IGBT junction temperature real-time decoupling modeling method and system based on dynamic electric heating compensation

Abstract

The invention discloses an IGBT junction temperature real-time decoupling modeling method and system based on dynamic electric heating compensation, and relates to the technical field of state monitoring of power electronic devices. At present, junction temperature estimation is not accurate due to IGBT on-state voltage drop measurement distortion and chip internal thermal gradient in a high-voltage environment. The method comprises the following steps: in an offline stage, constructing an ideal electrothermal characteristic database of an IGBT, a dynamic temperature gradient compensation database in a chip and Foster thermal network model parameters; in the on-line stage, synchronous sampling and preprocessing are carried out on IGBT module operation parameters; combining online real-time parameters with an offline database by using an adaptive Kalman filter, and iteratively and cooperatively estimating the real-time junction temperature and loss of the IGBT; and finally, outputting and applying the estimated accurate junction temperature and loss data. According to the technical scheme, the hot spot effect in the chip is accurately decoupled, the junction temperature estimation precision and the robustness of online monitoring are improved, and core data support is provided for fine thermal management and early fault early warning of the flexible direct current converter valve.

Description

Technical Field

[0001] The present invention relates to the technical field of power electronic device state monitoring, and in particular to a real-time decoupling modeling method and system for IGBT junction temperature based on dynamic electrothermal compensation. Background Art

[0002] The modular multilevel converter (MMC) in a flexible HVDC system is a key component for achieving high-voltage, high-capacity HVDC transmission. The operating status of its core power device, the insulated-gate bipolar transistor (IGBT), is crucial to the reliability and lifespan of the converter valve. IGBTs generate power losses during operation, causing their junction temperature to rise. Excessive junction temperature fluctuations under cyclic stress are the primary causes of IGBT failures, including thermal fatigue, solder layer degradation, and bond wire detachment. Therefore, accurate, real-time, online monitoring and modeling of IGBT junction temperature are crucial for ensuring safe and reliable operation of the converter valve, extending device lifespan, and enabling predictive maintenance.

[0003] At present, the online estimation method of IGBT junction temperature mainly relies on its electrical parameters (such as collector-emitter on-state voltage drop V CE(on) However, in the high-voltage and complex electromagnetic environment of the flexible DC converter valve, it is difficult to directly measure the IGBT's V CE(on) There are many challenges: First, the high common mode voltage and high frequency noise inside the converter valve V CE(on) This kind of weak signal acquisition at the millivolt level causes serious interference, resulting in large measurement errors, which in turn seriously affects the accuracy of junction temperature estimation; secondly, in the actual operation of the IGBT chip, due to factors such as uneven current distribution and differences in thermal resistance paths, uneven temperature distribution often occurs, that is, local hot spots and temperature gradients appear. V CE(on) The measured value does not simply represent the overall or maximum junction temperature of the chip, but is affected by the local temperature. As a result, traditional methods based on single-point measurement or simple lookup tables cannot achieve accurate "decoupling" of the junction temperature and cannot accurately reflect the actual junction temperature status. In addition, existing junction temperature estimation models (such as those based on simplified thermal network models or static lookup tables) often use static parameters and lack the real-time compensation and adaptive capabilities for the dynamic electrothermal coupling characteristics and internal temperature non-uniformity within the IGBT chip. As a result, under complex operating conditions, especially when the hot spot effect intensifies, the estimation accuracy is significantly reduced and the robustness is insufficient, which cannot meet the needs of high-reliability systems for refined junction temperature monitoring.

[0004] In summary, there is an urgent need for an advanced method and system that can overcome the measurement challenges in high-voltage environments, achieve dynamic compensation for temperature gradients within the chip, and provide high-precision, real-time decoupled junction temperature estimation, so as to effectively improve the operational reliability and intelligent operation and maintenance level of the flexible DC transmission system. Summary of the Invention

[0005] The technical problem to be solved and the technical task to be addressed by the present invention are to improve and enhance existing technical solutions by providing a method and system for real-time decoupling modeling of IGBT junction temperature based on dynamic electrothermal compensation, thereby achieving high-precision real-time decoupling modeling of IGBT junction temperature. To this end, the present invention adopts the following technical solutions.

[0006] A high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation, comprising: S1: Offline construction of the electrothermal characteristics and compensation database of the IGBT module, including: Construct a database of ideal electrothermal characteristics of IGBT body and anti-parallel diode; Build a dynamic temperature gradient compensation database inside the IGBT chip; Extracting Foster thermal network model parameters of the IGBT module; S2: Online real-time high-precision synchronous sampling and pre-processing of IGBT module operating parameters, including high-precision isolated sampling of the IGBT collector-emitter on-state voltage drop, and multi-channel high-speed synchronous sampling and signal pre-processing of the on-state voltage drop, collector current, and module case temperature; S3: Based on online real-time operating parameters and an offline database, it performs junction temperature-loss electrothermal coupling iterative calculation and dynamic compensation; S4: Output and apply the junction temperature and loss data obtained through iterative calculation; applications include online thermal management, operating condition monitoring, or fault warning.

[0007] By integrating offline modeling with online adaptive filtering, this method achieves high-precision, real-time estimation of IGBT junction temperature under complex operating conditions. This effectively decouples the chip's internal hotspot effects, significantly improving the accuracy and robustness of electrothermal state monitoring and providing a critical data foundation for intelligent operation, maintenance, and reliability assessment of converter valves. Specifically, by performing high-voltage, isolated acquisition of the collector-emitter on-state voltage drop, this method overcomes measurement distortion caused by the high common-mode voltage and electromagnetic interference of the converter valve, providing a reliable data foundation for online estimation. The impact of chip hotspot effects on the on-state voltage drop is pre-quantified. During the online phase, measurement deviations are dynamically compensated based on real-time operating conditions, ensuring that the on-state voltage drop measurement truly reflects the overall junction temperature rather than the temperature of a local hotspot. Junction temperature estimation and loss calculation are coupled into a closed-loop system, and real-time interactive corrections are implemented to improve model self-consistency, avoiding the error accumulation of traditional static models under varying operating conditions. Aligning the sampling timestamps of the on-state voltage drop, current, and case temperature ensures the temporal and spatial consistency of electrothermal parameters during transient processes, meeting microsecond-level real-time performance requirements.

[0008] As a preferred technical means: constructing the ideal electrothermal characteristics database in step S1 specifically includes: obtaining the relationship between the on-state voltage drop and current of the IGBT and the anti-parallel diode at multiple preset uniform junction temperature points through pulse testing, and using piecewise polynomial or logarithmic fitting to establish a three-dimensional mapping function of the on-state voltage drop as the current and junction temperature change.

