Method and system for predicting remaining service life of IGBT device

Abstract

The present application provides a method and system for predicting the remaining service life of an IGBT device, comprising: simulating an IGBT device power cycle aging experiment, collecting conduction voltage drop historical data on the basis of the experiment, and respectively obtaining a model parameter empirical value and a model parameter estimated value on the basis of the conduction voltage drop historical data; constructing a first state weight on the basis of the current operation state, and performing smooth adjustment on the model parameter estimated value on the basis of the model parameter empirical value and the first state weight to obtain a parameter correction result; and establishing a state transition equation on the basis of the parameter correction result, and predicting the remaining service life of an IGBT device on the basis of the state transition equation. Use of the present application can improve the accuracy of prediction of the remaining service life of the IGBT device.

Description

IGBT device remaining useful lifetime prediction method and system TECHNICAL FIELD

[0001] The present application relates to the technical field of power electronic devices, in particular to an IGBT device remaining useful lifetime prediction method and system. BACKGROUND

[0002] Insulated Gate Bipolar Transistor (IGBT) has become the core component of modern high-power power electronic equipment due to its small driving power, large current-carrying capacity and fast switching speed. However, in actual operation, as the service time of the IGBT device increases, its packaging structure will continue to age under the action of cyclic electro-thermal stress, resulting in gradual decline of the electrical and thermal performance of the device, which may not meet the design requirements of the power electronic equipment, ultimately affecting the reliability of the entire equipment or system, and in severe cases, it may have a huge impact on system safety. Therefore, accurately predicting the remaining useful lifetime (RUL) of the IGBT device is of great significance to improving the reliability of the system. Accurate acquisition of the remaining useful lifetime of the IGBT device not only helps to reliably evaluate its current health status, but also helps to maintain the system and replace the device that will fail in time, thereby ensuring system safety.

[0003] At present, the existing related technology has carried out extensive research on the life modeling of IGBT devices, and has proposed various IGBT device life models, such as the Coffin-Manson model based on the analytical method and the stress-strain model based on the failure physics method. The above life models use the operating condition parameters of the IGBT device as input, such as the junction temperature fluctuation ΔTj, the switching frequency f and the current size IC and other parameters when the device is running, and output the average service life of the IGBT device. However, the input parameters are operating condition parameters, and the output is the average service life, which makes it impossible to accurately predict the remaining useful lifetime of a specific IGBT device. SUMMARY

[0004] The purpose of the present application is to overcome the shortcomings of the above-mentioned existing related technology, and to provide an IGBT device remaining useful lifetime prediction method and system, which can improve the accuracy of the remaining useful lifetime prediction of the IGBT device.

[0005] In a first aspect, the present application provides an IGBT device remaining useful lifetime prediction method, comprising:

[0006] An experiment of simulating power cycle aging of the IGBT device is performed, on-off voltage drop historical data is collected according to the experiment, and model parameter empirical values and model parameter estimated values are obtained according to the on-off voltage drop historical data;

[0007] According to the current running state, a first state weight is constructed, and the model parameter estimated value is adjusted according to the model parameter empirical value and the first state weight, to obtain a parameter correction result;

[0008] According to the parameter correction result, a state transition equation is established, and the remaining life of the IGBT device is predicted according to the state transition equation.

[0009] The present application uses the collected conduction voltage drop historical data to obtain the model parameter empirical value and the model parameter estimated value, and adjusts the model parameter estimated value in combination with the model parameter empirical value and the first state weight constructed according to the current running state, so as to reduce the error of directly estimating the model parameter, to improve the accuracy of the remaining life prediction based on the adjusted input data; at the same time, according to the obtained parameter correction result, the state transition equation is established to predict the remaining life of the IGBT device, to realize the prediction accuracy adjustment based on the output data, and further improve the accuracy of the remaining life prediction.

[0010] Further, the model parameter estimated value is adjusted according to the model parameter empirical value and the first state weight to obtain a parameter correction result, including:

[0011] The first state weight is assigned to the model parameter estimated value to obtain a first weighted result, and the second state weight is assigned to the model parameter empirical value to obtain a second weighted result, and the sum of the first weighted result and the second weighted result is taken as the parameter correction result; wherein the sum of the first state weight and the second state weight is a preset value.

[0012] The present application assigns the first state weight to the model parameter estimated value, and assigns the second state weight to the model parameter empirical value, to realize the smooth adjustment of the model parameter estimated value based on the historical data and the current data, so as to reduce the error of directly estimating the model parameter, to improve the accuracy of the remaining life prediction based on the adjusted input data.

