PLANNING DISTRIBUTION SYSTEM RESOURCE ISLANDS CONSIDERING RELIABILITY, COST AND THE IMPACT OF PENETRATION OF PLUG-IN HYBRID ELECTRIC VEHICLES Julieta Giráldez Graduate Student Division of Engineering CSM.
Download ReportTranscript PLANNING DISTRIBUTION SYSTEM RESOURCE ISLANDS CONSIDERING RELIABILITY, COST AND THE IMPACT OF PENETRATION OF PLUG-IN HYBRID ELECTRIC VEHICLES Julieta Giráldez Graduate Student Division of Engineering CSM.
Slide 1
PLANNING DISTRIBUTION SYSTEM RESOURCE
ISLANDS CONSIDERING RELIABILITY, COST AND
THE IMPACT OF PENETRATION OF PLUG-IN
HYBRID ELECTRIC VEHICLES
Julieta Giráldez
Graduate
Student
Division of
Engineering
CSM
1
Slide 2
2
• Outline
o
Introduction
o
Design of distributed resource islands
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on distributed resource islands
o
Conclusions and future work
Slide 3
3
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 4
4
• Introduction
Smart Grid Initiative [1]: what is the evolution of
electric power distribution systems?
• Distributed
Energy
Resources
(DER)
or
Distributed
Generation (DG)
• Incorporate ways of physical and virtual storage to balance
consumption and production including PHEVs
• Increased used of technologies: advanced meters, advanced
inverters, distribution automation, communication systems, etc.
[1] 110thCongress of the United States, "Title XIII (Smart Grid)," in Energy Independence and Security Act of 2007. Washington,
DC: Dec. 2007, pp. 292 –303.
Slide 5
5
• Introduction
Why?
• Contribute to the load relief of the transmission system by
increasing the generation in the distribution system and new
ways of energy management
• Higher reliability and power quality
• Integration of green technologies into the grid
Slide 6
6
• Introduction
How to implement the smart grid?
Microgrid concept: a distributed resource island
• Self-contained autonomous subset of the area electric power
system
• Has local Distributed Energy Resources (DER)
• Operates semi-autonomously of the grid, being able to island and
reconnect as circumstances dictate
• Able to provide power quality and reliability different from
general macro-grid standards
[2] N.Hatziargyriou, H.Asano, R.Iravani, and C.Marnay , “Microgrids”, IEEE Power & Energy Magazine, pp.78-94, July/Aug.
2007.
Slide 7
[3] “Distributed Energy Resources Integration”, Consortium for Electric Reliability Technology Solutions (CERTS), [Online].
Available:http://certs.lbl.gov/certs-der.html
7
Slide 8
8
• Outline
o
Introduction
o
Design of distributed resource islands
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 9
9
• Design of distributed resource islands
Distribution systems
•Traditional
electric distribution systems:
Infinite bus
Transformer
Line
Grid
Grid
Load
Slide 10
10
• Design of distributed resource islands
Distribution systems
•Evolving
distribution systems:
Grid
Grid
Increase annual RELIABILITY at a feasible COST
Slide 11
11
• Design of distributed resource islands
Modeling of annual load
• Annual demand: two ways of modeling annual load
annual average demand at every load: i.e. 1 load level
representative of the annual demand
6 step-load duration curve representation (hourly demand
∆T2=1900h
reordered in increasing demand): i.e. 6 load levels
representative of the annual demand
∆T1 =100 h
[4] R. Billinton, S. Kumar, et al., "A Reliability Test System for Educational Purposes - Basic Data," IEEE Transactions on Power
Systems, vol. 4, pp. 1238-1244, August 1989.
Slide 12
12
• Design of distributed resource islands
Modeling of DG
• DG: aggregate power output of Renewable Energy (RE)
and Conventional Distributed Generation (CDG) [5]
Pout = CDG + RE + DS
Pout = R CDG * CF CDG
R RE * CF RE
(1 CF RE ) * 0 . 15 * L i
Capacity Factor: ratio of the actual output of a power
source and its output if it had operated at full capacity
Total DG rating R=RRE + RCDG
[5] H. Brown, “Implications on the Smart Grid Initiative on Distribution System Engineering: Improving Reliability on Islanded
Distribution Systems with Distributed Generation Sources, M.S thesis, Dept. Elec. Eng., Colorado School of Mines, 2010.
Slide 13
13
• Design of distributed resource islands
Basic Reliability Concepts:
• ASAI: The time as a fraction of a year for which the system is
available
• Annual Outage Time, U : Time as a fraction of a year for which
the system is NOT available ~ U 1 ASAI
• Power Not Supplied (PNS) [MW]: Unserved load or demand
that the system cannot attend
• Reliability metric: Energy Not Supplied [MWh]
8760
ENS U
PNS ( h )
h 1
Slide 14
14
• Design of distributed resource islands
I. Enter the power system component data
Slack
bustool:
is modeled as a
Powerbus:
systemsslack
simulation
generator
thatprogram
absorbs
supplies
• Computer
to solveor
a power
flow: generation
supplies
demand, and
to control
the frequency of the
in order1.toGeneration
balance
thethe load
generation
system
~ Power Not Supplied~
Bus
magnitudes
remainthree
closephase
to the
rated values
II.2.Solve
thevoltage
Power Flow
under balanced
conditions
2.3.
1.
Lines and transformers are not overloaded
• PowerWorld SimulatorTM is used
3.
Slide 15
15
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 16
16
• MOGA
Multi-objective redesign problem
•
Investment cost versus reliability: Pareto-optimality
~ no single optimal solutions but a set of alternative solutions ~
•
Non-linear problem, discrete and non-convex feasible region
•
Intractability of the problem as the size of the system grows [5]
•
Evolutionary methods
[5] H. Brown, “Implications on the Smart Grid Initiative on Distribution System Engineering: Improving Reliability on Islanded
Distribution Systems with Distributed Generation Sources, M.S thesis, Dept. Elec. Eng., Colorado School of Mines, 2010.
Slide 17
17
• MOGA
Mathematical formulation
•
Variables
•
Objective function 1: COST
Ng
Nc
f1 ( x )
C
i 1
CC :Cost of conductor [$/km]
li : Length of connection i [km]
l x
c c
i
c
i
C
DG
P
DG
j
g
j
(x )
j 1
CDG: Cost of DG [$/MW]
PDGj: Power output of DGlocated at bus j [MW]
Slide 18
18
• MOGA
Mathematical formulation
•
Objective function 2: RELIABILITY ~ Energy Not Supplied
8760
ENS f 2 ( x ) U
•
h 1
Annual average loads:
f 2 , Average
•Six
_ Loads
( x ) 8760 * U * PNS
step load duration curve:
6
f 2 , Load
PNS ( h )
(t )
U * PNS ( L Ti ) * Ti , with
i 1
6
T
i
i
8760 h
Slide 19
19
• MOGA
Mathematical formulation
•
Constraints
Voltage
within 5% of the nominal value at every bus j:
0 . 95 V j ( x ) 1 . 05
Loading of the Line from bus j to k:
S jk 1 . 00
Slide 20
20
• MOGA
*GAs
*
•
and the fitness function
A population is comprised of individuals or chromosomes ~
a potential
solution to the optimization problem
*
•
Evolutionary operators are used to create randomly individuals which may
* move to a higher level of fitness such as mutation, recombination, and
crossover.
•
MatlabTM Genetic Algorithm Optimization Toolbox (GAOT) inbuilt
functions
• The
fitness function determines how likely an individual is to survive to the
next generation ~ output of fitness function ~
f1 ( x )
f
f 2 ( x)
Slide 21
21
• MOGA
Importance of the initial population for convergence
• Explore 3 ways of selecting the initial population
Slide 22
22
• MOGA: RBTS test system
Application to a test system
•
RBTS System [6]:
Customer Type
Load points i
Average Load,
Max. Peak Load,
[MW]only the 164
[MW]
Possible Connections 302. We input
connections
Residential
1, 4-7, 20-24, 32-36
0.4684
which length is less
than 3km.
