Sie sind auf Seite 1von 8

1

Optimum Sizing of Energy Storage for an


Electric Ferry Ship
Salman Mashayekh, Student Member, IEEE, Zhenyuan Wang, Lisa Qi,
John Lindtjorn, and Tor-Arne Myklebust

Fuel oil consumption


Abstract -- Each vessel has a different load profile based on its Steady state specific fuel oil consumption
application. Among the vessels, the ones that have electrical Transient specific fuel oil consumption
propulsion systems with high fluctuations of propulsion demand Electricity price at the harbor
power incur poor fuel consumption efficiencies due to: a) low Number of trips per year
loading of the generators and the lower efficiency in lower loads,
Saved fuel per trip
and b) variations in the output power of the generators.
Therefore, it may be economically feasible to install energy Fuel price in the first year
storage units in such ships to improve generator efficiencies. Fuel cost increase rate per year
Given the load profile of a vessel, this paper formulates a Yearly interest rate
simple but general deterministic dynamic optimization problem Number of capacitor modules
to find the optimal loading strategy for the ship generators in the Capacitor cost per module
presence of an energy storage unit. It then adds an outer layer to Capacitor area cost per module
the optimization problem that finds the optimum capacity for the Capacity of each capacitor module
energy storage.
II. INTRODUCTION
Index Terms – Ferry Ship, Fuel Consumption Minimization,
Energy Storage, Supper Capacitor, Battery, Dynamic
Optimization
E ACH vessel has a different load profile. Depending on the
application, some of the ships have high fluctuating loads,
which is mainly due to high fluctuations in propulsion
demand. These high variations are problematic in electric
I. NOMENCLATURE ships because: a) in light power demands, ship generators do
Number of time steps not work efficiently because of being lightly loaded, and b)
Number of generators frequent variations in generator output power increases fuel
( ) Generation cost of the -th generator at time step n consumption and depreciates fuel consumption efficiency.
( ) Percentage loading of the -th generator Therefore, it may be advantageous to install energy storage
( ) Startup cost of the -th generator at time step n units in such ships to shape the load. This way, high frequency
∗ load variations are covered with the energy storage and system
Fixed startup cost of the -th generator
, , Cost function coefficients for the -th generator efficiency increases due to reducing generator power
Rated power of the -th generator variations.
Transient penalty factor for the -th generator The work presented in this article is mainly founded on a
Time step research reported in [1],[2]. This research targets the problem
Maximum allowable ramp rate of the -th generator of finding the optimum super capacitor size in vessels using
( ) Total load power in the -th time step VSI type VFD to control propulsion/thruster motors. The
( ) Energy of the energy storage unit at the end of the super capacitors in this work are used to provide an extra dc-
-th time step link energy buffer for the VFD, thereby reducing the power
, Minimum and maximum allowable powers for the fluctuations seen by the power plant. This work includes
-th generator detailed modeling and constraints associated with this
, Minimum and maximum allowable energy for the particular ship electric system architecture. However, it does
energy storage unit not explore the problem of finding the optimal generation
dispatch in the presence of energy storage.
Another research on energy storage utilization in shipboard
S. Mashayekh is with the Department of Electrical and Computer
Engineering, Texas A&M University, College Station, TX 77840 USA (email: power systems is reported in [3]. This paper formulates a
s.mashayekh@tamu.edu). discrete-time Markov Decision Process (MDP) for the
Zhenyuan Wang and Lisa Qi are with Corporate Research Center, ABB problem of optimal power generation scheduling both with
Inc., Raleigh, NC 27518 USA (email: zhenyuan.wang@us.abb.com,
lisa.qi@us.abb.com) and without energy storage. However, it does not explain how
John Lindtjorn is with Marine, ABB AS., Billingstad, Norway (email: the optimum energy storage size can be determined.
john.o.lindtjorn@no.abb.com) To cover the gap in literature, this paper solves the problem
Tor-Arne Myklebust is with Marine, ABB AS., Alesund, Norway
(email: tor-arne.mykelbust@no.abb.com) of finding the optimum energy storage size by formulating a

