Project Introduction and Motivation

Electricity system challenges have
emerged in Washington’s Olympic
Peninsula that make it an ideal
laboratory for advanced intelligent utility
network (IUN) experiments. The peninsula’s
transmission feed is approaching
capacity as a result of population growth,
but there are significant concerns over
siting additional transmission or central
generation assets there due to the ecologically
sensitive nature of the Olympic
National Forest, which comprises much
of the northern portion of the peninsula.
In 2001, in response to that situation,
Bonneville Power Administration (BPA)
began their Non-Wires Solutions program
of demand response (DR) and distributed
generation (DG) experiments and pilots
with the goal of minimizing the environmental
impact of new transmission and
generation construction.

The Pacific Northwest GridWise™ Testbed
program, part of the Department of
Energy’s GridWise™ initiative, was established
in 2005 to coordinate technology
projects supporting the GridWise™ vision
of a transactive electric infrastructure
leveraging advances in information technology
and communications. The goals
of GridWise™ and the challenges faced
by BPA led to the creation of the Olympic
Peninsula Project as part of the Testbed
program. This project combines DR and
DG control, using a real-time retail market
clearing mechanism to determine the
optimal balance of DR and DG dispatch
in response to the current state of the
system.

The yearlong Olympic Peninsula Project,
which concluded in March 2007, was
a partnership involving a number of organizations:
Pacific Northwest National Laboratory
(PNNL), IBM Research, Bonneville
Power Administration, Invensys Controls
and several public and private utilities
in the region. PNNL, working with BPA,
defined the project and acted as the overall
project manager. IBM Research was
responsible for the system integration of
the various components with the market
clearing system developed by PNNL.
Invensys Controls provided the residential
equipment, including communicating
meters, programmable thermostats, load
control modules for water heaters and
broadband gateways for communication.

Design of the Experiment – Participants and Structure

The Olympic Peninsula Project involves
about 120 residential customers and several
commercial/industrial customers.
The residential customers are divided into
several contract types and a small control
group. The contract types are fixed price,
time of use/critical peak price (TOU/CPP)
and real-time price (RTP). One of the goals
of the project is to compare the performance
of the three contract types, with
particular emphasis on the RTP versus
TOU/CPP comparison. The hypothesis
being tested is that a properly designed
and implemented RTP system will be more
effective at managing Distributed Energy
Resources (DER) such as DR and DG than
a TOU/CPP system.

Rather than work through the process
of establishing experimental tariffs for
the residential participants, PNNL implemented
a shadow billing system with
an online payment portal. Participants
receive their normal utility bill based
on actual consumption, and pay it as
they ordinarily would. They receive an
additional online bill for the experiment
that represents their consumption from
a virtual distribution feeder. That bill is
paid via the online portal using funds
seeded into an account at the start of
each quarter, so customers experience no
additional out-of-pocket expense (part of
the project budget was reserved for this
purpose). The amount of funds that are
seeded is based on historical consumption
for each customer.

Fi gure 1 Summary of Olympic Peninsula Project ComponentsTo give the experiment some level
of real-world motivation and benefit in
exchange for being responsive to critical
situations or high prices, PNNL refunds
any balance remaining at the end of each
quarter back to the residential customer.
Therefore, the more aggressively a customer
responds, the larger his refund
check will be. The experiment has been
designed to allow the most aggressive
participants to receive on the order of
$150 – roughly what it might cost to take
a family out to dinner, for example.

The virtual distribution feeder is implemented
with feeder modeling software at
PNNL using the real consumption information
being collected. This allows all the
participants to appear as if they’re on
the same feeder, and that feeder can be
artificially constrained to create critical
situations as part of the experiment.

For the residential customers, the
DR assets are the heating and cooling
systems and the water heaters. The
physical control components in the
homes have been provided by Invensys
Controls, and, in the case of the TOU/CPP
contract homes, the Invensys GoodWatts
pilot system is being used to implement
the control management. Homeowners
can set up occupancy profiles and timeof-
day schedules based on the predetermined
time-of-use tariffs. Critical peak
price events are scheduled as part of
the experiments based on expected constraint
times.

Fixed-price contract homes do not
have any DR automation, but homeowners
do have access to the same Web
portal as the TOU/CPP and RTP contract
homeowners, showing up-to-date consumption
information. Their equipment
also has the ability to help them manage
reduction of electricity consumption
through mechanisms such as night-time
thermostat set-backs, and part of what’s
being observed is how well those homeowners
reduced their overall load.

The DR assets of the RTP contract
homes, as well as the DG units, are
managed by their participation in an
artificial real-time market that clears
every five minutes. The market incorporates
information from multiple sources,
including the base transmission capacity
and price (provided by a live Dow Jones
feed of the wholesale price), virtual
distribution feeder capacity and actual
distribution cost, the current non-curtailable
load, the current curtailable load
and corresponding buy bids, and the
current dispatchable DG capacity and
corresponding sell bids submitted to the
market on each cycle. From this information,
the market determines a new clearing
price for retail electricity every five
minutes, and that clearing price is communicated
to the DR and DG assets. Each
of those assets then responds based on
its most recent bid and the parameters
defined by its owner.

