Online junction temperature calculation system and method for power semiconductor device

Abstract

The invention provides a junction temperature online calculation system and method for a power semiconductor device, and the system comprises a three-dimensional relation model module which is used for building a three-dimensional relation model of the junction temperature, the conduction voltage drop and the conduction current of a target device, storing the three-dimensional relation model as a three-dimensional lookup table, and obtaining the sensitivity of the junction temperature relative to the conduction voltage drop under different currents; the data acquisition module is used for acquiring conduction voltage drop and current when the device works in real time; the sensitive interval judgment module is used for determining a current sensitive interval, screening current and conducting voltage drop according to the real-time change of the conducting current; the temperature calculation module is only used for calculating a corresponding junction temperature candidate value through a look-up table or a formula based on a three-dimensional look-up table and the screened breakover voltage drop and breakover current in the current sensitive interval; and the data processing module is used for carrying out abnormal value filtering and mean value processing on the junction temperature candidate values in the preset time window to obtain a final junction temperature calculation result. Through sensitive interval selection and data screening, the junction temperature measurement precision of the power semiconductor device can be improved.

Description

Technical Field

[0001] The present application relates to the field of power electronics technology, and specifically, to an online junction temperature calculation system and method for a power semiconductor device. Background Art

[0002] The junction temperature of a power semiconductor device (such as an IGBT, SiC MOSFET, etc.) refers to the actual operating temperature of the PN junction or metal-semiconductor contact junction inside its chip, and it is a core parameter that determines the electrical performance, reliability, and lifespan of the device. During operation, an increase in the junction temperature will cause a decrease in the carrier mobility and an increase in the on-resistance, which will further lead to a reduction in device efficiency, thermal runaway, and even permanent failure. According to statistics, for every 10°C increase in the junction temperature of a power device, its service life will be shortened by approximately 50%. Therefore, accurately measuring the junction temperature is of great significance for optimizing the heat dissipation design, realizing the health status monitoring of the device, and improving the reliability of the power electronics system.

[0003] The prior art lacks an effective sensitive interval screening mechanism and does not optimize the selection of the current sensitive interval, resulting in the measurement data being easily interfered by noise and having low measurement accuracy and stability. The prior art does not introduce a dynamic calibration mechanism, which causes the calculated junction temperature result to deviate from the actual value when the power semiconductor device ages or the circuit structure changes.

[0004] After retrieval, Chinese Patent with Application No. 202410517247.4 discloses an online junction temperature detection circuit and method for a power semiconductor device, which can perform online measurement of the junction temperature. However, the calculation data interval used by it is relatively general, and it does not calibrate the power semiconductor device operating for a long time.

[0005] Application Content

[0006] Aiming at one of the defects in the prior art, the purpose of the present application is to provide an online junction temperature calculation system and method for a power semiconductor device.

[0007] In the first aspect of the present application, there is provided an online junction temperature calculation system for a power semiconductor device, including:

[0008] Three-dimensional relationship model module: A three-dimensional relationship model of the junction temperature, on-state current, and on-state voltage drop of the device is established in advance and stored as a three-dimensional look-up table to obtain the sensitivity of the junction temperature to the on-state voltage drop under different on-state currents;

[0009] Data acquisition module: The on-state voltage drop and on-state current during the operation of the device are collected in real time;

[0010] Sensitive interval determination module: According to the real-time change of the on-state current, the current sensitive interval is determined, and the on-state current and on-state voltage drop located in this interval are screened out;

[0011] Temperature calculation module: Using the filtered on-state voltage drop and on-state current, calculate through look-up table or formula to obtain the corresponding candidate junction temperature values;

[0012] Data processing module: Filter out outliers and perform mean processing on the candidate junction temperature values to obtain the final calculated result of the junction temperature.

[0013] Optionally, the three-dimensional relationship model module includes:

[0014] Data unit: Obtain the on-state voltage drop at different junction temperatures and on-state currents through experimental tests or data manuals;

[0015] Model unit: Perform polynomial interpolation fitting or formula fitting on the relationship between the junction temperature, on-state current, and on-state voltage drop to generate a continuous three-dimensional look-up table;

[0016] Sensitivity unit: In the three-dimensional look-up table, use different on-state voltage drops and junction temperatures at the same on-state current to calculate the sensitivity.

[0017] Optionally, the data acquisition module measures the on-state voltage drop and on-state current at the same measurement frequency and measurement timing, and the measurement timing is to avoid the transient state at the moment of turn-on or turn-off of the power semiconductor device. The transient state includes the overshoot, oscillation, or tailing of the device voltage drop and current caused by the device parasitic capacitance and circuit parasitic inductance during the switching state transition of the device.

