IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO.

3, SEPTEMBER 2009 661

Wind Farms as Reactive Power Ancillary Service

Providers—Technical and Economic Issues
Nayeem Rahmat Ullah, Member, IEEE, Kankar Bhattacharya, Senior Member, IEEE,
and Torbjörn Thiringer, Member, IEEE

Abstract—This paper examines the possibility of providing reac- wind turbines (WTs) can also provide temporary primary fre-
tive power support to the grid from wind farms (WFs) as a part of quency regulation and oscillation damping, as discussed in [9].
the ancillary service provisions. Detailed analysis of the WF capa- In the literature, reactive power as an ancillary service has
bility curve is carried out considering maximum hourly variation
of wind power from the forecasted value. Different cost components been mainly examined in the context of large thermal genera-
are identified, and subsequently, a generalized reactive power cost tors [10], [11], and not much work has been reported that exam-
model is developed for wind turbine generators that can help the ines how wind generators could contribute to system reactive
independent system operator (ISO) in managing the system and power requirements. Reactive power provision from wind farms
the grid efficiently. Apart from the fixed cost and the cost of loss (WFs) is rarely procured by the independent system operators
components, a new method is proposed to calculate the opportu-
nity cost component for a WF considering hourly wind variations. (ISOs). However, with increasing penetration of wind gener-
The Cigré 32-bus test system is used to demonstrate a case study ators into the power grid together with the increasing usage
showing the implementation of the developed model in short-term of power electronics in turbines, WFs could be useful reactive
system operations. A finding is that higher wind speed prediction power service providers in the future. Due to the advancement
errors (a site with high degree of wind fluctuations) may lead to in wind forecasting techniques in recent years, together with
increased payments to the WFs for this service, mainly due to the
increased lost opportunity cost (LOC) component. In a demon- the power smoothing effect within a WF, the WFs can now be
strated case, it is found that 2340 $/h is paid to the WF as the LOC considered as a dispatchable power source by the system oper-
payment only, when the wind prediction error is 0.5 per unit (p.u.), ators [12], [13]. Consequently, the reactive power support from
whereas 54 $/h is the expected total payment to the WF when the the voltage-source converters (VSCs) of WFs can also be treated
prediction error is 0.2 p.u. for its reactive power service. as dispatchable by the operators.
Index Terms—Ancillary service, reactive power, variable-speed The capability curve of these units needs to be defined taking
wind turbine (VSWT), voltage-source converter (VSC). the wind variations into account. Different cost components
associated with reactive power generation by these units need to
I. INTRODUCTION be examined as well. A knowledge of these cost components will
ARIABLE-SPEED wind turbines (VSWT) with power assist the ISO in formulating appropriate financial compensation
V electronic converters are increasingly penetrating the
power systems because of their flexibility in controlling the
mechanisms for their reactive power service provision. When
these issues are answered, WFs with reactive power capability
response of the turbine. In recent years, researchers have exam- can be treated by the ISOs as reactive power ancillary service
ined these wind turbine generators extensively, with regard to providers.
their grid integration issues and control of the converters [1]–[4]. The main objectives of this paper are to propose compre-
In the context of deregulation of the electricity sector, the market hensive technical models and detailed cost representation of
dispatch aspect of wind turbine generators and their role have reactive power generation from WFs. The research focusses on
also been studied [5], [6]. the development of the capability curve of a VSWT-based WF
In addition to providing active power to the grid, the wind with full-scale power electronic converter, and hence its reac-
generators with power electronic convertors can also provide tive power production taking into consideration the currently
reactive power to the system by incorporating minor modifica- existing grid codes and the wind variability factor. Representa-
tions to their design and/or control architecture [7], [8]. These tion of different cost components, and hence the development
of a reactive power cost model for WFs that can be used by
Manuscript received April 22, 2008; revised July 9, 2008. First published the ISOs for procurement of reactive power ancillary services
January 27, 2009; current version published August 21, 2009. This work was from these units is examined. A new method for defining the
supported by E.ON Sverige AB’s Research Foundation. Paper no. TEC-00137- lost opportunity cost (LOC) for WFs is proposed that takes the
2008.
N. R. Ullah was with the Department of Energy and Environment, Chalmers wind fluctuations into account. The goal is to consider WF sites
University of Technology, Göteborg 412 96, Sweden. He is now with E.ON with different degrees of wind power prediction errors and its
Elnät AB, Malmö 205 09, Sweden (e-mail: [email protected]). consequences on the payment to WFs for the reactive power.
K. Bhattacharya is with the Department of Electrical and Computer En-
gineering, University of Waterloo, Waterloo, ON N2L 3G1 Canada (e-mail: The paper is organized as follows. Section II provides the de-
[email protected]). tails of the development of the reactive power capability curve
T. Thiringer is with the Department of Energy and Environment, Chalmers of a WF considering wind variations under different grid con-
University of Technology, Göteborg 412 96, Sweden (e-mail: torbjorn.
[email protected]). nectivity standards. Section III identifies different cost compo-
Digital Object Identifier 10.1109/TEC.2008.2008957 nents (fixed cost, cost of loss, and LOC) associated with reactive