Residential
11, 12, 13, 18, 25
0.8367
0.4758
0.8500
0.4339
0.7750
Possible DG Location 27 buses
2, 15, 26, 30
•DG: ResidentialDESIGN
PARAMETERS
0.8472 0.25, 0.3,
1.0167
0.8
8, 9, 10
CF: Small
wind, solar, conventional
Industrial
DG
Commercial
16, 17, 19, 28, 29, 31,
Total DG 3,penetration
80%0.2886
Total Annual0.5222
Average Load
37, 38
RE penetration
20% Total DG penetration
Office
14,27
0.5680
0.9250
[6] R. Billinton and S.Buildings
Jonnavithula, "A Test System for Teaching Overall Power System Reliability assessment," IEEE Transactions on Power
Systems, vol. 11, pp. 1670-1676, November 1996.
Slide 23
Solution #
Connection (s)
DG (s) bus location #
1
2
Line 11-17
Line 1-7
15
4 & 15
Cost
[106
23
US $]
• MOGA: RBTS Test system
ENS [MWh]
17.97
18.06
31.40
21.96
Line 11-17
4 & 15
25 decision maker
18.12
Results:
“look-up table”
for&the
21.88
Application
to a test system
Line 11-17
Line 1-7
3
•
Line 23-29
• A more
4
expensive solution may be chosen if the Value of Lost Load
Line 11-17
13 & 29
18.31
21.62
7 & 13 & 23
18.47
21.36
11 & 13 & 23
18.91
21.33
(VOLL) [$] of the system is greater than the investment cost
Line 23-29
Line 11-17
5
Line 17-23
Line 23-29
Line 1-7
6
Line 9-15
Line 21-27
Slide 24
24
• MOGA: RBTS Test system
If VOLL ≤ Cost Solution 6 might be chosen
Slide 25
25
• MOGA: RBTS Test system
Application to a test system
Time [h]
•Very
40
similar redesign solutions for the RBTS with annual average loads
and
35 with step-load duration curve
• 30
ENS
overestimated with annual average demand
25
•
Computational time :
20
15
10
5
0
modeling of the annual load
connection from Matlab to PowerWorld Simulator
initial(i) population
RBTS Case I:Annual average loads
(ii)
(iii)
Initial population type
RBTS Case II: Step-load duration curve
Slide 26
26
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 27
27
• Impact of PHEVs in distributed resource islands
Introduction to PHEVs
•
IEEE definition: “vehicles that have a battery storage system rating of
4 kWh or more, a means of recharging the battery form an external
source, and the ability to drive at least 10 miles in all electric mode”
•
Vehicle-to-grid (V2G): using the battery of a vehicle as a Distributed
Energy Resource (DER)
•
New way of electric energy management
•
Existing power system infrastructure may not be adequate to deal with
the increased demand and new patterns of consumption and power
flows in the grid
[7] “Vehicle to Grid (V2G) Electricity” , Global Greenhouse Warming, [Online]. Available: http://www.global-greenhousewarming.com/vehicle-to-grid.html
Slide 28
28
• Impact of PHEVs in distributed resource islands
Modeling PHEVs in distribution systems
•
How many PHEVs?
•What
is the behavior of the driver?
•
For how long does a PHEV behave as a load?
•
For how long does a PHEV behave as DG?
~ KEY ASSUMPTIONS TO
STUDY THE IMPACT ~
Slide 29
29
• Impact of PHEVs in distributed resource islands
Modeling PHEVs in a distribution system
•How
•What
manykind
vehicles?
of
PHEVs?
• For how long does
• What design and
the PHEV behaves as
operational
a load?
•How
characteristics?
many PHEVs in
• … and as a
• What
theissystem?
the behavior
generator?
of the driver?
• Electric customer consumes
• Driving
2 kW
and has 1.5factors
vehicles for
residential;
38 workers per
1. Peak-shaving
office building and 17 workers
2.per• commercial
Owner’s
Probabilistic
benefit
and 1.5
•Linear
simulation
Programming
vehicles
per worker
(LP)
methodology
algorithms to
• optimize
30% penetration
of the
charging
total transportation
patterns fleet
Slide 30
Probabilistic simulation of PHEV fleet for 8760 hours [8]
30
PHEV Class
Class 4 islands
PHEV Class
PHEV Class 3 PHEV
• Impact
of 1PHEVs
in2 distributed
resource
Tools &
methods
Daily vehicle data for optimization
Energy
Miles driven
Arrival time
required
Methodology
Departure
time
•
Probabilistic simulation methodology [8]
•
LP Optimization of daily charging pattern of PHEVs for 1 year
Objective/s: maximize owners profit and/or utility peak shaving
(Demand response)
Contributions made by this thesis:
LP algorithm
~ determine
Incorporate
optimized PHEV
load (hourly) tothe
loadloading
duration
Impact
curve of distribution system
on
design
Impact of PHEV
fleet on annualresource
distributed
reliability of islanded legacy radial
distribution systems
and
reliability
of
Impact of PHEV fleet on annual
islands
reliability of islanded networked
distribution systems
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖
PSERC Document 09-12, Oct. 2009.
Results
Slide 31
31
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Four vehicle classes (types)
•
Design characteristics (SOC):
Vehicle class c
1
2
3
4
Battery size
Bc [kWh]
Max
12
14
21
23
Min
8
10
17
19
[8] S. Meliopoulos, J. Meisel and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖
PSERC Document 09-12, Oct. 2009.
Slide 32
32
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Amount of driving supplied from electric battery? From fuel?
•
kphev=0 represents a charge sustaining (CS) mode in which on
average all the drive energy comes from gasoline
•
kphev=1 represents a charge depleting (CD) mode, all of the drive
energy comes from electricity
•
•
Simulations run in Powerdrive Simulation Analysis Tool (PSAT)
Performance parameter Ec: required energy per mile [kWh/mi.]
E
bc
E
c
c
c
E a ( kphev )
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 33
33
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Vehicle control strategy: drive in CD from battery while in
SOC ranges and switch to CS to maintain SOC relying on gas
•
M
D
Charge depleting distance MD
Vehicle
class c
Bc
Ec
M
D
40 miles
E c a ( kphev c )
E
c
E
bc
kphevc
Max
min
1
0.5976
0.2447
2
0.6151
0.2750
3
0.5428
0.3217
4
0.4800
0.3224
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 34
34
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
•
Four random paramaters
1.
kphevc and Bc
2.
# Vehicles per class
3.
Daily Miles driven per vehicle
4.
Driver’s behavior ~ Time parameters
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 35
35
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
1.
Vehicle design characteristics kphevc and usable battery capacity Bc
are distributed according to a bivariate normal distribution with mean
vector μ and covariance matrix ∑ with 0.8 parameter correlation
•
Performance
parameter BC
Ec is[kWh]
determined knowingkphev
kphevc
Vehicle
c
class c
E c a ( kphev )
E
E
14.3015
c
14.1827
bc
0.5976
c
0.6151
3
19.1516
0.5428
4
21.3211
0.4800
1
2
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 36
36
• Impact of PHEVs in distributed resource islands
Probabilistic Sim. Meth. applied to the RBTS test system[8]
2.
Vehicles per class: normal distribution with mean #PHEVs*Probability vehicle
class and 1% standard deviation
• Total #
•
vehicles (light transportation fleet): 15, 269 = 14,925 res + 230 com+ 114 off
Uniform distribution of the #PHEVs throughout the load points of the RBTS per
demand type~ daily parameters generated only for the #PHEVs in one load type~
• Vehicle
population size per class:
Vehicle
Vehicles per load point type
class
• Approximate
c
Residential
Commercial
Office building
1
44
15
7
2
48
18
9
3
45
14
6
4
49
19
8
to the average #PHEV per class per load type
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 37
37
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
3.
Miles driven per vehicle per day Md,c,v : log normal
distribution with mean 3.37 and standard deviation of 0.5
•
Daily energy required per vehicle from the grid [kWh]:
DE
d ,c ,v
, if MD ≤ Md,c,v
Bc
D
M
*
E
,
if
M
≤
M
c
d,c,v
d ,c ,v
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 38
38
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
4.
