Online monitoring system adaptive to recycling scene of SiC power device

Abstract

The invention relates to the field of rail transit, and discloses an online monitoring system adapted to a SiC power device recycling scene, and the system comprises the steps: collecting key operation parameters of a SiC power device in real time, and when the collected parameters exceed a preset warning threshold value, judging whether the subsequent failure trend prediction needs to be carried out or not; constructing a device performance change curve based on the collected operation parameters, analyzing the curve type and the fluctuation trend, judging whether the device has a potential failure risk or not, and judging whether key monitoring or load regulation and control need to be performed on a specific module or not by combining the parallel operation state of the device and a historical abnormal fluctuation event; the influence degree of the potential failure trend is determined by analyzing the relevance among on-resistance sudden change, junction temperature fluctuation, welding layer cavity expansion and bonding interface aging, and according to the influence degree of the potential failure trend and a warning threshold value, an online early warning signal is output and a regulation and control instruction is executed. The method has the advantage of finding the potential failure trend in advance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail transit, in particular to an online monitoring system suitable for SiC power device recycling scenarios. BACKGROUND

[0002] SiC power devices are widely used in high-frequency power conversion systems in new energy vehicle driving, photovoltaic inversion, rail transit and other fields due to their high withstand voltage, low conduction loss and excellent high-temperature working characteristics. In some application scenarios, to reduce maintenance costs and spare parts occupation, the operation party will recycle the retired or repaired SiC power modules. However, the existing recycling mode mainly relies on factory detection and regular offline sampling inspection, and lacks continuous online sensing capability for performance degradation, packaging thermal fatigue and junction temperature fluctuation trend of the device during re-service. Especially in the power converter with multiple modules in parallel, the conduction resistance of individual recycled SiC modules may suddenly change in a short time due to chip bonding interface aging or weld layer cavity expansion, causing uneven current and even triggering protection action, resulting in intermittent shutdown of the system. The existing monitoring scheme cannot identify such transient degradation signs in advance, and can only replace the device after the fault occurs, increasing the risk of unplanned downtime. Therefore, it is necessary to design an online monitoring system suitable for SiC power device recycling scenarios to detect potential failure trends in advance. SUMMARY

[0003] In view of the deficiencies in the prior art, the present application provides an online monitoring system suitable for SiC power device recycling scenarios, which has the advantage of detecting potential failure trends in advance and solves the problems in the above background art.

[0004] To achieve the above purpose of detecting potential failure trends in advance, the present application provides the following technical solution: an online monitoring system suitable for SiC power device recycling scenarios, comprising: a state acquisition module: real-time acquisition of key operating parameters of SiC power devices, including junction temperature, conduction resistance and operating current fluctuation; when the acquired parameters exceed the preset warning threshold, it is determined whether subsequent failure trend prediction is needed, and if so, it enters the prediction analysis module; a prediction analysis module: based on the acquired operating parameters, a device performance change curve is constructed, the curve type and fluctuation trend are analyzed, and the future changes of junction temperature and conduction resistance are predicted to determine whether the device has potential failure risks. If the prediction result shows that there is a potential failure trend, it enters the regulation requirement analysis module; a regulation analysis module: combined with the parallel operation state of the device and the historical abnormal fluctuation events, it is determined whether the specific module needs to be monitored or the load needs to be regulated, and if so, it enters the risk correlation analysis module; Risk analysis module: determine the influence degree of potential failure trend by analyzing the relevance between the sudden change of on-resistance, the fluctuation of junction temperature and the expansion of solder voids, the aging of bonding interface, and send the analysis results to the early warning and control module; Early warning and control module: output online early warning signal and execute control instruction according to the influence degree of potential failure trend and warning threshold.

[0005] Preferably, the process of judging whether to perform subsequent failure trend prediction is: Calculate the deviation of real-time collected junction temperature, on-resistance and working current fluctuation from the preset warning threshold, and generate a comprehensive risk score combined with historical operation data and device aging curve; If the comprehensive risk score is higher than the set prediction trigger threshold, it is determined that subsequent failure trend prediction is needed; If the comprehensive risk score is lower than the set threshold, continue to maintain the normal monitoring mode.