[0013] Further, the parameter correction result includes a first parameter correction result and a second parameter correction result; the model parameter estimated value includes a first model parameter estimated value and a second model parameter estimated value; the model parameter empirical value includes a first model parameter empirical value and a second model parameter empirical value; wherein the first parameter correction result and the second parameter correction result are respectively represented as:

[0014] Wherein, α k and β k are the first parameter correction result and the second parameter correction result, respectively; and The first model parameter estimation value and the second model parameter estimation value are respectively; and both are the first model parameter empirical value and the second model parameter empirical value; The first state weight is The second state weight is

[0015] Further, the current operating state includes: a current voltage state of the IGBT device; and the first state weight is constructed according to the current operating state, including:

[0016] The first state weight is constructed according to a ratio of the current voltage state to a failure threshold of the IGBT device.

[0017] Further, the on-state voltage drop historical data is collected according to the experiment, and the model parameter empirical value and the model parameter estimation value are obtained according to the on-state voltage drop historical data, including:

[0018] The on-state voltage drop historical data is collected according to the experiment, and a plurality of groups of time series data of on-state voltage drop changes are obtained according to the on-state voltage drop historical data, and the model parameter empirical value is obtained according to the time series data.

[0019] Further, the model parameter empirical value is obtained according to the time series data, including:

[0020] The IGBT degradation process model is established by taking the on-state voltage drop data as a characteristic quantity, sampling in the time series data is taken as a training data set of the IGBT degradation process model, the empirical distribution of the model parameters is obtained according to the training data set and the IGBT degradation process model, and the model parameter empirical value is obtained according to the empirical distribution.

[0021] Further, the on-state voltage drop historical data is collected according to the experiment, and the model parameter estimation value is obtained according to the on-state voltage drop historical data, including:

[0022] The model parameters are constrained according to the time series data and based on a range value of the empirical distribution, the model parameters are estimated by a particle swarm optimization algorithm, and the model parameter estimation value is obtained.

[0023] Further, the state transition equation is established according to the parameter correction result, and the residual life of the IGBT device is predicted according to the state transition equation, including:

[0024] According to the current operating state and the parameter correction result, a particle filter state vector and a state transition equation are constructed together, and the particle filter state vector is updated according to the state transition equation for multiple times until a posterior distribution of the state vector under the condition of the on-state voltage drop historical data is obtained, and then the state transition equation at this time is used to predict the remaining life of the IGBT device.

[0025] Further, the remaining life prediction of the IGBT device by using the state transition equation at this time comprises:

[0026] The current particle filter state vector is obtained, and the future state distribution is predicted according to the state transition equation at this time, and if the remaining life is defined as the time required for the particle of the particle filter state vector to reach the failure threshold for the first time from the current time, then the remaining life of the IGBT device is obtained according to the future state distribution.

[0027] In a second aspect, the present application provides an IGBT device remaining life prediction system, comprising: a historical data acquisition module, a correction module and a prediction module; wherein,

[0028] The historical data acquisition module is configured to simulate an IGBT device power cycle aging experiment, collect on-state voltage drop historical data according to the experiment, and obtain model parameter empirical values and model parameter estimated values according to the on-state voltage drop historical data.

[0029] The correction module is configured to construct a first state weight according to a current operating state, and to adjust the model parameter estimated values by smoothing according to the model parameter empirical values and the first state weight, to obtain a parameter correction result.

[0030] The prediction module is configured to establish a state transition equation according to the parameter correction result, and to predict the remaining life of the IGBT device according to the state transition equation. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 is a flowchart of an IGBT device remaining life prediction method provided by an embodiment of the present application;

[0032] Fig. 2 is a flowchart of another IGBT device remaining life prediction method provided by an embodiment of the present application;

[0033] Fig. 3(a) is a schematic diagram of a remaining life prediction method provided by an embodiment of the present application;

[0034] Fig. 3(b) is a schematic diagram of a remaining life prediction result of t k = 140k;

[0035] FIG. 4 is a schematic diagram of the remaining life prediction results at different prediction times according to an embodiment of the present application;

[0036] FIG. 5 is a schematic diagram of the remaining life prediction deviation at different prediction times according to an embodiment of the present application;

[0037] FIG. 6 is a schematic diagram of the structure of the IGBT device remaining life prediction system according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0039] It is worth noting that the existing related technology uses the operating condition parameters as the input parameters and takes the average service life as the output, and fails to correct the model parameters of the IGBT device, and the predicted life is the average life, so that the remaining life of a specific IGBT device cannot be accurately predicted. Based on this, the present application provides an IGBT device remaining life prediction method and system, which can improve the accuracy of the remaining life prediction of the IGBT device.

[0040] In order to more clearly illustrate the solutions of the present application, the following embodiments will be described in detail.

[0041] Embodiment 1

[0042] Referring to FIG. 1, which is a flowchart of the IGBT device remaining life prediction method according to an embodiment of the present application, comprising steps S11-S13, specifically:

[0043] In step S11, an experiment of simulating the power cycle aging of the IGBT device is performed, the on-state voltage drop historical data is collected according to the experiment, and the model parameter empirical value and the model parameter estimated value are obtained according to the on-state voltage drop historical data.

