Reliability evaluation method and system for high-power charging module

Abstract

The invention relates to the technical field of electric data processing, and provides a reliability evaluation method and system for a high-power charging module, and the method comprises an obtaining module which is mainly used for obtaining a multi-dimensional data sequence of the high-power charging module in a plurality of operation states, then carrying out the preprocessing, and obtaining a multi-dimensional data sequence; a standardized sequence set is obtained to construct a multi-dimensional fault trend evolution graph and a transient anomaly distribution graph, an early warning sensitive area and an early warning steady-state area are generated based on the multi-dimensional fault trend evolution graph and the transient anomaly distribution graph, scale fusion is conducted on the early warning sensitive area, a steady-state offset path is obtained, and the early warning sensitive area and the early warning steady-state area are obtained. And performing time consistency matching on the steady-state offset path and the early-warning steady-state region to generate a reliability evaluation curve. The reliability evaluation curve is generated through multi-dimensional data sequence collection and standardization processing of the high-power charging module in multiple operation states, the accuracy and reliability of evaluation are improved, and the problem that under the complex operation condition, an existing evaluation method cannot fully reveal the potential fault evolution trend is solved.

Description

Technical Field

[0001] This application relates to the technical field of electrical data processing, and particularly to a method and system for reliability assessment of high-power charging modules. Background Art

[0002] With the rapid development of electric vehicles and intelligent devices, high-power charging technology has become an important research direction in the energy field. Especially the popularization of electric transportation means makes the reliability of high-power charging modules crucial for equipment operation and user experience. The industry's growing demand for fast, stable, and reliable charging systems has promoted technological iteration and innovation.

[0003] In related technical means, the reliability assessment of high-power charging modules is usually based on static performance tests and long-term operation observations. These methods mainly monitor parameters such as charging efficiency, internal temperature changes of the module, and current fluctuations to quantify the stability of the module under different operating environments. This assessment method can effectively detect conventional faults and provide a reference basis for module design optimization.

[0004] Regarding the above technical solution, although the preliminary evaluation of the module's reliability can be achieved through static performance tests and long-term observations, under complex operating conditions, such as when the module is subjected to transient shocks or multi-factor interferences, the existing assessment methods cannot fully reveal the problem of the potential fault evolution trend. Summary of the Invention

[0005] In order to improve the problem that existing assessment methods cannot fully reveal the potential fault evolution trend under complex operating conditions, this application provides a method and system for reliability assessment of high-power charging modules.

[0006] The present invention provides a method for reliability assessment of high-power charging modules, including: obtaining a multi-dimensional data sequence of the high-power charging module in multiple operating states, and preprocessing the multi-dimensional data sequence to obtain a standardized sequence set; constructing a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generating a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, performing collaborative analysis on the node influence factor matrix to obtain a key evolution segment; performing parameter perturbation analysis on the key evolution segment to obtain an incentive influence intensity matrix, performing cumulative deviation analysis on the incentive influence intensity matrix to obtain a trend deviation factor, performing sliding window statistics and dynamic adjustment of the coefficient of variation on the trend deviation factor to obtain an anomaly focus index; using the anomaly focus index to generate a warning sensitive area and a warning steady state area, performing scale fusion on the warning sensitive area to obtain a steady state offset path, and performing time consistency matching on the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

[0007] As a preferred solution, the steps of obtaining the multi-dimensional data sequence of the high-power charging module in multiple operating states and preprocessing the multi-dimensional data sequence to obtain a standardized sequence set include: collecting the real-time working current data, transient temperature rise data, and input and output voltage change data of the high-power charging module under different load conditions, ambient temperature conditions, and grid fluctuation conditions respectively to obtain a multi-dimensional data sequence, where the multi-dimensional data sequence includes a working state composite data block, an environmental response data group, and a power supply disturbance data sequence; performing time window segmentation and amplitude normalization processing on the working state composite data block to obtain an operation behavior sequence, performing temperature segment moving average and dynamic discretization rate calculation on the environmental response data group to obtain a thermal stability index sequence, and performing multi-channel alignment analysis on the operation behavior sequence and the thermal stability index sequence to obtain a behavior thermal response cross-spectrum diagram; performing disturbance segmentation and frequency statistics on the power supply disturbance data sequence to obtain a transient anomaly trigger sequence, performing principal component mapping on the behavior thermal response cross-spectrum diagram and the transient anomaly trigger sequence to obtain a normalized state set, and performing time series encoding on the normalized state set to obtain a standardized sequence set.

[0008] As a preferred solution, the steps of performing disturbance segmentation and frequency statistics on the power supply disturbance data sequence to obtain a transient anomaly trigger sequence, performing principal component mapping on the behavior thermal response cross-spectrum diagram and the transient anomaly trigger sequence to obtain a normalized state set, and performing time series encoding on the normalized state set to obtain a standardized sequence set include: identifying boundary conditions and extracting event labels from the power supply disturbance data sequence to generate a disturbance event index table and a disturbance peak distribution sequence, performing period analysis and frequency statistics on the disturbance peak distribution sequence to generate a frequency band feature mapping diagram; performing cross-checking on the frequency band feature mapping diagram and the disturbance event index table to obtain a transient anomaly trigger sequence, performing continuity verification on the transient anomaly trigger sequence, and generating an anomaly pattern annotation vector based on the verification result; performing principal component mapping on the anomaly pattern annotation vector and the behavior thermal response cross-spectrum diagram to obtain a joint anomaly influence map, extracting stable working segments and high-response fluctuation segments based on the joint anomaly influence map to generate a feature uniformity matrix and an anomaly conflict map; performing grouping aggregation on the feature uniformity matrix and the anomaly conflict map to obtain a normalized state set, and performing time series index encoding on the normalized state set to obtain a standardized sequence set.

[0009] As a preferred solution, the steps of constructing a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generating a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, and performing collaborative analysis on the node influence factor matrix to obtain key evolution segments include: constructing a trend change tensor using the change rates between different dimensions in the standardized sequence set, and unfolding the trend change tensor along the time dimension to obtain a multi-dimensional fault trend evolution map and a transient anomaly distribution map, and performing adjacent node correlation aggregation on the multi-dimensional fault trend evolution map to obtain a trend clustering path set; identifying an incremental mutation region based on the transient anomaly distribution map to generate an anomaly dense region annotation map, and performing associated projection on the trend clustering path set and the anomaly dense region annotation map to generate a node influence factor matrix; identifying a node group with a change mutation rate greater than the adjacent interval based on the node influence factor matrix to obtain a cross-drift flag, and performing fuzzy density filtering and spatial connectivity analysis on the anomaly dense region annotation map and the cross-drift flag to obtain an anomaly trend region and a slow-varying continuous region; performing section matching and boundary collaborative analysis on the anomaly trend region and the slow-varying continuous region to obtain key evolution segments.

[0010] As a preferred solution, the steps of identifying a node group with a change mutation rate greater than the adjacent interval based on the node influence factor matrix to obtain a cross-drift flag, and performing fuzzy density filtering and spatial connectivity analysis on the anomaly dense region annotation map and the cross-drift flag to obtain an anomaly trend region and a slow-varying continuous region include: performing differential calculation and window clustering analysis on the change rates of each node in the node influence factor matrix to obtain a set of fusion nodes, identifying local change rate abnormal transition points from the set of fusion nodes to construct a time transition spectrum map and a structural discontinuity matrix; extracting mutation dense regions and interference boundary segments using the time transition spectrum map to construct a jump region coding map, and performing drift consistency voting analysis on the structural discontinuity matrix and the jump region coding map to obtain a cross-drift flag; performing fuzzy density fitting on the anomaly dense region annotation map and the cross-drift flag to obtain a jump continuity map and an intensity curvature distribution, and performing regional morphology filtering and boundary simplification and reconstruction on the jump continuity map to obtain an anomaly trend region and a trend extension candidate segment; performing spatial connectivity analysis and curvature coupling fusion on the intensity curvature distribution and the trend extension candidate segment to obtain a slow-varying continuous region.