[0009] Measuring the on-state voltage drop at extremely short pulse widths suppresses chip self-heating, ensuring that the data strictly reflects the intrinsic electrical characteristics at a preset uniform junction temperature, avoiding the temperature rise distortion caused by traditional DC testing. To address the strong nonlinear characteristics of the IGBT's on-state voltage drop, which varies exponentially with current and is negatively correlated with junction temperature, a piecewise polynomial or logarithmic function is used for three-dimensional mapping. This method better reflects the physical nature of semiconductor devices than linear interpolation or simple table lookup methods, significantly improving the accuracy of function expression. Through systematic testing at multiple junction temperature points, a continuous function of current and junction temperature is constructed, overcoming the limitation of discrete data points in representing intermediate states and providing a complete mathematical benchmark for online real-time decoupling calculations.

[0010] As a preferred technical means: constructing a dynamic temperature gradient compensation database in step S1 specifically includes: using multi-physics field finite element simulation software to establish a three-dimensional electrothermal coupling model of the IGBT module, simulating the transient temperature field distribution of the chip under different working conditions; extracting the average junction temperature, maximum junction temperature and local junction temperature of the chip at the measurement point; defining the temperature gradient compensation amount as the difference between the simulated on-state voltage drop considering temperature non-uniformity and the ideal on-state voltage drop in the ideal electrothermal characteristic database, and establishing a mapping relationship between the compensation amount and the current, average junction temperature and thermal gradient characteristic quantity.

[0011] Three-dimensional electrothermal coupling simulation replaces physical probes. Under the constraints of chip structures that make it difficult to deploy temperature sensors, finite element simulation is used to accurately reconstruct the transient temperature field distribution. This directly obtains the maximum junction temperature and local measurement point temperatures, which are unmeasurable using traditional methods, revealing microscopic thermal inhomogeneities. Abstract thermal gradients are converted into quantifiable voltage compensation quantities, establishing a physical causal relationship for compensation. The compensation quantities are correlated with current, average junction temperature, and thermal gradient characteristics to form a compensation rule library, providing a theoretically complete input-output relationship for online dynamic compensation.

[0012] As a preferred technical means: in the Foster thermal network model parameters extracted in step S1, the transient thermal impedance is characterized by a combination of an N-order thermal resistance and a corresponding thermal time constant, where N is the model order.

[0013] Through the combination of multi-order thermal resistance and thermal time constant, the thermal response process at different time scales can be fully covered, providing an accurate heat transfer model foundation that conforms to physical laws for online real-time junction temperature estimation, ensuring the accuracy of the prediction of dynamic changes in junction temperature.

[0014] As a preferred technical means: in step S2, a high-voltage isolated differential amplifier circuit is used to instantaneously collect the collector-emitter on-state voltage drop; the rising edge or falling edge of the IGBT gate drive signal is used as the sampling trigger point, and the on-state voltage drop, collector current and module case temperature are synchronously collected through a multi-channel high-speed analog-to-digital converter; the collected data is digitally filtered and time stamp aligned to ensure time synchronization.

[0015] The use of a high-voltage isolated differential amplifier circuit effectively protects against high-voltage common-mode interference and high-frequency electromagnetic noise during IGBT operation, capturing millivolt-level collector-emitter on-state voltage drop transient signals and addressing the issue of insufficient measurement accuracy for weak electrical parameters in high-voltage environments. Using the rising or falling edge of the IGBT gate drive signal as the sampling trigger point, the critical moment of device turn-on / off can be precisely locked, ensuring that the collected on-state voltage drop, collector current, and module case temperature truly reflect the device characteristics under the same operating state, avoiding parameter mismatches caused by sampling timing deviations. The synchronous acquisition, digital filtering, and timestamp alignment of multi-channel high-speed analog-to-digital converters eliminate transient noise and measurement glitches, while strictly ensuring the temporal consistency of on-state voltage drop, current, and case temperature. This provides high-quality, time-synchronized, noise-suppressed input data for subsequent junction temperature-loss coupling calculations, laying the foundation for accurate modeling.

[0016] As a preferred technical means: Step S3 includes: Construct a Kalman filter state space model, with junction temperature as the state vector, Foster thermal network model as the state transition equation, and on-state voltage drop as the observation value; Iterate in each sampling period: a) State prediction: Based on the junction temperature, current, and case temperature at the previous moment, the current junction temperature is predicted using the Foster model. b) Observation update: i. Obtaining the temperature gradient compensation amount from the dynamic temperature gradient compensation database based on the current and predicted junction temperature; ii. Dynamically adjust the measurement noise covariance matrix so that it increases as the temperature gradient compensation amount increases; iii. Subtract the sum of the ideal on-state voltage drop and the temperature gradient compensation amount from the real-time on-state voltage drop measurement value to obtain the compensated observation value; iv. Update the junction temperature estimate in combination with the Kalman gain; c) Output the real-time junction temperature after dynamic compensation and recalculated loss.

[0017] The sum of the ideal on-state voltage drop and the temperature gradient compensation is subtracted from the real-time measured on-state voltage drop to eliminate systematic measurement bias caused by local chip hotspots, ensuring that the observed value truly reflects the overall junction temperature state. The measurement noise covariance matrix is ​​adjusted in real time based on the compensation amount, achieving an intelligent balance between model predictions and measured values, improving adaptability to varying operating conditions. The output junction temperature is fed back into the loss calculation in real time, and the updated losses are re-entered into the Foster model for prediction in the next cycle, forming a self-correcting loop to prevent error accumulation. This ensures that the junction temperature estimation and loss calculation can feedback and be corrected in real time, so that the output results can track the dynamic electrothermal state of the IGBT in real time, meeting the real-time requirements of online monitoring.

[0018] As a preferred technical means: in step S3, Dynamically adjusting the measurement noise covariance matrix includes: querying a dynamic temperature gradient compensation database based on the current current and the predicted junction temperature, and linearly or nonlinearly increasing the value of the measurement noise covariance matrix when the absolute value of the temperature gradient compensation amount increases; Loss recalculation includes: based on the updated junction temperature estimate, combined with the real-time current and switching frequency, the loss is calculated using the conduction loss formula and the switching energy formula.

[0019] The design of dynamically adjusting the measurement noise covariance matrix adaptively adjusts the trust weight of the on-state voltage drop measurement based on the actual internal chip temperature gradient (reflected by the temperature gradient compensation). As the absolute value of the temperature gradient compensation increases, the linear or nonlinear increase in the measurement noise covariance matrix reduces the weight of the measured value in the junction temperature estimation, reducing errors caused by interference from local hotspots. This enhances the stability and anti-interference capability of the junction temperature estimation under complex operating conditions.

[0020] Recalculating losses based on the updated junction temperature estimate ensures real-time coordination between losses and junction temperature. Because losses and junction temperature are strongly coupled, using the latest junction temperature to calculate losses ensures loss results more closely match the device's actual operating conditions. Accurate loss data, in turn, provides reliable input for the next round of junction temperature predictions, forming a dynamic closed-loop correction between junction temperature and losses, further improving the consistency and accuracy of both estimates.