[0018] Optionally, the sensitive interval determination module includes:

[0019] X-axis determination unit: Real-time monitor the waveform of the on-state current, and identify its current peak and several sampling points near the peak;

[0020] Y-axis determination unit: Combine the sensitivity distribution of the junction temperature to the on-state voltage drop in the three-dimensional look-up table, and select the current interval with a sensitivity higher than the preset threshold;

[0021] Sensitive interval unit: Combine the sampling points and the current interval to obtain the current sensitive interval.

[0022] Optionally, the data processing module includes one or more of the following units for outlier processing:

[0023] - Statistic-based anomaly detection unit: Dynamically set the threshold interval using the 3σ principle, move the sliding window within the threshold interval, and mark a candidate junction temperature value as an anomaly when it deviates from the mean of the sliding window by ±3 times the standard deviation;

[0024] - Sliding window dynamic update unit: Set a dual threshold mechanism to automatically expand the sliding window length to b times the original value when the number of consecutive abnormal values appears ≥ a times;

[0025] -Data correction strategy unit: For the junction temperature candidate value marked as abnormal, linear interpolation of the forward effective value is used to replace it;

[0026] -Multi-method cross-validation unit: Simultaneously runs the Z-score algorithm and quantile method on candidate junction temperature values, and eliminates data when both methods identify abnormalities.

[0027] -Historical data tracing unit: stores the most recent N valid temperature values to build a reference sequence; a secondary check is triggered when the correlation coefficient between the junction temperature candidate value and the reference sequence is lower than a certain threshold.

[0028] Optionally, the data processing module includes one or more of the following units for performing mean processing:

[0029] - Sliding window configuration unit: uses a variable-length time window, the window length is configurable, and the number of data points N in the window is programmable;

[0030] -Dynamic weight allocation unit: using exponential decay weighting strategy, weight coefficient w i =β N-i , where β is the weight coefficient, i represents the i-th data in the window;

[0031] -Standard deviation check unit: When the standard deviation of the data in the window exceeds the set threshold, the window length is automatically shortened to half of the original value and recalculated;

[0032] - Collaborative processing unit: works in conjunction with the outlier filtering module to replace the marked outliers with the moving average of the first three valid data points or eliminate extreme values.

[0033] Optionally, a safety protection module is further included, which receives the junction temperature data calculated by the data processing module and triggers protection when the junction temperature data is higher than a set threshold;

[0034] The security protection module includes:

[0035] Dynamic threshold setting unit: sets the temperature warning threshold based on the device parameter manual value and historical aging data;

[0036] Hierarchical early warning trigger unit: obtains the current device junction temperature from the data processing module. When the junction temperature exceeds a certain proportion of the threshold for N consecutive power frequency cycles, the current soft derating control is activated; when the instantaneous value of the junction temperature exceeds the threshold and the rate of change remains positive, the hardware protection circuit is immediately activated.

[0037] Optionally, it further includes one or both of the following modules:

[0038] A junction temperature calibration module, which calibrates the three-dimensional relationship among the junction temperature, on-state voltage drop, and on-state current when there is a deviation between the measured values of the on-state voltage drop and on-state current and the three-dimensional look-up table, and sends the corrected three-dimensional look-up table to the three-dimensional relationship model module.

[0039] A fault locking module, which obtains real-time current and on-state voltage drop data from the data acquisition module, and locks the use of the three-dimensional relationship model module when it detects that the characteristic curves of the device on-state voltage drop and on-state current deviate from the three-dimensional look-up table by more than a certain proportion.

[0040] Optionally, the junction temperature calibration module includes:

[0041] A reference temperature calibration unit: during the initialization stage of the device or when triggered by a preset working condition, it collects the temperature at a certain position of the device in real time through an external temperature sensing device as the reference temperature;

[0042] A dynamic data acquisition unit: at the reference temperature, it uses the data acquisition module to synchronously collect on-state voltage drop parameters at different currents to generate an on-state voltage drop - current characteristic curve;

[0043] A model dynamic correction unit: extracts the interface function of the look-up table or fitting formula in the three-dimensional relationship model at the reference temperature, introduces a compensation value to make the corrected model a model about the compensation value, and the cross-sectional function value at the reference temperature is equal to the value of the on-state voltage drop - current characteristic curve at the reference temperature.

[0044] In the second aspect of the present application, a method for online calculating the junction temperature of a power semiconductor device is provided, including:

[0045] Pre-establish a three-dimensional relationship model of the junction temperature, on-state current, and on-state voltage drop of the device, store it as a three-dimensional look-up table, and obtain the sensitivity of the junction temperature relative to the on-state voltage drop at different on-state currents;

[0046] Real-time collect the on-state voltage drop and on-state current when the device is working;

[0047] According to the real-time change of the on-state current, determine the current sensitive interval, and screen out the on-state current and on-state voltage drop located in this interval;

[0048] Adopt the screened on-state voltage drop and on-state current, and obtain the corresponding junction temperature candidate value through look-up table or formula calculation;

[0049] Filter out outliers and perform mean processing on the junction temperature candidate value to obtain the final junction temperature calculation result.