978-1-4673-2729-9/12/$31.00 ©2012 IEEE


2

two-layer optimization problem in which the inner layer finds IV. DYNAMIC DISPATCH: PROBLEM FORMULATION
the optimal power generation scheduling for a particular In this work, a very simple power system model is
energy storage capacity and the outer layer goes over all considered for the ship. As shown in Fig. 2, all of the system
possible designs (storages capacities) and determines the net components are connected to a common bus. This system can
saving (saving minus cost) for each design. Finally, the design generally be composed of several generators with various
with maximum net saving is chosen as the optimum design. sizes, 1 energy storage unit, and 1 lumped load which
The remaining of this paper is organized as follows: represents all system loads plus losses.
chapter III introduces the 2-lyaered optimization approach
used in this work. In the next chapter, the dynamic dispatch
problem for the inner layer optimization is formulated. An G G ... G
Energy
Storage
example case is introduced in the next chapter and then,
chapter VI presents the solution methodology. Then, chapter
VII shows how the results can improve by separating the unit M Loss
commitment problem from the optimal dispatch problem. The
next chapter, chapter VIII, formulates the outer loop. Finally, Total System Load

the paper is summarized and concluded in chapter IX. Fig. 2. Simplified shipboard power system model

III. TWO-LAYERED OPTIMIZATION APPROACH


A. Objective Function
It is desired to determine the optimum capacity for the
The objective of the (inner loop) optimization is to
energy storage in a ship, given its load profile. To achieve this
minimize total generation cost over the whole study period.
goal: the optimal control strategy for the generators and the
System cost can be either in terms of money or in terms of
installed energy storage capacity needs to be determined, first
fuel. The operation cost of the system, shown in (1), is
(inner layer). Then, considering the maximum possible saving
composed of 3 terms: 1) generation fuel consumption, 2)
for each energy storage capacity (which is associated with the
startup cost, and 3) cost of initial charge for the energy
optimal control strategy for that capacity), the best design will
storage. Note that the energy storage unit is charged to
be chosen (outer layer).
level when the ship is at harbor. This initial charge is also a
Fig. 1 shows the 2-layer optimization algorithm developed
decision variable and costs × .
for this problem. In this approach an outer loop goes over all
possible energy storage capacities (all possible designs). Then,
for each capacity, a dynamic optimization problem is min = Δ × ( )+ ( ) + × (1)
formulated and solved which determines the optimal control
strategy for the installed capacity. Having the optimal control It is common practice to approximate the steady state fuel
strategy, the maximum possible savings for that particular consumption characteristic of a generator as a second order
capacity can be calculated. Finally, the maximum possible [4],[5] or a third order [1],[2] polynomial function of the
saving minus the investment cost is compared for all of the output power. Moreover, to take into consideration extra fuel
designs and the best one is chosen. Note that the grey blocks consumption needed to change the generator output level,
in this figure represent the inner loop dynamic optimization some references add a transition penalty to the steady state
problem and the white blocks form the outer loop. fuel consumption [1],[2]. Thus, as shown in the following
equation, a second order polynomial fuel consumption with
some transient penalty factor is considered for the generators.
( )= ( ) × ( ) + ( )+
( ) − ( − 1) (2)
× 1+
Δ

Bringing the generator online from the offline mode


requires some extra fuel. As a result, a startup cost is usually
considered for off to on transition. Some references consider
the startup cost to be a function of down time [5] and some
other references assume a fixed startup cost regardless of the
down time [4]. As implementing the latter is much easier in an
optimization problem, a fixed startup cost is considered for the
generators, which is shown in (3).

