You are on page 1of 9

Intelligent Distributed Autonomous Power Systems (IDAPS) and their Impact

on Critical Electrical Loads

Manisa Pipattanasomporn and Saifur Rahman


Department of Electrical and Computer Engineering, Virginia Tech, USA
E-mail: mpipatta@vt.edu and srahman@vt.edu

Abstract circuits rely entirely on electricity from the grid and no


internal generation is installed in a particular circuit to
The transition toward deregulation and profit- serve all critical loads when the grid fails. In addition,
driven operation of power systems has led to a there is no local supervisory control unit responsible
decrease of reserves and reliability levels in various for establishing the optimum operating strategy to
parts of the electric power network which is the dispatch internal generation.
backbone of the nationwide critical infrastructure. In To tackle these challenges, we propose the concept
order to improve the adequacy and security of power of Intelligent Distributed Autonomous Power Systems
systems, we propose the concept of Intelligent (IDAPS). This concept integrates two areas of research,
Distributed Autonomous Power Systems (IDAPS). namely characterizing a generic architecture of
IDAPS is a network of several interconnected sub- intelligent agents for the cooperative operation of
networks that are cellular in structure and can operate elements in an autonomous power system (APS), and
autonomously in the event of a system fault. Each sub- quantifying the optimum size and type of generation to
network (cell) is given enough generating capacity to meet critical loads.
secure critical loads, as well as the ability to
communicate internally among generation, loads and 2. Overview of IDAPS
electronic control devices. The IDAPS concept,
therefore, ensures an uninterrupted supply of IDAPS is a system of systems [1] that looks at the
electricity to the high priority loads, thus maintaining operation of power systems of the future. It comprises a
the integrity of the critical infrastructure. This paper group of loosely connected APS’s. To start building the
defines the concept, components and architecture of IDAPS from scratch, existing electric power infrastruc-
IDAPS, and demonstrates how the electronic control ture must first be examined to identify internal sub-
concept can be applied. networks or APS’s that can be candidates for
autonomous operation - independent of the grid - in the
event of a grid failure. In simple terms, an APS can
1. Introduction consist of the loads fed by one circuit from a substation
transformer including all the switches, distribution
Electricity outages and blackouts have highlighted transformers and control devices, as shown in Figure 1.
the fragility of critical electricity infrastructure. This In a typical electric power distribution feeder, there
degradation of electricity service is mainly due to the are generally no utility-owned and only a few
structural changes in the industry, whereby electricity is customer-owned generators installed. This makes the
looked upon as a marketable commodity rather than a critical loads in a particular circuit significantly
service, as well as the aging of electricity infrastructure susceptible to outages and disturbances in power
in many developed countries. systems. Since the theme of IDAPS is to secure all
In the traditional power system, central-station critical loads located in each APS during outages,
generating plants supply electricity through the distributed generation (DG) must be placed in each
transmission and distribution networks to end-users. distribution circuit if no existing back-up generation
This conventional practice is often not capable of exists. This gives all APS’s an ability to be autonomous
delivering secure and uninterruptible power supply to so that they can be responsible for their own internal
critical loads. This is mainly because most distribution critical loads and operate independently from the grid