Implementing the System Using Event-Based Middleware

The most innovative functionality in the
Olympic Peninsula Project is the market-
based control of DR assets in the
RTP contract homes. The experiment
requires each programmable thermostat
to understand how to create a buy
bid into the real-time market based on
homeowner comfort goals, the current
state of the device (e.g., the temperature
in the home) and the current trends of
the market. Note that the water heaters
are not bidding into the market; they are
using an open-loop control scheme in
which they only respond to price signals
from the market. Both the thermostats
and water heaters then have to respond
appropriately to the clearing price generated
every five minutes before the cycle
starts again. (The DG assets are also bidding
and responding but using a different
software and communication approach,
so they are not discussed here.)

The design and implementation of this
part of the system is based on a prototype
event-based programming framework
called Internet-scale Control Systems
(iCS) developed at IBM’s T.J. Watson
Research Center. iCS is an example of
what is currently emerging in the academic
and research community under
the category of cyber-physical systems,
reflecting the increasing importance of
monitoring and managing the physical
world in a much richer and more detailed
fashion. The physical world, in this case,
is the electric grid and its associated enduse
loads (such as heating and air-conditioning
systems).

In practical applications, there is an
additional aspect that needs addressing.
Cyber-physical systems will almost always
operate within the context of some business
environment. In the Olympic Peninsula
Project, the business environment
is reflected in the use of a market-based
control scheme and the need for devices
to be virtually augmented with business
domain awareness – they need to bid into
the market and react to clearing events.
For this reason, IBM Research is focusing on an expanded scope of investigation
referred to as cyber-physical business
(CPB) systems, which includes both
discrete, event-based systems and continuous
data stream-based systems. The
objective is to define a middleware framework
that addresses solutions, such as
intelligent utility networks, which involve
the integration of the physical operations
domain with the business domain, deals
with the challenges of a highly distributed
and heterogeneous environment, and also
reflects the time-dependent nature of
such integrated solutions.

By using a loosely coupled event-programming
approach with a very simple
programming model, iCS enabled all
the components of the system to be
abstracted and represented as simple
sensor, actuator or control objects in
the market-control application – from
the market itself all the way down to the
heating element relays and load-monitoring
sensors on the water heaters. The
market-awareness and bidding/response
algorithms were added in the middleware
framework without modification of the
devices themselves, essentially creating
new, more intelligent virtual devices from
the perspective of the market. This also
allowed the design to be easily adapted
in response to requirement adjustments
or other issues that surfaced during
development. In addition, iCS enabled the
overall market-control application to be
structured as a layered set of hierarchical
control loops linked together through the
event communication model.

Another benefit of using a loosely coupled
event-based approach is the level of
resiliency it affords. The system was implemented
so devices fall into a safe degraded
mode if there is any loss of communication
or failure in the market clearing signal.
Operations continue in a less-than-optimal
mode, but there is no catastrophic failure.
Once the problem is resolved, the system
returns to its optimal state.

Early Results

In spring 2007, the Olympic Peninsula
Project is just completing the region’s
winter electricity-constraint season.
Several cold periods occurred, and the
real-time market system operated as
designed. It limited total load on the virtual
distribution feeder to the configured
capacity through price-based DR management,
and it dispatched DG units when
the market price for electricity met the
sell bids. When no constraints existed,
the market price was stable and in the
expected range. In terms of performance,
it appears there was more optimal shifting
of load with the RTP contracts than with
the TOU/CPP contracts.

Another intermediate observation
was that the fixed price contract homes
sometimes shifted more load than the
TOU/CPP homes. That may have been an
artifact of the small sample size and the
individual homeowners involved, but it
is still an important result in that it indicates
customers will take advantage of
well-designed energy management tools,
and may have further been motivated
by having real-time information sources
about their consumption and cost easily
accessible.

An additional positive indicator surfaced
as a result of extending the project
from its original end date of September
2006 to March 2007: Most of the residential
participants agreed to continue their
involvement, presumably because they
were satisfied with the system’s positive
impact on their electricity consumption
and, to some extent, on their overall electricity
cost.

Conclusion

The Olympic Peninsula Project has
already succeeded in a number of dimensions,
from demonstrating the effectiveness
of a cyber-physical business system
approach in designing such solutions, to
generating a great deal of interest in and
awareness of the optimal management of
combined demand response and distributed
generation in dealing with a variety
of issues in electric grid operations. The
goal of using this project as both an
experiment and an educational tool is
already starting to be realized.

The project also represents an important
milestone in the realization of the
GridWise™ vision of an intelligent utility
network – by integrating utility and customer
assets; by leveraging advanced
information technologies; and by bridging
the operations and business domains of
the utility environment. Further, it has
been an extremely successful collaboration
of industry and public organizations,
including equipment and information
technology vendors, investor-owned
utilities, public utility districts, and the
Department of Energy’s National Laboratory
system.