[0011] As a preferred solution, the steps of performing parameter perturbation analysis on the key evolution segment to obtain an incentive influence intensity matrix, performing cumulative deviation analysis on the incentive influence intensity matrix to obtain a trend deviation factor, and performing sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor to obtain an abnormal focus index include: extracting a high-heat response section, a current rebound feature section, and a voltage instability slope section based on the key evolution segment to construct a perturbation feature input set, performing parameter perturbation simulation on the perturbation feature input set to obtain a response offset matrix and a perturbation sensitivity distribution map; performing cross-section cumulative offset evaluation on the response offset matrix to obtain a response fluctuation factor set, and performing interactive clustering analysis on the perturbation sensitivity distribution map and the response fluctuation factor set to generate an incentive influence intensity matrix; performing multi-dimensional time window weighted superposition on the incentive influence intensity matrix to obtain an offset density spectrogram, calculating the local deviation mean and weighted variation rate within each window of the offset density spectrogram to generate a trend deviation factor sequence; performing multi-scale sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor sequence to obtain a focused abnormal intensity map and a jump breakpoint distribution, and performing joint threshold fusion on the focused abnormal intensity map and the jump breakpoint distribution to obtain an abnormal focus index.

[0012] As a preferred solution, the steps of using the abnormal focus index to generate a warning sensitive area and a warning steady state area, performing scale fusion on the warning sensitive area to obtain a steady state offset path, and performing time consistency matching on the steady state offset path and the warning steady state area to generate a reliability evaluation curve include: calculating the fluctuation frequency and abnormal level ranking of each time segment of the high-power charging module using the abnormal focus index to obtain a local peak aggregation area and a periodic flat section, performing regional convolution and scale transformation on the local peak aggregation area to generate a rank index map; performing boundary fusion on the rank index map and the periodic flat section to obtain a warning sensitive area and a warning steady state area, and generating a scale distribution group and a continuous fault mapping set based on the warning sensitive area; performing regional coding and structure compression on the scale distribution group and the continuous fault mapping set to generate a steady state offset path, and performing time axis synchronization alignment and jump segment cross-over analysis on the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

[0013] The present application also provides a reliability evaluation system for a high-power charging module, including: an acquisition module, configured to acquire a multi-dimensional data sequence of the high-power charging module in multiple operating states, and perform preprocessing on the multi-dimensional data sequence to obtain a standardized sequence set; an analysis module, configured to construct a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generate a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, perform collaborative analysis on the node influence factor matrix to obtain a key evolution segment; an adjustment module, configured to perform parameter perturbation analysis on the key evolution segment to obtain an incentive influence intensity matrix, perform cumulative deviation analysis on the incentive influence intensity matrix to obtain a trend deviation factor, perform sliding window statistics and dynamic adjustment of the coefficient of variation on the trend deviation factor to obtain an anomaly focusing index; a generation module, configured to generate a warning sensitive area and a warning steady state area by using the anomaly focusing index, perform scale fusion on the warning sensitive area to obtain a steady state offset path, perform time consistency matching on the steady state offset path and the warning steady state area, and generate a reliability evaluation curve.

[0014] Compared with the prior art, the present application has the following beneficial effects: high accuracy and strong flexibility. Through the acquisition and standardized processing of the multi-dimensional data sequence of the high-power charging module in multiple operating states, a comprehensive standardized sequence set is generated, providing a basis for subsequent map construction. Based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, a key evolution segment is extracted through collaborative analysis of the node influence factor matrix, and combined with parameter perturbation experiments and cumulative deviation analysis, the incentive influence intensity and the trend deviation factor are deeply mined, and an anomaly focusing index is generated to accurately locate the fault incentive. Through the fusion of the warning area and the steady state path, a reliability evaluation curve is generated, improving the accuracy and reliability of the evaluation and solving the problem that the existing evaluation method cannot fully reveal the potential fault evolution trend under complex operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the conditions for the implementation of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0017] Figure 1 It is a schematic flowchart of a method for evaluating the reliability of a high-power charging module provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of a system for evaluating the reliability of a high-power charging module provided by an embodiment of the present invention.

[0018] Explanation of reference numerals: 10. System for evaluating the reliability of a high-power charging module; 11. Acquisition module; 12. Analysis module; 13. Adjustment module; 14. Generation module. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] The flowchart shown in the drawings is only an example, and does not necessarily include all the content and operations / steps, nor does it necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may change according to the actual situation.

[0021] It should also be understood that the terms used in this specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification of the present application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0022] It should be further understood that the term " / and" used in this specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0023] Next, the technical solutions of the present invention will be further described in conjunction with the drawings and through specific implementation manners.

[0024] Embodiment 1: As Figure 1 shown, the present application provides a reliability evaluation method for a high-power charging module, including steps S100 to S400.

[0025] Step S100: Obtain the multi-dimensional data sequences of the high-power charging module in multiple operating states, and preprocess the multi-dimensional data sequences to obtain a standardized sequence set.

[0026] In this step, the current, voltage, temperature and other multi-dimensional data sequences of the high-power charging module in multiple operating states are collected through the built-in sensors of the high-power charging module, and the normalization algorithm is used to standardize the parameters in the multi-dimensional data sequences to eliminate the dimension differences between the data, and finally a standardized sequence set is formed. Specifically, the standard deviation normalization method or the extreme value normalization method can be used to process the data sequences to ensure that the data in each dimension have the same dimension and comparability.

[0027] For example, when in the operating state of "low-temperature charging", the collected current sequence is {30A, 32A, 31A}, which is transformed into {0.93, 1, 0.97} after normalization processing to ensure the consistency of the data under different states.

[0028] Step S200: Construct a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generate a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, and perform collaborative analysis on the node influence factor matrix to obtain the key evolution segment.

[0029] In this step, the standardized sequence set is used to generate a multi-dimensional fault trend evolution map to show the dynamic change trend of each parameter of the high-power charging module; at the same time, the transient anomaly distribution map is used to capture the distribution state of parameter mutation or deviation. Subsequently, based on the node influence analysis in the map, a node influence factor matrix is generated, and matrix collaborative calculation is performed on it to identify the key and influential stages in the evolution process. Specifically, the matrix decomposition algorithm can be used to analyze the key fault causes and correlations in the module to efficiently extract the key evolution segments.

[0030] For example, in the fault trend evolution map, it is found that the temperature of the charging module fluctuates abnormally in the operating state of "high-temperature charging" and is closely related to the current change. Further, through the node influence factor matrix, it is obtained that the key evolution segment at this time is the stage of the module temperature rising.

[0031] Step S300: Perform parameter perturbation analysis on the key evolution segment to obtain an inducement influence intensity matrix, perform cumulative deviation analysis on the inducement influence intensity matrix to obtain a trend deviation factor, and perform sliding window statistics and coefficient of variation dynamic adjustment on the trend deviation factor to obtain an anomaly focus index.

[0032] In this step, for the key evolution segments, external and internal changes are simulated through parameter perturbation experiments to generate an incentive influence intensity matrix, and the cumulative deviation analysis is used to capture the influence degree of incentive changes on the trends of multi-dimensional data sequences, forming a trend deviation factor. Subsequently, by dynamically adjusting the sliding window size and the coefficient of variation, the abnormal data in the trend deviation factor are statistically screened to obtain an abnormal focusing index that can accurately locate faults. Specifically, the high-frequency abnormalities of multi-dimensional data sequences can be captured through a time series analysis model, and the sensitivity of fault location can be enhanced through the abnormal focusing index.

[0033] For example, the perturbation range {50°C, 55°C, 60°C} is set for the temperature parameter in the key evolution segment, and it is found that the temperature change produces a significant deviation in the current fluctuation. After further calculation, the abnormal focusing index value is 0.85.