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
in the event of a grid failure. We discuss this issue in total annual levelized capital and operating costs of the
more detail in section 3. candidate generators. The former represents the
In addition to the necessary generation, supporting annualized fixed cost ($ per year) of installing the
technologies in sensing, metering, communicating and selected generating units in the APS; and the latter is
controlling the operation of elements within a cell must the total production cost of energy ($ per year) that is
be identified for the IDAPS to succeed. All elements in generated by the selected units if they do not suffer any
the APS, including generators, loads and switching forced outages. The derivation of this factor involves a
devices, must be able to communicate with other well-known reliability index of electric power systems
elements within the cell through intelligent agents. The called loss of load probability (LOLP) [2]. LOLP is
agents are responsible for coordinating the operation of defined as the fraction of time in which the available
generators, loads and switches in a smart and efficient capacity is insufficient to serve the hourly demand.
manner in order to minimize the loss of critical loads in The decision variables of this problem are the sizes
the event of outages, and to shave the peak during of generating units (kW) to be installed in the APS.
normal operating conditions. This subject is further These are of continuous type. DG candidates can be
discussed in section 5. solar photovoltaic systems, diesel engines, fuel cells,
electric vehicles, etc. We minimize the annualized
3. Necessary generation to secure critical investment and operating costs of DG, subject to a
loads constraint that these dispersed generators must be able
to deliver uninterruptible power supply to serve the
To quantify the generation needed to meet critical critical loads at the level of specified reliability. In
loads in a particular APS, we develop a non-linear addition, the maximum size and number of each
optimization model as presented in Appendix A. The candidate generator must be specified. For example, if
objective function of this problem is to minimize the the maximum size of an IC engine is set at 500 kW and

Figure 1. IDAPS components

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
the maximum of three units are allowed, the model can month of August is shown in Figure 3. At present, no
select up to three IC units, each of which can be of any existing DG is installed in this circuit, and electricity is
size between 0 to 500 kW. The actual size of IC engine entirely purchased from the regional power company at
to be installed can then be rounded to the closest 69 kV. This voltage is transformed to a distribution
available size. We can also add environmental voltage of 12.47kV at a substation, which is reduced
constraints for limiting emissions from these distributed further at the end-user transformers to meet the
generators. Doing so can make cleaner but more customer voltage requirements.
expensive technologies, like solar cells and fuel cells,
3500
acceptable as viable solutions.
Datasets required to run this optimization problem 3000
include: size of critical loads, load duration curve of
2500
critical loads, existing generating capacity within an

Demand (kW)
APS, availability of local energy resources, technical 2000
and economic performance of candidate DG’s and
allowable annual emission caps. 1500

1000
4. Improving reliability of critical loads in
500
a test case
0
4.1. Electricity demand 24 25 26 27 28 29 30
-A -A -A -A -A -A -A
ug ug ug ug ug ug ug
-0 -0 -0 -0 -0 -0 -0
In order to construct our test case, we have used 3 3 3 3 3 3 3.
data from the electric power system in the campus town
of Blacksburg [3] in Virginia, USA. There are twelve Figure 3. Chronological load duration curve
distribution circuits as shown in Figure 2. An APS, as (seven days) as of August 2003
defined earlier, is equivalent to a distribution circuit,
consisting of the loads fed by one circuit from a 4.2. Electricity supply candidates
substation transformer.
Since our test circuit does not have existing internal
generation, we consider installing four DG candidates
to serve the critical loads, namely diesel engines (IC),
fuel cells (FC), photovoltaic systems (PV) and
hydrogen fuel cell cars (EV). Technical and economic
data of each candidate is summarized in Table 1.

Table 1. Technical and economic data of DG


candidates
Capital Eff
DG O&M
cost Life icie Fuel cost
units ($/kWh)
($/kW) ncy
IC 1,000 20 Diesel: 0.015 35%
yrs $1.28/gal
FC 5,000 10 30% Gas: 0.010
yrs $6/MMBTU
PV 8,000 20 - - 0.010
yrs
Figure 2. Electric distribution circuits in
EV Lease a fuel cell car at $500 per month, including
Blacksburg town fuel and maintenance (size 90kW)

We select the circuit number 8 as the APS of A diesel engine is the best-known and most mature
interest. This circuit serves mostly small commercial technology with the cheapest capital cost of about
loads with a peak and off-peak demands of 3,007 kW $1,000 per kW installed. A fuel cell is operated using
and 1,098 kW, respectively as of 2003. A sample of the natural gas which is transformed into hydrogen by a
chronological load curve of this circuit during the reformer. The capital cost of a fuel cell unit with a