[0050] The online calculation system and method for the junction temperature of a power semiconductor device provided in the present application adopt a technical means of filtering data within a sensitivity range and then calculating the junction temperature, which brings about a technical effect that can improve the measurement accuracy and stability of the junction temperature of the power semiconductor device. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0052] Figure 1 1 is a schematic structural diagram of an online calculation system for junction temperature of a power semiconductor device according to an exemplary embodiment;

[0053] Figure 2 is a schematic diagram of a current sensitive interval according to an exemplary embodiment;

[0054] Figure 3 4 is a flow chart showing a method for online calculation of junction temperature of a power semiconductor device according to an exemplary embodiment;

[0055] Figure 4 is a three-dimensional relationship diagram showing the measured junction temperature, on-state current, and on-state voltage drop of a certain type of device according to an exemplary embodiment;

[0056] Figure 5 FIG1 is a diagram showing the relationship between sensitivity and current of a certain type of device according to an exemplary embodiment;

[0057] Figure 6 The figure is a comparison diagram of junction temperature calculation effects before and after data screening according to an exemplary embodiment. DETAILED DESCRIPTION

[0058] The present application is described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present application, but are not intended to limit the present application in any form. It should be noted that, without departing from the concept of the present application, a number of variations and improvements may be made by those skilled in the art, and these all fall within the scope of protection of the present application. Parts not described in detail in the following examples may be implemented using existing technologies.

[0059] Explanation of terms:

[0060] Sensitivity refers to the partial derivative of the on-state voltage drop relative to the junction temperature in the three-dimensional relationship of Tj(Vce,I) (Tj represents junction temperature, Vce represents on-state voltage drop, and I represents on-state current), that is, In this application, after obtaining the three-dimensional lookup table, the corresponding relationship of different (Vce, Tj) under the same current can be used to find out, that is, ΔVce / ΔTj.

[0061] The current sensitive interval refers to the current range in which the conduction current is near the peak value of the alternating current and the sensitivity of the junction temperature relative to the conduction voltage drop is higher than a preset threshold value.

[0062] The three-dimensional look-up table refers to the relationship table of the three dimensions of junction temperature, current, and conduction voltage drop.

[0063] The characteristic curve refers to the relationship curve between the conduction voltage drop, conduction current, and junction temperature measured during the actual operation of the device, which reflects the electrical characteristics of the device in the real working state.

[0064] In the prior art, the accuracy of junction temperature measurement is insufficient. Based on the above problems, the embodiments of the present application provide an on-line calculation system and method for the junction temperature of a power semiconductor device to solve the above existing problems.

[0065] Refer to Figure 1 As shown in

[0066] Three-dimensional relationship model module: A three-dimensional relationship model of the junction temperature, conduction current, and conduction voltage drop of the device is established in advance and stored as a three-dimensional look-up table to obtain the sensitivity of the junction temperature relative to the conduction voltage drop under different conduction currents;

[0067] Data acquisition module: Real-time acquisition of the conduction voltage drop and conduction current during the operation of the device;

[0068] Sensitive interval determination module: According to the real-time change of the conduction current, determine the current sensitive interval, and screen out the conduction current and conduction voltage drop located in this interval;

[0069] Temperature calculation module: Use the screened conduction voltage drop and conduction current to calculate the corresponding junction temperature candidate value through table lookup or formula;

[0070] Data processing module: Filter out outliers and perform mean processing on the junction temperature candidate values to obtain the final junction temperature calculation result.

[0071] In the embodiments of the present application, by carefully selecting the sensitive interval and performing data screening, the error of the calculation result is minimized, the noise is effectively reduced, and the stability is significantly improved, thereby enhancing the accuracy of the junction temperature measurement of the power semiconductor device.

[0072] In order to obtain a more accurate three-dimensional relationship model, in some specific embodiments of the present application, for the three-dimensional relationship model module, data units and model units can be used.

[0073] Specifically, data unit: Obtain the conduction voltage drop data under different junction temperatures and conduction currents through experimental tests or data manuals.

[0074] It should be noted that a forward voltage drop can be determined by a forward current and a junction temperature. Therefore, whether it is a data sheet or experimental testing, the values of M * N forward voltages can be obtained at M forward currents and N junction temperatures.

