Capacitor health state evaluation method and device, storage medium and computer device

Abstract

This application discloses a method, apparatus, storage medium, and computer equipment for assessing the health status of capacitors, belonging to the field of power module technology. Based on the operating data of the power module, thermal analysis is performed to determine a first parameter estimation set and a first quality metric; based on the operating data of the power module, electrical analysis is performed to determine a second parameter estimation set and a second quality metric; based on the first and second quality metrics generated by the thermal and electrical analyses respectively, the first and second parameter estimation sets are adaptively weighted and fused to obtain a high-confidence unified parameter estimation set, and a health status index is generated accordingly. When any measurement method is subjected to specific interference leading to a quality degradation, its weight in the final fusion result is automatically and non-linearly reduced, thereby avoiding the impact of interference on the assessment result, and realizing online, low-disturbance, high-precision, and highly robust assessment of the health status of energy storage capacitors.

Description

Technical Field

[0001] This application relates to the field of power module technology, specifically to capacitor health status assessment methods, devices, storage media, and computer equipment. Background Technology

[0002] Energy storage capacitors in power modules, especially film capacitors or aluminum electrolytic capacitors on the DC bus, are core passive components in power electronic converters. In critical areas such as new energy grid-connected inverters, electric vehicle drive systems, data center uninterruptible power supplies (UPS), and industrial frequency converters, they play a crucial role in stabilizing bus voltage, absorbing high-frequency ripple, and providing instantaneous high power. Therefore, the health status of energy storage capacitors directly determines the performance, reliability, and even safety of the entire power conversion system. Currently, methods for assessing the health status of energy storage capacitors struggle to accurately identify early, minor increases in equivalent series resistance (ESR) online with low disturbance, limiting the long-term reliability of energy storage capacitors in power modules. Summary of the Invention

[0003] This application provides a method, apparatus, storage medium, and computer equipment for assessing the health status of capacitors, aiming to solve the problems of low online assessment accuracy and susceptibility to operating conditions and environmental disturbances in existing energy storage capacitor health status assessment methods.

[0004] Firstly, a method for assessing the health status of energy storage capacitors in power modules is provided, including the following steps:

[0005] Acquire the operating data of the power module, which includes a set of baseline parameters; Based on the operational data, thermal analysis is performed to determine a first set of parameter estimates and a first mass metric. The first set of parameter estimates includes thermal estimates of the equivalent series resistance. Based on the operational data, an electrical analysis is performed to determine the second parameter estimation set and the second quality metric. The second parameter estimation set includes electrical estimates of capacitance and equivalent series resistance. Based on the first quality metric and the second quality metric, a weighted fusion process is performed on the first parameter estimation set and the second parameter estimation set to determine the parameter estimation set. Based on the comparison results between the parameter estimation set and the baseline parameter set, a health status index of the power module is generated by mapping.

[0006] In some embodiments, a weighted fusion process is performed on the first parameter estimation set and the second parameter estimation set based on a first quality metric and a second quality metric to determine the parameter estimation set, including: By utilizing nonlinear mapping relationships, the first and second quality metrics are transformed into a set of fusion weights; Based on the fusion weight set, an optimization algorithm based on a robust loss function is used to adaptively weight the first parameter estimation set and the second parameter estimation set to determine the parameter estimation set.

[0007] In some embodiments, performing a thermal analysis includes: An envelope excitation with a preset envelope frequency is injected into the power module. The envelope excitation is used to modulate the effective value of the ripple current to generate periodic power loss. Acquire the capacitor housing temperature response and current signal of the power module under envelope excitation; Phase-sensitive demodulation with a preset envelope frequency is performed on the temperature response of the capacitor casing to extract the thermal response amplitude and thermal response phase; Based on the thermal response amplitude, thermal response phase, and effective value change of the current signal, the thermal estimate of the equivalent series resistance in the first parameter estimation set is determined.

[0008] In some embodiments, before injecting an envelope excitation with a preset envelope frequency into the power module, the method further includes: Based on the prior interval of thermal time constant in the baseline parameter set, an envelope frequency scanning matching strategy is adopted to screen out a candidate set of envelope frequencies that are sensitive to thermal response. Based on power disturbance limits, control stability and electromagnetic compatibility constraints, the value of the preset envelope frequency is determined from the candidate envelope frequency set, and the upper limit of the envelope amplitude is determined. An envelope excitation with a preset envelope frequency is generated and injected by superimposing an envelope less than or equal to a frequency threshold at the zero average power position of the modulation duty cycle, or by implementing phase micro-jitter at the carrier layer.

[0009] In some embodiments, performing phase-sensitive demodulation of the capacitor case temperature response with reference to a preset envelope frequency includes: Based on the capacitor case temperature and the ambient reference temperature in the operating data, a temperature difference signal sequence is constructed to suppress common-mode temperature drift. Generate a reference orthogonal basis signal with the same preset envelope frequency; The thermal response amplitude and phase are obtained by performing correlation integration on the temperature difference signal sequence and the reference orthogonal basis signal and then demodulating them.

[0010] In some embodiments, determining the thermal estimate of the equivalent series resistance in the first parameter estimation set based on the thermal response amplitude, thermal response phase, and effective value change of the current signal includes: The equivalent thermal time constant is calculated based on the functional relationship between the thermal response phase and the preset envelope frequency. By combining the equivalent thermal time constant, thermal response amplitude, and thermal resistance parameters in the baseline parameter set, the injected power envelope amplitude is solved in reverse. The thermal estimate of the equivalent series resistance is determined based on the amplitude of the injected power envelope, the effective value change of the current signal, and the waveform coefficient.

[0011] In some embodiments, performing thermal analysis further includes a dual-frequency thermal excitation self-calibration process, specifically including: At the first envelope frequency, excitation injection, phase-sensitive demodulation, and parameter acquisition are performed to obtain the first set of thermal responses; At the second envelope frequency, excitation injection, phase-sensitive demodulation, and parameter acquisition are performed to obtain the second set of thermal responses; Based on the first set of thermal responses and the second set of thermal responses, the thermal estimate of the equivalent series resistance is determined.

[0012] In some embodiments, determining a thermal estimate of the equivalent series resistance based on a first set of thermal responses and a second set of thermal responses includes: Based on the first set of thermal responses and the second set of thermal responses, a system of simultaneous equations is constructed, which includes equivalent thermal resistance and equivalent series resistance as common unknowns. Based on the first and second sets of thermal responses, solve the simultaneous equations to determine the thermal estimate of the equivalent series resistance.

[0013] In some embodiments, performing electrical analysis includes: A composite electrical signal disturbance is injected into the power module once. The composite electrical signal includes a first identification frequency point and a second identification frequency point. The frequency range of the composite electrical signal is 1kHz to 10kHz. Acquire the voltage and current responses of the power module under complex electrical signal disturbances; Parametric spectrum estimation is performed on the voltage and current responses to obtain complex voltages and complex currents at the first and second identification frequency points, thus forming the response spectrum; Based on the equivalent impedance model and response spectrum, a system of dual-frequency complex equations is constructed, and the second parameter estimate set is obtained by solving the system.

[0014] In some embodiments, determining the first identification frequency point and the second identification frequency point includes: Based on the equivalent impedance model, sensitivity analysis is performed on the transfer function of the equivalent impedance model with respect to capacitance and equivalent series resistance, respectively, to identify multiple frequency points with different responses to changes in the parameters of the two, and to form a suggested set of identification frequency points. Two frequency points from the suggested identification frequency point set that maximize the condition number of solving the dual-frequency complex equation system and suppress intermodulation interference are selected as the first identification frequency point and the second identification frequency point, respectively.

[0015] In some embodiments, the second parameter estimate set is obtained by solving, including: Condition number evaluation is performed on the two-frequency complex equation system to generate numerical stability evaluation results; Based on the numerical stability assessment results, physical boundary constraints are applied to the dual-frequency complex equation system, where both the capacitance value and the equivalent series resistance value are greater than zero. Under physical boundary constraints, the least squares method is used to solve the dual-frequency complex equation system, identify the high-frequency estimate of the capacitance value and the high-frequency cross estimate of the equivalent series resistance, and thus obtain the second parameter estimate set.

[0016] In some embodiments, determining a first quality metric and a second quality metric includes: The thermal response amplitude, thermal response phase and thermal channel signal-to-noise ratio obtained from thermal analysis are tested for model consistency, and a first quality metric characterizing the confidence level of thermal analysis is generated. Amplitude-phase residual analysis is performed on the response spectrum obtained from the electrical analysis and the second parameter estimation set to generate a second quality metric characterizing the consistency of the electrical analysis spectrum.

[0017] In some embodiments, after determining the parameter estimation set, the method further includes: Substitute the parameter estimation set back into the thermal analysis model and the electrical analysis model for consistency testing, and calculate the thermoelectric consistency residual; If the thermoelectric consistency residual is lower than the preset consistency threshold and the health status index is in the stable assessment range, a baseline update allow flag is generated.

[0018] In some embodiments, after generating the baseline update permission flag, the process includes: In response to the baseline update allow flag being true, a slow update algorithm with a guard factor is adopted, and the baseline parameter set is iteratively corrected using the parameter estimation set to obtain the updated baseline parameter set.

[0019] In some embodiments, acquiring the operating data of the power module includes: The original sampling data set, system configuration parameter set, and baseline parameter set of the power module are obtained. After clock calibration and dimensional unification processing, the aligned data set and effective configuration parameter set are obtained. Based on the aligned dataset and the effective set of configuration parameters, thermal and electrical analyses are performed.

[0020] Secondly, a capacitor health status assessment device is also provided, comprising: The data acquisition module is used to acquire the operating data of the power module, which includes a set of baseline parameters. The thermal analysis module is used to perform thermal analysis based on the running data to determine a first set of parameter estimates and a first mass metric. The first set of parameter estimates includes thermal estimates of the equivalent series resistance. The electrical analysis module is used to perform electrical analysis based on the running data, determine the second parameter estimation set and the second quality metric, the second parameter estimation set including electrical estimates of capacitance and equivalent series resistance; The fusion processing module is used to perform weighted fusion processing on the first parameter estimation set and the second parameter estimation set based on the first quality metric and the second quality metric to determine the parameter estimation set. The health assessment module is used to map and generate health status indicators based on the comparison results between the parameter estimation set and the baseline parameter set.

[0021] Thirdly, a storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described capacitor health status assessment method.

[0022] Fourthly, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the capacitor health status assessment method described above.