0885-8969/$26.00 © 2009 IEEE

662 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

Fig. 1. VSWT with a full-scale back-to-back power electronic converter com-

prising a machine-side converter (MSC) and a GSC.
Fig. 3. Capability curve of a VSWT with full-scale power electronic converter.

hardware setup, Vc has a certain maximum value. X1 represents

the total reactance of the wind turbine’s step-up transformers
and the grid filters (10%), X2 represents the reactance of the
transformer transforming the wind turbine’s medium voltage to
400 kV (20%).

B. Limiting Factors
The grid-side converter (GSC) has a certain maximum
current-carrying capacity, which will impose a limit on the P -
and Q-capability of the turbine. In the PQ plane, it will be a cir-
cle similar to that of the armature current limit of a synchronous
generator, as shown in Fig. 3. The relationship between the
active and reactive power at the converter current limit is
P 2 + Q2 = (Vg Ic )2 (1)
where Ic is the converter current.
The converter voltage Vc will impose another limit on P -
and Q-capability of the wind turbine, which is similar to that
of the field current limit of a synchronous generator. Based on
these analogies of the capability curve of a wind turbine with
full power converter (see Fig. 1) with that of a conventional
synchronous generator [14], the relationship between P and Q
Fig. 2. Typical wind park layout considered in this paper. at the converter voltage limit is given by

2  2
2
Vg2 Vc Vg
power support from a WF. Section IV brings out the various cost P + Q+ = (2)
X X
components of reactive power from WFs, and then proposes a
unified cost model for the same, which can be used by ISOs where X is the total equivalent reactance from the wind turbine
in the procurement of these services. Section V presents a case low-voltage terminal to the grid connection point (see Fig. 2).
study demonstrating the integration of the WF reactive power Equation (2) represents a circle in the PQ plane with center (c)
capability (Q-capability) in short-term system operations. and radius (r) given as


Vg2 Vc Vg
II. REACTIVE POWER (Q) CAPABILITY OF WFS UNDER c = 0, − , r= . (3)
X X
DIFFERENT GRID CODES
A. Investigated WF Layout C. Design Values
A VSWT with a full-scale power electronic converter is Grid codes of utilities often require a WF to be able to op-
shown in Fig. 1, where Ic is the converter current. The typi- erate continuously at the rated value at a certain power fac-
cal layout of a WF is shown in Fig. 2 wherein Vg , P , and Q tor (pf) within a certain band of voltages and frequencies
represent the voltage at the grid connection point, WF active and at the grid connection point, see [15] and [16], for exam-
reactive power production at the grid connection point, respec- ples. Let us define the upper and lower values of voltage and
tively. Vc represents the converter voltage, which depends on frequency as Vg ,m ax = 1.05, Vg ,m in = 0.9, fm ax = 1.01, and
the dc-link voltage of the converter (Vdc ), the modulation tech- fm in = 0.98, respectively [all in per unit (p.u.)], and the rated
nique used, and the amplitude modulation index [4]. For a given pf as cos θR , where θR is the rated pf angle of the WF. These
ULLAH et al.: WIND FARMS AS REACTIVE POWER ANCILLARY SERVICE PROVIDERS—TECHNICAL AND ECONOMIC ISSUES 663

