With the global transformation and upgrading of the energy structure, utility interactive inverters, as critical interfaces between renewable energy sources and the grid, have become increasingly important. The large-scale integration of renewable energy sources, such as wind and solar power, poses new challenges to grid stability and reliability. Traditional utility interactive inverters often exhibit poor operational performance due to insufficient control strategies, lacking the inherent inertia and damping capabilities of synchronous generators, which hinders effective synchronization with the grid. Virtual synchronous generator (VSG) control has garnered widespread attention because it can emulate the external characteristics of traditional synchronous generators and provide essential frequency and voltage support to the grid.
As a core component within utility interactive inverters, the reliability of the insulated gate bipolar transistor (IGBT) is crucial. Active thermal control methods for IGBTs have emerged to mitigate internal electrothermal stresses. Generally, electrothermal stresses in IGBTs are categorized into two types: steady-state thermal stress caused by long-term thermal cycling and transient overstress caused by short-term thermal cycling. Some studies focus on active thermal control strategies targeting steady-state thermal stress under long-term operation, such as by adding gate drivers, modifying modulation strategies, or adjusting switching frequencies. However, transient overstress due to events like grid faults or shutdowns is equally critical over short timescales and should not be overlooked.
In motor drive applications, research has addressed short-term overstress during motor stall conditions. However, due to the complexity and brief duration of transient faults in grid-connected converters, few studies have focused on thermal control of IGBTs during these transient periods. This paper proposes an active thermal control method for IGBTs in utility interactive inverters during transient faults. We develop a utility interactive inverter model based on VSG control and a corresponding electrothermal model for the internal IGBT. By analyzing the relationship between the IGBT junction temperature and electrical parameters, we propose an active thermal control method that integrates virtual impedance, adaptive power adjustment, and switching frequency variation. This approach aims to alleviate internal overstress during transient faults while enhancing inverter stability. Finally, we validate the feasibility and accuracy of the proposed method through simulations and experiments.
The basic structure of a utility interactive inverter is shown in the figure above. Here, \(U_{dc}\) represents the DC bus voltage, \(R_f\) is the equivalent filter resistance of the inverter, \(L_f\) and \(C_f\) are the output filter inductance and capacitance, \(R_g\) and \(L_g\) are the grid-side equivalent resistance and inductance, \(i_{abc}\) and \(u_{abc}\) are the inverter output filter inductor current and capacitor voltage, \(P_{cal}\) and \(Q_{cal}\) are the real-time calculated output power, \(P_{ref}\) and \(Q_{ref}\) are the input power commands, \(\omega_{ref}\) and \(V_{ref}\) are the reference angular frequency and internal electromotive force in VSG control, \(\theta_{inv}\) and \(V_{inv}\) are the equivalent output power angle and internal electromotive force of the inverter, \(I_{dqref}\) and \(E_{dq}\) are the dq-axis current and voltage reference values output by the voltage and current control loops, and \(f_{sw}\) is the switching frequency.
For utility interactive inverters, VSG control is typically designed to enable self-synchronization and grid connection, ensuring more stable and continuous operation. VSG control simulates the rotor equations of traditional synchronous generators, allowing the inverter to exhibit inertia and damping characteristics, thereby significantly improving the inverter’s disturbance resistance and overall grid system stability. VSG control primarily consists of two parts: the P-f loop and the Q-V loop. The P-f loop stabilizes the inverter’s output active power and angular frequency, while the Q-V loop stabilizes the output reactive power and internal electromotive force. The expressions for VSG control are as follows:
$$ \begin{cases} P_{ref} – P_{cal} – D_p(\omega_{ref} – \omega) = J_p \frac{d\omega}{dt} \\ Q_{ref} – Q_{cal} = D_q(V_{inv} – V_{ref}) \end{cases} $$
where \(D_p\) and \(D_q\) are the damping coefficients for the P-f and Q-V loops, respectively, and \(J_p\) is the introduced virtual inertia. The real-time output power of the inverter is calculated as:
$$ \begin{cases} P_{cal} = i_a u_a + i_b u_b + i_c u_c \\ Q_{cal} = \frac{i_a(u_b – u_c) + i_b(u_c – u_a) + i_c(u_a – u_b)}{\sqrt{3}} \end{cases} $$
To perform thermal analysis on the IGBT within the utility interactive inverter and implement appropriate active thermal control, it is essential to first establish a junction temperature monitoring model. We employ an electrothermal model to estimate the junction temperature without increasing measurement complexity. This electrothermal model comprises three parts: a device model characterizing the IGBT on-state voltage drop and switching losses, a loss model calculating IGBT power losses, and a thermal model estimating the junction temperature.
