Method for predicting health state of key element of frequency converter

Abstract

The invention discloses a health state prediction method for a frequency converter key element, and belongs to the technical field of electric power. Direct current capacitor data and IGBT data are collected, a constraint prediction model C-S2 is constructed, dynamic calculation of capacitor degradation and residual life prediction are performed, and the problem that a constraint model between a capacitor and IGB T is established is solved. In order to realize degradation prediction and health state evaluation of the direct current bus capacitor, the invention provides a constraint modeling thought of a main element and a related element, the working characteristics of the IGBT are introduced into a capacitor degradation calculation model, required parameters can be acquired through a conventional sensor, and the method has good engineering realizability and can be applied to the technical field of direct current bus capacitor degradation prediction and health state evaluation. A time windowing feature extraction and dynamic updating mechanism is adopted, real-time prediction of the health state and the residual life of the capacitor can be achieved, and dependence on field parameters such as ESR which are difficult to measure is reduced.

Description

Technical Field

[0001] This invention belongs to the field of power technology, and in particular relates to a method for predicting the health status of key components of frequency converters. Background Technology

[0002] As a key device in industrial electrical control systems, the reliability of frequency converters directly depends on the health status of power devices and energy storage components.

[0003] In the main circuit of a frequency converter, the DC bus capacitor and the IGBT module are two of the most critical and easily aged components:

[0004] DC bus capacitors play a vital role in smoothing DC voltage and suppressing current ripple. Their performance degradation can cause DC bus voltage fluctuations, increased ripple, and even system instability.

[0005] As a main power switching device, the thermal and electrical stress changes of IGBTs can affect the on-state voltage drop, switching losses, and device lifespan.

[0006] Current health status prediction methods mostly employ a unit-by-unit modeling approach, such as establishing a lifetime model based solely on capacitor ripple current, temperature, or ESR parameters. However, these methods have the following shortcomings:

[0007] 1. Lack of multi-component coupling modeling: Capacitor degradation is coupled with the thermal and electrical stress of IGBTs, but existing models do not consider the associated effects, resulting in large prediction bias;

[0008] 2. Single feature and reliance on difficult-to-measure parameters: Most models rely on features such as ESR and power loss that are difficult to collect in real time, which is not conducive to engineering implementation;

[0009] 3. Inability to achieve real-time prediction and updates: Traditional methods are mostly offline analyses and lack a real-time data stream health status update mechanism. Summary of the Invention

[0010] The purpose of this invention is to provide a method for predicting the health status of key components of a frequency converter, which solves the technical problem of establishing a constraint model between capacitors and IGBTs to achieve degradation prediction and health status assessment of DC bus capacitors.

[0011] To achieve the above objectives, the present invention adopts the following technical solution:

[0012] A method for predicting the health status of key components of a frequency converter includes the following steps:

[0013] Step 1: Establish a data preprocessing module in the data center. This module acquires real-time data from various sensors in the inverter system via the Internet of Things (IoT) to construct a real-time dataset. The real-time dataset includes DC capacitor data and IGBT data. The DC capacitor data includes ripple current I0. ripple Ripple voltage V ripple and bus voltage fluctuation V bus IGBT data includes bridge arm current I arm , conduction voltage V ce and switching transient voltage V switch The data in the real-time dataset is filtered, denoised, and windowed to extract the data features of each time window, and a feature vector F is constructed with the window as the unit. k :

[0014] F k =[I rms{ripple,k} V rms{ripple,k} V var{bus,k} ,I rms{arm,k} V avg{ce,k} V peak{switch,k} ];

[0015] Among them, I rms{ripple,k} and V rms{ripple,k} The ripple current I of the k-th window are respectively ripple and ripple voltage V ripple RMS value, V var{bus,k} The bus voltage fluctuation V in the k-th window bus The mean of I rms{arm,k} Let I be the bridge arm current of the k-th window. arm RMS value, V avg{ce,k} The on-state voltage V of the k-th window ce The mean of V peak{switch,k} V is the switching transient voltage of the k-th window. switch The peak value;