Driver’s behavior ~ time parameters: Gaussian distribution
Departure (am)
•
Arrival (pm)
Parameter
Weekday
Weekend
Weekday
Weekend
μc
σc
7
1.73
9
2.45
6
1.73
15
2.45
Only residential charging in [8], what about office and commercial loads?
Dep
Arr
non res
d ,c ,v
non res
d ,c ,v
Arr
Dep
res
d ,c ,v
res
d ,c ,v
M
d ,c ,v
S
M
d ,c ,v
S
average urban driving speed
25 [mi./h]
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 39
39
• Impact of PHEVs in distributed resource islands
LP algorithms
•
By now we know:
Size and design characteristics of the PHEV fleet
Daily energy required per vehicle from the grid
Daily available time for charging per vehicle
DETERMINE DAILY CHARGING
PATTERNS: Utility peak shaving or benefit
of the owner
Slide 40
40
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
•
Sets:
I = set of load types , from 1 … NI
C= set of PHEV classes, from 1…NC
V= set of PHEVs per class, from 1 … NV
D=set of days in a year, from 1…ND
T= set of hours in a day, from 1 … NT
Slide 41
41
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
•
Parameters:
B
c = Battery size per vehicle class c [kWh]
max
C c = Maximum hourly charge rate per vehicle class c [kW]
DE
d,c,v = Daily energy required per day d, vehicle class c and vehicle v
base
L d,i,t = Base load (without PHEVs) on day d, load type i and hour t [kW]
[kWh]
Lavd,i = Average base load (without PHEVs) on day d and load type i [kW]
Ad,c,v = Daily arrival time per day d, vehicle class c and vehicle v [h]
Pd,t = Price of energy on day d and hour t [$/kWh]
Dd,c,v = Daily departure time per day d, vehicle class c and vehicle v [h]
From the Probabilistic Simulation Methodology
Slide 42
[9] Reliability Test System Task Force of the Application of Probability Methods Subcommittee, “IEEE reliability test system,” IEEE
Transactions on Power Apparatus and Systems, vol. PAS-98, no. 6, pp. 2047-54, November 1979.
42
• Impact of PHEVs in distributed resource islands
Day system % Annual Peak Load
Application to the RBTS test
• Base load of the system:
Customer
Load points i
Max. Annuak
Type
Peak Load,
Residential
[MW]
0.8367
Residential
1, 4-7, 20-24, 3236
11, 12, 13, 18, 25
Week
0.8500
Residential
2, 15, 26, 30
0.7750
Small
Industrial
8, 9, 10
1.0167
3, 16, 17, 19, 28,
29, 31, 37, 38
0.5222
14,27
0.9250
Commercial
Office
Buildings
1
2
3
4
5
6
7
8
9
10
11
12
13
Peak
Load
82.2
90
87.8
83.4
88
84.1
83.2
80.6
74
73.7
71.5
72.7
70.4
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday
Week
Peak
14
15
16
17
18
19
20
21
22
23
24
25
26
Load
75
72.1
80
75.4
83.7
87
88
85.6
81.1
90
88.9
89.6
86.1
93
100
98
96
94
77
75
Week
Peak
27
28
29
30
31
32
33
34
35
36
37
38
39
Load
75.5
81.6
80.1
88
72.2
77.6
80
72.9
72.6
70.5
78
69.5
72.4
Week
Peak
40
41
42
43
41
45
46
47
48
49
50
51
52
Load
72.4
74.3
74.4
80
88.1
88.5
90.9
94
89
94.2
97
100
95.2
Slide 43
43
• Impact of PHEVs in distributed resource islands
Midnight
Midnight
0.052
Application to the RBTS test system
0.052
0.19
•Charge
0.19
0.85
rates assumptions:
0.21
Residential: Classes 1&2 ~ Level0.95
1 (120V;15A)
Classes 3&4 ~ Level 2 (240V;30A)
Non-residential: all classes at Level 2
Noon
*the numbers inside the pie charts express the energy rate in $/kWh
•
Noon
Price of energy: Time Of Use (TOU) pricing
2 seasons
3 price levels: on-peak, medium peak and off-peak
Slide 44
44
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Variables:
C+d,c,v,t= Amount charged on day d, vehicle class
Hourly charge
If positive,
c, vehicle v and time t [kW]
(+) or a
change
in the
discharge
(-)
+W
=
Absolute discharged
value of the difference
d,c,v,t
C
=
Amount
on day d, between
vehicle
direction
d,c,v,t
Lof
=
New
load
on
day
d,
load
type
i
and
hour
t
[kW]
d,i,t
power in the
+
Cclass
and
C+d,c,v,t+1
d,c,v,t
c,
vehicle
v and[kW]
time
t [kW]
Z
=
Absolute
value
of
the
difference
between
Ld,i,t and Lavd,i [kW]
battery d,i,t
Energy
Cd,c,v,t = Energy stored on day d, vehicle class c,
inventory
vehicle v and time t [kWh]
Slide 45
45
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Battery constraints:
Limit the
charge/discharge to the
available connection
Energy in the
battery when the
PHEV arrives
home
Inventory
balance
C+d,c,v, t ≤ Cmaxc for every d, c, v, t
C-d,c,v, t ≤ Cmaxc for every d, c, v, t
Cd,c,v, t = Bc - DEd,c,v for t=Ad,c,v – 1 and every d,c,v
Cd,c,v, t = Cd,c,v, t-1 + C+d,c,v, t - C-d,c,v, t for Ad,c,v ≤ t ≤ Dd,c,v
and every d,c,v
Battery fully charged
by dep. time
Cd,c,v, t = Bc for t=Dd,c,v and every d,c,v
Slide 46
46
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Battery constraints:
-W+d,c,v,t ≤ C+d,c,v, t - C+d,c,v, t+1 ≤ W+d,c,v,t for Ad,c,v ≤ t ≤ Dd,c,v -1
Hour
C+d,c,v,t [kW]
t1
7
t2
7
t3
C-d,c,v,t [kW]
and every+ d, c, v
C+d,c,v,t
W
d,c,v,t
max for A
∑W+d,c,v,t
≤
3*C
c
d,c,v ≤ t ≤ Dd,c,v -1 and every c, v
0
7
0
0
7
7
0
7
0
t4
0
7
0
t5
7
0
7
t6
7
0
7
t7
0
7
0
t8
0
7
0
W+
d,c,v,t?
0
7
0
7
0
Slide 47
47
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Load constraints:
New load
with PHEVs
Peakshaving
measure
C d ,c ,v ,t C d , c , v ,t
v 1
N
L d, i, t L
Base
d ,i ,t
V
- Z d, i, t L d , i , t L d , i Z d , i , t
av
for every d, i, t
for every d, i, t
Slide 48
48
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Objective function:
1.
Utility peak-shaving
Peak shaving Minimize Z d , i , t
d ,i ,t
2.