[0006] Preferably, the process of constructing device performance change curve based on the collected operation parameters is: Extract the key points of junction temperature, on-resistance and current fluctuation in the continuously collected time series data; Generate multi-parameter performance change curve by using smoothing filtering, interpolation and trend fitting method; Perform time segmentation analysis on each performance curve to identify fluctuation peak, trend slope and abnormal deviation interval.

[0007] Preferably, the process of analyzing curve type and fluctuation trend is: Classify and process the performance change curve to identify linear rise, exponential growth, periodic oscillation and mutation, etc. Calculate the fluctuation amplitude, frequency and change rate by statistical analysis and machine learning method to judge the stability of device operation state; If the curve type and fluctuation trend show abnormal mode, mark it as potential failure interval and enter the potential failure risk prediction link.

[0008] Preferably, the process of judging whether the device has potential failure risk is: Calculate the potential failure index based on the fluctuation amplitude and trend rate of the performance change curve, combined with the device manufacturing characteristics and historical life data; If the potential failure index exceeds the preset threshold, it is determined that the device has potential failure risk, and the control demand analysis is triggered; If the potential failure index does not exceed the threshold, maintain the normal monitoring state and continue to collect data.

[0009] Preferably, the process of combining the parallel operation state of the device and the historical abnormal fluctuation event is: Obtain real-time operating parameters of each SiC device in the same power module, and analyze the load distribution and temperature difference between parallel devices; In combination with the historical abnormal event database, it is evaluated whether the current device fluctuation is consistent with the past failure mode; A module-level risk aggregation score is generated for determining whether special devices or modules need to be monitored.

[0010] Preferably, the process of determining whether special modules need to be monitored or load regulation is: According to the module-level risk aggregation score and real-time load condition, it is determined whether a single device or module has an overload or thermal stress concentration risk; If the risk score exceeds the set threshold, the module is marked as a monitoring object, and a load regulation suggestion is generated; If the risk score is lower than the threshold, the normal monitoring and load distribution strategy is maintained.

[0011] Preferably, the process of analyzing the correlation between the on-resistance mutation, junction temperature fluctuation, and the expansion of the solder layer cavity and the aging of the bonding interface is: A device multi-physical quantity correlation model is established, and the on-resistance, junction temperature, solder layer cavity size and bonding interface aging degree are included in the unified analysis framework; Statistical correlation analysis and multi-factor regression model are used to evaluate the causal relationship and sensitivity between parameters; According to the analysis result, a quantitative risk index of potential failure trend is generated, and input into the early warning and regulation module.

[0012] Preferably, the process of outputting online early warning signals and executing regulation instructions is: Based on the quantitative index of potential failure trend generated by the risk correlation analysis module, in combination with the real-time operating parameters and historical operating data of each SiC device and module, a comprehensive risk score is calculated; The comprehensive risk score is compared with the preset multi-level warning threshold to determine the risk level of the device or module, and three-level online early warning signals of high risk, medium risk and low risk are generated; The system synchronously records the early warning signal, regulation instruction and execution result to the online monitoring database to form a closed loop feedback.