[0044] In some embodiments, the on-state voltage drop historical data is collected according to the experiment, and the model parameter empirical value and the model parameter estimated value are obtained according to the on-state voltage drop historical data, comprising: collecting a plurality of groups of on-state voltage drop change time series data according to the on-state voltage drop historical data, and obtaining the model parameter empirical value according to the time series data.

[0045] In some embodiments, obtaining empirical values ​​of model parameters based on time series data includes: establishing an IGBT degradation process model using on-state voltage drop data as a feature quantity; sampling the time series data as a training dataset for the IGBT degradation process model; obtaining the empirical distribution of model parameters based on the training dataset and the IGBT degradation process model; and obtaining empirical values ​​of model parameters based on the empirical distribution.

[0046] In some embodiments, power cycling accelerated aging experiments are conducted on N+1 groups of IGBT devices, and the on-state voltage drop VCE at each cycle is extracted simultaneously. The on-state voltage drop time series of N groups of IGBT devices are selected to establish a model training dataset V = {V1, V2, ..., V...} N Each sequence represents the change in the on-state voltage drop of an IGBT device.

[0047] In some embodiments, the IGBT degradation process model can be represented as: V CE =1+αln(βt+1);

[0048] Where α and β are model parameters, t is any time, and in the experiment, it is the number of aging cycles of the IGBT device.

[0049] In some embodiments, model parameters are obtained by fitting the on-state voltage drop time series dataset V. The obtained parameters are empirical distributions of model parameters α and β, which are used to estimate the model parameters in the subsequent particle filtering process.

[0050] In some embodiments, the IGBT degradation process model can define state transition equations and observation equations.

[0051] In some embodiments, defining state transition equations and observation equations includes: establishing the space-state equations of the particle filter model, and then defining the state transition equations and state transition equations required by the particle filter model.

[0052] In some implementations, the general form of the space-state equations can be expressed as:

[0053] Where, x k Indicates t k The current IGBT device state value, y k Represents the state measurement value, u k v k These represent process noise and observation noise, respectively.

[0054] In some embodiments, based on the established IGBT degradation process model, the state transition equation required for the particle filter model can be obtained. First, its derivative is taken to obtain:

[0055] In some embodiments, the state transition equations required for the particle filter model can be written in the following discrete form:

[0056] Where Δt is the time step after discretization.

[0057] In some embodiments, the measurement equations required for the particle filter model can be written as: V CE_m,k =V CE,k +v k ;

[0058] Among them, V CE_m,k This represents the measured value of the on-state voltage drop of an IGBT device, v. k ~N(0,Q), where Q is the variance of the observation noise.

[0059] In some embodiments, historical data of conduction voltage drop are collected experimentally, and model parameter estimates are obtained based on the historical data of conduction voltage drop. This includes: constraining the model parameters based on time series data and the range of empirical distribution values, estimating the model parameters using a particle swarm optimization algorithm, and obtaining the model parameter estimates.

[0060] In some embodiments, the historical dataset {V} obtained from the on-state voltage drop measurement of the IGBT device is used as the basis. CE_m} 1:k The model parameters are estimated under constraints using the particle swarm optimization algorithm. The constraints are the empirical range of the empirical distribution of the model parameters. The model parameter estimates are obtained, including a first model parameter estimate and a second model parameter estimate, denoted as […]. and

[0061] Step S12: Based on the current running state, construct the first state weights, and based on the empirical values ​​of the model parameters and the first state weights, smoothly adjust the estimated values ​​of the model parameters to obtain the parameter correction results.

[0062] In some embodiments, the current operating state includes: the current voltage state of the IGBT device; constructing a first state weight based on the current operating state includes: constructing the first state weight based on the ratio of the current voltage state to the failure threshold of the IGBT device.

[0063] In some embodiments, based on the current voltage state V of the IGBT device CE,k The weights of the first state can be calculated as follows:

[0064] Among them, w _th The failure threshold for the IGBT device is set; when the device is operating normally, w_th >V CE,k After the device fails, there is w _th ≤V CE,k .

[0065] In some embodiments, the estimated values ​​of model parameters are smoothly adjusted based on empirical values ​​of model parameters and first state weights to obtain parameter correction results, including: assigning first state weights to the estimated values ​​of model parameters to obtain a first weighted result, and assigning second state weights to empirical values ​​of model parameters to obtain a second weighted result, and using the sum of the first weighted result and the second weighted result as the parameter correction result; wherein the sum of the first state weights and the second state weights is a preset value.

