An IGBT Junction Temperature Estimation Method Based on Big Data Analysis

Abstract

The present invention discloses an IGBT junction temperature estimation method based on big data analysis, which relates to the field of power electronics technology. The method includes the following steps: S1: Graphically illustrate the influence of the output current frequency on the IGBT junction temperature; S2: Division and quantization processing of the IGBT junction temperature rise; S3: Junction temperature change trend lines of different modules; S4: Influence of the output current amplitude on the IGBT junction temperature; S5: Graphically illustrate the influence of the bus voltage and the carrier frequency on the IGBT junction temperature rise; S6: Summary of the estimation method and error evaluation. The difference between the present invention and the prior art is that, based on the output current amplitude and frequency, bus voltage, radiator temperature and carrier frequency detected by the frequency converter, and applying the estimation coefficients deduced from the simulation results and a general frequency coefficient table, the IGBT junction temperature can be quickly estimated. This algorithm consists only of addition, subtraction, multiplication and division, occupying very little software code and hardware computing power resources. Under most whole-machine working conditions, the estimation error can meet the requirements of engineering applications.

Description

Technical Field

[0001] The present invention relates to the technical field of power electronics, and particularly to an IGBT junction temperature estimation method based on big data analysis. Background Technique

[0002] In the industry related to power electronics, the chip junction temperature has long been a key indicator for evaluating the thermal stress of power semiconductor devices and defining the safe operating area.

[0003] When multiple boundary conditions appear superimposed, if the control mode is not adjusted in time, the junction temperature of the IGBT may soar to the non-safe operating area, affecting the reliable operation of the product. Although the conventional junction temperature algorithm has high accuracy, it takes a lot of time and cannot meet the requirements of real-time adjustment of the control mode and rapid shutdown protection. Summary of the Invention

[0004] The purpose of the present invention is to provide an IGBT junction temperature estimation method based on big data analysis to solve the problems raised in the above background technique.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An IGBT junction temperature estimation method based on big data analysis, including the following steps:

[0006] S1: Graphically illustrate the influence of the IGBT junction temperature on the output current frequency;

[0007] S2: Division and quantization processing of the IGBT junction temperature rise;

[0008] S3: Junction temperature change trend lines of different modules;

[0009] S4: Influence of the IGBT junction temperature on the output current amplitude;

[0010] S5: Graphically illustrate the influence of the IGBT junction temperature rise on the bus voltage and carrier frequency;

[0011] S6: Summary of the estimation method and error evaluation.

[0012] Furthermore, in the step S1, under various working conditions of the same half-bridge module, when the output current frequency is lower than 25 Hz, the IGBT junction temperature rise will increase with the decrease of the output current frequency, and the change trend is the same.

[0013] Furthermore, in the step S2, the fluctuation component in the chip junction temperature changes regularly with the output current frequency.

[0014] Compared with the prior art, the beneficial effects of the present invention are:

[0015] The IGBT junction temperature peak estimation method based on simulation data analysis is different from the existing technologies in that, according to the output current amplitude, frequency, bus voltage, radiator temperature, and carrier frequency detected by the frequency converter, by applying the estimation coefficients deduced from the simulation results and a general frequency coefficient table, the junction temperature of the IGBT can be quickly estimated. This algorithm consists only of addition, subtraction, multiplication, and division, occupying very little software code and hardware computing power resources. Under most operating conditions of the whole machine, the estimation error can meet the requirements of engineering applications. Description of the Drawings

[0016] Figure 1 It is a graph showing the change trend of the junction temperature rise with the output current frequency of the 200A module of the present invention under various operating conditions of the whole machine;

[0017] Figure 2 It is a graph showing the change trend of the junction temperature rise with the output current frequency of the 450A module of the present invention under various bus voltages;

[0018] Figure 3 It is a graph showing the fluctuation of the virtual junction temperature of the IGBT chip of the present invention at different output current frequencies;

[0019] Figure 4 It is a component diagram for dividing the junction temperature rise of the chip of the present invention;

[0020] Figure 5 It is a diagram of the Infineon 450A module of the present invention;

[0021] Figure 6 It is a diagram of the Fuji 200A module of the present invention;