[0023] This application embodiment acquires the operating data of the power module, performs thermal analysis based on the operating data to determine a first parameter estimation set and a first quality metric; performs electrical analysis based on the operating data to determine a second parameter estimation set and a second quality metric; and performs adaptive weighted fusion processing on the first and second parameter estimation sets generated by the thermal and electrical analyses respectively to obtain a unified parameter estimation set with high confidence, and generates a health status index accordingly. When any measurement method is subjected to specific interference that leads to a quality degradation, its weight in the final fusion result will be automatically and non-linearly reduced, thereby avoiding the impact of interference on the overall evaluation result, and realizing online, low-disturbance, high-precision and high-robust evaluation of the health status of energy storage capacitors. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating the capacitor health status assessment method provided in an exemplary embodiment of this application; Figure 2This is a schematic flowchart of the thermal analysis method provided in an exemplary embodiment of this application; Figure 3 This is a schematic flowchart illustrating the method for performing electrical analysis provided in an exemplary embodiment of this application; Figure 4 This is a schematic diagram of the structure of the capacitor health status assessment device provided in an exemplary embodiment of this application; Figure 5 This is a schematic diagram of the internal structure of a computer device provided in an exemplary embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0028] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not exclude the applicability to or configuration to devices performing additional tasks or steps. Additionally, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more of the stated conditions or values ​​may in practice be based on additional conditions or values ​​beyond those stated.

[0029] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0030] The applicant notes that energy storage capacitors in power modules, especially film capacitors or aluminum electrolytic capacitors on the DC bus, are core passive components in power electronic converters. They play crucial roles in key areas such as new energy grid-connected inverters, electric vehicle drive systems, data center uninterruptible power supplies (UPS), and industrial frequency converters, stabilizing bus voltage, absorbing high-frequency ripple, and providing instantaneous high power. Therefore, the health status of energy storage capacitors directly determines the performance, reliability, and even safety of the entire power conversion system. Statistical data shows that capacitors are among the components with the highest failure rate in power electronic systems; their failure often leads to system shutdowns or even catastrophic accidents. Therefore, developing online, accurate, and early warning methods for capacitor health status assessment is of engineering and economic value for ensuring the reliable operation of critical equipment, reducing the total lifecycle maintenance cost, and realizing a shift from post-failure repair to predictive maintenance.

[0031] Currently, research on methods for assessing the health status of energy storage capacitors has made some progress. Offline measurement is the most traditional and direct method, which involves removing the capacitor after system shutdown and using specialized equipment such as an LCR (Inductance, Capacitance, and Resistance) tester to accurately measure its capacitance (C) and equivalent series resistance (ESR). Online assessment methods avoid shutdown and removal, and are mainly divided into two categories: electrical parameter identification methods and thermal analysis methods. Online electrical methods typically identify the values ​​of C and ESR by injecting electrical signals of a specific frequency into the system or by utilizing the voltage and current ripple generated during system operation, based on the equivalent circuit model of the capacitor, and calculating the impedance response of the system at a specific frequency. Thermal analysis methods are based on the physical principle that the power loss inside the capacitor is mainly generated by ESR, which is ultimately dissipated as heat. Therefore, by monitoring the temperature change of the capacitor casing, the health status of its internal ESR can be reflected to some extent. For example, the surface temperature of a capacitor can be obtained using an infrared thermal imager or a patch temperature sensor, and a correlation model between its temperature rise and ESR under a specific ripple current can be established.

[0032] However, online evaluation methods still face several technical challenges when confronted with increasingly stringent industrial application requirements. These challenges mainly focus on the accurate detection capability of early-stage faults, the robustness of measurements under complex operating conditions, and the self-consistency verification of evaluation results.

[0033] First, online assessment methods struggle to accurately identify early, minute increases in ESR online with minimal disturbance. These minute increases in ESR are often the earliest precursors to capacitive thermal runaway and eventual failure, with changes potentially only on the milliohm scale. For online electrical methods, such minute ESR changes are easily masked by noise in the complex electromagnetic environment of the system. Furthermore, at high frequencies, stray parameters such as bus parasitic inductance significantly affect impedance measurements, leading to insufficient ESR identification accuracy. Traditional thermal methods, such as those based on steady-state temperature rise analysis, face even greater challenges. The power loss changes caused by minute ESR increases are extremely weak, and the resulting temperature rise signal (potentially on the order of 0.01 K) is completely masked by ambient temperature fluctuations, temperature drift caused by load changes, and variations in cooling conditions (such as wind speed), resulting in an extremely low signal-to-noise ratio and making early warning impossible.

[0034] Secondly, single-physical-dimensional measurement methods lack inherent robustness and a self-consistent verification mechanism for results. Evaluations relying solely on electrical methods may be biased due to unmodeled resonant points, electromagnetic interference, or model mismatch. Similarly, relying solely on thermal methods can lead to significant errors due to variations in thermal resistance R_th. Factors such as dust accumulation on the heatsink surface, aging thermal grease, and fan speed reduction can all alter R_th. In such cases, even if the ESR has not deteriorated, the measured temperature may be abnormal, leading to misjudgments. Related technologies lack a mechanism for deeply integrating electrical and thermal measurements, making it impossible to cross-validate measurement results from two independent physical principles, thus failing to provide highly confident and systematic health assessment conclusions.

[0035] Finally, many online evaluation methods heavily rely on a set of precise offline calibration parameters that may change throughout the device's lifecycle. Whether it's the bus parasitic inductance L in the electrical model or the equivalent thermal resistance R_th and heat capacity C_th in the thermal model, these are typically assumed to be fixed baseline values. This assumption of static models is fragile in practical applications; slow parameter drift accumulates errors, ultimately leading to a deterioration in the accuracy of the health assessment model over time. Related technical solutions generally lack the ability to perform online self-calibration or adaptive correction of these critical model parameters, limiting their long-term reliability.

[0036] Example 1: This application provides an overall framework for a method to assess the health status of energy storage capacitors in power modules. Specifically, this application provides a method for assessing the health status of energy storage capacitors in power modules, such as... Figure 1 As shown, it includes the following steps: Step S100: Obtain the operating data of the power module, which includes a set of baseline parameters.

[0037] In some embodiments, the baseline parameter set is the capacitance parameters in a healthy state (e.g., at the time of manufacture), and the capacitance parameters in a healthy state include the capacitance value baseline, ESR baseline, thermal time constant baseline, etc.

[0038] Step S200: Based on the running data, perform thermal analysis to determine the first parameter estimation set and the first quality measure. The first parameter estimation set includes thermal estimates of the equivalent series resistance.

[0039] In some embodiments, the ESR of a capacitor is a key factor in the generation of Joule heating; even a small change in ESR can directly lead to a change in the heat generation power under the same ripple current. The value of the ESR can be deduced by accurately measuring this minute temperature rise caused by the change in ESR. In this embodiment, the thermal analysis is a low-frequency detection method based on the lock-in amplification principle, and the frequency of the detection excitation is preferably 0.2-1Hz. This low-frequency range is chosen because the thermal system response of a capacitor is slow, and low-frequency excitation ensures sufficient heat transfer to the casing surface for detection by the temperature sensor, while simultaneously being completely separated from high-frequency electrical noise (such as switching ripple) in frequency.

[0040] Step S300: Based on the operating data, perform electrical analysis to determine the second parameter estimation set and the second quality metric. The second parameter estimation set includes electrical estimates of capacitance and equivalent series resistance.

[0041] In some embodiments, thermal analysis is sensitive to equivalent series resistance but cannot directly measure the capacitance value C, while electrical analysis can simultaneously identify both capacitance and equivalent series resistance, thus complementing thermal analysis. In this application embodiment, electrical analysis is a high-frequency injection method that injects kHz-level composite signal electrical disturbances into the system and measures the system's voltage and current responses (i.e., impedance responses) at these frequencies to solve for the equivalent circuit model parameters of the capacitor. The advantages of high-frequency analysis are its speed, high signal-to-noise ratio, and greater sensitivity to changes in capacitance.

[0042] Step S400: Based on the first quality metric and the second quality metric, perform weighted fusion processing on the first parameter estimation set and the second parameter estimation set to determine the parameter estimation set.

[0043] In some embodiments, thermal analysis and electrical analysis each generate a first quality metric and a second quality metric to evaluate the reliability of their estimation results. For example, a low signal-to-noise ratio in thermal analysis results in a low first quality metric, and poor spectral consistency in electrical analysis results in a low second quality metric. The purpose of the weighted fusion process is to adaptively weight the first and second parameter estimation sets based on the first and second quality metrics. It is understood that a result with a higher quality metric means it was less affected by interference in the measurement and is more reliable; therefore, it will be given a higher weight and contribute more to the final parameter estimation set. This fusion method effectively combines the advantages of both methods, resulting in more robust and accurate evaluation results than a single method.

[0044] Step S500: Based on the comparison results between the parameter estimation set and the baseline parameter set, generate a health status index for the power module.

[0045] In some embodiments, the parameter estimation set consists of physical quantities (e.g., ESR of 2.5 mΩ and capacitance of 980 μF), while the health status index is a normalized, intuitive score. Specifically, this can be achieved by calculating the deviation rate of the current parameters relative to baseline parameters, for example, an increase in ESR of 25% or a decrease in capacitance of 2%. These deviation rates are then weighted and aggregated according to their importance to the health status, ultimately mapping them to a health status index. This setup enables online, low-disturbance, high-precision, and highly robust assessment of the health status of energy storage capacitors.

[0046] In some embodiments, the health status indicator can be an intuitive score, such as 0-100. Further, embodiments of this application use health level labels to qualitatively classify the health status indicator; for example, a health status indicator of 90-100 is considered healthy, 70-90 is considered concerning, and below 70 is considered deteriorating, thereby providing clear and easy-to-understand decision support for operations and maintenance personnel.

[0047] In some embodiments, obtaining the operating data of the power module in step S100 includes: Step S110: Obtain the original sampling data set, system configuration parameter set, and baseline parameter set of the power module. After clock calibration and dimension unification processing, obtain the aligned data set and effective configuration parameter set.

[0048] In some embodiments, a raw sampled data set, a system configuration parameter set, and a baseline parameter set are acquired. After clock calibration and dimensional unification processing, an aligned data set and a valid configuration parameter set are obtained. The raw sampled data set includes time-series data such as bus voltage, bus current, capacitor case temperature, and ambient reference temperature. The system configuration parameter set contains information such as the power module model and rated parameters. The baseline parameter set consists of capacitor parameters in a healthy state (e.g., at the time of manufacture), such as the capacitance value baseline, equivalent series resistance (ESR) baseline, and thermal time constant baseline.