[0066] In some embodiments, the parameter correction results include: a first parameter correction result and a second parameter correction result; the model parameter estimates include: a first model parameter estimate and a second model parameter estimate; and the model parameter empirical values ​​include: a first model parameter empirical value and a second model parameter empirical value.

[0067] It is worth noting that, in some embodiments, the first state weight is constructed based on the ratio of the current motion state to the failure threshold of the IGBT device, thus estimating the first model parameters. The weighted state is used The second state weights are estimates of the parameters of the second model. The weighted state weights, the sum of the first state weight and the second state weight, are preset values, which can be used... To express, This is a preset value.

[0068] In some embodiments, It can be 1 or any other positive number.

[0069] This application achieves smooth adjustment of model parameter estimates based on historical and current data by assigning first state weights to the model parameter estimates and second state weights to the empirical values ​​of the model parameters. This reduces the error of directly estimating model parameters and improves the accuracy of remaining life prediction by adjusting the input data.

[0070] In some embodiments, obtaining model parameter estimates during model parameter estimation includes: a first model parameter estimate. Second model parameter estimates In the experiment simulating the power cycling aging of IGBT devices, empirical values ​​of model parameters were obtained by collecting historical on-state voltage drop data. The collected historical on-state voltage drop data included: the empirical value of the first model parameter α and the empirical value of the second model parameter β. When smoothing the estimated model parameters, parameter correction results were obtained, specifically by adjusting the estimated value of the first model parameter... Second model parameter estimates A smooth adjustment is performed; therefore, the parameter correction results include: the first parameter correction result α. k Second parameter correction result β k The first weighted result refers to the weighting of the first state. Assign the parameter estimates to the first model Second model parameter estimates Then, the corresponding estimated values ​​of the first model parameters are obtained. Second model parameter estimates The weighted result, that is: the weights of the first state Assign the parameter estimates to the first model The first model parameter estimate can be obtained. The weighted result; weights in the first state Assign the parameter estimates to the second model After that, the estimated values ​​of the second model parameters can also be obtained. The first weighted result; similarly, the second weighted result refers to the weighting of the second state. After assigning empirical values ​​α to the first model parameter and β to the second model parameter, respectively, we obtain the weighted results of the first model parameter empirical value α and the second model parameter empirical value β, that is: the weights of the second state. After assigning empirical values ​​α to the first model parameters, we can obtain the weighted result of the empirical values ​​α for the first model parameters; then we can assign the weights of the second state. After assigning the empirical value β to the second model parameter, we can obtain the weighted result of the empirical value β of the second model parameter.

[0071] In some embodiments, the first parameter correction result and the second parameter correction result are respectively expressed as:

[0072] Where, α k and β k These are the results of the first parameter correction and the second parameter correction, respectively. and These are the estimated values ​​of the first and second model parameters, respectively; α and β are both empirical values ​​of the first and second model parameters. These are the first state weights. It is the weight of the second state.

[0073] Step S13: Based on the parameter correction results, establish the state transition equation and predict the remaining lifetime of the IGBT device based on the state transition equation.

[0074] In some embodiments, a state transition equation is established based on the parameter correction results, and the state of the IGBT device can be estimated and the remaining lifetime can be predicted using a particle filter algorithm.

[0075] In some embodiments, a state transition equation is established based on the parameter correction results, and the remaining lifetime of the IGBT device is predicted based on the state transition equation. This includes: constructing a particle filter state vector and a state transition equation based on the current operating state and the parameter correction results, and updating the particle filter state vector multiple times based on the state transition equation until the posterior distribution of the state vector under the condition of historical on-state voltage drop data is obtained, and then predicting the remaining lifetime of the IGBT device based on the state transition equation at this time.

[0076] In some embodiments, the IGBT device voltage V CE Model parameter α k and β k Together, we construct the particle filter state vector x, which can be expressed as: x = [V CE ,α,β].

[0077] In some embodiments, the particle filter state vector is updated multiple times according to the state transition equation until the posterior distribution of the state vector under the condition of historical on-voltage drop data is obtained. This includes: initializing the particle state of the particle filter state vector to obtain the particle state result, and initializing the particle weight of the particle filter state vector to obtain the particle weight result; updating the particle state result according to the state transition equation, and updating the particle weight result according to the observation equation. If the number of valid particles in the updated weight result is less than a threshold, the particles are resampled until the posterior distribution of the state vector under the condition of historical on-voltage drop data is obtained.

[0078] In some embodiments, initial particles can be randomly generated based on the initial state distribution, and the particle filter state vector x can be initialized as follows:

[0079] Where i = 1, 2, ..., Ns, Ns represents the total number of particles, and the particle weight ω of each initial particle is set to be the same.