[0013] Compared with the prior art, the present application provides an online monitoring system suitable for the recycling use scene of SiC power devices, which has the following beneficial effects: The application can realize all-around, dynamic and intelligent monitoring of the running state of SiC power devices according to the state acquisition module, the prediction analysis module, the regulation demand analysis module, the risk correlation analysis module and the early warning regulation module. The real-time parameter acquisition of the state acquisition module can accurately capture the key indicators such as the junction temperature, the on-resistance and the working current fluctuation of the device, and ensure the first-time perception of abnormal conditions; the prediction analysis module uses historical and real-time data to construct the performance change curve of the device, deeply analyzes the curve type and fluctuation trend, realizes the scientific prediction of the future change of the junction temperature and the on-resistance, and thus identifies the potential failure risk in advance; the regulation demand analysis module can implement key monitoring and load regulation on the high-risk module on the basis of comprehensively considering the parallel running state of the device and historical abnormal events, optimizes the running conditions of the device, and reduces the local stress concentration and failure probability; the risk correlation analysis module realizes the quantitative evaluation of the influence degree of the potential failure trend through the correlation analysis between the on-resistance mutation, the junction temperature fluctuation and the expansion of the solder layer cavity, the aging of the bonding interface and other factors, and provides data support for scientific decision-making; the early warning regulation module generates multi-level online early warning signals based on the comprehensive analysis results, and automatically executes the corresponding regulation instructions, realizes the active protection and load optimization of the device. Overall, the online monitoring system realizes the closed-loop processing of data acquisition, trend prediction, regulation analysis, correlation evaluation and intelligent early warning, not only improves the reliability and safety of SiC power devices in the recycling process, but also discovers the potential failure trend in advance, and provides an efficient, intelligent and operable solution for the online monitoring system adapted to the recycling scenario of SiC power devices. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The figure is a structural schematic diagram of the application. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0016] Embodiment 1: refer to Figure 1 The online monitoring system adapted to the recycling scenario of SiC power devices described in the embodiments of the application comprises: A state acquisition module: real-time acquisition of key running parameters of SiC power devices, including junction temperature, on-resistance and working current fluctuation; when the acquired parameters exceed the preset warning threshold, it is determined whether subsequent failure trend prediction is needed, and if so, the prediction analysis module is entered; The state acquisition module determines whether the subsequent failure trend prediction process needs to be performed: The deviation of the real-time collected junction temperature, on-state resistance and working current fluctuation from the preset warning threshold is calculated, and the historical operation data and the device aging curve are combined to generate a comprehensive risk score; Real-time acquisition of key operating parameters such as junction temperature, on-state resistance and working current fluctuation of SiC power devices, filtering and denoising of collected data, and corresponding matching of processed values with system preset warning thresholds to form standardized input data sequences required for deviation calculation, calculation of deviation from corresponding warning thresholds for junction temperature, on-state resistance and working current fluctuation, quantification using standardized difference or percentage deviation, and obtaining of deviation quantification values of each indicator under the current operating state, combination of real-time deviation data with historical operation data and device aging characteristic curve, comparison of time series and trend fitting method to form the relative position of the current device state in the historical evolution curve, and obtaining of aging sensitivity and risk contribution of each indicator, and finally weighting and integrating of each indicator deviation and aging sensitivity, using a risk score model to calculate a comprehensive risk score to generate a single quantitative value for determining whether to trigger the subsequent failure trend prediction.

[0017] If the comprehensive risk score is higher than the set prediction trigger threshold, it is determined that the subsequent failure trend prediction needs to be performed; The generated comprehensive risk score is compared with the system set prediction trigger threshold: if the comprehensive risk score is higher than the trigger threshold, it is determined that the device needs to enter the subsequent failure trend prediction module to start the prediction analysis process; if the comprehensive risk score is lower than the trigger threshold, the monitoring mode under the normal operating state is continued, and the deviation calculation and risk score are periodically updated.

[0018] If the comprehensive risk score is lower than the set threshold, the normal monitoring mode is continued; After each calculation is completed, the real-time deviation, comprehensive risk score and determination result are recorded to the monitoring system database, and are transmitted to the subsequent prediction analysis module or normal monitoring module through the interface to realize data continuity and state tracking.

[0019] Prediction analysis module: based on the collected operating parameters, a device performance change curve is constructed, the curve type and fluctuation trend are analyzed, and the future change of junction temperature and on-state resistance is predicted to determine whether the device has a potential failure risk; if the prediction result shows that there is a potential failure trend, it enters the regulation demand analysis module; The process of constructing a device performance change curve in the prediction analysis module based on the collected operating parameters is: Key points of junction temperature, on-state resistance and current fluctuation are extracted from the continuously collected time series data; The continuously collected SiC power device operating parameter time series data, including junction temperature, on-state resistance and operating current fluctuation, are recorded point by point according to fixed sampling interval, key points are extracted by numerical division and feature detection algorithm, including local extreme value, inflection point and mutation point, to provide original data nodes for subsequent curve generation.

[0020] Smooth filtering, interpolation and trend fitting methods are used to generate multi-parameter performance change curves. The extracted key point data is subjected to smooth filtering processing, moving average, Savitzky-Golay filtering or exponential smoothing method is used to reduce instantaneous noise interference, interpolation algorithm is applied to complete the data points for missing sampling or abnormal interval data, to ensure the continuity of the curve and the integrity of the time series, the junction temperature, on-state resistance and operating current fluctuation parameters are subjected to trend fitting, linear regression, piecewise linear fitting or nonlinear curve fitting method is used to generate multi-parameter performance change curves, each curve retains time series information and numerical amplitude, forming a complete performance curve data set that can be used for subsequent analysis.

