Sei sulla pagina 1di 5

1

GridAgentsTM: Intelligent Agent Applications for


Integration of Distributed
Energy Resources within Distribution Systems
David A. Cohen, CEO Infotility, dave@infotility.com

AbstractInfotilitys GridAgentsTM software is being used for


large-scale integration of distributed energy and renewable
energy resources into distribution systems with specific
applications on SmartGrid based energy networks including
Microgrid management, intelligent load control and smart
charging applications, including photovoltaics & storage, load
control, and future assets such as plug-in hybrid cars. The
GridAgents EAM Suite is an advanced software foundation
(based on multi-agent system technology) designed to work in
concert with Service-oriented Architectures such as Web
Services. The architecture enables the deployment of
communities and eco-systems of intelligent grid resources which
can adapt to changing run-time conditions collaborate to solve
complex problems, and flexibly accommodate new behavior
through the use of plug-in architectures. This paper will discuss
the use of multi-agent systems used in the GridAgents software
and will discuss some example applications and deployments
currently being used with distribution utilities as part of
distribution automation applications.
Index TermsMulti-Agent Systems, Distribution Automation,
SmartGrid. MicroGrid, Renewables. Distributed Energy
Resources.

I. NOMENCLATURE
DER - Distributed Energy Resources
MAS - Multi-Agent System
DA - Distribution Automation
DNC Distribution Network Control

II. INTRODUCTION

his paper briefly discusses some of the advanced


demonstrations of the Infotility GridAgents software platform
being deployed within Con Edison of New Yorks distribution
system. The GridAgents software and applications are being
installed within Con Edisons 3G: System of the Future
(3G/SOF) development efforts. This Smart Grid
development convergence will continue to be guided by
ongoing industry consortium efforts such as GridWise,
GridWorks, and Intelligrid with the goal of achieving both
This work was partially7supported in part by the U.S. Department of Energy
under Grant BS123.

2008 IEEE.

ConEdisons and the DOE Office of Electricitys long term


electric delivery goals.
III. CON EDISON NETWORK
Con Edison services New York Citys five boroughs and
Westchester County a territory with a very high load density
that covers more than six hundred square miles and 8.7M
people. Con Edison maintains one of the most sophisticated
secondary network systems in the world, with one of the
highest levels of service reliability and resulting high
infrastructure and maintenance costs in the U.S. The Con
Edison electric grid includes approximately 27,000 km of
primary distribution cables, supplying 24,000 network
transformers in 63 secondary networks. By some estimates
Con Edison has half the electrical distribution networks in the
world.
3G: System of the Future
The Con Edison 3G: System of the Future is an ongoing
project to design the Third Generation electric system that
will be the basis for power delivery into the future, using
advanced technologies to meet new operating and customer
electric demands, maintain reliability, optimize costs, and
ensure security. The 3G System of the Future is implementing
a modernization of its electric network design to optimize its
configurations of substations, and Primary and secondary
distributions systems.
The general design principles
supported by the 3G/SOF project include:

Inclusion of distributed resources


Consider Substation load transfer
Application of HVDC
Higher distribution voltage (27kV, 33kV, 33+kV)
Minimal secondary grid
Advanced intelligence and distributed software
Company-deployed mobile DER assets
TABLE I

IV. GRIDAGENTSTM: DER AND


DISTRIBUTION NETWORK CONTROL
FRAMEWORK
The Infotility GridAgentsTM Framework is currently under
development by Infotility as a platform for building advanced,
large-scale Distributed Energy Resource (DER) and
Distribution Network Control (DNC) solutions. GridAgents is
a novel, advanced software foundation (based agent and
multi-agent system technology) for developing optimized
DER control in varying operational configurations such as
Microgrids, demand response automation, DNC, and
intelligent monitoring applications. It is designed to integrate
with SCADA systems and Internet-enabled applications.
The GridAgentsTM Framework is well suited to the evolving
advanced distribution automation (such as the 3G/SOF
applications) and distributed problem solving applications.
Some defining architectural characteristics of the GridAgents
Framework include:
Adaptive: Adaptive to changing and unforeseen run-time
conditions for distribution control objectives.
Robustness: Designed to survive the loss of individual
components with minimal loss of functionality.

power production or for demand limiting based on the current


ambient conditions, customer preferences, etc. Other agent
modules which utilize a multi-agent system approach such as
the Planning and Optimization agents include broker agents
that manage aggregate DER devices under various operational
scenarios such as optimal Microgrid control strategies. Other
Plug-ins includes a Least-Cost Linear Optimization planning
routine and a Genetic Algorithm planning routine under
development. A self-organizing Coordination Plug-in
algorithm is under development which allows bottom-up
aggregation of resources based on internal planning and
reasoning. New agent plug-ins can be developed to build
customized applications. Integration with internet-based
systems can be achieved through the Web Services gateway
via the Blackboard agents. The blackboard agent can also
provide redundant volatile storage and provide a pathway to
persistent storage and information channels through various
publish/subscribe applications (InfoNow is a Web-based
middleware built by Infotility). GridAgents is currently being
deployed on mobile devices, gateways, meters, as well as PCs
and Servers. By design, the Framework supports any device
and operating system as long as the memory and processor
capacity meet the design specifications. As well, the
Framework does not depend on any particular GUI to stay
running; however, there is currently a web client, windows
client, and PDA client GUIs which are supported.