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
reformer is typically around $5,000 per kW installed. It selected and the rest of critical loads are fulfilled by
is assumed that Blacksburg community can purchase increasing the size of IC engines. If more than two fuel
diesel oil and natural gas at the wholesale prices of cell vehicles are available in this circuit, the model
$1.28/gallon and $6/MMBTU, respectively. tends to select them since their fuel costs and
Solar cell is another possible alternative. However, maintenance costs are already included in the leased
the investment cost of a solar cell system is still high. price.
We also consider as a DG candidate a new and
emerging technology, hydrogen fuel cell cars. The fuel

Critical Load = 500 kW Critical Load = 400 kW Critical Load = 300 kW


EV
cell power plant in the car can be connected to the LOLP=0.001
IC
distribution system when the car is not on the road. At
present, there is an initiative in the United States to put LOLP=0.0001
hydrogen fuel cell cars on the road by 2020. A pilot
project has been launched in California allowing to
lease Honda FCX, the latest fuel cell car, at $500 per LOLP=0.00001
month including fuel and maintenance [4].
LOLP=0.001
4.3. Solutions for reliability improvement

Scenarioes
To secure the critical loads within our test circuit, LOLP=0.0001
we employ the optimization technique discussed in
Appendix A. The model is run to quantify the optimum
LOLP=0.00001
(least cost) size and type of necessary generation. As
for parameters used in the optimization model, the
performance of four DG candidates as summarized in LOLP=0.001
Table 1 is used. We also assume that maximum of three
diesel engines and three fuel cells can be installed in
this circuit, and only two 90kW fuel cell vehicles are LOLP=0.0001
available. The maximum size of diesel engines and fuel
cells is limited by the size of the critical loads. For LOLP=0.00001
example, for the system with 300kW critical loads, the
size of IC engines and fuel cells can be anything
between 0kW and 300 kW. 0 200 400 600 800 1000 1200
Given that the grid reliability in our test circuit is Additional Generation Required (kW)
99%, Figure 4 summarizes the results of optimization
Figure 4. Additional generation required to
problems when the size of the critical demands range
serve critical loads at various reliability levels
from 300kW to 400kW and 500kW. And the reliability
requirement to serve these demands are increased
Notice that fuel cells and solar cells are not viable at
incrementally to 99.9%, then 99.99% and finally to
this time and are not selected by the model. This is
99.999%. Full results are presented in Table A2 in the
because of their high first costs. However, a customer
Appendix.
may be willing to pay the higher price and install FC or
The result indicates that, to improve the reliability
PV due to their more desirable features, such as
level of critical loads from 99% to at least 99.9%, the
producing fewer or no emissions and lower
size of DG should be at least equivalent to the size of
maintenance costs.
the critical loads in our test circuit. And, the size of DG
installed increases with the increase in the required
reliability. For example, to secure the critical loads of 4.4. Value of DG for peak shaving
300kW, roughly 300kW, 360kW and 500kW of
generation is required to reduce the outage probability Based on the calculation discussed in Appendix A
of critical loads to 0.001, 0.0001 and 0.00001, (section A4), Figure 5 presents the accumulated
respectively. demand charges that can be avoided in one year if
The least cost solutions suggested by the model are 300kW, 400kW, or 500kW DG’s are installed to shave
to select fuel cell vehicles and diesel engines as back- the peak demand. These numbers represent the actual
up generation. Two 90kW fuel cell cars are always saving for DG owners provided that the costs to run
DG match the cost to purchase electricity from the grid.