[0021] Each performance curve is subjected to time segmentation analysis to identify fluctuation peak value, trend slope and abnormal deviation interval. Each generated performance curve is subjected to fixed time window or adaptive segmentation processing to identify fluctuation peak value, slope and change rate of each segment. Local trend information is extracted using statistical methods or sliding window analysis to provide a basis for abnormal deviation interval identification. Based on the segmentation analysis results, the deviation amount of the curve change amplitude from the preset threshold in each time interval is calculated, and the interval exceeding the threshold is marked as an abnormal deviation interval. The abnormal interval, fluctuation characteristics and trend parameters of each curve are recorded in a structured manner to provide input data for subsequent failure trend prediction.

[0022] The analysis curve type and fluctuation trend process in the prediction analysis module is as follows: The performance change curves are classified and processed to identify linear rise, exponential growth, periodic oscillation and mutation. The generated performance change curves are input into the curve feature extraction module according to time sequence, the slope change characteristics and acceleration characteristics of the curve in different time intervals are extracted by using first derivative, second derivative and curvature calculation, the curve is classified according to the matching degree of the feature parameters and the preset mode template, and the curve types including linear rise, exponential growth, periodic oscillation and instantaneous mutation are identified. For curves with mixed modes, the curve type labels of each time interval are labeled by segment matching method.

[0023] The fluctuation amplitude, frequency and change rate are calculated by statistical analysis and machine learning method to judge the stability of the device operating state. On the basis of the classified curves, statistical analysis methods are used to calculate the fluctuation amplitude, fluctuation frequency and change rate of the curves. For multi-parameter curves, the fluctuation characteristics of the junction temperature, on-resistance and working current fluctuation are calculated respectively, and are stored uniformly through a multi-dimensional feature vector. If necessary, the frequency domain analysis method is called to extract the implicit periodic characteristics to supplement the time domain statistical results.

[0024] If the curve type and fluctuation trend show an abnormal pattern, it is marked as a potential failure interval and enters the potential failure risk prediction link; The curve type and fluctuation feature vector of each time interval are input into the pattern matching and difference detection algorithm, the deviation degree from the normal operation mode template is calculated, and for the time interval whose deviation degree exceeds the preset threshold, the interval marking operation is performed, the interval is recorded as the input data of potential failure analysis, and the interval index and corresponding feature vector are transmitted to the potential failure risk prediction module to support the subsequent prediction calculation process.

[0025] The process of determining whether the device has a potential failure risk in the prediction analysis module is: Based on the fluctuation amplitude and trend rate of the performance change curve, combined with the device manufacturing characteristics and historical life data, the potential failure index is calculated; The fluctuation amplitude, trend rate and other characteristic parameters extracted from the performance change curve are matched with the manufacturing characteristic parameters and historical life data of the device. A weighted feature fusion method is used to combine the time domain features and frequency domain features to form a unified feature vector input into the failure risk calculation model. The model outputs a quantitative potential failure index through a regression algorithm or a multi-dimensional mapping function, which is used to represent the closeness between the current operating state and the known failure mode.

[0026] If the potential failure index exceeds the preset threshold, it is determined that the device has a potential failure risk, and the regulation demand analysis is triggered; The calculated potential failure index is compared with the preset risk judgment threshold one by one, and the comparison process is executed under the control of numerical accuracy to ensure the stability and consistency of the threshold judgment. When the potential failure index is higher than the threshold in the continuous sampling period, a risk trigger flag is generated, and the flag is stored with the current collected data as the input condition of the subsequent regulation demand analysis module.

[0027] If the potential failure index does not exceed the threshold, the normal monitoring state is maintained and the data collection continues; When the potential failure index is lower than the preset threshold in the continuous sampling period, the risk flag is set to a non-trigger state, and the running feature data of this time period is recorded in the system to update the normal operation mode reference library. The system continues to perform real-time data collection and feature updating operations to ensure that the subsequent analysis can be based on the latest operating state.