can be used to calculate the design value or the maximum value

of the converter voltage (Vc,m ax ), which determines the maxi-
mum dc-link voltage (Vdc,m ax ) of the converter, as well as the
maximum current rating (Ic,m ax ) of the converter. Note that it
is assumed that the grid reactive power requirement is met by
the built-in converters of the wind turbines, and accordingly, the
design values Vc,m ax , Ic,m ax , and Vdc,m ax are determined.
A balanced design of the wind turbines should make sure
that the rated values of the active and reactive power at the grid
connection point can be maintained at the most demanding grid Fig. 4. WF’s capability diagram (design pf = 1.0) for different grid voltages.
situations.
The converter current Ic is expressed as

P 2 + Q2
Ic = . (4)
Vg
The converter current is maximum when the active and reactive
power from the WF are at the rated value and the grid voltage
is at the minimum value, which is

PR2 + Q2R
Ic,m ax =
Vg ,m in
 Fig. 5. WF’s capability diagram (design pf = 0.95) for different grid voltages.
PR2 + (PR tan θR )2
= (5)
Vg ,m in
where PR and QR are the rated active and reactive power of the tively, at a certain level of active power production. Qc and Qv
WF, respectively. If PR is taken as the mega volt ampere (MVA) are given by
base of the system, then Ic,m ax (5) in p.u. reduces to 
 Qc = (Vg Ic,m ax )2 − P 2 (11)
1 + tan2 θR
Ic,m ax = . (6)  2
Vg ,m in Vc,m ax Vg Vg2
Qv = − P2 − . (12)
The converter voltage Vc can be found from the relation X X
 2
2
Vc Vg 2
Vg2 1) Design pf = 1.0: If the specified requirement for con-
= P + P tan θ +
X X nectivity of the WF to the grid is to maintain a continuous zero
 reactive power exchange at the connection point (usually at the
 
X Vg2 2 transformers), then the current and voltage limit curves of the
⇒ Vc = P 2 + P tan θ + . (7) WF should intersect the real axis of the PQ plane at point (1, 0)
Vg X
(where Q = 0) and is the most demanding grid conditions for
The converter voltage is maximum when the grid voltage and a balanced design (see “×” in Fig. 4). For this particular case,
system frequency are at maximum, and the active and reactive Ic,m ax = 1.11 and Vc,m ax = 1.09 calculated from (6) and (8).
power from the WF are at the rated values. So, the maximum With this design, the wind turbine is nevertheless capable of
converter voltage (line-line rms) in p.u. is producing reactive power at different wind speeds as determined

  by different limits of the converters, as shown in Fig. 4. This
fm ax X Vg2,m ax 2
Vc,m ax = 1 + tan θR + . (8) reactive power can be provided by the WF to the system at low
Vg ,m ax fm ax X additional cost required for software modification. This aspect
The corresponding maximum dc-link voltage to obtain will be further discussed in a later section.
Vc,m ax , assuming the space vector modulation (SVM) tech- 2) Design pf = 0.95: In this case, we consider that the grid
nique or the nonsinusoidal pulsewidth modulation (PWM) tech- connection requirement is that the WF should have the capacity
nique [17], is to operate at leading/lagging pf = 0.95 at rated operation. In
√ order to fulfil this requirement, the maximum converter volt-
Vdc,m ax = 2Vc,m ax . (9)
age (Vc,m ax ) and the maximum current-carrying capacity of the
The maximum reactive power injection capability from the converter (Ic,m ax ) should be larger compared to pf = 1.0 case,
WF is now calculated as which will require additional cost. The new values for Ic,m ax and
Vc,m ax are 1.17 and 1.17, respectively, calculated from (6) and
Q = min{Qc , Qv } (10)
(8). The new capability curve for the WF is shown in Fig. 5. It
where Qc and Qv are the converter-current-rating- and voltage- can be seen that the WF can now inject or absorb reactive power
rating-limited reactive power production from the WF, respec- in excess of QR (see “×” in Fig. 5) during low wind speeds at
664 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

