IEEE Power & Energy Magazine - January/February 2020 - 72

Even though remote-control capabilities have been enhanced
through smart grid efforts, industry practices remain mostly
manual and use mostly backup feeders for restoration.
Long-Term ADMS Applications
and Enablers
Added system complexity and the need for resilient opera-
tions result in new operational challenges for future distribu-
tion systems. To address these challenges, an interdisciplin-
ary approach is needed. Concepts and methods from data,
computer, and network sciences are expected to play a piv-
otal role in efficiently operating future distribution systems.
In the next three sections, we describe three key enablers of
future ADMS applications: adequate control architecture,
intelligence-based decision-making concepts, and data qual-
ity and consistency.

Adequate Control Architecture
It is often desirable to have a control architecture where the
operation of the entire distribution system is managed cen-
trally, giving the operator a complete view of the network.
The sheer size of typical distribution systems, however,
makes this a very challenging task due to the computa-
tional burden and the requirement of adequate information
and communication technology infrastructure to connect
the ADMS with customers and network assets. Depending
on the ADMS application, the level of complexity in the
distribution system modeling can be drastically different.
From the transmission system operator's perspective, the
primary interest is the behavior at the transmission/distri-
bution interface (e.g., requiring a specific power factor),
and it may be sufficient to obtain only aggregated informa-
tion at the corresponding substations. When dealing with
MV-connected DERs, such as actively managing a PV
farm, considering each individual LV-connected customer
may not be necessary. In such cases, it may be sufficient
to consider only the MV network, which is significantly
simpler than an integrated MV and LV model.
The hierarchical structure (based on different voltage levels)
of power systems can be exploited to address the scalability
challenges of a fully centralized architecture. Instead of con-
sidering the entire distribution system in a single model, in the
near future, ADMS applications are likely to adopt a hierar-
chical architecture. The whole distribution system can be bro-
ken down into multiple subsystems that decouple the original
problem (e.g., separating the problem at the interfaces between
different voltage levels and/or creating multiple zones within
the same voltage level). As this control architecture involves
certain assumptions at the interfaces of these newly defined
subsystems, it will require a process to check the validity of the
assumptions and the optimal set points across all subsystems.
72

ieee power & energy magazine

Although a hierarchical architecture offers scalability
improvements over a fully centralized one, they still share simi-
lar characteristics because the operation of each subsystem is
still managed centrally. As a result, apart from the computa-
tional requirement, both architectures are affected by a single
point of failure, particularly if this happens in the control room
(within the ADMS). With the increasing awareness of data pri-
vacy, customers may not wish to share their data (which are cru-
cial in a centralized architecture) with another entity; depend-
ing on the situation, other control architectures that do not rely
on a single entity to determine the set points are also desirable.
In the long run, ADMS applications may benefit from
adopting distributed or decentralized architectures by making
it possible for DERs and network assets (commonly referred
to as agents) to determine their own set points based on the
information they have locally. In essence, these architectures
leverage the processing power of many agents to determine
the set points. At the same time, because each agent has only
access to limited information about the entire network, an iter-
ative process is typically required to determine the optimal set
points; some impact on overall performance can be expected.
The exact definition of these control architectures can
vary, as shown in the available literature. A typical differen-
tiating factor is whether the information is exchanged among
participants. In a distributed architecture, local information
can be shared with neighboring participants; in a decentral-
ized architecture, each party will act solely based on the local
information. Other variations also exist, such as indirectly
influencing the behavior of participants through time-vary-
ing signals (e.g., using pricing signals to incentivize demand
reduction of households).
Figure 6 illustrates the differences among each of the con-
trol architectures: fully centralized, hierarchical, distributed,
and fully decentralized. Having access to all information from
key substations (e.g., primary substations) down to individual
customers may equip the network operator with sufficient
information to determine the best set points. Yet, there is an
inherent tradeoff between scalability, performance, and many
other aspects (e.g., customers' privacy). These different archi-
tectures will play a significant role in addressing the various
needs of future distribution systems.

Intelligence-Based
Decision-Making Concepts
The current practices and technologies used for managing
the grid's resources are largely driven by models that may not
be able to meet the control requirements of large, complex,
january/february 2020
IEEE Power & Energy Magazine - January/February 2020

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - January/February 2020