Flexible: Expandability through plug-ins (algorithms,


planning routines, device models, and control logic); Resolve
multiple operational goals with business rule and physical
constraints by encoding domain specific information
(expression evaluator).
Scalable: Designed with an open system architecture and
well documented component-based interface
Efficient: Supports device portability, multiple user
interface support, low deployment and maintenance cost
through the use of automatic and remote configuration.
The diagram (below) illustrates the major elements of the
GridAgentsTM Framework. Applications and plug-in agent
modules that have been developed include device Sense and
Control and Resource agents (Photovoltaic, generator,
Microturbine, CHP, load control strategies, and sensing)
which have embedded analytic methods to calculate optimal
response based on pricing signals subject to constraints.
These agent operational models can utilize business rules,
physical constraints, and statistical and AI-based learning
routines to optimize operation and/or response over time. The
Resource Controller agent has two flavors: the Load Control
class is used to interact with non-generation demand response
systems. This includes technology such as HVAC setpoint reset, zone thermostat reset, lighting dimming, motor and pump
staging, etc. The Generator base class describes the behavior
of on-site power generation, including fuel consumption and
avail-able heat recovery. One of the primary roles of the
Resource development of model coefficients that can be
used by higher-level agents to determine the potential for

3G/SOF: Advanced Distribution Operation with DER


Integration
The project objectives are to perform research, development,
and pilot-scale testing of advanced, next-generation
distribution operational strategies using ConEdisons 3G:
Distribution System of the Future and associated
infrastructure as the real-world Test Bed (demonstration
network) combined with the Infotility GridAgents software
Platform as the embedded intelligence software platform for
advanced operational strategies development. The project is
being deployed within the Fulton Corridor Network and the
Madison Square Network over a 3 year period.

V. INTEGRATION OF DERS WITHIN


DISTRIBUTION SYSTEMS: ADVANCED
DISTRIBUTION OPERATIONS STRATEGIES
A brief summary of some of the Advanced Distribution
Operation (ADO) applications being applied to the 3G SOF
are being developed with the GridAgents software including:

Dispatch of customer load controllers for compact


network load alleviation
Physical assurance (current limiting) meter to
dispatch customer generator
Identification of non-functioning limiters on area
network
Large-Scale Demand Response Integration
Smart charging of PHEVs and PV Systems
Thermal aging calculations for feeders /
transformers
DER assets (historical profile and real-time status
/ availability)
Exercising network protectors to gauge
functionality
DER preservation during voltage reduction events
VAR control via transformer reconfiguration
VAR dispatch using DER via synchronous
generator
Software demonstration of "intentional islanding"
on 460 V network

Figure 1 Example components and agent configuration


The basic function of this system of agents is as follows:

A single example will be discussed in this paper for


example.
EXAMPLE: Dispatch of customer load controllers for
compact network load alleviation
Description / scenarios / use cases
In this application a system of agents are used to identify
loads on the distribution transformers and to attempt demand
response by activating some form of load controller. The key
components for such a system are illustrated in Figure 1
below.

Agents A1 through A4 are monitoring the current,


load, temperature or combination of values from
each of the transformers that supply power to the
network. Each of these agents also contains a set of
rules related to the maximum allowable value for a
particular parameter. That is, each agent is able to
determine if the transformer is running too hot or is
approaching its operating limits.
The status of each transformer is regularly reported
to the management agent A5. The status values are
reports about the state of each transformer, for
example, off-line, normal, requesting load reduction,
etc.
The load controller agents A8, A9, A10, and A11 are
able to report the status of and provide short-term
load reduction seen at the associated customer meter.
The generator agent A7 is able to report the status of
and dispatch the generator associated with that
specific customer.
The management agent A5 acts as the broker between
the transformer agents and the load controller and
generator agents. When the transformers request
load reduction in the network, agent A5 will negotiate
with the demand response agents to determine the
actions that result in least-cost operation of the
compact network.

This application has the potential to reduce outages and to


provide demand response. The outage probability reduction is
accomplished through the imposition of a price on such an
event and allowing the management agent to force action that
would limit load to prevent transformer failure.
High loads at transformer banks
The primary scenario for this application is one where the
transformers are each approaching their temperature or

current limits, or where an anticipated load extreme or cycling


of load will cause an unacceptable degree of damage or aging
to a transformer.
Human Interactions
The human actors in this application are

System configuration personnel enter the initial


information about the capacity and operating limits
of the equipment.
Load controller rule writers act on behalf of the
customers associated with each meter. Certain
customers may choose to respond to or ignore
requests for load reduction.
Generator rule writers set the interconnection
requirements for the dispatch of the generator(s) on
the compact network. These rules may also
correspond to customer preferences with regard to
times when the generator(s) may be turned on or off.