( ) > 0 ( − 1) = 0
( )= (3)
0 ℎ

Fig. 1. Flowchart of the 2-layered optimization approach


3

B. Generator Constraints Moreover, the transient fuel consumption corrective factor


Two types of constraints are considered for the generators: ( ) is considered to be a multiple of the following :
a) generator power limits and b) generator ramp rate
constraints. The former constraint, shown in (4), simply limits = × , = 7 × 10 (9)
the output power to be within a range. Note that is usually
more than zero, i.e. 10% or 20%. Also the fuel oil price is considered to be 500$ per ton.
The other constraint, generator ramp rate constraint, does Therefore, for this example problem, the following values are
not allow very sharp changes in the output power of the assumed:
generator by defining the maximum allowable ramp rate. Note ( )= 320 × − 496 × + 382.5
that in a discrete simulation, derivative of ( ) can be ℎ
320 496 382.5 (10)
approximated by (5). × − × +
60 60 60
≤ ( )≤ (4)
Δ 7 −4 Δ
, = ( )× 1+ × × (11)
( ) − ( − 1) Δ 3600 Δ
Δ
≤ (5)
This reference does not consider a ramp rate constraint.
Thus, an arbitrary maximum ramp rate of 30% per minute is
considered for the generators in this example:
C. Energy Storage Constraint
In this formulation the only constraint considered for the Δ 30%
≤ (12)
Δ
energy storage is the minimum and maximum allowable
stored energy. This constraint is shown in the following For the sake of simplicity, startup cost is neglected in this
equation: example.
≤ ( )≤ (6)
B. Load Characteristics
The load profile of a ferry ship from BC Ferries, named
D. System Dynamics
Queen of Oak Bay, is arbitrarily chosen for this example. This
Any dynamic optimization problem needs to have the
ferry ship goes back and forth between Nanaimo and
dynamic equations of the system under study. Likewise in this
Vancouver 8 times a day. The daily schedule and the path for
problem, the dynamics of the system need to be added to the
this ferry ship can be found at BC Ferries website.
optimization problem. Dynamic equations simply define how
The online ferry ship tracking feature from the BC Ferries
system state at each time is related to the previous state(s).
website is used to come up with the speed profile of the Queen
As shown in (7), in this problem the dynamics are defined
of Oak Bay during one trip. This speed profile is shown in Fig.
as a single algebraic equation which means the following:
3.
stored energy of the energy storage at the end of the current
Having this speed profile, the power profile shown in Fig. 4
time step is its energy at the previous time step plus the
is assumed for 1 trip for this ship. This figure shows that the
difference between the total generation and the total load,
power profile for this ship is composed of 4 intervals:
multiplied by the time step length.
Ramping up: when the ship leaves the harbor and for a
period of minutes, the power increases and supplies the
( ) = ( − 1) + Δ × ( ) − ( ) (7) propellers enough energy to reach the almost constant cruising
speed. An overshoot of is considered for this subinterval.
In this example = 10 , = 1.15 .
V. EXAMPLE CASE Cruising: in the cruising mode, the speed of the ship is
To be able to perform some studies using the proposed almost constant, but it is assumed that some small sinusoidal
solution methodology, an example case needs to be prepared. power variations with amplitude and frequency exist in
For this example case, generator characteristics are chosen the load profile. This interval starts at and ends at . In this
from [1],[2]. Moreover, the load profile of a ferry ship is used example = 90 , = 1 , = 0.05 , and
as the example load profile. = 1/15 ( / ).
Ramping down: when the ferry is close to the harbor at the
A. Generator Characteristics destination, it slows down and therefore the power drops. In
this example = 100 .
It is assumed that 2 6 generators exist in the example Harbor: when the ship is at harbor, it is lightly loaded.
ferry ship. For a 2005 size generator, [1] uses the Therefore, the load power is assumed to be a sinusoidal wave
following specific fuel oil consumption characteristics: with amplitude and frequency around a small average
power of . In this example = 130 , =
( )= × + × +
ℎ (8) 0.1 , = 0.04 , and = 1/15 ( / ).
= 320, = −496, = 382.5
Note that the per-unit values are defined based on the
4

average load power in the cruise mode which is 8 in this Δ ×( ( ) )×( ( ) + ( )+ )


example. ( ) − ( − 1)
=
× 1+
+ (15)
Δ
25
20
Ship Speed (Knots)