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
60,000 The interface layer connects all data sources, data
Annual avoided demand charges ($) $52,180 storage units and data users to their associated agents.
50,000 For example, a user can have access to a user agent via
$41,744 a user console or a computer in which a particular
40,000 software package is installed. Such software package
$31,308
may include a human-machine interface (HMI) or a
web-based interface with GUI, applets, hypertext
30,000
preprocessors, or active server pages embedded in
HTML. This user console has the capability to sense
20,000
price signals from the utility, as well as display real-
time information on available generation and load in its
10,000 neighborhood. This data can either be instantaneously
gathered from the database via a database agent or
0 other agents in the same network. The database is
300 kW DGs 400 kW DGs 500 kW DGs linked to a database agent through a software package
like SQL or Microsoft Access. Similarly, DG’s, loads
Figure 5. Annual avoided demand charge with and other devices can connect with their associated DG
300kW, 400kW, or 500kW DG’s agents, load agents and device agents, respectively, via
an Ethernet network. For short range communications
Section A3 in the Appendix summarizes the net cost where an agent is located at or near a device, direct
per year to secure critical loads in these three scenarios. connections with physical wires or wireless Bluetooth
The net cost here is defined as the difference between and 802.11 are feasible.
the total annual levelized costs of installing and Next to the interface layer is the agent layer. The
operating DG’s and the avoided annual demand agent layer comprises of multiple agents that are
charges. interconnected through the LAN. These agents are
responsible for connecting all elements in a particular
5. IDAPS agent technology APS and consolidating information to protect critical
loads when the grid fails. All agents communicate with
5.1. Agents in IDAPS concept every other agent using the agent communication
language (ACL) or protocols [6]. The ACL gives all
Historically, data communications in traditional agents in the APS a means to exchange necessary
electric power systems has been based on the industry information and knowledge to accomplish the given
standard called the supervisory control and data tasks.
acquisition (SCADA) system. The SCADA system can
centrally gather information from remote terminals and 5.3. Agent’s responsibility
control the operation of electronic devices in the
system. In the case of the IDAPS, each APS will User agents are basically gateways to connect users
employ its own SCADA capability and the multi-agent to the rest of the system. In order to gain access to the
concept that operates on an IP network. An agent is a database, it is necessary to have a database agent which
software-based entity that can sense and respond to is responsible for collecting and storing information in
changes in signals to accomplish its individual goal [5]. a database, and providing access to the database for
More complex tasks and problems involved in the other agents.
IDAPS lead to the need for utilizing more than one DG agents, load agents and device agents connect all
agents, constituting a group of agents that can work in a DG’s, loads and electronic devices, i.e., circuit
cooperative manner in a particular APS. breakers and protective relays, to the system. They
gather the following information from all generators,
5.2. A multi-agent architecture in IDAPS loads and devices: ID (identification number, owner
name, address, etc); LOCATION (circuit number,
A generic architecture of cooperation among distribution branch, etc); SIZE (kW rating, nameplate,
multiple IDAPS agents is illustrated in Figure 6. The etc); TYPE (internal combustion engines, fuel cells,
architecture consists of the interface layer, the agent critical/non-critical loads, switches, etc); STATUS
layer, and the inner-most layer, which is the local area (on/off, down, standby, etc); PERFORMANCE
network (LAN). (efficiency, heating value, forced outage rate, etc); and

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
Figure 6. Architecture of the multi-agent system in an IDAPS

PRICE (DG variable costs, price of purchasing grid Securing critical loads
electricity, demand charge, outage costs, etc).
The data analysis and IDAPS management agents When a device agent senses a failure in the electric
are the central processing units that can reside power system from the abnormal status of a device
anywhere in the APS. These agents are in charge of located in an APS, the device agent sends a set of
developing an instantaneous optimum dispatch scheme ACL’s to the data analysis agent. The data analysis
based on the information provided by other agents. agent then instantaneously requests information from
Then the signal is sent to the control agent to distribute DG agents, load agents, and device agents to check the
to all elements in the APS. availability of current generation and loads. Upon
receiving information from the data analysis agent, the
5.4. Applications IDAPS management agent calculates the optimum
dispatch strategy to ensure availability of electricity to
By employing the multi-agent architecture, instant the high-priority loads (i.e., hospital or loads with high
communications can be initiated among generators, outage costs) and minimize load curtailment. Then, the
loads and electronic control devices by appropriate IDAPS management agent sends ACL’s to the control
agents within an APS. This will help to achieve IDAPS agent, which in turn forwards this information to DG
given goals, i.e., to secure critical loads during faults agents, load agents and device agents. These agents
and to employ demand side management during normal therefore dispatch their associated elements by starting
operating conditions. standby generators, or shed any unnecessary low-
priority loads.