[0021] Another technical solution of the present invention is: a high-precision real-time decoupling modeling system for IGBT junction temperature based on dynamic electrothermal compensation, which includes: An offline modeling module is used to build an electrothermal characteristics and compensation database for the IGBT module offline, including: building an ideal electrothermal characteristics database for the IGBT body and anti-parallel diode, building an internal dynamic temperature gradient compensation database for the IGBT chip, and extracting the Foster thermal network model parameters of the IGBT module; Online sampling and preprocessing module, used for online real-time high-precision synchronous sampling and preprocessing of IGBT module operating parameters, including: high-precision isolated sampling of IGBT collector-emitter on-state voltage drop; multi-channel high-speed synchronous sampling and signal preprocessing of on-state voltage drop, collector current and module case temperature; an electrothermal coupling estimation module, connected to the offline modeling module and the online sampling and preprocessing module, for performing junction temperature-loss electrothermal coupling iterative calculation and dynamic compensation based on online real-time operating parameters and an offline constructed database; The data output and application module is used to output and apply the junction temperature and loss data obtained by iterative calculation, and the applications include online thermal management, working condition monitoring or fault warning.

[0022] The system's offline modeling module pre-builds three core databases: ideal electrothermal characteristics of IGBTs, dynamic temperature gradient compensation, and Foster thermal network model parameters. This provides comprehensive and accurate offline basic data support for online estimation. The online sampling and preprocessing module, specifically designed for parameter measurement in high-voltage environments, implements high-precision isolated sampling of on-state voltage drop and simultaneous multi-parameter preprocessing, ensuring input data accuracy. The electrothermal coupling estimation module connects offline data with online parameters, performing dynamic coupling calculations and compensation. The data output and application module directly connects to actual operation and maintenance needs, enabling effective data application. This creates a complete, highly systematic closed loop of "offline modeling - online acquisition - dynamic estimation - actual application." The online sampling and preprocessing module, utilizing high-precision isolated sampling technology and multi-channel simultaneous processing, accurately captures weak signals such as millivolt-level on-state voltage drop in complex high-voltage electromagnetic environments. Signal preprocessing also ensures strict temporal synchronization of on-state voltage drop, current, and case temperature, providing high-quality input data for subsequent estimation, overcoming the issues of susceptibility to interference and insufficient accuracy in high-voltage measurements. The electrothermal coupling estimation module, based on an offline database and online real-time parameters, effectively corrects the impact of internal chip temperature gradients on measurements through iterative calculations and dynamic compensation. This decouples local hotspot effects, ensuring junction temperature and loss estimates are closer to the actual IGBT operating state, improving estimation accuracy under dynamic conditions. The data output and application module applies accurate junction temperature and loss data to practical scenarios such as online thermal management, operating condition monitoring, and fault warning. This directly supports the refined operation and maintenance of flexible DC converter valves, enhancing the system's engineering practicality and reliability.

[0023] As a preferred technical means: the offline modeling module includes: an industrial computer or workstation equipped with simulation software and data analysis tools, which is used for characteristic calibration, simulation and database generation of IGBT modules.

[0024] Industrial computers or workstations possess powerful computing capabilities, efficiently supporting multi-physics finite element simulations of refined three-dimensional electrothermal coupled models of IGBT modules, as well as the processing and analysis of large amounts of test data, ensuring the efficiency of the offline modeling process and the smooth completion of complex computational tasks. Simulation software (such as multi-physics coupled simulation tools) accurately simulates the electrothermal characteristics and temperature gradient distribution within the IGBT chip. Professional data analysis tools (such as data fitting and regression analysis software) enable high-precision processing of test and simulation data (for example, constructing a three-dimensional mapping function of ideal electrothermal characteristics and fitting Foster thermal network model parameters), ensuring the accuracy and refinement of offline databases (such as the ideal electrothermal characteristics database and the dynamic temperature gradient compensation database). This integrated hardware and software configuration is specifically designed for IGBT module characteristic calibration, simulation, and database generation. It effectively integrates and calibrates test and simulation data, ensuring that the offline model parameters and database conform to both theoretical simulation principles and actual device characteristics, providing high-quality and reliable offline data support for subsequent online real-time estimation.

[0025] As a preferred technical means: the online sampling and preprocessing module includes: High-voltage isolated on-state voltage drop sampling unit, used to collect instantaneous collector-emitter on-state voltage drop; Current acquisition unit, used to measure the instantaneous collector current of IGBT in real time; Module shell temperature acquisition unit, used to measure module shell temperature in real time; A high-speed data acquisition and synchronization processing unit, configured to perform instantaneous synchronous sampling of the on-state voltage drop, collector current, and module shell temperature, and perform digital filtering and time stamp alignment; The electrothermal coupling estimation module includes: Kalman filter core algorithm implementation unit, used to implement state prediction and observation update of Kalman filter; A loss calculation unit, used for calculating the conduction loss and switching loss of the IGBT in real time; A database interface for interacting with the ideal electrothermal characteristics database and the dynamic temperature gradient compensation database generated by the offline modeling module; A Foster thermal network model solver, configured to implement the Foster thermal network model and perform state prediction of the junction temperature; The data output and application module includes: A data storage unit, used for storing the historical junction temperature and loss data; Communication interface unit, used for exchanging data with the converter valve control system, SCADA system or remote monitoring and diagnostic platform; A human-computer interaction interface for displaying the electrical and thermal status of the IGBT in real time; Control and protection interface, used to control the cooling system or protection device.

[0026] The high-voltage isolation on-state voltage drop sampling unit specifically captures instantaneous collector-emitter on-state voltage drop, accurately capturing weak millivolt-level signals in high-voltage environments and ensuring the measurement accuracy of key electrical parameters. The current acquisition unit and module case temperature acquisition unit focus on real-time acquisition of instantaneous current and case temperature, respectively, ensuring that core operating parameters are not missed. The high-speed data acquisition and synchronization processing unit uses instantaneous synchronous sampling, digital filtering, and timestamp alignment to strictly ensure the temporal consistency of on-state voltage drop, current, and case temperature, effectively eliminating noise interference and providing high-quality, time-synchronized input data for subsequent electrothermal coupling calculations. The Kalman filter core algorithm implementation unit ensures efficient execution of the dynamic iteration process for junction temperature estimation, supporting real-time junction temperature updates. The loss calculation unit focuses on real-time calculation of conduction and switching losses, dynamically coupled with junction temperature. A database interface ensures efficient access to the offline ideal electrothermal characteristics database and dynamic temperature gradient compensation database, providing a precise basis for calculations. The Foster thermal network model solver implements junction temperature prediction based on the physical laws of heat conduction, ensuring physical consistency of calculations and jointly improving the accuracy and dynamic response of junction temperature and loss estimation. The data storage unit effectively stores historical junction temperature and loss data, supporting trend analysis and backtracking. The communication interface unit is compatible with multiple platforms, including converter valve control systems and SCADA systems, expanding the application scope of the data. The human-machine interaction interface intuitively displays the IGBT electrical and thermal status, allowing operation and maintenance personnel to understand the equipment status in real time. The control and protection interface directly connects to the cooling system and protection device, allowing junction temperature and loss data to be directly converted into control instructions, supporting real-time thermal management and protection actions, and improving the engineering practicality and system reliability of the technical solution.