[0034] Step S400: Use the abnormal focusing index to generate a warning sensitive area and a warning steady state area, perform scale fusion on the warning sensitive area to obtain a steady state offset path, and perform time consistency matching between the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

[0035] In this step, the sensitive interval and the stable interval of the module operation state are divided based on the abnormal focusing index, and multi-dimensional scale fusion is performed on the sensitive interval to construct a steady state offset path. Then, a time correlation analysis is performed between the steady state offset path and the stable interval to generate a reliability evaluation curve of the module. Specifically, potential fault trends can be identified through the interactive mapping between the offset path and the warning area, and a quantitative reliability score can be output.

[0036] For example, in the "low-temperature charging" state, the matching result between the steady state offset path and the warning area shows that the trend deviation index exceeds the threshold, and the generated reliability evaluation curve indicates that the reliability level of the module in this state is "good".

[0037] In this embodiment, by acquiring the multi-dimensional data sequences of the high-power charging module in multiple operating states and performing normalization processing on the multi-dimensional data sequences to form a normalized sequence set. Based on the normalized sequence set, a multi-dimensional fault trend evolution map and a transient anomaly distribution map are constructed. By collaboratively analyzing the node influence factor matrix, key evolution segments are extracted. After performing parameter perturbation analysis on the key evolution segments, an inducement influence intensity matrix is generated, and a trend deviation factor is obtained through cumulative deviation analysis. Then, through sliding window statistics and dynamic adjustment of the coefficient of variation, an anomaly focusing index is generated, and the early warning sensitive area and the early warning steady state area are divided using the anomaly focusing index. Finally, a steady state offset path is obtained through scale fusion and time consistency matching with the early warning steady state area, thereby generating a reliability evaluation curve. Through the in-depth analysis of the multi-dimensional data sequences and the construction of the multi-dimensional fault trend evolution map, the fault trend of the high-power charging module under complex operating conditions can be comprehensively revealed. By introducing the anomaly focusing index and its dynamic adjustment technology, the accurate positioning and dynamic tracking of the fault inducement are realized, effectively improving the accuracy of the reliability evaluation and improving the problem that the existing evaluation methods cannot fully reveal the potential fault evolution trend under complex operating conditions.

[0038] Embodiment 2: In step S100, by respectively collecting the real-time working current data, transient temperature rise data, and input and output voltage change data of the high-power charging module under different load conditions, environmental temperature conditions, and grid fluctuation conditions, a multi-dimensional data sequence is obtained, where the multi-dimensional data sequence includes a working state composite data block, an environmental response data group, and a power supply disturbance data sequence.

[0039] By respectively collecting the real-time working current data, transient temperature rise data, and input and output voltage change data of the high-power charging module under different load conditions, environmental temperature conditions, and grid fluctuation conditions, a multi-dimensional data sequence is formed. Among them, the multi-dimensional data sequence includes a working state composite data block, an environmental response data group, and a power supply disturbance data sequence. Specifically, for the working state composite data block, a continuous load test bench can be used to simulate the operating behavior of the module under different load conditions, and its working current data and the corresponding change in power loss inside the module are recorded; for the environmental response data group, the change in environmental temperature can be monitored through a high-precision temperature acquisition device, and the transient temperature rise data inside the module is obtained synchronously; for the power supply disturbance data sequence, a grid disturbance simulator is used to generate voltage fluctuation conditions with different intensities and frequencies, and the voltage change conditions at the input and output ends of the module are recorded. The above collection process sets a dynamic adjustment period and a storage depth to ensure the integrity and continuity of the data.

[0040] For example, under different load conditions, the load resistances are set to 10Ω, 15Ω, and 20Ω respectively, and the corresponding working current data {50A, 35A, 25A} is recorded to form a working state composite data block; under environmental temperature conditions, the temperature is adjusted to 25°C, 35°C, and 45°C, and the transient temperature rise data {5°C, 10°C, 15°C} is recorded respectively to form an environmental response data group; under power grid fluctuation conditions, the input voltages are set to 220V±5% and 220V±10% respectively, and the output voltage changes {215V, 210V} are recorded to form a power supply disturbance data sequence.

[0041] The working state composite data block is subjected to time window segmentation and amplitude normalization processing to obtain an operation behavior sequence. The environmental response data group is subjected to temperature segment moving average and dynamic discrete rate calculation to obtain a thermal stability index sequence. The operation behavior sequence and the thermal stability index sequence are subjected to multi-channel alignment analysis to obtain a behavior thermal response cross-spectrum diagram.

[0042] By performing time window segmentation on the working state composite data block, the continuous working current data is grouped according to a fixed time period, and amplitude normalization processing is performed on each group of data to eliminate the differences between different load conditions, finally forming an operation behavior sequence; for the environmental response data group, temperature segment moving average is performed in combination with dynamic discrete rate calculation to obtain the thermal stability performance of the module under different environmental temperature conditions, and finally a thermal stability index sequence is obtained. Subsequently, the operation behavior sequence and the thermal stability index sequence are subjected to multi-channel alignment analysis, and by synchronously comparing the change trends and correlations of different data sequences during operation, a behavior thermal response cross-spectrum diagram is generated.

[0043] For example, in the working state composite data block, the time period is set to 10 seconds, and the working current data after segmentation is {50A, 35A, 25A}. After amplitude normalization processing, the operation behavior sequence obtained is {1.0, 0.7, 0.5}; after performing moving average calculation on the environmental response data group, the thermal stability index sequence is {0.8, 0.6, 0.4}; through alignment analysis, the synchronous change trends of the operation behavior sequence and the thermal stability index sequence are presented in the behavior thermal response cross-spectrum diagram.

[0044] The power supply disturbance data sequence is subjected to disturbance segmentation and frequency statistics to obtain a transient anomaly trigger sequence. The behavior thermal response cross-spectrum diagram and the transient anomaly trigger sequence are subjected to principal component mapping to obtain a normalized state set. The normalized state set is subjected to time series encoding to obtain a standardized sequence set.

[0045] By segmenting the power supply disturbance data sequence, the power supply disturbance data is divided into multiple disturbance intervals along the time axis, and frequency statistics are performed for each interval to generate a transient anomaly trigger sequence. Specifically, first, analyze the disturbance boundaries of each interval to determine the time range and intensity change of each disturbance; then record the trigger times of high-frequency disturbances and their corresponding time points through frequency statistics, mark abnormal disturbance events, and form a complete transient anomaly trigger sequence. Next, perform principal component mapping on the behavioral thermal response cross-spectrum diagram and the transient anomaly trigger sequence. By constructing a multi-dimensional space transformation, integrate the comprehensive effects of high-response behaviors and transient anomaly events to generate a normalized state set. Finally, perform time series encoding on the normalized state set, encoding the time series features and data distribution rules into a standardized sequence set to ensure the unity of the data sequence and the accuracy of the analysis.

[0046] For example, for the power supply disturbance data sequence, set the time axis to be divided into 10-second intervals, and identify five disturbance events occurring at time points such as the 5th second, the 20th second, and the 45th second. The frequency statistics record the intensity change range as {10V, 15V, 20V} to generate a transient anomaly trigger sequence. Subsequently, perform principal component mapping on the thermal response change amplitudes {0.2, 0.3, 0.5} in the behavioral thermal response cross-spectrum diagram and the above transient anomaly trigger sequence, comprehensively analyze the correlation between the thermal response behavior and the intensity of abnormal events, finally generate a normalized state set, and obtain a standardized sequence set {0.1, 0.3, 0.5, 0.7} after performing time series encoding on the normalized state set.

[0047] Among them, the steps of performing disturbance segmentation and frequency statistics on the power supply disturbance data sequence to obtain a transient anomaly trigger sequence, performing principal component mapping on the behavioral thermal response cross-spectrum diagram and the transient anomaly trigger sequence to obtain a normalized state set, and performing time series encoding on the normalized state set to obtain a standardized sequence set include: identifying boundary conditions and extracting event labels from the power supply disturbance data sequence to generate a disturbance event index table and a disturbance peak distribution sequence, performing periodic analysis and frequency statistics on the disturbance peak distribution sequence, and generating a frequency band feature mapping diagram.