Energybill
Customer profit
Minimize Pd , t * L d , i , t
d ,i ,t
SOLVE ONE OBJECTIVE AT A TIME AND
COMPARE IMPACT IN RELIABILITY
Slide 49
49
• Impact of PHEVs in distributed resource islands
Results
•
Loading of the RBTS system with PHEVs
RBTS Power demand [kW]
(1) RBTS Base Load
Peak demand [kW]
(2) RBTS Base load + PHEV for peak
shaving
(3) RBTS Base load + PHEV for
customer benefit
(4) RBTS Base load + PHEV
uncontrolled charging & no V2G
Base Load[kW]
Time [h]
Slide 50
50
• Impact of PHEVs in distributed resource islands
Results
•
Loading of the RBTS system with PHEVs
RBTS Power demand [kW]
(1) RBTS Base Load
(2) RBTS Base load + PHEV
uncontrolled charging & no V2G
Peak demand [kW]
(3) RBTS Base load + PHEV
delayed charging & no V2G
Base Load[kW]
Time [h]
Slide 51
51
• Impact of PHEVs for peak-shaving
Results
Some charging before Daily peak demand shifted
base load peak• Individual charging patterns and
demand
daily load with PEAK-SHAVING:
General charging
during the night
Daily base load shifted
Slide 52
• Impact of PHEVs for PEAK-SHAVING versus
52
UNCONTROLLED charging
Results
•Daily
load with PEAK-SHAVING versus UNCONTROLLED charging:
Daily average of
the base load
with no PHEVs
Slide 53
53
• Impact of PHEVs for customer benefit
Daily peak demand shifted
Results
• Individual
charging patterns and
Charging before
base
load peak demand
daily
with TOU PRICING:
Noload
valley-filling
Discharge in the morning
Daily base load shifted
Slide 54
54
• Impact of PHEVs in distributed resource islands
Reliability impact in the RBTS radial system
•
Same annual average loads for RBTS test system with PHEVs optimized for
peak-shaving & benefit of PHEV owner
•
RBTS
Base load
Base load + PHEVs
ENS [MWh]
44.52
47.45
Using step-load duration curve modeling:
6
ENS
Load ( t )
U * PNS ( L t i ) * T i
i 1
Load levels
B
∆Tβ [hours]
1
2
3
4
5
6
3*10-4
ENS [MWh]
Base load + PHEVs
Peak shaving
Base load + PHEVs
Customer benefit
PNS [MW]
∆Tβ [hours]
PNS [MW]
∆Tβ [hours]
PNS [MW]
23.64
20.20
16.76
13.32
9.88
6.45
3*10-4
26.27
21.96
17.64
13.33
9.21
4.71
1
87
1992
2817
2783
1056
27.85
22.86
17.88
12.90
7.92
2.95
Base load
2225
2531
1975
1583
422
39.88
249
2632
3235
1965
654
45.87
39.26
Slide 55
55
• Impact of PHEVs in distributed resource islands
Results
•
Step-load duration curve:
RBTS Power demand [kW]
(1) RBTS Base Load
(2) RBTS Base load + PHEV for peak
shaving
(3) RBTS Base load + PHEV for
customer benefit
(4) RBTS Base load + PHEV
uncontrolled charging & no V2G
Valley filling of
Peak-shaving
Reduce consumption
for customer benefit
Time [h]
Slide 56
56
• Impact of PHEVs in distributed resource islands
Reliability impact in the RBTS with DG + feeder
interties
• ENS reduced in the redesigned RBTS with PHEVs
•
However, the optimal solutions for the base load of the RBTS system
without PHEVs and the cost and reliability are directly influenced by
the demand per load point which has changed
•
MOGA applied to the RBTS with PHEVs
OPTIMAL SOLUTIONS CHANGE?
Slide 57
57
• Impact of PHEVs in distributed resource islands
MOGA applied to the RBTS with PHEVs
• Annual average modeling ~ same for peak-shaving and TOU pricing
Solution #
Connection (s)
DG (s) bus location #
Cost [106 US $]
ENS [MWh]
1
Line 11-17
15
17.60
30.72
2
Line 1-7
5 & 14
Linepoints
11-17i
Load
Base load
Base load + PHEVs
Base load + PHEVs
Line 1-7
5 & 11
peak 17.87
shaving
TOU
pricing
21.31
Residential
1, 4-7, 20-24, 32-36
0.4684
0.4939
0.4940
Residential 4
Line
11, 12
, 13,1-7
18, 25
4 &0.4758
14& 23
18.64
0.5011
21.62
0.5012
0.4606
0.4607
1.0167
1.0167
1.0167
3, 16, 17, 19, 28, 29, 31, 37, 38
0.1889
0.2024
0.2024
Line 11-17
5 & 11 & 19
Customer Type
3
Line 11-17
No PHEVs: 2nd0.4339
highest
Residential
Line
11-17
2,
15, 26,
30
Small Industrial
Line
8, 9,23-29
10
Line 1-7
Commercial
5
Office Buildings
14,27
Line 17-24
0.3345
Annual average load, [MW]
17.70PHEVs: highest
21.55
With
18.33
0.4778
21.00
0.4778
Slide 58
58
• Impact of PHEVs in distributed resource islands
Conclusions in the RBTS test system
• Several assumptions required…
•
Peak demand may be increased and shifted in time
•
Charging patterns for customer benefit (TOU pricing) without
demand charges increase the peak-demand by 25% but
increase the reliability of the system (reduce energy
consumption)
•
Charging pattern for peak-shaving increase the peak demand
by 8% and reduce the reliability (valley filling)
• The
redesign solutions of distribution systems considering
PHEVs may change
Slide 59
• Future work
59
MOGA methodology
• Time dependency on the power output of DG (Stochastic approach)
•
JISEA
project on “Verifiable Decision-Making Algorithms for
Reconfiguration of Electric Microgrids” in collaboration with
University of Colorado-Boulder: Acceleration technique for filtering
potentially infeasible and/or suboptimal inputs, based on Machine
Learning [10]
•
Explore other evolutionary approaches to the redesign problem
10] J. Giráldez, A. Jaintilal , J. Walz, H. E. Brown, S. Suryanarayanan, S. Sankaranarayanan, E. Chang, “An evolutionary algorithm
and acceleration approach for topological design of distributed resource island,” accepted in Proc. 2011 IEEE PES PowerTech,
Trondheim, Norway, Jun 2011.
Slide 60
• Future work
60
Study of PHEVs
•
Develop a study or a survey on how a future vehicle fleet in distributions
systems will look like
• Acquire
PHEV simulation software to run performance, design and
behavioral simulations
•Modeling of
a vehicle battery in the LPs can be extended and more detail on
the operation included
•
Refine the LP algorithms:
peak-shaving: define a new average load, explore dynamic approach
customer benefit: explore other demand response pricing schemes
•
Probabilistic based methodology to model the distribution of PHEVs
throughout the load points of a medium voltage system
Slide 61
• Accomplishments
61
Presentations
J. Giráldez, “A multi-objective genetic algorithmic approach for optimal allocation of
distributed generation and feeder interties considering reliability and cost,” student poster
contest, IEEE PES Power Systems Conference and Exposition, Phoenix, AZ, Mar 2011.
S.
Suryanarayanan, J. Giráldez , S. Rajopadhye, S. Natarajan, S. Sankaranarayanan, E. Chang,
D. Grunwald, J. Walz, A. Jaintilal “Verifiable Decision-Making Algorithms for Reconfiguration
of Electric Microgrids,” poster presentation at JISEA Annual Meeting, Mar. 2011.
Giráldez, “An evolutionary algorithm for planning distributed resource islands,”
presentation, IEEE Powel Electronics Society (PELS) , Colorado School of Mines, Golden CO,
Nov. 2010
J.
Giráldez, S. Suryanarayanan, S. Sankaranarayanan, “Modeling and simulation aspects of
topological design of distributed resource islands,” presentation, Joint Institute for Strategic
Energy Analysis (JISEA), Nat’l Renewable Energy Lab (NREL). [Online] Available
http://www.jisea.org/pdfs/20101214_seminar.pdf (Dec 2010).
J.
Publications
J. Giráldez, A. Jaintilal , J. Walz, H. E. Brown, S. Suryanarayanan, S. Sankaranarayanan, E.
Chang, “An evolutionary algorithm and acceleration approach for topological design of
distributed resource island,” accepted in Proc. 2011 IEEE PES PowerTech, Trondheim, Norway,
Jun 2011.
Chapter
4 is leading to a paper that will be submitted to IEEE International Conference or
Transactions
Slide 62
• Accomplishments
Unique contributions
•
Enhancement of an existing technique (MOGA) for planning
distributed resource islands:
Simultaneous location of DG and feeder interties in a given radial
distribution system
Exploration of
2 ways of modeling the annual load and its effect in the
redesign
Redesign
of distribution systems considering PHEV penetration with
V2G technology:
o methodology to model the behavior of a PHEV fleet as load and as
generation in residential and non-residential demand types
o impact on the reliability of distributed resource islands of different
charging strategies of a PHEV fleet
62
Slide 63
Thank you!
Questions?