[0027] Beneficial effects: 1. Significantly improve junction temperature estimation accuracy: through innovative high-precision on-state voltage drop V CE(on) Instantaneous sampling technology overcomes the challenges of electrical parameter measurement in high-voltage environments. Combined with the offline dynamic temperature gradient compensation database and online adaptive Kalman filter, it can dynamically compensate for the hot spot effect inside the chip. V CE(on) The influence of the measured value is decoupled to obtain a more realistic and representative IGBT junction temperature, reducing the estimation error to an extremely low level.

[0028] 2. Realize electrothermal coupled collaborative estimation of junction temperature and losses: Establish a junction temperature-loss iterative closed loop with a Kalman filter as the core, tightly coupling the junction temperature estimation and loss calculation, so that the two can correct each other and converge collaboratively. The estimation results are more consistent with the actual electrothermal coupling characteristics of the IGBT, improving the physical consistency and accuracy of the model.

[0029] 3. Improving the robustness and adaptability of online monitoring: The adaptive Kalman filter can effectively suppress measurement noise and model uncertainty, and adjust its filtering strategy based on the dynamic temperature gradient changes caused by electrothermal coupling, so that this method can still maintain high accuracy and high robustness under complex and changing operating conditions (such as load fluctuations and ambient temperature changes).

[0030] 4. Provide core data for refined thermal management of flexible DC converter valves: The resulting high-precision real-time junction temperature and loss data can be directly used to implement precise overheating protection, optimize cooling strategies, evaluate device thermal stress accumulation, and provide early fault warnings. This can effectively support the development of flexible DC transmission systems towards intelligent and highly reliable operation and maintenance, extend equipment life, and reduce operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a general block diagram of the present invention; Figure 2 is a schematic diagram of an off-line test bench of the present invention; Figure 3 This is a flow chart of the construction of a dynamic temperature gradient compensation database within a chip according to the present invention; Figure 4 This is a block diagram of the online operating parameter sampling circuit module of the present invention; Figure 5 It is a Kalman filter flow chart of the present invention. DETAILED DESCRIPTION

[0032] The technical solution of the present invention is further described in detail below with reference to the accompanying drawings.

[0033] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] Example 1: like Figure 1 As shown, the present invention provides a high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation, comprising the following steps: S1: Build an offline electrothermal characteristics and compensation database for the IGBT module. This step is performed in a controlled laboratory environment and aims to provide an accurate electrothermal characteristics model and compensation database for online real-time estimation.

[0035] S1.1 Build an ideal electrothermal characteristics database for the IGBT body and anti-parallel diode.

[0036] Specifically, the present invention uses high-precision testing equipment to perform pulse testing on the IGBT module to obtain its electrical characteristics at a uniform junction temperature. Figure 2 ) mainly include: High-power pulse power supply: Provides controllable, short, high-current pulses (for example, peak current can reach twice the module's rated current, with an adjustable pulse width from 10µs to 100µs) to turn on the IGBT. The pulse width is short enough to ensure negligible device self-heating during measurement, keeping the junction temperature consistent with the case temperature.

[0037] Precision temperature control chamber (constant temperature oil bath or TEC temperature control system): Used to precisely control the temperature of the IGBT module substrate or heat sink under test. Temperature control accuracy must be within ±0.1°C to ensure uniformity and stability of the IGBT chip junction temperature during measurement. Test temperature points can be selected at multiple representative temperatures, such as 25°C, 75°C, 125°C, and 150°C.

[0038] High-speed oscilloscope: with high bandwidth (e.g. above 1 GHz), high sampling rate (e.g. above 5 GS / s) and high vertical resolution (e.g. 12 bits or higher), used to synchronously capture the instantaneous collector current (IGBT collector current I C or anti-parallel diode forward current I F ) waveform and collector-emitter on-state voltage drop (IGBT collector-emitter on-state voltage drop V CE(on) or anti-parallel diode forward voltage V F ) waveform. Current measurement is usually achieved through a high-bandwidth current probe or precision shunt, while voltage measurement uses a four-wire Kelvin connection method, connecting the voltage measurement point directly to the emitter and collector pins of the IGBT chip to eliminate the resistance of the leads and test fixture. V CE(on) The impact of measurement ensures measurement accuracy.

[0039] Data acquisition and processing system: responsible for synchronously triggering the pulse power supply and oscilloscope, and storing and analyzing the collected waveform data. Data processing includes: extracting the stable conduction platform from the pulse waveform I C and V CE(on) Instantaneous value, multiple measurements are averaged to further reduce random noise.

[0040] For multiple groups collected V CE(on) - I C and V F - I F The data is then mathematically fitted. The fitting method can be piecewise polynomial fitting or logarithmic fitting. For example, at each fixed junction temperature T j Next, right I C and V CE(on) The relationship is fitted to obtain the coefficient A ( T j ), B ( T j ), C ( T j );right I F and V F The relationship between the coefficients D ( T j ), E ( T j ), F ( T j ). These coefficients themselves are then used as the junction temperature T j Fitting a function, such as Finally, a three-dimensional mapping function or a multi-dimensional lookup table is formed that can accurately describe the ideal electrical characteristics of the IGBT and the anti-parallel diode FWD under uniform junction temperature conditions.

[0041] Subsequent applications: This "ideal electrothermal characteristics database" is the ideal V CE(on),ideal ( I C , T j ) is the basis of the dynamic temperature gradient compensation Δ V CE(on),comp A calculated benchmark used to compare actual measurements to an ideal situation, thereby quantifying the effects of temperature non-uniformity.

[0042] S1.2 Build a dynamic temperature gradient compensation database inside the chip.

[0043] Specifically, the present invention utilizes professional multi-physics finite element simulation software, such as COMSOL Multiphysics. The present invention selects COMSOL Multiphysics because of its powerful multi-physics bidirectional coupling simulation capability, which can simultaneously solve the electric field, current field and temperature field problems inside power electronic devices, and accurately simulate the generation, transfer and distribution of heat, rather than just simple heat conduction calculations. (For the specific process, refer to Figure 3 Flowchart shown) Model building method: 3D Geometric Model Reconstruction: Based on detailed IGBT module structural drawings or X-ray tomography data, accurately reconstruct the 3D geometry of the IGBT module in COMSOL. This model should include all key layers, such as the IGBT chip, anti-parallel diode chip, chip solder layer (e.g., silver sintering layer, soft solder layer), ceramic copper-clad laminate (DBC), copper substrate, bonding wires, package housing, and attached heat sink. The geometric dimensions of each layer must be accurately modeled according to the actual device.