The IGBT junction temperature primarily results from the accumulation of internal power losses. Generally, power losses include conduction losses and switching losses. Conduction losses occur when current flows through the device and are related to the transient current through the chip and the conduction voltage drop. Using conventional PWM modulation, the conduction losses for the IGBT and its antiparallel diode can be expressed as:
$$ \begin{cases} P_{IC} = i_{ce} \cdot v_{ce} \cdot \left( \frac{1}{2} + \frac{M \sin(\omega t + \phi)}{2} \right) \\ P_{DC} = i_{ce2} \cdot v_{ce2} \cdot \left( \frac{1}{2} – \frac{M \sin(\omega t + \phi)}{2} \right) \end{cases} $$
where \(P_{IC}\) and \(P_{DC}\) are the conduction losses of the IGBT and diode, respectively, \(i_{ce}\), \(v_{ce}\) and \(i_{ce2}\), \(v_{ce2}\) are the transient currents and conduction voltage drops of the IGBT and diode, \(M\) is the voltage modulation ratio, and \(\phi\) is the phase difference between the inverter output voltage and current.
For utility interactive inverters based on VSG control, the switching frequency is typically high to ensure output power quality. During IGBT turn-on and turn-off, overlapping voltage and current waveforms result in significant switching losses due to frequent high-frequency switching. The switching energy during these transitions is obtained by integrating the product of transient current and voltage. The switching losses for the IGBT and diode are given by:
$$ \begin{cases} P_{IS} = f_{sw} \cdot (E_{ons} + E_{offs}) \\ P_{DS} = f_{sw} \cdot E_{rr} \end{cases} $$
where \(P_{IS}\) and \(P_{DS}\) are the switching losses of the IGBT and diode, respectively, \(E_{ons}\) and \(E_{offs}\) are the turn-on and turn-off energies of the IGBT, and \(E_{rr}\) is the reverse recovery energy of the diode.
Heat flow, like electrical energy flow, is a form of energy transfer. According to thermal analogy theory, thermal circuits can be equivalently compared to electrical circuits: power loss corresponds to current, temperature corresponds to voltage, and thermal resistance corresponds to electrical resistance. Two primary thermal models describe power device thermal characteristics: the Cauer model and the Foster model. The Cauer model accurately reflects the heat conduction process through the material layers but is complex. The Foster model, while not corresponding to specific material layers and having nodes without physical meaning, is suitable for analytical calculation of temperature distribution within power modules. Therefore, we establish a Foster thermal network model for the power module to estimate the IGBT junction temperature. The thermal network is shown below, where \(T_{jI}\) and \(T_{jD}\) are the junction temperatures of the IGBT and diode, \(T_c\) and \(T_a\) are the case and ambient temperatures, \(R_{jc}\), \(C_{jc}\) and \(R_{jc}’\), \(C_{jc}’\) are the thermal resistances and capacitances of the IGBT and diode chips, \(R_{ch}\) and \(C_{ch}\) are the thermal resistance and capacitance from the module case to the heat sink, and \(R_{ha}\) and \(C_{ha}\) are the thermal resistance and capacitance between the heat sink and ambient environment.