15
C. Decision Variable Bounds
10
The following vectors define the bounds for the decision
variables:
5
0 = … | … | … | (16)
-5
0 20 40 60 80 100 120 = … | … | … | (17)
Time (Minutes)
Fig. 3. Queen of Oak Bay speed profile
D. Ramp Rate Constraint
Ship Power (Normalized by

OS ω1 Each generator at each time step has to satisfy the ramp


the Avg Cruise Power)

1 rate constraint, which can be shown as the 2 equations of (18).


Avg1 A1 Harbor
( ) − ( − 1) + ( )− ( − 1) ≤ Δ
Cruise
Δ
≤ ⇒
− ( )+ ( − 1) ≤ Δ
(18)
0.5
Ramp down ω2
The 2 equations in (18) can be written in the following
Ramp up Avg2 A2 matrix forms:
0
0 T1 20 40 60 80 T2 T3 120 T4 140 ≤ (19)
Time (minutes)
Fig. 4. Assumed power profile for Queen of Oak Bay ferry ship ≤ (20)
where,
VI. DYNAMIC DISPATCH: SOLUTION METHODOLOGY +1 | | | 0
The most common method of solving dynamic −1 +1 | | | 0
optimization problems is to convert them into their equivalent ⋱ | | | ⋮
−1 +1 | | | 0
static forms. This chapter discusses the process of converting − − − − + − + − − − − + −
the problem from a dynamic optimization form to a static = | ⋱ | | ⋮ (21)
optimization form. After converting the problem into an − − − − + − + − − − − + −
| | +1 | 0
equivalent static optimization problem, any optimization
| | −1 +1 | 0
solver can be used to solve the problem. In this work | | ⋱ | ⋮
’s (constrained minimization) function is | | −1 +1 | 0
used. and =− . Moreover,
In order to use this solver, the dynamic optimization
= +Δ × … | … | …
problem has to be converted into the following static form: (22)
+ (0) 0 … 0 | … | (0) 0 … 0
min ( )
. . ≤ = +Δ × … | … | …
(13) − (0) 0 … 0 | … | (0) 0 … 0
(23)
=
≤ ≤
E. System Dynamic Constraint
A. Decision Variables In this problem, system dynamic equation is the equation
When converting an optimization problem from dynamic that relates generator powers and the state of charge of the
form to static form, any decision variable at any time step will energy storage unit, as shown in (24).
be an individual decision variable. Therefore, the vector of
decision variables for this problem is defined as follows: ≤ ( ) = ( − 1) + Δ ( ) − ( ) ≤ (24)
= ⋯ | … | ( ) … |
= (1) ⋯ ( ) | … | (1) … ( ) | (14) The recursive term in (24) can be eliminated so that ( ) is
written as a function of generator powers only. It is shown in
The decision variable vector contains the power of each
(25).
generator at each time step and also the initial charge of the
energy storage. ( ) (0)
≤ = + ( ) − ( ) ≤ (25)
Δ Δ Δ Δ
B. Objective Function
The objective function is as follows: Finally, the 2 inequality constraints of (25) can be written
in matrix form as follows:
5

≤ (26) for the energy storage unit. For this example, Fig. 5 and Fig. 7
show that to minimize the trip cost, a fully charged capacitor
≤ (27) is required when = 8 / ℎ. In contrast, if the electricity
costs 10 / ℎ at the harbor, it is better not to charge the
where , , and are defined as follows:
energy storage and start the trip with a fully discharged unit.
0 ⋯ 0 | ⋯ | 0 ⋯ 0 | 1
Δ 1.0