[0016] Step 2: Establish a model building module in the data center, and the model building module retrieves the feature vector F. k A capacitor degradation prediction model C-S2 is constructed, and the correlation constraints between IGBTs and DC bus capacitors are established:

[0017] The C-S2 model for predicting capacitance degradation is as follows:

[0018] ΔCap k+1 =a×I rms{ripple,k} +b×V rms{ripple,k} +c×V var{bus,k} +d×F IGBT,k ;

[0019] F IGBT,kThe association constraints are as follows:

[0020] F IGBT,k =k1×I rms{arm,k} +k2×V avg{ce,k} +k3×V peak{switch,k} ;

[0021] Where a, b, c, k1, k2, and k3 are all weighting coefficients; k represents the window index; ΔCap k+1 The calculation results of the C-S2 capacitance degradation prediction model are expressed as capacitance degradation in units of a window.

[0022] Step 3: The prediction output module retrieves the capacitor degradation prediction model C-S2 and uses it to calculate the capacitance degradation ΔCap of the DC bus capacitor. k+1 Dynamically update the capacitor health status Cap k+1 Calculate the remaining lifetime (RUL) cap,k ;

[0023] Based on the calculation results of the capacitor degradation prediction model C-S2, a bar chart of capacitor health status is constructed and displayed through a visualization interface.

[0024] Preferably, step 2 includes the following steps:

[0025] Step 2-1: Construct the capacitor degradation prediction model C-S2, and establish the correlation constraint F between the IGBT and the DC bus capacitor. IGBT,k , will ripple current I ripple Ripple voltage V ripple and bus voltage fluctuation V bus As key data, it directly affects the calculation of capacitance degradation value, and will be associated with constraint F. IGBT, k serves as a constraint condition in the calculation of the corrected capacitance degradation value, ultimately yielding the capacitance degradation amount ΔCap. k+1 Capacitor degradation ΔCap k+1 It visually reflects the health status of the DC bus capacitors;

[0026] Step 2-2: Update the health status of the DC bus capacitor using the following state update model:

[0027] Cap k+1 =Cap k -ΔCap k+1 ;

[0028] Among them, Cap k+1 Cap is the healthy value of the DC bus capacitance in the (k+1)th window. k Let be the health value of the DC bus capacitance in the k-th window;

[0029] Steps 2-3: Construct a prediction model for the remaining lifetime of the DC bus capacitor:

[0030] RUL cap,k =(Cap k ÷ΔCap k+1 )×T win ;

[0031] Among them, RUL cap,k T represents the predicted remaining useful life. win Indicates the duration of a single time window.

[0032] Preferably, step 3 includes the following steps:

[0033] Step 3-1: Retrieve the capacitor degradation prediction model C-S2, the remaining lifetime prediction model, and the state update model. Use these three models to perform joint calculations to obtain the updated state Cap of the DC bus capacitor. k+1 and remaining useful life prediction value RUL cap,k ;

[0034] Step 3-2: On a window-by-window basis, update the Cap status. k+1 and remaining useful life prediction value RUL cap,k Generate a bar chart;

[0035] Step 3-3: Update the bar chart and status Cap. k+1 Remaining life prediction value (RUL) cap,k Both the real-time dataset and the actual dataset are sent to the web interface for display.

[0036] Preferably, each sensor in the frequency converter system specifically includes:

[0037] The device used to collect ripple current I in the DC bus circuit ripple Hall current sensor;

[0038] A device for acquiring ripple voltage V is set across the DC bus capacitor. ripple Voltage sampling sensor;

[0039] A sensor installed at the positive and negative ends of the busbar is used to collect busbar voltage fluctuations. bus Voltage sampling sensor;

[0040] A sensor located on the output side of the IGBT arm is used to collect the arm current I. arm Hall current sensor;

[0041] The on-state voltage V is located at both ends of the IGBT bridge. ce Voltage sampling sensor.

[0042] The voltage V is set at both ends of the IGBT bridge to collect the transient switching voltage. switch Voltage sampling sensor.

[0043] Preferably, each sensor communicates with the data center via the Internet of Things, and the data center adopts a B / S architecture.