[0048] By identifying the boundary conditions of the power supply disturbance data sequence, detect the start and end positions of each disturbance event and mark its trigger time point, and at the same time extract the corresponding event label to label the disturbance category, such as short-term fluctuations, long-term interference, etc., and finally generate a disturbance event index table. Subsequently, perform periodic analysis on the disturbance peak data in the disturbance event index table to identify periodic fluctuation rules, and calculate the trigger frequencies of various types of events in different periods through frequency statistics to form a complete disturbance peak distribution sequence. Based on the frequency band analysis, map the disturbance event index table and the disturbance peak distribution sequence to generate a frequency band feature mapping diagram for comprehensively analyzing the disturbance peak distribution rules and their influence characteristics.

[0049] For example, three short-time fluctuation events are identified in the power supply disturbance data sequence, which occur at the 10th second, 30th second, and 50th second respectively, and are marked as "Fluctuation Event 1", "Fluctuation Event 2", and "Fluctuation Event 3", generating a disturbance event index table. Subsequently, a periodic analysis is performed on the peaks of these events, and the periodic trigger frequencies are found to be {10Hz, 20Hz, 30Hz}. Finally, a frequency band feature mapping diagram is generated, where the distribution law of the fluctuation intensity shows that the average intensity of the short-time fluctuation events is 15V.

[0050] The frequency band feature mapping diagram and the disturbance event index table are cross-checked to obtain a transient anomaly trigger sequence. The transient anomaly trigger sequence is continuously checked, and an anomaly pattern annotation vector is generated based on the verification result.

[0051] By performing point-by-point cross-checking on the frequency band feature mapping diagram and the disturbance event index table, the consistency between the two in terms of time and intensity is detected. Specifically, first, the high disturbance intensity regions in the frequency band feature mapping diagram are marked, and the trigger conditions are identified by matching the corresponding event labels in the disturbance event index table. Then, a transient anomaly trigger sequence is generated based on the matching result, recording the time sequence and the change process of the disturbance amplitude of all trigger events. Subsequently, the transient anomaly trigger sequence is continuously checked, analyzing the intervals and intensity changes between trigger events, and generating a verification result for the abnormal events. Finally, the verification result is used to construct an anomaly pattern annotation vector, and each abnormal event in the annotation vector is classified and marked to ensure the accuracy of the abnormal feature annotation.

[0052] For example, three high disturbance intensity regions are detected in the frequency band feature mapping diagram, corresponding to the 15th second, 35th second, and 50th second respectively. By matching with the disturbance event index table, these regions are marked as "Transient Event A", "Transient Event B", and "Transient Event C" respectively. The continuous check finds that the intervals between events are 20 seconds, and the intensity changes are {10V, 15V, 20V} respectively. Finally, an anomaly pattern annotation vector is generated, where the vector content is {A: 10V, B: 15V, C: 20V}.

[0053] Principal component mapping is performed on the anomaly pattern annotation vector and the behavioral thermal response cross-spectrum diagram to obtain a combined anomaly influence mapping diagram. Based on the combined anomaly influence mapping diagram, a stable working section and a high response fluctuation section are extracted to generate a feature uniformity matrix and an anomaly conflict mapping diagram.

[0054] By performing principal component mapping on the anomaly pattern annotation vector and the cross-spectrum diagram of behavioral thermal response, a comprehensive analysis model is constructed to reveal the correlation between anomaly-triggering events and thermal response behaviors. Specifically, the anomaly event attributes in the annotation vector and the thermal response behaviors in the cross-spectrum diagram are mapped into a multi-dimensional space, and a joint anomaly influence diagram is generated through data decomposition and clustering analysis. Subsequently, based on the joint anomaly influence diagram, a stable working section and a high-response fluctuation section are extracted. Statistical mean analysis is performed on the stable working section to generate a feature uniformity matrix, while anomaly characteristic aggregation analysis is performed on the high-response fluctuation section to generate an anomaly conflict diagram.

[0055] For example, mapping the "transient event B" in the anomaly pattern annotation vector to the corresponding change amplitude of the thermal response behavior in the cross-spectrum diagram of behavioral thermal response is 0.3. Through principal component analysis, a joint anomaly influence diagram is generated, and the diagram shows that "transient event B" has a high correlation with the high-response fluctuation section. Subsequently, the mean analysis of the behavioral characteristics of the stable working section yields a feature uniformity matrix with the content {stability: 0.85}, and the anomaly aggregation analysis of the high-response fluctuation section generates an anomaly conflict diagram with the content {conflict intensity: 0.7}.

[0056] Grouping and aggregating the feature uniformity matrix and the anomaly conflict diagram to obtain a normalized state set, and performing time-series index coding on the normalized state set to obtain a standardized sequence set.

[0057] By grouping and aggregating the feature uniformity matrix and the anomaly conflict diagram, hierarchical classification is performed on the data characteristics of the stable working section and the high-response fluctuation section to form a normalized state set. Specifically, according to the aggregation relationship of each data attribute in the matrix and the diagram, the stable working section is classified into a group with consistent characteristics, and the high-response fluctuation section is classified into a group with different characteristics. Then, normalization processing is performed on the data of each group to eliminate data deviation. Subsequently, time-series index coding is performed on the normalized state set, and an index coding sequence is constructed based on the event characteristics order on the time axis, finally forming a unified standardized sequence set.

[0058] For example, the stability attribute value of {0.85} in the feature uniformity matrix is classified into the group with consistent characteristics, and the conflict intensity attribute value of {0.7} in the anomaly conflict diagram is classified into the group with different characteristics. After grouping and aggregating, the content of the normalized state set is {consistent group: [0.85], different group: [0.7]}, the result of time-series index coding is {stable: 0.85, conflict: 0.7}, and the final standardized sequence set is {0.85, 0.7}.

[0059] In step S200, a trend change tensor is constructed using the change rates between different dimensions in the standardized sequence set, and the trend change tensor is unfolded along the time dimension to obtain a multi-dimensional fault trend evolution diagram and a transient anomaly distribution diagram. Adjacent node correlation aggregation is performed on the multi-dimensional fault trend evolution diagram to obtain a trend clustering path set.

[0060] By utilizing the change rates among various dimensions in the standardized sequence set, a trend change tensor is constructed to determine the main characteristic change trend in the module operating state. Specifically, the change rates of multi-dimensional data are calculated to form a time series tensor structure, and the trend change tensor is unfolded along the time dimension to map the mutual dynamic relationship between parameters, generating a multi-dimensional fault trend evolution map. At the same time, through the feature extraction of transient anomalies, a transient anomaly distribution map is constructed to record the concentrated distribution regions of mutations in the data of each dimension during operation. Subsequently, the adjacent nodes in the multi-dimensional fault trend evolution map are subjected to correlation aggregation, the dynamic association between the nodes is analyzed, and a trend clustering path set is generated based on the aggregation result.

[0061] For example, by calculating the change rates of the "working current" and "input voltage" dimensions in the standardized sequence set, the time change sequence of the "current-voltage relationship" in the trend change tensor is obtained as {0.2, 0.3, 0.1}; through time unfolding, a fault trend evolution map is generated, showing that the voltage change rate increases significantly between the 10th second and the 20th second, and the trend clustering path set obtained after the aggregation of node correlations is {Path 1: [Node A, Node B], Path 2: [Node C, Node D]}.

[0062] Based on the transient anomaly distribution map, the incremental mutation region is identified to generate an abnormal dense region annotation map, and the trend clustering path set is associated and projected with the abnormal dense region annotation map to generate a node influence factor matrix.

[0063] Through the transient anomaly distribution map, the incremental mutation region is identified, and the abnormal mutation phenomena of each dimension are marked to generate an abnormal dense region annotation map. Specifically, the boundary regions of the mutation points in the distribution map are detected, the time segments and intensity parameters of the concentrated distribution region are extracted, and the characteristic distribution of the dense abnormal region is recorded through the annotation map. Subsequently, the trend clustering path set is associated and projected with the abnormal dense region annotation map, and through the time alignment between paths and the analysis of the dynamic relationship of nodes within the region, a node influence factor matrix is generated. This matrix includes the influence weights and spatial distributions of each node on the abnormal mutation phenomenon.