The case temperature and junction temperatures of the IGBT and diode can be expressed as:
$$ T_c = T_a + P_{tol} \left( \frac{R_{ch}}{1 + s\tau_{ch}} + \frac{R_{ha}}{1 + s\tau_{ha}} \right) $$
$$ \begin{cases} T_{jI} = T_c + (P_{IC} + P_{IS}) \sum_{i=1}^{4} \frac{R_{jci}}{1 + s\tau_i} \\ T_{jD} = T_c + (P_{DC} + P_{DS}) \sum_{i=1}^{4} \frac{R_{jci}’}{1 + s\tau_i’} \end{cases} $$
where \(P_{tol}\) is the total loss generated by the chips, \(\tau_{ch}\) and \(\tau_{ha}\) are the time constants for \(R_{ch}\) and \(R_{ha}\), and \(\tau_i\) and \(\tau_i’\) are the time constants for \(R_{jci}\) and \(R_{jci}’\). The relationship between thermal resistance and time constant is \(\tau = R \cdot C\).
The thermal impedance parameters for the power module are summarized in the table below.
| Parameter | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| \(R_{jci}\) (K/W) | 0.0174 | 0.0957 | 0.0928 | 0.0841 |
| \(\tau_i\) (s) | 0.01 | 0.02 | 0.05 | 0.1 |
| \(R_{jci}’\) (K/W) | 0.03 | 0.165 | 0.16 | 19.5214 |
| \(\tau_i’\) (s) | 0.01 | 0.02 | 0.05 | 0.1 |
During transient faults in utility interactive inverters, the output characteristics change significantly. Taking the grid voltage amplitude and phase as reference, the inverter output current can be expressed as:
$$ L_{eq} \frac{di_{abc}}{dt} + R_{eq} i_{abc} = V_{inv} – V_g $$
where \(R_{eq}\) and \(L_{eq}\) are the equivalent resistance and inductance of the line between the inverter and grid, and \(V_g\) is the grid voltage amplitude. When a transient fault causes grid voltage sag, the inverter’s equivalent internal electromotive force amplitude and phase do not change abruptly, creating a potential difference between the inverter and grid. Due to the small equivalent line impedance, even a small potential difference leads to rapid growth in inverter output current. The phasor diagram before and after a transient voltage sag fault shows that when the grid voltage sag depth is large, the current surge from the inverter is also larger. Additionally, the output current increases further with the power angle during grid faults.
The rapid increase in output fault current during transient faults is a key factor causing the rise in IGBT junction temperature. Simultaneously, transient fault current peaks can easily destroy IGBTs. Therefore, controlling the peak output current during transient faults is crucial for improving IGBT reliability and lifespan. Introducing line impedance is one solution to suppress overcurrent. However, conventional current-limiting impedance requires additional voltage compensation during steady-state operation. To avoid real-time compensation, virtual impedance can be utilized to activate current-limiting impedance only during transient faults, generating a voltage drop to limit current. The virtual impedance control strategy is expressed as:
$$ \begin{cases} R_v = k_R \cdot \text{Max}(I_{amp} – I_{lim}, 0) \\ X_v = k_X \cdot R_v \end{cases} $$
where \(I_{amp}\) is the amplitude of current \(i_{abc}\), \(I_{lim}\) is the virtual impedance current-limiting threshold, \(k_R\) and \(k_X\) are the proportional gains for virtual resistance and reactance, \(R_v\) and \(X_v\) are the virtual resistance and reactance. The dq-axis voltage drops due to virtual impedance are:
$$ \begin{cases} E_{dv} = -i_d R_v – i_q X_v \\ E_{qv} = -i_q R_v + i_d X_v \end{cases} $$
and the current amplitude is \(I_{amp} = \sqrt{i_d^2 + i_q^2}\). Compared to other methods, this virtual impedance control is relatively simple and ensures it is not activated during stable inverter operation, avoiding additional compensation.
