Documente Academic
Documente Profesional
Documente Cultură
r
|
c
e
(
]
M
W
h
)
Date
Max rlce
AvC rlce
Mln rlce
Fig. 1. Belpex prices for month of April in 2013
IV. GENERIC ENERGY STORAGE MODEL
Before mathematical formulation of the optimization prob-
lem, a mathematical model for energy storage system is given
in this section. The following generic discrete-time model for
energy storage can be employed to perform optimization [12]:
E
s
(k + 1) = E
s
(k) + P
s
(k).T (1)
ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia
978-1-4799-2449-3/14/$31.00 2014 IEEE 573
where
=
_
c
, if P
s
(k) 0 (charging)
d
=
1
c
, otherwise (discharging)
(2)
subject to
0 E
s
(k) E
max
s
(3)
|P
s
(k)| P
max
s
(4)
where k denotes the present time instant t
k
= k.T, E
s
(k)
the level of stored energy at present time k, P
s
(k) the power
exchange (charging/discharging) with the storage at time k,
c
resp.
d
charging resp. discharging efciencies. T =
t
k+1
t
k
is the duration of each interval (time difference
between two subsequent time instants e.g. k +1 and k). Note
that depending on the type of the storage technology, different
operational constraints on maximum power exchange as well
as on maximum energy capacity itself may apply. E
max
s
denotes the maximum energy capacity of the storage, and
P
max
s
is the maximum rate of charge/discharge. We assume in
this study that the (absolute value of) maximum rate of charge
P
maxc
s
and maximum rate of discharge P
maxd
s
are equal.
Thus,
|P
maxc
s
| = P
max
s
, |P
maxd
s
| = P
max
s
(5)
V. FINANCIAL VIABILITY OF A STORAGE PROJECT
Due to high capital cost of energy storage systems, and
limited lifetime of most of storage technologies and their
(rather) low efciency, an in-depth economic analysis is
needed to determine whether or not a new storage project is
nancially viable. Difculties in accurately estimating capital
cost (especially for state-of-the-art storage technologies), and
their expected lifetime (considerably effected by depth-of-
discharge (DOD)) makes this problem even more challenging.
We will use net present value (NPV) as an economic measure
for this purpuse. Note that NPV is a measure to compare the
value of money at present time with the value of the same
amount of money in the future, taking ination into account.
An investment on a prospective project is protable if its NPV
is positive, while a negative NPV means that the project should
be rejected.
NPV = C
S
+
N
n=1
R
n
(1 + r)
n
(6)
where C
s
is the initial investment cost (capital + O&M), N
is the overall lifetime of the project (storage in this case), r
is the (annual) discount/interest rate, and R
n
is the revenue
generated by storage at year/month n.
VI. FORMULATION OF THE OPTIMIZATION PROBLEM
A. Stage I: design optimization for sizing the storage
The design optimization is to determine storage maximum
energy capacity E
max
s
and maximum charge/discharge power
rate P
max
s
. On the other hand, the total investment cost
of energy storage system C
S
is very much technology-
dependent, and can signicantly vary from one to another
[13]. The latter two design parameters E
max
s
, P
max
s
will
jointly dene C
S
by imposing costs associated with P
max
s
(an indication of how big the power conversion unit is), plus
costs associated with E
max
s
(e.g. size of electrolyte storage
tank in ow battery).
C
S
= C
g
P
. P
max
s
+ C
g
E
. E
max
s
. .
initial capital cost
+ C
g
O&M
. .
O&M cost
(7)
where C
g
P
is the capital cost per each unit of power of the
storage technology g (in e/KW), and C
g
E
is the capital cost
per each unit of energy of the technology g (in e/KWh). As
mentioned before, the values of C
g
E
and C
g
P
can be very dif-
ferent for each storage technology. Note that beside one-time
capital cost certain annual operation and maintenance costs
C
g
O&M
(in e/KW-yr) also apply for each storage technology.
It is a common practice to assume the annual C
g
O&M
constant
over the lifetime of the energy storage (e.g. equal to 1% of
the capital cost per year).
We size the storage based on the following:
when congestion:
- charge only from wind unit
- discharge to the grid at market price as
soon as congestion is relieved to gener-
ate revenue
This can be mathematically expressed by the following
maximization problem:
(P
max
s
, E
max
s
) = argmax
P
max
s
,E
max
s
NPV (P
max
s
, E
max
s
) (8)
= (C
g
P
. P
max
s
+ C
g
E
. E
max
s
+ C
g
O&M
)
+ R
1
N
n=1
1
(1 + r)
n
where
R
1
= R
n
, n = 2, . . . , N
R
1
=
n
T
k=1
(k). min
_
_
E
S
(0) +
k
i=1
(i). sign(P
S
(i)).
min(|P
S
(i)|, P
max
s
).T
_
, E
max
S
_
.(k)
and
=
_
0, if P
s
0 (charging)
1, otherwise (discharging)
n
T
=
T
T
is the total number of 15-minute intervals, and
T total length of study horizon. is a binary variable which
ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia
978-1-4799-2449-3/14/$31.00 2014 IEEE 574
is introduced to distinguish between charging and discharging
modes. (k) refers to the external spot market price at time k.