[0044] The present invention provides a method for predicting the health status of key components of a frequency converter, which solves the technical problem of establishing a constraint model between capacitors and IGBTs to predict the degradation and assess the health status of DC bus capacitors. The present invention proposes a constraint modeling approach for main components and related components, and introduces the operating characteristics of IGBTs into the capacitor degradation calculation model. The required parameters can all be collected by conventional sensors, which has good engineering feasibility. By adopting a time window feature extraction and dynamic update mechanism, the health status and remaining life of capacitors can be predicted in real time, reducing the dependence on difficult-to-measure parameters such as ESR in the field. Attached Figure Description

[0045] Figure 1 This is the main flowchart of the present invention;

[0046] Figure 2 This is a flowchart of step 1 of the present invention;

[0047] Figure 3 This is a flowchart of step 2 of the present invention;

[0048] Figure 4 This is a flowchart of step 3 of the present invention. Detailed Implementation

[0049] Depend on Figures 1-4 The method for predicting the health status of a key component of a frequency converter, as shown, includes the following steps:

[0050] Step 1: Establish a data preprocessing module in the data center. The data preprocessing module acquires real-time data from various sensors in the inverter system through the Internet of Things and constructs a real-time dataset.

[0051] All sensors communicate with the data center via the Internet of Things (IoT), and the data center adopts a B / S architecture.

[0052] The real-time dataset includes DC capacitor data and IGBT data. The DC capacitor data includes ripple current Ic. ripple Ripple voltage V ripple and bus voltage fluctuation V bus IGBT data includes bridge arm current I arm , conduction voltage V ce and switching transient voltage V switch ;

[0053] The sensors in the inverter system specifically include:

[0054] The device used to collect ripple current I in the DC bus circuit ripple Hall current sensor;

[0055] A device for acquiring ripple voltage V is set across the DC bus capacitor. ripple Voltage sampling sensor;

[0056] A sensor installed at the positive and negative ends of the busbar is used to collect busbar voltage fluctuations. bus Voltage sampling sensor;

[0057] A sensor located on the output side of the IGBT arm is used to collect the arm current I. arm Hall current sensor;

[0058] The on-state voltage V is located at both ends of the IGBT bridge. ce Voltage sampling sensor.

[0059] The voltage V is set at both ends of the IGBT bridge to collect the transient switching voltage. switch Voltage sampling sensor.

[0060] In this embodiment, no additional sensor acquisition system is required. The analysis of relevant data on the DC bus capacitor and the prediction of its lifespan can be completed using only the sensors built into the inverter system, which greatly improves the applicability of the product and reduces costs.

[0061] The data in the real-time dataset is filtered, denoised, and windowed to extract the data features of each time window, and a feature vector F is constructed with the window as the unit. k :

[0062] F k =[I rms{ripple,k} V rms{ripple,k} V var{bus,k} ,I rms{arm,k} V avg{ce,k} V peak{switch,k} ];

[0063] Among them, I rms{ripple,k} and V rms{ripple,k} The ripple current I of the k-th window are respectively ripple and ripple voltage V ripple RMS value, V var{bus,k} The bus voltage fluctuation V in the k-th window bus The mean of I rms{arm,k} Let I be the bridge arm current of the k-th window. arm RMS value, V avg{ce,k} The on-state voltage V of the k-th windowce The mean of V peak{switch,k} V is the switching transient voltage of the k-th window. switch The peak value;

[0064] In this embodiment, the filtering, denoising, and windowing processing of real-time data specifically includes filtering and denoising, RMS (Real Mean Square) calculation, mean calculation, peak value calculation, and fluctuation amplitude calculation. The processing model used is an existing technology and has universality. The specific model is as follows:

[0065] The filtering and noise reduction process is a moving average filter:

[0066]

[0067] Among them, X filt，k This represents the result of the moving average filtering; N represents the number of sample points in the moving average window, i represents the intermediate variable, and X represents the result of the moving average filtering. raw，k-i X represents the ki-th sampling point of the original time-series signal, where k represents the current sample position and X represents a general variable.