For transient faults, instantaneous current spikes occur at the moment of grid voltage sag, and sustained overcurrent persists during the fault. Virtual impedance, with its dynamic adjustment capability, can quickly respond to current surges and suppress transient spikes, but it has limitations in suppressing sustained overcurrent during faults. Therefore, we introduce an adaptive power adjustment strategy that flexibly reduces the input active power command during transient faults to lower the output current level, thereby reducing the peak junction temperature. Reducing the active power command during faults can also improve the transient synchronization stability of the utility interactive inverter. The adaptive power adjustment strategy is expressed as:
$$ P_{ref} = \begin{cases} P_n, & V_{g0} < V_g < V_{gn} \\ \frac{K_f V_g P_n}{V_{gn}}, & V_g < V_{g0} \end{cases} $$
where \(P_n\) is the initial input active power command, \(V_{g0}\) and \(V_{gn}\) are the lower voltage limit for activating the control strategy and the rated grid voltage amplitude, and \(K_f\) is the power adjustment coefficient. To minimize overcurrent impact and power angle instability, the unbalanced active power difference before and after the transient fault should be zero, leading to:
$$ \Delta P_{out} = \frac{V_{inv0} V_{gn} \sin \delta_{out0} – V_{inv1} V_g \sin \delta_{out1}}{\omega (L_{eq})} = 0 $$
Thus, the power adjustment coefficient \(K_f\) can be derived as \(K_f = V_{inv1} / V_{inv0}\). For practical implementation, \(K_f\) can be set as a constant. By setting an appropriate power adjustment coefficient, the utility interactive inverter can achieve fault ride-through and limit current through the IGBT during transient faults, thereby suppressing the peak junction temperature and improving reliability.
The inherent reliability and lifespan of IGBTs primarily depend on the maximum thermal cycling and peak junction temperature within the device. As analyzed, the large overcurrent during transient faults induces higher junction temperatures than steady-state operation. The occurrence of transient faults further exacerbates the maximum thermal cycling inside the IGBT. Therefore, limiting the maximum thermal cycling is also crucial for extending lifespan. According to the switching loss equation, the PWM switching frequency is proportional to IGBT chip losses, and the transient current through the IGBT is also positively correlated. Thus, the switching frequency can be reduced during transient overcurrent to decrease switching losses and increased when the fault recovery current has defects to increase losses, thereby suppressing further growth of the maximum thermal cycling while the overall junction temperature changes rapidly with transient current. The switching frequency variation is expressed as:
$$ f_{sw} = \begin{cases} f_{sw0}, & V_{g0} < V_g < V_{gn} \\ k_s I_{amp} + b_s, & V_g < V_{g0} \end{cases} $$
where \(f_{sw0}\) is the initial switching frequency, \(k_s\) and \(b_s\) are the proportional gain and adjustment coefficient for switching frequency, with \(k_s < 0\). Parameters can be determined by limiting the resonant frequency of the LC filter between \(f_{sw}/6\) and \(f_{sw}/2\). The switching frequency changes only during transient faults, not affecting normal grid-connected operation. By detecting the output current value during transient faults and appropriately adjusting the IGBT switching frequency, the maximum thermal cycling growth can be further limited.
The integrated control structure of the proposed active thermal control method for utility interactive inverters is summarized in the diagram below, combining virtual impedance, adaptive power adjustment, and switching frequency variation.
To validate the effectiveness of the proposed active thermal control method, we conducted simulations using PLECS software. The simulation model is based on a three-phase two-level DC/AC utility interactive inverter with an IGBT electrothermal model. The power module used is Infineon’s FS100R12KT4G three-phase full-bridge IGBT module, with a maximum collector-emitter voltage of 1200 V and a continuous collector current of 100 A. In the simulation, the inverter is directly connected to the grid at t=0 s, and the IGBT junction temperature rises steadily, reaching stability at approximately t=1.2 s. At t=2 s, a transient grid voltage sag fault occurs, with the point of common coupling voltage dropping to about 86% of the rated voltage, and clears at t=2.5 s. After fault clearance, the junction temperature re-enters steady state at t=3.1 s.