[0075] Model unit: According to the operating junction temperature and current range of the power semiconductor device, polynomial interpolation fitting or formula fitting is performed on the forward current, forward voltage, and junction temperature data obtained from the data unit and their three-dimensional correspondence to generate a continuous three-dimensional look-up table.

[0076] In the above embodiments of the present application, the three-dimensional relationship model (three-dimensional look-up table) is used to describe the relationship between the junction temperature of the device under test and the forward current and forward voltage drop. This relationship is different for each device and needs to be determined through experiments or data sheets. During the actual operation of the device, the junction temperature can be calculated by measuring the forward current and forward voltage drop.

[0077] In order to obtain more accurate real-time data, in some specific embodiments of the present application, the data acquisition module needs to measure the forward voltage drop and forward current simultaneously at the same measurement frequency and measurement timing, and the measurement timing needs to avoid the transient state at the moment when the power semiconductor device is turned on or off.

[0078] Specifically, the transient state includes the overshoot, oscillation, or tailing of the device voltage drop and current caused by the parasitic capacitance of the device and the parasitic inductance in the circuit when the device switches between the on and off states.

[0079] The sensitivity of the junction temperature to the forward voltage drop increases as the current increases. In order to effectively reduce the influence of sampling noise and error on the calculation of the junction temperature and calculate the junction temperature more accurately. In some specific embodiments of the present application, as Figure 2 shown, for the sensitive interval determination module, an x-axis determination unit and a y-axis determination unit can be used.

[0080] x-axis determination unit: Monitor the waveform of the forward current in real time and identify the current peak and several sampling points near the peak.

[0081] Specifically, the several sampling points here refer to the set of sampling points with a certain length before and after the sampling point corresponding to the current peak, and the specific length can be determined according to the actual operation of the device.

[0082] y-axis determination unit: Combining the sensitivity distribution of the junction temperature to the forward voltage drop in the three-dimensional look-up table, select the current interval with a sensitivity higher than the preset threshold as the current sensitive interval.

[0083] Among them, the preset threshold is jointly determined by the sensitivity of the junction temperature to the forward voltage drop in the three-dimensional relationship model, the sampling error of the forward voltage drop, and the tolerance for calculating the junction temperature.

[0084] By screening the sensitive interval determined by the above embodiments, current and conduction voltage sampling points with as large a current as possible and a sensitivity greater than a certain threshold can be selected, which can effectively reduce the influence of sampling noise and errors on the junction temperature calculation and calculate the junction temperature more accurately.

[0085] In some specific embodiments of the present application, in order to obtain accurate and reliable junction temperature calculation results, for the temperature calculation module, bilinear interpolation, bicubic spline interpolation, or triangulation interpolation can be used for the look-up table calculation, and the formula calculation includes obtaining a functional relationship between the junction temperature, conduction voltage, and conduction current by fitting an empirical formula or a semiconductor physics model.

[0086] Exemplarily, a feasible empirical formula is: where T j is the junction temperature, V CE is the conduction voltage, I is the conduction current, and a, b, c, d, e are fitting constants.

[0087] Exemplarily, a feasible physical model is: where, V CE is the conduction voltage, I is the conduction current, T j is the junction temperature, k is the Boltzmann constant, q is the carrier charge, and R CH and I0 are constants related to the device.

[0088] In the above embodiments of the present application, through the conduction current value and conduction voltage data obtained by actual measurement, calculations are performed to determine the accurate junction temperature condition of the device under normal operating conditions.

[0089] In some specific embodiments of the present application, in order to further ensure the reliability of the calculated data, the data processing module can use one or more of the units such as a statistical-based outlier detection unit, a sliding window dynamic update unit, a data correction strategy unit, a multi-method cross-validation unit, and a historical data traceability unit to filter outlier values.

[0090] Statistical-based outlier detection unit: Dynamically set the threshold interval using the 3σ principle, and mark as an outlier when the junction temperature candidate value deviates from the sliding window mean by ±3 times the standard deviation;

[0091] Sliding window dynamic update unit: Set a double-threshold mechanism, and automatically expand the sliding window length to b times the original value when the number of consecutive occurrences of outlier values ≥ a times;

[0092] Specifically, the double threshold here refers to a and b. The two thresholds correspond to the following trigger conditions respectively:

[0093] Outlier persistence determination threshold (a times)

[0094] Definition: The lower limit of the number of consecutive occurrences of exceptions that trigger window expansion.

[0095] Function: When the number of consecutive occurrences of abnormal values ≥ a, it is determined to be in a continuous abnormal state, and the sliding window needs to be adjusted to avoid misjudgment or missed detection. Exemplarily, a = 5 means that the expansion is triggered only when an abnormality is detected 5 times consecutively.

[0096] Window expansion multiple threshold (b times)

[0097] Definition: The upper limit of the ratio of window length adjustment.