[0049] Specifically, this step involves preprocessing the aforementioned data and parameter set. First, source auditing is completed by comparing the version number and checksum of each data source, and the integrity and boundary range of the baseline parameters are verified to ensure the validity of the data source.

[0050] Subsequently, clock calibration and trigger alignment are performed on sensor data at different sampling rates. For example, this is achieved by uniformly upsampling to the highest sampling rate or by using interpolation algorithms to achieve precise alignment of the time axis, thereby obtaining data that is strictly synchronized in time. At the same time, the units and dimensions of all parameters are standardized, for example, all temperature units are standardized to degrees Celsius (°C) and resistance units are standardized to milliohms (mΩ), in order to generate a parameter set with unified dimensions.

[0051] Furthermore, to ensure the reliability of subsequent analyses, the system operating conditions can be assessed before execution. For example, by monitoring bus voltage fluctuation rate and current harmonic content, an identification window can be established where the system is in a statically stable or lightly loaded state with low electromagnetic interference. Subsequent analysis steps are only triggered within this window. Finally, the processed data is packaged into an aligned dataset and a set of valid configuration parameters for use in subsequent steps.

[0052] Step S120: Perform thermal and electrical analyses based on the aligned dataset and the set of valid configuration parameters.

[0053] In some embodiments, thermal analysis and electrical analysis are performed based on the aligned data set and the effective configuration parameter set, respectively. The thermal analysis and electrical analysis are complementary. It is understood that thermal analysis is sensitive to ESR but cannot directly measure the capacitance value C, while electrical analysis can identify both C and ESR.

[0054] Example 2: This application's embodiments illustrate the specific implementation of the thermal analysis steps in Embodiment 1.

[0055] In some embodiments, please refer to Figure 2 The thermal analysis performed in step S200 includes: Step S210: Inject an envelope excitation with a preset envelope frequency into the power module. The envelope excitation is used to modulate the effective value of the ripple current to generate periodic power loss.

[0056] In some embodiments, a low-frequency power envelope excitation with a preset envelope frequency is injected into the power module. This excitation generates small periodic power losses by modulating the effective value of the ripple current, and simultaneously acquires an aligned set of data including the temperature response of the capacitor casing.

[0057] Step S220: Obtain the capacitor housing temperature response and current signal of the power module under envelope excitation.

[0058] Step S230: Perform phase-sensitive demodulation on the temperature response of the capacitor casing with a preset envelope frequency as a reference, and extract the thermal response amplitude and thermal response phase.

[0059] In some embodiments, the acquired capacitor casing temperature response is subjected to phase-sensitive demodulation with reference to the envelope frequency, the thermal response amplitude and thermal response phase are separated and extracted, and the thermal channel signal-to-noise ratio, the change in the effective value of the current and the waveform coefficients are estimated.

[0060] Step S240: Based on the thermal response amplitude, thermal response phase, and effective value change of the current signal, determine the thermal estimate of the equivalent series resistance in the first parameter estimation set.

[0061] In some embodiments, the equivalent series resistance thermal estimates constituting the first parameter estimation set are obtained by inversion calculation using a first-order thermal model. After obtaining the thermal response amplitude A_T and phase φ, inversion calculation is performed based on a first-order thermal model of capacitance.

[0062] By injecting a weak power disturbance and using phase-locked demodulation technology to extract the extremely weak temperature response synchronized with the disturbance from the strong noise background, the ESR can be accurately inverted. A reliable estimate of the ESR can be obtained without disassembling the device and with low disturbance.

[0063] In some embodiments, before injecting envelope excitation with a preset envelope frequency into the power module in step S210, the method further includes: Step S201: Based on the prior interval of thermal time constant in the baseline parameter set, an envelope frequency scanning matching strategy is adopted to screen out a candidate set of envelope frequencies that are sensitive to thermal response.

[0064] In some embodiments, a preset envelope frequency is first determined, and an envelope frequency scanning matching strategy is adopted. Based on the prior interval of the thermal time constant in the baseline parameters (e.g., 10s-40s), a candidate set of envelope frequencies sensitive to thermal response is selected. The range of the candidate set of envelope frequencies is usually 0.2-1Hz.

[0065] Step S202: Based on power disturbance limits, control stability and electromagnetic compatibility constraints, determine the value of the preset envelope frequency from the candidate envelope frequency set, and determine the upper limit of the envelope amplitude.

[0066] In some embodiments, the final envelope frequency f_env is determined from the candidate envelope frequency set by comprehensively considering power disturbance limits, such as additional injected power being less than 0.5W, and joint constraints on control system stability and electromagnetic compatibility. For example, the envelope frequency f_env is 0.5Hz.

[0067] Step S203: Generate and inject an envelope excitation with a preset envelope frequency by superimposing an envelope less than or equal to a frequency threshold at the zero average power position of the modulation duty cycle, or by implementing phase micro-jitter at the carrier layer.

[0068] In some embodiments, the excitation is generated by superimposing a small envelope with a frequency of f_env at the zero average power position of the modulation duty cycle, or by implementing phase jitter at the zero vector injection time in the carrier layer, while keeping the system average power constant, to generate an excitation that causes the effective value of the power module ripple current I_rms to change slightly, slowly, and periodically.

[0069] In some embodiments, step S230, performing phase-sensitive demodulation of the capacitor casing temperature response with reference to a preset envelope frequency, includes: Step S231: Based on the capacitor case temperature and the ambient reference temperature in the operating data, construct a temperature difference signal sequence to suppress common-mode temperature drift.

[0070] In some embodiments, in order to suppress common-mode temperature drift caused by ambient temperature fluctuations, a temperature difference signal time series ΔT(t) is constructed, i.e., ΔT(t) = T_c(t) - T_ref(t); Where T_c(t) is the time series of the capacitor casing temperature, and T_ref(t) is the temperature time series of an environmental reference point far from the capacitor.

[0071] Step S232: Generate a reference orthogonal basis signal that is consistent with the preset envelope frequency.

[0072] In some embodiments, two reference orthogonal basis signals consistent with the envelope frequency f_env are generated, namely sin(2π×f_env×t) and cos(2π×f_env×t).

[0073] Step S233: Perform correlation integration on the temperature difference signal sequence and the reference orthogonal basis signal, and demodulate to obtain the thermal response amplitude and thermal response phase.

[0074] In some embodiments, by performing correlation integration operations (i.e., multiplying and integrating over one or more periods) on the temperature difference signal ΔT(t) and the two reference orthogonal basis signals respectively, the in-phase component X and the orthogonal component Y can be demodulated.

[0075] Therefore, the thermal response amplitude A_T and the thermal response phase φ can be calculated: A_T=sqrt(X²+Y²)φ=atan2(Y,X); Where A_T is the temperature rise fluctuation amplitude of the capacitor casing at the frequency f_env, in K or °C; φ is the phase delay of this temperature rise relative to the power excitation, in radians. It should be noted that the temperature response here is characterized by two real numbers, amplitude and phase, not complex numbers.

[0076] In other embodiments, the signal-to-noise ratio of the thermal channel is obtained by performing noise variance estimation on the time series of the temperature difference signal, and the effective value change of the current and the waveform coefficients are calculated by performing narrowband power spectrum estimation on the current signal in the aligned dataset. Specifically, by performing narrowband power spectrum estimation on I_rms(t), the amplitude ΔI_rms of the effective value change of the current and the waveform coefficient k can be calculated. The waveform coefficient k is related to the shape of the current envelope; for example, for a sine wave envelope, k=1 / 2; for a triangular wave envelope, k=1 / 3; and for a square wave envelope, k=1.

[0077] In some embodiments, determining the thermal estimate of the equivalent series resistance in the first parameter estimation set based on the thermal response amplitude, thermal response phase, and effective value change of the current signal in step S240 includes: Step S241: Calculate the equivalent thermal time constant based on the functional relationship between the thermal response phase and the preset envelope frequency.

[0078] In some embodiments, the equivalent series resistance thermal estimates constituting the first parameter estimation set are obtained by inversion calculation using a first-order thermal model. After obtaining the thermal response amplitude A_T and phase φ, inversion calculation is performed based on a first-order thermal model of capacitance. First, the equivalent thermal time constant τ is obtained analytically by utilizing the arctangent function relationship between the thermal response phase and the envelope frequency: τ = tan(-φ) / (2π×f_env); Where τ is the equivalent thermal time constant in seconds (s); φ is the measured thermal response phase in radians; and f_env is the excitation envelope frequency in Hertz (Hz).

[0079] Step S242: Combine the equivalent thermal time constant, thermal response amplitude, and thermal resistance parameters in the baseline parameter set to solve for the injected power envelope amplitude in reverse.

[0080] In some embodiments, the amplitude P_1 of the periodic thermal power envelope generated by the ESR can be estimated. Based on the amplitude-frequency relationship A_T = (P_1 × R_th) / sqrt(1 + (ωτ)²) and the phase-frequency relationship cos(φ) = 1 / sqrt(1 + (ωτ)²), where ω = 2π × f_env, P_1 = A_T / (R_th × cos(φ)) can be derived. In some embodiments, the equivalent thermal resistance R_th can be used as a baseline parameter input. In other embodiments, R_th can be calculated from τ and the baseline equivalent heat capacity C_th, i.e., R_th = τ / C_th.

[0081] Step S243: Determine the thermal estimate of the equivalent series resistance based on the injected power envelope amplitude, the effective value change of the current signal, and the waveform coefficient.

[0082] In some embodiments, based on the definition of thermal power P_1=k×ESR×(ΔI_rms)², the thermal estimate of the equivalent series resistance ESR_th can be obtained by inverse solution: ESR_th=P_1 / (k×(ΔI_rms)²)=(A_T / (R_th×cos(φ))) / (k×(ΔI_rms)²); Where ESR_th is the final thermal estimate of the equivalent series resistance, in ohms (Ω); A_T is the measured thermal response amplitude; R_th is the equivalent thermal resistance; φ is the measured thermal response phase; k is the waveform coefficient; and ΔI_rms is the amplitude of the change in the effective value of the current caused by the injected excitation.

[0083] By following the steps above, a reliable estimate of ESR can be obtained without disassembling the device and with minimal disturbance.