[0022] Figure 7 It is a K_Fo curve diagram of different modules of the present invention;

[0023] Figure 8 It is a junction temperature rise curve, two fitting schemes and error diagram of the Infineon 450A module of the present invention;

[0024] Figure 9 It is a diagram of the junction temperature rise and fitting line of the Fuji 200A module of the present invention under various carrier frequencies;

[0025] Figure 10 It is a distribution diagram of the newly introduced junction temperature estimation coefficient with the operating conditions of the present invention;

[0026] Figure 11 It is a junction temperature estimation coefficient diagram of the ceramic substrate module of the Fuji 200A module of the present invention;

[0027] Figure 12 It is a junction temperature estimation coefficient diagram of the ceramic substrate module of the Infineon 35A of the present invention;

[0028] Figure 13Summary diagram of the error in estimating the junction temperature of the Fuji 200A half-bridge module according to the present invention;

[0029] Figure 14 Junction temperature estimation diagram under the rated light load current of the whole machine according to the present invention;

[0030] Figure 15 Junction temperature estimation diagram under 1.15 times the rated light load current of the whole machine according to the present invention;

[0031] Figure 16 Junction temperature estimation diagram under 1.3 times the rated light load current of the whole machine according to the present invention;

[0032] Figure 17 Junction temperature estimation diagram under 0.85 times the rated light load current of the whole machine according to the present invention;

[0033] Figure 18 Junction temperature estimation diagram under 0.7 times the rated light load current of the whole machine according to the present invention;

[0034] Figure 19 Value diagram corresponding to the key frequency points according to the present invention;

[0035] Figure 20 Power function diagram fitted with the influence of the bus voltage according to the present invention. Detailed implementation manners

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] As Figures 1 - 20 shown, the present invention provides a technical solution: an IGBT junction temperature estimation method based on big data analysis.

[0038] It mainly includes the following three aspects:

[0039] 1. Introduce a general frequency coefficient table - used to decouple the non-linear influence of the output current frequency on the junction temperature fluctuation component. This is the core of the algorithm and lays the foundation for subsequent quantitative analysis.

[0040] 2. Confirm that the bus voltage linearly affects the slope of the junction temperature estimation coefficient (for the carrier frequency). This eliminates the trouble caused by the use of power function interpolation in the switching loss estimation formula by some manufacturers.

[0041] 3. Confirm that the junction temperature rise can be linearly fitted with the output current - the estimation error meets the engineering application in most whole machine working conditions.

[0042] I. Diagram of the Influence of IGBT Junction Temperature on Output Current Frequency

[0043] Experience shows that when the load motor operates with a heavy load at a low speed, the inverter unit is more likely to be damaged. Therefore, first investigate the variation law of IGBT junction temperature or temperature rise (ΔTj - heatsink) with the output current frequency (Fo).

[0044] Taking the commonly used Fuji 200A (2MBI200XBE120 - 50) and Infineon 450A (FF450R12KT4) half - bridge modules as examples, use the simulation software provided by the manufacturer to obtain the junction temperature rise of the frequency converter under various working conditions (the radiator temperature is uniformly set to 90 degrees), and summarize it into Figure 1 and Figure 2 for analysis in the curves.

[0045] To clearly see the drastic change of the junction temperature rise in the low - frequency band of the output current, Figure 1 and Figure 2 the abscissa in does not use a linear coordinate; it can be seen that for the same half - bridge module under various working conditions, when the output current frequency is lower than 25Hz, the IGBT junction temperature rise will increase with the decrease of the output current frequency, and the change trend is the same; when the output current approaches zero frequency, the chip junction temperature rises to 2 - 3 times that at 25Hz!

[0046] II. Division and Quantification of IGBT Junction Temperature Rise

[0047] Figure 3 is a schematic diagram of the IGBT virtual junction temperature provided by the module manufacturer: when outputting the same amplitude of current at three current frequencies, although the average chip junction temperature is the same and the equivalent loss waveforms only have different periods, the lower the output current frequency, the larger the amplitude and peak value of the virtual junction temperature fluctuation.