[0080] In some embodiments, particle weights can be expressed as:

[0081] In some embodiments, the particle state of the particle filter state vector updated according to the state transition equation can be expressed as:

[0082] In some embodiments, updating particle weights according to the observation equation can be expressed as:

[0083] In some embodiments, normalizing the particle weights can be expressed as:

[0084] In some embodiments, the number of effective particles N is calculated. eff , can be represented as:

[0085] In some embodiments, to address the particle degradation problem, when the effective number of particles... When the time comes, the particles are resampled, and the weights of the resampled particles are reset to 1 / Ns.

[0086] Therefore, we can obtain the historical data {V}. CE_m} 1:k Posterior distribution of the state vector under given conditions:

[0087] Where δ(·) represents the Dirac function.

[0088] In some embodiments, predicting the remaining lifetime of an IGBT device using the current state transition equation includes: obtaining the current particle filter state vector and predicting the future state distribution based on the current state transition equation. If the remaining lifetime is defined as the time required for the particles in the particle filter state vector to first reach the failure threshold from the current moment, then the remaining lifetime of the IGBT device is obtained based on the future state distribution.

[0089] In some embodiments, based on the current state, the future state can be predicted according to the state transition equation, thus revealing the state distribution at time k+l as follows:

[0090] in,

[0091] In some embodiments, the remaining lifetime L is defined k For the particle to first reach the failure threshold w at the current time _th The required time, i.e.: L k =inf{l|V CE,k+l ≥w _th |V CE_m,1:k}

[0092] Where l represents the time required for the particle to go from initial time to failure.

[0093] In some embodiments, the distribution of future remaining lifetimes can be represented as:

[0094] Compared to traditional analytical methods for predicting IGBT lifetime, this application incorporates the current state and historical data of the device during the lifetime prediction process, enabling prediction of the remaining lifetime of a specific device. Furthermore, in the process of predicting the remaining lifetime of IGBT devices using the particle filter algorithm, this application introduces state-determined weights to correct the estimated parameters during the parameter estimation stage. This results in the corrected parameters that primarily reflect empirical parameters in the early stages, while better reflecting the specific failure state parameters of the device in the later stages of failure.

[0095] Example 2

[0096] This application provides specific examples of IGBT device lifetime distribution probability prediction application. Referring to Figure 2, which is a flowchart illustrating another IGBT device remaining lifetime prediction method provided in this application, the specific application example is as follows:

[0097] (1) Six groups of IGBT devices were subjected to accelerated power cycling aging experiments. The on-state voltage drop measurement data were divided into a training dataset and a validation dataset with N=5. The training dataset contained five sets of on-state voltage drop variation data. The on-state voltage drop V was established based on the training dataset. CE Degradation process model: V CE =1+αln(βt+1);

[0098] Where t is any time, and α and β are model parameters that can be obtained by fitting the on-state voltage drop sequence.

[0099] After fitting, the model parameters of the five IGBT devices were obtained, as shown in Table 1.

[0100] Table 1 Model parameter values

[0101] By performing normal statistical fitting on the model parameters α and β respectively, the prior distribution of the model parameters can be obtained:

[0102] Where N(μ,σ) represents a normal distribution with mean μ and variance σ.

[0103] (2) Establish the state equation based on the IGBT device degradation process model:

[0104] Where Δt represents the discretized time step, VCE,k Indicates the t-th k The on-state voltage drop of an IGBT device over time, α k and β k Indicates the t-th k Estimated model parameters over time.

[0105] Establish the measurement equation: V CE_m,k =V CE,k +v k ;

[0106] Among them, V CE_m,k Indicates t k The measured value of the on-state voltage over time, v k This represents measurement noise, which is affected by the measuring equipment. Here, it is considered to be white noise with zero mean. k ~N(0,0.001).

[0107] (3) Use the particle filter algorithm to validate the data set at the selected monitoring time t. k To predict remaining lifetime, the validation set data will be used for the first t... k Time data is considered as known historical data, t k Data after a certain time is considered as unknown future data; therefore, we can define the IGBT device at the current time t. k Lifespan L over time k For: L k =inf{l k |V CE,k ≥ω|V CE_m,1:k};

[0108] Among them, {V CE_m} 1:k Indicates t k k sets of historical voltage measurement data before time point, w _th This represents the set voltage failure threshold for the IGBT device. Here, we take 1.05 times the initial voltage value based on experience. inf{·} is used to find the minimum lower limit of the set.

[0109] (4) The model parameters are estimated using a constrained particle swarm optimization algorithm, where the search range of the model parameters is constrained by empirical values ​​of the parameters to avoid overfitting.

[0110] The fitness function fopt of the algorithm is described using the sum of squared errors:

[0111] in, This represents the voltage value calculated based on the current parameter estimation results. and This represents the currently obtained model parameter estimates. The particle swarm optimization algorithm solves for the optimal parameter estimates by minimizing the fitness function within the constraints.

