research highlights
Hardware Testbed
CURENT has developed a hardware demonstration platform using power electronic converters for
demonstrating wide-area voltage and secondary frequency control in its hardware test bed. The test bed
uses two power electronic clusters to emulate different parts of a small scale power system. Each of the
power electronic clusters consists of several power electronic converters emulating different components
in the system. Load is emulated using ZIP+ induction machine models, which includes the dynamic
behavior of the induction machine. The sources range from typical synchronous generators to full scale
emulation of photovoltaic and wind turbine models. The entire system is controlled using a LabView
interface incorporating many of the features found in modern control rooms. The control room also allows
for incorporating real-time wide-area measurements and closed loop operation of voltage and frequency to
demonstrate techniques developed by other thrusts within the center.
As an example, one of the wide-area closed loop voltage control demonstrations in the hardware test bed is
done by emulating the load center and boundary busses in an area and using PMU measurements to predict
the voltage stability margin in real time. By knowing the stability margin in real-time, actuators can be
controlled before the system reaches instability.
Robust Dynamic State Estimation Using Wide-area Synchronized Measurements
Accurate dynamic estimation of generator states and, in particular, of its frequency, is essential to efficient
control of ultra wide-area electric power grids. Such state estimates can also be used in prevention of
cascading failures and dynamic security analysis.
The primary objective of this research is to
provide dynamic estimates of generator states
that are robust with respect to timing and system
parameter inaccuracies and, in addition, can
minimize the effects of network perturbations
such as transmission delays, corrupted sensor
measurements (“bad data”) and information
packet drops. To achieve this objective we rely on
a recently developed robust version of the Kalman
filter, which we augment with delay mitigation and
bad data detection capabilities.
We are also developing compact performance
metrics that can predict the quality of our
dynamic state estimates in the presence of
various network perturbations. Our goal is to use
such metrics to evaluate and compare alternatives in sensor deployment, so as to provide guidelines for
optimal deployment.
3