[0048] From Figure 3 it can be seen that the chip junction temperature rise can be regarded as consisting of two parts. One is the average chip temperature rise independent of the output current frequency, and the other is the increment caused by the virtual junction temperature fluctuation.

[0049] Use the Infineon online simulation software to investigate the junction temperature rise of the 450A module under the rated current of the whole machine with a light load. Figure 4 The horizontal axis in is the frequency of the bridge - arm output current, and the solid line in the figure is the junction temperature rise obtained by simulation at each frequency point; the long dashed line is the average chip temperature rise obtained by simulation (denoted as ΔT_ac in the following text).

[0050] Figure 4There is also a short dashed line passing through the junction temperature rise points corresponding to 2 Hz and 3 Hz. The intersection of this line with the vertical axis can be regarded as the assumed junction temperature rise when outputting direct current. After subtracting the average junction temperature rise from the ordinate of the intersection point, the increment of the junction temperature rise (ΔT_dc) brought about by the zero-frequency output current is obtained. This is used as a reference benchmark to quantitatively evaluate the change trend of the fluctuating component in the junction temperature rise, that is, it is quantitatively fitted using the following formula Figure 4 The simulation results in

[0051] ΔTj-h = ΔT_ac + ΔT_dc * K_fo / 100 …………②’

[0052] K_fo in formula ②’ is a coefficient table that varies with the output current frequency and can be deduced inversely from the simulation results of each output frequency point. For example, in Figure 4 :

[0053] K_Fo(@2Hz) = (ΔTj-h - ΔT_ac) / ΔT_dc * 100 = (29.66 - 17.4) / 17 * 100 = 72

[0054] K_Fo(@3Hz) = (ΔTj-h - ΔT_ac) / ΔT_dc * 100 = (27.4 - 17.4) / 17 * 100 = 59

[0055] K_Fo(@10Hz) = (ΔTj-h - ΔT_ac) / ΔT_dc * 100 = (21.4 - 17.4) / 17 * 100 = 24

[0056] ……

[0057] Some people may question why Figure 4 when determining the reference quantity (ΔT_dc), the dashed line passing through the temperature rise points of 2 Hz and 3 Hz is selected in? And the absolute values of the reference quantity and the frequency coefficient (K_fo) deduced from dashed lines with different slopes will be different. In fact, only the proportional relationship between the frequency coefficients can reflect the change trend of the fluctuating component of the junction temperature rise - the absolute value of the reference quantity is not crucial. However, in order to compare the frequency coefficient tables under different working conditions, once the drawing method of the dashed line is determined, it should not be changed

[0058] If the fluctuating component in the IGBT junction temperature rise indeed changes regularly with the output current frequency, then the frequency coefficient tables deduced under different whole-machine working conditions should be the same. Select other whole-machine working conditions and other modules, repeat the above quantitative process, and obtain new K_Fo coefficient curves as shown in Figure 5 and Figure 6 shown

[0059] Figure 5 and Figure 6All show that the frequency coefficient curves of the same module under different working conditions are approximately coincident in the frequency band of 0 - 15 Hz. That is, in the low-frequency band where the IGBT junction temperature changes drastically, the trend of junction temperature change has nothing to do with the overall machine working conditions other than the output current frequency; that is to say, the fluctuation component in the chip junction temperature changes regularly with the output current frequency.

[0060] III. Junction temperature change trend lines of different modules

[0061] Continuing to compare IGBT modules with different current capacities, it is found that for most general-purpose modules with copper substrates, the frequency coefficient curves deduced from their junction temperature simulation results are approximately coincident in the range of 0.5 - 15 Hz, as Figure 7 shown.

[0062] "The frequency coefficient curves of general-purpose modules with copper substrates are approximately coincident" can be understood as follows: The general-purpose modules launched by each manufacturer in the same era have similar internal chip materials, junction layer materials and thicknesses, and processing technologies, resulting in similar trends in the thermal impedance curves of the chips - this is the fundamental reason for the drastic change in the chip junction temperature in the low-frequency band and the approximately coincident change trends. The different current capacities of the modules mainly affect the chip area or quantity and do not change the trend of the thermal impedance curve.

