged according to the working characteristics of the connector to ensure comprehensive monitoring of the operation status. The current and voltage sensors are used to detect the current and voltage data of the connector input and output respectively; the temperature and humidity sensors are used to monitor the ambient temperature and humidity to evaluate the impact of the connector working environment on its performance; and the pressure sensor is used to detect the pressure fluctuation inside the connector. The data collection frequency of the sensor needs to be adjusted according to the frequency of change of the connector working state, usually collecting data hundreds of times per second to ensure that the fluctuations that occur in the short term can be captured in real time. All collected data are transmitted to the central data processing node through the wireless communication module for aggregation. These data do not contain timestamp information, so the next step will be further processed in the data cleaning and feature extraction process. Data cleaning is performed on the connector operation status data. The purpose of data cleaning is to remove noise data caused by sensor failure, environmental interference or communication transmission errors. First, the data collected by each sensor is detected for abnormality by simple statistical methods (such as mean, standard deviation, etc.). If the fluctuation range of the data exceeds the standard deviation of the normal working range by 3 times, it is regarded as abnormal data and removed. Secondly, interpolation processing is performed on the missing values ​​in the data. If the collected data at a certain moment is lost, the missing value is filled by the linear interpolation algorithm based on the linear relationship between the data values ​​at adjacent moments and the working status of the equipment. In the specific implementation, if the collected data of a certain sensor is missing for a long time, the time series interpolation algorithm is used to predict and fill it according to the collected data before and after it. During the data cleaning process, it is also necessary to check the duplicates in the data. If it is found that the data from the same sensor at the same time has repeated records, deduplication processing is performed to retain valid data. After the above cleaning method, the "connector operation status cleaning data" obtained will be more accurate and provide more reliable basic data for subsequent feature extraction. Feature extraction is performed on the "connector operation status cleaning data". The purpose of feature extraction is to extract key information for fault monitoring and risk warning from the original data. First, for time series data such as current and voltage, basic statistical characteristics such as mean, maximum value, minimum value, and standard deviation of the data are extracted through time domain analysis. These features can reflect the fluctuation range and deviation of the equipment during normal operation. Next, the current and voltage data are analyzed in the frequency domain. The fast Fourier transform (FFT) algorithm is used to convert the time domain signal into a frequency domain signal and extract the frequency components of the data. The high-frequency signals in the frequency components usually represent signs of equipment failure, such as overload, short circuit, etc. The features obtained based on the spectrum analysis, such as the main frequency, bandwidth, spectrum energy, etc., will further assist in determining whether the connector is in an abnormal working state.For environmental factors such as temperature and humidity, the sliding window method is used to smooth the data. By analyzing the change trend of the data, the long-term change trend and short-term fluctuation of the data are extracted. The fluctuation of these environmental data often has an important impact on the performance of the connector, so its change characteristics also need to be included in the fault monitoring indicator system. The "connector operation status feature data" is timestamped to obtain the "operation status feature timing tag data". The goal of this process is to associate the feature data with the actual time point to ensure that each data has a clear time dimension for subsequent timing analysis and real-time monitoring. First, the data of all sensors are obtained and synchronized with the system clock. The synchronization method includes unified time tagging of different sensors through the synchronization signal of the hardware clock to ensure that the data collected by the sensor accurately matches the actual time. Each data point will be added with a high-precision timestamp, and the accuracy of the timestamp needs to reach the millisecond level. Then, the data after feature extraction is arranged in chronological order, and a time tag is assigned to each data point. The time tag includes specific information such as year, month, day, hour, minute, second and millisecond of the acquisition time. These marks will provide basic data for subsequent timing analysis, so that the operating status of the connector can be traced back and analyzed according to time. Through the data marked with timestamps, the system can realize real-time monitoring of the connector, analyze its operating status in different time periods, and compare it with historical data to identify potential abnormalities or failure modes. This step provides time-related data support for subsequent fault identification and risk prediction of the system.

[0073] Preferably, step S2 comprises the following steps:

[0074] Step S21: using a preset abnormal risk identification model to identify the abnormal state of the operating state feature time series mark data, and obtaining operating state abnormal feature identification data;

[0075] Step S22: extracting abnormal current / voltage fluctuations from the abnormal operating state feature recognition data to obtain abnormal current / voltage fluctuation data;

[0076] Step S23: performing performance release imbalance analysis on the connector operation status data according to the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data;

[0077] Step S24: performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data to obtain the fault triggering pre-risk factor.

[0078] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0079] Step S21: using a preset abnormal risk identification model to identify the abnormal state of the operating state feature time series mark data, and obtaining operating state abnormal feature identification data;

[0080] In the embodiment of the present invention, the "operation state characteristic time series marking data" is identified as abnormal by using a preset abnormal risk identification model. The purpose of this step is to identify the abnormal state of the connector during operation, and then provide basic data for subsequent fault warning. First, a statistical method is used to perform a preliminary analysis on the operation state characteristic data, and the fluctuation range under normal operation is determined by calculating the mean, standard deviation, maximum value, minimum value and other indicators. Then, by setting an abnormal threshold, when the collected data of a certain sensor deviates from the normal fluctuation range, the data is marked as abnormal. For current and voltage data, the abnormal fluctuation of current or voltage can be identified based on the threshold method by comparing the fluctuation range of each period in the historical data. In addition, a rule-based abnormal detection method can also be introduced. For example, the upper and lower limits of current and voltage are set. If the current or voltage value at a certain moment exceeds these set values, the state at that moment is considered to be abnormal. In the implementation process, the collected data is usually analyzed step by step using a sliding window technology, and the data in the window is compared with the normal fluctuation range to detect whether there is a sudden abnormal fluctuation. Finally, the label data of each sensor is combined to output "operating status abnormal feature recognition data", which includes not only the abnormal label at each time point, but also the specific type of abnormality of each sensor, such as overcurrent, overvoltage, etc.