Julieta Giráldez
Graduate
Student
Division of
Engineering
CSM
63
PLANNING DISTRIBUTION SYSTEM RESOURCE
ISLANDS CONSIDERING RELIABILITY, COST AND
THE IMPACT OF PENETRATION OF PLUG-IN
HYBRID ELECTRIC VEHICLES
Julieta Giráldez
Graduate
Student
Division of
Engineering
CSM
1
Slide 2
2
• Outline
o
Introduction
o
Design of distributed resource islands
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on distributed resource islands
o
Conclusions and future work
Slide 3
3
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 4
4
• Introduction
Smart Grid Initiative [1]: what is the evolution of
electric power distribution systems?
• Distributed
Energy
Resources
(DER)
or
Distributed
Generation (DG)
• Incorporate ways of physical and virtual storage to balance
consumption and production including PHEVs
• Increased used of technologies: advanced meters, advanced
inverters, distribution automation, communication systems, etc.
[1] 110thCongress of the United States, "Title XIII (Smart Grid)," in Energy Independence and Security Act of 2007. Washington,
DC: Dec. 2007, pp. 292 –303.
Slide 5
5
• Introduction
Why?
• Contribute to the load relief of the transmission system by
increasing the generation in the distribution system and new
ways of energy management
• Higher reliability and power quality
• Integration of green technologies into the grid
Slide 6
6
• Introduction
How to implement the smart grid?
Microgrid concept: a distributed resource island
• Self-contained autonomous subset of the area electric power
system
• Has local Distributed Energy Resources (DER)
• Operates semi-autonomously of the grid, being able to island and
reconnect as circumstances dictate
• Able to provide power quality and reliability different from
general macro-grid standards
[2] N.Hatziargyriou, H.Asano, R.Iravani, and C.Marnay , “Microgrids”, IEEE Power & Energy Magazine, pp.78-94, July/Aug.
2007.
Slide 7
[3] “Distributed Energy Resources Integration”, Consortium for Electric Reliability Technology Solutions (CERTS), [Online].
Available:http://certs.lbl.gov/certs-der.html
7
Slide 8
8
• Outline
o
Introduction
o
Design of distributed resource islands
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 9
9
• Design of distributed resource islands
Distribution systems
•Traditional
electric distribution systems:
Infinite bus
Transformer
Line
Grid
Grid
Load
Slide 10
10
• Design of distributed resource islands
Distribution systems
•Evolving
distribution systems:
Grid
Grid
Increase annual RELIABILITY at a feasible COST
Slide 11
11
• Design of distributed resource islands
Modeling of annual load
• Annual demand: two ways of modeling annual load
annual average demand at every load: i.e. 1 load level
representative of the annual demand
6 step-load duration curve representation (hourly demand
∆T2=1900h
reordered in increasing demand): i.e. 6 load levels
representative of the annual demand
∆T1 =100 h
[4] R. Billinton, S. Kumar, et al., "A Reliability Test System for Educational Purposes - Basic Data," IEEE Transactions on Power
Systems, vol. 4, pp. 1238-1244, August 1989.
Slide 12
12
• Design of distributed resource islands
Modeling of DG
• DG: aggregate power output of Renewable Energy (RE)
and Conventional Distributed Generation (CDG) [5]
Pout = CDG + RE + DS
Pout = R CDG * CF CDG
R RE * CF RE
(1 CF RE ) * 0 . 15 * L i
Capacity Factor: ratio of the actual output of a power
source and its output if it had operated at full capacity
Total DG rating R=RRE + RCDG
[5] H. Brown, “Implications on the Smart Grid Initiative on Distribution System Engineering: Improving Reliability on Islanded
Distribution Systems with Distributed Generation Sources, M.S thesis, Dept. Elec. Eng., Colorado School of Mines, 2010.
Slide 13
13
• Design of distributed resource islands
Basic Reliability Concepts:
• ASAI: The time as a fraction of a year for which the system is
available
• Annual Outage Time, U : Time as a fraction of a year for which
the system is NOT available ~ U 1 ASAI
• Power Not Supplied (PNS) [MW]: Unserved load or demand
that the system cannot attend
• Reliability metric: Energy Not Supplied [MWh]
8760
ENS U
PNS ( h )
h 1
Slide 14
14
• Design of distributed resource islands
I. Enter the power system component data
Slack
bustool:
is modeled as a
Powerbus:
systemsslack
simulation
generator
thatprogram
absorbs
supplies
• Computer
to solveor
a power
flow: generation
supplies
demand, and
to control
the frequency of the
in order1.toGeneration
balance
thethe load
generation
system
~ Power Not Supplied~
Bus
magnitudes
remainthree
closephase
to the
rated values
II.2.Solve
thevoltage
Power Flow
under balanced
conditions
2.3.
1.
Lines and transformers are not overloaded
• PowerWorld SimulatorTM is used
3.
Slide 15
15
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 16
16
• MOGA
Multi-objective redesign problem
•
Investment cost versus reliability: Pareto-optimality
~ no single optimal solutions but a set of alternative solutions ~
•
Non-linear problem, discrete and non-convex feasible region
•
Intractability of the problem as the size of the system grows [5]
•
Evolutionary methods
[5] H. Brown, “Implications on the Smart Grid Initiative on Distribution System Engineering: Improving Reliability on Islanded
Distribution Systems with Distributed Generation Sources, M.S thesis, Dept. Elec. Eng., Colorado School of Mines, 2010.
Slide 17
17
• MOGA
Mathematical formulation
•
Variables
•
Objective function 1: COST
Ng
Nc
f1 ( x )
C
i 1
CC :Cost of conductor [$/km]
li : Length of connection i [km]
l x
c c
i
c
i
C
DG
P
DG
j
g
j
(x )
j 1
CDG: Cost of DG [$/MW]
PDGj: Power output of DGlocated at bus j [MW]
Slide 18
18
• MOGA
Mathematical formulation
•
Objective function 2: RELIABILITY ~ Energy Not Supplied
8760
ENS f 2 ( x ) U
•
h 1
Annual average loads:
f 2 , Average
•Six
_ Loads
( x ) 8760 * U * PNS
step load duration curve:
6
f 2 , Load
PNS ( h )
(t )
U * PNS ( L Ti ) * Ti , with
i 1
6
T
i
i
8760 h
Slide 19
19
• MOGA
Mathematical formulation
•
Constraints
Voltage
within 5% of the nominal value at every bus j:
0 . 95 V j ( x ) 1 . 05
Loading of the Line from bus j to k:
S jk 1 . 00
Slide 20
20
• MOGA
*GAs
*
•
and the fitness function
A population is comprised of individuals or chromosomes ~
a potential
solution to the optimization problem
*
•
Evolutionary operators are used to create randomly individuals which may
* move to a higher level of fitness such as mutation, recombination, and
crossover.
•
MatlabTM Genetic Algorithm Optimization Toolbox (GAOT) inbuilt
functions
• The
fitness function determines how likely an individual is to survive to the
next generation ~ output of fitness function ~
f1 ( x )
f
f 2 ( x)
Slide 21
21
• MOGA
Importance of the initial population for convergence
• Explore 3 ways of selecting the initial population
Slide 22
22
• MOGA: RBTS test system
Application to a test system
•
RBTS System [6]:
Customer Type
Load points i
Average Load,
Max. Peak Load,
[MW]only the 164
[MW]
Possible Connections 302. We input
connections
Residential
1, 4-7, 20-24, 32-36
0.4684
which length is less
than 3km.
Residential
11, 12, 13, 18, 25
0.8367
0.4758
0.8500
0.4339
0.7750
Possible DG Location 27 buses
2, 15, 26, 30
•DG: ResidentialDESIGN
PARAMETERS
0.8472 0.25, 0.3,
1.0167
0.8
8, 9, 10
CF: Small
wind, solar, conventional
Industrial
DG
Commercial
16, 17, 19, 28, 29, 31,
Total DG 3,penetration
80%0.2886
Total Annual0.5222
Average Load
37, 38
RE penetration
20% Total DG penetration
Office
14,27
0.5680
0.9250
[6] R. Billinton and S.Buildings
Jonnavithula, "A Test System for Teaching Overall Power System Reliability assessment," IEEE Transactions on Power
Systems, vol. 11, pp. 1670-1676, November 1996.