Generator Power
⋯ 0 | ⋯ | ⋯ 0 | 1

(Percentage)
= Δ (28) P1
⋮ ⋮ ⋱ ⋮ | | ⋮ ⋮ ⋱ ⋮ | ⋮ 0.5 P2
⋯ | ⋯ | ⋯ | 1
Δ
0.0
0 20 40 60 80 100 120
=− (29)

20

State of Charge
+ + ( )
Δ

(MWxmin)
10
= +Δ + ( )
(30)
⋮ 0
0 20 40 60 80 100 120
+ + ( ) Time (Minutes)
Δ
Fig. 6. Optimum generator powers for =8 / ℎ, = 20′000

− − ( ) 1.0
Δ

Generator Power
(Percentage)
P1

= −Δ − ( )
(31) 0.5 P2

− − ( ) 0.0
Δ 0 20 40 60 80 100 120

Therefore, the complete optimization problem is composed


2.0
of the objective function in (15) and the constraints in (16),
State of Charge
(MWxmin)

(17), (19), (20), (26), and (27).


1.0

F. Results
0.0
The results for energy storage capacities of 2 000 0 20 40 60 80 100 120
and 20′000 for a harbor electricity cost of 8 / ℎ Time (Minutes)
Fig. 7. Optimum generator powers for = 10 / ℎ, = 2′000
are shown in Fig. 5 and Fig. 6, respectively. It can be seen that
as the energy storage size increases, the oscillations is VII. SEPARATING OPTIMAL DISPATCH AND UNIT
generator powers decrease and the load fluctuations are COMMITMENT PROBLEMS
covered more by the energy storage. It is worth mentioning
that load fluctuations being covered by the energy storage was Some references consider a minimum acceptable generator
the main idea behind using the energy storage units in ships. loading of more than 0, for example 10% or 20%. It means
Also, Fig. 7 shows the results for a 2 000 energy that if the generator is on, it has to generate more than 10% of
storage, assuming the electricity costs 10 / ℎ at the harbor. its rated power and the reason is higher maintenance costs for
the generator when being lightly loaded. Now assume such a
1.0 limitation needs to be implemented in this formulation.
Generator Power

Therefore:
(Percentage)

P1

0.5 P2
( )∈ 0 ∪ , , >0 (32)
0.0
Implementing this disjoint range in this optimization
0 20 40 60 80 100 120 problem is very difficult. Therefore, to limit generator powers
to a non-zero and still keep the optimization simple, it is
2.0
assumed that the generator on/off statuses are determined in
State of Charge
(MWxmin)

advance. In other words, the unit commitment problem is


1.0
solved before the dispatch problem. Thus, when solving the
generation dispatch problem, the generator on/off statuses are
0.0
0 20 40 60 80 100 120 known (they can be considered as input parameters to this
Time (Minutes) problem). This way, when the generator is off its power is 0
Fig. 5. Optimum generator powers for =8 / ℎ, = 2′000
and when it is on the power is more than .
Clearly, the difference between the electricity price at the To explain how this new input parameter affects the
harbor and on the ship determines the optimum initial charge formulation, the following operator needs to be defined first:
assume is a vector and is a vector of zeros and ones, with
6

the same size as vector . Then = is a vector D. Results


composed of the elements of with a 1 in the associated For the example case introduced earlier, the minimum
position in . To make it clear, consider the following acceptable power for the generators is arbitrarily chosen to be
example: 10% of the nominal power, and the electricity cost at the
= 5 2 3 4 8 harbor is assumed to be 8 / ℎ. Moreover, considering how
= 1 1 0 1 0
⇒ = = 5 2 4 (33) the load profile looks like, ship’s 2 generators are committed
to be online as shown in Fig. 8. It can be seen that generator 1
Likewise for matrices, assume is an × matrix
is in service all the time. In contrast, generator 2 is off in some
and and are binary × 1 and × 1 vectors,
periods (from 0 − 3 and also 99 − 130 ) due to
respectively. Then = , is a matrix that its
light system loading.
rows are the rows of with 1’s in the associated places in
and its columns are the columns of with 1’s in the 10
associated places in .