[0081] Step S22: extracting abnormal current / voltage fluctuations from the abnormal operating state feature recognition data to obtain abnormal current / voltage fluctuation data;

[0082] In the embodiment of the present invention, the "operating state abnormal feature identification data" is used to extract the current / voltage abnormal fluctuation data. Current and voltage fluctuations are important indication signals of connector failure, so it is necessary to extract the abnormal fluctuations of these signals in detail. First, by screening the time period marked as abnormal in the "operating state abnormal feature identification data", the abnormal data of current and voltage are further analyzed. For the abnormal fluctuation characteristics of current and voltage, a waveform analysis method, such as a waveform comparison method, can be used. By comparing the characteristic parameters of the current and voltage waveforms under normal working conditions (such as the amplitude, period, phase, etc. of the waveform), the specific characteristics of the abnormal fluctuations are identified. Secondly, signal processing methods, such as wavelet transform, can be used to perform time-frequency domain analysis on the current and voltage signals to extract the high-frequency components of the signals. These high-frequency components usually represent abnormal fluctuations of the equipment in the short term, such as the moment when the current or voltage changes dramatically, reflecting the fault signs such as connector load changes and poor contact. After extracting the abnormal fluctuation characteristics of current and voltage, they are saved as "current / voltage abnormal fluctuation data" and provide key data support for subsequent performance analysis.

[0083] Step S23: performing performance release imbalance analysis on the connector operation status data according to the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data;

[0084] In the embodiment of the present invention, the connector operation status data is analyzed for performance release imbalance based on the "current / voltage abnormal fluctuation data". Performance release imbalance refers to the instability of the internal working state of the connector due to abnormal fluctuations in current and voltage during operation. The core of this process is to evaluate whether there is an imbalance in the performance of the connector by comparing the current / voltage characteristics during normal operation with the characteristics during abnormal fluctuations.

[0085] Step S24: performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data to obtain the fault triggering pre-risk factor.

[0086] In the embodiment of the present invention, based on the "abnormal current / voltage performance imbalance impact data", the fault triggering pre-risk factor identification is performed. The fault triggering pre-risk factor refers to some potential risk signals that exist before the connector fails, and these signals are usually reflected by the fluctuation of parameters such as current and voltage. Therefore, identifying these pre-risk factors is crucial for early warning of connector failures. First, using the "abnormal current / voltage performance imbalance impact data" and combining with the known connector failure modes, the risk factors that cause the failure are defined. For example, the sudden change and frequent fluctuation of current and voltage can be defined as pre-risk factors. Then, using multivariate analysis techniques, such as principal component analysis (PCA), the most representative risk factors are extracted from data of multiple dimensions such as current, voltage, and temperature. These factors are usually closely related to the specific fault type of the connector, such as some fluctuation modes are related to short circuit faults, and some frequency abnormalities are related to overheating faults. Further, time series analysis methods, such as autoregressive integrated moving average model (ARIMA), etc., can be used to predict the pre-risk factors and analyze the changing trends of these risk factors before the failure occurs. Through these methods, a set of “fault triggering pre-risk factors” with strong early warning capabilities can be extracted from historical data.

[0087] Preferably, step S23 includes the following steps:

[0088] Step S231: performing current abnormal fluctuation spectrum analysis on the current / voltage abnormal fluctuation data to obtain a current abnormal fluctuation spectrum diagram;

[0089] Step S232: Calculate the frequency mutation point mean difference of the abnormal current fluctuation spectrum to obtain abnormal current frequency mutation point mean difference data;

[0090] Step S233: performing voltage over-limit abnormality identification on the current / voltage abnormal fluctuation data to obtain voltage over-limit abnormality data;

[0091] Step S234: Perform performance release imbalance analysis based on the abnormal current frequency mutation point average difference data and the voltage over-limit abnormal data to obtain abnormal current / voltage performance imbalance impact data.

