IEEE Power & Energy Magazine - November/December 2020 - 42

take place between a cell-level control platform and individual customers; these are necessary to optimize customerlevel objectives while respecting electrical limits within a
cell. Message passing among cells to optimize the flow of
power is based on economic and reliability targets. These
levels of hierarchy allow for scalable distributed optimization algorithms to be designed and implemented in AEGs.
Figure 5 presents three levels of hierarchy. The top level,
level 3, aggregates neighborhoods to achieve an optimization objective, such as voltage regulation or power balancing.
This level communicates to level 2 (e.g., a single neighborhood) about the aggregated power designated for that neighborhood. This information is passed from the neighborhood
level down to the homeowner, level 1, as a power set point
to track. The homeowners might accordingly coordinate
their own distributed wind or solar, smart-home devices,
and EVs to optimally balance the grid needs and their own
usage preferences. Communications run in both directions,
as indicated in Figure 5. For example, if homeowners are
unable to meet their power set points, information is passed
back up to level 2 (e.g., via monitoring the aggregate power
of the neighborhood) to indicate this, and the optimization
is repeated until each agent in the cell has reached a feasible
solution that achieves the global objective as well as individual satisfaction.

After demonstrating that distributed control concepts
can work on a single distribution circuit, the goal is to
implement a hierarchical control scheme that would allow
true scalability. This work considered a potentially large
distribution network controlled cooperatively by several
networked AEGs. Figure 6 is an illustration of this work.
A regional coordinator communicates with all the dispatchable nodes within each AEG cell, and a central coordinator communicates with all the regional coordinators.
Each regional coordinator knows only the topology and
line parameters of the cell that it controls, and the central
coordinator knows only the topology and line parameters
of the reduced network, which treats each cell as a node
and connects all the cells. Given such information availability, we explored the topological structure of the linearized power flow model to derive a hierarchical, distributed
implementation of the primal-dual gradient optimization
algorithm that solves an OPF problem. The OPF problem
minimizes the total cost of all the controllable DERs and a
cost associated with the total network load subject to voltage regulation constraints. The proposed implementation
significantly reduced the computational burden compared
with the centrally coordinated implementation of the primal-dual algorithm, which requires a central coordinator
for the whole network.

Additional
Levels
cc
Level 3
(Multiple
Neighborhoods)

cc
cc

P, Q Setting
P, Q Response

Level 2
(Neighborhood)
P, Q Setting
P, Q Response
Level 1
(Home)
(cc) = Cell Controller
P = Real Power
Q = Reactive Power

figure 5. The communications architecture for distributed and real-time optimization of AEGs. In the figure, Level 1
would be at a home or business, Level 2 would be at a neighborhood, and Level 3 would be multiple neighborhoods,
all on a single distribution circuit.
42

ieee power & energy magazine

november/december 2020

IEEE Power & Energy Magazine - November/December 2020

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2020