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
Employing demand side management commercial/industrial facilities”, IEEE Transactions on
Power Systems, Vol. 20, No. 1, February 2005.
During normal operating conditions, when a user [3] Virginia Tech Utilities [Online]. Available: http://
console senses a high price signal from the grid, the facilities.vt.edu/utilities/.
customer may want to reduce the peak load to avoid
[4] Honda Fuel Cell Cars [Online]. Available: http://world.
high demand charges. In that case, the console sends a honda.com/FuelCell/.
set of ACL’s to the data analysis agent to initiate
demand side management. The data analysis agent then [5] D. P. Buse, P. Sun, Q.H. Wu and J. Fitch, “Agent-based
instantly requests information from DG agents, load substation automation”, Power and Energy Magazine, IEEE
agents, and device agents to check the availability of Vol. 1, Issue 2, pp. 50-55, March-April 2003.
standby generation, loads, and operating costs of all [6] Y. Labrou, T. Finin and P. Yun, “Agent communication
available DG’s. The data analysis agent then sends the languages: the current landscape”, Intelligent Systems and
necessary information to the IDAPS management Their Application, IEEE, Vol. 14, pp. 45-52, March-April
agent. The management agent then determines which 1999.
internal generators to call upon to produce the [7] Virginia Dominion Power, Electricity tariff for
necessary generation to shave the peak and sell intermediate general service – time of usage (schedule GS-
electricity back to the grid, if extra generation is 2T) [Online]. Available http://www.dom.com/customer/pdf/
available. va/vabgs2t.pdf.

6. Summary
Biographies
This paper presents a concept of IDAPS that
outlines the intelligent, distributed and autonomous Manisa Pipattanasomporn (S’01 – IEEE) received a B.S.
aspects of power systems of the future. The architecture degree from the Electrical Engineering Department, Faculty
of IDAPS is composed of several autonomous power of Engineering, Chulalongkorn University, Bangkok,
systems (APS) that can independently meet their own Thailand in 1999. She received the M.S. degree from the
critical loads during outages. To ensure that the Energy Program, School of Environmental Resources and
Development, Asian Institute of Technology (AIT),
reliability of critical loads can be increased to the
Bangkok, Thailand in 2001. She received her Ph.D. in
required level, a methodology has been presented that electrical engineering from Virginia Polytechnic Institute and
estimates the size and type of DG’s required, in State University (Virginia Tech), U.S.A in 2004. She is
addition to the electricity purchased from the grid. We currently working as a post-doctoral associate at the
also present the generic architecture of implementing Alexandria Research Institute in Alexandria, Virginia. Her
the multi-agent module in an APS, and propose the fields of interest are renewable energy systems and
application of the agent-based system during faults as distributed generation.
well as normal operating conditions. With the adequate
generation to secure critical loads and the multi-agent
architecture, autonomy can be provided to each APS to Saifur Rahman (S’75, M’78, SM’83, F’98 – IEEE) is the
Director of Alexandria Research Institute at Virginia Tech
be self-sufficient during grid failures, constituting an
where he is the Joseph R. Loring professor of electrical and
inherently distributed power system that can maintain computer engineering. He also directs the Center for Energy
the integrity of the critical electric power infrastructure. and the Global Environment at the University. Professor
Rahman is currently serving as the vice chair of the IEEE
7. References Publications Board. He is an IEEE Distinguished Lecturer.
He has served on the IEEE Power Engineering Society
Governing Board for five years first as the Vice President for
[1] S. Rahman, "Intelligent Distributed Autonomous Power
Education and Industry Relations and then as the VP for
Systems - A concept paper", presented at the Workshop on
Technical Information Services. He served as the chairman of
Intelligent Distributed Autonomous Power Systems,
the IEEE Lifelong Learning Council in 2002. He is also a
sponsored by the US National Science Foundation and the
member-at-large of the IEEE-USA Energy Policy
Japan Society for the Promotion of Science, jointly organized
Committee. He has published over 300 papers on
by Virginia Tech and the Yokohama National University,
conventional and renewable energy systems, load
Hakone, JAPAN, July 1998.
forecasting, uncertainty evaluation and infrastructure
[2] M. Pipattanasomporn, M. Willingham, and S. Rahman, planning.
“Implications of on-site distributed generation for