[0068] In this embodiment, the value of N is determined based on the specific data types involved in the filtering and denoising process, N = f s / f cut ;f s f represents the sampling rate. cut Indicates the highest noise frequency that is expected to be filtered out;

[0069] Ripple current I ripple Ripple voltage V ripple and bus voltage fluctuation V bus Bridge arm current I arm , conduction voltage V ce and switching transient voltage V switch Substituting each value into X yields the results of the moving average filtering for each data point.

[0070] For example, if f is used for slow signals such as ripple... s =10kHz, expected to filter out >500Hz noise, N≈20 can be selected (i.e. 2ms smoothing).

[0071] The principle of substituting X for RMS, mean, peak, and fluctuation amplitude calculations is the same as that for X in filtering and denoising.

[0072] The formula for calculating RMS (Real Mean Squared Value) is as follows:

[0073]

[0074] Among them, X filt，i X represents the i-th filtered sampling point within the k-th window; rms{k}This represents the RMS effective value of the k-th window, where M represents the number of samples within the window, and M = T. win ×f s In this embodiment, T win The window duration should cover the main ripple frequencies of several cycles, and can specifically be taken between [0.1s, 5s]. For example, if f... s =10kHz and T win =1s, then M=10,000. The specific value of M can be customized according to the scene and the type of data.

[0075] The formula for calculating the mean is as follows:

[0076]

[0077] Among them, X avg{k} This is the mean.

[0078] The formula for calculating the peak value is as follows:

[0079]

[0080] Among them, X peak{k} This is the peak value.

[0081] The formula for calculating the fluctuation range is as follows:

[0082]

[0083] Among them, X var{k} This refers to the fluctuation range.

[0084] Step 2: Establish a model building module in the data center, and the model building module retrieves the feature vector F. k A capacitor degradation prediction model C-S2 is constructed, and the correlation constraints between IGBTs and DC bus capacitors are established:

[0085] The C-S2 model for predicting capacitance degradation is as follows:

[0086] ΔCap k+1 =a×I rms{ripple,k} +b×V rms{ripple,k} +c×V var{bus,k} +d×F IGBT,k ;

[0087] F IGBT,k The association constraints are as follows:

[0088] F IGBT,k =k1×I rms{arm,k} +k2×V avg{ce,k} +k3×V peak{switch,k} ;

[0089] Where a, b, c, k1, k2, and k3 are all weighting coefficients; k represents the window index; ΔCap k+1 The calculation results of the C-S2 capacitance degradation prediction model are expressed as capacitance degradation in units of a window.

[0090] Capacitor degradation is a thermo-electric stress coupled process, caused by long-term effects such as ripple current, ripple voltage, and bus voltage fluctuations, leading to electrolyte evaporation and dielectric aging. In this embodiment, linear combination characteristics (RMS, mean, peak value, and fluctuation amplitude) are used to quantify the degradation amount ΔCap for each time window. k+1 Among them, parameters a, b, and c are used to control the direct influence of capacitor terminal parameters on degradation. a characterizes the influence of ripple current on loss, b characterizes the influence of dielectric stress caused by ripple voltage, and c reflects the thermal shock effect caused by bus voltage fluctuation.

[0091] F IGBT,k This reflects the correlation constraint between the IGBT and the capacitor, that is, the additional impact of the dynamic behavior of the switching device on the capacitor stress. The current pulses and voltage spikes during the IGBT's turn-on / turn-off transients will exacerbate capacitor losses, therefore its weighted summation F IGBT,k Used to correct the degradation estimate of the main model.