[0084] In other embodiments, this application provides a more detailed and robust, safer implementation of the thermal analysis steps compared to Embodiment 2, with the specific steps as follows: Step one, injecting a low-frequency power envelope excitation with a preset envelope frequency into the power module, specifically includes: based on the prior interval of the thermal time constant in the baseline parameter set, using an envelope frequency scanning matching strategy to select a candidate set of envelope frequencies sensitive to thermal response; from the candidate set of envelope frequencies, comprehensively considering the joint constraints of power disturbance limits, control stability, and electromagnetic compatibility, determining the final value of the preset envelope frequency, and calculating the upper limit of the allowable envelope amplitude; while keeping the average power constant, generating and injecting the low-frequency power envelope excitation by superimposing an envelope less than or equal to the frequency threshold at the zero average power position of the modulation duty cycle, or by implementing phase micro-jitter at the carrier layer.

[0085] In this embodiment, the selected candidate envelope frequencies are preferably in the range of 0.3 Hz to 0.8 Hz. Furthermore, to ensure sufficient time energy accumulation of the measurement signal for high signal-to-noise ratio demodulation, the duration of each excitation injection is preferably 40 to 80 seconds.

[0086] Furthermore, calculating the upper limit of the allowable envelope amplitude is a computational process that requires comprehensive consideration of various engineering constraints. Specifically, the system needs to calculate the largest possible envelope amplitude that simultaneously satisfies the following conditions: Power disturbance limit: Ensure that the additional heat loss caused by this amplitude is less than a safe threshold, such as 0.5W, to avoid unnecessary thermal stress on the capacitor itself.

[0087] Control stability: Ensure that the bus voltage fluctuation caused by the periodic change of the effective value of the ripple current does not exceed ±2% of the rated DC bus voltage. This is to avoid triggering the overvoltage or undervoltage protection of the system or interfering with the control stability of the system.

[0088] Electromagnetic compatibility (EMC): Ensures that the low-frequency envelope itself and its low-order harmonic components do not overlap or cross over with other control frequencies in the system (such as the bandwidth of the power control loop or synchronous phase-locked loop) to prevent control resonance or interference.

[0089] In this embodiment, before injecting the stimulus, the system also performs a mandatory feasibility self-check and boundary condition verification as a safety safeguard. Specifically: First, check the reading of the capacitor housing temperature sensor to confirm that it is within the preset safe operating range, such as -20℃ to 85℃.

[0090] Confirm that the fluctuation rate of the current DC bus voltage is less than a stable threshold, such as 5%, to ensure that the measurement is not performed when the system operating conditions change drastically.

[0091] If the power module is equipped with an active air-cooling system, the system will verify that the air-cooling system is in normal working condition to ensure that the heat dissipation conditions meet expectations.

[0092] If any of the above conditions are not met, this measurement task will be postponed and the specific reason will be recorded. It will be triggered again when the conditions are met.

[0093] Step 2: Perform phase-sensitive demodulation with the envelope frequency as a reference on the collected capacitor shell temperature response, separate and extract the thermal response amplitude and thermal response phase, and estimate the thermal channel signal-to-noise ratio, the change in the effective value of the current, and the waveform coefficients.

[0094] In this embodiment, the generation process of the first quality metric (i.e., the phase-locked thermal quality metric) is specified. The signal-to-noise ratio of the thermal channel is estimated by performing spectral analysis on the temperature difference signal ΔT(t) before demodulation, estimating the average power spectral density of the noise, i.e., the noise variance, in the frequency band far from the excitation frequency f_env, and then comparing it with the signal energy at the excitation frequency.

[0095] In some embodiments, the generation of the first quality metric further includes an algorithm for model consistency testing and phase residual scoring. Specifically, after obtaining an estimate of the thermal time constant τ using the method of Embodiment 2, the theoretical thermal response phase at that τ value can be predicted according to the relationship φ_predicted = -arctan(2π × f_env × τ). Then, the difference between the theoretically predicted phase and the actual measured phase φ is calculated, i.e., the phase residual. The smaller the residual, the higher the agreement between the measurement result and the first-order thermal model, and the better the model consistency. Finally, this phase residual score is weighted and combined with the aforementioned thermal channel signal-to-noise ratio to form a more reliable and comprehensive first quality metric Q_th.

[0096] Step 3: Through inversion calculation using a first-order thermal model, the equivalent series resistance thermal estimates constituting the first parameter estimation set are obtained.

[0097] In this embodiment, the calculation process of this step is basically the same as in Embodiment 2. However, through the more rigorous pre-checks and quality assessments in Steps 1 and 2, the measurement data (A_T,φ) input to this step has higher reliability, thus ensuring the accuracy of the final inverted ESR_th. Furthermore, the entire injection and demodulation process includes interlocking start / stop and failure rollback strategies. An independent monitoring task monitors the bus safety parameters in real time. Once any abnormality is detected during the measurement process (e.g., current or voltage exceeding limits), the excitation injection is immediately stopped, and the system control state is safely rolled back to the state before the measurement, thus providing hardware-level safety assurance for the entire testing process.

[0098] Example 3: This application describes the specific implementation of the electrical analysis steps in Embodiment 1. The specific steps are as follows: In some embodiments, please refer to Figure 3 The electrical analysis performed in step S300 includes: Step S310: Inject a composite electrical signal disturbance into the power module once. The composite electrical signal includes a first identification frequency and a second identification frequency. The frequency range of the composite electrical signal is 1kHz to 10kHz.

[0099] In some embodiments, a composite electrical signal disturbance including a first identification frequency and a second identification frequency is injected into the power module in a single injection, and an aligned data set including voltage and current responses is recorded simultaneously. It is understood that this electrical analysis method is a high-frequency supplement to the low-frequency thermal analysis method of Embodiment 2. To implement this step, it is first necessary to determine the first identification frequency f_1 and the second identification frequency f_2.

[0100] In some embodiments, determining the first identification frequency point and the second identification frequency point in step S310 includes: Step S311: Based on the equivalent impedance model, perform sensitivity analysis on the transfer function of the equivalent impedance model with respect to the capacitance value and the equivalent series resistance respectively, in order to identify multiple frequency points with different responses to changes in the two parameters, and form a suggested set of identification frequency points.

[0101] In some embodiments, the determination of the first and second identification frequency points follows a sensitivity separation principle. Specifically, based on an equivalent impedance model incorporating bus parasitic inductance, sensitivity analyses are performed on the transfer function of this model with respect to capacitance C and ESR, respectively. For example, it can be found that in lower kHz frequency bands (e.g., 1-2 kHz), impedance is more sensitive to changes in capacitance C; while in higher kHz frequency bands (e.g., 5-10 kHz), impedance is more significantly affected by ESR. Accordingly, multiple frequency points with significantly different responses to changes in these two parameters can be identified, forming a suggested set of identification frequency points.

[0102] Step S312: Select two frequency points from the suggested identification frequency point set that maximize the condition number of solving the dual-frequency complex equation system and suppress intermodulation interference, and use them as the first identification frequency point and the second identification frequency point, respectively.

[0103] In some embodiments, a pair of frequency points that maximize the condition number of subsequent equation solving and suppress intermodulation interference are selected from the set of suggested frequency points as the final f_1 and f_2. For example, f_1 can be selected as 1.5kHz and f_2 can be selected as 8kHz.

[0104] Step S320: Obtain the voltage and current responses of the power module under composite electrical signal disturbances.

[0105] In some embodiments, after selecting a first identification frequency and a second identification frequency, the sinusoidal signals of these two frequencies are synthesized with a specific amplitude ratio to form a composite electrical signal disturbance. This composite electrical signal disturbance can be injected into the DC bus of the power module by superimposing it on the modulation wave of the inverter or by directly injecting it into the command of the current control loop. Simultaneously with the injection, the voltage and current response time series on the bus are synchronously recorded at a sampling rate (e.g., 100 kS / s) much higher than the injection frequency.

[0106] Step S330: Perform parameterized spectrum estimation on the voltage response and current response to obtain complex voltage and complex current at the first identification frequency and the second identification frequency, thus forming the response spectrum.

[0107] In some embodiments, a narrowband discrete Fourier transform or Goertzel algorithm is used to perform parameterized spectral estimation on the recorded voltage and current responses, obtaining complex voltages and complex currents at the first and second identification frequency points, which together constitute the response spectrum. In this embodiment, since only the responses at two specific frequency points f_1 and f_2 are of interest, using the Goertzel algorithm is a more efficient choice than performing a full Fast Fourier Transform (FFT). By inputting the recorded voltage and current time series into the Goertzel algorithm, the complex responses at these two frequency points can be efficiently calculated. The result is four complex quantities: V_c(f_1), I_c(f_1), V_c(f_2), I_c(f_2); Where V_c(f) is the complex voltage at frequency f and I_c(f) is the complex current at frequency f, together forming the response spectrum required for subsequent solutions.

[0108] Step S340: Based on the equivalent impedance model and response spectrum, construct a system of dual-frequency complex equations and solve them to obtain the second parameter estimate set.

[0109] In some embodiments, based on an equivalent impedance model that includes the parasitic inductance of the bus, a dual-frequency complex equation system is constructed using the response spectrum. By solving the equation system by least squares, the high-frequency estimated values ​​of the capacitance and the high-frequency cross-estimated values ​​of the equivalent series resistance that constitute the second parameter estimation set are simultaneously identified.

[0110] First, an equivalent impedance model Z(jω) describing the characteristics of the DC bus is established. A simplified equivalent impedance model can be expressed as: Z(jω)=R_L+jωL+(1 / (jωC+1 / ESR)); After sorting, we get: Z(jω)=R_L+jωL+(ESR / (1+jωC×ESR)); Where Z(jω) is the complex impedance at frequency ω; ω is the angular frequency, ω=2πf; R_L is the equivalent resistance of the bus line; L is the parasitic inductance of the bus; C is the capacitance to be determined; ESR is the equivalent series resistance to be determined; j is the imaginary unit.

[0111] Using the response spectrum obtained in step two, the experimentally measured impedances at the two frequency points can be calculated: Z_m(jω_1)=V_c(f_1) / I_c(f_1) and Z_m(jω_2)=V_c(f_2) / I_c(f_2).

[0112] Thus, a system of equations is constructed, consisting of two complex equations and two unknowns (C and ESR): {Z_m(jω_1)=R_L+jω_1×L+(ESR / (1+jω_1×C×ESR))}; {Z_m(jω_2)=R_L+jω_2×L+(ESR / (1+jω_2×C×ESR))}; In this system of equations, R_L and L can typically be input as known parameters through offline calibration.

[0113] With this setup, by injecting a composite electrical signal containing two different high-frequency components in a single injection, the capacitance value C and the equivalent series resistance ESR can be solved synchronously and accurately by utilizing the different impedance responses of the system at the two frequency points.

[0114] In some embodiments, the solution obtained in step S340 to obtain a second parameter estimation set includes: Step S341: Perform condition number evaluation on the two-frequency complex equation system and generate numerical stability evaluation results.