[0092] In the embodiment of the present invention, in step S231, the spectrum analysis of abnormal current fluctuation is performed to obtain a spectrum diagram of abnormal current fluctuation. This process aims to identify potential abnormal fluctuation patterns by analyzing the characteristics of the current signal in the frequency domain. First, the collected abnormal current fluctuation data is converted into frequency domain data. This step uses the fast Fourier transform (FFT) algorithm to perform frequency decomposition on the current signal and convert the time domain signal into a frequency domain signal. Through FFT, the frequency components of the current signal will be decomposed into components of different frequencies, each frequency component corresponds to a certain amplitude, and the change in amplitude reflects the fluctuation intensity of the current signal. Next, the frequency components of the current fluctuation are analyzed, with special attention paid to those frequency regions where the fluctuation amplitude is significantly increased. At this time, the spectrum diagram will be displayed as a graph representing the relationship between frequency and amplitude, where under normal conditions, the frequency interval with a larger amplitude represents the normal operating frequency, and the prominent part with a frequency distribution deviating from the normal range corresponds to abnormal fluctuations. The output of the abnormal current fluctuation spectrum diagram is a data set containing the amplitude information of each frequency component, which can be used for subsequent analysis. In step S232, the frequency mutation point mean difference calculation is performed based on the "abnormal current fluctuation spectrum diagram" to obtain the abnormal current frequency mutation point mean difference data. The core of the frequency mutation point mean difference analysis is to capture the amplitude of the frequency characteristic mutation and help identify the areas of drastic changes in current fluctuations, which often indicate abnormal currents. First, select all the frequency points with significant mutations in the current abnormal fluctuation spectrum and calculate their mean difference (i.e., the amplitude difference between adjacent frequency points). This process is achieved by directly performing differential calculations on the spectrum, and the difference results can be used to evaluate the severity of the amplitude change in the spectrum. Next, the mean difference of the mutation points is statistically analyzed, its mean and standard deviation are calculated, and compared with the normal fluctuation range in the historical data. If the mean difference value exceeds the set threshold, it means that the current fluctuation in this area has a large change, indicating that there is some abnormality inside the system. Through this process, the "abnormal current frequency mutation point mean difference data" is obtained, which can reveal the moment when the frequency characteristics in the current signal change drastically, and provide an important basis for further performance analysis. In step S233, the "current / voltage abnormal fluctuation data" is subjected to voltage over-limit abnormal identification to obtain "voltage over-limit abnormal data". Voltage over-limit abnormality means that the voltage value exceeds the safety threshold set when the connector is working, which is usually a precursor to a fault. Therefore, the goal of this step is to promptly identify the risk of voltage over-limit. First, the voltage data is threshold judged based on the preset voltage operating range. The set upper and lower voltage limits are based on the design specifications of the connector and the actual operating conditions. By comparing the voltage data with these set thresholds, any voltage value that exceeds the upper or lower limit will be marked as "over-limit abnormality". The identification of voltage over-limit abnormality not only relies on simple threshold judgment, but also requires in-depth analysis of the dynamic changes of voltage fluctuations.For example, by calculating the rate of voltage change, it can be determined whether the voltage has risen or fallen sharply, and then it can be inferred whether a voltage mutation event has occurred. The voltage over-limit abnormal data can include specific over-limit time points, voltage values, deviation degrees and other information, providing detailed voltage abnormal records for fault analysis. In step S234, according to the "abnormal current frequency mutation point average difference data" and "voltage over-limit abnormal data", a performance release imbalance analysis is performed to obtain "abnormal current / voltage performance imbalance impact data". Performance release imbalance refers to the imbalance of the internal circuit and system working state of the connector due to abnormal fluctuations in current and voltage. First, the frequency mutation of current and voltage over-limit abnormality are used as the main input data, and these abnormal features are comprehensively analyzed using the weighted average method. In this process, each abnormal point will be weighted according to its impact on the performance of the connector. For example, the frequency mutation point of current has a higher risk weighting coefficient than the voltage over-limit. Then, by constructing a performance imbalance analysis model based on current and voltage abnormal data, the operating state of the connector is evaluated. Dynamic simulation methods can be used to simulate the impact of abnormal current / voltage fluctuations on the thermal effect, electromagnetic interference and other factors inside the connector based on the current, voltage and electrical characteristics of the connector. By calculating the comprehensive impact of current and voltage fluctuations on connector performance, the "abnormal current / voltage performance imbalance impact data" is ultimately output. This data not only reveals the severity of the performance imbalance, but also provides data support for subsequent fault diagnosis and fault warning, helping to predict the future fault trend of the connector.

[0093] Preferably, step S232 includes the following steps:

[0094] Perform frequency amplitude change analysis on the abnormal current fluctuation spectrum to obtain frequency amplitude change data;

[0095] The frequency negative sequence component is calculated for the abnormal current fluctuation spectrum according to the frequency amplitude change data to obtain the frequency negative sequence component data;

[0096] Based on the frequency negative sequence component data, the frequency amplitude change data is marked with frequency mutation point distribution to obtain frequency mutation point distribution data;

[0097] The phase difference of the mutation point is calculated for the frequency mutation point distribution data to obtain the phase difference data of the mutation point;

[0098] The frequency mutation point mean difference data is calculated based on the mutation point phase difference data and the frequency mutation point distribution data to obtain the abnormal current frequency mutation point mean difference data.