TABLE I
Ic , m a x , V c , m a x , P c , m a x , AND V d c , m a x FOR DIFFERENT DESIGN CASES

Fig. 6. Maximum capacitive reactive power capability with different design cases are summarized in Table I. In the table, the corresponding
pf as a function of wind speed. Approximate active power is also shown. maximum dc-link voltage is also given [calculated from (9)].
Note that converter voltage limitation will not occur if the
converter maximum voltage Vc,m ax is such that (from Fig. 3)
Vc,m ax Vg Vg2
≥ + Vg Ic,m ax
X X
⇒ Vc,m ax ≥ Vg + Ic,m ax X. (14)
Within the continuous operating region of the WF, if Vc,m ax ≥
1.4 (i.e., Vdc,m ax ≥ 2), then the converter voltage will not im-
pose any limit on the capability of the WF, only the converter’s
maximum current-carrying capacity will limit its capability. See
Fig. 7(b) for the new set of capability curves where voltage limit
does not occur. Note that the reactive power margin of the WF
Fig. 7. (a) WF’s capability diagram for different design pf values when V g = will largely depend on the initial reactive power output from the
1.0 p.u. (b) WF’s capability diagram for different design pf values at V g = 1.0 WF.
when the converter voltage limit does not occur.

D. WF Q-Capability Considering Wind Variations

the most demanding grid condition when the grid voltage is at
the maximum. The power generated from a WF varies considerably over
3) Design pf = 0.9: Additional reactive power support can the daily cycle. The reported wind plant integration study [1]
be provided by the WF with appropriate modifications to its by Xcel Energy North considered wind forecast errors in the
hardware (such as modifications of converter voltage and cur- range of ±10% to ±50%. It is reported in [12] that a reasonable
rent ratings). These will involve additional costs and may be un- day-ahead hourly wind power forecast, using the state-of-the-
dertaken by the WF if there are appropriate mechanisms for cost art forecasting tools, will have a mean absolute error (MAE)
recovery. If the WF is designed to operate at 0.9 pf at the most of 10%–15% of the rated capacity of the WF. The frequency
demanding grid conditions, which is in excess of the grid code distribution of the day-ahead wind power forecasting error in
requirements of 0.95 pf operation, then the required converter Germany reveals a maximum ±25% error [13]. Such a high
current and voltage ratings are Ic,m ax = 1.24 and Vc,m ax = 1.21 degree of variability of the actual hourly wind power from the
calculated from (6) and (8). forecasted value reduces the available Q-capability of the WF
Finally, the active and the maximum reactive power gener- from the actual capability, which is derived in this section. This
ation from the WF as a function of wind speed are shown in reduced capability should be reported to the ISO as the available
Fig. 6. The wind speed is divided into three regions—region (I) hourly Q-capability of the WF.
covers from zero to the cut-in wind speed vi , region (II) covers The actual Q-capability of the WF (Qac,k ), when the fore-
from vi to the rated wind speed vr , and region (III) covers wind casted wind power is Pk∗ , at kth hour of operation is obtained
speeds above vr . In these calculations, a piecewise linear active from (10) as
power generation for the WF is used, given as
Qac,k = min{Qc,ac,k , Qv ,ac,k } (15)
P =0 ∀v ≤ vi where
  
v − vi
= PR ∀vi < v < vr Qc,ac,k = (Vg Ic,m ax )2 − Pk∗2 (16)
vr − v i
 2
= PR ∀v ≥ vr (13) Vc,m ax Vg Vg2
Qv ,ac,k = − Pk∗2 − (17)
X X
where PR is the rated power of the wind turbine. The capability
curves of the WF for three different design pfs are summarized where Qc,ac,k and Qv ,ac,k are the actual converter-current-
in Fig. 7(a) for grid voltage (Vg ) of 1.0 p.u. rating- and voltage-rating-limited Q-capability, respectively,
4) Hardware Comparison: The maximum converter current at kth hour of operation. However, due to the high degree of
and voltage ratings as well as the power rating for these three variability of actual wind power in both directions from the
ULLAH et al.: WIND FARMS AS REACTIVE POWER ANCILLARY SERVICE PROVIDERS—TECHNICAL AND ECONOMIC ISSUES 665

