IEEE Electrification - September 2021 - 23

In the 1980s, utilities started experimenting with the
potential of recruiting small residential loads for DR. The
controllable loads were the HVAC units, electric water
heaters, and large white appliances such as dryers, washers,
and dishwashers. However, those attempts usually
failed to sustain after the initial honeymoon period ended.
A few weeks into the program, many consumers chose to
leave the control settings fixed or override the control.
Looking back, there were several key hurdles for customers
to embrace the concept of LC:
Timeline
TABLE 3. The evolution of load control methods.
Past
Now
Available resources
Decision makers
A limited amount of large
industrial or commercial
loads
Utility
Resource selectivity Nonselective
Customer
participation
Mandatory
Communication need One-way
Communication
protocols/standards
Proprietary communication
protocols, Energy Interoperation
1.0 for demand
response
Most residential, commercial, and
industrial loads are now eligible
for LC
Utilities, load serving entities, load
service aggregators, customers, preprogrammed
software agents
Resources selected well ahead of the
time
A mix of mandatory and voluntary
participation with an opt-out option
No communication, one-way, two-way
OpenADR 2.0 standard, IEEE Standard
1547 for interconnection and
interoperability of distributed energy
resources. Other related standards:
IEEE 1815.1-2015, IEC 61850, IEC
61968, IEC 61970, IEC TC 57, IEEE
2030.5
Control hierarchy
Control signals
Control mechanism
Centralized
Direct control through
circuit breakers or control
feeder voltage
Set up predetermined,
fixed voltage or frequency
thresholds on protection
relays
Control algorithm
mainly rule-based
Centralized, decentralized, autonomous
Direct
or indirect
Turn ON/OFF through relay, set
thermostat set point and bandwidth,
select the mode of operation
Rule- or optimization-based
Flexibly switch to different control
hierarchies or form a new control
hierarchy
Compatible with all control signals
Customizable thresholds; allow
baseline operation and follow
load-following up and down signals
Reinforce learning for adaptability.
Meta-learning for learning from past
experiences. Transfer learning to
learn from peers.
Reward mechanisms Fixed rate
Time-vary rates, coupons, rebates,
credits, discounts, and so on. Rewards
are usually not known until a much
later time.
Rewards are calculated based on
performance. Rewards to resources
are instantaneous.
x low monetary return
x degradation in convenience and comfort
x semimanual, inflexible control options
x losing control to utilities and invasion of customer privacy
x
reward mechanism not reflecting quality of performance
and degree of contribution.
After all, adjusting electricity consumptions when
there is a shortage of generation or transmission capacity
is viewed as a public service that brings societal
Future Trends
The number of available LC
resources will greatly exceed the
LC needs most of the time
Artificial-intelligence-powered
adaptive agents
Near real-time resource selection
to minimize uncertainty, cost, and
impact to customer comfort
Bid-for-services (only the best
customer resources are selected)
Near real-time communication
(<10 ms)
IEEE Electrification Magazine / SEPTEMBER 2021
23
IEEE Electrification - September 2021

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