[0064] For example, it is detected that the incremental mutation region in the transient anomaly distribution map is located between the 25th second and the 35th second, and the intensity changes within the region are {5%, 10%, 20%}. An abnormal dense region annotation map is generated to record the mutation occurrence time and location information; by associating and projecting the trend clustering path set with the annotation map, the dynamic relationship between Node A and Node B within the abnormal region is analyzed, and the generated node influence factor matrix is {Node A: 0.8, Node B: 0.9}.

[0065] Identify the node group with a change mutation rate greater than that of the adjacent interval based on the node influence factor matrix to obtain the cross-drift flag. Perform fuzzy density filtering and spatial connectivity analysis on the abnormal dense area annotation map and the cross-drift flag to obtain the abnormal trend area and the slow-varying continuous area.

[0066] By analyzing the change rate in the node influence factor matrix, identify the node group with a change mutation rate greater than that of the adjacent interval and mark it as the cross-drift flag. Specifically, perform a difference calculation on the change rate of the influence factor to determine the intensity of the mutation point and the corresponding time interval, and generate a cross-drift flag to record the triggering conditions for the dynamic changes of the nodes. Then perform fuzzy density filtering on the abnormal dense area annotation map and the cross-drift flag to eliminate interfering data points, and through spatial connectivity analysis, dynamically link each abnormal node in the area to divide the abnormal trend area and the slow-varying continuous area.

[0067] For example, through the calculation of the change rate of node B in the node influence factor matrix, it is found that the mutation intensity is {15%, 20%}, and the time interval is from the 30th second to the 40th second. Generate a cross-drift flag to record the mutation point. Subsequently, perform density filtering on the mutation points in the abnormal dense area annotation map to retain the key area. After spatial connectivity analysis, the abnormal trend area is {the associated area of node B}, and the slow-varying continuous area is {the stable area of node A}.

[0068] Perform section matching and boundary coordination analysis on the abnormal trend area and the slow-varying continuous area to obtain the key evolution section.

[0069] By performing section matching on the abnormal trend area and the slow-varying continuous area, identify the boundary transition characteristics between the two, and determine the time range and change intensity of the key evolution section through boundary coordination analysis. Specifically, through the comparison of the dynamic characteristics between sections, extract the area with the most significant influence, and combine with the coordination analysis tool to measure the concentrated manifestation of the fault trend in the area, and finally generate the key evolution section.

[0070] For example, through the coordination analysis of the boundary sections of the abnormal trend area and the slow-varying continuous area, it is found that the boundary transition time is from the 20th second to the 30th second, and the change intensity increases to {10%, 15%}. Finally, determine that the key evolution section is the high-fault concentration area from the 25th second to the 30th second.

[0071] Among them, the steps of identifying the node group with a change mutation rate greater than that of the adjacent interval based on the node influence factor matrix to obtain the cross-drift flag, and performing fuzzy density filtering and spatial connectivity analysis on the abnormal dense area annotation map and the cross-drift flag to obtain the abnormal trend area and the slow-varying continuous area include: performing difference calculation and window clustering analysis on the change rate of each node in the node influence factor matrix to obtain the fusion node set, and identifying the local change rate abnormal transition points from the fusion node set to construct the time transition spectrum diagram and the structure discontinuous matrix.

[0072] By performing differential calculations on the change rates of each node in the node influence factor matrix to analyze the gradient characteristics of data changes in the time series, and combining the window clustering analysis method, nodes with similar change trends are aggregated to form a fusion node set. Specifically, the continuous change rates within the matrix are grouped by sliding a time window, and node groups with abnormal aggregation of change rates are identified through a clustering algorithm, and local change rate abnormal jump points are screened out from them. Then, based on the jump points, a time jump spectrum diagram is constructed to show the distribution of mutation events in the time dimension, and a structure discontinuous matrix is generated to record the discontinuous characteristics of the dynamic associations between nodes.

[0073] For example, for the change rates in the node influence factor matrix, the differential calculation results in a time series change of {0.05, 0.15, 0.25}. Through window clustering analysis, a fusion node set {Node A, Node C} is extracted, and the local abnormal jump point is identified as the change rate of Node C being 0.25. The constructed time jump spectrum diagram shows that the jump occurs at the 15th second, and at the same time, a structure discontinuous matrix is generated to mark the interruption of the association between Node A and Node C.

[0074] The mutation dense region and interference boundary segments are extracted using the time jump spectrum diagram to construct a jump region coding map, and a drift consistency voting analysis is performed on the structure discontinuous matrix and the jump region coding map to obtain a cross-drift flag.

[0075] The mutation dense region is extracted through the time jump spectrum diagram to identify the aggregation positions of abnormal change nodes, and at the same time, interference boundary segments are detected to construct a jump region coding map. Specifically, the mutation events in the time jump spectrum diagram are analyzed, the time range and intensity distribution of the regions where they are located are extracted, and the interference boundaries are coded to mark the characteristics of abnormal regions. Subsequently, a drift consistency voting analysis is performed on the structure discontinuous matrix and the jump region coding map, the consistency of their spatial associations is calculated, the dynamic relationships between abnormal nodes are determined, and finally, a cross-drift flag is generated.

[0076] For example, it is detected in the time jump spectrum diagram that the mutation dense region is located between the 20th second and the 25th second, the interference boundary segments are extracted as {Node B, Node D}, and the constructed jump region coding map records the change amplitudes as {0.3, 0.4}; through the drift consistency voting analysis with the structure discontinuous matrix, the result shows that the dynamic association consistency between Node B and Node D is 0.85, and finally, a cross-drift flag is generated to record the associated mutation events.

[0077] The abnormal dense region annotation map and the cross-drift flag are subjected to fuzzy density fitting to obtain a jump continuity map and an intensity curvature distribution. The jump continuity map is subjected to regional morphological filtering and boundary simplification and reconstruction to obtain an abnormal tendency region and a trend extension candidate segment.

[0078] By performing fuzzy density fitting on the abnormal dense area annotation map and the cross-drift flag, fitting analysis is carried out on the density data within the mutation region to generate a jump continuity map to characterize the continuous changes within the region. At the same time, the intensity curvature distribution is calculated to quantify the intensity of regional changes. Specifically, regional morphological filtering is performed on the abnormal nodes in the jump continuity map to remove noise data points, and the morphological characteristics of the mutation region are optimized through boundary simplification and reconstruction technology. Finally, the abnormal trend area and the trend expansion candidate segment are divided to ensure the effectiveness and concentration of data changes within the region.

[0079] For example, the mutation point intensities {0.15, 0.25, 0.35} in the abnormal dense area annotation map are subjected to fuzzy density fitting with the cross-drift flag, and the generated jump continuity map shows a dense distribution of changes within the region; the calculated intensity curvature distribution gives curvature values {0.8, 0.9}; noise data points are removed through regional morphological filtering, and after boundary simplification, the abnormal trend area is determined to be from the 18th second to the 24th second, and the trend expansion candidate segment is from the 25th second to the 30th second.

[0080] Spatial connectivity analysis and curvature coupling fusion are performed on the intensity curvature distribution and the trend expansion candidate segment to obtain a slow-varying continuous area.

[0081] Through spatial connectivity analysis of the intensity curvature distribution and the trend expansion candidate segment, the dynamic association between the two is detected to identify the spatial characteristics within the region. Specifically, the connectivity of the nodes in the intensity curvature distribution is measured, the coupling relationship within the trend expansion candidate segment is analyzed, and fusion processing is carried out in combination with the curvature change characteristics to generate a region with high continuity and slow change, which is finally divided into a slow-varying continuous area to represent the stable state characteristics of module operation.

[0082] For example, the coupling association of the nodes with curvature values {0.7, 0.8} in the intensity curvature distribution within the trend expansion candidate segment is 0.9. After spatial connectivity analysis, the time range of the slow-varying continuous area is determined to be from the 28th second to the 35th second, and the intensity of the stable state characteristics within the region is recorded as 0.75.