[0063] The simulation verification results also show that for modules with different structural packages, such as low-power modules without copper substrates, some 600A modules with special process support, and single-tube packaged IGBTs, their frequency coefficient curves deviate significantly from those of general-purpose copper substrate modules ( Figure 7 ).

[0064] Since the frequency coefficients of most copper substrate modules are approximately coincident in the range of 0.5 - 15 Hz, we can use a general frequency coefficient table to estimate the junction temperature; to improve the accuracy of the estimation, in the frequency band of 0.5 - 5.0 Hz where the junction temperature changes drastically, the frequency coefficient can be simulated and calculated every 0.1 Hz, and every 1 - 2 Hz above 5 Hz; from the previous simulation results ( Figure 1 , Figure 2 and Figure 4 ), the fluctuation component of the junction temperature has tended to be stable in the frequency band above 15 Hz. Even if the relative error of the deduced frequency coefficient becomes larger at this time, it does not affect the use of formula ②'.

[0065] Figure 19 Lists the values corresponding to the key frequency points in the general frequency coefficient table.

[0066] If it is confirmed that the frequency coefficient of a certain module is approximate to the general frequency coefficient table, it is only necessary to simulate the junction temperature rise at two frequency points (such as 3 Hz and 10 Hz) under specific working conditions, and refer to the general frequency coefficient table and formula ②', then the average temperature rise (ΔT_ac) and the fluctuation reference quantity (ΔT_dc) of the chip under this working condition can be deduced, and based on this, the junction temperature rise within the entire frequency band under this working condition can be predicted.

[0067] To determine whether the frequency coefficient table of the module is approximate to the general frequency coefficient table, the following simple method can be adopted: simulate the junction temperature rise at six frequency points (such as 1 Hz, 2 Hz, 3 Hz, 5 Hz, 10 Hz, and 15 Hz) under a certain working condition; first assume that the module is the same as the general frequency coefficient table at the two frequency points of 3 Hz and 10 Hz, substitute the corresponding simulation results into formula ②', and calculate ΔT_ac and ΔT_dc under this working condition; then, based on the junction temperature rise at the other four frequency points (1 Hz, 2 Hz, 5 Hz, and 15 Hz), inversely deduce the corresponding frequency coefficients and compare them with the general frequency coefficient table. If the errors at the four frequency points do not exceed 10%, then the module is applicable to the general frequency coefficient table.

[0068] IV. Influence of IGBT Junction Temperature by Output Current Amplitude

[0069] The theoretical calculation formula of IGBT junction temperature is as follows:

[0070] The loss P in formula ①' is Figure 3 an equivalent periodic sinusoidal half-wave, and the peak power is 3.14 times the average power. The thermal impedance is a single-pulse power response curve provided by the manufacturer and can be fitted by summing multiple exponential functions.

[0071] First, exclude the influence of the output current frequency and assume that the thermal impedance amplitude is fixed. Then formula ①' can be arranged as:

[0072] ΔTj-h = P(t)*Z(t) ∝ P(t) ∝ P_avg = Pcond + Psw

[0073] The calculation formulas for the average values of conduction loss and switching loss provided by most IGBT manufacturers are as follows:

[0074]

[0075]

[0076] In the above formula, except that the loss of the internal leads of the module and its contact resistance (rce) is quadratic with the output current, both the ideal conduction loss and the switching loss of the IGBT are proportional to the output current of the bridge arm. If the proportion of the lead resistance loss in the above formula is low, or the error after linear fitting is acceptable, then: the chip loss and the junction temperature rise are proportional to the amplitude of the output current. After referring to the thermal resistance label (Rth) and introducing a new estimation coefficient (R’th), formula ① is obtained:

[0077] ΔTj_h≈R’th*Io……①

[0078] Simulating the junction temperature rise of the Infineon 450A module at different output currents, the Figure 8 quasi-parabolic (solid) line in is obtained; it shows that the loss and thermal stress contributed by the internal lead resistance of the module at this time (which is a quadratic function of the output current) cannot be ignored.