Slide 23
Solution #
Connection (s)
DG (s) bus location #
1
2
Line 11-17
Line 1-7
15
4 & 15
Cost
[106
23
US $]
• MOGA: RBTS Test system
ENS [MWh]
17.97
18.06
31.40
21.96
Line 11-17
4 & 15
25 decision maker
18.12
Results:
“look-up table”
for&the
21.88
Application
to a test system
Line 11-17
Line 1-7
3
•
Line 23-29
• A more
4
expensive solution may be chosen if the Value of Lost Load
Line 11-17
13 & 29
18.31
21.62
7 & 13 & 23
18.47
21.36
11 & 13 & 23
18.91
21.33
(VOLL) [$] of the system is greater than the investment cost
Line 23-29
Line 11-17
5
Line 17-23
Line 23-29
Line 1-7
6
Line 9-15
Line 21-27
Slide 24
24
• MOGA: RBTS Test system
If VOLL ≤ Cost Solution 6 might be chosen
Slide 25
25
• MOGA: RBTS Test system
Application to a test system
Time [h]
•Very
40
similar redesign solutions for the RBTS with annual average loads
and
35 with step-load duration curve
• 30
ENS
overestimated with annual average demand
25
•
Computational time :
20
15
10
5
0
modeling of the annual load
connection from Matlab to PowerWorld Simulator
initial(i) population
RBTS Case I:Annual average loads
(ii)
(iii)
Initial population type
RBTS Case II: Step-load duration curve
Slide 26
26
• Outline
o
Introduction
o
Optimization of islanded distribution systems
from a design perspective
o
Multi-Objective Genetic Algorithm (MOGA)
o
Impact of Plug-in Hybrid Electric Vehicles
(PHEVs) on an electric distributed island
o
Conclusions and future work
Slide 27
27
• Impact of PHEVs in distributed resource islands
Introduction to PHEVs
•
IEEE definition: “vehicles that have a battery storage system rating of
4 kWh or more, a means of recharging the battery form an external
source, and the ability to drive at least 10 miles in all electric mode”
•
Vehicle-to-grid (V2G): using the battery of a vehicle as a Distributed
Energy Resource (DER)
•
New way of electric energy management
•
Existing power system infrastructure may not be adequate to deal with
the increased demand and new patterns of consumption and power
flows in the grid
[7] “Vehicle to Grid (V2G) Electricity” , Global Greenhouse Warming, [Online]. Available: http://www.global-greenhousewarming.com/vehicle-to-grid.html
Slide 28
28
• Impact of PHEVs in distributed resource islands
Modeling PHEVs in distribution systems
•
How many PHEVs?
•What
is the behavior of the driver?
•
For how long does a PHEV behave as a load?
•
For how long does a PHEV behave as DG?
~ KEY ASSUMPTIONS TO
STUDY THE IMPACT ~
Slide 29
29
• Impact of PHEVs in distributed resource islands
Modeling PHEVs in a distribution system
•How
•What
manykind
vehicles?
of
PHEVs?
• For how long does
• What design and
the PHEV behaves as
operational
a load?
•How
characteristics?
many PHEVs in
• … and as a
• What
theissystem?
the behavior
generator?
of the driver?
• Electric customer consumes
• Driving
2 kW
and has 1.5factors
vehicles for
residential;
38 workers per
1. Peak-shaving
office building and 17 workers
2.per• commercial
Owner’s
Probabilistic
benefit
and 1.5
•Linear
simulation
Programming
vehicles
per worker
(LP)
methodology
algorithms to
• optimize
30% penetration
of the
charging
total transportation
patterns fleet
Slide 30
Probabilistic simulation of PHEV fleet for 8760 hours [8]
30
PHEV Class
Class 4 islands
PHEV Class
PHEV Class 3 PHEV
• Impact
of 1PHEVs
in2 distributed
resource
Tools &
methods
Daily vehicle data for optimization
Energy
Miles driven
Arrival time
required
Methodology
Departure
time
•
Probabilistic simulation methodology [8]
•
LP Optimization of daily charging pattern of PHEVs for 1 year
Objective/s: maximize owners profit and/or utility peak shaving
(Demand response)
Contributions made by this thesis:
LP algorithm
~ determine
Incorporate
optimized PHEV
load (hourly) tothe
loadloading
duration
Impact
curve of distribution system
on
design
Impact of PHEV
fleet on annualresource
distributed
reliability of islanded legacy radial
distribution systems
and
reliability
of
Impact of PHEV fleet on annual
islands
reliability of islanded networked
distribution systems
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖
PSERC Document 09-12, Oct. 2009.
Results
Slide 31
31
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Four vehicle classes (types)
•
Design characteristics (SOC):
Vehicle class c
1
2
3
4
Battery size
Bc [kWh]
Max
12
14
21
23
Min
8
10
17
19
[8] S. Meliopoulos, J. Meisel and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖
PSERC Document 09-12, Oct. 2009.
Slide 32
32
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Amount of driving supplied from electric battery? From fuel?
•
kphev=0 represents a charge sustaining (CS) mode in which on
average all the drive energy comes from gasoline
•
kphev=1 represents a charge depleting (CD) mode, all of the drive
energy comes from electricity
•
•
Simulations run in Powerdrive Simulation Analysis Tool (PSAT)
Performance parameter Ec: required energy per mile [kWh/mi.]
E
bc
E
c
c
c
E a ( kphev )
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 33
33
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
• Vehicle control strategy: drive in CD from battery while in
SOC ranges and switch to CS to maintain SOC relying on gas
•
M
D
Charge depleting distance MD
Vehicle
class c
Bc
Ec
M
D
40 miles
E c a ( kphev c )
E
c
E
bc
kphevc
Max
min
1
0.5976
0.2447
2
0.6151
0.2750
3
0.5428
0.3217
4
0.4800
0.3224
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 34
34
• Impact of PHEVs in distributed resource islands
Parameters for the Prob. Sim. Methodology [8]
•
Four random paramaters
1.
kphevc and Bc
2.
# Vehicles per class
3.
Daily Miles driven per vehicle
4.
Driver’s behavior ~ Time parameters
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 35
35
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
1.
Vehicle design characteristics kphevc and usable battery capacity Bc
are distributed according to a bivariate normal distribution with mean
vector μ and covariance matrix ∑ with 0.8 parameter correlation
•
Performance
parameter BC
Ec is[kWh]
determined knowingkphev
kphevc
Vehicle
c
class c
E c a ( kphev )
E
E
14.3015
c
14.1827
bc
0.5976
c
0.6151
3
19.1516
0.5428
4
21.3211
0.4800
1
2
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 36
36
• Impact of PHEVs in distributed resource islands
Probabilistic Sim. Meth. applied to the RBTS test system[8]
2.
Vehicles per class: normal distribution with mean #PHEVs*Probability vehicle
class and 1% standard deviation
• Total #
•
vehicles (light transportation fleet): 15, 269 = 14,925 res + 230 com+ 114 off
Uniform distribution of the #PHEVs throughout the load points of the RBTS per
demand type~ daily parameters generated only for the #PHEVs in one load type~
• Vehicle
population size per class:
Vehicle
Vehicles per load point type
class
• Approximate
c
Residential
Commercial
Office building
1
44
15
7
2
48
18
9
3
45
14
6
4
49
19
8
to the average #PHEV per class per load type
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 37
37
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
3.
Miles driven per vehicle per day Md,c,v : log normal
distribution with mean 3.37 and standard deviation of 0.5
•
Daily energy required per vehicle from the grid [kWh]:
DE
d ,c ,v
, if MD ≤ Md,c,v
Bc
D
M
*
E
,
if
M
≤
M
c
d,c,v
d ,c ,v
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC Document 0912, Oct. 2009.
Slide 38
38
• Impact of PHEVs in distributed resource islands
Probabilistic simulation methodology [8]
4.
Driver’s behavior ~ time parameters: Gaussian distribution
Departure (am)
•
Arrival (pm)
Parameter
Weekday
Weekend
Weekday
Weekend
μc
σc
7
1.73
9
2.45
6
1.73
15
2.45
Only residential charging in [8], what about office and commercial loads?
