IEEE Power & Energy Magazine - Grid Edge 2023 - 44

Distributed Intelligence
and Multiagent Systems
The study of market-based approaches as a distributed intelligence
mechanism for solving multiobjective optimization
problems have their roots in economic theory. With
the advent of information and communications technology
and the growth of robotics and intelligent systems,
their application to solve complex systems of systems
problems is expanding. The increasing pervasiveness of
computational resources in devices enables local intelligence,
and their communications connectivity allows
them to interact. This interaction of intelligent devices is
foundational to the discipline of multiagent systems and
the methods and tools that support them.
For the purposes of this article, devices or systems are
intelligent if they are able to perceive their surroundings
to gather information about the context of their operation.
Like people, they can reason autonomously by processing
information in a goal-directed way to forecast, plan,
and act. Distributed multiagent systems are characterized
by a population of intelligent devices that communicate
with each other to exchange information (measured or
derived) to better accomplish their independent objectives
through cooperation and joint action. Multiagent
systems are constrained by their environment, which imposes
physical, temporal, and policy (economic or other)
conditions on their interaction. In this regard, intelligent
agents reside in an ecosystem of products, services, deployment
platforms, and other supporting infrastructure.
Like societies of people, they interact with more or less
independent decision-making authority within cultural
and governing policy structures. As their intelligence is
captured in cyber-based programs, multiagent systems
can reside in the real world or run in a simulated cyberworld.
This is particularly important for complex systems
design and testing, and simulation.
certainty about the momentary system reaction, realize an
efficient market with proper incentives, and protect the privacy
of the end user whose devices participate in the energy
management task.
TE Systems Implementations
in terms of customer privacy, scalability, and efficiency, te
systems have clear advantages over more common smart
grid coordination, such as price reactive systems and centralized
optimization. Both in the united States and europe, te
research has had a strong focus on intelligent agent-based innovation
in household equipment and field demonstrations involving
grid operators, energy supply companies, power technology
companies, and regulators.
TE Systems Implementations
in the United States
the u.S. department of energy partnered with several organizations
in three major demonstration projects using te
mechanisms to coordinate distributed energy resources with
system operations.
Olympic Peninsula Demonstration, 2006-2007
this first proof-of-concept te project was located in an
area of the olympic Peninsula of Washington state, which
receives electricity through a radial transmission connection
to the Pacific northwest power grid. the project tested
the potential for flexibility offered in coordinating distributed
energy resources to postpone or remove the need for
a transmission upgrade. the project used a 5-min doubleauction
market technique to coordinate four large municipal
water pumps, two backup diesel generators, and residential
demand response from electric water and space heating systems
in 112 homes. the project established the viability of
te to achieve multiple objectives: system peak load and distribution
constraint management; wholesale price purchases
by the utility; and residential, commercial, and municipal
energy cost savings.
the market received supply bids from the utility based
of devices participating. further, the approach protects the
end user's privacy as the bids communicate only information
about energy quantities and prices. When these bids are
aggregated on the level of a house, a building, or an industrial
site before being communicated externally, the information
exchanged is comparable to that of a metering system
collecting near-real-time data as described for the price
reaction approach above. and unlike the centralized optimization
approach, no complicated models of the devices,
consumer behavior, or preferences are exchanged or maintained.
in summary, te approaches are able to access the
full response potential of flexible devices, provide greater
44
ieee power & energy magazine
upon a markup of the wholesale price of energy in the area.
the diesel generators' bid was based on the actual fixed and
variable costs incurred for operation. the pumps' bid into
the market was based on water-reservoir levels that they were
designed to regulate. and the residential demand-response
equipment allowed the households to specify their automatic
price-response preferences. to capture their preferences, a
list of comfort settings named to indicate ranges between
most comfortable (nonprice responsive) to greatest economy
(highly price responsive). the 5-min market determined the
clearing price for energy and broadcast that to the market
participants. each participant's bidding equipment would
operate based on whether their bid was higher or lower than
the market-clearing price.
figure 2 is the operational dashboard for the demonstration.
Besides coordinating the price-responsive resources
may/june 2016
IEEE Power & Energy Magazine - Grid Edge 2023

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - Grid Edge 2023