[0079] Figure 8 There are also two fitting schemes for the junction temperature rise in: one is the linear (dash-dotted line) fitting with the smallest comprehensive error, and the other is the (dotted line) scheme of connecting the zero point to the rated current junction temperature point; the errors of both fitting schemes are less than 2.5 degrees, but the linear fitting has a lower estimation when the current is relatively large; considering that the underestimated scheme will delay the protection of the chip, it is finally decided to adopt the second (dotted line) scheme to fit the junction temperature rise.

[0080] Figure 9 The solid line in is the junction temperature rise of the Fuji 200A module. When operating at a high carrier frequency, the fitting line almost coincides with the simulation result (the thinnest solid line at the top) - the smaller the rated output current determined by the junction temperature rise limit, the lower the proportion of the internal lead resistance loss of the module, and the closer the junction temperature rise curve is to a straight line; when operating at a low carrier frequency, the corresponding rated output current increases, the proportion of the internal lead resistance loss of the chip increases, and the junction temperature rise is closer to a parabola (the thickest solid line at the bottom). Using this fitting method for the IGBT junction temperature rise, the maximum error will occur at 50% of the full machine rated current: Figure 8 in it does not exceed 2.5 degrees, Figure 9 in it does not exceed 2 degrees and is even lower under high carrier frequency conditions; in the 85% - 100% rated current range, the fitting error does not exceed 7%.

[0081] If it can be confirmed that in the 85% - 115% rated current range (corresponding to the rated output current range of approximately 115% - 154% for heavy-duty models) and under various combined operating conditions, the fitting error does not exceed 10%, this fitting method can be used for engineering estimation (such as Figures 13 - 18 ).

[0082] V. Illustrating the Influence of the Bus Voltage and Carrier Frequency on the IGBT Junction Temperature Rise

[0083] From the calculation formula of the average loss of IGBT, the bus voltage and the switching frequency (Vcc and fsw in the following formula) only affect the switching loss. Some manufacturers fit the influence caused by the bus voltage into a power function, which seems difficult to handle, such as Figure 20 shown.

[0084] —— Infineon believes that there may be an error of 20%

[0085] —— The interpolation formula recommended by Semikron

[0086] Review Figure 1 In the junction temperature rise curve in , when the output current is 150 A and the switching frequency changes from 1 to 3 kHz, the corresponding junction temperature rise curve moves up regularly; Figure 2 The junction temperature rise curve in also moves up regularly with the increase of the bus voltage. It may be easier to find the internal law by directly analyzing the simulation results.

[0087] It has been confirmed in the previous section that the junction temperature rise and the output current can be linearly fitted, that is: ΔTj-h≈R’th*Io……① - Although there is an error, the influence of the output current is quickly decoupled; in the third section, the junction temperature rise is divided into two parts, one is the average junction temperature rise, and the other is the increment caused by the fluctuation, that is: ΔTj-h=ΔT_ac+ΔT_dc*K_fo / 100……②’ - A general frequency coefficient table is introduced to decouple the influence of the output current frequency; after substituting formula ① into formula ②’, we can get:

[0088] R’th=ΔT_ac / Io+ΔT_dc*K_fo / 100 / Io

[0089] The average junction temperature rise is originally the product of the average loss and the thermal resistance (ΔT_ac=P_avg*Rth∝Io); the fluctuation component of the junction temperature rise is also positively correlated with the peak power of the equivalent sine half-wave (ΔT_dc∝P∝P_avg∝Io); after introducing the new junction temperature estimation coefficients R’th_ac(=ΔT_ac / Io) and R’th_dc(=ΔT_dc / Io), the above formula is rewritten as:

[0090] R’th=R’th_ac+R’th_dc*K_fo / 100……②

[0091] Since the frequency coefficient table (K_fo) in formula ② has nothing to do with the operating conditions other than the output current frequency (the second section), the changes in the switching frequency and the bus voltage only affect the two newly introduced estimation coefficients Rth’_ac and R’th_dc.