[0028] The regulation analysis module: combine the parallel operation state of the device and the historical abnormal fluctuation event, judge whether it is necessary to monitor or load regulation for a specific module, if necessary, enter the risk correlation analysis module; The process of combining the parallel operation state of the device and the historical abnormal fluctuation event in the regulation analysis module is: Obtain the real-time operation parameters of each SiC device in the same power module, and analyze the load distribution and temperature difference between the parallel devices; Synchronize the real-time operation parameters from each SiC power device in the same power module, including junction temperature, on-resistance, working current and switching loss data. After aligning the collected data according to the time stamp, the load distribution calculation between parallel devices is performed, and by comparing the device output current proportion, power loss proportion and junction temperature difference, the electrical and thermal distribution characteristics under parallel operation state are determined. The distribution characteristics are stored in a multi-dimensional matrix form for subsequent risk aggregation analysis.

[0029] Combine the historical abnormal event database to assess whether the current device fluctuation is consistent with the past failure mode; Call the historical abnormal event database to retrieve records similar to the current device operation characteristics. The matching process is based on a multi-parameter similarity calculation method, which calculates the similarity score between the key feature vector of the current collected data and the feature vector stored in the database, and labels the corresponding known failure mode type, occurrence frequency and associated operating environment conditions in the matching result.

[0030] Generate a module-level risk aggregation score to determine whether a specific device or module needs to be monitored; Fuse the parallel device operation difference analysis results and the matching results of historical abnormal events to generate a module-level risk aggregation score. The fusion method is based on a weighted scoring model, which gives different weights to real-time operation state characteristics and historical event similarity characteristics to obtain a single quantitative result. The score result is stored in the risk assessment data structure of the system and used as an input parameter for subsequent key monitoring decision logic.

[0031] The process of determining whether a specific module needs to be monitored or load regulated in the regulation analysis module is: Determine whether a single device or module has an overload or thermal stress concentration risk based on the module-level risk aggregation score and real-time load conditions; The risk aggregation score data of the current module is called from the risk assessment subsystem, and the real-time load parameters of each parallel SiC device in the module are extracted from the real-time monitoring database, including instantaneous current value, average power and junction temperature distribution, the risk aggregation score is associated and mapped to the same analysis table with the real-time load parameters by using the data matching program, so as to perform unified comparison processing in the same calculation period, calculate the deviation rate of the load proportion of each device from its rated bearing capacity, and generate a thermal distribution matrix according to the temperature sensor data, locate the temperature peak area and current density concentration area through matrix analysis algorithm, form the identification marker matrix of overload and thermal stress concentration risk, and cross verify with the risk aggregation score.

[0032] If the risk score exceeds the set threshold, mark the module as a key monitoring object and generate a load regulation suggestion; The risk score trigger threshold is read from the system parameter library, and the numerical comparison is performed with the current module risk score, when the risk score is greater than or equal to the threshold, the device ID and related risk parameters of the module are written into the key monitoring object list and updated to the monitoring task scheduling table, at the same time, the load optimization algorithm is called, the load regulation suggestion table is generated based on the existing running state, including current redistribution scheme, power limitation parameter and temperature equalization strategy, and the suggestion table is sent to the control execution unit for scheduling execution.

[0033] If the risk score is lower than the threshold, the normal monitoring and load distribution strategy is maintained; When the risk score is less than the threshold, the module continues to be registered in the normal monitoring task queue, the original acquisition period, sampling accuracy and data recording rules are maintained, the current load distribution table and power management strategy parameters are called, and the control execution unit is directly issued as a maintenance instruction to ensure that each parallel device operates according to the predetermined distribution ratio, and the risk score calculation and threshold comparison operation are continuously repeated in the subsequent monitoring period to ensure that the state change can be captured in time.

[0034] Risk analysis module: by analyzing the correlation between the sudden change of on-resistance, junction temperature fluctuation, solder layer cavity expansion and bonding interface aging, the influence degree of potential failure trend is determined, and the analysis result is sent to the early warning control module; The correlation analysis process between the sudden change of on-resistance, junction temperature fluctuation, solder layer cavity expansion and bonding interface aging in the risk analysis module is: A device multi-physical quantity correlation model is established, and the on-resistance, junction temperature, solder layer cavity size and bonding interface aging degree are included in the unified analysis framework; In the multi-physical quantity analysis framework, real-time monitoring data of on-state resistance, junction temperature, solder layer cavity size and bonding interface aging degree are collected, time alignment, outlier rejection, data interpolation and smoothing processing are performed to ensure that each parameter is comparable on a unified time sequence, and a correlation model is constructed with a unified data structure to provide basic data support for subsequent causal analysis.

