Page 247 - Modern Control of DC-Based Power Systems

Control Approaches for Parallel Source Converter Systems 211

advantage of ensuring, by definition, the global asymptotic stability of the
system. However, it has a complicated control law whose implementation
might be challenging in a digital controller.
The fourth approach was an observer-based control which relies on
the concept of 2DOF controller. Here, only the bus voltage is measured
while the local inductor current is estimated. This is because in systems
with switching converters the current has a strong ripple and a direct
measurement has to be filtered; thus, it could be replaced by a software
measurement in form of the Kalman filter. Its design compared to the
LSF is more complex as steady-state Kalman filter, covariance matrices,
and correlation parameters have to be selected. The usage of the Kalman
filter is not limited to software measurement; it also performs in its aug-
mented version the additional estimation of the nonlocal part of the
power system (i.e., the estimation of the other converter currents), and
generates, in a 2DOF structure, also the set-point trajectory. It should be
noted that such set-point trajectory generation relies on phase-minimal
system behavior. Furthermore, due to the application of steady-state
Kalman estimation, the estimated system is always linearized in the oper-
ating point. Both centralized and decentralized solutions are possible; a
centralized control has knowledge about all inductor current estimates
while the decentralized control has only knowledge of the local inductor
current estimate. Both variations of this control approach rely fully on lin-
ear control theory.
The fifth approach combines the virtual disturbance concept with a
nonlinear control law, namely, Backstepping. Backstepping is a procedure
which is based on Lyapunov Control Function; its synthesis procedure is
less complex than the previous decentralized approach, especially as it
does not need a set-point trajectory while it also uses estimated inductor
currents.
The sixth approach is Adaptive Backstepping, where the nonlocal part
is no longer assumed as a current source and estimated by a Kalman filter,
but corresponds to Power Estimation. This leads to a complete nonlinear
decentralized controller while also having a nonlinear plant. For both
Backstepping approaches a centralized realization does not seem possible
as those control laws require the system matrix to be written in a lower
triangular form.
The seventh approach was the H N approach where the derived con-
trollers are internally stable and provide guaranteed stability margins. One
of the drawbacks of the LQG control approach is that it seeks to optimize