The table below summarizes the simulation results for IGBT output junction temperature under different active power commands before and after applying the active thermal control.
| Active Power Command | Peak Junction Temperature Before Control (°C) | Peak Junction Temperature After Control (°C) | Maximum Thermal Cycling Before Control (°C) | Maximum Thermal Cycling After Control (°C) |
|---|---|---|---|---|
| 5.5 kW | 65.59 | 61.24 | 22.04 | 18.68 |
| 6.5 kW | 73.09 | 67.87 | 23.50 | 18.87 |
| 7.5 kW | 81.05 | 74.59 | 25.68 | 19.49 |
As shown, before applying active thermal control, significant instantaneous junction temperature spikes occur at the inception and clearance of the grid transient fault, with peak temperature increases up to 28%. Deeper grid voltage sags lead to larger instantaneous junction temperature spikes. During the fault, due to the regulation of the Q-V loop, the potential difference between the inverter and grid decreases, causing some reduction in junction temperature, but it remains higher than during steady-state operation. After applying the proposed active thermal control method, the overall junction temperature level during the fault state decreases, with more pronounced suppression during transient faults. The addition of virtual impedance suppresses transient current and junction temperature spikes at fault inception and clearance, while adaptive power adjustment effectively reduces output current and junction temperature levels during transient faults. However, to maintain power transmission capability during transients, the active power command is not reduced excessively, limiting the control effect on junction temperature.
Under the synergistic effect of virtual impedance and adaptive power adjustment, the output current level decreases due to the generated current-limiting voltage drop and reduced overall power level during transient faults, leading to a drop in overall IGBT junction temperature. During the grid transient voltage fault, the peak IGBT junction temperature decreases by 4.35°C for a 5.5 kW command. Simultaneously, due to the coordinated switching frequency variation strategy—reducing switching frequency when IGBT current exceeds the rated value and increasing it otherwise—the excessive growth of internal maximum thermal cycling is suppressed, effectively limiting the maximum thermal cycling amplitude, corresponding to a reduction of 3.36°C. For a 6.5 kW command, the maximum thermal cycling decreases by 4.63°C, and the peak junction temperature during the grid transient voltage sag decreases by 5.22°C. Similarly, for a 7.5 kW command, the active thermal control reduces the maximum thermal cycling amplitude and peak junction temperature by 6.19°C and 6.46°C, respectively. Thus, under different active power commands, the proposed active thermal control method effectively suppresses the peak junction temperature level and maximum thermal cycling amplitude. Larger input active power commands yield more noticeable control effects.
To further verify the theoretical analysis, we built an experimental platform for a photovoltaic utility interactive inverter, as shown in the setup. The control algorithm is implemented on a DSP TMS320F28335. Key experimental parameters are listed in the table below.
| Parameter | Value | Parameter | Value |
|---|---|---|---|
| Ambient Temperature \(T_a\) | 25 °C | DC Bus Voltage \(U_{dc}\) | 650 V |
| Reactive Power Command \(Q_{ref}\) | 2000 var | Virtual Inertia \(J_p\) | 0.01 |
| Damping Coefficient \(D_p\) | 9 | Damping Coefficient \(D_q\) | 200 |
| Reference Angular Frequency \(\omega_{ref}\) | 100π rad/s | Reference Internal EMF \(V_{ref}\) | 320 V |
| Virtual Resistance Gain \(k_R\) | 1 | Virtual Reactance Gain \(k_X\) | 0.01 |
| Current Limit Threshold \(I_{lim}\) | 30 A | Power Adjustment Coefficient \(K_f\) | 0.7 |
| Lower Voltage Limit \(V_{g0}\) | 300 V | Initial Switching Frequency \(f_{sw0}\) | 10 kHz |
| Switching Frequency Gain \(k_s\) | -0.1 | Switching Frequency Adjustment \(b_s\) | 12 |
The IGBT module is Infineon’s FS100R12KT4G, with thermal impedance parameters as previously listed. For junction temperature monitoring, the IGBT module is opened to allow temperature sensor contact. Temperature measurement uses AMEC Lab high-precision fiber optic temperature sensors. After the IGBT junction temperature stabilizes, a transient voltage sag fault is set at t=4.5 s, with the grid three-phase voltage RMS simultaneously dropping by 30 V, and recovers at t=6 s, lasting 1.5 s. This fault represents the maximum stable operation achievable on the experimental platform. During the transient fault, due to high output current, the IGBT experiences overcurrent surges.
