IEEE Electrification - September 2020 - 98

control nodes in the cyber layer, i.e., each control node is a
neighbor to every other control node, and in this case, the
communication graph forms a complete graph.
In terms of implementation of a distributed control
architecture on the C-HIL testbed, six Arduino Due microcontrollers serve as the distributed control nodes. As
described before, each device is interfaced with an Ethernet shield that allows the control nodes to communicate
with the Typhoon HIL device via the Modbus TCP/IP protocol so as to enable the monitoring and control of the emulated microgrid assets. In addition, each control node also
has a wireless module that allows the control nodes to
communicate and exchange information with their
neighbors. The C-HIL setup for the distributed architecture
is illustrated in Figure 8(b).
Each control node implements several algorithms, e.g.,
the ratio consensus algorithm and the accelerated primal-
dual algorithm, which enable the distributed implementation of different control functions. In particular, the
ratio-consensus algorithm serves as a primitive for implementing a wide range of control functions, including secondary frequency control, secondary voltage control,
optimal generation asset dispatch, and the provision of
ancillary services to the bulk grid. The accelerated distributed primal-dual algorithm serves as a primitive for
implementing several control functions, including the
selection of optimal DER set points, voltage control, and
provision of ancillary services to the bulk grid. The speed
at which the distributed architecture carries out the power
system control function depends on how fast the distributed algorithms converge. The rate of convergence heavily
depends on the connectivity of the communication network. For example, for the case where each distributed
control node is directly connected to every other control
node in the system, the convergence speed is the same as
that achieved with the centralized scheme. Each control
node uses the previously mentioned algorithms with the
information they acquire locally and from neighboring
exchanges to calculate new set points for the controllable
assets of the microgrid. To close the loop, they send out
the new set points to the DERs and controllable loads
within a microgrid emulated in the Typhoon device.

Case Study: Benchmarking Distributed and
Centralized Architectures
This section presents one of the applications of the C-HIL
testbed. We describe the testing and performance comparison of the centralized and distributed control architectures
when utilized to provide secondary frequency control to an
islanded ac microgrid. We start out by describing the
microgrid secondary frequency control and the motivation
behind carrying out such a comparative case study. We
describe the testing setup and provide the testing results,
comparing the performance (in terms of the system
response time) and resilience (in terms of withstanding the
failure of a control device) of both control architectures.

I E E E E l e c t r i f i cati o n M agaz ine / SEPTEMBER 2020

Microgrid Secondary Frequency Control
A microgrid can operate in both grid-connected and
islanded modes. In islanded mode, frequency control is
a major problem; this is due to the intermittent nature
of renewable-based DERs, e.g., photovoltaic installations,
and the utilization of power electronic inverters to interface DERs to the microgrid, leading to low or no rotating
inertia. Among the various frequency-control objectives,
a key one is the secondary frequency control, which
entails ensuring that, following an operating point
change, the system-wide frequency returns to its nominal value.
Over the years, several coordination and control
schemes have primarily utilized centralized and decentralized decision-making approaches for microgrid secondary frequency control. These schemes have several
limitations: e.g., the centralized decision-making
approach is susceptible to a single point of failure, while
the decentralized decision-making approach typically
lacks the ﬂexibility that is necessary for a seamless integration of additional resources. An alternative, the distributed decision-making ap--proach, has gained some
popularity among researchers in the last decade. In theory, coordination and control schemes based on the distributed decision-making approach should overcome the
limitations of its other counterparts, but it is important to
quantitatively assess the performance of each approach
to identify its merits and demerits. Through this case
study, we show how our C-HIL testbed can be utilized to
test, validate, and compare the performance of a distributed decision-making approach to that of a centralized
decision-making approach in the context of performing
secondary frequency control.
The objective for the secondary frequency control is
to ensure that the system-wide frequency returns to
normal as the system operating points change due to
some perturbations. These perturbations can be classified as small perturbations, e.g., a change in load over
time, or large, e.g., the loss of generation and so forth.
For secondary frequency control scheme, under any
decision-making approach, the goal for the scheme is to
compute the so-called average frequency error (AFE).
Using the AFE, the control scheme implements a simple
proportional-integral (PI) controller to compute new
operating points for the controllable assets in the
microgrid. Under the new operating points, the controllable assets bring the system-wide frequency to nominal
by driving the AFE to zero.

Test Setup
We consider a six-bus islanded ac microgrid with three
DERs and three loads as depicted in Figures 7 and 8. The
setup for testing the centralized control scheme is similar to the one described in the centralized coordination
and control architecture. It comprises the cRIO 9068
device that carries out secondary frequency control in

IEEE Electrification - September 2020

Table of Contents for the Digital Edition of IEEE Electrification - September 2020