IEEE Electrification - March 2021 - 39

xx
Small-scale: An individual contribution of residential-

scale loads is determined by typically low-wattage
consumer appliances, with EVs being a notable
exception. As a result, effective DR programs require a
large number of enrolled DR participants. This, however, implies that any profits from deploying DR
resources must also be shared among a large number
of participants. This profit, partially distributed in the
form of monetary incentives, might not be sufficient
to engage and retain DR participants. However, some
nonfinancial incentives, e.g., an awareness of environmental impacts and the potential to mitigate climate change, may convince more customers to enroll
in DR programs.
xx
Distributed connectivity: Distributed DR participants have
various means of connectivity (e.g., cellular network,
Wi-Fi, and Bluetooth) and levels of cyberhygiene. This
diversity in cyberawareness incurs dealing with both the
explored and novel (zero-day) vulnerabilities. Furthermore, due to the distributed nature of DR customers, utilities may need to uneconomically expand their
infrastructure and scope to capture a small DR flexibility.
xx
Stochasticity: Residential DR is subject to systemic and
behavioral uncertainty. First, residential DR participants are not obliged to follow DR call signals. Even if
the operator is able to directly control appliances, customers are always able to manually interfere and
override the DR control signals. Additionally, even if
some preferences have been communicated or even
somehow committed, some aspects of real-time preferences might be unknown to DR participants themselves due to limited rationality and changing
environmental circumstances (e.g., weather).
xx
Heterogeneity: Residential DR participants are heterogeneous in nature, with unique load profiles and different preferences and behaviors that complicate the
implementation and standardization of residential
DR programs.

account for uncertainty caused by the random, intentional or unintentional, interference of DR customers.
The available DR capacity depends on the physical
characteristics of DR resources and the individual preferences of their users. For example, the DR participation of
an EV requires the DR operator to know or estimate its
state of charge, the desirable time of readiness, and battery-specific characteristics (e.g., degradation curves and
charging history). Similarly, the thermal inertia of cooling
and heating systems can be exploited to temporally shift
power-consumption patterns. However, estimating the
kW-capacity that can be extracted from this thermal inertia requires information about the technical characteristics of these systems and the temperature preferences of
their users. While some system settings can be obtained
using a two-way communication infrastructure and digitized appliance interfaces, individual preferences or comfort zones are rarely observable. Furthermore, acquiring,
processing, and storing behavioral data involves effort for
both the DR operator and participants, which may outweigh the benefits of the DR program.
DR participants expect some remuneration in return for
their participation in DR programs as a compensation for
lost utility due to the need to depart from their regular consumption patterns during DR events and also as an incentive to purchase new controllable appliances or necessary
periphery. Therefore, DR operators aim to establish a remuneration scheme that incentivizes sufficient DR enrollment
and systematic participation. At the same time, the overall
cost of the total remuneration must be kept at a reasonable
level to ensure a long-term profitability of the DR program.
The reliability of real-time DR deployments depends on
the accuracy of DR capacity estimates and the sufficiency
of the offered incentives. Scheduled load reductions may
be insufficient if the called appliances are not operated as
estimated, they fail to communicate with the DRAS, or DR
participants suddenly opt out from the DR event. Such
uncertain behaviors are difficult to predict ahead of time,
which can reduce the effectiveness of DR programs.

Learning Optimal DR Decisions
The goal of the DR program is to produce DR control or
incentive signals to achieve a desirable change in the
observed system demand. The ability to meet this goal
depends on three attributes that characterize each DR
resource but are not exactly known to the DR operator:
xx
Available capacity: Commit and dispatch DR resources
with respect to their spatio-temporal restrictions and
for particular applications, e.g., peak-load shaving,
mitigation of intermittent injections from wind and
solar, or other ancillary services.
xx
Cost: Determine the short- and long-term costs of dispatching and enrolling DR resources, respectively, and
weigh them against other dispatchable resources
available to the system.
xx
Reliability: Evaluate the projected real-time effectiveness of the scheduled DR dispatch decisions and

Distributed
Connectivity

Small Scale

Residential
DR Challenges

Heterogeneity

Stochasticity

Figure 2. The challenges of residential DR programs.

IEEE Electrific ation Magazine / MARCH 2 0 2 1

IEEE Electrification - March 2021

Table of Contents for the Digital Edition of IEEE Electrification - March 2021