[0112] The particle swarm optimization algorithm uses historical measurement data to estimate model parameters. However, when historical data is limited, the estimation accuracy may be poor. Therefore, first-state weights are introduced. Correct the estimated model parameters:

[0113] in, Indicates the current t k The state over time determines the weight, and its value is defined as the ratio of the current on-state voltage of the IGBT device to the failure threshold. k When the voltage is relatively small, the difference between the voltage of the IGBT device and the failure threshold is large, therefore The value is less than 1, thus the corrected model parameters are closer to the empirical values, reducing the problem of parameter overfitting when the amount of data is small; when t k When the voltage is large, the actual voltage of the IGBT device is closer to the failure threshold, therefore The value is closer to 1, thus the corrected model parameters are closer to the estimated values, further increasing the accuracy of the model parameter estimation results.

[0114] If t k If we use 140k as the current time point to predict the remaining lifetime of the device, then the parameter estimates for the particle swarm optimization algorithm can be obtained. The current state determines the weight. The corrected parameter is α k ~N(-0.0093,0.0014) 2 ), β k ~N(-0.0061,0.00046) 2 Thus, its probability density function can be written:

[0115] in, and The model parameters α are respectively k and β k The posterior probability distribution, μ k and Let these represent their expected value and variance, respectively, and in t k =140 when

[0116] (4) The particle filter algorithm is used to predict the remaining lifetime of the IGBT device. First, an initial particle set is generated based on the above parameter estimation results:

[0117] Where Ns is the number of particles generated in the simulation, here we take Ns = 5000. and From their modified empirical distributions respectively and Importance sampling is performed, and each particle is assigned an initial weight. When the IGBT device is acquired at t k Voltage measurement value V at time CE_m,k Then, the state and weights of all particles are updated.

[0118] The particle state update equation is:

[0119] The particle weight update equation is:

[0120] Normalize the updated particle weights:

[0121] If the updated effective particle count A resampling operation is required to improve the accuracy of the prediction results. The particle weights after resampling are redefined as 1 / Ns.

[0122] Among them, the number of effective particles N eff for:

[0123] Therefore, we can obtain the historical data {V}. CE_m} 1:k Posterior distribution of IGBT device on-state voltage drop under the given conditions:

[0124] Where δ(·) represents the Dirac function.

[0125] Furthermore, based on the current state, future states can be predicted according to the state transition equation. Thus, the state distribution at time k+l when failure occurs is:

[0126] in,

[0127] Furthermore, the future remaining lifetime distribution can be calculated based on the time it takes for particles to reach the failure threshold:

[0128] Thus, we can obtain the result at t. kThe remaining lifetime at the start of prediction at 140k is approximately 27.5k. The remaining lifetime prediction result is shown in Figure 3(a), which is a schematic diagram of the remaining lifetime prediction method provided in this embodiment. Figure 3(b) shows the remaining lifetime prediction result provided in this embodiment. k The diagram shows the remaining lifetime prediction results at 140k. According to the validation set, the actual remaining lifetime of the IGBT device is approximately 26k, with a prediction error of 5.6%, demonstrating the accuracy of the proposed method. Furthermore, Figures 4 and 5 respectively illustrate the remaining lifetime prediction results at different prediction times and the remaining lifetime prediction deviation at different prediction times provided by the embodiments of this application. These figures also demonstrate the high accuracy of the proposed results.

[0129] Example 3

[0130] Referring to Figure 6, it is a schematic diagram of the structure of the IGBT device remaining lifetime prediction system provided in the embodiment of this application, including: historical data acquisition module 31, correction module 32 and prediction module 33.

[0131] It is worth noting that when the IGBT device remaining lifetime prediction system simulates the power cycle aging experiment of IGBT devices in the historical data acquisition module 31, it can be simulated using an online software system or conducted offline.

[0132] In some embodiments, the historical data acquisition module 31 is used to output the empirical value and the estimated value of the model parameter to the correction module 32. After receiving the empirical value and the estimated value of the model parameter, the correction module 32 outputs the parameter correction result and transmits the parameter correction result to the prediction module 33. After receiving the parameter correction result, the prediction module 33 performs the remaining lifetime prediction of the IGBT device and outputs the lifetime prediction result.

[0133] The historical data acquisition module 31 is used to simulate the power cycle aging experiment of IGBT devices. It collects historical data of conduction voltage drop based on the experiment, and obtains empirical values ​​and estimated values ​​of model parameters based on the historical data of conduction voltage drop.

[0134] In some embodiments, historical conduction voltage drop data is collected experimentally, and empirical values ​​and estimated values ​​of model parameters are obtained based on the historical conduction voltage drop data. This includes: collecting historical conduction voltage drop data experimentally, obtaining multiple sets of time series data on the changes in conduction voltage drop based on the historical conduction voltage drop data, and obtaining empirical values ​​of model parameters based on the time series data.