The experimental results for IGBT output junction temperature under different active power commands during grid transient voltage faults are summarized in the table below.
| Active Power Command | Peak Junction Temperature Before Control (°C) | Peak Junction Temperature After Control (°C) | Maximum Thermal Cycling Before Control (°C) | Maximum Thermal Cycling After Control (°C) |
|---|---|---|---|---|
| 5.5 kW | 62.29 | 61.08 | 3.27 | 2.06 |
| 6.5 kW | 68.46 | 66.92 | 3.15 | 1.61 |
| 7.5 kW | 73.59 | 72.19 | 2.88 | 1.48 |
Before applying active thermal control, the maximum thermal cycling amplitudes reach 3.27°C, 3.15°C, and 2.88°C under different active power commands, with corresponding increases in peak junction temperature. In actual setups, IGBT junction temperature is influenced by cooling conditions, so transient temperature spikes are less pronounced than in simulations. Additionally, the fiber optic temperature sensor’s fast response time of 100 ms cannot detect transient spikes from fundamental frequency current-induced periodic thermal fluctuations. Thus, the measured maximum thermal cycling amplitude is smaller than the actual transient spikes experienced by the IGBT. However, since the focus is on the overall effect of active thermal control, this error is acceptable.
Output voltage and current waveforms from the utility interactive inverter during grid transient faults, with and without the active thermal control method, were captured. Without any active thermal control, the inverter output current becomes uncontrolled after fault clearance, exhibiting sustained oscillations and eventual instability. The oscillating current can reach up to 2.5 times the rated output current. With the proposed active thermal control method, the inverter maintains stable grid connection during faults, achieving low-voltage fault ride-through. The method effectively suppresses overcurrent during transient faults, limiting output current to 30 A and quickly returning to normal after fault recovery. Since the method is applied during transients, the temporary impact on THD is negligible compared to long-term steady-state operation, so detailed THD analysis is not provided.
Output waveforms during grid transient faults with the active thermal control method but without switching frequency variation show that the absence of switching frequency changes does not directly affect grid connection reliability. The inverter can still suppress overcurrent and maintain connection. However, without switching frequency variation, current spikes are more noticeable during fault ride-through. According to the switching frequency equation, overcurrent during transient faults reduces IGBT switching frequency. While higher switching frequencies ensure better power quality and lower current harmonics, they also lead to faster dynamic response and higher switching losses, potentially causing current spikes and transients. During transient faults, lower switching frequency reduces the current change rate (di/dt), resulting in smoother inverter output current and improved waveforms. Additionally, lower switching frequency implies longer switching periods, giving the control system more time for feedback regulation during transients, reducing system response speed and introducing lag, thereby mitigating overshoot and oscillations from rapid reactions. Therefore, the proposed active thermal control method is effective, and the included switching frequency variation strategy further improves output current waveforms during transient faults, enhancing inverter transient stability.
In conclusion, as utility interactive inverters with VSG control gain attention, their power supply reliability and system stability become increasingly important. As a key core component within inverters, improving IGBT reliability is crucial. For utility interactive inverters, the overstress on IGBTs during grid transient voltage sag faults is particularly significant. This paper proposes an active thermal control method for IGBTs in utility interactive inverters during transient periods, with simulation and experimental validation yielding the following conclusions:
- During transient faults, the potential difference between the utility interactive inverter and grid, combined with small line equivalent impedance, causes large output current surges, which is a primary reason for increased IGBT junction temperature during faults.
- The proposed active thermal control method reduces the peak junction temperature level and maximum thermal cycling amplitude in utility interactive inverters during transient faults. Larger input active power commands result in more pronounced control effects. For example, with a 7.5 kW command, the maximum thermal cycling amplitude decreases by nearly 50% after applying the method.
- The method enhances IGBT reliability while ensuring stable grid connection of the utility interactive inverter, avoiding output current oscillations after fault clearance. Additionally, the switching frequency variation strategy further improves output current waveforms during transient faults, enhancing inverter transient stability.
Future work could explore optimizing control parameters for different fault scenarios and extending the method to other types of utility interactive inverters or power electronic converters. The integration of advanced thermal monitoring techniques and machine learning for predictive thermal management could also be investigated to further improve reliability and lifespan in renewable energy systems.