[0044] Material property definition: define the temperature-dependent thermophysical parameters (such as thermal conductivity) for each material in the model (e.g. Si chip, Ag sintering, Cu, Al2O3 ceramic, Si3N4 ceramic, etc.) k ( T ), specific heat capacity C p ( T ),density ρ ( T )) and electrical parameters (such as resistivity ρ e ( T )). Temperature dependence is key to accurately capturing the electrothermal coupling effects.

[0045] Physics coupling settings: Establish the "Electric Current-Heat Transfer" multiphysics interface.

[0046] Set up a current field in the AC / DC module and calculate the Joule heat loss density in the IGBT chip and bonding wires based on the current density distribution (as a heat source term).

[0047] In the heat conduction module, the temperature field is set up and the losses calculated from the current field are input into the heat conduction equation as heat sources.

[0048] Set up bidirectional coupling: the losses calculated from the current field are used as a heat source. At the same time, the calculated results of the temperature field (junction temperature) are fed back into the update of electrical parameters (such as chip resistivity and bond wire resistance), affecting the loss calculation at the next moment.

[0049] Boundary Conditions and Meshing: Set realistic heat dissipation conditions, such as the convective heat transfer coefficient h between the heat sink base and the coolant (accounting for flow and temperature effects), or natural convection and radiation heat transfer with the ambient air. Perform local mesh refinement (e.g., using tetrahedral or hexahedral meshes with boundary layer meshing for critical heat flow paths) in critical areas such as the IGBT chip, solder layer, and bond wires to accurately capture subtle temperature gradients and hot spots in these areas.

[0050] Transient simulation settings: Simulate the transient electrothermal behavior of the IGBT under varying operating currents (e.g., from light load to heavy load), switching frequencies, duty cycles, and transient thermal shocks (e.g., short-term overload). The simulation time step should be small enough to capture the rapid dynamic changes in junction temperature.

[0051] Data extraction and compensation parameter construction: Post-process the simulation results. For each simulation condition and time step, extract the instantaneous three-dimensional temperature field distribution of the IGBT chip. Based on this, accurately calculate: the average junction temperature of the chip ( T j avg,sim ), the maximum junction temperature inside the chip ( T j max,sim ), and with S2 V CE(on) The local junction temperature corresponding to the measurement point position ( T j local,sim By setting virtual voltage measurement points in the simulation model, the simulated voltage is obtained, which takes into account the influence of the temperature non-uniformity inside the chip. V CE(on) value( V CE(on),simulated ).

[0052] Then, the present invention defines the temperature gradient compensation amount Δ V CE(on),comp , and its calculation formula is:

[0053] The “thermal gradient characteristic” is a parameter used to quantify the temperature non-uniformity inside the chip. For example, it can be the difference between the highest temperature and the average temperature of the chip (Δ T j grad = T j max,sim - T j avg,sim), or a more complex spatial temperature gradient feature vector (such as temperature differences at multiple key points). By performing nonlinear regression fitting or machine learning training (such as radial basis function neural network or support vector regression machine) on these simulation data, Δ V CE(on),comp and I C 、 T j avg,sim The complex mapping relationship between the temperature gradient and the thermal gradient characteristic quantity is stored in the form of a mathematical model or a multi-dimensional lookup table, forming a "dynamic temperature gradient compensation database".

[0054] Subsequent application: This database is implemented by the present invention V CE(on) FEM simulation provides internal details that are difficult to obtain through traditional testing, allowing us to accurately quantify the impact of temperature non-uniformity on the V CE(on) and introduce it into the online estimation model to improve the accuracy of online estimation.

[0055] S1.3 extracts the Foster thermal network model parameters of the IGBT module.

[0056] Specifically, the present invention obtains its transient thermal impedance characteristic curve by referring to the IGBT module data sheet provided by the manufacturer. Z thjc ( t Alternatively, the thermal transient test equipment (e.g., Mentor Graphics' T3Ster thermal tester) can be used for experimental measurement. This equipment can measure the IGBT's transient state after applying a power pulse. V CE(on) The transient change of the junction temperature can be inferred from the transient response of the junction temperature, and then the accurate Z thjc ( t )curve.

[0057] This measured or datasheet curve is then fitted to the Foster thermal network model using a curve fitting algorithm (e.g., a nonlinear optimization algorithm based on least squares or particle swarm optimization):

[0058] in, N is the order of the model (usually 3 to 5 based on fitting accuracy and computational efficiency), R i For the i Step thermal resistance, τ i For thei The thermal time constant of the order. These parameters are set { R i , τ i} will be used for dynamic estimation of online real-time junction temperature as the core component of the state transfer equation in the Kalman filter.

[0059] Subsequent applications: The Foster thermal network model provides an accurate physical model basis for online junction temperature estimation, ensuring the accuracy of the junction temperature dynamic response.

[0060] S2: Online, real-time, high-precision synchronous sampling and preprocessing of IGBT module operating parameters. This step aims to obtain high-quality electric and thermal parameter inputs at the actual operation site of the flexible DC converter valve.

[0061] S2.1 V CE(on) High-precision isolated sampling.

[0062] Specifically, the present invention deploys a specially designed high-voltage isolation and high common-mode rejection ratio differential amplifier circuit at the collector and emitter pins of each IGBT module. This circuit uses a precision differential operational amplifier (such as ADI's AD8479) combined with a high-speed analog isolator (such as Analog Devices' ADuM7420 or Texas Instruments' ISO224) to withstand transient common-mode voltages of up to several thousand volts when the converter valve is operating and ensure electrical isolation between the measurement circuit and the control circuit. The differential amplifier adopts a design with high bandwidth, low input bias current, and high common-mode rejection ratio (CMRR is usually required to be greater than 80dB@100kHz) to ensure that the measurement circuit is electrically isolated at the millivolt level. V CE(on) During the signal acquisition process, high-frequency noise and common-mode interference are effectively suppressed. The bandwidth of the sampling circuit needs to match the switching speed of the IGBT, usually requiring a range of several MHz to tens of MHz. (For the specific process, refer to Figure 4 Circuit module block diagram shown) S2.2 Multi-channel high-speed synchronous sampling and signal preprocessing.

[0063] Specifically, the present invention uses a high-performance, multi-channel, high-speed analog-to-digital converter (ADC) system. The system can be composed of a high-speed ADC chip integrated with an FPGA (Field-Programmable Gate Array) or a high-performance digital signal processor (DSP) (for example, TI's ADS8688 series, which provides 16-bit resolution and a single-channel sampling rate of 500 kSPS to 1 MSPS, or higher). The ADC system receives the data from the above-mentionedV CE(on) Isolation sampling circuit, high-precision Hall effect current sensor (such as LEM's LA series, used to measure I C ) and thermistor (NTC) or fiber optic temperature sensor mounted on the IGBT module housing or heat sink substrate (used to measure module housing temperature) T C ) of the analog signal.