Dep
Arr
non res
d ,c ,v
non res
d ,c ,v
Arr
Dep
res
d ,c ,v
res
d ,c ,v
M
d ,c ,v
S
M
d ,c ,v
S
average urban driving speed
25 [mi./h]
[8] S. Meliopoulos, J. Meiselrge and T. Overbye, ―Power System Level Impacts of Plug-In Hybrid Vehicles (Final Project Report),‖ PSERC
Document 09-12, Oct. 2009.
Slide 39
39
• Impact of PHEVs in distributed resource islands
LP algorithms
•
By now we know:
Size and design characteristics of the PHEV fleet
Daily energy required per vehicle from the grid
Daily available time for charging per vehicle
DETERMINE DAILY CHARGING
PATTERNS: Utility peak shaving or benefit
of the owner
Slide 40
40
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
•
Sets:
I = set of load types , from 1 … NI
C= set of PHEV classes, from 1…NC
V= set of PHEVs per class, from 1 … NV
D=set of days in a year, from 1…ND
T= set of hours in a day, from 1 … NT
Slide 41
41
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
•
Parameters:
B
c = Battery size per vehicle class c [kWh]
max
C c = Maximum hourly charge rate per vehicle class c [kW]
DE
d,c,v = Daily energy required per day d, vehicle class c and vehicle v
base
L d,i,t = Base load (without PHEVs) on day d, load type i and hour t [kW]
[kWh]
Lavd,i = Average base load (without PHEVs) on day d and load type i [kW]
Ad,c,v = Daily arrival time per day d, vehicle class c and vehicle v [h]
Pd,t = Price of energy on day d and hour t [$/kWh]
Dd,c,v = Daily departure time per day d, vehicle class c and vehicle v [h]
From the Probabilistic Simulation Methodology
Slide 42
[9] Reliability Test System Task Force of the Application of Probability Methods Subcommittee, “IEEE reliability test system,” IEEE
Transactions on Power Apparatus and Systems, vol. PAS-98, no. 6, pp. 2047-54, November 1979.
42
• Impact of PHEVs in distributed resource islands
Day system % Annual Peak Load
Application to the RBTS test
• Base load of the system:
Customer
Load points i
Max. Annuak
Type
Peak Load,
Residential
[MW]
0.8367
Residential
1, 4-7, 20-24, 3236
11, 12, 13, 18, 25
Week
0.8500
Residential
2, 15, 26, 30
0.7750
Small
Industrial
8, 9, 10
1.0167
3, 16, 17, 19, 28,
29, 31, 37, 38
0.5222
14,27
0.9250
Commercial
Office
Buildings
1
2
3
4
5
6
7
8
9
10
11
12
13
Peak
Load
82.2
90
87.8
83.4
88
84.1
83.2
80.6
74
73.7
71.5
72.7
70.4
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday
Week
Peak
14
15
16
17
18
19
20
21
22
23
24
25
26
Load
75
72.1
80
75.4
83.7
87
88
85.6
81.1
90
88.9
89.6
86.1
93
100
98
96
94
77
75
Week
Peak
27
28
29
30
31
32
33
34
35
36
37
38
39
Load
75.5
81.6
80.1
88
72.2
77.6
80
72.9
72.6
70.5
78
69.5
72.4
Week
Peak
40
41
42
43
41
45
46
47
48
49
50
51
52
Load
72.4
74.3
74.4
80
88.1
88.5
90.9
94
89
94.2
97
100
95.2
Slide 43
43
• Impact of PHEVs in distributed resource islands
Midnight
Midnight
0.052
Application to the RBTS test system
0.052
0.19
•Charge
0.19
0.85
rates assumptions:
0.21
Residential: Classes 1&2 ~ Level0.95
1 (120V;15A)
Classes 3&4 ~ Level 2 (240V;30A)
Non-residential: all classes at Level 2
Noon
*the numbers inside the pie charts express the energy rate in $/kWh
•
Noon
Price of energy: Time Of Use (TOU) pricing
2 seasons
3 price levels: on-peak, medium peak and off-peak
Slide 44
44
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Variables:
C+d,c,v,t= Amount charged on day d, vehicle class
Hourly charge
If positive,
c, vehicle v and time t [kW]
(+) or a
change
in the
discharge
(-)
+W
=
Absolute discharged
value of the difference
d,c,v,t
C
=
Amount
on day d, between
vehicle
direction
d,c,v,t
Lof
=
New
load
on
day
d,
load
type
i
and
hour
t
[kW]
d,i,t
power in the
+
Cclass
and
C+d,c,v,t+1
d,c,v,t
c,
vehicle
v and[kW]
time
t [kW]
Z
=
Absolute
value
of
the
difference
between
Ld,i,t and Lavd,i [kW]
battery d,i,t
Energy
Cd,c,v,t = Energy stored on day d, vehicle class c,
inventory
vehicle v and time t [kWh]
Slide 45
45
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Battery constraints:
Limit the
charge/discharge to the
available connection
Energy in the
battery when the
PHEV arrives
home
Inventory
balance
C+d,c,v, t ≤ Cmaxc for every d, c, v, t
C-d,c,v, t ≤ Cmaxc for every d, c, v, t
Cd,c,v, t = Bc - DEd,c,v for t=Ad,c,v – 1 and every d,c,v
Cd,c,v, t = Cd,c,v, t-1 + C+d,c,v, t - C-d,c,v, t for Ad,c,v ≤ t ≤ Dd,c,v
and every d,c,v
Battery fully charged
by dep. time
Cd,c,v, t = Bc for t=Dd,c,v and every d,c,v
Slide 46
46
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Battery constraints:
-W+d,c,v,t ≤ C+d,c,v, t - C+d,c,v, t+1 ≤ W+d,c,v,t for Ad,c,v ≤ t ≤ Dd,c,v -1
Hour
C+d,c,v,t [kW]
t1
7
t2
7
t3
C-d,c,v,t [kW]
and every+ d, c, v
C+d,c,v,t
W
d,c,v,t
max for A
∑W+d,c,v,t
≤
3*C
c
d,c,v ≤ t ≤ Dd,c,v -1 and every c, v
0
7
0
0
7
7
0
7
0
t4
0
7
0
t5
7
0
7
t6
7
0
7
t7
0
7
0
t8
0
7
0
W+
d,c,v,t?
0
7
0
7
0
Slide 47
47
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Load constraints:
New load
with PHEVs
Peakshaving
measure
C d ,c ,v ,t C d , c , v ,t
v 1
N
L d, i, t L
Base
d ,i ,t
V
- Z d, i, t L d , i , t L d , i Z d , i , t
av
for every d, i, t
for every d, i, t
Slide 48
48
• Impact of PHEVs in distributed resource islands
Mathematical formulation of the LPs
• Objective function:
1.
Utility peak-shaving
Peak shaving Minimize Z d , i , t
d ,i ,t
2.