[0098] Function: Expand the sliding window length from the original value N to N×b, improve statistical stability by increasing the sample size, and reduce short-term noise interference. Exemplarily, b = 1.5 means the window is enlarged by 50%.

[0099] The above double-threshold mechanism constitutes progressive conditions. First, it satisfies that the number of consecutive abnormal occurrences ≥ a, and then the window is expanded by b times. It can trigger a larger window for high-frequency abnormalities and suppress false alarms; for low-frequency abnormalities, a small window is maintained to retain sensitivity. Therefore, the final obtained junction temperature data value has a dynamic balance between sensitivity and robustness.

[0100] Data correction strategy unit: Use linear interpolation of the forward valid value to replace the marked abnormal value.

[0101] Specifically, the substitution weight w = (1 - α) + α×(T n -T n-1 ), where α is the weight coefficient; Tn represents the junction temperature calculated in the current calculation period, and T n-1 represents the junction temperature calculated in the previous calculation period.

[0102] Multi-method cross-validation unit: Simultaneously run the Z-score algorithm and the quantile method, and perform data elimination when both methods determine an abnormality at the same time;

[0103] Historical data traceability unit: Store the last 100 valid temperature values to construct a reference sequence, and trigger a secondary verification when the correlation coefficient between the newly measured data and the reference sequence is lower than a certain threshold.

[0104] Specifically, the correlation coefficient here refers to the Pearson correlation coefficient r, which quantifies the linear correlation between the junction temperature candidate value and the historical pattern. A coefficient close to 1 indicates a strong positive correlation, close to -1 indicates a strong negative correlation, and close to 0 indicates no linear association.

[0105] Exemplarily, the threshold θ = 0.5. When r < 0.5, it is considered that the current temperature change pattern deviates from the historical law and triggers a secondary verification.

[0106] Here, secondary verification refers to using multiple other units to perform calculations simultaneously to further confirm the authenticity of the anomaly and avoid misjudging a single indicator. If the verification passes, the new data point is marked as a credible anomaly; if it fails, it is attributed to noise / interference and ignored.

[0107] It should be noted that any one or more of the above-mentioned units for filtering outliers can be used. When multiple units are used, there is a close logical connection between these units, forming a closed-loop anomaly detection and correction process. Specifically, dynamic adjustment of detection sensitivity (window expansion), multi-level verification (cross-validation + historical tracing), data repair and feedback form an adaptive anomaly handling chain, ultimately improving the reliability of junction temperature calculation.

[0108] Exemplary:

[0109] The initial statistics-based anomaly detection unit detected an anomaly and continued for 5 times with the sliding window expanded to 1.5 times.

[0110] If an anomaly is flagged by the 3σ test, but not confirmed by both the Z-score and the quantile method → retain the data and only record the event.

[0111] The corrected data (interpolation substitution) enters the historical sequence, and then the reference library is updated, affecting the subsequent correlation coefficient calculation.

[0112] Similarly, in order to further ensure the reliability of the calculated data, in some specific embodiments of the present application, the data processing module can use one or more units such as a sliding window configuration unit, a dynamic weight allocation unit, a standard deviation verification unit, and a collaborative processing unit to perform mean processing.

[0113] Sliding window configuration unit: uses a variable length time window, the window length is configurable, and the number of data points N in the window is programmable;

[0114] Dynamic weight allocation unit: using exponential decay weighting strategy, weight coefficient w i =β N-i , where β is the weight coefficient and N represents the number of data points in the window.

[0115] Standard deviation check unit: When the standard deviation of the data in the window exceeds the set threshold, the window length is automatically shortened to 50% of the original value and recalculated;

[0116] Collaborative processing unit: works in conjunction with the outlier filtering module to replace the marked outliers with the moving average of the first three valid data points or eliminate extreme values.

[0117] It should be noted that any one or more of the above units for mean processing can be used for mean processing. When multiple units are used, there is a close logical relationship between these units. Window configuration, data weighting, fluctuation detection, length adjustment, anomaly handling, data update, and re-weighting calculation form a dynamic optimization chain to ensure the robustness of mean calculation against noise, anomalies, and operating condition changes.

[0118] Exemplarily:

[0119] Initial window length = 100. When the standard deviation check detects that the standard deviation σ exceeds the limit, the window is shortened to 50%.

[0120] The data fluctuation of the shortened window is reduced, and the dynamic weight allocation focuses on more recent data, improving the mean accuracy.

[0121] The collaborative processing unit replaces the outlier with the mean of the first three valid points to further smooth the data within the window.

[0122] In the above embodiments of the present application, the abnormal calculated junction temperature values are filtered to improve the stability of the junction temperature calculation result. Combined with mean processing, the adaptability in a noisy environment can be improved.