[0135] In some embodiments, obtaining empirical values ​​of model parameters based on time series data includes: establishing an IGBT degradation process model using on-state voltage drop data as a feature quantity; sampling the time series data as a training dataset for the IGBT degradation process model; obtaining the empirical distribution of model parameters based on the training dataset and the IGBT degradation process model; and obtaining empirical values ​​of model parameters based on the empirical distribution.

[0136] In some embodiments, historical data of conduction voltage drop are collected experimentally, and model parameter estimates are obtained based on the historical data of conduction voltage drop. This includes: constraining the model parameters based on time series data and the range of empirical distribution values, estimating the model parameters using a particle swarm optimization algorithm, and obtaining the model parameter estimates.

[0137] The correction module 32 is used to construct the first state weights based on the current running state, and to smoothly adjust the estimated values ​​of the model parameters based on the empirical values ​​of the model parameters and the first state weights, so as to obtain the parameter correction results.

[0138] In some embodiments, the current operating state includes: the current voltage state of the IGBT device; constructing a first state weight based on the current operating state includes: constructing the first state weight based on the ratio of the current voltage state to the failure threshold of the IGBT device.

[0139] In some embodiments, the estimated values ​​of model parameters are smoothly adjusted based on empirical values ​​of model parameters and first state weights to obtain parameter correction results, including: assigning first state weights to the estimated values ​​of model parameters to obtain a first weighted result, and assigning second state weights to empirical values ​​of model parameters to obtain a second weighted result, and using the sum of the first weighted result and the second weighted result as the parameter correction result; wherein the sum of the first state weights and the second state weights is a preset value.

[0140] In some embodiments, the parameter correction results include: a first parameter correction result and a second parameter correction result; the model parameter estimates include: a first model parameter estimate and a second model parameter estimate; and the model parameter empirical values ​​include: a first model parameter empirical value and a second model parameter empirical value.

[0141] In some embodiments, the first parameter correction result and the second parameter correction result are respectively expressed as:

[0142] Where, α k and β k These are the results of the first parameter correction and the second parameter correction, respectively. and These are the estimated values ​​of the first and second model parameters, respectively; α and β are both empirical values ​​of the first and second model parameters. These are the first state weights. It is the weight of the second state.

[0143] The prediction module 33 is used to establish a state transition equation based on the parameter correction results, and to predict the remaining lifetime of the IGBT device based on the state transition equation.

[0144] In some embodiments, a state transition equation is established based on the parameter correction results, and the remaining lifetime of the IGBT device is predicted based on the state transition equation. This includes: constructing a particle filter state vector and a state transition equation based on the current operating state and the parameter correction results, and updating the particle filter state vector multiple times based on the state transition equation until the posterior distribution of the state vector under the condition of historical on-state voltage drop data is obtained, and then predicting the remaining lifetime of the IGBT device based on the state transition equation at this time.

[0145] In some embodiments, the particle filter state vector is updated multiple times according to the state transition equation until the posterior distribution of the state vector under the condition of historical on-state voltage drop data is obtained. This includes: initializing the particle state of the particle filter state vector to obtain the particle state result, and initializing the particle weights of the particle filter state vector to obtain the initial weight result; updating the particle state result according to the state transition equation, and updating the initial weight result according to the observation equation. If the number of valid particles in the updated weight result is less than a threshold, the particles are resampled until the posterior distribution of the state vector under the condition of historical on-state voltage drop data is obtained.

[0146] In some embodiments, predicting the remaining lifetime of an IGBT device using the current state transition equation includes: obtaining the current particle filter state vector and predicting the future state distribution based on the current state transition equation. If the remaining lifetime is defined as the time required for the particles in the particle filter state vector to first reach the failure threshold from the current moment, then the remaining lifetime of the IGBT device is obtained based on the future state distribution.

[0147] This application uses historical on-state voltage drop data to obtain empirical and estimated values ​​of model parameters. It then combines these empirical values ​​with first-state weights constructed based on the current operating state to adjust the estimated model parameters. This reduces the error of directly estimating model parameters, thereby improving the accuracy of remaining lifetime prediction based on adjusted input data. Simultaneously, based on the obtained parameter correction results, a state transition equation is established to predict the remaining lifetime of the IGBT device, achieving prediction accuracy adjustment based on output data and further improving the accuracy of remaining lifetime prediction.

[0148] Those skilled in the art will understand that embodiments of this application may also include computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0150] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0151] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0152] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for predicting the remaining lifetime of an IGBT device, characterized in that, include: An experiment simulating the power cycling aging of IGBT devices was conducted. Historical on-state voltage drop data was collected based on the experiment, and empirical values ​​and estimated values ​​of model parameters were obtained based on the historical on-state voltage drop data. Based on the current operating state, construct the first state weights, and based on the empirical values ​​of the model parameters and the first state weights, smoothly adjust the estimated values ​​of the model parameters to obtain the parameter correction results; Based on the parameter correction results, a state transition equation is established, and the remaining lifetime of the IGBT device is predicted based on the state transition equation.