B. Stage II: operational optimization for maximizing the rev-
enue generated by storage
After sizing the storage in Stage I, in order to obtain the
maximum revenue generated by storage, we consider another
operational optimization according to the following:
when no congestion:
- charge from both wind unit and the grid
at favorable moments
- discharge when favorable (not as soon as
congestion is relieved)
This can be mathematically expressed by the following opti-
mization problem:
P
s
= argmax
P
s
R
1
(P
s
) (9)
where
R
1
=
n
T
k=1
(k).P
S
(k).(k)
and
=
_
1, if P
s
0 (charging)
+1, otherwise (discharging)
where the optimal discharge (and also charging from the grid)
time can be decided based on the other parameters such as
market price, wind forecast, DLR, and DSM.
This optimization algorithm will iteratively -for every dis-
crete time instant- dictate the optimal schedule for charg-
ing/discharging power rates of storage.
The proposed optimization algorithm employs the histori-
cal time-series data for wind generation units, demand, and
market price, for every 15-minute interval, of the past years.
This would mean a perfect-forecast scenario for a real-time
optimization algorithm to which possible prediction errors of
those parameters apply. Thus the optimization algorithm based
on historical data will nd the absolute optimal size of the
storage, and the maximum possible revenue available during
that past year. Expectedly, the results given by a real-time
optimization algorithm may be different from those given by
an algorithm based on historical data. However, the historical-
data based algorithm provides an effective measure to evaluate
the performance of a real-time optimization algorithm by
comparing the results.
VII. TEST SYSTEM
Figure 2 shows a simple prototypical wind-storage system
which is connected to the main grid via some weak trans-
mission lines. The 5 MW wind turbine generator is located
far away from load area (in main grid), and the 15 kV
(distribution) line is subject to voltage and current congestion.
At the presence of no storage device, one would need to
simply curtail the congestion-induced wind power, let alone
reinforcing the transmission lines, in order to relieve the
congestion. Figure 3 shows the excessive power production (in
KW) of the wind turbine on an 15-minute basis that causes
voltage and/or current congestion. There are 6695 quarters
(out of 35040) with excessive production, i.e. about 19% of
the overall time (entire year 2011). The prole is averaged
around 265 KW across the entire year, and around 1385 KW
across periods with exessive power production. To relive the
congestion, the surplus of power needs either to be curtailed or
to be stored in this case. The aim of this section is to nd the
optimal size of a storage device to avoid congestion. Note that
the economic operation of the storage device is not considered
in this paper. Thus it is assumed that energy storage is only
able to generate revenue by reducing the amont of congestion-
induced wind power curtailemnt, and cannot participate in
generation time-shifting and price arbitrage. As shown in
Fig. 4, the imbalance prices for year 2011 (in e/MWh) in
Belgian electricity market is employed on an 15-minute basis
to perform the proposed optimization algorithm. As can be
seen, the price is averaged around 30 e/MWh across the entire
year, with very high peak over 2750 e/MWh in March 28 [14].
Beside pumped hydro and compressed air energy storage
technologies as mature solutions, the utility-scale Sodium-
Sulfur (Na-S) and Lead-Acid (Pb) batteries are commercially
available. We will consider Na-S battery type for the calcula-
tion, with the parameters given in Table I [15], [16], [17].
TABLE I
CHARACTERISTICS OF THE CONSIDERED NA-S BATTERY
C
g
P
255 e/KW
C
g
E
255 e/KWh
C
g
O&M
10 e/KW-yr
(%) 85
N 15
The exchange rate 1US$ = 0.73 e is applied.
This economical analysis reveals that the total investment
cost of the Na-S battery considered in this paper is much
higher than the revenue that could be generated by employing
only a single revenue stream (curtailment reduction). Thus, the
algorithm conrms that it is not economically viable to size
the storage device based on only curtailment reduction. The
storage project may become protable either by reducing the
total investment cost of the storage device, or by including
other possible revenue streams to maximize the monetary
value that can be generated by storage (such as generation
time-shifting, price arbitrage, or any governmental subsidy
in terms of feed-in tariffs etc.) The NPV for a 1.5MW
10MWh Na-S battery turns out to be -e2953400 in the studied
case (note the negative sign), i.e. the difference of the total
investment cost e3157500 and the revenue that could be
generated only by curtailment reduction e204120. Assuming
that O&M cost remains the same (10 e/KW-yr), the storage
under consideration would be marginally protable if the
capital costs would be dramatically reduced down to values
ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia
978-1-4799-2449-3/14/$31.00 2014 IEEE 575
C
g
P
= 3.2123 e/KW and C
g
E
= 1.6062 e/KWh.