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
Appendix A 1. Reliability constraint: reliability requirement of
critical loads must be met.
A1. Optimization model 1 − LOLP ≥ required _ reliability (5)
N
where LOLP = ELDC N (¦ X i ) (6)
The objective function shown below represents the i =1
non-linear optimization of the total levelized annual
capital and operating costs. 2. Maximum number of units to be installed.
Minimize: 0 ≤ no _ IC ≤ Max _ noIC
N ­ ½°
° TCi
0 ≤ no _ FC ≤ Max _ noFC (7)
¦ ®FCi ⋅ X i + 8760 ⋅ pi ⋅ VCi ⋅ ³ ELDCi −1 ( L)dL¾ (1)
i =1 °̄ TCi −1 °¿ 0 ≤ no _ EV ≤ Max _ noEV
where
FCi : the annual levelized fixed cost of 3. Maximum size of each technology.
generating unit i ($/kW per year); 0 ≤ X IC ≤ MaxIC (kW )
Xi : decision variable representing the (8)
0 ≤ X FC ≤ MaxFC (kW )
size of generating unit i (kW);
VCi : the variable cost, including fuel cost Available _ Area
0 ≤ X PC ≤
and maintenance cost, of generating size _ of _ 1kWPV
unit i ($/kWh); 0 ≤ X EV ≤ MaxEV (kW )
pi : 1-qi when qi is the forced outage rate
(FOR) of generator i;
4. The annual emission caps for carbon dioxide
ELDCi (L) : equivalent load duration curve seen
(CO2), nitrogen oxide (NOX) and sulfur oxide
by unit i;
(SOX).
TCi : the total capacity of i generating TC i
i
units. TC i = ¦ X k 8760 ⋅ p i ⋅ EFi ,CO ⋅ ³ ELDC i −1 ( L)dL ≤ EC CO
2 2
(9)
TC i −1
k =1
where
FCi and VCi are calculated according to (2) and (3), EFi,CO2 : CO2 emission factor of generating
respectively. unit i (lb/kWh);
ECCO2 : CO2 annual emission cap (lb).
r (r + 1) n § $ ·
FCi = n
⋅ INVi ¨ ¸ (2)
(1 + r ) − 1 © kW ¹ A2. Demand and supply assumptions
§ $ · 1 § gallon · 3412 § BTU · § $ ·
VC i = f i ¨¨ ¸¸ ⋅ ¨ ¸⋅ ¨ ¸ + OM i ¨ ¸
© gallon ¹ HV i © BTU ¹ eff i © kWh ¹ © kWh ¹ Critical loads: 300kW, 400kW and 500kW with the
inversed load duration curves (ELDC0) as shown in
(3)
Figure A1. Grid reliability used is 99% and the
where
required reliability levels of critical loads are 99.9%,
r : interest rate (assuming at 10%);
99.99% and 99.999%. Supply assumptions are
n : lifetime of generating unit i (year);
summarized in Table A1.
INVi : the investment cost of generating unit i as
listed in Table 1;
fi : the fuel cost of generating unit i;
HVi : the heating value of fuel input to
generating unit i;
effi : the efficiency of generating unit i;
OMi : the operation and maintenance cost of
generating unit i.

ELDCi (L) is calculated using (4).