[0064] The sampling trigger uses the rising or falling edge of the IGBT gate drive signal as the precise synchronization trigger point, and is controlled by the high-precision timer inside the FPGA to ensure V CE(on) 、 I C and T C Strictly synchronized acquisition in the time dimension. During the extremely short plateau period (usually 1-2 microseconds) when the IGBT is turned on, the ADC system performs high-density instantaneous sampling to obtain the data that best represents the device's on-state. V CE(on) Instantaneous value.

[0065] The acquired raw digital data stream is digitally filtered in real time (for example, using a median filter to remove spike noise or a low-pass filter to smooth high-frequency noise). A precise timestamp alignment algorithm is then implemented, using interpolation or resampling techniques to ensure strict temporal consistency of all acquired parameters, providing a clean, synchronized, and highly accurate data stream for subsequent electrothermal coupling estimation.

[0066] Subsequent applications: High-quality online measurement data is the basis for high-precision junction temperature estimation. It overcomes the difficulties of traditional measurement in high-voltage environments and provides reliable observation values ​​for subsequent Kalman filtering.

[0067] S3: Based on the online real-time operating parameters and the offline database, iterative calculation and dynamic compensation of junction temperature and loss electrothermal coupling are performed. This step is the core intelligent processing part of the present invention and is typically executed in real time on a high-performance embedded controller (such as a DSP or FPGA) or industrial PC.

[0068] S3.1 Construct the state space model of the Kalman filter.

[0069] System state vector x k : Defined as the instantaneous junction temperature of the IGBT, that is, x k =[ T j ] k T .

[0070] State transition equation: Based on the Foster thermal network model and IGBT loss calculation model extracted from S1.3. t s The dynamic change of junction temperature can be approximately expressed as:

[0071] in, T j (k) is the junction temperature at the current moment, T C (k) It is the module shell temperature measured by S2 in real time. is the total IGBT loss estimated at the last moment (including conduction loss and switching loss), Z thjc (Δ t s ) is the transient thermal impedance calculated based on the Foster model parameters (S1.3) in Δ t s The response within a time step, w k is process noise representing model uncertainty or unmodeled dynamics, typically modeled as zero-mean Gaussian white noise.

[0072] The calculation depends on the junction temperature at the previous moment and the current current, which is obtained through the "ideal electrothermal characteristics database" of S1.1 and the loss calculation formula:

[0073] in, P cond is the conduction loss, P sw is the switching loss, V bus is the DC bus voltage, f sw is the switching frequency. Conduction losses are usually calculated based on I C and V CE(on) Calculate the switching loss based on the switching energy E on, E off (These also depend on I C , V bus, T j ) and switching frequency calculation.

[0074] Observation equation: The high-precision sampling in S2 is obtained As an observed value, it is related to the junction temperature:

[0075] in, V CE(on),ideal Provided by the "Ideal Electrothermal Characteristics Database" of S1.1, I C (k) is the current collected by S2 in real time, v k is the measurement noise (residual white noise). Δ V CE(on),comp It is the key dynamic temperature gradient compensation item, which is determined by the "dynamic temperature gradient compensation database" of S1.2 according to the current I C and the junction temperature predicted by the Kalman filter (or a priori estimate ) query or calculation.

[0076] S3.2 Iterative calculation process based on adaptive Kalman filter.

[0077] This method performs the following Kalman filter iteration steps in each real-time sampling period (see Figure 5 Kalman filter flowchart shown in the figure) to achieve coordinated estimation and dynamic compensation of junction temperature and loss: Initialization: At the beginning of each sampling period, the current real-time I C (k) 、 T C (k) and . Set the initial state estimate of the Kalman filter and the error covariance matrix P k-1 If it is the first time running, Can be set to T C (0) .

[0078] State prediction step: Predicted junction temperature: estimated based on the junction temperature at the previous moment and current I C (k) 、 T C (k) First, use the "ideal electrothermal characteristics database" of S1.1 and the loss calculation formula to calculate the loss at the previous moment. Then, the Foster thermal network model of S1.3 is used to predict the junction temperature prior estimate at the current moment. .

[0079] Forecast error covariance: Update the prior error covariance matrix:

[0080] in, A k is the state transfer matrix derived based on the state transfer equation, Q k is the process noise covariance matrix, which reflects the uncertainty of the model itself and unmodeled dynamics.

[0081] Observation update and dynamic compensation: Calculate the Kalman gain K k :

[0082] in, H k is the observation matrix derived from the observation equation (i.e. V CE(on) right T j partial derivatives of ). R k is the dynamically adjusted measurement noise covariance matrix. R k The value of is adjusted in real time based on the information provided by the dynamic temperature gradient compensation database constructed in S1.2. Specifically, according to the current I C (k) and predicted junction temperature , query or calculate the corresponding Δ from the database V CE(on),comp When the temperature gradient inside the chip is larger (Δ V CE(on),comp The larger the absolute value of R k The larger the value of , the greater the V CE(on) The "reliability" or "representativeness" of the measurement decreases, and the Kalman filter will rely more on the model prediction value. Conversely, when the temperature gradient is small, R k The smaller the value, the more the filter depends on the measured value. This dynamic adjustment enables the filter to adaptively deal with the measurement error and non-uniformity caused by electrothermal coupling, thus enhancing the robustness of the filter. State update (junction temperature decoupling): Using the calculated Kalman gain, combined with the high-precision real-time acquisition in S2 and the model predicted VCE(on) , update the junction temperature estimate:

[0083] right Perform "calibration" or "compensation" to "decouple" V CE(on) The influence of the temperature non-uniformity inside the chip during measurement makes It can more accurately reflect the overall or critical area junction temperature of the IGBT, rather than just the temperature of the local measurement point affected by hot spots.

[0084] Update the posterior error covariance matrix:

[0085] Loss recalculation and output: using the latest and most accurate junction temperature estimate , recalculate the exact loss at the current moment Outputs the current high-precision, dynamically compensated IGBT junction temperature and accurate loss This process is carried out in real time during each sampling period, forming a closed loop and continuously providing updated electric and heating status information.

[0086] S4: Output and apply the precise junction temperature and loss data obtained through iterative calculation. This step aims to use the real-time estimated precise IGBT junction temperature and loss data to guide the refined operation management and reliability improvement of the flexible DC converter valve.

[0087] Specifically, this precise junction temperature and loss data can be directly applied to the following aspects: Online thermal management and heat dissipation optimization: Based on real-time feedback of the precise IGBT junction temperature, the converter valve control system can intelligently adjust the cooling system (for example, regulating coolant flow, controlling fan speed, and switching cooling modes) to ensure that the IGBT always operates within the safe operating area, avoiding performance degradation or damage due to overheating.

[0088] Detailed device condition monitoring: The system monitors the actual IGBT loss and junction temperature data in real time and compares them with design values, historical health data, or preset thresholds. If any abnormal fluctuations or deviations occur, the system will promptly issue an alarm, helping maintenance personnel identify potential operational anomalies or early signs of failure.