[0083] In step S300, based on the key evolution segment, a high-heat response segment, a current rebound feature segment, and a voltage instability slope segment are extracted to construct a perturbation feature input set. Parameter perturbation simulation is performed on the perturbation feature input set to obtain a response offset matrix and a perturbation sensitivity distribution map.

[0084] Through comprehensive analysis of key evolution segments, the high - heat response section, the current rebound characteristic section, and the voltage instability slope section are extracted, which represent the rapid temperature rise, the drastic change of transient current, and the unstable trend of input - output voltage during module operation, respectively. Specifically, thermal power consumption estimation and temperature - rise curve fitting are performed on the high - heat response section, current - time series fluctuation analysis is carried out on the current rebound characteristic section, input - output voltage difference evaluation is conducted on the voltage instability slope section, and all paragraph data are integrated to generate a perturbation - feature input set. Subsequently, parameter perturbation simulation is carried out on the perturbation - feature input set. By setting the perturbation range of different environmental parameters, the dynamic response behaviors of each characteristic section are simulated to obtain a response deviation matrix and a perturbation sensitivity distribution map.

[0085] For example, in the high - heat response section, the rapid temperature - rise range is {50°C to 70°C}, in the current rebound characteristic section, the current change amplitude is {15 A to 30 A}, and in the voltage instability slope section, the voltage difference amplitude is {10 V to 20 V}. Through parameter perturbation simulation, the response deviation matrix is {heat - zone deviation: 5%, current - fluctuation deviation: 10%, voltage - slope deviation: 15%}, and a perturbation sensitivity distribution map is generated to record the distribution density of the deviation data.

[0086] Perform cross - section cumulative deviation evaluation on the response deviation matrix to obtain a response - fluctuation factor set, and conduct interactive clustering analysis on the perturbation sensitivity distribution map and the response - fluctuation factor set to generate an inducement influence intensity matrix.

[0087] Through cross - section cumulative deviation evaluation of the response deviation matrix, integrate the dynamic deviation conditions of each characteristic section, quantify the cumulative influence between key parameters to generate a response - fluctuation factor set. Specifically, use the parameter fluctuation conditions in the deviation matrix for cumulative calculation, and determine the comprehensive influence factor across sections through correlation analysis. Subsequently, conduct interactive clustering analysis on the perturbation sensitivity distribution map and the response - fluctuation factor set, identify the dynamic correlation characteristics of sensitive parameters, extract the strong correlations between inducements through clustering algorithms, and finally generate an inducement influence intensity matrix.

[0088] For example, in the response deviation matrix, the deviation value of the high - heat response section is {5%}, the deviation value of the current rebound characteristic section is {10%}, and the deviation value of the voltage instability slope section is {15%}. After cumulative calculation, the content of the response - fluctuation factor set is {heat - zone influence intensity: 0.85, current influence intensity: 0.75, voltage influence intensity: 0.9}; the inducement influence intensity matrix generated through interactive clustering analysis is {heat - zone current correlation: 0.6, voltage - heat - zone correlation: 0.8}.

[0089] Perform multi - dimensional time - window weighted superposition on the inducement influence intensity matrix to obtain a deviation density spectrogram, and calculate the local deviation from the mean and the weighted variation rate within each window of the deviation density spectrogram to generate a trend - deviation factor sequence.

[0090] By performing multi-dimensional time-window weighted superposition on the incentive influence intensity matrix, calculating the weights of the dynamic influences of each incentive within the time window, analyzing the overall trend of incentive changes, and generating an offset density spectrogram to display the density distribution characteristics of each incentive. Specifically, calculate the deviation from the mean of the local data in the offset density spectrogram to identify significant change trends within the region, and evaluate the degree of change in combination with the weighted coefficient of variation, and finally form a trend deviation factor sequence to quantify the deviation characteristics of each time window.

[0091] For example, set the time window width in the incentive influence intensity matrix to {5 seconds, 10 seconds, 15 seconds}, and the calculated offset density spectrogram after weighted superposition shows that the offset density within the time period is {0.6, 0.8, 0.9}; the results of local deviation from the mean calculation are {mean deviation: 0.7, weighted coefficient of variation: 0.15}, and the generated trend deviation factor sequence content is {factor 1: 0.7, factor 2: 0.85, factor 3: 0.9}.

[0092] Perform multi-scale sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor sequence to obtain a focused anomaly intensity map and a jump breakpoint distribution, and perform joint threshold fusion on the focused anomaly intensity map and the jump breakpoint distribution to obtain an anomaly focus index.

[0093] By performing multi-scale sliding window statistics on the trend deviation factor sequence, analyze the dynamic change characteristics under different window sizes, and at the same time perform dynamic coefficient of variation adjustment for the variation of the deviation factors to ensure the sensitivity of data analysis. Subsequently, generate a focused anomaly intensity map to record the change intensity of the anomaly factors, and perform feature statistics on the jump breakpoint distribution to mark the significant change positions. Combine the joint threshold fusion calculation of the focused anomaly intensity map and the jump breakpoint distribution, extract the anomaly characteristics of the high-sensitivity region, and finally generate an anomaly focus index to characterize the key anomaly range during the operation of the module.

[0094] For example, the content of the trend deviation factor sequence is {factor 1: 0.7, factor 2: 0.85, factor 3: 0.9}, the multi-scale sliding window is set to {window 1: 5 seconds, window 2: 10 seconds}, and the focused anomaly intensity map shows that the intensity distribution is {0.8, 0.9}; the jump breakpoint distribution records the breakpoint positions as {the 20th second, the 35th second}; the anomaly focus index generated after the joint threshold fusion calculation is {intensity index: 0.9, sensitive region: [from the 20th second to the 35th second]}.

[0095] In step S400, use the anomaly focus index to calculate the fluctuation frequency and sort the anomaly levels of each time segment of the high-power charging module to obtain a local peak aggregation region and a periodic flat section, and perform regional convolution and scale transformation on the local peak aggregation region to generate a rank index map.

[0096] Calculate the fluctuation frequency of each time segment of the high-power charging module through abnormal focusing indicators, analyze the abnormal fluctuation intensity and frequency characteristics during operation, and rank the abnormal levels according to the calculation results to identify the key abnormal ranges. Specifically, divide the time segment into multiple intervals, respectively count the fluctuation intensity and frequency, and combine the abnormal level indicators to rank and form data priorities. Subsequently, perform regional convolution analysis on the local peak aggregation area, perform convolution superposition processing on the abnormal distribution within the area, optimize the hierarchy of the abnormal distribution through scale transformation, and finally generate a level index map to display the abnormal level distribution and influence intensity of different time segments of the module.

[0097] For example, in the calculation of the fluctuation frequency of the time segment {5 seconds, 10 seconds, 15 seconds}, it is found that the intensities are {0.3Hz, 0.5Hz, 0.8Hz} respectively, and the abnormal level ranking results are {Level 1: 0.8Hz, Level 2: 0.5Hz, Level 3: 0.3Hz}; after performing regional convolution processing on the local peak aggregation area {10 seconds to 15 seconds}, the convolution intensity is {0.6}; the level index map generated through scale transformation shows that the abnormal intensity distribution is {Region 1: Level 1 intensity, Region 2: Level 2 intensity}.

[0098] Fuse the boundary of the level index map with the periodic flat section to obtain the warning sensitive area and the warning steady state area, and generate a scale distribution group and a continuous fault mapping set based on the warning sensitive area.

[0099] Through the fusion analysis of the boundary data of the level index map and the periodic flat section, divide the abnormal and stable areas in the operation state of the module. Specifically, according to the abnormal distribution characteristics of different level areas in the level index map, compare with the recorded operation stable time range in the periodic flat section to generate the warning sensitive area and the warning steady state area to clarify the abnormal risk level of the module operation state. Subsequently, perform scale distribution analysis based on the warning sensitive area to generate a scale distribution group to record the spatial distribution characteristics of the sensitive area, and at the same time generate a continuous fault mapping set through the fault characteristic analysis of the warning sensitive area to mark the boundary and internal structure changes of the sensitive area.