[0115] Step S342: Based on the numerical stability assessment results, apply physical boundary constraints to the dual-frequency complex equation system, where both the capacitance value and the equivalent series resistance value are greater than zero. Step S343: Under physical boundary constraints, the least squares method is used to solve the dual-frequency complex equation system to identify the high-frequency estimated value of the capacitance and the high-frequency cross-estimate of the equivalent series resistance, thereby obtaining the second parameter estimate set.

[0116] In some embodiments, a condition number evaluation is performed on the system of equations before solving to predict the stability of the numerical solution. An excessively large condition number may indicate poor selection of the two frequency points, making the solution highly sensitive to noise. By imposing positive definite physical boundary constraints on the two-frequency complex equations, where both capacitance and equivalent series resistance are greater than zero, and performing nonlinear least squares solutions under these constraints, the high-frequency estimate of capacitance C_e and the high-frequency cross-estimate of ESR ESR_e can be obtained simultaneously.

[0117] In some embodiments, to ensure the reliability of the results, this application further includes a verification step. Specifically, a spectral consistency index can be obtained by calculating the amplitude-phase residual and coherence coefficient of the solution results. If the index is lower than a preset threshold, the system can determine that the current electrical measurement is invalid and automatically trigger a retest to avoid using interfered or unstable results in subsequent fusion steps.

[0118] Example 4: This application illustrates how to systematically and robustly integrate the thermal analysis results obtained in Example 2 with the electrical analysis results obtained in Example 3, and finally output an intuitive health status.

[0119] The step of determining the first quality metric in step S200 includes: The thermal response amplitude, thermal response phase, and thermal channel signal-to-noise ratio obtained from thermal analysis are tested for model consistency, and a first quality metric characterizing the confidence level of thermal analysis is generated.

[0120] In some embodiments, the first quality metric is a lock-in thermal quality metric. Its generation steps are as follows: Model consistency testing and phase residual scoring are performed on the thermal response amplitude A_T, thermal response phase φ, and thermal channel signal-to-noise ratio obtained from thermal analysis. Specifically, the identified thermal time constant τ and thermal resistance R_th can be substituted back into the first-order thermal model to predict the theoretical amplitude and phase under this excitation, and compared with the actually measured A_T and φ. The smaller the difference, the better the model consistency. Combined with the measured thermal channel signal-to-noise ratio, a lock-in thermal quality metric Q_th between 0 and 1 is obtained through a comprehensive score.

[0121] The step of determining the second quality metric in step S300 includes: Amplitude-phase residual analysis is performed on the response spectrum obtained from the electrical analysis and the second parameter estimation set to generate a second quality metric characterizing the consistency of the electrical analysis spectrum.

[0122] In some embodiments, the second quality metric is a spectral consistency index. Its generation involves synthesizing the response spectrum obtained from electrical analysis with the identified C_e and ESR_e. This process combines least-squares amplitude-phase residual analysis, coherence coefficient assessment of voltage and current signals at the excitation frequency, and calculation of the utilization rate of injected energy. Similarly, after these indices are synthesized, a spectral consistency index Q_el between 0 and 1 is obtained.

[0123] In some embodiments, step S400, which involves performing a weighted fusion process on the first parameter estimation set and the second parameter estimation set based on the first quality metric and the second quality metric to determine the parameter estimation set, includes: Step S410: Using a nonlinear mapping relationship, the first quality metric and the second quality metric are transformed into a set of fusion weights.

[0124] Step S420: Based on the fusion weight set, an optimization algorithm based on a robust loss function is used to adaptively weight the first parameter estimation set and the second parameter estimation set to determine the parameter estimation set.

[0125] In some embodiments, a nonlinear mapping relationship is used to transform the first quality metric and the second quality metric into a set of fusion weights. An optimization solution based on a Huber-type robust loss function is then employed to implement adaptive weighting, resulting in a unified parameter estimation set. After obtaining the quality metrics Q_th and Q_el, a nonlinear mapping (e.g., the Sigmoid function) is used to transform them into normalized fusion weights w_th and w_el, such that w_th + w_el = 1.

[0126] Next, a unified set of parameter estimates is obtained by solving a robust optimization problem. Taking ESR fusion as an example, the goal is to find a fused ESR_fused that minimizes the weighted loss function. This loss function adopts a Huber-type robust loss function, which has the following specific form: Loss(ESR_fused)=w_th×ρ(ESR_th-ESR_fused)+w_el×ρ(ESR_e-ESR_fused); Where ρ(x) is the Huber loss function: ρ(x) = {x² / 2, if |x| ≤ δ; δ(|x| - δ / 2), if |x| > δ}; x is the residual between the estimated value and the fused value; δ is a robust parameter used to distinguish normal values ​​and outliers in the residuals, and its value can be dynamically set according to the standard deviation σ of historical data, for example, δ = 1.345σ. The ESR_fused can be obtained by minimizing the loss function. The advantage of the Huber loss function is that when the residuals are small (|x| ≤ δ), its effect is similar to the L2 norm (squared loss), which is sensitive to the data; when the residuals are large (|x| > δ), its effect is similar to the L1 norm (absolute value loss), which can effectively suppress the excessive influence of outliers (such as a method being severely inaccurate in this measurement) on the fusion result. Similar processing is performed on other parameters (such as C, τ, etc.) to finally obtain the parameter estimation set.

[0127] After obtaining a reliable set of parameter estimates, the standardized offset of each parameter is calculated by comparing it with the baseline parameter set. For example, the offset of ESR is: d_ESR=(ESR_fused-ESR_baseline) / ESR_baseline; Then, based on the contribution of each parameter to the overall health of the capacitor, these offsets are weighted and aggregated to form a single health index (HI): HI=1-(w_ESR×d_ESR+w_C×d_C); Here, HI represents the health status indicator; d_ESR and d_C are the offsets for ESR and capacitance, respectively; w_ESR and w_C are the corresponding weights, and w_ESR + w_C = 1. For example, w_ESR = 0.6 and w_C = 0.4 can be set, because an increase in ESR is usually a more critical precursor to failure. Finally, based on the range of HI values, health level labels such as Healthy, Attention, and Deterioration are output.

[0128] Example 5: This embodiment illustrates how to achieve adaptive and closed-loop updates of system baseline parameters after obtaining reliable health status assessment results, and how to conduct systematic result management.

[0129] In some embodiments, after determining the parameter estimation set in step S420, the following steps are further included: Step S421: Substitute the parameter estimation set back into the thermal analysis model and the electrical analysis model for consistency verification, and calculate the thermoelectric consistency residual.

[0130] In some embodiments, the unified parameter estimation set is substituted back into the models of thermal and electrical analysis for consistency testing, quantifying and generating a thermoelectric consistency residual characterizing the degree of self-consistency in thermoelectric measurements. This is the final verification of the fusion result. For example, substituting the fused ESR_fused into the first-order thermal model, combined with the known current excitation ΔI_rms, the theoretical temperature rise amplitude A_T_predicted can be predicted positively. Then, this predicted value is compared with the actual measured thermal response amplitude A_T_measured, and the normalized difference is the thermoelectric consistency residual Res_TE. Res_TE=|A_T_predicted-A_T_measured| / A_T_measured.

[0131] A small Res_TE (e.g., less than 5%) indicates that the ESR estimates obtained through two completely independent physical principles (thermal and electrical) are highly consistent, greatly enhancing the credibility of the final results.

[0132] Optionally, if the thermoelectric consistency residual Res_TE exceeds a preset threshold, the system can determine that the result of this fusion assessment is at risk and trigger an anomaly handling strategy, such as temporarily not updating the health status, or marking the data point and suggesting manual review.

[0133] Step S422: If the thermoelectric consistency residual is lower than the preset consistency threshold and the health status index is in the stable assessment range, then a baseline update allow flag is generated.

[0134] In some embodiments, a stable evaluation interval is identified based on health status indicators, and it is determined whether the thermoelectric consistency residual is below a preset consistency threshold, thereby generating an update permission flag. It is understood that the baseline parameters are the benchmark for all subsequent evaluations, and their updates must be extremely cautious to prevent a single, noise-contaminated measurement result from having a long-term adverse effect on the entire system. Therefore, this embodiment sets up a strict update admission mechanism. Specifically, the system continuously monitors the historical sequence of the health status indicator HI obtained from Embodiment 4. Only when HI remains stable over multiple consecutive evaluation periods (e.g., 5 consecutive times), i.e., the fluctuation range is less than a small threshold (e.g., ±2%), will the system identify this period as a stable evaluation interval. This aims to ensure that the data used for updates does not come from a fault development period with drastic parameter changes. Simultaneously, even within a stable evaluation interval, the thermoelectric consistency residual Res_TE generated by each evaluation must be below a preset consistency threshold (e.g., Res_TE < 5%). This ensures that the data used for updates is highly consistent and reliable. The system will only generate a true update flag and authorize subsequent baseline update operations when both conditions are met: it is in a stable evaluation range and thermoelectric consistency is high.

[0135] Following the generation of baseline update allow tags in step S422, the following is also included: Step S423: In response to the baseline update allow flag being true, a slow update algorithm with a guard factor is used to iteratively correct the baseline parameter set using the parameter estimation set to obtain the updated baseline parameter set.

[0136] In some embodiments, when the update permission flag is true, a slow update algorithm with a guard factor is employed to iteratively correct the baseline parameter set using a unified parameter estimation set. After obtaining update permission, to avoid impacting the baseline due to unknown biases that may exist in a single measurement, this embodiment employs a slow update algorithm with a guard factor. The core of this algorithm is to ensure that changes to the baseline are gradual and smooth.

[0137] Specifically, the method for calculating the guardian factor α can be as follows: α=min(α_max,σ_est / |P_fused-P_baseline|); Where α is the guardian factor for this update, or learning rate; α_max is the upper limit of the guardian factor, which is an empirical constant, such as 0.1, used to limit the maximum update magnitude in a single update; σ_est is the standard deviation of the parameter estimate, which can be obtained by estimating the parameter covariance in Example 4, and characterizes the uncertainty of the current estimate; P_fused is the parameter estimate obtained by this fusion (e.g., ESR_fused); P_baseline is the baseline value currently used.

[0138] The formula for updating the baseline parameters is: P_baseline_new=(1-α)×P_baseline_old+α×P_fused; Where P_baseline_new is the new baseline value after the update; P_baseline_old is the old baseline value before the update. It can be seen that the smaller the protection factor α, the smaller the impact of the current measurement result on the baseline, and the smoother the update process.

[0139] Step S424: Generate a structured diagnostic report and archive and index the data.