TABLE II
INVESTMENT COST OF SOME WFS [21]–[23]

Fig. 8. Available Q-capability considering wind variations compared to actual

Q-capability at V g = 0.95 p.u., design pf = 0.95 (considering ∆P m a x = 0.2 5) decommissioning (2%);
and 0.5 p.u.). Only the capacitive capability is shown. 6) WF design (1%).
To extract the investment cost components, only the following
forecasted value, delivery of Qac,k cannot be guaranteed by the components need to be considered:
WF over the entire hour of operation. The WF may lose signifi- 1) hardware (including onshore transportation) (82%);
cant amount of revenue from its energy sold if Qac,k is reported 2) assembly, transport, and installation (16%);
to the ISO as the available amount of reactive power (Qav,k ) at 3) WF design (2%).
the kth hour of operation based on forecasted wind power Pk∗ . In The breakdown of the hardware cost at the second level consists
that case, the WF has to limit its active power production to Pk∗ . of mainly the wind energy converter system (WECS) cost, which
The wind variation has to be taken into account while de- is around 80% of the hardware cost component, and the electric
termining the available/guaranteed reactive power. Based on collection and transmission system cost, which is around 12%
available meteorological data, the maximum hourly variation of the hardware cost [18]. The breakdown of the WECS cost at
of wind power generated from the forecasted value (∆Pm ax in the third level consists of mainly three components [18], which
p.u. of installed capacity of the WF) can be determined or be are:
agreed upon, which can be updated once every year or month. 1) support structure (44%);
∆Pm ax will depend on the specific site [12]. The available reac- 2) nacelle (32%);
tive power at the kth hour (Qav,k ) can now be determined from 3) rotor (24%).
(18) as follows: The electric conversion system, i.e., the power electronic con-
verter cost, and the computer and sensors cost are around 5% and
Qav,k = min{Qc,av,k , Qv ,av,k } (18) 26% of the nacelle cost, respectively [19]. Note that the reported
cost model analysis in [18] and [19] is done based on a DFIG
where
 type of WT, where the converter handles only around 30% of
Qc,av,k = (Vg Ic,m ax )2 − Pm∗2ax,k (19) the rated power. However, in the case of a full power converter
 system, the converter handles the full rated power of the WT.
2 So, we can expect that, in such cases, the converter cost can be
Vc,m ax Vg Vg2
Qv ,av,k = − Pm∗2ax,k − (20) around 15% of the nacelle cost, assuming that the converter cost
X X
varies linearly with the power rating of the converter [20].
Pm∗ ax,k = Pk∗ + ∆Pm ax (21) From the aforementioned cost figures, we can conclude that
the back-to-back full power converter (carrying only the rated
∆Pm ax ≤ Pm∗ ax,k ≤ 1. (22)
active power of the turbine) cost is around 3% of the total invest-
The available Q-capability of the WF and its comparison to the ment cost, and the converter control system cost is embedded
actual Q-capability is shown in Fig. 8 for design pf = 0.95. as a small part of the computer and sensors cost, as can be ex-
pected. The GSC cost is half of the total back-to-back converter
III. COST COMPONENTS OF REACTIVE POWER FROM WF cost, i.e., the GSC cost is 1.5% of the total investment cost of
the WF.
A. Fixed Cost Component
If the GSCs are designed to comply with the reactive power
The breakdown of the levelized production cost of a base- requirements of the WF at the grid connection point, then the
line offshore WF, equipped with doubly fed induction generator cost will increase. Note that the additional reactive power sup-
(DFIG) type wind turbines, is reported in [18]. Six main cost port is provided by the grid-side VSC. The cost of the GSC, in
components are identified in [18] as the first-level components, percentage of the total investment cost of the WF, to comply
which are: with 1.0, 0.95, and 0.9 pf requirement at the grid connection
1) hardware (including onshore transportation) (51%); point are 1.8%, 2%, and 2.25%, respectively, calculated from
2) yearly operation and maintenance (O&M) (29%); the required power rating of the converter (see Table I).
3) assembly, transport, and installation (10%); The total investment cost of some of the recently completed
4) retrofit and overhaul (7%); WFs are listed in Table II.
666 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

