Active thermal management for a single-phase H-Bridge inverter employing switching frequency control

A thermal controller can be employed to reduce the thermal swing consequent to power cycling. This is demonstrated in this work for a full bridge inverter without a priori knowledge of the mission profile to which the converter is a subsystem, with the objective to reduce the thermal cycling without measurement of the junction temperature. The performance of the thermal controller is experimentally demonstrated by comparing a system, which operates with constant switching frequency with a system equipped with differently tuned thermal controllers. The impact of active thermal control on lifetime is investigated. 1) Introduction Power electronic converters are widely used for electric vehicles, aircrafts and grid connected applications, such as renewable energies. The reliability of these systems is essential, because failures cause downtimes and thus costs [1]. The most important failure mechanism is based on thermal cycling and is addressed in [2],[3]. Thermal cycling is affected by ambient temperature variations and power cycling, leading to mechanical stress within the module, which causes aging and finally failures. Semiconductor manufacturers optimize hardware to increase the reliability. Beside improved hardware, active thermal control enables to decrease the effects of power cycling [4]. However, in many systems, the main target is to maximize the efficiency, which is especially true in photovoltaic applications. Another way to improve the cost efficiency is to increase the lifetime of the system by a reduction of thermal cycling. In [5] active thermal control has been introduced to drive the system in safe operation during a fault of a subsystem, but not to increase lifetime during normal operation. This work proposes to modify the switching frequency to control the semiconductor losses, leading to reduced thermal cycles and increased lifetime under fast changing load conditions. The thermal control relies on electrical measurements and does not require additional sensors to reduce the thermal cycling. It is shown that the appropriate tuning is mandatory to find the optimal system behavior. The active thermal control is first described in theory and then tested on an experimental setup consisting of an H-bridge inverter connected to a passive load. 2) Active thermal control The junction temperature Tj of the power semiconductors depends on the losses of the device Ploss, the thermal resistance between junction to ambient Rth,ja and the ambient temperature Ta. Tj = Ta + Ploss ⋅ Rth,ja (1) Since the ambient temperature cannot be influenced and the thermal resistance is dependent on the hardware, the losses are controlled to reduce the thermal cycling. In high power applications, usually the switching losses Psw and the conduction losses Pcond are dominant, while driving or blocking losses are neglected: Ploss = Psw + Pcond (2) The conduction losses can be approximated with a second order polynomial, where iload pk is the peak alternating current, m is the modulation factor and cos (φ) is the power factor: Pcond = ( 1 2π + m⋅cos(φ) 8 ) ⋅ uce,0 ⋅ iload pk + ( 1 8 + m⋅cos(φ) 3π ) ⋅ rce(iload pk ) 2 (3) The switching losses can be expressed with the turn on losses Eon, the turn off losses Eoff, the switching frequency fsw, the dc-link voltage Udc, the load current and an empirical factor c, suggested to be chosen to 1.3 in [7] Psw = fsw ⋅ (Eon + Eoff) ⋅ i load pk Iref ( Udc Udc,ref ) c (4) Based on the bondwire liftoff failure mechanism, which is characterized by the Coffin-Manson equation, the power semiconductors can undergo an specific number of thermal cycles Nf, which is exponentially dependent on the magnitude of the thermal cycles ΔTJ and average temperature Tj . A = 302500, α = −5.039 and Ea = 9.89E − 20 are the Scheuermann fitting parameters, found for average power modules while k is the Boltzmann constant. As can be seen from the α value, the importance of the amplitude of the thermal for lifetime is critical. The impact of Tj is less relevant, but an increased temperatures reduces the expected lifetime as well. Nf = A(ΔTJ) e ( Ea kTj ) (5) The idea proposed in this paper is to reduce the thermal cycling by regulating determined profiles of losses via switching frequency variations. Some hypothesis will be considered: The full-bridge (shown in Figure 1) operates in hard-switching conditions at variable modulation ratio on a RL load. The H-bridge is driven by a unipolar modulation (with freewheeling in the upper and lower sides of the H-bridge). The modulation index is changed according to a mission profile. Figure 2 shows the proposed control method. The common operation resides in the losses model, shown in equations (3) and (4), that allows estimating the average losses over a fundamental output period. The difference between the actual losses and a low-pass filtered version constitutes the power difference that the active thermal control aims at reducing. A look-up table that links the power difference to the frequency increase is adopted, so that a higher switching frequency is selected when the output power is reducing (with gain ∆fmax ∆Pmax ). Instead, when the output power is increasing, the switching frequency is locked to the minimum. This non-linear control aims at preventing the cooling of the power module when the output power is reduced, but does not work when the modules is heating up. This behavior