[0140] In some embodiments, after the assessment and update are completed, the system generates a structured diagnostic report. This report includes not only the final health level label, but also preferably the unified parameter estimation set obtained from this assessment, a snapshot of the effective configuration parameters used for this assessment, thermoelectric consistency residuals, and confidence intervals estimated based on parameter covariance. In particular, the report includes timestamps of key operating conditions during the measurement, enabling the reproduction and tracing of problems.

[0141] Meanwhile, to enable long-term asset health management and algorithm model iteration, the system archives and stores the diagnostic report, the updated baseline parameter set, and the corresponding original data fragments. A retrieval index is created for this archived record, with index fields including device ID, assessment time, health level, and the flag that triggered the update. This systematic archiving and indexing mechanism provides a solid data foundation for subsequent large-scale statistical analysis of device group health status, tracing specific failure modes, and training degradation models.

[0142] Example 6: This application embodiment is an alternative to the phase-locked thermal method of Embodiment 2, and the specific method is as follows: In some embodiments, performing thermal analysis in step S200 further includes a dual-frequency thermal excitation self-calibration process, specifically including: Step S250: At the first envelope frequency, perform excitation injection, phase-sensitive demodulation and parameter acquisition to obtain the first set of thermal responses.

[0143] In some embodiments, at a first envelope frequency, excitation injection, phase-sensitive demodulation, and parameter acquisition are performed to obtain a first set of thermal responses, which include a first thermal response amplitude and a first thermal response phase. Specifically, a first envelope frequency f_env,1 (e.g., 0.3Hz) is first selected, and a low-frequency power envelope excitation of the corresponding frequency is injected into the power module. Through phase-sensitive demodulation, a first set of thermal responses at this frequency is obtained, denoted as (A_T1,φ_1).

[0144] Step S260: At the second envelope frequency, perform excitation injection, phase-sensitive demodulation and parameter acquisition to obtain the second set of thermal responses.

[0145] In some embodiments, the aforementioned operation is repeated at a second envelope frequency different from the first envelope frequency to obtain a second set of thermal responses, which includes the second thermal response amplitude and the second thermal response phase. Specifically, the envelope frequency is switched to a second envelope frequency f_env,2 (e.g., 0.8Hz) that is significantly different from the first envelope frequency, and all operations in step one are repeated. A second set of thermal responses is obtained at this frequency, denoted as (A_T2,φ_2).

[0146] Step S270: Based on the first set of thermal responses and the second set of thermal responses, determine the thermal estimate of the equivalent series resistance.

[0147] In some embodiments, the equivalent series resistance thermal estimate is obtained by jointly processing the first group and the second group of thermal responses.

[0148] In some embodiments, determining the thermal estimate of the equivalent series resistance based on the first set of thermal responses and the second set of thermal responses in step S270 includes: Step S271: Based on the first set of thermal responses and the second set of thermal responses, construct a system of simultaneous equations that includes equivalent thermal resistance and equivalent series resistance as common unknowns.

[0149] In some embodiments, ESR and R_th are treated as common unknowns, and a system of simultaneous equations is constructed and solved. Based on the first-order thermal model, the following relationship exists: A_T=|H_th(jω)|×P_1=(R_th / sqrt(1+(ωτ)²))×(k×(ΔI_rms)²×ESR)φ=-arctan(ωτ); Among them, τ=R_th×C_th, ω=2π×f_env.

[0150] Substituting the two sets of measurement results, we obtain a system containing four correlation equations. More directly, we can construct the following system of simultaneous equations: A_T1×sqrt(1+tan²(φ_1))=k×(ΔI_rms)²×ESR×R_th A_T2×sqrt(1+tan²(φ_2))=k×(ΔI_rms)²×ESR×R_th; tan(-φ_1) / (2π×f_env,1)=R_th×C_th tan(-φ_2) / (2π×f_env,2)=R_th×C_th.

[0151] Step S272: Solve the simultaneous equations based on the first set of thermal responses and the second set of thermal responses to determine the thermal estimate of the equivalent series resistance.

[0152] In some embodiments, by solving the above set of equations, the values ​​of ESR and R_th can be identified simultaneously without prior knowledge of the baseline values ​​of the equivalent heat capacity C_th or the equivalent thermal resistance R_th. For example, the value of R_th × C_th (i.e., the thermal time constant τ) can be solved first from the latter two equations and cross-validated. Then, τ can be substituted into the system composed of the first two equations to separate and solve for ESR and R_th.

[0153] This setup, by measuring at two different envelope frequencies, achieves self-calibration of the equivalent thermal resistance R_th, thereby eliminating the method's dependence on baseline parameters of thermal resistance or thermal capacity and significantly improving measurement robustness under varying environmental conditions (such as wind speed and ambient temperature). This dual-frequency self-calibration method, by adding an extra measurement, achieves complete decoupling of the system's thermal baseline parameters, making it particularly suitable for applications where heat dissipation conditions are prone to change or thermal parameters are unknown, offering higher accuracy and environmental adaptability.

[0154] Example 7: This application describes an embodiment that, based on a precise assessment of the current health status, further predicts the remaining useful life (RUL) of the energy storage capacitor. This enables this application to possess an early warning system with predictive maintenance capabilities.

[0155] Step S600: Obtain historical health status sequence and stress data.

[0156] In some embodiments, RUL prediction is based on the accumulation of historical data. This step reads the historical sequence of a unified parameter estimation set (including ESR_fused, C_fused, etc.) aligned with timestamps for a specific power module over a past period from the archived database of Embodiment 5, as well as the historical sequence of the corresponding health status index HI. Simultaneously, using the operating condition segment index in the archived records, the operating stress data corresponding to each historical data point is extracted, such as the average operating temperature of the capacitor, the RMS value of the ripple current, and the DC bus voltage.

[0157] Step S610: Construct and fit a stress-related degradation model.

[0158] In some embodiments, a mathematical model is established to describe how key capacitor parameters (such as ESR) degrade over time and under operating stress. For example, an empirical or semi-empirical model based on physical mechanisms can be used. A widely accepted capacitor lifetime model is: L=L_0×(V / V_0)^(-n)×exp[(E_a / k_B)×(1 / T_0-1 / T)]; Where L is the expected lifespan; L_0 is the lifespan under rated conditions; V and T are the actual operating voltage and temperature, respectively; V_0 and T_0 are the rated voltage and temperature; n is the voltage stress exponent; E_a is the activation energy; and k_B is the Boltzmann constant.

[0159] Using the historical ESR_fused sequence and corresponding temperature and voltage stress data obtained in step S600, the parameters (such as E_a,n) of the above model are fitted and corrected online to obtain a personalized degradation model for the specific device.

[0160] In other implementations, purely data-driven models can be used, such as the Autoregressive Integrated Moving Average (ARIMA) model or the Long Short-Term Memory (LSTM) network that takes into account long-term dependencies, using historical parameter sequences and stress sequences as inputs and learning their degradation patterns through training.

[0161] Step S620: Perform degradation trajectory extrapolation and RUL estimation.

[0162] In some embodiments, the Remaining Life (RUL) can be predicted after obtaining a well-fitted, personalized degradation model. First, one or more failure thresholds need to be set. For example, the end of capacitor life can be defined as an ESR value reaching or exceeding 200% of its baseline value or a Health Status Index (HI) below 60. Based on a hypothetical future mission profile, such as assuming the equipment will operate under typical average conditions in the future, the degradation trajectory of ESR or HI is extrapolated forward using the fitted degradation model until it intersects with the preset failure threshold. The time span from the current moment to this intersection point is the predicted Remaining Life (RUL). This RUL estimate can be output along with a confidence interval, providing users with quantitative, forward-looking information about the future risks of the equipment. This enables a shift from reactive, post-failure maintenance to proactive, condition- and predictive maintenance, thereby maximizing equipment availability and reducing total lifecycle costs.

[0163] In some embodiments, the system may maintain a unit rule mapping table internally. For example, this table stipulates that all temperature inputs (regardless of whether the original unit is Celsius, Fahrenheit, or Kelvin) must be uniformly converted to Kelvin for physical calculations; all pressure units are uniformly converted to Pascals. When a raw parameter with a unit identifier is read, the system will query this mapping table and call the corresponding conversion function to complete the normalization.

[0164] In some embodiments, statistical methods, such as the 3-sigma principle, can be used to test the validity of baseline parameters. The system calculates the mean and standard deviation of all historical baseline values. If a newly read baseline value deviates from the mean by more than three times the standard deviation, it is identified as an outlier and automatically reverts to using the valid baseline value from the previous version.

[0165] In some embodiments, before each evaluation process begins, the system serializes the currently used set of valid configuration parameters (e.g., converts it to JSON or XML format) and calculates its hash value (such as SHA-256) as a digital signature. The serialized text of this parameter set, along with its signature and current timestamp, is stored as an immutable snapshot. Subsequent diagnostic reports are forcibly associated with the hash value of this snapshot, ensuring that any result can be precisely traced back to the original configuration it was based on, eliminating analysis confusion caused by configuration changes.

[0166] In some embodiments, to suppress intermodulation distortion between the dual-frequency injected signals, the preferred first identification frequency f1 and the second identification frequency f2 should avoid a simple integer multiple relationship. For example, frequencies that are coprime can be chosen, or their frequency intervals can be far from any known system resonant point. The amplitude ratio can be optimized based on the impedance response sensitivity at different frequencies. For example, a slightly larger injection amplitude can be allocated at the lower frequency f1, which is more sensitive to capacitance C, to obtain a higher signal-to-noise ratio.

[0167] In some embodiments, the selection of the injection channel depends on the inverter's topology and control strategy. For voltage source inverters, superimposing a small signal onto the carrier or modulated wave of a pulse-width modulation (PWM) signal is a direct and efficient method. For systems employing current closed-loop control, directly superimposing the composite disturbance signal onto the current command I_cmd is a more stable and accurate injection method. The system can automatically select the optimal injection channel based on the read configuration parameters.

[0168] In some embodiments, such as those recommended by the audit, the system will perform a rigorous security boundary check before injection: Confirm that the total injected power is less than the preset hardware safety threshold, such as 0.5W.

[0169] Spectrum analysis verifies that the frequency components of the injected signal do not overlap with the system's critical control frequencies (such as current loop bandwidth and phase-locked loop bandwidth), thus preventing control instability.

[0170] By using simulation or lookup tables, pre-evaluate the voltage or current spikes that the injected signal may cause in the worst case, ensuring that they do not trigger the hardware's overcurrent or overvoltage protection.