the total losses in the GSC for the three transistor legs become
Ploss,GSC = 3(Pc + Ps,T + Ps,D ). (26)

Fig. 9. Model of the transistor (T) and the reverse diode (D) of each switch
Equation (26) can also be written as
for the loss modeling of the converter. 2
Ploss,GSC = al Irm s + bl Irm (27)
s
TABLE III where al and bl are the power module constants and can be
EXTRACT FROM SEMIKRON POWER MODULE SKIIP 1803GB172-3DW DATA
SHEET [26]
expressed as
√  
6 2 EON + EOFF Er r
al = VIGBT + fsw + fsw (28)
π IC ,nom IC ,nom
bl = 3rIGBT . (29)
The increased active power loss (∆Ploss ) due to the increased
reactive power production can be expressed as
∆Ploss = al (Irm s,1 − Irm s,0 ) + bl (Irm
2
s,1 − Irm s,0 )
2
(30)
where Irm s,1 and Irm s,0 are the rms values of the current with
and without the increased reactive power, respectively, and can
be calculated as follows:
B. Cost of Losses 
P02 + Q2m an
The additional active power losses incurred by the WF be- Irm s,0 = (31)
Vg
cause of increased reactive power demand from the ISO need to 
be calculated for financial compensation purpose, if it accounts P12 + Q21
Irm s,1 = . (32)
for a significant value. In cases where the reactive power de- Vg
mand is met by the grid-side VSC of the built-in back-to-back In (31) and (32), P0 and P1 are the active power produc-
converter of WTs, as considered in this paper, the additional ac- tion before and after the increased reactive power production
tive power losses due to increased reactive power supply from (P0 ≥ P1 ), respectively, Qm an is the mandatory reactive power
the WF mainly occur in the GSC switches and the resistances production set out by the grid connectivity standard or by the
of transformers/cables. grid code, and Q1 is the enhanced reactive power production
Each switch of the GSC consists of a transistor (T) and a (Q1 > Qm an ).
reverse diode (D). The losses of a converter is divided into The power module, whose parameters are shown in Table III,
switching losses and conduction losses. The switching losses of can be seen as a 1-MW module when connected to a 690-V
the transistors are due to turn-on and turn-off losses, while for grid and 900 A is taken out of it. The losses are calculated as a
diodes, these are due to the turn-off losses [24]. The turn-on and percentage of the rated power of the converter for the following
turn-off losses for transistors and diodes can be found from data scenarios, assuming unity pf design case and Vg = 1.0 p.u.:
sheets. The conduction losses arise from the current flowing 1) rated operation, unity pf. Loss = 1.15%;
through the transistors and diodes. 2) 50% rated operation, unity pf. Loss = 0.52%;
For the purpose of modeling of losses, the transistors and 3) 50% rated operation, 0.9 pf. Loss = 0.58%.
diodes can be modeled as constant voltage drops (VC E O and The losses in the power module increases by 12% when 25%
VT O ) with series resistances (rC E and rT ; see Fig. 9) [25]. The reactive power is supplied at 50% rated operation in scenario 3
calculation of losses of a typical industrial power module used as compared to scenario 2. The increase in active power loss is
in WTs is presented in [25]. Different parameters of the power 0.06% of the rated or the base value. On the other hand, when
module are shown in Table III [26]. the reactive power demand is met by reducing the active power
The conduction and switching losses can be expressed as [25] production (reducing the active power production to 50% of the
√
2 2 rated value from the 100% operation to provide 25% reactive
2
Pc = VIGBT Irm s + rIGBT Irm s (23) power), the power loss decreased by 0.57% of the rated or the
π
√ base value.
2 2 Irm s As the power module will be used in a parallel configuration
Ps,T = (EON + EOFF ) fsw (24)
π IC ,nom in a multimegawatt WT, the percentage or the p.u. losses in the
√ GSC, on the WT base quantities, will be the same. This value
2 2 Irm s
Ps,D = Er r fsw (25) will increase or decrease depending on the amount of reactive
π IC ,nom
power supply from the module, as can be expected. Although
where Pc is the conduction losses in Ts and Ds, Ps,T is the not significant, a certain amount of increased active power loss
switching losses in Ts, Ps,D is the switching losses in Ds, Irm s can be incurred by the WF or can be reduced by the WF, while
is the rms value of the sinusoidal current to the grid or the ma- fulfilling the ISO’s request to increase the reactive power pro-
chine, VIGBT = VC E O ≈ VT O , and rIGBT = rC E ≈ rT . Now, duction from the mandatory value (grid code requirement), with
ULLAH et al.: WIND FARMS AS REACTIVE POWER ANCILLARY SERVICE PROVIDERS—TECHNICAL AND ECONOMIC ISSUES 667