[0123] To ensure the safety when using power semiconductor devices, in some specific embodiments of the present application, a safety protection module is adopted. This module receives the junction temperature data calculated by the data processing module and triggers protection when the junction temperature data is higher than the set threshold.

[0124] Specifically, the safety protection module includes a dynamic threshold setting unit and a hierarchical warning trigger unit.

[0125] Dynamic threshold setting unit: Based on the device parameter manual value and historical aging data, comprehensively determine the temperature warning threshold;

[0126] Hierarchical warning trigger unit: Obtain the current device junction temperature from the data processing module. When the junction temperature exceeds the set value of the threshold for N consecutive power frequency cycles, activate the current soft derating control; when the instantaneous value of the junction temperature exceeds the threshold and the change rate remains positive, immediately start the hardware protection circuit.

[0127] Exemplarily, the set value can be 80%.

[0128] In the above embodiments of the present application, through the dynamic threshold setting unit and the hierarchical warning trigger unit, device damage can be effectively prevented.

[0129] When the power semiconductor device ages or the circuit structure changes, there will be a deviation between the measured values of the on-state voltage drop and on-state current and the three-dimensional look-up table. To ensure the accuracy of the three-dimensional relationship model, in some specific embodiments of the present application, a junction temperature calibration module is adopted.

[0130] Specifically, the junction temperature calibration module obtains real-time current, on-state voltage drop, and reference temperature data from the data acquisition module, calibrates the three-dimensional relationship among the junction temperature, on-state voltage drop, and on-state current, and sends the corrected three-dimensional look-up table to the three-dimensional relationship model module.

[0131] Exemplarily, the junction temperature calibration module includes:

[0132] Reference temperature calibration unit: During the initialization stage of the device or when triggered by a preset working condition, the temperature at a certain position of the device is collected in real time through an external temperature sensing device as the reference temperature T ref ;

[0133] Dynamic data acquisition unit: Under the temperature reference, the data acquisition module is used to synchronously acquire on-state voltage drop parameters at different currents to generate an on-state voltage drop-current characteristic curve dataset Q(T ref )

[0134] Model dynamic correction unit: Extract the look-up table or fitting formula in the three-dimensional relationship model at the reference temperature T ref to obtain the cross-sectional function Φ(I, Vce, T = T ref ), introduce a compensation value ΔV(Vce, I) = F(Vce, I), so that the corrected model Φ'(I, Vce, T) = Φ(I, Vce + ΔV(Vce, I), T) satisfies Q(T ref ) ≡ Φ'(I, Vce, T = T ref ) ref .

[0135] Among them, the cross-sectional function represents a two-dimensional sub-model extracted from the three-dimensional model Φ(I, Vce, T) at a fixed reference temperature. F(Vce, I) represents the compensation value generation function, and the input quantities are the on-state voltage drop Vce and the on-state current I. The function form can be: linear function, non-linear function, piecewise function, etc.

[0136] In the above embodiments of the present application, when the three-dimensional look-up table cannot accurately describe the relationship between the junction temperature of the device under test, the on-state current, and the on-state voltage drop, the characteristic points (that is, Q(T ref )) are used to correct the original look-up table, so that the corrected look-up table can accurately represent the actual device characteristics.

[0137] To ensure the accuracy of the junction temperature calculation of the power semiconductor device, in some specific embodiments of the present application, a fault locking module is adopted.

[0138] Specifically: The fault locking module obtains real-time current and on-state voltage drop data from the data acquisition module. When it detects that the characteristic curves of the on-state voltage drop and on-state current of the device deviate from the three-dimensional look-up table by more than a certain proportion, it locks the use of the three-dimensional relationship model module.

[0139] Specifically, when comparing the characteristic curve with the three-dimensional look-up table, the on-state voltage - current data (characteristic curve) collected in real time is matched with the pre-established three-dimensional look-up table (the corresponding relationship between the junction temperature, on-state current, and on-state voltage based on experiments / data manuals), and the deviation value of the on-state voltage under the same junction temperature and on-state current conditions is calculated.

[0140] Combined with practical engineering experience, when the on-state voltage deviation value is within the set proportional range, it indicates that the measured characteristics match the original model.

[0141] Of course, this proportional range is usually set according to the device type and accuracy requirements. More rigorously, it needs to be calibrated through experiments.

[0142] Exemplarily, the proportional range can be set to ±5% or ±10%.

[0143] When the on-state voltage deviation value exceeds the proportion, it indicates that the device is aging, the packaging material is degrading, or the circuit parameters are changing, resulting in the mismatch between the measured characteristics and the original model (three-dimensional look-up table). It is necessary to lock the current look-up table for use to prevent incorrect junction temperature calculation hijacking, and at the same time trigger the calibration module to correct the model to ensure the accuracy of the junction temperature calculation.