2. The method for predicting the remaining lifetime of an IGBT device as described in claim 1, characterized in that, The step of smoothly adjusting the estimated model parameters based on the empirical values ​​of the model parameters and the first state weights to obtain the parameter correction result includes: The first state weight is assigned to the estimated value of the model parameters to obtain the first weighted result, and the second state weight is assigned to the empirical value of the model parameters to obtain the second weighted result. The sum of the first weighted result and the second weighted result is the parameter correction result; wherein the sum of the first state weight and the second state weight is a preset value.

3. The method for predicting the remaining lifetime of an IGBT device as described in claim 2, characterized in that, The parameter correction results include: a first parameter correction result and a second parameter correction result; the model parameter estimates include: a first model parameter estimate and a second model parameter estimate; the model parameter empirical values ​​include: a first model parameter empirical value and a second model parameter empirical value; wherein, the first parameter correction result and the second parameter correction result are respectively expressed as: Where, α k and β k These are the results of the first parameter correction and the second parameter correction, respectively. and These are the estimated values ​​of the first and second model parameters, respectively; α and β are both empirical values ​​of the first and second model parameters. These are the first state weights. It is the weight of the second state.

4. The method for predicting the remaining lifetime of an IGBT device as described in claim 1, characterized in that, The current operating state includes: the current voltage state of the IGBT device; the construction of the first state weight based on the current operating state includes: The first state weight is constructed based on the ratio of the current voltage state to the failure threshold of the IGBT device.

5. The method for predicting the remaining lifetime of an IGBT device as described in claim 1, characterized in that, Based on the historical conduction voltage drop data collected in the experiment, empirical values ​​and estimated values ​​of model parameters are obtained, including: Based on the historical data of conduction voltage drop collected in the experiment, and based on the historical data of conduction voltage drop, multiple sets of time series data of conduction voltage drop changes are obtained, and empirical values ​​of model parameters are obtained based on the time series data.

6. The method for predicting the remaining lifetime of an IGBT device as described in claim 5, characterized in that, The step of obtaining empirical values ​​for model parameters based on the time series data includes: An IGBT degradation process model is established using on-state voltage drop data as a feature quantity. Samples are taken from the time series data to serve as the training dataset for the IGBT degradation process model. Based on the training dataset and the IGBT degradation process model, the empirical distribution of the model parameters is obtained, and the empirical values ​​of the model parameters are obtained based on the empirical distribution.

7. The method for predicting the remaining lifetime of an IGBT device as described in claim 6, characterized in that, Based on the historical conduction voltage drop data collected in the experiment, and based on the historical conduction voltage drop data, estimated values ​​of model parameters are obtained, including: Based on the time series data and constraining the model parameters according to the range of the empirical distribution, the model parameters are estimated using the particle swarm optimization algorithm to obtain the estimated model parameter values.

8. The method for predicting the remaining lifetime of an IGBT device as described in claim 1, characterized in that, The step of establishing a state transition equation based on the parameter correction result, and predicting the remaining lifetime of the IGBT device based on the state transition equation, includes: Based on the current operating state and the parameter correction results, a particle filter state vector and a state transition equation are jointly constructed. The particle filter state vector is then updated multiple times according to the state transition equation until the posterior distribution of the state vector under the condition of the historical on-state voltage drop data is obtained. The remaining lifetime of the IGBT device is then predicted using the state transition equation at this point.

9. The method for predicting the remaining lifetime of an IGBT device as described in claim 8, characterized in that, The prediction of the remaining lifetime of the IGBT device based on the current state transition equation includes: Obtain the current particle filter state vector and predict the future state distribution based on the current state transition equation. If the remaining lifetime is defined as the time required for the particles in the particle filter state vector to first reach the failure threshold from the current moment, then the remaining lifetime of the IGBT device is obtained based on the future state distribution.

10. A system for predicting the remaining lifetime of an IGBT device, characterized in that, include: The system comprises a historical data acquisition module, a correction module, and a prediction module; among which, The historical data acquisition module is used to simulate the power cycle aging experiment of IGBT devices. It collects historical on-state voltage drop data based on the experiment and obtains empirical values ​​and estimated values ​​of model parameters based on the historical on-state voltage drop data. The correction module is used to construct a first state weight based on the current running state, and to smoothly adjust the estimated value of the model parameters based on the empirical value of the model parameters and the first state weight, so as to obtain the parameter correction result. The prediction module is used to establish a state transition equation based on the parameter correction result, and to predict the remaining lifetime of the IGBT device based on the state transition equation.

Images