[0171] In some embodiments, a parameterized sigmoid function can be used to perform this nonlinear mapping. For example, the relationship between the weight w and the quality metric Q can be expressed as: w = 1 / (1 + exp(-a × (Qb))); Here, 'a' is the gain factor, which controls the steepness of the mapping curve; 'b' is the bias or center point, typically set to an acceptable median value for the quality metric (e.g., 0.8). This function allows small changes in the quality metric near the center point to cause significant changes in the weights, while in the saturation region close to 0 or 1, the changes are more gradual, aligning with engineering intuition.

[0172] In some embodiments, the system can be configured with a tiered handling strategy. For example, if the thermoelectric consistency residual Res_TE exceeds a first-level threshold (e.g., 10%), the system will automatically trigger a complete retest. If the Res_TE of three consecutive retests exceeds the threshold, the system determines that there is a persistent measurement anomaly and will trigger a second-level handling: no more measurements will be performed, but a warning log containing all current diagnostic information will be generated, the relevant data segments will be marked as pending review, and an alarm will be sent to the upper-level monitoring system.

[0173] In some embodiments, the upper limit α_max of the safeguard factor can be automatically adjusted. Within the identified stable evaluation interval, if the system continuously monitors a decreasing trend in the standard deviation σ_est of the parameter estimates, this indicates that the measurement results are becoming increasingly stable and consistent. In this case, the system can determine that the reliability of the current model is improving, and therefore the value of α_max can be moderately and slowly increased (e.g., gradually increasing from 0.05 to 0.1) to accelerate the baseline's tracking of the system's true state.

[0174] In some embodiments, the system creates a structured retrieval index for each record during archiving. This index should include at least the following fields: Device ID, precise timestamp of report generation, report unique ID, software firmware version number used during evaluation, hash signature of the associated effective configuration parameter snapshot, Health Status Indicator (HI), health level label, and a boolean flag indicating whether a baseline update was triggered. This detailed indexing mechanism enables advanced applications such as large-scale data analysis, horizontal comparison of devices of the same model, or tracing issues in specific batches.

[0175] In some embodiments, dual-frequency self-calibration is described as follows: Perform the first thermal analysis: At the first envelope frequency f1 (preferably in the range of 0.2-0.5Hz), perform a complete lock-in thermal measurement to obtain the first set of thermal responses (A_T1,φ_1).

[0176] Perform a second thermal analysis: At the second envelope frequency f2 (preferably in the range of 0.5-1Hz and significantly different from f1), perform the measurement again to obtain the second set of thermal responses (A_T2,φ_2).

[0177] Simultaneous Solution: Based on the first-order thermal model, a system of simultaneous equations is constructed with ESR and equivalent thermal resistance R_th as common unknowns. By solving this system of equations, the current values ​​of ESR and R_th can be directly analyzed without prior knowledge of the baseline values ​​of heat capacity C_th or R_th, thus eliminating the dependence on the baseline values ​​of thermal resistance.

[0178] In other embodiments, the following analysis is performed for an aluminum electrolytic capacitor rated at 1000μF / 450V.

[0179] Injected excitation: f_env=0.5Hz, effective value change of injected current ΔI_rms=5A, waveform is square wave (k=1).

[0180] Response measurement: The measured thermal response amplitude A_T = 0.08K and the thermal response phase φ = -0.35rad were obtained.

[0181] The process of step-by-step calculation to solve for the thermal time constant τ, the injected power envelope P1, the equivalent thermal resistance R_th, and the ESR_th yields intermediate and final results such as τ≈35s, R_th≈2.5K / W, and ESR_th≈2.1mΩ.

[0182] Injected excitation: f1 = 1.5 kHz, f2 = 8 kHz. The results of solving for C_e and ESR_e are given.

[0183] Set up a scenario where the thermal analysis quality metric is high, while the electrical analysis quality metric is slightly lower.

[0184] The process of weight calculation and weighted fusion is demonstrated to obtain the final fusion result, such as ESR_fused=2.05mΩ and C_fused=985μF. Compared with the baseline values ​​(e.g., ESR_baseline=2.0mΩ, C_baseline=1000μF), the final health index HI=94 and the health level label are calculated and obtained.

[0185] In some embodiments, by constructing a closed-loop evaluation system that adaptively fuses and self-verifies the results of thermal and electrical analyses, this application fundamentally solves the technical problems of poor robustness and inability to perform online self-verification for single-physical-dimensional measurement methods. Specifically, this effect is achieved through the synergistic effect of the following technical features: First, the system obtains ESR estimates from two completely independent physical domains, thermal and electrical, in parallel; second, it uses a phase-locked loop thermal quality metric and a spectral consistency index to quantify and score the real-time reliability of the two signals; finally, it innovatively employs the Huber robust fusion algorithm to dynamically weight the two results based on the quality metric score. This mechanism ensures that when either measurement method is subjected to specific interference (such as thermal analysis being affected by sudden changes in ambient temperature, or electrical analysis being affected by electromagnetic noise) leading to a quality degradation, its weight in the final fusion result is automatically and non-linearly reduced, thereby avoiding the contamination of the overall evaluation result by inferior data. Furthermore, through the calculation of thermoelectric consistency residuals, the system obtains an intrinsic scale for evaluating the reliability of the final result, achieving self-consistency verification of the evaluation result. Therefore, the beneficial effect of this application is that the accuracy and reliability of the assessment results have been significantly improved. It can always output a health status judgment with high confidence in complex and ever-changing industrial environments, and greatly reduce the risk of false alarms or omissions caused by the limitations of the measurement method itself.

[0186] In some embodiments, by creatively applying lock-in amplification technology from the field of precision physical measurement to online ESR identification of capacitors within power modules, the core technical challenge of accurately detecting early, minute ESR degradation under low-disturbance and low-signal-to-noise ratio conditions using traditional methods is solved. This achievement stems from the synergistic effect of the following technological innovations: First, by generating a low-frequency power envelope excitation through minute modulation of the PWM waveform, the ESR degradation information to be measured is transferred to a selected specific frequency point (e.g., 0.3-0.8Hz) far removed from various types of noise. Second, differential temperature measurement suppresses common-mode ambient temperature drift noise at its source. Most importantly, by performing correlation integration (i.e., phase-sensitive demodulation) on this weak temperature difference signal and a quadrature reference base signal of the same frequency, extremely narrowband filtering and huge gain amplification of the target frequency signal are achieved, enabling the effective extraction of milliKelvin (mK) level target temperature signals from background noise several times or even tens of times stronger. It achieves high-sensitivity and high-precision online monitoring of ESR, with a detection resolution far exceeding that of the traditional temperature rise method. Moreover, the injected energy throughout the process is extremely low (<1W), causing almost no disturbance to the normal operation of the system. It is especially suitable for standby or light-load conditions that are sensitive to disturbances, thus enabling the capture of the first signs of degradation in the health of capacitors, providing the possibility for truly predictive maintenance.

[0187] In some embodiments, a composite electrical disturbance identification method with single injection and dual-frequency decoupling is designed to resolve the contradiction between the long time consumption and unsuitability for rapid online diagnosis of traditional electrical identification methods (such as frequency sweeping methods). While ensuring identification speed, it improves the separation and accuracy of the two key parameters, capacitance (C) and ESR. This achievement is attributed to the deep optimization of the injection signal design and signal processing algorithm: First, through sensitivity analysis of the system model, two optimal frequency points for C and ESR identification are pre-selected, ensuring the robustness of subsequent solutions. Second, the signals at these two frequencies are composited and injected in a single operation, compressing the frequency sweeping process, which originally required several seconds or even minutes, into a snapshot measurement at the hundred-millisecond level. Finally, at the signal processing end, the Goertzel algorithm, with a computational cost far less than FFT, is used to efficiently extract the complex responses of the two target frequency points and directly construct a dual-frequency complex equation system for solution. This provides a rapid electrical identification method that perfectly complements the slow, high-precision phase-locked loop thermal method. Its speed meets the real-time requirements of online applications, and its accuracy (thanks to frequency optimization) provides high-quality cross-validation data for thermoelectric fusion. Together, they form the foundation for the efficient and reliable operation of the entire evaluation system.

[0188] In some embodiments, an adaptive baseline closed-loop update mechanism based on health status and consistency verification is introduced to address the fundamental deficiency of assessment models that rely on static, offline calibration parameters and cannot adapt to individual differences and long-term changes. This effect is achieved through a prudent and intelligent closed-loop update logic: First, the triggering conditions for the update operation are extremely strict, requiring the simultaneous fulfillment of two preconditions: stable assessment conditions and thermoelectric consistency residuals below a threshold. This mechanism ensures that only the highest quality assessment results are eligible to modify the baseline. Second, the update process is not a simple direct replacement, but rather employs a slow update algorithm with a safeguard factor, ensuring that each baseline adjustment is small and gradual, effectively preventing single measurement deviations from impacting the system benchmark. This endows the entire health assessment system with self-learning and adaptive capabilities. During use, the system can continuously and safely self-calibrate its internal model parameters (such as thermal resistance R_th), enabling it to closely track parameter drift caused by equipment aging and environmental changes, thereby maintaining the accuracy and effectiveness of the assessment throughout the entire lifespan of the equipment.

[0189] In summary, to address the issues of low accuracy and susceptibility to operating conditions and environmental disturbances in related online assessment technologies, this paper proposes a method to obtain a thermal estimate of the equivalent series resistance (ESR). This is achieved by injecting a low-frequency (e.g., 0.3-0.8Hz) power envelope excitation into the power module and employing phase-sensitive demodulation technology to accurately separate the amplitude and phase of the thermal response from the weak temperature response. Simultaneously, a composite small-signal disturbance containing two high-frequency (e.g., kHz) components is injected into the module, and the electrical estimates of the capacitance (C) and ESR are simultaneously identified based on the system impedance model. Furthermore, based on the quality metrics generated by the two methods, this application uses an optimization algorithm based on the Huber robust loss function to adaptively weight and fuse the thermal and electrical estimation results, obtaining a high-confidence unified parameter estimation set, and generating a health status index accordingly. This application achieves online, low-disturbance, high-precision, and highly robust assessment of capacitor health status.