[0092] First, simulate the junction temperature rise at two frequency points. Then, with the help of the general frequency coefficient table and formula ②’, after calculating ΔT_ac and ΔT_dc corresponding to this working condition, the newly introduced junction temperature estimation coefficient can be obtained (only differing by a coefficient Io). Simulate the junction temperature at different bus voltages and 4 - 5 carrier frequencies most likely to occur in the whole machine (select the rated output current corresponding to each carrier frequency for the current amplitude). After calculating the estimation coefficients (Rth’_ac and R’th_dc) under each working condition from the simulation results, summarize them into Figure 10 in.

[0093] Figure 10 It shows that at any bus voltage, the two junction temperature estimation coefficients are linearly distributed with respect to the carrier frequency; at different bus voltages, the extension lines of the two estimation coefficient lines have a common intersection point on the vertical axis; in addition, the slope of the estimation coefficient line changes proportionally with the bus voltage!

[0094] From Figure 10 it is easy to write the linear function analytical expressions for the two junction temperature estimation coefficients:

[0095] R’th_ac(fsw,v) = R’th_ac(@Fsw = 0) + K_v_ac(v) * Fsw... Formula ③

[0096] R’th_dc(fsw,v) = R’th_dc(@Fsw = 0) + K_v_dc(v) * Fsw... Formula ③’

[0097] K_v_ac(v) = K_v_ac(@520V) + (v - 520) / 50 * K_Δv_ac... Formula ④

[0098] K_v_dc(v) = K_v_dc(@520V) + (v - 520) / 50 * K_Δv_dc... Formula ④’

[0099] Formula ④ is used to calculate the influence of the change in bus voltage on the slope of the connection line of the estimation coefficient (R’th_ac), and K_Δv_ac is the variation coefficient of the slope of the connection line of the estimation coefficient with respect to the bus voltage (for every 50V increase).

[0100] Under various bus voltage conditions, the connection lines of the estimation coefficients have a common intersection point with the vertical axis. This phenomenon can find a physical explanation: when the carrier frequency approaches zero, the chip loss and junction temperature rise are mainly determined by the conduction loss and are independent of the bus voltage (which only affects the switching loss). Therefore, the extension lines of the estimation coefficients under each bus voltage have a common intersection point at zero frequency.

[0101] If the expression of the average chip loss is brought into the formula for calculating the average junction temperature rise, after simplification, an analytical expression consistent with the form of the junction temperature estimation formula ③ can be obtained (omitted); this can not only obtain the expression of each junction temperature estimation coefficient in formula ③, but also help understand the role of each estimation coefficient, for example:

[0102] R'th_ac(@Fsw=0)=Pcond×R th / Ip——Conduction loss affects the average junction temperature coefficient K_v_ac(v) = P sw ×R th / (Ip×F sw )——The coefficient of switching loss affecting the average junction temperature

[0103] ——The coefficient of switching loss affecting the average junction temperature at rated 520V bus voltage

[0104] …

[0105] Then, I also verified the Fuji 200A module and the Infineon 35A ceramic substrate module. Their junction temperature estimation coefficients showed the same pattern, such as Figure 11 , Figure 12 shown.

[0106] At this point, the intrinsic connection between the junction temperature (rise) of the IGBT chip and the main working conditions of the module has been solved! The estimated coefficients of the average value component of the junction temperature and the fluctuation reference quantity (Rth'_ac and R'th_dc) are both linear functions of the carrier frequency (Formula ③ and ③'). Their values ​​at zero carrier frequency are mainly determined by the conduction loss of the IGBT - which can be inferred from the graphical estimated coefficient connection line; the bus voltage linearly affects the slope of the linear function (Formula ④ and ④'); most copper-based modules can use a universal frequency coefficient table (K_Fo in Formula ②) to decouple the influence of the output current frequency on the fluctuation component in the junction temperature; the junction temperature rise and the bridge arm output current are approximately linearly fitted (Formula ①).

