REPOSITORY > RESULTS

Abstract

Despite their environmental benefits and high energy conversion efficiencies, solid oxide
fuel cells (SOFCs) have yet to achieve their full market potential. Currently, the main
focus of SOFC research is on extending the lifespan of the fuel cells. One way of achieving
higher reliability and durability of the fuel cells is through the development and application
of diagnostic, monitoring, and control algorithms.
This dissertation proposes an integrated control of the SOFC system, where longevity of
the stack is directly incorporated in the control strategy, which is a major novelty compared
to the conventional control systems. Conventional controllers typically manipulate flows
and temperatures of the stack to keep the system running in safe operation. Conversely,
this thesis develops an integrated control system that has direct impact on the durability
of the stack, thus making the system health-aware. The implementation of the proposed
solution requires development of several conceptually different, yet interconnected modules.
The controller consists of two layers. The low-level part keeps the system operating
in the corresponding set-point, while the supervisory controller (SC) defines the said setpoints.
The SC is formulated as a nonlinear program which minimises the cost function
of two conflicting objectives: (i.) the degradation rate of the SOFC stack, and (ii.) system
efficiency.
While the system efficiency can be easily calculated, the degradation rate has to be
estimated via structure of Kalman filters (KFs). The so called state of health (SoH) estimator
implements a lumped stack model in an unscented Kalman filter (UKF) algorithm
to recursively identify the area specific resistance (ASR) of the stack. In the subsequent
step, a model describing temporal evolution of the ASR is identified from the semi-past
data, and remaining useful life (RUL) predictions are obtained accordingly. The developed
approach is invariant to operating conditions which gives it a major advantage over the
conventional solutions.
Physical sensors are essential in control and monitoring of any system. However, installing
in-situ instrumentation becomes impractical on an industrial scale. Therefore, this
thesis develops soft sensors capable of estimating process variables such as stack temperature
gradients and fuel gas composition. The temperature soft sensor can be directly
exploited by the low-level controller to constrain the stack gradients, while the gas composition
estimates are directly employed by the SoH estimator.
Both the soft sensors and SoH estimator are experimentally validated on a 10 kWSOFC
system. The integrated control strategy, however, is fully validated on the model of the
said 10 kW system. The simulation study shows the capability of the developed control
system to minimise the degradation rate of the stack, thus extending the total life span of
the system. It is shown that an increase in lifespan comes at the cost of system efficiency.
Furthermore, the control system with an integrated SoH estimator and soft sensors
were implemented on standard industrial equipment. All of the algorithms were tuned to
the experimental system and run on-line. Due to the issues related to poor condition of
the system, only a preliminary experimental validation is discussed in the thesis.