[0190] Please see Figure 4This application also provides a capacitor health status assessment device, including a data acquisition module, a thermal analysis module, an electrical analysis module, a fusion processing module, and a health assessment module: the data acquisition module is used to acquire operating data of a power module, the operating data including a baseline parameter set; the thermal analysis module is used to perform thermal analysis based on the operating data to determine a first parameter estimation set and a first quality metric, the first parameter estimation set including a thermal estimate of the equivalent series resistance; the electrical analysis module is used to perform electrical analysis based on the operating data to determine a second parameter estimation set and a second quality metric, the second parameter estimation set including an electrical estimate of the capacitance value and the equivalent series resistance; the fusion processing module is used to perform weighted fusion processing on the first parameter estimation set and the second parameter estimation set based on the first quality metric and the second quality metric to determine a parameter estimation set; the health assessment module is used to map and generate a health status index based on the comparison result between the parameter estimation set and the baseline parameter set.

[0191] This setup acquires the operating data of the power module, performs thermal analysis based on the data to determine the first parameter estimation set and the first quality metric, and performs electrical analysis based on the data to determine the second parameter estimation set and the second quality metric. Based on the first and second quality metrics generated by the thermal and electrical analyses, the first and second parameter estimation sets are adaptively weighted and fused to obtain a unified parameter estimation set with high confidence. A health status index is then generated based on this set. When any measurement method is subjected to specific interference that leads to a decrease in quality, its weight in the final fusion result will be automatically and non-linearly reduced, thereby avoiding the impact of interference on the overall evaluation result. This achieves online, low-disturbance, high-precision, and highly robust evaluation of the health status of energy storage capacitors.

[0192] In some embodiments, this application provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 5 As shown. The computer device includes a processor, internal memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating devices and computer programs in the non-volatile storage medium. The database stores motion detection data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in any of the above embodiments of the power module energy storage capacitor health status assessment method.

[0193] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0194] In some embodiments, this application provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in any of the above embodiments of the power module energy storage capacitor health status assessment method.

[0195] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0196] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0197] The above provides a detailed description of a capacitor health status assessment method, apparatus, storage medium, and computer device provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for assessing the health status of a capacitor, characterized in that, Includes the following steps: Acquire the operating data of the power module, the operating data including a set of baseline parameters; Based on the operational data, thermal analysis is performed to determine a first parameter estimation set and a first mass metric, wherein the first parameter estimation set includes thermal estimates of the equivalent series resistance. Based on the operational data, an electrical analysis is performed to determine a second set of parameter estimates and a second quality metric. The second set of parameter estimates includes electrical estimates of capacitance and equivalent series resistance. Based on the first quality metric and the second quality metric, the first parameter estimation set and the second parameter estimation set are weighted and fused to determine the parameter estimation set; Based on the comparison results between the parameter estimation set and the baseline parameter set, a health status index of the power module is generated.

2. The method according to claim 1, characterized in that, The step of performing a weighted fusion process on the first parameter estimation set and the second parameter estimation set based on the first quality metric and the second quality metric to determine the parameter estimation set includes: By utilizing a nonlinear mapping relationship, the first quality metric and the second quality metric are transformed into a set of fusion weights; Based on the fusion weight set, an optimization algorithm based on a robust loss function is used to adaptively weight the first parameter estimation set and the second parameter estimation set to determine the parameter estimation set.

3. The method according to claim 1, characterized in that, The thermal analysis includes: An envelope excitation with a preset envelope frequency is injected into the power module. The envelope excitation is used to modulate the effective value of the ripple current to generate periodic power loss. The temperature response and current signal of the capacitor housing of the power module under the envelope excitation are obtained; The phase-sensitive demodulation of the temperature response of the capacitor casing is performed with reference to the preset envelope frequency to extract the thermal response amplitude and thermal response phase. Based on the thermal response amplitude, the thermal response phase, and the effective value change of the current signal, the thermal estimate of the equivalent series resistance in the first parameter estimation set is determined.

4. The method according to claim 3, characterized in that, Before injecting the envelope excitation with a preset envelope frequency into the power module, the method further includes: Based on the prior interval of thermal time constant in the baseline parameter set, an envelope frequency scanning matching strategy is used to screen out a candidate set of envelope frequencies that are sensitive to thermal response. Based on power disturbance limits, control stability and electromagnetic compatibility constraints, the value of the preset envelope frequency is determined from the candidate envelope frequency set, and the upper limit of the envelope amplitude is determined. An envelope excitation at the preset envelope frequency is generated and injected by superimposing an envelope less than or equal to a frequency threshold at the zero average power position of the modulation duty cycle, or by implementing phase micro-jitter at the carrier layer.

5. The method according to claim 3, characterized in that, The step of performing phase-sensitive demodulation with reference to the preset envelope frequency in response to the temperature of the capacitor casing includes: Based on the capacitor casing temperature and ambient reference temperature in the operational data, a temperature difference signal sequence is constructed to suppress common-mode temperature drift. Generate a reference orthogonal basis signal that matches the preset envelope frequency; The thermal response amplitude and thermal response phase are obtained by performing correlation integration on the temperature difference signal sequence and the reference orthogonal basis signal, and then demodulating them.

6. The method according to claim 3, characterized in that, The step of determining the thermal estimate of the equivalent series resistance in the first parameter estimation set based on the thermal response amplitude, the thermal response phase, and the effective value change of the current signal includes: The equivalent thermal time constant is calculated based on the functional relationship between the thermal response phase and the preset envelope frequency; By combining the equivalent thermal time constant, the thermal response amplitude, and the thermal resistance parameters in the baseline parameter set, the injected power envelope amplitude is solved in reverse. Based on the injected power envelope amplitude, the effective value change of the current signal, and the waveform coefficient, the thermal estimate of the equivalent series resistance is determined.

7. The method according to claim 3, characterized in that, The thermal analysis also includes a dual-frequency thermal excitation self-calibration process, specifically including: At the first envelope frequency, excitation injection, phase-sensitive demodulation, and parameter acquisition are performed to obtain the first set of thermal responses; At the second envelope frequency, excitation injection, phase-sensitive demodulation, and parameter acquisition are performed to obtain the second set of thermal responses; Based on the first set of thermal responses and the second set of thermal responses, the thermal estimate of the equivalent series resistance is determined.

8. The method according to claim 7, characterized in that, The step of determining the thermal estimate of the equivalent series resistance based on the first set of thermal responses and the second set of thermal responses includes: Based on the first set of thermal responses and the second set of thermal responses, a system of simultaneous equations is constructed, including equivalent thermal resistance and equivalent series resistance as common unknowns. Based on the first set of thermal responses and the second set of thermal responses, the simultaneous equations are solved to determine the thermal estimate of the equivalent series resistance.

9. The method according to claim 1, characterized in that, The electrical analysis includes: A composite electrical signal disturbance is injected into the power module once. The composite electrical signal includes a first identification frequency point and a second identification frequency point. The frequency range of the composite electrical signal is 1 kHz to 10 kHz. The voltage and current responses of the power module under the combined electrical signal disturbance are obtained; The voltage response and the current response are parametrically spectrum estimated to obtain complex voltage and complex current at the first identification frequency point and the second identification frequency point, thus forming the response spectrum; Based on the equivalent impedance model and the response spectrum, a system of dual-frequency complex equations is constructed, and the second parameter estimation set is obtained by solving it.

10. The method according to claim 9, characterized in that, The determination of the first identification frequency point and the second identification frequency point includes: Based on the equivalent impedance model, sensitivity analysis is performed on the transfer function of the equivalent impedance model with respect to capacitance and equivalent series resistance, respectively, to identify multiple frequency points that have different responses to changes in the two parameters, and to form a suggested set of identification frequency points. Two frequency points from the suggested identification frequency point set that maximize the solution condition number of the dual-frequency complex equation system and suppress intermodulation interference are selected as the first identification frequency point and the second identification frequency point, respectively.

11. The method according to claim 9, characterized in that, The solution yields the second parameter estimate set, including: The condition number of the two-frequency complex equation system is evaluated to generate numerical stability evaluation results. Based on the numerical stability evaluation results, physical boundary constraints are applied to the dual-frequency complex equation system, where both the capacitance value and the equivalent series resistance value are greater than zero. Under the physical boundary constraints, the least squares method is used to solve the dual-frequency complex equation system to identify the high-frequency estimate of the capacitance value and the high-frequency cross estimate of the equivalent series resistance, thereby obtaining the second parameter estimate set.

12. The method according to claim 1, characterized in that, The determination of the first quality metric and the second quality metric includes: The thermal response amplitude, thermal response phase and thermal channel signal-to-noise ratio obtained from the thermal analysis are tested for model consistency, and the first quality metric characterizing the confidence level of the thermal analysis is generated. Amplitude-phase residual analysis is performed on the response spectrum obtained from the electrical analysis and the second parameter estimation set to generate the second quality metric characterizing the consistency of the electrical analysis spectrum.

13. The method according to claim 1, characterized in that, After determining the parameter estimation set, the method further includes: The parameter estimation set is substituted back into the thermal analysis model and the electrical analysis model for consistency verification, and the thermoelectric consistency residual is calculated. If the thermoelectric consistency residual is lower than the preset consistency threshold and the health status index is in the stable evaluation range, a baseline update allow flag is generated.

14. The method according to claim 13, characterized in that, Following the generation of the baseline update allow flag, the following is included: In response to the baseline update permission being marked as true, a slow update algorithm with a guard factor is used to iteratively correct the baseline parameter set using the parameter estimation set to obtain the updated baseline parameter set.

15. The method according to claim 1, characterized in that, The acquisition of the power module's operating data includes: The original sampling data set, system configuration parameter set, and baseline parameter set of the power module are obtained. After clock calibration and dimension unification processing, the aligned data set and effective configuration parameter set are obtained. Based on the aligned data set and the effective configuration parameter set, the thermal analysis and the electrical analysis are performed.

16. A capacitor health status assessment device, characterized in that, include: The data acquisition module is used to acquire the operating data of the power module, the operating data including a set of baseline parameters; The thermal analysis module is used to perform thermal analysis based on the operating data to determine a first parameter estimation set and a first mass metric, wherein the first parameter estimation set includes thermal estimates of the equivalent series resistance. An electrical analysis module is used to perform electrical analysis based on the operating data to determine a second parameter estimation set and a second quality metric. The second parameter estimation set includes electrical estimates of capacitance and equivalent series resistance. The fusion processing module is used to perform weighted fusion processing on the first parameter estimation set and the second parameter estimation set based on the first quality metric and the second quality metric to determine the parameter estimation set. The health assessment module is used to map and generate health status indicators based on the comparison results between the parameter estimation set and the baseline parameter set.

17. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the capacitor health status assessment method as described in any one of claims 1 to 15.

18. A computer device, characterized in that, The device includes a memory and a processor, the memory storing a computer program, characterized in that, when the processor executes the computer program, it implements the steps of the capacitor health status assessment method as described in any one of claims 1 to 15.

Images