[0144] Based on the same technical concept, in some specific embodiments of the present application, an on-line junction temperature calculation method for a power semiconductor device, as Figure 3 shown, includes the following steps:

[0145] Step 1, pre-establish a three-dimensional relationship model of the junction temperature, on-state current, and on-state voltage of the device, and store it as a three-dimensional look-up table to obtain the sensitivity of the junction temperature to the on-state voltage under different on-state currents;

[0146] Step 2, collect the on-state voltage and on-state current of the device during operation in real time;

[0147] Step 3, determine the current sensitive interval according to the real-time change of the on-state current, and screen out the on-state current and on-state voltage located in this interval;

[0148] Step 4, use the screened on-state voltage and on-state current to obtain the corresponding candidate junction temperature value through look-up table or formula calculation;

[0149] Step 5, filter out the outliers and perform mean processing on the candidate junction temperature value to obtain the final junction temperature calculation result.

[0150] In the above examples of the present application, each step can refer to the specific implementation technologies of the corresponding modules / units in the on-line junction temperature calculation system for power semiconductor devices in the above embodiments, which will not be elaborated here.

[0151] In the above embodiments, each preferred feature can be used alone in any one of the embodiments, and can also be used in any combination on the premise of non - conflict. In addition, the parts not described in detail in the embodiments can be implemented by using the prior art.

[0152] As Figure 4 shown, the measured three - dimensional relationship diagram of the junction temperature, conduction current, and conduction voltage drop of a certain type of device was obtained. Subsequently, based on this three - dimensional relationship diagram, the sensitivity - current relationship diagram as Figure 5 shown was drawn. Further, the sensitive interval was determined from the sensitivity - current relationship diagram, and the data was screened using this interval. The effect of data screening is as Figure 6 shown. Before screening, there were a large number of incorrect data points and high noise in the calculated junction temperature, while the junction temperature data after screening was more stable.

[0153] Some specific embodiments of the present application have been described above. It should be understood that the present application is not limited to the above - mentioned specific implementation manners. Those skilled in the art can make various deformations or modifications within the scope of the claims, which does not affect the essence of the present application. The above - mentioned preferred features can be used in any combination on the premise of non - conflict.

Claims

1. An on-line calculation system for the junction temperature of a power semiconductor device, characterized in that Comprising: Three-dimensional relationship model module: A three-dimensional relationship model of the junction temperature, conduction current, and conduction voltage drop of the device is pre-established and stored as a three-dimensional lookup table to obtain the sensitivity of the junction temperature relative to the conduction voltage drop at different conduction currents; Data acquisition module: Real-time acquisition of the conduction voltage drop and conduction current during device operation; Sensitive interval determination module: According to the real-time change of the conduction current, determine the current sensitive interval, and screen out the conduction current and conduction voltage drop within this interval; Temperature calculation module: Use the filtered conduction voltage drop and conduction current to obtain the corresponding junction temperature candidate value through look-up table or formula calculation; Data processing module: Filter out outliers and perform mean processing on the junction temperature candidate value to obtain the final junction temperature calculation result.

2. The on-line junction temperature calculation system for a power semiconductor device according to claim 1, characterized in that, The three-dimensional relationship model module includes: Data unit: Obtain the conduction voltage drop at different junction temperatures and conduction currents through experimental tests or data manuals; Model unit: Perform polynomial interpolation fitting or formula fitting on the relationship between the junction temperature, conduction current, and conduction voltage drop to generate a continuous three-dimensional lookup table; Sensitivity unit: In the three-dimensional lookup table, use different conduction voltage drops and junction temperatures at the same conduction current to calculate the sensitivity.

3. The on-line junction temperature calculation system of a power semiconductor device according to claim 1, characterized in that The data acquisition module measures the conduction voltage drop and conduction current at the same measurement frequency and measurement timing, and the measurement timing is to avoid the transient state at the moment of turning on or off the power semiconductor device. The transient state includes the overshoot, oscillation, or tailing of the device voltage drop and current caused by the parasitic capacitance of the device and the parasitic inductance in the circuit during the switching of the device state.

4. The on-line junction temperature calculation system of a power semiconductor device according to claim 1, characterized in that, The sensitive interval determination module includes: X-axis determination unit: Real-time monitor the waveform of the conduction current, and identify its current peak and several sampling points near the peak; Y-axis determination unit: Combine the sensitivity distribution of the junction temperature to the conduction voltage drop in the three-dimensional lookup table, and select the current interval with a sensitivity higher than the preset threshold; Sensitive interval unit: Combine the sampling points and the current interval to obtain the current sensitive interval.