ELDC i ( L) = p i ⋅ ELDC i −1 ( L) + q i ⋅ ELDC i −1 ( L − X i )
(4)
Subject to the following constraints: Figure A1. Inversed load duration curves of
critical loads (300kW, 400kW and 500kW)

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE
A3. Summary of results • The distribution demand charge is $3.387 per kW;
• The electricity supply demand charge is $6.249 per
The results from the optimization model is shown in kW for the billing months of June to September;
Table A2, summarizing additional generating capacity and
required to increase the reliability of electricity supply • The electricity supply demand charge is $4.840 per
of critical loads to the level of 99.9%, 99.99% and kW for the billing months of October to May.
99.999%, given that the current reliability level The distribution demand charge is calculated based
available from the grid is 99%. on the highest average kW measured during any 30-
It must be pointed out the negative costs (or minute interval of the current and previous 11 billing
incomes) reflected in Table A2 will be more than months. On the other hand, the electricity supply
neutralized when the cost of the agent based control demand is the highest average kW measured in any 30-
system is included. For the lack of cost information on minute interval of the current billing month during on-
such systems, these costs were not included here. peak hours.
For example, with 295kW DG’s installed to shave
A4. Calculation of avoided demand charges the peak demand (Table A2 column 2), during June to
September the avoided demand charge is calculated as
According to the electricity tariff of a power 295kW x (3.387+6.249) x 4 months; similarly during
company in Virginia, the following fees are the demand October to May, the avoided demand charge is
charges for a 30-day period for an intermediate general calculated as 295kW x (3.387+4.84) x 8 months. Thus,
service (30-500kW) [7]: the total avoided demand charge is $30,786 per year.

Table A1. Supply assumptions


Levelized annual (1) (2) (1)+(2)
Generating Max pi (1- capital cost (FCi) fuel cost O&M Variable cost
Maximum size
units units FOR) ($/kW per year) ($/kWh) cost (VCi)
($/kWh) ($/kWh)
IC  95.3% Critical loads $117.46/kW/yr 0.09562 0.015 $0.11062/kWh
FC 3 99.5% Critical loads $813.73/kW/yr 0.06824 0.010 $0.07824/kWh
PV ∞ 33.3%* α area $939.70/kW/yr - 0.010 $0.010/kWh
EV 2 99.5% 90kW each $66.67/kW/yr - - $0/kWh
* This number is based on the assumption that the capacity of PV is available only 8 hours per day.

Table A2. Additional generators required to secure critical loads to the levels of specified reliability
Scenario I: Scenario II: Scenario III:
Critical load is 300kW Critical load is 400kW Critical load is 500kW
LOLP 0.001 0.0001 0.00001 0.001 0.0001 0.00001 0.001 0.0001 0.00001
EV1 capacity (kW) 90 90 90 90 90 90 90 90 90
EV2 capacity (kW) 90 90 90 90 90 90 90 90 90
PV capacity (kW) 0 0 0 0 0 0 0 0 0
FC1 capacity (kW) 0 0 0 0 0 0 0 0 0
FC2 capacity (kW) 0 0 0 0 0 0 0 0 0
FC3 capacity (kW) 0 0 0 0 0 0 0 0 0
IC1 capacity (kW) 115 60 106 215 109 205 315 160 307
IC2 capacity (kW) 0 60 106 0 109 205 0 160 307
IC3 capacity (kW) 0 60 106 0 109 205 0 160 307
Total capacity (kW) 295 360 498 395 507 795 495 660 1101
Total cost ($/yr) 26,260 33,600 50,130 38,930 52,140 86,520 51,600 70,720 122,800
Avoided demand 30,786 37,569 51,971 41,222 52,910 82,966 51,658 68,877 114,900
charges ($/yr)
Net cost ($/yr) -4,526 -3,969 -1,841 -2,292 -770 3,554 -58 1,843 7,900

Proceedings of the 2005 First IEEE International Workshop on Critical Infrastructure Protection (IWCIP’05)
0-7695-2426-5/05 $20.00 © 2005 IEEE