Machine Interactions

Transformer agents determine the state of the


transformer based on the measured values and the
known operation limits of the devices.
The management agent uses the published
transformer states to determine if the load controllers
should take action. The decision of the management
agent is made using estimated costs of specific
circumstances and actions. For example, the
management agent may attempt to minimize a cost
function such as
J = C1 x P1 + C2 x D2 + C3 x D3 +

where C1 is the cost of transformer replacement, P1 is


the probability of a transformer failure as determined
from the equipment models within the transformer
agent, C2 is the cost (or perhaps customer rebate) per
depth of achieved load reduction D2, and so forth.
Load controller agents respond to requests from the
management agent

Assumptions
Communication network
Figure 1 shows the agents hosted on the network
protectors based on the understanding that the
network protectors may have sufficient on-board
processing capability. The network protectors are
assumed to be close enough to the transformers to
allow for direct wired or short-range RF sensor
monitoring.
It is assumed that transformer agents have access to
temperature, current, and/or power measurements
taken at the transformer.

Bandwidth
It is not expected that the transformer values and load
controller signals would need to be sent with urgent
frequency. It is likely that messages could be sent between
the agents on a time scale measured in minutes or perhaps
even tens of minutes. The largest use of bandwidth is likely
the management agent requesting the costs of actions from the
load controller and generator agents.
Depending on the number of allowable permutations of each
device, the matrices of actions and associated costs could be
sizable and numerous. For example, a load controller may be
able to exercise five different demand response options (such
as temperature reset, lighting reduction, etc.) If there are
three levels to each of these options (such as low, medium,
and high) then there are 35 = 243 total number of
combinations. One must also account for the time window
over which the actions are evaluated. Calculating the number
of permutations over a time window of, say, three hours into
the future with a 15-minute control update interval implies
that the true number of results in the cost matrix is in the tens
or hundreds of thousands.
The bandwidth required can be estimated from the number of
values that must be transmitted within a certain time. As an
example we assume that a matrix of actions and costs
contains 250,000 values and must be transmitted every ten
minutes. If the numbers are four-byte integers then it is
implied that the allowable transfer rate must be in excess of
1300 baud for each load controller. It is safe to assume that
each compact network would contain multiple load controller
agents, possibly tens or hundreds. The communication
bandwidth would have to accommodate all of the messages
send at this rate as well as the conventional messages used by
the utility for normal operations.
VI.

CONCLUSIONS

This project is ongoing and the GridAgents software is being


used for large-scale integration of distributed energy and
renewable energy resources into distribution systems with
electric utilities with specific applications on SmartGrid based
energy networks including Microgrid management, intelligent
load control and smart charging applications, including
photovoltaics & storage, load control, and future assets such
as plug-in hybrid cars. In the future, more applications will be
developed as well as embedded device OEM models
supported to support SmartGrid deployments and advanced
distribution automation applications.
VII. ACKNOWLEDGMENTS
The author gratefully acknowledges the contributions of
Arthur Kressner and Serena Lee of Con Edison, and Peter
Curtiss and Robert Dodier for their participation in this work.
.

VIII. BIOGRAPHIES

David Cohen is co-founder and Chief Executive Officer of


Infotility, Inc. and is currently leading the development of a next-generation,
intelligent agent-based software infrastructure (GridAgents) for SmartGrid
applications. David Cohen has over 20 years of experience in the energy and
software industries specializing in renewable energy, distributed computing, and
distributed energy. Prior to co-founding Infotility, David managed the
development and market launch of Silicon Energy's industry leading Distributed
Energy Manager (DEM) product, a web-based software platform for
networking, managing, and controlling distributed energy resources, part of
Silicon Energy's EEM Suite. Using the DEM product, David co-developed the
industries first Systems Operation Center (SOC) software with DTE Energy
Technologies branded as EnergyNow. He also co-developed ABBs Virtual
Power Plant application based on the DEM software. Prior to co-founding
Silicon Energy, David was a Principal with ICF Consulting in Washington, D.C.
David has held various senior-level positions at Cable and Wireless
Communications, Architectural Energy Corporation, Highland Energy Group
and the Joint Center for Energy Management. David is actively involved in the
Renewables, CleanTech, and SmartGrid industries including his current board
position on the GridWise Architecture Council (GWAC), team member of
Modern Grid Initiative (MGI), the Galvin Initiative Microgrid working group,
and a member of the American Council for Renewable Energy's (ACORE) CEO
council. David was a founding member of the Colorado Renewable Energy
Society (CRES), Vice President within the Solar Energy Industries Association
(SEIA), and currently serves on the board of Green Building Studio, Inc, and
Aura Renewable Energy Corporation.

Potrebbero piacerti anche