[0089] Valve Operation Efficiency Assessment: Accurately calculating the power loss of each IGBT module is crucial for evaluating the efficiency of the entire valve system and, ultimately, the HVDC Flexible system. Accurate loss data can be used to optimize system scheduling strategies and improve overall energy conversion efficiency.

[0090] Early failure warning and predictive maintenance: Monitoring long-term trends in junction temperature and losses, such as unusually large junction temperature fluctuations or consistently increasing losses under the same operating conditions, can serve as early warning signals of IGBT internal aging (e.g., bond wire fatigue or solder void expansion leading to increased thermal resistance). This enables predictive maintenance of the device, avoiding unexpected failures, reducing operating and maintenance costs, and improving system availability.

[0091] Example 2: The embodiment of the present invention also provides a high-precision real-time decoupling modeling system for IGBT junction temperature based on dynamic electrothermal compensation. The system can implement all the processes of the above method.

[0092] Please refer to the attached figure, which shows a high-precision real-time decoupling modeling system for IGBT junction temperature based on dynamic electrothermal compensation, including: Offline modeling module: used to execute the above-mentioned S1 offline electrothermal characteristics and compensation database construction. This module is a high-precision test and simulation integrated platform, usually composed of high-performance industrial computers or dedicated workstations, equipped with: High-precision pulse test unit: includes a high-power pulse power supply, precision temperature control box, high-speed oscilloscope and data acquisition card, used to perform S1.1, ideal electrothermal characteristic calibration tests of IGBT body and anti-parallel diode.

[0093] Multi-physics simulation workstation: Equipped with professional multi-physics finite element simulation software such as COMSOL Multiphysics and its related modules (such as AC / DC module and heat conduction module), it is used to perform the simulation calculations required to build the dynamic temperature gradient compensation database inside the S1.2 chip.

[0094] Data analysis and database management unit: Integrates data analysis tools and database management systems such as MATLAB and Python to perform mathematical fitting, regression analysis, or machine learning training on test and simulation data, ultimately generating and managing the "ideal electrothermal characteristics database" and "dynamic temperature gradient compensation database," as well as storing Foster thermal network model parameters.

[0095] Online sampling and preprocessing module: It is used to perform online real-time high-precision synchronous sampling and preprocessing of IGBT module operating parameters in S2. This module is physically deployed near the IGBT module of the flexible direct current converter valve and mainly includes: High-voltage isolated on-state voltage drop sampling unit: Contains high-voltage isolation, high common-mode rejection ratio differential amplifier circuit, directly connected to the collector and emitter of the IGBT, and transmits signals to the main processing unit through optical fiber or high-speed digital isolator to achieve millivolt level V CE(on) Safe and accurate signal acquisition in high voltage environments.

[0096] Current acquisition unit: uses high-bandwidth, high-precision Hall effect current sensor or Rogowski coil to measure the instantaneous current of IGBT in real time I C .

[0097] Module shell temperature acquisition unit: uses a platinum resistance temperature detector (RTD) sensor or a fiber optic temperature sensor, installed on the IGBT module housing or the surface of the radiator substrate, to measure the module shell temperature in real time and stably. T C .

[0098] High-speed data acquisition and synchronous processing unit: usually implemented by FPGA or DSP, integrating multi-channel high-speed ADC chip, which can be triggered synchronously with the IGBT gate drive signal, with high precision of microsecond or even nanosecond level. V CE(on) 、 I C and T C The unit also has built-in digital filtering algorithms and time stamp alignment algorithms to pre-process the collected raw data and output a clean, synchronized digital signal stream.

[0099] The electrothermal coupling estimation module is used to perform the aforementioned S3 iterative calculation and dynamic compensation of junction temperature-loss electrothermal coupling based on online real-time operating parameters and an offline database. This module is the core intelligent computing unit of the entire system and is typically implemented by a high-performance digital signal processor (DSP, such as the TIC2000 series) or an industrial-grade embedded computer (equipped with a multi-core processor). It has powerful floating-point computing capabilities and real-time operating system (RTOS) support. The functional logic within this module includes: Kalman filter core algorithm implementation unit: Programming implements the state prediction and observation update of the Kalman filter to perform real-time and dynamic estimation of the IGBT junction temperature.

[0100] Loss calculation unit: Receives the junction temperature estimated by this module and the real-time current provided by the online sampling and pre-processing module, and calculates the conduction loss and switching loss of the IGBT in real time based on the "ideal electrothermal characteristics database" constructed by S1.1.

[0101] Database interface and access logic: Provides an efficient access mechanism to the "ideal electrothermal characteristics database" and "dynamic temperature gradient compensation database" built by S1 to obtain the required characteristic parameters and compensation amounts.

[0102] Foster thermal network model solver: implements the Foster thermal network model extracted by S1.3 for state prediction of junction temperature in the Kalman filter.

[0103] Dynamic compensation logic: As the core component of the Kalman filter observation update, this logic obtains the corresponding Δ from the "dynamic temperature gradient compensation database" based on the current current and predicted junction temperature V CE(on),comp . And dynamically adjust the measurement noise covariance matrix in the Kalman filter accordingly R k The value of , while executing in the observation update formula V CE(on) The measured value is compensated and corrected, thus achieving a "decoupled" and accurate estimation of the junction temperature.

[0104] Data output and application module: It is used to execute the above S4 to output and apply the accurate junction temperature and loss data obtained by iterative calculation. This module is responsible for managing the output data (accurate junction temperature, accurate loss) of the electrothermal coupling estimation module and providing external interfaces, usually including: Data storage unit: Uses high-speed, large-capacity non-volatile memory (such as SSD or NAND Flash) for long-term storage of real-time calculated junction temperature and loss data, supporting historical trend analysis, data backtracking, and fault diagnosis.

[0105] Communication interface unit: supports multiple standard industrial communication protocols (such as Modbus TCP / RTU, IEC 61850, CANbus, and Ethernet / IP), enabling real-time data exchange with converter valve station-level control systems, SCADA systems, and remote monitoring and diagnostic platforms.

[0106] Human-Machine Interface (HMI): Provides a visual interface that displays the IGBT's current junction temperature, loss, operating status, alarm information, and historical trend charts in real time, allowing operation and maintenance personnel to intuitively understand the equipment's health status.

[0107] Control and protection interface (optional): Provides a direct control interface with the cooling system or protection system, such as through digital output or analog output, to achieve accurate overtemperature protection action or active thermal management control based on real-time junction temperature.

[0108] In specific implementation, the working principle, control process and technical effects of the high-precision real-time decoupling modeling system for IGBT junction temperature based on dynamic electrothermal compensation provided by the embodiment of the present invention are the same as those of the high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation in the above embodiment, and will not be repeated here.

[0109] The above-mentioned method and system for real-time decoupling modeling of IGBT junction temperature based on dynamic electrothermal compensation are specific embodiments of the present invention, which have reflected the substantial characteristics and progress of the present invention. They can be modified equivalently according to actual usage needs under the guidance of the present invention, and all of them are within the scope of protection of this solution.