[0092] Step 2 includes the following steps:

[0093] Step 2-1: Construct the capacitor degradation prediction model C-S2, and establish the correlation constraint F between the IGBT and the DC bus capacitor. IGBT,k , will ripple current I ripple Ripple voltage V ripple and bus voltage fluctuation V bus As key data, it directly affects the calculation of capacitance degradation value, and will be associated with constraint F. IGBT, k serves as a constraint condition in the calculation of the corrected capacitance degradation value, ultimately yielding the capacitance degradation amount ΔCap. k+1 Capacitor degradation ΔCap k+1 It visually reflects the health status of the DC bus capacitors;

[0094] In this embodiment, a capacitor degradation prediction model C-S2 is established to quantitatively predict the performance degradation of the DC bus capacitor during operation. Capacitor degradation is mainly caused by the combined effects of electrical and thermal stresses, with ripple current I... ripple This will cause the internal temperature of the capacitor to rise and accelerate the evaporation of the electrolyte, resulting in a ripple voltage V. ripple This will cause electric field stress fatigue in the dielectric layer, and bus voltage fluctuation V bus This can lead to periodic thermal shocks and material expansion stress.

[0095] To improve the accuracy of degradation prediction, a correlation constraint term F between the IGBT and the DC bus capacitance is introduced in this step. IGBT,k This describes the indirect effect of IGBT switching characteristics on capacitor stress. The IGBT's arm current I... arm , conduction voltage V ce and switching transient voltage V switch These can respectively reflect the conduction loss of the device, the switching speed, and the additional thermal stress on the capacitor caused by transient spikes.

[0096] Using the C-S2 capacitance degradation prediction model, the instantaneous degradation amount ΔCap of the capacitor can be calculated within each time window k. k+1 This enables dynamic quantification of capacitor health status, providing a foundation for subsequent status updates and lifetime prediction.

[0097] Step 2-2: Update the health status of the DC bus capacitor using the following state update model:

[0098] Cap k+1 =Cap k -ΔCap k+1 ;

[0099] Among them, Cap k+1 Cap is the healthy value of the DC bus capacitance in the (k+1)th window. k Let be the health value of the DC bus capacitance in the k-th window;

[0100] This embodiment employs a discrete-time recursive model to dynamically update the capacitor's health status. During long-term operation, the effective capacitance Cap of the DC bus capacitor gradually decays over time, and the capacitance degradation ΔCap reflects the health loss within a unit time window. The state update model enables iterative calculation of the capacitor's health status, equivalent to discretizing and integrating the continuous degradation process. Each calculation of ΔCap updates the health status Cap, thus forming a complete health change curve on the time axis. This model allows the system to track changes in the remaining effective capacitance in real time and promptly identify sudden changes or abnormal declines in health status.

[0101] Steps 2-3: Construct a prediction model for the remaining lifetime of the DC bus capacitor:

[0102] RUL cap,k =(Cap k ÷ΔCap k+1 )×T win ;

[0103] Among them, RUL cap,k T represents the predicted remaining useful life. win Indicates the duration of a single time window.

[0104] This embodiment uses the known current health status of the capacitor, Cap. k With the current window degradation rate ΔCap k+1 To estimate its future lifespan, if we assume that the degradation rate of the capacitor remains stable in the short term, its usable lifespan is approximately proportional to its current health and inversely proportional to the degradation rate. The Cap in the formula... k ÷ΔCap k+1 This represents the number of windows that the capacitor can sustain under the current degradation trend, multiplied by the time length T corresponding to each time window. win The predicted remaining lifetime (RUL) can then be obtained. cap,k .

[0105] The remaining useful life prediction model can achieve online life assessment when RUL cap,k When the voltage drops below a preset threshold, the system can issue an early warning for maintenance or replacement, thereby effectively preventing inverter system shutdown caused by capacitor failure.

[0106] Step 3: The prediction output module retrieves the capacitor degradation prediction model C-S2 and uses it to calculate the capacitance degradation ΔCap of the DC bus capacitor. k+1 Dynamically update the capacitor health status Cap k+1 Calculate the remaining lifetime (RUL) cap,k ;

[0107] Based on the calculation results of the capacitor degradation prediction model C-S2, a bar chart of capacitor health status is constructed and displayed through a visualization interface.