5. The on-line junction temperature calculation system of a power semiconductor device according to claim 1, characterized in that, The data processing module includes one or more of the following units for outlier processing: - Statistic-based anomaly detection unit: Dynamically set the threshold interval using the 3σ principle, move the sliding window within the threshold interval, and mark the junction temperature candidate value as an anomaly when it deviates from the mean of the sliding window by ±3 times the standard deviation; - Sliding window dynamic update unit: Set a double-threshold mechanism, and automatically expand the length of the sliding window to b times the original value when the number of consecutive occurrences of outliers ≥ a times; - Data correction strategy unit: Use linear interpolation of the forward effective value to replace the marked abnormal junction temperature candidate value; - Multi-method cross-validation unit: Synchronously run the Z-score algorithm and the quantile method on the junction temperature candidate value, and perform data elimination when both methods determine an anomaly at the same time; - Historical data traceability unit: Store the latest N valid temperature values to construct a reference sequence; trigger secondary verification when the correlation coefficient between the junction temperature candidate value and the reference sequence is lower than a certain threshold.

6. The on-line junction temperature calculation system of a power semiconductor device according to claim 1, characterized in that The data processing module includes one or more of the following units for mean processing: - Sliding window configuration unit: uses a variable-length time window, the window length is configurable, and the number of data points N in the window is programmable; - Dynamic weight allocation unit: adopting an exponential decay weighting strategy, with the weight coefficient w i = β N-i , where β is the weight coefficient and i represents the i-th data within the window; -Standard deviation check unit: When the standard deviation of the data in the window exceeds the set threshold, the window length is automatically shortened to half of the original value and recalculated; - Collaborative processing unit: works in conjunction with the outlier filtering module to replace the marked outliers with the moving average of the first three valid data points or eliminate extreme values.

7. The online calculation system for the junction temperature of a power semiconductor device according to claim 1, characterized in that, It also includes a safety protection module, which receives the junction temperature data calculated by the data processing module and triggers protection when the junction temperature data is higher than a set threshold; The security protection module includes: Dynamic threshold setting unit: sets the temperature warning threshold based on the device parameter manual value and historical aging data; Hierarchical early warning trigger unit: obtains the current device junction temperature from the data processing module. When the junction temperature exceeds a certain proportion of the threshold for N consecutive power frequency cycles, the current soft derating control is activated; when the instantaneous value of the junction temperature exceeds the threshold and the rate of change remains positive, the hardware protection circuit is immediately activated.

8. The on-line junction temperature calculation system of a power semiconductor device according to claim 1, characterized in that, Also includes one or both of the following modules: The junction temperature calibration module calibrates the three-dimensional relationship among the junction temperature, the on-state voltage drop and the on-state current when the measured values of the on-state voltage drop and the on-state current deviate from the three-dimensional lookup table, and sends the corrected three-dimensional lookup table to the three-dimensional relationship model module. The fault locking module obtains real-time current and on-state voltage drop data from the data acquisition module. When it detects that the characteristic curves of the device's on-state voltage drop and on-state current differ from the three-dimensional lookup table by more than a certain ratio, the module locks the use of the three-dimensional relationship model module.

9. The on-line junction temperature calculation system of a power semiconductor device according to claim 8, characterized in that The junction temperature calibration module includes: Reference temperature calibration unit: During the initialization phase of the device or when a preset operating condition is triggered, the temperature of a certain position of the device is collected in real time through an external temperature sensor as a reference temperature; Dynamic data acquisition unit: generating a conduction voltage drop-current characteristic curve using the conduction voltage drop parameters synchronously acquired by the data acquisition module at different currents at the reference temperature; Model dynamic correction unit: extracts the interface function of the lookup table or fitting formula in the three-dimensional relationship model at the reference temperature, introduces the compensation value, makes the correction model a model about the compensation value, and its cross-sectional function value at the reference temperature is equal to the value of the on-state voltage drop-current characteristic curve at the reference temperature.

10. An online calculation method for the junction temperature of a power semiconductor device, characterized in that, include: A three-dimensional relationship model of the device's junction temperature, on-state current, and on-state voltage drop is pre-established and stored as a three-dimensional lookup table to obtain the sensitivity of junction temperature to on-state voltage drop under different on-state currents; Real-time acquisition of the on-state voltage drop and on-state current of the device during operation; Determine a current sensitive interval according to the real-time change of the on-current, and screen out the on-current and on-voltage drop within the interval; Using the screened on-state voltage drop and on-state current, a corresponding candidate junction temperature value is obtained by looking up a table or calculating with a formula; The candidate junction temperature values are subjected to outlier filtering and mean processing to obtain a final junction temperature calculation result.

Images