Claims

1. A high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation, characterized by: Including steps: S1: Offline construction of the electrothermal characteristics and compensation database of the IGBT module, including: Construct a database of ideal electrothermal characteristics of IGBT body and anti-parallel diode; Build a dynamic temperature gradient compensation database inside the IGBT chip; Extracting Foster thermal network model parameters of the IGBT module; S2: Online real-time high-precision synchronous sampling and pre-processing of IGBT module operating parameters, including high-precision isolated sampling of the IGBT collector-emitter on-state voltage drop, and multi-channel high-speed synchronous sampling and signal pre-processing of the on-state voltage drop, collector current, and module case temperature; S3: Based on online real-time operating parameters and an offline database, it performs junction temperature-loss electrothermal coupling iterative calculation and dynamic compensation; S4: Output and apply the junction temperature and loss data obtained through iterative calculation; applications include online thermal management, operating condition monitoring, or fault warning.

2. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 1 is characterized in that: Constructing the ideal electrothermal characteristics database in step S1 specifically includes: obtaining the relationship between the on-state voltage drop and current of the IGBT and the anti-parallel diode at multiple preset uniform junction temperature points through pulse testing, and using piecewise polynomial or logarithmic fitting to establish a three-dimensional mapping function of the on-state voltage drop as the current and junction temperature change.

3. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 1, characterized in that: Constructing a dynamic temperature gradient compensation database in step S1 specifically includes: using multi-physics field finite element simulation software to establish a three-dimensional electrothermal coupling model of the IGBT module to simulate the transient temperature field distribution of the chip under different working conditions; extracting the average junction temperature, maximum junction temperature and local junction temperature of the chip at the measurement point; defining the temperature gradient compensation amount as the difference between the simulated on-state voltage drop considering temperature non-uniformity and the ideal on-state voltage drop in the ideal electrothermal characteristic database, and establishing a mapping relationship between the compensation amount and the current, average junction temperature and thermal gradient characteristic amount.

4. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 1, characterized in that: In the Foster thermal network model parameters extracted in step S1, the transient thermal impedance is characterized by a combination of an N-order thermal resistance and a corresponding thermal time constant, where N is the model order.

5. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 1 is characterized in that: In step S2, a high-voltage isolated differential amplifier circuit is used to instantaneously acquire the collector-emitter on-state voltage drop. The rising or falling edge of the IGBT gate drive signal is used as the sampling trigger point, and the on-state voltage drop, collector current, and module case temperature are synchronously acquired through a multi-channel high-speed analog-to-digital converter. The acquired data is digitally filtered and timestamp aligned to ensure time synchronization.

6. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 1, characterized in that: Step S3 includes: Construct a Kalman filter state space model, with junction temperature as the state vector, Foster thermal network model as the state transition equation, and on-state voltage drop as the observation value; Iterate in each sampling period: a) State prediction: Based on the junction temperature, current, and case temperature at the previous moment, the current junction temperature is predicted using the Foster model. b) Observation update: i. Obtaining the temperature gradient compensation amount from the dynamic temperature gradient compensation database based on the current and predicted junction temperature; ii. Dynamically adjust the measurement noise covariance matrix so that it increases as the temperature gradient compensation amount increases; iii. Subtract the sum of the ideal on-state voltage drop and the temperature gradient compensation amount from the real-time on-state voltage drop measurement value to obtain the compensated observation value; iv. Update the junction temperature estimate in combination with the Kalman gain; c) Output the real-time junction temperature after dynamic compensation and recalculated loss.

7. The high-precision real-time decoupling modeling method for IGBT junction temperature based on dynamic electrothermal compensation according to claim 6, characterized in that: In step S3, Dynamically adjusting the measurement noise covariance matrix includes: querying a dynamic temperature gradient compensation database based on the current current and the predicted junction temperature, and linearly or nonlinearly increasing the value of the measurement noise covariance matrix when the absolute value of the temperature gradient compensation amount increases; Loss recalculation includes: based on the updated junction temperature estimate, combined with the real-time current and switching frequency, the loss is calculated using the conduction loss formula and the switching energy formula.

8. An IGBT junction temperature high-precision real-time decoupling modeling system using the IGBT junction temperature high-precision real-time decoupling modeling method based on dynamic electrothermal compensation according to claim 1, characterized in that: include: An offline modeling module is used to build an electrothermal characteristics and compensation database for the IGBT module offline, including: building an ideal electrothermal characteristics database for the IGBT body and anti-parallel diode, building an internal dynamic temperature gradient compensation database for the IGBT chip, and extracting the Foster thermal network model parameters of the IGBT module; Online sampling and preprocessing module, used for online real-time high-precision synchronous sampling and preprocessing of IGBT module operating parameters, including: high-precision isolated sampling of IGBT collector-emitter on-state voltage drop; multi-channel high-speed synchronous sampling and signal preprocessing of on-state voltage drop, collector current and module case temperature; an electrothermal coupling estimation module, connected to the offline modeling module and the online sampling and preprocessing module, for performing junction temperature-loss electrothermal coupling iterative calculation and dynamic compensation based on online real-time operating parameters and an offline constructed database; The data output and application module is used to output and apply the junction temperature and loss data obtained by iterative calculation, and the applications include online thermal management, working condition monitoring or fault warning.

9. The IGBT junction temperature high-precision real-time decoupling modeling system according to claim 8, characterized in that: The offline modeling module includes: an industrial computer or workstation equipped with simulation software and data analysis tools, which is used for characteristic calibration, simulation and database generation of IGBT modules.

10. The IGBT junction temperature high-precision real-time decoupling modeling system according to claim 8, characterized in that: The online sampling and preprocessing module includes: High-voltage isolated on-state voltage drop sampling unit, used to collect instantaneous collector-emitter on-state voltage drop; Current acquisition unit, used to measure the instantaneous collector current of IGBT in real time; Module shell temperature acquisition unit, used to measure module shell temperature in real time; A high-speed data acquisition and synchronization processing unit, configured to perform instantaneous synchronous sampling of the on-state voltage drop, collector current, and module shell temperature, and perform digital filtering and time stamp alignment; The electrothermal coupling estimation module includes: Kalman filter core algorithm implementation unit, used to implement state prediction and observation update of Kalman filter; A loss calculation unit, used for calculating the conduction loss and switching loss of the IGBT in real time; A database interface for interacting with the ideal electrothermal characteristics database and the dynamic temperature gradient compensation database generated by the offline modeling module; A Foster thermal network model solver, configured to implement the Foster thermal network model and perform state prediction of the junction temperature; The data output and application module includes: A data storage unit, used for storing the historical junction temperature and loss data; Communication interface unit, used for exchanging data with the converter valve control system, SCADA system or remote monitoring and diagnostic platform; A human-computer interaction interface for displaying the electrical and thermal status of the IGBT in real time; Control and protection interface, used to control the cooling system or protection device.

Images