[0108] Step 3 includes the following steps:

[0109] Step 3-1: Retrieve the capacitor degradation prediction model C-S2, the remaining lifetime prediction model, and the state update model. Use these three models to perform joint calculations to obtain the updated state Cap of the DC bus capacitor. k+1 and remaining useful life prediction value RUL cap,k ;

[0110] In this embodiment, the capacitor health characteristics collected during long-term operation are input into the degradation prediction model C-S2 to fit the nonlinear change trend of the capacitor's equivalent parameters over time. Subsequently, the remaining lifetime prediction model estimates the remaining usable lifetime of the capacitor under set threshold conditions through time series inference and health factor weighting mechanism. The state update model adopts a recursive correction strategy to feed back the prediction residuals of the new round of monitoring data into the model parameters, thereby realizing dynamic updating and adaptive correction of the capacitor state.

[0111] Step 3-2: On a window-by-window basis, update the Cap status. k+1and remaining useful life prediction value RUL cap,k Generate a bar chart;

[0112] This embodiment utilizes a sliding window mechanism to visualize the prediction results in segments, with each time window representing a stage of the capacitor's operating state. By using Cap... k+1 With RUL cap,k The numerical values ​​are mapped to bar chart heights, enabling a trend-based representation of capacitor health. The changing patterns of the bar chart intuitively reflect the rate of capacitor degradation and lifespan decline, providing maintenance personnel with graphical decision-making support. Unlike traditional static numerical outputs, this invention can display the dynamic evolution of lifespan prediction in real time, facilitating rapid identification of predicted anomalies.

[0113] Step 3-3: Update the bar chart and status Cap. k+1 Remaining life prediction value (RUL) cap,k Both the real-time dataset and the actual dataset are sent to the web interface for display.

[0114] This embodiment adopts a deployment method based on a B / S (Browser / Server) architecture. The overall architecture consists of three parts: a data acquisition terminal, a central server, and a web client. The data acquisition terminal is used to collect operating signals of IGBTs and DC bus capacitors, including parameters such as voltage, current, and temperature, and performs preprocessing and feature extraction on the collected data. The central server is used for model calculation and data storage, and has built-in degradation prediction model C-S2, state update model, and lifetime prediction model to complete real-time prediction of capacitor health and remaining lifetime. The web client accesses the server's page interface through a browser, and realizes data calling and visualization display based on HTTP and WebSocket protocols, dynamically presenting capacitor health status, lifetime prediction curves, and operating history, realizing remote visual monitoring and operation and maintenance management.

[0115] The present invention provides a method for predicting the health status of key components of a frequency converter, which solves the technical problem of establishing a constraint model between capacitors and IGBTs to predict the degradation and assess the health status of DC bus capacitors. The present invention proposes a constraint modeling approach for main components and related components, and introduces the operating characteristics of IGBTs into the capacitor degradation calculation model. The required parameters can all be collected by conventional sensors, which has good engineering feasibility. By adopting a time window feature extraction and dynamic update mechanism, the health status and remaining life of capacitors can be predicted in real time, reducing the dependence on difficult-to-measure parameters such as ESR in the field.

Claims

1. A method for predicting the health status of key components of a frequency converter, characterized in that: Includes the following steps: Step 1: Establish a data preprocessing module in the data center. This module acquires real-time data from various sensors in the inverter system via the Internet of Things (IoT) to construct a real-time dataset. The real-time dataset includes DC capacitor data and IGBT data. The DC capacitor data includes ripple current I0. ripple Ripple voltage V ripple and bus voltage fluctuation V bus IGBT data includes bridge arm current I arm , conduction voltage V ce and switching transient voltage V switch The data in the real-time dataset is filtered, denoised, and windowed to extract the data features of each time window, and a feature vector F is constructed with the window as the unit. k : F k =[I rms{ripple,k} ,V rms{ripple,k} ,V var{bus,k} ,I rms{arm,k} ,V avg{ce,k} ,V peak{switch,k} ]; Among them, I rms{ripple,k} and V rms{ripple,k} The ripple current I of the k-th window are respectively ripple and ripple voltage V ripple RMS value, V var{bus,k} The bus voltage fluctuation V in the k-th window bus The mean of I rms{arm,k} Let I be the bridge arm current of the k-th window. arm RMS value, V avg{ce,k} The on-state voltage V of the k-th window ce The mean of V peak{switch,k} V is the switching transient voltage of the k-th window. switch The peak value; Step 2: Establish a model building module in the data center, and the model building module retrieves the feature vector F. k A capacitor degradation prediction model C-S2 is constructed, and the correlation constraints between IGBTs and DC bus capacitors are established: The C-S2 model for predicting capacitance degradation is as follows: ΔCap k+1 =a×I rms{ripple,k} +b×V rms{ripple,k} +c×V var{bus,k} +d×F IGBT,k ; F IGBT,k The association constraints are as follows: F IGBT,k =k1×I rms{arm,k} +k2×V avg{ce,k} +k3×V peak{switch,k} ; Where a, b, c, k1, k2, and k3 are all weighting coefficients; k represents the window index; ΔCap k+1 The calculation results of the C-S2 capacitance degradation prediction model are expressed as capacitance degradation in units of a window. Step 3: The prediction output module retrieves the capacitor degradation prediction model C-S2 and uses it to calculate the capacitance degradation ΔCap of the DC bus capacitor. k+1 Dynamically update the capacitor health status Cap k+1 Calculate the remaining lifetime (RUL) cap,k ; Based on the calculation results of the capacitor degradation prediction model C-S2, a bar chart of capacitor health status is constructed and displayed through a visualization interface.