Load Power
(MW)
5
A. Decision Variables and Objective Function
When a generator is forced to be off, its power is forced to 0
0 20 40 60 80 100 120
0 and therefore no decision variable is required for that
generator for that particular time step. Thus, the new vector of
decision variables is obtained as follows: assume
1

Generator
, ,…, are the on/off vectors of generators 1 to .

Status
G1
Then vectors and are defined as: 0 G2

−− 0 20 40 60 80 100 120
Time (Minutes)
= ⋮ , ′ = − − (34)
−− 1 Fig. 8. Load power and generator on/off commitment for the example case

1.0
Generator Power

And the new vector of decision variables, , is:


(Percentage)

P1

0.5 P2

−− −−
⋮ ⋮ 0.0
= = − − = −− (35) 0 20 40 60 80 100 120

−− — 2.0
State of Charge

1
(MWxmin)

Note that objective function remains the same, and the only 1.0
difference is that the eliminated ( )’s are replaced with 0.
0.0
0 20 40 60 80 100 120
B. Ramp Rate Constraint Time (Minutes)
The ramp rate constraint for the -th generator at time n is Fig. 9. Optimum generator powers for = 8 / ℎ, = 2′000
eliminated from the optimization if ( ) = 0. Therefore, the when the generator on/off commitments are as shown in Fig. 8
original , , and are modified as shown in (36) and 1.0
Generator Power

(37).
(Percentage)

P1
0.5 P2
= , , = , (36)

= , = (37) 0.0
0 20 40 60 80 100 120

C. System Dynamic Constraint 20


State of Charge

System dynamic constraints guarantee that ( ) remains


(MWxmin)

within the limits during the whole study period. Therefore, 10

none of the rows in the matrices associated with the dynamic


constraints are eliminated and only the columns associated 0
0 20 40 60 80 100 120
with the eliminated ( )’s are dropped from the original Time (Minutes)
Fig. 10. Optimum generator powers for = 8 / ℎ, = 20′000
and matrices. Therefore: when the generator on/off commitments are as shown in Fig. 8
= , , = , (38)
Fig. 9 and Fig. 10 show the results for energy storage
= = , = = (39) capacities of 2 and 20 , respectively. These
figures clearly show that generator powers are more than 10%
7

when they are on. Note that generation cost would increase in by using super capacitors. From this figure, it can be seen that
this case compared to the results shown in the previous the savings decrease as the capacitor size increases. Moreover,
chapter, because the optimization problem has more it shows that the total saving is a negative number, meaning
constraints. However, if the startup cost and maintenance costs that the cost of the installed capacitor is higher than the value
are added to the problem, the new solutions would be more of the saved fuel. Therefore, it is not economically
cost efficient. advantageous to install super capacitors in the ship of this
example. Note that this is mainly due to the very expensive
VIII. OUTER LOOP cost of the capacitors.
The objective of the outer loop, as shown in (40), is to 40
minimize total system cost, which is actually the cost minus

Cost per Year (k$)


saving. Equation (41) shows how the present value of saving 30

Saved Running
is calculated and (42) shows the equipment (energy storage)
20
costs. Note that the area cost is ignored in this example,
because no information was found on a reasonable penalty for 10
the occupied area.
0
0 5 10 15 20
min ( − )
(40) EM (MWmin)
max ( − ) Fig. 11. Saved fuel cost per year as a function of energy storage size

0
( )
( × )× ×

= (41)

in 10 Years (M$)
(1 + )

