Supporting Automotive Fuel Cell Applications

Fuel cell electric vehicles (FCEVs) have advantages over the internal combustion engine (ICE) in terms of emissions, fuel economy, and fuel flexibility. To understand the operation of the power electronics portion of the fuel cell system, we first have to look at a typical electric power distribution system that consists of two voltage buses, as you can see in Fig. 1.

The fuel cell's high-voltage bus (typically 200V to 400V) supplies the power for traction drive and other high power loads such as air compressor drive, air conditioner compressor drive, etc. Also, a low-voltage bus with a 12V conventional vehicle battery supplies the conventional automotive loads such as lamps, and control electronics, etc. Similar to an ICE vehicle, the 12V battery also provides the energy to start the fuel cell engine system.

The bidirectional dc-dc converter is a key subsystem managing energy flow between two buses within the FCEV.

Design Challenge

A major challenge in designing a bidirectional 3kW dc-dc converter with 12V battery is high battery current. For example, assuming in the boost mode, the battery voltage is 10V, to obtain 3kW output at 85% efficiency requires an input current as high as 350A.

This high current makes it a nontrivial task to design the power stage for the low-voltage side of the converter. The switching device is the primary source for power loss. Conduction loss is the dominant loss component when using MOSFETs in this low-voltage application, thus, it's the determining factor for device sizing.

Conduction loss in a MOSFET switch is:

Where:
I_RMS = rms drain current
R_DS(on) = On-resistance

The rms current in a switching device depends on the average power the converter processes. It also depends on the converter topology, because different topologies cause different current waveshapes — even with the same average current. The best topology is one that produces lower rms current for the same processed power.

Use current- or voltage-fed configurations for the low-voltage side converter. Fig. 2, on page 34, shows a current-fed full bridge (top) and voltage-fed full bridge (bottom).

Assuming battery, V_b, provides the same power (i.e. the same current), we can derive the current waveforms of Q1-Q4 shown in Fig. 3, on page 34. With this, we can derive the rms value of device current:

For a current-fed full bridge, the rms current for each is:

For the voltage-fed full bridge:

Where:
I_bav = Average battery current
D = Duty cycle determined by voltage control
The ratio of I_RMSV/I_RMSI is:

Fig. 4, on page 37, is a plot of Equation (4). It shows that over a wide range of duty cycles, the device rms current of a voltage-fed full bridge is at least 1.5 times that of the current-fed full bridge. This is significant in the low-voltage and high-current application such as this one. As a result, we selected the current-fed arrangement for the low-voltage full bridge.

Unfortunately, the circuit (top) in Fig. 2 will not work properly because it has a current source inverter directly connected to a transformer. Due to the transformer's leakage inductance, at every commutation instant, excessive voltage spikes occur across the inverter bridge. Without a proper voltage clamping snubber, the bridge could be damaged by the overvoltage. Therefore, we must add a passive clamp snubber (Fig. 5, on page 38) to suppress the overvoltage. Here, the resistor dissipates a significant portion of the energy stored in the clamping snubber, which adversely affects the converter's efficiency. This is particularly severe for low-voltage, high-current application because the energy processed by the clamp snubber is proportional to the square of the input current.

One approach for improving the efficiency of the passive clamp snubber is to use an active clamp snubber. This involves the addition of a clamping capacitor and a clamping switch across the full bridge. During the commutation period, the clamping switch transfers the energy in the transformer's leakage inductance to the clamping capacitor, limiting the voltage spike. The circuit then recycles the energy stored in the clamping capacitor back to the transformer to enhance converter efficiency.

However, the active clamp snubber has some limitations. The clamping switch must handle the same full load current as the main converter switches — but at twice the frequency, possibly overstressing the clamp switch. Also, the clamping capacitor design requires some difficult trade-offs. To minimize the transformer's leakage inductance, we want to minimize the energy circulating through the clamping capacitor, thereby minimizing the rating for all clamp circuit components. However, to maintain proper operation of the active clamp snubber, the leakage inductance and the switching frequency determine the clamping capacitor value.