[0099] In the embodiment of the present invention, at the beginning of step S232, the frequency amplitude change analysis is first performed on the abnormal current fluctuation spectrum. The abnormal current fluctuation spectrum is obtained by fast Fourier transform (FFT) in step S231, and the figure reflects the amplitude distribution of the current signal at different frequency components. The purpose of the frequency amplitude change analysis is to capture the changing trend of the frequency components of the current signal over time, especially those frequency components that show significant fluctuations. The specific operation is to select the frequency amplitude data in the abnormal current fluctuation spectrum and perform time domain segmentation analysis on it. During the analysis, the sliding window technology is used to set a fixed time window, and the amplitude change rate, that is, the difference in frequency amplitude in adjacent time periods, is calculated by analyzing the change of frequency amplitude in each time period. Through this analysis method, detailed data of frequency amplitude change in each time period can be obtained, thereby obtaining "frequency amplitude change data". This data provides basic information for subsequent abnormality detection, especially in identifying frequency fluctuation anomalies. After obtaining the frequency amplitude change data, the next step is to calculate the frequency negative sequence component of the abnormal current fluctuation spectrum. The frequency negative sequence component is the frequency component relative to the fundamental frequency in the spectrum. These negative sequence components are usually abnormal signals generated when the system is unbalanced. The size and change of the negative sequence component can reflect the abnormal degree of the current signal. The calculation process first identifies the components of the negative sequence frequency in the spectrum based on the frequency amplitude change data. The negative sequence component is usually located in the symmetrical position of the spectrum and forms a pair with the positive sequence component. Therefore, the frequency negative sequence component can be extracted by performing a symmetrical analysis of the frequency components of the spectrum graph. At this time, the amplitude of the negative sequence component is calculated and compared with the positive sequence component to obtain the amplitude data of the negative sequence component. This process can locate the negative sequence frequency and extract the corresponding amplitude information by calculating the spectrum data after Fourier transformation, and finally generate "frequency negative sequence component data". These data can reveal whether there are abnormal frequency components caused by imbalance in the current. After obtaining the frequency negative sequence component data, the next step is to perform the frequency mutation point distribution marking operation. The frequency mutation point refers to the moment when the frequency component of the current signal changes dramatically within a certain period of time. Frequency mutation usually occurs when the load of the power system changes, equipment fails or the system is unstable. The specific operation is to identify the moments when the amplitude changes dramatically in the current signal by analyzing the relationship between the "frequency negative sequence component data" and the "frequency amplitude change data", and mark these moments as frequency mutation points. At this time, a change threshold needs to be set. If the frequency amplitude change is greater than the threshold and the amplitude of the negative sequence component increases significantly, it is considered that a frequency mutation has occurred. This process compares the mutation point positions of the frequency amplitude change data and the negative sequence component data, marks the position and distribution of the frequency mutation point, and finally generates the "frequency mutation point distribution data". After obtaining the frequency mutation point distribution data, the next step is to calculate the phase difference of the frequency mutation point.The phase difference of the frequency mutation point refers to the difference in the signal phase before and after the mutation point. In the spectrum of the current signal, the phase difference can reflect the imbalance degree and abnormal fluctuation of the signal, especially when there is a fault or system instability, the phase difference of the mutation point will increase. By performing phase calculation on the frequency mutation point distribution data, the phase difference before and after each mutation point is compared. In the specific operation, the Fourier transform is used to obtain the phase information of the frequency corresponding to each mutation point, and the phase change between adjacent mutation points is calculated. If the phase change exceeds the predetermined threshold, it is considered that there is a large phase difference at the mutation point. Through this calculation process, the "mutation point phase difference data" is obtained, which provides detailed information on the phase change between frequency mutation points and can further reveal the severity of current fluctuations. After completing the calculation of the mutation point phase difference, the last step is to calculate the frequency mutation point mean difference of the frequency amplitude change data based on the mutation point phase difference data and the frequency mutation point distribution data. The core of this step is to further confirm the abnormal fluctuation in the current signal by calculating the mean difference of the amplitude change between the frequency mutation points. First, the frequency amplitude change of the mutation point is calculated, and the mean difference of the amplitude change between each pair of mutation points is calculated, that is, the average difference of the frequency amplitude between adjacent mutation points is calculated. If the difference is greater than a preset threshold, it is considered that the frequency mutation in this area is relatively abnormal, indicating that there is a risk of failure in the connector. Through this method, the "abnormal current frequency mutation point mean difference data" is obtained. These data can help further identify the abnormal degree of current fluctuations and provide a basis for subsequent fault detection and early warning.

[0100] Preferably, step S233 includes the following steps:

[0101] Perform voltage amplitude outlier detection on current / voltage abnormal fluctuation data to obtain voltage amplitude outlier data;

[0102] Performing time series concentration analysis on voltage amplitude outlier data to obtain amplitude outlier time series concentration data;

[0103] Performing spatial interpolation processing on the voltage amplitude outlier data according to the amplitude outlier time series concentrated data to obtain the voltage outlier spatial interpolation data;

[0104] Voltage over-limit anomaly identification is performed based on the voltage outlier spatial interpolation data to obtain voltage over-limit anomaly data.

[0105] In the embodiment of the present invention, at the initial stage of step S233, the voltage amplitude outlier detection is first performed on the current / voltage abnormal fluctuation data. The voltage amplitude outlier refers to the amplitude abnormality that suddenly appears in the voltage signal and is obviously inconsistent with the normal working state. This process is intended to identify the voltage value that is obviously deviated from the normal range from the current / voltage fluctuation data. The specific operation is to first set a normal fluctuation range of the voltage amplitude, which can be obtained based on historical operating data or standard operating parameters. Then, the voltage fluctuation data is analyzed point by point, and the standard deviation of each data point and its surrounding data is calculated. When the difference between a data point and the surrounding data points exceeds the set threshold, it is marked as an outlier. This detection method can be implemented based on statistical principles, such as Z-score or anomaly detection algorithm (for example, anomaly detection based on box plots). The specific steps are to calculate the Z-score value of each voltage data point. If the Z-score exceeds the set threshold, the data point is considered to be an outlier. Through this method, outliers in the voltage amplitude can be accurately identified to obtain "voltage amplitude outlier data". After obtaining the voltage amplitude outlier data, the next step is to conduct a time series concentration analysis. This step aims to study the distribution characteristics of voltage amplitude outliers in the time series, especially to determine whether these outliers show a certain concentration pattern, or whether they appear frequently in certain time periods, indicating that the system has a certain periodic or sudden failure risk. The specific operation is to first map the outlier data to the time axis and analyze its distribution. By calculating the frequency of occurrence of outliers in different time periods, use time series analysis methods such as autocorrelation analysis to determine the time series concentration of outliers. If outliers appear frequently in certain time periods, it indicates that there are instability or potential failures in these time periods. By performing cluster analysis on the time intervals of outliers, it is possible to further determine which time periods have a higher concentration of outliers. This analysis can be performed by calculating the time concentration of outliers or using the sliding window method to analyze their time distribution density, and finally obtain "amplitude outlier time series concentration data". Next, the "amplitude outlier time series concentration data" is spatially interpolated. Spatial interpolation refers to the use of existing outlier data to infer the voltage values ​​of missing or unobserved spatial areas. The purpose of this step is to extrapolate the spatial distribution pattern of outliers to other uncovered areas based on the concentration of outliers in the time series in order to detect the spatial distribution characteristics of voltage anomalies. The specific operation is to first determine the spatial coordinates of voltage outliers, combine the results of time series concentration analysis, and use interpolation methods such as nearest neighbor interpolation (Nearest Neighbor Interpolation) or Kriging Interpolation (Kriging Interpolation) and other technologies to extend the outlier information to other areas. Through interpolation, it is possible to generate estimated values ​​at the location of missing voltage data and obtain the spatial distribution of voltage fluctuations.During the interpolation process, the results of the concentration analysis in the time series are considered to infer which areas have voltage anomaly risks. This method helps to determine whether the abnormal patterns of voltage amplitudes are spatially clustered and whether these anomalies occur frequently in certain specific areas, thereby obtaining "voltage outlier spatial interpolation data". After obtaining the voltage outlier spatial interpolation data, the last step is to identify voltage over-limit anomalies. Voltage over-limit anomaly refers to the voltage signal exceeding the set safe working range, which usually indicates that the system has a fault or unstable operation. The goal of this step is to identify abnormal areas where the voltage exceeds the standard from the spatial interpolation results. The specific operation is to first set the safe working range or threshold of the voltage, which is usually based on the rated voltage range of the system or industry standards. When some voltage values ​​in the spatial interpolation data exceed the safe threshold, these data points are marked as voltage over-limit anomalies. In order to improve the recognition accuracy, a variety of threshold determination methods can be used, such as dynamic threshold adjustment, calibration combined with historical data, or threshold determination in a local range through a sliding window method. The final output of voltage over-limit anomaly identification is a set of all voltage over-limit data, forming "voltage over-limit anomaly data".

