Page 272 - Modern Control of DC-Based Power Systems

Simulation 233

and settling time while the LQG-Centralized control exhibits the highest
voltage drop. The LSF exhibits the slowest dynamics due to the chosen
dynamics through pole placement and the nonlinearity compensation. All
three decentralized virtual disturbance-based controls display a behavior
comparable to that already observed in the cascaded system setup and
show in all aspects better performance compared to the centralized LQG
formulation. Characteristic values, like those calculated in the cascaded
system, are presented here in Table 6.6. It should be clarified that due to
the steady-state error the numerical values are calculated for the end value
of voltage and current, taken at t 5 0.39 s.
To avoid the steady-state error after the load increase two options are
possible. One could be to design a prefilter, which actually can only be
designed for a specific load level and would mean that at other load levels
an over- or undervoltage appears. The alternative is the implementation
of a secondary control loop consisting of a PI controller with the feed-
back of the bus voltage as mentioned in Subsection 6.2.1. The latter
option was chosen.
It’s to be pointed that the introduction of a Droop PI outer control
loop for steady-state error suppression will lead to a change in transient
behavior as another control loop interaction is introduced. The corre-
sponding results can be analyzed in Figs. 6.12 and 6.13 where the reader
can observe that due to the additional PI-Loop the Centralized LQG
improves in performance; this resides in the fact that the centralized for-
mulation profits more from an adjusted voltage setpoint. For the other
controllers the decrease in voltage drop corresponds directly with the
reduction in steady-state error. The rise times for the decentralized virtual
disturbance and Synergetic control are 0.0022 s, while for the Centralized
LQG it is 0.0032 s, and for the LSF it corresponds to 0.0037 s.
After presenting the possibility of mitigating the steady-state error in
the bus voltage, all further test cases are performed without the
Droop PI as its inclusion would deviate the analysis from a pure stabiliz-
ing action point-of-view towards a system level analysis, and in certain
combinations retuning control coefficients would be necessary.
In an ISPS consisting of multiple LRCs also the case of a loss of avail-
able generation capability needs to be considered. At the time instant
t 5 0.3 s LRC 3 is disconnected and LRC 1,2 are accountable to supply the
30.9 MW load. Thus, the load factor is 0.515 of the installed power
before and 0.775 after the reduction of installed power. The results for
the control strategies are shown in Figs. 6.14 and 6.15. Those results