Where:
C_c = Clamp capacitance
L_lk = Transformer leakage inductance
f_s = Switching frequency of converter and clamp switch

Equation (5) suggests that the smaller L_lk, the higher C_c — if f_s is a constant. This contradicts the claim that the lower the leakage, the lower capacitance rating for energy handling capability.

This trade-off is more significant for high power applications. Another approach is a passive clamp snubber with an energy-recycling converter, as shown in Fig. 6. This new clamp circuit addresses the limitations of the simple passive clamp snubber (Fig. 5) and the active clamp snubber.

Compared with the passive clamp snubber shown in Fig. 5, we can replace the dissipative resistor with an energy-recycling buck converter (Fig. 6). Therefore, almost all the energy stored in the transformer's leakage inductance recycles back to the source (battery), which significantly improves converter efficiency.

Making it Work

The dc-dc converter has two modes of operation for the vehicle: start-up and normal drive. A hydrogen fuel cell vehicle stores hydrogen in a high-pressure container and an air compressor supplies oxygen. In start-up, the system transfers energy stored in the 12V battery to the high-voltage bus, so the air compressor drive can deliver oxygen to the fuel cell stack. Oxygen interacts with the hydrogen supplied from the high-pressure storage component and produces the dc electricity for the high-voltage bus. During the starting process, the energy flows from low voltage to high voltage using the dc-dc boost converter.

Once the system establishes a stable dc-voltage output, the vehicle can enter the normal drive mode. In this mode, the fuel cell supplies all the vehicle's energy, including the low-voltage load. Now, the dc-dc converter works as a buck converter, transferring energy from the high-voltage bus to the low-voltage bus.

Design features of engineering prototypes include the following:

Modular design to achieve high levels of flexibility and reconfigure the unit for various voltage levels.
Single stage, symmetrical power converter topology allows bidirectional power transfer using one set of power components.
High efficiency in boost and buck operational mode.
CAN capability to facilitate system integration needs.
Air cooled stand-alone unit for easy unit placement.

A stand-alone bidirectional dc-dc converter is a repackaged version of the converter integrated with the traction inverter for FCEVs. Fig. 7, on page 38, shows the integrated powertrain with the dc-dc converter in it.

The converter uses a symmetrical topology capable of fully bidirectional operation using one set of power switches and magnetics. As shown in Fig. 6, on page 38, the low-voltage side contains an inductor to limit the peak current on the 12V side, significantly reducing the rating of the transistors on the low-voltage side.

The converter achieves high efficiency through the use of single-stage power conversion and synchronous rectification. Figs. 8(a) and (b) show the measured efficiency for buck and boost operations, respectively. For most areas of the operation region, the efficiency is better than 90% — with peak efficiency approaching 95%.

Stand-Alone, Air-Cooled Unit

You can employ this high-performance, high-power, bidirectional dc-dc converter in automotive and other applications, which requires repackaging the integrated, liquid cooled dc-dc converter into a stand-alone, air-cooled unit.

To enhance its flexibility, this new package is modular. Fig. 9 shows the major modules. By changing different transformer turns ratio and power boards, you can configure the unit for a range of input and output voltages.

We achieved 200mV output voltage regulation in the prototype design for the high- and lower-voltage versions of the stand-alone converter and maintained the output voltage regulation at two-thirds specified value over a range of inductive and resistive loads — with 90% efficiencies. We'll perform more testing of transient response and output voltage ripple under a range of environmental conditions.

The use of lower cost automotive-grade power electronics in an integrated package provides the basis for higher performance, cost-effective, dual-voltage power networks. Special control circuits furnish the means to control resistive, capacitive, and inductive loads under dynamic load and varying environmental conditions. The development of water- and air-cooled models expand the potential use of these converters from automotive applications into material handling, and other industrial applications.

Notes

The actual performance of the existing prototype compared with customer target in parenthesis. It's possible to refine the existing design to achieve the target.
Quantities in the table represent the customer-specified target. Actual performance must be established.

For more information on this article, CIRCLE 333 on Reader Service Card