[0106] Preferably, step S3 comprises the following steps:

[0107] Step S31: normalizing the fault triggering pre-risk factor to obtain the fault triggering pre-risk normalization factor;

[0108] Step S32: performing time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk normalization factor to obtain pre-risk current / voltage lag data;

[0109] Step S33: performing risk-fault linkage evaluation on the pre-risk current / voltage hysteresis data to obtain risk-fault linkage data;

[0110] Step S34: predicting the fault hysteresis severity level according to the risk fault linkage data and the preceding risk current / voltage hysteresis data to obtain the fault hysteresis severity level prediction data.

[0111] In the embodiment of the present invention, at the initial stage of step S233, the voltage amplitude outlier detection is first performed on the current / voltage abnormal fluctuation data. The voltage amplitude outlier refers to the amplitude abnormality that suddenly appears in the voltage signal and is obviously inconsistent with the normal working state. This process is intended to identify the voltage value that is obviously deviated from the normal range from the current / voltage fluctuation data. The specific operation is to first set a normal fluctuation range of the voltage amplitude, which can be obtained based on historical operating data or standard operating parameters. Then, the voltage fluctuation data is analyzed point by point, and the standard deviation of each data point and its surrounding data is calculated. When the difference between a data point and the surrounding data points exceeds the set threshold, it is marked as an outlier. This detection method can be implemented based on statistical principles, such as Z-score or anomaly detection algorithm (for example, anomaly detection based on box plots). The specific steps are to calculate the Z-score value of each voltage data point. If the Z-score exceeds the set threshold, the data point is considered to be an outlier. Through this method, outliers in the voltage amplitude can be accurately identified to obtain "voltage amplitude outlier data". After obtaining the voltage amplitude outlier data, the next step is to conduct a time series concentration analysis. This step aims to study the distribution characteristics of voltage amplitude outliers in the time series, especially to determine whether these outliers show a certain concentration pattern, or whether they appear frequently in certain time periods, indicating that the system has a certain periodic or sudden failure risk. The specific operation is to first map the outlier data to the time axis and analyze its distribution. By calculating the frequency of occurrence of outliers in different time periods, use time series analysis methods such as autocorrelation analysis to determine the time series concentration of outliers. If outliers appear frequently in certain time periods, it indicates that there are instability or potential failures in these time periods. By performing cluster analysis on the time intervals of outliers, it is possible to further determine which time periods have a higher concentration of outliers. This analysis can be performed by calculating the time concentration of outliers or using the sliding window method to analyze their time distribution density, and finally obtain "amplitude outlier time series concentration data". Next, the "amplitude outlier time series concentration data" is spatially interpolated. Spatial interpolation refers to the use of existing outlier data to infer the voltage values ​​of missing or unobserved spatial areas. The purpose of this step is to extrapolate the spatial distribution pattern of outliers to other uncovered areas based on the concentration of outliers in the time series in order to detect the spatial distribution characteristics of voltage anomalies. The specific operation is to first determine the spatial coordinates of voltage outliers, combine the results of time series concentration analysis, and use interpolation methods such as nearest neighbor interpolation (Nearest Neighbor Interpolation) or Kriging Interpolation (Kriging Interpolation) and other technologies to extend the outlier information to other areas. Through interpolation, it is possible to generate estimated values ​​at the location of missing voltage data and obtain the spatial distribution of voltage fluctuations.During the interpolation process, the results of the concentration analysis in the time series are considered to infer which areas have voltage anomaly risks. This method helps to determine whether the abnormal patterns of voltage amplitudes are spatially clustered and whether these anomalies occur frequently in certain specific areas, thereby obtaining "voltage outlier spatial interpolation data". After obtaining the voltage outlier spatial interpolation data, the last step is to identify voltage over-limit anomalies. Voltage over-limit anomaly refers to the voltage signal exceeding the set safe working range, which usually indicates that the system has a fault or unstable operation. The goal of this step is to identify abnormal areas where the voltage exceeds the standard from the spatial interpolation results. The specific operation is to first set the safe working range or threshold of the voltage, which is usually based on the rated voltage range of the system or industry standards. When some voltage values ​​in the spatial interpolation data exceed the safe threshold, these data points are marked as voltage over-limit anomalies. In order to improve the recognition accuracy, a variety of threshold determination methods can be used, such as dynamic threshold adjustment, calibration combined with historical data, or threshold determination in a local range through a sliding window method. The final output of voltage over-limit anomaly identification is a set of all voltage over-limit data, forming "voltage over-limit anomaly data".

