Views: 0 | Downloads: 0
The proton exchange membrane (PEM) fuel cells are close to be competitive with the
conventional energy converting devices. However, the durability and reliability of PEM fuel
cell systems have to be increased in order to meet end-user’s expectations. The condition
monitoring (CM) of PEM fuel cells is considered as one of the required tools to enhance
the operation performance of the PEM fuel cell power systems.
This dissertation addresses the problem of reliable CM of PEM fuel cells. It provides
a novel CM methodology, which manages each step of the CM, from the initial data
acquisition to the final output in the form of a unit-free condition indicator (CI). The
methodology is based on the well adopted electrochemical impedance spectroscopy (EIS)
diagnostic method.
The conventional EIS method relies on the consecutive probing of the fuel cell with a
set of single-sine signals of particular frequencies and computation of the corresponding
points in the impedance curve. The approach proposed in this dissertation differs from
the conventional one in (i) the use of frequency-rich excitation signal and (ii) handling
impedance as a complex random variable. The excitation signal is the pseudorandom
binary sequence (PRBS) waveform. Use of such an excitation signal results in a substantial
reduction in the time required for conducting the impedance estimation. The short probing
time is especially convenient for the CM systems. Moreover, in order to achieve high timefrequency
resolution of the impedance estimates, the continuous wavelet transform (CWT)
with the Morlet mother wavelet is used as the focal signal processing tool. The statistical
properties of the employed PRBS signal, together with the CWT, enables the impedance at
individual frequency to be treated as a complex random variable. This allows for rigorous
selection of the initial CM alarm thresholds as well as for the information aggregation
into a single CI. The strength of such an approach arises from two important engineering
aspects. First, at commissioning or after a maintenance action, the tuning of the CM
system requires only the data from normal (i.e., fault-free) operation of the PEM fuel cell
system and the tolerated probability of false alarm (PFA). Second, the ultimate output of
the CM system is a unit-free CI that describes the current condition of a PEM fuel cell.
The dissertation also presents the designed 90-channel fuel cell voltage monitor (FCVM)
and modular direct-current-to-direct-current (DC-DC) converter. The interconnected FCVM
and DC-DC converter, together with the CM algorithm, form an embedded CM system for
PEM fuel cells. The interconnection enables the injection of the excitation signals and the
acquisition of measurements that are required for fast online estimation of the impedance
of PEM fuel cells. The system is both commercially viable and accurate enough to perform
the measurements required for the CM purposes.
The complete CM setup was tested on a commercially available PEM fuel cell power
system. For the validation of the CM methodology, the PEM fuel cells were subjected to
water management faults, which were successfully detected and evaluated by the CI.