[0035] Statistical correlation analysis and multi-factor regression model are used to evaluate the causal relationship and sensitivity between parameters. Statistical correlation analysis is used to identify the synchronous change characteristics between parameters, and multi-factor regression model is used to quantify the sensitivity and potential causal relationship of on-state resistance change on junction temperature fluctuation, solder layer cavity expansion and bonding interface aging, and to extract the change amplitude, change rate and mutual influence weight of each parameter to provide quantitative basis for risk index generation.

[0036] According to the analysis results, the quantitative risk index of potential failure trend is generated and input into the early warning and control module. The multi-physical quantity correlation analysis results are integrated according to the predefined weight to form the quantitative risk index of potential failure trend, and are input into the early warning and control module in real time to provide continuous data support for subsequent online early warning and control strategy. According to the comparison between the quantitative risk index and the set warning threshold, online early warning signals of high, medium and low levels are automatically generated, and corresponding control instructions are triggered, including power adjustment, load limitation or switch optimization strategy. At the same time, the control execution and feedback data are recorded for continuous update of device state and support for subsequent failure trend tracking.

[0037] Early warning and control module: according to the influence degree of potential failure trend and warning threshold, output online early warning signal and execute control instruction.

[0038] The process of outputting online early warning signal and executing control instruction in the early warning and control module is: Based on the potential failure trend quantitative index generated by the risk correlation analysis module, the real-time running parameters and historical running data of each SiC device and module are combined to calculate the comprehensive risk score. In the early warning and control module, first, the potential failure trend quantitative index generated by the risk correlation analysis module is obtained, and the real-time running parameters of each SiC device and module are collected, including junction temperature, on-state resistance, current load, etc. Time alignment and normalization processing are performed combined with historical running data, then through weight weighting and multi-factor comprehensive analysis method, the comprehensive risk score of each device and module is formed to provide quantitative basis for subsequent risk level determination.

[0039] Compare the comprehensive risk score with the preset multi-level warning threshold to determine the risk level of the device or module, and generate three-level online early warning signals of high risk, medium risk and low risk. The comprehensive risk score is compared with preset multi-level alert thresholds, classified and determined according to high, medium and low risk levels, and corresponding three-level online early warning signals are generated. During the signal generation process, the system records the risk score, trigger time and trigger threshold information of each device and module, and ensures that the signal can be real-time issued to the monitoring interface and automatic control interface, providing executable input for control decision.

[0040] The system simultaneously records the early warning signals, control instructions and execution results to the online monitoring database, forming a closed-loop feedback; The system generates online early warning signals, corresponding control instructions, and control execution status and results, and synchronously writes them into the online monitoring database, completing the closed-loop feedback; at the same time, time stamp labeling and data integrity checking are performed to ensure the traceability and continuity of subsequent analysis, trend tracking and control effect evaluation.

[0041] It should be noted that, in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply that these entities or operations exist in any such actual relationship or order. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0042] Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An online monitoring system adapted to the cycle use scenario of SiC power devices, characterized by: include: Status acquisition module: real-time acquisition of key operating parameters of SiC power devices, including junction temperature, on-resistance, and operating current fluctuations; When the acquisition parameters exceed the preset warning threshold, it is determined whether subsequent failure trend prediction is required. If so, the prediction analysis module is entered; Prediction and Analysis Module: This module constructs a device performance change curve based on the collected operating parameters, analyzes the curve type and fluctuation trend, and predicts future changes in junction temperature and on-resistance to determine whether the device has potential failure risks. If the prediction results show a potential failure trend, the module enters the control demand analysis module. Regulation and Analysis Module: Based on the parallel operation status of devices and historical abnormal fluctuation events, it determines whether key monitoring or load regulation is required for specific modules. If necessary, it enters the risk association analysis module; Risk Analysis Module: This module analyzes the correlation between on-resistance mutation, junction temperature fluctuation, solder void expansion, and bonding interface aging to determine the impact of potential failure trends and sends the analysis results to the early warning and control module. Early warning and control module: Outputs online early warning signals and executes control instructions based on the impact of potential failure trends and warning thresholds.