[0112] Preferably, step S32 includes the following steps:

[0113] Step S321: Perform multi-dimensional factor sensitivity analysis on the normalized factor of the pre-risk of fault triggering to obtain multi-dimensional sensitive data of the pre-risk factor;

[0114] Step S322: performing critical state identification on the multi-dimensional sensitive data of the pre-risk factors to obtain critical state data of the risk factors;

[0115] Step S323: calculating partial autocorrelation functions of abnormal current / voltage fluctuation data according to the risk factor critical state data, and obtaining abnormal partial autocorrelation functions of current / voltage under different risk factor states;

[0116] Step S324: performing time lag analysis on the abnormal current / voltage fluctuation data based on the abnormal current / voltage partial autocorrelation function under different risk factor states to obtain the pre-risk current / voltage lag data.

[0117] In the embodiment of the present invention, a multi-dimensional factor sensitivity analysis is first performed on the "fault triggering pre-risk normalization factor". The purpose of this analysis is to identify the key factors that affect the connector fault triggering and quantify the contribution of each factor to the overall risk, so as to provide a more accurate reference for subsequent fault warning. The implementation method of the multi-dimensional factor sensitivity analysis adopts a common technique in sensitivity analysis - a combination of variance analysis (ANOVA) and principal component analysis (PCA). Variance analysis is used to evaluate the impact of different factor changes on the fault triggering risk factor; principal component analysis is used to reduce the redundancy of high-dimensional data, project multiple related variables onto a few unrelated principal components, and then reveal the dimension that has the greatest impact on the fault risk. In actual operation, first collect a variety of factor data that affect the fault, including physical quantities such as voltage, current, temperature, and external load, and normalize these data. Then, variance analysis is applied to evaluate the variance contribution of each factor, and by comparing the contributions of different factors, the factors with higher sensitivity are screened out. Subsequently, the multi-dimensional data is reduced in dimension by principal component analysis, and the principal component that best represents the pre-risk factor is extracted. Finally, the results of variance analysis and principal component analysis are combined to obtain the "pre-risk factor multi-dimensional sensitive data". The critical state is identified based on the "pre-risk factor multi-dimensional sensitive data". The purpose of this process is to determine which specific risk factor value combinations will lead to the critical point of fault triggering, especially to find out the extreme abnormal state caused by the current, voltage and other indicators exceeding the normal range. The technical method of critical state identification adopts the statistical threshold method combined with cluster analysis (K-means). First, through the statistical analysis of the "pre-risk factor multi-dimensional sensitive data", the abnormal threshold range of each risk factor is defined. Specifically, the historical data of each risk factor (such as current, voltage, etc.) is statistically analyzed, and the upper and lower limits based on the standard deviation are set as the abnormal threshold by calculating its mean and standard deviation. For example, for the voltage factor, if its value exceeds the mean ± 2 times the standard deviation range, it is considered to be in an abnormal state. Then, the K-means clustering method is applied to cluster the risk factor data, identify the clustered abnormal points in the data, and then define the extreme points in these clusters as critical states. When a certain risk factor combination falls into these critical states, it is marked as "risk factor critical state data". In step S323, the "risk factor critical state data" is used to calculate the partial autocorrelation function (PACF) of the "current / voltage abnormal fluctuation data". The partial autocorrelation function is used to describe the correlation of current / voltage data after removing the influence of other time points, especially the size of its lag effect. Therefore, calculating the partial autocorrelation function can help identify the dependence of current / voltage fluctuations in different time periods, and further analyze the relationship between current / voltage data and risk factors. The process of calculating the partial autocorrelation function adopts the Yule-Walker equation.By this method, the autocorrelation coefficient of the abnormal current / voltage fluctuation data under the condition of lag time is calculated to remove the influence of other lag times on the current time point. First, the current / voltage fluctuation data is modeled in time series, and the autocorrelation coefficient of each lag time point is calculated using the Yule-Walker equation. At this time, according to the "risk factor critical state data", the abnormal current / voltage fluctuation data is divided into different states, and the partial autocorrelation function in each state is calculated respectively. By comparing the partial autocorrelation functions in different states, it can be analyzed whether the time lag effect of the current / voltage fluctuation data changes significantly under different risk factor states. Finally, the "abnormal partial autocorrelation function of current / voltage under different risk factor states" is obtained. In step S324, the abnormal current / voltage fluctuation data is further analyzed for time lag based on the "abnormal partial autocorrelation function of current / voltage under different risk factor states". At this time, the time lag analysis not only focuses on the lag effect of the time series data itself, but also considers the influence of the lag effect under different risk factor states on fault warning. This analysis method evaluates the impact of different lag times by taking a weighted average of the results of the partial autocorrelation function and combining the lag characteristics under each risk factor state. First, based on the calculation results of the partial autocorrelation function, the degree of response of the abnormal current / voltage fluctuation data to the fault under different lag times is identified. Next, the lag effects under different risk factor states are weighted, considering the weight differences of the lag effects under different states. At this point, the weighted time lag correlation coefficient is used to calculate the lag effect under each state to obtain more accurate lag effect data. Finally, based on the weighted lag coefficient, the "pre-risk current / voltage lag data" is obtained.

[0118] Preferably, step S33 includes the following steps:

[0119] Step S331: performing risk timeliness segmentation processing on the pre-risk current / voltage hysteresis data to obtain risk timeliness segmentation data;

[0120] Step S332: performing fault linkage deduction on the pre-risk current / voltage hysteresis data according to the risk time-effect segmented data to obtain time-effect segmented fault linkage deduction data;

[0121] Step S333: Perform risk fault linkage evaluation on the time-sensitive segmented fault linkage deduction data to obtain risk fault linkage data.

