IEEE Electrification Magazine - December 2017 - 80

The use of the
emulator is less
restrictive than
using a fuel cell in
terms of the cost
of gas and the risk
of damage.

calculated for its converter, full
compensation for the load power
high frequencies is close to being
achieved. The mean value for these
losses does not affect the control loop
for the state of charge of the stored
energy, as the losses are considered to
be an additional load and go through
the high-pass filter for the frequency
sharing in Figure 11(a). The disadvantage of this method is that, if the estimation of these losses is incorrect or
if they change over time (through
component wear and tear), compensation no longer functions. The ideal solution would be a
bidirectional dc-dc current converter for which we manage the current on the high-voltage side.
There is a similar situation for the strategy shown in
[Figure 11(b)] because the converter associated with the
fuel cell is a dc-dc unidirectional current boost converter (in the fuel cell to bus direction). It enables the management of the fuel cell current or the bus voltage.
Here, we want to manage the current sent to the bus.
As for the previous case, the losses are compensated by
a theoretical estimation and considered to be an additional load. They therefore go through the low-pass filter for frequency sharing and are compensated using a
mean value. The compensation for these losses has
less of an effect on the hybridization performance as
they have a low-frequency effect, which would be compensated for by the control loop for the state of charge
of the stored energy.

energy cannot fulfill this mission for
long, it becomes fully discharged
when the fuel cell reaches the mean
current required by the frequencysharing filter. Meanwhile, the bus
voltage is no longer controlled as the
stored energy is discharged and the
mission is not fulfilled. At startup,
there is an issue initializing this filter. Let us suppose that we are not
aware of the steady state; we, therefore, cannot initialize the filter at its
value in the steady state or at the
value of the current at the startup
time (in which case, we would risk starting on a peak). To
compensate for these difficulties, we suggest successively switching to slower and slower filters to arrive at the
final desired filter. Therefore, the chosen solution (shown
in Figure 12) proposes starting with a fast filter (for which
the cutoff frequency depends on the dynamics acceptable by the fuel cell and its fluidic regulations), while
other slower filters start at the same time. Given the
order of two and the damping factor ( 2 /2) of the filters,
they naturally have a slight exceedance, which means
that their values will inevitably coincide for a limited
profile. This will enable switching to the lower-frequency
filter with no value jumping. To prevent a return from a
low-frequency filter to a higher-frequency filter, switching should be irreversible.

Small-Scale Experimental Validation
Description of the Experiment Performed

Problem Starting the Frequency Sharing

Amplitude (p.u.)

Effective filtering for the fuel cell requires a low cutoff
frequency for the frequency-sharing filter (here, 10 mHz).
This means a long starting time for the filter. Used as is,
the filter rise corresponds to the rise in the fuel cell current, i.e., during that whole time, the stored energy provides the power required for the loads. As the stored

2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0

100
Time (s)

ILoad
ISlow Filter

IFast Filter
IMed Filter
ISwitched Filter

Figure 12. An example of filter switching. p.u.: per unit.

I E E E E l e c t r i f i cati o n M a gaz ine / DECEMBER 2017

150

The experimental validation shown in Figure 13 takes
place in two stages. The first is an experiment with the
fuel cell replaced by an emulator, and the second is with
an actual fuel cell. These trials will be used, among other
things, to assess the relevance of an experiment employing an emulator and will guide the choice for full-scale
validation. In addition, the use of the emulator is less
restrictive than using a fuel cell in terms of the cost of
gas and the risk of damage. It is used to adjust and validate all the control loops and push the strategies to their
limits, with minimal risks for the equipment.
For the fuel cell, we have a stack of 50 cells measuring
130 cm2, with H2/O2 operation, provided by AREVA Energy
Storage Inc. [Figure 14(a)]. For the stored energy, we are
equipped with 16-V, 58-F ultracapacitors (Maxwell Technologies' BMOD0058 E016). They will be connected in
series, in a sufficient number to adapt to the level of the
reference voltage for the stored energy [Figure 14(b)].
The load profile will be to scale based on the bench elements, i.e., 4 kW mean power and 8 kW peak power. Given
the unique elements of the bench, the experiment could
not take place in the exact sizing conditions resulting
from the optimization.

Table of Contents for the Digital Edition of IEEE Electrification Magazine - December 2017