Power grid
(electricity market)
Possible directions
for power flow
Fig. 2. One-line diagram of the test system
0 0.5 1 1.5 2 2.5 3 3.5 4
x 10
4
0
500
1000
1500
2000
2500
3000
Fig. 3. Curtailed power of a wind turbine
0 0.5 1 1.5 2 2.5 3 3.5 4
x 10
4
-500
0
500
1000
1500
2000
2500
3000
Fig. 4. Imbalance prices for year 2011
VIII. CONCLUSION
This paper deals with the exploitation of storage in low
voltage (LV) wind-storage distribution grids with the aim of
alleviating the transmission congestion, and to minimize the
amount of curtailed wind. A generic model for energy storage
is derived which is suitable to be used in optimization for-
mulations. From DSO point of view a two-stage optimization
problem is formulated to optimally size the storage, and later
to maximize the revenue that can be generated by it. To
perform the optimization, a simple prototypical wind-storage
system is considered which is connected to the main grid via
some weak transmission lines. The results show that due to
high investment costs of the energy storage (a Na-S battery in
this case), it is not economically viable to size the storage only
for curtailment reduction purpose. The results also suggest
that, in the studied case, it is necessary to include other
possible revenue streams to increase the monetary value of
the wind-storage system.
The important issue of storage placement is also discussed
by introducing the concept of PTDF. The paper under consid-
eration in fact provide the foundation for another upcoming
paper which will represent the simulation/optimization results
using the real-time time-series data of wind, price and demand.
ACKNOWLEDGMENT
The research was carried out in the frame of the Inter-
university Attraction Poles program IAP-VII-02, funded by
the Belgian Government.
REFERENCES
[1] Y. R. Sood and R. Singh, Optimal model of congestion management in
deregulated environment of power sector with promotion of renewable
energy sources, Renewable Energy, vol. 35, no. 8, pp. 1828 1836,
2010.
[2] S. Gill, G. W. Ault, and I. Kockar, The optimal operation of energy
storage in a wind power curtailment scheme, in 2012 IEEE Power and
Energy Society General Meeting, 2012.
[3] A. Michiorri, R. Currie, P. Taylor, F. Watson, and D. Macleman,
Dynamic line ratings deployment on the orkney smart grid. CIRED,
2011.
[4] P. Schell, B. Godard, J.-L. Lilien, H.-M. Nguyen, V. De Wilde,
O. Durieux, and J.-J. Lambin, Large penetration of distributed pro-
ductions: dynamic line rating and exible generation, a must regarding
investment strategy and network reliability, in CIRED 2012 Workshop
on Integration of Renewables into the Distribution Grid, 2012.
[5] U.S. Department of Energy. [Online]. Available:
http://www.eia.gov/electricity/data/eia861/dsm/index.html
[6] M. Moradzadeh, B. Zwaenepoel, R. Boel, and L. Vandevelde, Use of
energy storage for Belgian power network, in EPEC2013, 2013.
[7] J. Rogers, S. Fink, and K. Porter, Examples of wind energy curtailment
practices, NREL, Subcontract Report NREL/SR-550-48737, 2010.
[8] H. K. Jacobsen and S. T. Schr oder, Curtailment of renewable genera-
tion: Economic optimality and incentives, Energy Policy, vol. 49, pp.
663 675, 2012.
[9] Y. Gu, L. Xie, B. Rollow, and B. Hesselbaek, Congestion-induced wind
curtailment: Sensitivity analysis and case studies, in North American
Power Symposium (NAPS), 2011.
[10] S. Gill, E. Barbour, I. A. G. Wilson, and D. Ineld, Maximising
revenue for non-rm distributed wind generation with energy storage
in an active management scheme, in IET Conference on Renewable
Power Generation (RPG 2011), 2011.
[11] Y. Zhou, A. Scheller-Wolf, N. Secomandi, and S. Smith, Managing
wind-based electricity generation in the presence of storage and trans-
mission capacity, working paper, 2013.
[12] A. Parisio and L. Glielmo, A mixed integer linear formulation for
microgrid economic scheduling, in 2011 IEEE International Conference
on Smart Grid Communications (SmartGridComm), 2011, pp. 505510.
[13] S. Faias, J. Sousa, and R. Castro, Embedded energy storage systems
in the power grid for renewable energy sources integration, Renewable
Energy, pp. 6388, 2009.
[14] Elia, Belgiums electricity transmission system operator. [Online].
Available: http://www.elia.be/
[15] R. Carnegie, D. Gotham, D. Nderitu, and P. V. Preckel, Utility scale
energy storage systems, 2013.
[16] S. Schoenung, Energy storage systems cost update, Sandia National
Laboratories, Albuquerque, 2011.
[17] K. Bradbury, Energy storage technology review, Duke University, pp.
134, 2010.
ENERGYCON 2014 May 13-16, 2014 Dubrovnik, Croatia
978-1-4799-2449-3/14/$31.00 2014 IEEE 576