[0122] In the embodiment of the present invention, the "pre-risk current / voltage hysteresis data" is first segmented according to the risk timeliness. The goal of this process is to segment the current / voltage hysteresis data according to its change trend in the time dimension, identify the risk characteristics of the current / voltage fluctuation data in different time periods, and provide a basis for subsequent fault deduction and linkage evaluation. The segmentation process adopts a strategy combining the sliding window method and the threshold judgment method. First, by setting a fixed time window, the current / voltage hysteresis data is subjected to sliding analysis, and the window size is selected according to the working cycle of the connector and the average duration of the fault. Each time the window slides, the risk value of the data in the window is checked. According to the amplitude and duration of the current / voltage fluctuation, by setting a predetermined threshold, it is determined whether there is a potential risk in the period. Specifically, if the current or voltage value exceeds its normal range and the duration exceeds the set minimum time threshold, it is determined as a risk period. The data in each period will be marked as "high risk" or "low risk". Through this process, the entire time series is divided into multiple risk timeliness segments, and finally "risk timeliness segmented data" is generated. In step S332, the fault linkage deduction is performed on the "pre-risk current / voltage hysteresis data" based on the "risk time segmentation data". The purpose of this deduction is to simulate whether the current / voltage fluctuation will cause a fault in different risk periods, and to analyze the correlation between these faults. This process can help determine which risk periods have a significant impact on the triggering of connector faults. The implementation method of fault linkage deduction uses causal analysis and Markov chain model. First, combined with the high-risk periods marked in the "risk time segmentation data", causal deduction is performed based on the current / voltage fluctuation characteristics in these periods. For each high-risk period, the current / voltage hysteresis data before and after it are analyzed to identify the causal relationship between these data. For example, if the voltage fluctuation in a certain period exceeds the threshold and lasts for a long time, it will cause a failure of an internal component of the connector. By mapping these causal relationships into time series data, the linkage effect between different risk periods is deduced. Next, the Markov chain model is used to assume that the transfer of the system state (such as the working state of the connector) between different time points conforms to the characteristics of the Markov process. According to the risk time segmentation data, the state transition probability is calculated using historical data to deduce the potential impact of current / voltage fluctuations on the occurrence of faults at different time points. Finally, the "time-sensitive segmented fault linkage deduction data" is obtained. In step S333, the risk fault linkage is evaluated based on the "time-sensitive segmented fault linkage deduction data". The goal of this process is to determine the strength of the linkage between current / voltage fluctuations and faults in different time periods, thereby providing more accurate evaluation indicators for fault warning. The risk fault linkage evaluation adopts a combination of correlation analysis and grey correlation analysis.First, by calculating the correlation between the current / voltage data and the fault state in the "time-sensitive segmented fault linkage deduction data", the correlation value in each time period is obtained. The specific operation is to use the Pearson correlation coefficient or the Spearman rank correlation coefficient to calculate the data of each time-sensitive segment based on the relationship between the lagging current / voltage data and the known fault events to evaluate the strength of its linkage. The higher the correlation value, the closer the relationship between the current / voltage fluctuation in the period and the occurrence of the fault, and vice versa. The impact is small. Then, the gray correlation analysis is used to further quantify the linkage between different risk periods and faults. Gray correlation analysis can analyze the relative relationship between various factors and determine the relative impact of current / voltage fluctuations on faults in each period when the system information is incomplete. By calculating the gray correlation degree, the contribution of each time-sensitive segment to the occurrence of the fault is evaluated to obtain the "risk fault linkage data".

[0123] Preferably, the present invention further provides a real-time monitoring system for connector faults, which is used to execute the real-time monitoring method for connector faults as described above. The real-time monitoring system for connector faults comprises:

[0124] The connector operation status acquisition module is used to collect the operation status data of the connector through distributed sensors to obtain the connector operation status data; timestamp the connector operation status data to obtain the operation status characteristic time series mark data;

[0125] The pre-factor identification module is used to identify the abnormal state of the operating state characteristic time series mark data using a preset abnormal risk identification model to obtain current / voltage abnormal fluctuation data; perform performance release imbalance analysis based on the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data; identify the fault trigger pre-risk factor of the abnormal current / voltage performance imbalance impact data to obtain the fault trigger pre-risk factor;

[0126] The fault lag severity level prediction module is used to perform time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk factor to obtain pre-risk current / voltage lag data; perform fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data;

[0127] The fault identification model is constructed to construct a connector fault identification model based on the fault lag severity prediction data and the fault triggering pre-risk factors based on the policy gradient algorithm, and the connector fault identification model is obtained. The connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

[0128] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0129] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may 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 the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for real-time monitoring of connector faults, characterized in that: The following steps are involved: Step S1: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector; Timestamping the connector operation status data to obtain operation status characteristic time series marking data; Step S2: using a preset abnormal risk identification model to identify abnormal conditions of the operating state characteristic time series mark data, and obtaining current / voltage abnormal fluctuation data; Perform performance release imbalance analysis based on abnormal current / voltage fluctuation data to obtain abnormal current / voltage performance imbalance impact data; Identify the fault triggering pre-risk factor for the abnormal current / voltage performance imbalance impact data to obtain the fault triggering pre-risk factor; Step S3: Performing time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk factor to obtain pre-risk current / voltage lag data; Predicting the severity level of fault hysteresis according to the pre-risk current / voltage hysteresis data to obtain the prediction data of the severity level of fault hysteresis; Step S4: Based on the policy gradient algorithm, a connector fault identification model is constructed for the fault lag severity level prediction data and the fault triggering pre-risk factors to obtain the connector fault identification model, and the connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

2. The method for real-time monitoring of connector faults according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting the operation status data of the connector through the distributed sensor to obtain the operation status data of the connector; Step S12: Cleaning the connector operation status data to obtain connector operation status cleansing data; Step S13: extracting features from the connector operation status cleaning data to obtain connector operation status feature data; Step S14: Timestamp the connector operation status characteristic data to obtain operation status characteristic timing mark data.