2. The method for predicting the health status of key components of a frequency converter as described in claim 1, characterized in that: Step 2 includes the following steps: Step 2-1: Construct the capacitor degradation prediction model C-S2, and establish the correlation constraint F between the IGBT and the DC bus capacitor. IGBT,k , will ripple current I ripple Ripple voltage V ripple and bus voltage fluctuation V bus As key data, it directly affects the calculation of capacitance degradation value, and will be associated with constraint F. IGBT, k serves as a constraint condition in the calculation of the corrected capacitance degradation value, ultimately yielding the capacitance degradation amount ΔCap. k+1 Capacitor degradation ΔCap k+1 It visually reflects the health status of the DC bus capacitors; Step 2-2: Update the health status of the DC bus capacitor using the following state update model: Cap k+1 =Cap k -ΔCap k+1 ; Among them, Cap k+1 Cap is the healthy value of the DC bus capacitance in the (k+1)th window. k Let be the health value of the DC bus capacitance in the k-th window; Steps 2-3: Construct a prediction model for the remaining lifetime of the DC bus capacitor: NUMBER cap,k =(Head k ÷ΔHead k+1 )×T win ; Among them, RUL cap,k T represents the predicted remaining useful life. win Indicates the duration of a single time window.

3. The method for predicting the health status of key components of a frequency converter as described in claim 2, characterized in that: Step 3 includes the following steps: Step 3-1: Retrieve the capacitor degradation prediction model C-S2, the remaining lifetime prediction model, and the state update model. Use these three models to perform joint calculations to obtain the updated state Cap of the DC bus capacitor. k+1 and remaining useful life prediction value RUL cap,k ; Step 3-2: On a window-by-window basis, update the Cap status. k+1 and remaining useful life prediction value RUL cap,k Generate a bar chart; Step 3-3: Update the bar chart and status Cap. k+1 Remaining life prediction value (RUL) cap,k Both the real-time dataset and the actual dataset are sent to the web interface for display.

4. The method for predicting the health status of key components of a frequency converter as described in claim 1, characterized in that: The sensors in the inverter system specifically include: The device used to collect ripple current I in the DC bus circuit ripple Hall current sensor; A device for acquiring ripple voltage V is set across the DC bus capacitor. ripple Voltage sampling sensor; A sensor installed at the positive and negative ends of the busbar is used to collect busbar voltage fluctuations. bus Voltage sampling sensor; A sensor located on the output side of the IGBT arm is used to collect the arm current I. arm Hall current sensor; The on-state voltage V is located at both ends of the IGBT bridge. ce Voltage sampling sensor. The voltage V is set at both ends of the IGBT bridge to collect the transient switching voltage. switch Voltage sampling sensor.

5. The method for predicting the health status of key components of a frequency converter as described in claim 1, characterized in that: All sensors communicate with the data center via the Internet of Things (IoT), and the data center adopts a B / S architecture.

Images