Energybill
Customer profit
Minimize Pd , t * L d , i , t
d ,i ,t
SOLVE ONE OBJECTIVE AT A TIME AND
COMPARE IMPACT IN RELIABILITY
Slide 49
49
• Impact of PHEVs in distributed resource islands
Results
•
Loading of the RBTS system with PHEVs
RBTS Power demand [kW]
(1) RBTS Base Load
Peak demand [kW]
(2) RBTS Base load + PHEV for peak
shaving
(3) RBTS Base load + PHEV for
customer benefit
(4) RBTS Base load + PHEV
uncontrolled charging & no V2G
Base Load[kW]
Time [h]
Slide 50
50
• Impact of PHEVs in distributed resource islands
Results
•
Loading of the RBTS system with PHEVs
RBTS Power demand [kW]
(1) RBTS Base Load
(2) RBTS Base load + PHEV
uncontrolled charging & no V2G
Peak demand [kW]
(3) RBTS Base load + PHEV
delayed charging & no V2G
Base Load[kW]
Time [h]
Slide 51
51
• Impact of PHEVs for peak-shaving
Results
Some charging before Daily peak demand shifted
base load peak• Individual charging patterns and
demand
daily load with PEAK-SHAVING:
General charging
during the night
Daily base load shifted
Slide 52
• Impact of PHEVs for PEAK-SHAVING versus
52
UNCONTROLLED charging
Results
•Daily
load with PEAK-SHAVING versus UNCONTROLLED charging:
Daily average of
the base load
with no PHEVs
Slide 53
53
• Impact of PHEVs for customer benefit
Daily peak demand shifted
Results
• Individual
charging patterns and
Charging before
base
load peak demand
daily
with TOU PRICING:
Noload
valley-filling
Discharge in the morning
Daily base load shifted
Slide 54
54
• Impact of PHEVs in distributed resource islands
Reliability impact in the RBTS radial system
•
Same annual average loads for RBTS test system with PHEVs optimized for
peak-shaving & benefit of PHEV owner
•
RBTS
Base load
Base load + PHEVs
ENS [MWh]
44.52
47.45
Using step-load duration curve modeling:
6
ENS
Load ( t )
U * PNS ( L t i ) * T i
i 1
Load levels
B
∆Tβ [hours]
1
2
3
4
5
6
3*10-4
ENS [MWh]
Base load + PHEVs
Peak shaving
Base load + PHEVs
Customer benefit
PNS [MW]
∆Tβ [hours]
PNS [MW]
∆Tβ [hours]
PNS [MW]
23.64
20.20
16.76
13.32
9.88
6.45
3*10-4
26.27
21.96
17.64
13.33
9.21
4.71
1
87
1992
2817
2783
1056
27.85
22.86
17.88
12.90
7.92
2.95
Base load
2225
2531
1975
1583
422
39.88
249
2632
3235
1965
654
45.87
39.26
Slide 55
55
• Impact of PHEVs in distributed resource islands
Results
•
Step-load duration curve:
RBTS Power demand [kW]
(1) RBTS Base Load
(2) RBTS Base load + PHEV for peak
shaving
(3) RBTS Base load + PHEV for
customer benefit
(4) RBTS Base load + PHEV
uncontrolled charging & no V2G
Valley filling of
Peak-shaving
Reduce consumption
for customer benefit
Time [h]
Slide 56
56
• Impact of PHEVs in distributed resource islands
Reliability impact in the RBTS with DG + feeder
interties
• ENS reduced in the redesigned RBTS with PHEVs
•
However, the optimal solutions for the base load of the RBTS system
without PHEVs and the cost and reliability are directly influenced by
the demand per load point which has changed
•
MOGA applied to the RBTS with PHEVs
OPTIMAL SOLUTIONS CHANGE?
Slide 57
57
• Impact of PHEVs in distributed resource islands
MOGA applied to the RBTS with PHEVs
• Annual average modeling ~ same for peak-shaving and TOU pricing
Solution #
Connection (s)
DG (s) bus location #
Cost [106 US $]
ENS [MWh]
1
Line 11-17
15
17.60
30.72
2
Line 1-7
5 & 14
Linepoints
11-17i
Load
Base load
Base load + PHEVs
Base load + PHEVs
Line 1-7
5 & 11
peak 17.87
shaving
TOU
pricing
21.31
Residential
1, 4-7, 20-24, 32-36
0.4684
0.4939
0.4940
Residential 4
Line
11, 12
, 13,1-7
18, 25
4 &0.4758
14& 23
18.64
0.5011
21.62
0.5012
0.4606
0.4607
1.0167
1.0167
1.0167
3, 16, 17, 19, 28, 29, 31, 37, 38
0.1889
0.2024
0.2024
Line 11-17
5 & 11 & 19
Customer Type
3
Line 11-17
No PHEVs: 2nd0.4339
highest
Residential
Line
11-17
2,
15, 26,
30
Small Industrial
Line
8, 9,23-29
10
Line 1-7
Commercial
5
Office Buildings
14,27
Line 17-24
0.3345
Annual average load, [MW]
17.70PHEVs: highest
21.55
With
18.33
0.4778
21.00
0.4778
Slide 58
58
• Impact of PHEVs in distributed resource islands
Conclusions in the RBTS test system
• Several assumptions required…
•
Peak demand may be increased and shifted in time
•
Charging patterns for customer benefit (TOU pricing) without
demand charges increase the peak-demand by 25% but
increase the reliability of the system (reduce energy
consumption)
•
Charging pattern for peak-shaving increase the peak demand
by 8% and reduce the reliability (valley filling)
• The
redesign solutions of distribution systems considering
PHEVs may change
Slide 59
• Future work
59
MOGA methodology
• Time dependency on the power output of DG (Stochastic approach)
•
JISEA
project on “Verifiable Decision-Making Algorithms for
Reconfiguration of Electric Microgrids” in collaboration with
University of Colorado-Boulder: Acceleration technique for filtering
potentially infeasible and/or suboptimal inputs, based on Machine
Learning [10]
•
Explore other evolutionary approaches to the redesign problem
10] J. Giráldez, A. Jaintilal , J. Walz, H. E. Brown, S. Suryanarayanan, S. Sankaranarayanan, E. Chang, “An evolutionary algorithm
and acceleration approach for topological design of distributed resource island,” accepted in Proc. 2011 IEEE PES PowerTech,
Trondheim, Norway, Jun 2011.
Slide 60
• Future work
60
Study of PHEVs
•
Develop a study or a survey on how a future vehicle fleet in distributions
systems will look like
• Acquire
PHEV simulation software to run performance, design and
behavioral simulations
•Modeling of
a vehicle battery in the LPs can be extended and more detail on
the operation included
•
Refine the LP algorithms:
peak-shaving: define a new average load, explore dynamic approach
customer benefit: explore other demand response pricing schemes
•
Probabilistic based methodology to model the distribution of PHEVs
throughout the load points of a medium voltage system
Slide 61
• Accomplishments
61
Presentations
J. Giráldez, “A multi-objective genetic algorithmic approach for optimal allocation of
distributed generation and feeder interties considering reliability and cost,” student poster
contest, IEEE PES Power Systems Conference and Exposition, Phoenix, AZ, Mar 2011.
S.
Suryanarayanan, J. Giráldez , S. Rajopadhye, S. Natarajan, S. Sankaranarayanan, E. Chang,
D. Grunwald, J. Walz, A. Jaintilal “Verifiable Decision-Making Algorithms for Reconfiguration
of Electric Microgrids,” poster presentation at JISEA Annual Meeting, Mar. 2011.
Giráldez, “An evolutionary algorithm for planning distributed resource islands,”
presentation, IEEE Powel Electronics Society (PELS) , Colorado School of Mines, Golden CO,
Nov. 2010
J.
Giráldez, S. Suryanarayanan, S. Sankaranarayanan, “Modeling and simulation aspects of
topological design of distributed resource islands,” presentation, Joint Institute for Strategic
Energy Analysis (JISEA), Nat’l Renewable Energy Lab (NREL). [Online] Available
http://www.jisea.org/pdfs/20101214_seminar.pdf (Dec 2010).
J.
Publications
J. Giráldez, A. Jaintilal , J. Walz, H. E. Brown, S. Suryanarayanan, S. Sankaranarayanan, E.
Chang, “An evolutionary algorithm and acceleration approach for topological design of
distributed resource island,” accepted in Proc. 2011 IEEE PES PowerTech, Trondheim, Norway,
Jun 2011.
Chapter
4 is leading to a paper that will be submitted to IEEE International Conference or
Transactions
Slide 62
• Accomplishments
Unique contributions
•
Enhancement of an existing technique (MOGA) for planning
distributed resource islands:
Simultaneous location of DG and feeder interties in a given radial
distribution system
Exploration of
2 ways of modeling the annual load and its effect in the
redesign
Redesign
of distribution systems considering PHEV penetration with
V2G technology:
o methodology to model the behavior of a PHEV fleet as load and as
generation in residential and non-residential demand types
o impact on the reliability of distributed resource islands of different
charging strategies of a PHEV fleet
62
Slide 63
Thank you!
Questions?
Julieta Giráldez
Graduate
Student
Division of
Engineering
CSM
63