GaN Components Utilized in Pre-Sizing a Modular High-Power DC/DC Converter
Significant revisions to an aircraft’s electric power supply architecture ensure optimal performance.
Recent trends in engineering aircraft are shifting toward a more electric aircraft (MEA) paradigm—increasing the use of electric energy over pneumatic and hydraulic energies to power on-board systems. Implementation of MEA has some undeniable advantages, such as energy efficiency and low-cost maintenance, which are imminent to see from its increasing implementation in recent aircraft like the Airbus A380 or the Boeing 787. To incorporate the same, significant revisions to the electric power supply architecture are inevitable to ensure the best performance.
The electric power architecture is based on multiple on-board grids, both interconnected and galvanically isolated. Also, there are two network voltage levels: a 28-VDC standard low-voltage DC (LVDC) bus to power the control devices and a 270-VDC/540-VDC high-voltage DC (HVDC) bus for power actuators. Power conversion is key to the implementation of a more electric approach, outlining the quantifiable benefits of switching from a single converter based on directly adapted components to a multi-converter structure adopting the most appropriate components to achieve a given specification.
Major challenges when it comes to MEA are developing smaller and lighter electric power systems; hence, the architecture is to be centered around power density and power-to-weight ratio. According to the literature, wide-bandgap technologies can be credited to achieving up to a 2-kW/kg power-to-weight ratio with the use of a combination of GaN and SiC components. The use of GaN components also proposes lower power losses, in turn contributing to conversion efficiency and reducing the size of the thermal management system. They can also withstand higher junction temperatures.
Designing a dual active bridge (DAB) in this case suits the contextual needs, as it further meets the requirements of galvanic isolation, high power density, high power ratings and high efficiency. Considering the use of GaN components, a modular architecture based on the series and/or parallel connection of lower-power elementary bricks is necessary (Figure 1).
Simulating the modular electric power architecture
Rather than carrying out tedious and expensive implementations and tests to bring about technological advancements, a simulation-based design of the power electronics involved in the interconnection between LVDC and HVDC networks in MEA implementation is more convenient. The electrical power-conversion function is performed by a combination of elementary bricks. These are based on DAB technology and are controlled using a single-phase-shift control strategy. Figure 2 outlines the 11 macro-parameters of the DAB converter in the model.
The models generated also account for several losses associated with the semiconductors (the GaN components), the magnetic components and so on. Considering transistor conduction losses, reverse conduction losses during dead time, iron losses, copper losses and more are also crucial in determining the power-to-weight ratio of models, as it is the main performance factor in focus. The paper in the references walks through several equations to calculate power losses due to conduction losses, iron losses and more. These equations play a cumulative role in calculating the power-to-weight ratio.
Likewise, several design decisions were made to significantly minimize losses:
- The use of planar transformers, as they meet the requirements of modern power electronics to limit skin effect and proximity effects
- Parallel windings to reduce copper losses and current density at a low voltage
The simulated model design also made computations for mass estimation models. The generated models must pass several criteria for performance to be used in experiments—optimization is a subsequent requirement.
Inculcating optimization into the models
To proceed with the optimization process, adding physical constraints must be given precedence, especially considering the transformer. This process constitutes introducing coefficients and constants, which ground the equations in reality and also aid them to suit the power electronics requirements.
The optimization of the simulated model designs was conducted using a particle-swarm–optimization algorithm. The required optimization is with respect to the power-to-weight ratio. While executing the optimization algorithm, the parameters were adjusted to generate three optimal solutions and one sub-optimal solution. Among the optimal solutions, Case A used an E43 ferrite core, Case B used an E58 ferrite core and Case C used an E64 core.
Case B resulted in a promising 4.5-kW/kg power-to-weight ratio—performing twice as high as the state-of-the-art. Hence, GaN transistors can forge a path toward significant advancements in power electronics, irrespective of their association with aircraft implementations and tests. Meanwhile, Case C had a bigger core, resulting in higher low-voltage currents along with unnecessary winding capacitors, which in turn limits the switching frequency. Case A, with a smaller core, had a significantly lower low-voltage current, leading to lower copper losses but increasing the GaN driving energy and core losses.
The sub-optimal solution had a lower power density along with an additional 0.5 points of efficiency compared with Case B.
Note that there were two winding configurations suiting the scrutiny that contributed to achieving a high resonant frequency and therefore a high switching frequency.
Experimentation validation and result of the proposed models
Validating the implementation hypotheses is the subsequent crucial step. For this, it must be checked that at least four GaN transistors may be driven using the same gate driver and routed with insignificant parasitic inductances and low resistive connections. Thermal management based on a single heatsink per bridge should also be validated.
The first experimental trial included:
- High-voltage and low-voltage bridges based on GaN systems and EPC components
- A DAB of 1.9 kW with a frequency of 304 kHz
It resulted in a difference of approximately 15% between the measurement and model at a frequency of 300 kHz. Such a difference is attributed to the fact that the modeling was done in 2D, while a transformer is in 3D. However, the objective to reduce the mass remains largely accomplished. The simulation-based approach, being generic and quick, can be used in examining the optimum defined regarding the evolution of the technological characteristics of the various components used in the power converter. The results of the performance index achieved here are the consequence of using the new GaN components, especially given their parallelization.