3. The method for real-time monitoring of connector faults according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: using a preset abnormal risk identification model to identify the abnormal state of the operating state feature time series mark data, and obtaining operating state abnormal feature identification data; Step S22: extracting abnormal current / voltage fluctuations from the abnormal operating state feature recognition data to obtain abnormal current / voltage fluctuation data; Step S23: performing performance release imbalance analysis on the connector operation status data according to the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data; Step S24: performing fault triggering pre-risk factor identification on the abnormal current / voltage performance imbalance impact data to obtain the fault triggering pre-risk factor.

4. The method for real-time monitoring of connector faults according to claim 3, characterized in that: Step S23 includes the following steps: Step S231: performing current abnormal fluctuation spectrum analysis on the current / voltage abnormal fluctuation data to obtain a current abnormal fluctuation spectrum diagram; Step S232: Calculate the frequency mutation point mean difference of the abnormal current fluctuation spectrum to obtain abnormal current frequency mutation point mean difference data; Step S233: performing voltage over-limit abnormality identification on the current / voltage abnormal fluctuation data to obtain voltage over-limit abnormality data; Step S234: Perform performance release imbalance analysis based on the abnormal current frequency mutation point average difference data and the voltage over-limit abnormal data to obtain abnormal current / voltage performance imbalance impact data.

5. The method for real-time monitoring of connector faults according to claim 4, characterized in that: Step S232 includes the following steps: Perform frequency amplitude change analysis on the abnormal current fluctuation spectrum to obtain frequency amplitude change data; The frequency negative sequence component is calculated for the abnormal current fluctuation spectrum according to the frequency amplitude change data to obtain the frequency negative sequence component data; Based on the frequency negative sequence component data, the frequency amplitude change data is marked with frequency mutation point distribution to obtain frequency mutation point distribution data; The phase difference of the mutation point is calculated for the frequency mutation point distribution data to obtain the phase difference data of the mutation point; The frequency mutation point mean difference data is calculated based on the mutation point phase difference data and the frequency mutation point distribution data to obtain the abnormal current frequency mutation point mean difference data.

6. The method for real-time monitoring of connector faults according to claim 4, characterized in that: Step S233 includes the following steps: Perform voltage amplitude outlier detection on current / voltage abnormal fluctuation data to obtain voltage amplitude outlier data; Performing time series concentration analysis on voltage amplitude outlier data to obtain amplitude outlier time series concentration data; Performing spatial interpolation processing on the voltage amplitude outlier data according to the amplitude outlier time series concentrated data to obtain the voltage outlier spatial interpolation data; Voltage over-limit anomaly identification is performed based on the spatial interpolation data of voltage outliers to obtain voltage over-limit anomaly data.

7. The method for real-time monitoring of connector faults according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: normalizing the fault triggering pre-risk factor to obtain the fault triggering pre-risk normalization factor; Step S32: performing time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk normalization factor to obtain pre-risk current / voltage lag data; Step S33: performing risk-fault linkage evaluation on the pre-risk current / voltage hysteresis data to obtain risk-fault linkage data; Step S34: predicting the fault hysteresis severity level according to the risk fault linkage data and the preceding risk current / voltage hysteresis data to obtain the fault hysteresis severity level prediction data.

8. The method for real-time monitoring of connector faults according to claim 7, characterized in that: Step S32 includes the following steps: Step S321: Perform multi-dimensional factor sensitivity analysis on the normalized factor of the pre-risk of fault triggering to obtain multi-dimensional sensitive data of the pre-risk factor; Step S322: performing critical state identification on the multi-dimensional sensitive data of the pre-risk factors to obtain critical state data of the risk factors; Step S323: calculating partial autocorrelation functions of abnormal current / voltage fluctuation data according to the risk factor critical state data, and obtaining abnormal partial autocorrelation functions of current / voltage under different risk factor states; Step S324: performing time lag analysis on the abnormal current / voltage fluctuation data based on the abnormal current / voltage partial autocorrelation function under different risk factor states to obtain the pre-risk current / voltage lag data.

9. The method for real-time monitoring of connector faults according to claim 7, characterized in that: Step S33 includes the following steps: Step S331: performing risk timeliness segmentation processing on the pre-risk current / voltage hysteresis data to obtain risk timeliness segmentation data; Step S332: performing fault linkage deduction on the pre-risk current / voltage hysteresis data according to the risk time-effect segmented data to obtain time-effect segmented fault linkage deduction data; Step S333: Perform risk fault linkage evaluation on the time-sensitive segmented fault linkage deduction data to obtain risk fault linkage data.

10. A real-time monitoring system for connector faults, characterized in that: Used to execute the method for real-time monitoring of connector faults according to claim 1, the real-time monitoring system for connector faults comprises: The connector operation status acquisition module is used to collect the operation status data of the connector through distributed sensors to obtain the connector operation status data; timestamp the connector operation status data to obtain the operation status characteristic time series mark data; The pre-factor identification module is used to identify the abnormal state of the operating state characteristic time series mark data using a preset abnormal risk identification model to obtain current / voltage abnormal fluctuation data; perform performance release imbalance analysis based on the current / voltage abnormal fluctuation data to obtain abnormal current / voltage performance imbalance impact data; identify the fault trigger pre-risk factor of the abnormal current / voltage performance imbalance impact data to obtain the fault trigger pre-risk factor; The fault lag severity level prediction module is used to perform time lag analysis on the current / voltage abnormal fluctuation data according to the fault triggering pre-risk factor to obtain pre-risk current / voltage lag data; perform fault lag severity level prediction according to the pre-risk current / voltage lag data to obtain fault lag severity level prediction data; The fault identification model is constructed to construct a connector fault identification model based on the fault lag severity prediction data and the fault triggering pre-risk factors based on the policy gradient algorithm, and the connector fault identification model is obtained. The connector fault identification model is sent to the data cloud platform to perform real-time monitoring of connector faults.

Images