2. The online monitoring system adapted to the cyclic use scenario of SiC power devices according to claim 1, characterized in that: The process of determining whether subsequent failure trend prediction is necessary is as follows: Calculates the deviation between the real-time collected junction temperature, on-resistance, and operating current fluctuations and the preset warning thresholds, and combines historical operating data and device aging curves to generate a comprehensive risk score; If the comprehensive risk score is higher than the set prediction trigger threshold, it is determined that subsequent failure trend prediction is required; If the comprehensive risk score is lower than the set threshold, the routine monitoring mode will continue.

3. The online monitoring system adapted to the cyclic use scenario of SiC power devices according to claim 2, characterized in that: The process of constructing the device performance change curve based on the collected operating parameters is as follows: Extract key points of junction temperature, on-resistance and current fluctuations from continuously acquired time series data; Generate multi-parameter performance change curves using smoothing filtering, interpolation and trend fitting methods; Perform time segment analysis on each performance curve to identify fluctuation peaks, trend slopes, and abnormal deviation intervals.

4. The online monitoring system adapted for the cyclic use scenario of SiC power devices according to claim 3, characterized in that: The process of analyzing curve types and fluctuation trends is as follows: Classify performance change curves to identify linear rise, exponential growth, periodic oscillation, and mutation; Calculate the fluctuation amplitude, frequency, and rate of change through statistical analysis and machine learning methods to determine the stability of the device's operating status; If the curve type and fluctuation trend show abnormal patterns, it will be marked as a potential failure interval and enter the potential failure risk prediction stage.

5. The online monitoring system adapted to the cycle use scenario of SiC power devices according to claim 4, characterized in that: The process of determining whether a device has a potential failure risk is as follows: Calculate the potential failure index based on the fluctuation amplitude and trend rate of the performance change curve, combined with device manufacturing characteristics and historical life data; If the potential failure index exceeds the preset threshold, the device is judged to have a potential failure risk and a control demand analysis is triggered; If the potential failure index does not exceed the threshold, maintain the normal monitoring state and continue to collect data.

6. The online monitoring system adapted to the cyclic use scenario of SiC power devices according to claim 5, characterized in that: Combining the parallel operation status of the devices and the historical abnormal fluctuation event process is as follows: Obtain real-time operating parameters of each SiC device in the same power module and analyze load distribution and temperature differences between parallel devices; Combined with the historical abnormal event database, evaluate whether the current device fluctuations are consistent with past failure modes; Generate module-level risk aggregation scores to determine whether specific devices or modules need to be monitored.

7. The online monitoring system adapted for the cyclic use scenario of SiC power devices according to claim 6, characterized in that: The process of determining whether a specific module needs to be monitored or load regulated is as follows: Determine whether a single device or module has overload or thermal stress concentration risks based on module-level risk aggregation scores and real-time load conditions; If the risk score exceeds the set threshold, the module will be marked as a key monitoring target and load regulation suggestions will be generated; If the risk score is below the threshold, maintain normal monitoring and load distribution strategies.

8. The online monitoring system adapted to the cyclic use scenario of SiC power devices according to claim 7, characterized in that: By analyzing the correlation between on-resistance mutation, junction temperature fluctuation, solder layer void expansion, and bonding interface aging, the following process is obtained: Establish a multi-physical quantity correlation model for devices, incorporating on-resistance, junction temperature, solder void size, and bonding interface aging into a unified analysis framework; Statistical correlation analysis and multivariate regression model were used to evaluate the causal relationship and sensitivity among various parameters; Based on the analysis results, quantitative risk indicators of potential failure trends are generated and input into the early warning and control module.

9. The online monitoring system adapted for the cyclic use scenario of SiC power devices according to claim 8, characterized in that: The process of outputting online warning signals and executing control instructions is as follows: Based on the quantitative indicators of potential failure trends generated by the risk association analysis module, a comprehensive risk score is calculated by combining the real-time operating parameters and historical operating data of each SiC device and module; Compare the comprehensive risk score with the preset multi-level warning threshold to determine the risk level of the device or module and generate three-level online warning signals: high risk, medium risk, and low risk; The system also records early warning signals, control instructions and execution results to the online monitoring database, forming a closed-loop feedback loop.

Images