[0107] 6. Summary of Estimation Methods and Error Assessment

[0108] The formulas required to estimate the IGBT junction temperature rise are summarized as follows:

[0109] ΔTj-h≈R'th*Io……①

[0110] R'th=R'th_ac+R'th_dc*K_fo / 100……②

[0111] R'th_ac(fsw,v)=R'th_ac(@Fsw=0)+K_v_ac(v)*Fsw……③

[0112] R’th_dc(fsw,v) = R’th_dc(@Fsw = 0) + K_v_dc(v) * Fsw……③’

[0113] K_v_ac(v) = K_v_ac(@520V) + (v - 520) / 50 * K_Δv_ac……④

[0114] K_v_dc(v) = K_v_dc(@520V) + (v - 520) / 50 * K_Δv_dc……④’

[0115] Before applying the above estimation formula, it is necessary to confirm that the frequency coefficient of the module is approximate to the general frequency coefficient table. The specific method can be referred to in Section 3; then, according to the junction temperature limit and simulation, determine the rated output current corresponding to 4 - 5 common carrier frequencies; taking these rated output currents and the common bus voltage (520V) as the basis points, simulate and deduce the six estimation coefficients required in Formulas ③ and ④.

[0116] For example, the estimation coefficients deduced from the simulation results of Fuji 200A module are as follows:

[0117] R’th_ac(fsw,v) = R’th_ac(@Fsw = 0) + K_v_ac(v) * Fsw……Formula ③

[0118] = 0.07873 + [0.0286 + (v - 520) / 50 * 0.00271] * Fsw

[0119] R’th_dc(fsw,v) = R’th_dc(@Fsw = 0) + K_v_dc(v) * Fsw……Formula ③’

[0120] = 0.06332 + [0.0411 + (v - 520) / 50 * 0.00456] * Fsw

[0121] Applying the above formula to estimate the junction temperature of the 200A module under various whole - machine working conditions, it is quickly found that when the boundary conditions are superimposed, the IGBT junction temperature rise may exceed the limit of fifty degrees within the full - current range, and there are even cases where it exceeds one hundred degrees, which highlights the necessity of junction temperature estimation!

[0122] Among them, according to Figures 13 - 18 , the summarized estimation error shows that even considering the comprehensive influence of the bus voltage, carrier frequency, and current frequency changes, within the range of 0.85 - 1.15 times the whole - machine light - load current with a relatively high occurrence probability, the error of this algorithm is less than 10% and meets the requirements of engineering applications; within the range of 0.7 - 1.3 times the whole - machine light - load current, the error is also less than 15%.

[0123] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An IGBT junction temperature estimation method based on big data analysis, characterized in that, It includes the following steps: S1: Illustrate the influence of the output current frequency Fo on the IGBT junction temperature. Among them, in the same half-bridge module under various working conditions, when the output current frequency is lower than 25 Hz, the IGBT junction temperature rise will increase with the decrease of the output current frequency, and the change trend is the same; S2: Division and quantization processing of the IGBT junction temperature rise. Divide the IGBT junction temperature rise ΔTj-h into the average chip temperature rise component ΔT_ac independent of the output current frequency and the increment ΔT_dc caused by the virtual junction temperature fluctuation, and perform quantization processing through the formula ΔTj-h = ΔT_ac + ΔT_dc * K_fo / 100. Among them, K_fo is the general frequency coefficient table, which is deduced from the simulation results of each output frequency point; S3: Verify the junction temperature change trend lines of different modules, and confirm that the frequency coefficient curves of the general module with a copper substrate are approximately coincident in the range of Fo equal to 0.5 - 15 HZ, and use the K_fo general frequency coefficient table to decouple the non-linear influence of Fo on the ΔTj-h fluctuation component; S4: Establish an approximate relationship between ΔTj-h and the output current amplitude Io through linear fitting, ΔTj_h ≈ R’th * Io, where R’th is the junction temperature estimation coefficient; S5: Analyze the influence of the bus voltage and the carrier frequency on ΔTj-h, and determine that the average temperature rise component coefficient R’th_ac and the fluctuation reference quantity coefficient R’th_dc of the estimation coefficient are linear functions of the carrier frequency, and their slopes are linearly adjusted by the bus voltage, and are integrated through the formula R’th = R’th_ac + R’th_dc * K_fo / 100; S6: Based on ΔTj_h ≈ R’th * Io and the correlation coefficients R’th_ac, R’th_dc, K_fo, combined with the general frequency coefficient table and the key working condition simulation data, realize the rapid estimation of the junction temperature.

Images