Total Savings
-5

=( × )+( × )
= ×
(42)
-10

A. Super-capacitor -15
Super-capacitors and batteries are 2 energy storage 0 5 10 15 20

alternatives for ships. However, they are very different in E M (MWmin)


Fig. 12. Total savings (including investment costs) in 10 years as a function of
terms of costs. Therefore, the case for using super capacitors super capacitor size
and batteries are considered, separately. The financial
parameters used for this study are as follows: B. Battery
TABLE I
The previous section showed that installing super
FINANCIAL PARAMETERS FOR FEASIBILITY EVALUATION OF SUPER- capacitors in the example ferry ship cannot save money,
CAPACITORS because the installation cost is more than the value of the
Parameter Value saved fuel. As the battery price is much cheaper than the super
365 × 10 = 3650
1 500 $/ [1]
capacitor price, this section evaluates the feasibility of using
5% [1] batteries for this example.
8% [1] All the parameters for this study are the same as the ones
50, 100, 200, 500, 1000, 2000 shown in the previous section, except the energy storage cost
5833 $ [1],[6]
which is considered to be 500 $/ ℎ [8]. It is worth
0
10 × mentioning that the solution of the optimal dispatch problem is
12 / ℎ the same whether super-capacitor or battery is used, due to the
very general formulation presented in this work.
Regarding , while [1],[6] assume a cost of 35’000 $/ Consequently, the values of the saved fuel as a function of the
ℎ and 32’000 $/ ℎ for Maxwell super-capacitors, installed energy storage capacity (batteries in this case)
respectively, [7] uses 80’000 $/ ℎ. The 5833 $ cost for remains the same (Fig. 11).
each capacitor module shown in Table I is equivalent to Fig. 13 shows the total savings in 10 years (fuel saving
35’000 $/ ℎ. value minus the battery cost) as a function of the battery size.
For this example, Fig. 11 shows the value of the saved fuel In contrast with the super capacitor case, it shows positive
per year as a function of the energy storage size, obtained by savings. Moreover, an optimum battery size can be extracted
solving the dynamic dispatch problem. Clearly, as the size of from this figure. It is 10’000 × or 167 ℎ for this
the energy storage increases, more savings are achieved. It can example.
also be seen that this curve shows a saturation behavior. It Therefore, for this example the optimum battery capacity is
means that as the storage size gets bigger and bigger, the rate 167 ℎ and the optimum generator powers during each trip
of increase of the saved fuel decreases. are as shown in Fig. 11.
Fig. 12 integrates the fuel saving value of Fig. 11 with the
super capacitor cost to show the total money that can be saved
8

0.25 XI. REFERENCES


in 10 Years (M$)
0.20 [1] M. B. Solhaug, "Fuel Efficient Operation of DP Thruster Drives by use
Total Savings

of Super Capacitors," Master of Science, Dept. Department of Electric


Power Engineering, Norwegian University of Science and Technology,
0.15
2010.
[2] M. B. Solhaug, "Fuel efficient operation of DP thruster drives by use of
0.10 super capacitors," ABB 2009.
[3] W. Wei, W. Daifeng, A. Arapostathis, and K. Davey, "Optimal Power
0 5 10 15 20 Generation Scheduling of a Shipboard Power System," in Proc. 2007
EM (MWmin) IEEE Electric Ship Technologies Symposium, pp. 519-522.
Fig. 13. Total savings (including investment costs) in 10 years as a function of [4] H. Ma and S. M. Shahidehpour, "Unit commitment with transmission
battery size security and voltage constraints," IEEE Transactions on Power Systems,
vol. 14, pp. 757-764, 1999.
[5] C. C. A. Rajan and M. R. Mohan, "An evolutionary programming-based
1.0
Generator Power

tabu search method for solving the unit commitment problem," IEEE
(Percentage)