[0100] For example, the abnormal level areas shown in the level index map are {Region A, Region B}, and the recorded stable time range in the periodic flat section is {20 seconds to 30 seconds}; after the fusion analysis, the warning sensitive area is {Region A: 25 seconds to 30 seconds}, and the warning steady state area is {Region B: 20 seconds to 25 seconds}; the content of the scale distribution group generated through scale distribution analysis is {sensitive area scale: 0.4, 0.6}; the continuous fault mapping set generated through fault analysis shows that the boundary intensity is {0.3, 0.5}.

[0101] Perform regional coding and structural compression on the scale distribution group and the continuous fault mapping set to generate a steady-state offset path. Synchronize the steady-state offset path with the warning steady-state region on the time axis and perform cross-over analysis on the jump segments to generate a reliability evaluation curve.

[0102] By performing regional coding on the scale distribution group and the continuous fault mapping set, encode and identify the boundaries of sensitive regions and fault characteristics, and optimize the storage of data through a compression algorithm to generate a steady-state offset path, which is used to describe the stability change trend of the module operation state. Specifically, perform synchronous alignment analysis on the steady-state offset path and the warning steady-state region on the time axis to ensure the consistency of their time characteristics, and at the same time perform cross-over analysis on the jump segments to reveal the dynamic abnormal impacts during the module operation. Finally, generate a reliability evaluation curve based on the results of the cross-over analysis to quantify the reliability levels of the module under different operation states.

[0103] For example, the sensitive region coding in the scale distribution group is {Region A: 0.4, Region B: 0.6}, and the fault identification in the continuous fault mapping set is {Fault 1: 0.3, Fault 2: 0.5}; after regional coding and structural compression, the steady-state offset path records the stability change intensity as {0.8, 0.9}; the results of the cross-over analysis of the jump segments after time axis synchronous alignment show that the dynamic abnormal impact intensity is {Region A: 0.7, Region B: 0.6}; the content of the generated reliability evaluation curve is {Time period 1: High reliability, Time period 2: Medium reliability}.

[0104] In this embodiment, by collecting the real-time working current data, transient temperature rise data, and input / output voltage change data of the high-power charging module under different load conditions, ambient temperature conditions, and grid fluctuation conditions, a multi-dimensional data sequence is formed, and it is normalized and feature-extracted to construct a working state composite data block, an environmental response data group, and a power supply disturbance data sequence. By means of time window segmentation, normalization processing, moving average, and dynamic discrete rate calculation, a behavioral thermal response cross-spectrum diagram and a transient anomaly trigger sequence are generated, and principal component mapping is performed on them to obtain a normalized state set and a standardized sequence set. In the fault trend evolution analysis stage, based on the standardized sequence set, a multi-dimensional fault trend evolution map and a transient anomaly distribution map are generated, and through the node influence factor matrix and the cross-drift flag, the abnormal tendency area and the slow-varying continuous area are identified to extract the key evolution segments. In the process of disturbance characteristic analysis, through parameter perturbation simulation, cumulative offset evaluation, and interactive clustering, an inducement influence intensity matrix and a trend deviation factor sequence are generated, and using a multi-scale sliding window and dynamic coefficient of variation adjustment, an abnormal focus index is finally extracted. In the evaluation stage, through the abnormal focus index, the fluctuation frequency of each time segment is calculated and the abnormal level is ranked, the early warning sensitive area and the early warning steady state area are divided, and based on regional scale analysis and structure compression, a steady state offset path and a reliability evaluation curve are generated. Through the comprehensive multi-dimensional data processing, map construction, and dynamic analysis technology, a comprehensive, accurate, and efficient solution is provided for the reliability evaluation of the high-power charging module.

[0105] Embodiment 3: As Figure 2 shown, the present application also provides a reliability evaluation system 10 for a high-power charging module, including an acquisition module 11, an analysis module 12, an adjustment module 13, and a generation module 14.

[0106] The acquisition module 11 is mainly used to acquire the multi-dimensional data sequence of the high-power charging module in multiple operating states, and preprocess the multi-dimensional data sequence to obtain a standardized sequence set.

[0107] The analysis module 12 is mainly used to construct a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generate a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, and perform collaborative analysis on the node influence factor matrix to obtain the key evolution segments.

[0108] The adjustment module 13 is mainly used to perform parameter perturbation analysis on the key evolution segments to obtain an inducement influence intensity matrix, perform cumulative deviation analysis on the inducement influence intensity matrix to obtain a trend deviation factor, perform sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor to obtain an abnormal focus index.

[0109] The generation module 14 is mainly used to generate a warning sensitive area and a warning steady state area by using the anomaly focus index, perform scale fusion on the warning sensitive area to obtain a steady state deviation path, perform time consistency matching on the steady state deviation path and the warning steady state area, and generate a reliability evaluation curve.

[0110] In this embodiment, the acquisition module 11 collects the multi-dimensional data sequences of the high-power charging module in multiple operating states in real time, and generates a standardized sequence set through preprocessing methods such as normalization and feature extraction for the multi-dimensional data sequences to ensure the integrity and consistency of the data. The analysis module 12 constructs a multi-dimensional fault trend evolution map and a transient anomaly distribution map through the standardized sequence set to comprehensively reveal the dynamic trend of fault evolution and the concentrated distribution of transient anomalies during the operation of the module. Through the collaborative analysis of the node influence factor matrix, the key evolution segments are accurately extracted, providing basic support for further fault cause analysis. The adjustment module 13 generates an influence intensity matrix of causes through parameter perturbation analysis of the key evolution segments, and generates a trend deviation factor and an anomaly focus index through cumulative deviation analysis and dynamic adjustment methods to achieve in-depth mining of fault characteristics and accurate positioning of sensitive areas. The generation module 14 divides the operating state of the high-power charging module by using the anomaly focus index, generates a warning sensitive area and a warning steady state area, constructs a steady state deviation path through scale fusion and time consistency matching, and finally generates a reliability evaluation curve. Through the cooperation of multiple modules, a comprehensive evaluation of the operating reliability of the high-power charging module is realized, potential fault trends are accurately identified, providing a scientific basis for module design optimization and fault prediction, and effectively improving the reliability and stability of the system.

[0111] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the foregoing Embodiment 1 and will not be repeated here.

[0112] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those who are familiar with this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that the technical content disclosed by the present invention can cover.

[0113] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A reliability evaluation method for a high-power charging module, characterized in that Including: Obtain the multi-dimensional data sequence of the high-power charging module in multiple operating states, and preprocess the multi-dimensional data sequence to obtain a standardized sequence set; Construct a multi-dimensional fault trend evolution map and a transient anomaly distribution map based on the standardized sequence set, generate a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient anomaly distribution map, and perform collaborative analysis on the node influence factor matrix to obtain a key evolution segment; Perform parameter perturbation analysis on the key evolution segment to obtain an inducement influence intensity matrix, perform cumulative deviation analysis on the inducement influence intensity matrix to obtain a trend deviation factor, perform sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor to obtain an anomaly focus index; Generate a warning sensitive area and a warning steady state area using the anomaly focus index, perform scale fusion on the warning sensitive area to obtain a steady state offset path, and perform time consistency matching on the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

2. The reliability evaluation method of the high-power charging module according to claim 1, wherein The step of obtaining the multi-dimensional data sequence of the high-power charging module in multiple operating states and preprocessing the multi-dimensional data sequence to obtain a standardized sequence set includes: Collect the real-time working current data, transient temperature rise data, and input and output voltage change data of the high-power charging module under different load conditions, environmental temperature conditions, and grid fluctuation conditions respectively to obtain a multi-dimensional data sequence, where the multi-dimensional data sequence includes a working state composite data block, an environmental response data group, and a power supply perturbation data sequence; Perform time window segmentation and amplitude normalization processing on the working state composite data block to obtain an operation behavior sequence, perform temperature segment moving average and dynamic discretization rate calculation on the environmental response data group to obtain a thermal stability index sequence, and perform multi-channel alignment analysis on the operation behavior sequence and the thermal stability index sequence to obtain a behavior thermal response cross-spectrum diagram; Perform perturbation segmentation and frequency statistics on the power supply perturbation data sequence to obtain a transient anomaly trigger sequence, perform principal component mapping on the behavior thermal response cross-spectrum diagram and the transient anomaly trigger sequence to obtain a normalized state set, and perform time series encoding on the normalized state set to obtain a standardized sequence set.