P1
Transactions on Power Systems, vol. 19, pp. 577-585, 2004.
0.5 P2 [6] "Maxwell Ultra Capacitors," [online] Available:
http://energystoragenews.com.
0.0 [7] L. M. P. Fanjul, "Some New Applications of Supercapacitors in Power
0 20 40 60 80 100 120 Electronic Systems," Master of Science, Dept. Electrical Engineering,
Texas A&M University, College Station, 2003.
[8] R. Lache, D. Galves, and P. Nolan, "Vehicle Electrification: More Rapid
10 Growth; Steeper Price Declines for Batteries," Deutsche Bank March
State of Charge

2010.
(MWxmin)

0 XII. BIOGRAPHIES
0 20 40 60 80 100 120
Time (Minutes) Salman Mashayekh (S'09) received his B.S. and M.S. in Electrical Power
Fig. 14. Optimum generator powers for the optimum battery size ( = Systems from University of Tehran, Iran, in 2006 and 2008, respectively. He
10′000 ) when = 12 / ℎ and generator on/off commitments are joined the Power System Automation Lab in Texas A&M University as a PhD
as shown in Fig. 8. student in 2008. His research interests are in power management system for
isolated power systems. His job focuses on contingency analysis, dynamic
stability studies, security enhancement, etc.
IX. SUMMARY AND CONCLUSION Zhenyuan Wang (M’2000) joined ABB US Corporate Research Center in
This paper showed how the optimal generation dispatch Raleigh, North Carolina in 2000, where he is currently a group leader. His
research interests include electric power equipment condition
problem can be solved for a particular energy storage size. For monitoring/assessment/diagnosis, system monitoring, control and automation
this formulation, this paper used a simple but general for a smart grid. His experiences include asset management IT applications in
shipboard architecture which can be applied to almost any ship the electric power industry, power system transient analysis,
substation/distribution automation, data integration/warehousing/mining
with any energy storage type. Then, the optimal dispatches for applications, and renewable integration..
various energy storage sizes were used to find the maximum Li “Lisa” Qi (SM’07) received her B. Eng and M.Sc. from Xi’an Jiaotong
possible fuel saving for each design. Finally, a simple University and Zhejiang University, China and Ph.D. from Texas A&M
economical analysis was used to determine the optimum University, College Station, TX, USA. All are in Electrical Engineering. She
is currently an R&D Engineer with ABB Corporate Research (US Center).
energy storage size, i.e. the one with maximum net saving Her expertise and research interests include power system modeling and
over the ship life time. simulation, power system dynamics, stability analysis, power system
The studies performed showed that the results highly protection, DC grids, and electricity market.
John Olav Lindtjørn received his M.S. in Electrical Engineering from
depend on the load profile, and generation fuel consumption the Norwegian University of Science and Technology in Trondheim in 2004.
characteristics, especially the transient inefficient penalty He joined ABB shortly afterwards where his main technical focus has been
factor ( ). It is worth mentioning that a good and accurate diesel electric propulsion systems onboard Oil and Gas related vessels. His
most recent work has been related to the ABB Onboard DC Grid concept.
model for the transient energy consumption of generators does Tor-Arne Myklebust received his M.S. in Electrical Engineering from
not exist in literature and improvements in this area can help the Norwegian University of Science and Technology in Trondheim in 2003.
in finding a more accurate economical figure for the energy After graduation he joined ScanWind AS, where he worked on the electrical
power systems and controls of a new wind turbine concept. In 2005 he started
storage problem. his career at ABB. Besides working with analyses and design of diesel electric
Formulation simplicity was one of the main criteria in this propulsion systems, he has been involved in various development projects,
work which resulted in a simple yet general optimization including the ABB Onboard DC Grid concept.
formulation with linear constraints. However, this formulation
can improve by including startup cost and energy storage
charge/discharge efficiencies into the problem, which both
will introduce some nonlinear constraints into the problem.

X. ACKNOWLEDGMENT
The first author would like to acknowledge ABB’s support
on this research work.

Das könnte Ihnen auch gefallen