3. The reliability evaluation method of the high-power charging module according to claim 2, wherein The step of performing perturbation segmentation and frequency statistics on the power supply perturbation data sequence to obtain a transient anomaly trigger sequence, performing principal component mapping on the behavior thermal response cross-spectrum diagram and the transient anomaly trigger sequence to obtain a normalized state set, and performing time series encoding on the normalized state set to obtain a standardized sequence set includes: Identify boundary conditions and extract event labels from the power supply perturbation data sequence to generate a perturbation event index table and a perturbation peak distribution sequence, perform periodic analysis and frequency statistics on the perturbation peak distribution sequence to generate a frequency band feature mapping diagram; Perform cross-checking on the frequency band feature mapping diagram and the perturbation event index table to obtain a transient anomaly trigger sequence, perform continuity verification on the transient anomaly trigger sequence, and generate an anomaly pattern annotation vector based on the verification result; Perform principal component mapping on the abnormal pattern annotation vector and the behavior thermal response cross-spectrum diagram to obtain a joint abnormal influence diagram. Based on the joint abnormal influence diagram, extract the stable working section and the high-response fluctuation section to generate a feature uniformity matrix and an abnormal conflict diagram; Group and aggregate the feature uniformity matrix and the abnormal conflict diagram to obtain a normalized state set, and perform time-series index coding on the normalized state set to obtain a standardized sequence set.

4. The reliability evaluation method of the high-power charging module according to claim 1, wherein The steps of constructing a multi-dimensional fault trend evolution diagram and a transient abnormal distribution diagram based on the standardized sequence set, generating a node influence factor matrix based on the multi-dimensional fault trend evolution diagram and the transient abnormal distribution diagram, and performing collaborative analysis on the node influence factor matrix to obtain a key evolution section include: Construct a trend change tensor using the change rates between different dimensions in the standardized sequence set, and unfold the trend change tensor along the time dimension to obtain a multi-dimensional fault trend evolution diagram and a transient abnormal distribution diagram. Aggregate the adjacent node correlations of the multi-dimensional fault trend evolution diagram to obtain a trend clustering path set; Identify the incremental mutation region based on the transient abnormal distribution diagram to generate an abnormal dense region annotation map, and perform associated projection on the trend clustering path set and the abnormal dense region annotation map to generate a node influence factor matrix; Identify the node group with a change mutation rate greater than the adjacent interval according to the node influence factor matrix to obtain a cross-drift flag, and perform fuzzy density filtering and spatial connectivity analysis on the abnormal dense region annotation map and the cross-drift flag to obtain an abnormal tendency region and a slow-varying continuous region; Perform section matching and boundary collaborative analysis on the abnormal tendency region and the slow-varying continuous region to obtain a key evolution section.

5. The reliability evaluation method of the high-power charging module according to claim 4, wherein The steps of identifying the node group with a change mutation rate greater than the adjacent interval according to the node influence factor matrix to obtain a cross-drift flag, and performing fuzzy density filtering and spatial connectivity analysis on the abnormal dense region annotation map and the cross-drift flag to obtain an abnormal tendency region and a slow-varying continuous region include: Perform differential calculation and window clustering analysis on the change rates of each node in the node influence factor matrix to obtain a fusion node set, and identify local change rate abnormal transition points from the fusion node set to construct a time transition spectrum diagram and a structure discontinuity matrix; Extract the mutation dense region and the interference boundary segment using the time transition spectrum diagram to construct a jump region coding map, and perform drift consistency voting analysis on the structure discontinuity matrix and the jump region coding map to obtain a cross-drift flag; Perform fuzzy density fitting on the abnormal dense region annotation map and the cross-drift flag to obtain a jump continuity map and an intensity curvature distribution, and perform region morphology filtering and boundary simplification reconstruction on the jump continuity map to obtain an abnormal tendency region and a trend expansion candidate segment; Perform spatial connectivity analysis and curvature coupling fusion on the intensity curvature distribution and the trend expansion candidate segment to obtain a slow-varying continuous region.

6. The reliability evaluation method of the high-power charging module according to claim 1, characterized in that The step of performing parameter perturbation analysis on the key evolution segment to obtain an inducement influence intensity matrix, performing cumulative deviation analysis on the inducement influence intensity matrix to obtain a trend deviation factor, and performing sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor to obtain an abnormal focusing index includes: Extracting a high-heat response section, a current rebound characteristic section, and a voltage instability slope section based on the key evolution segment to construct a perturbation characteristic input set, and performing parameter perturbation simulation on the perturbation characteristic input set to obtain a response offset matrix and a perturbation sensitivity distribution map; Performing cross-section cumulative offset evaluation on the response offset matrix to obtain a response fluctuation factor set, and performing interactive clustering analysis on the perturbation sensitivity distribution map and the response fluctuation factor set to generate an inducement influence intensity matrix; Performing multi-dimensional time window weighted superposition on the inducement influence intensity matrix to obtain an offset density spectrogram, and calculating the local deviation mean and weighted variation rate within each window of the offset density spectrogram to generate a trend deviation factor sequence; Performing multi-scale sliding window statistics and dynamic coefficient of variation adjustment on the trend deviation factor sequence to obtain a focused abnormal intensity map and a jump breakpoint distribution, and performing joint threshold fusion on the focused abnormal intensity map and the jump breakpoint distribution to obtain an abnormal focusing index.

7. The reliability evaluation method of the high-power charging module according to claim 1, characterized in that The step of generating a warning sensitive area and a warning steady state area using the abnormal focusing index, performing scale fusion on the warning sensitive area to obtain a steady state offset path, and performing time consistency matching on the steady state offset path and the warning steady state area to generate a reliability evaluation curve includes: Calculating the fluctuation frequency and abnormal level ranking of each time segment of the high-power charging module using the abnormal focusing index to obtain a local peak aggregation area and a periodic flat section, and performing regional convolution and scale transformation on the local peak aggregation area to generate a level index map; Performing boundary fusion on the level index map and the periodic flat section to obtain a warning sensitive area and a warning steady state area, and generating a scale distribution group and a continuous fault mapping set based on the warning sensitive area; Performing region coding and structure compression on the scale distribution group and the continuous fault mapping set to generate a steady state offset path, and performing time axis synchronization alignment and jump segment cross-over analysis on the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

8. A reliability evaluation system for a high-power charging module, characterized in that, Including: An acquisition module for acquiring a multi-dimensional data sequence of the high-power charging module in multiple operating states and preprocessing the multi-dimensional data sequence to obtain a standardized sequence set; An analysis module for constructing a multi-dimensional fault trend evolution map and a transient abnormal distribution map based on the standardized sequence set, generating a node influence factor matrix based on the multi-dimensional fault trend evolution map and the transient abnormal distribution map, and performing collaborative analysis on the node influence factor matrix to obtain a key evolution segment; An adjustment module is used to perform parameter perturbation analysis on the key evolution segment to obtain an inducement influence intensity matrix, perform cumulative deviation analysis on the inducement influence intensity matrix to obtain a trend deviation factor, and perform sliding window statistics and dynamic adjustment of the coefficient of variation on the trend deviation factor to obtain an anomaly focus index; A generation module is used to generate a warning sensitive area and a warning steady state area by using the anomaly focus index, perform scale fusion on the warning sensitive area to obtain a steady state offset path, and perform time consistency matching on the steady state offset path and the warning steady state area to generate a reliability evaluation curve.

Images