Fig. 10. Concept of LOC component calculation for a WF.

or without reducing the present active power production. The highest limit on the active power production (Plim ,k ). In other
cost of this power loss can be considered by the ISO while words, to fulfill the ISO reactive power demand of Qr,k , which
formulating mechanisms for reactive power procurement from is ∆Qk in excess of the available reactive power capability, the
WFs. WF may lose ∆Pk active power production, where
∆Pk = Pm∗ ax,k − Plim ,k . (34)
C. Opportunity Cost Component
If the WF has to decrease its active power generation, and ∆Pk is the amount of the lost active power production that the
hence forgo market opportunity in order to fulfill ISO request WF should be compensated for in the form of LOC.
for additional reactive power, then the WF should receive a During the operational planning phase, prior to the kth hour,
payment to cover its LOC. Conventional generators receive such the ISO can calculate the LOC in the following way:
payment in some electricity markets [27]. LOC = ρk ∆Pk (35)
Because of the high degree of variability of wind speed during
an hour, it is difficult for the WF to guarantee the provision where ρk is the spot price of the electricity at hour k. Note that,
of a certain amount of reactive power to the ISO based on the since the WF does not save any fuel cost by reducing its active
forecasted power. The maximum hourly variation of wind power power production, the LOC should only reflect the revenue lost
from historical data needs to be considered, and the method of due to the nondelivered energy.
calculating the available amount of reactive power considering In the context of the operation of the Swedish power sys-
the hourly wind variation has been introduced in Section II-D. tem, and the energy balance service market functioning [28],
Furthermore, it would be difficult for the ISO to determine at the end of the kth hour of operation, when the WF genera-
the LOC for a WF, since the ISO will not have information on tion imbalance Pimb,k and the market regulation state (up- or
whether the wind speed was naturally low (so that the WF could down-regulation) are known, then the account for the LOC can
provide enhanced reactive power supply) or it acted on ISO’s be settled as
request for increased reactive power supply. In this section, a LOC = ρk ∆Pk (36)
procedure to calculate the LOC for a WF is proposed.
When the forecasted wind power is Pk∗ at the kth hour of op- where ρk can be equal to the spot price ρk , the up-regulation
eration, the actual Q-capability is Qac,k (see Fig. 10). However, price ρup,k or the down-regulation price ρdn,k depending on the
this much Q cannot be uniformly guaranteed by the WF dur- WF imbalance Pimb,k and the market regulation state (up- or
ing the whole kth hour of operation. Due to the high degree of down-regulation).
variability of wind, the actual production of the WF can exceed Now, consider the situation when the WF imbalance Pimb,k is
the forecasted value during the hour. Taking such variations into positive, i.e., the WF produces more energy than scheduled. The
account, a value ∆Pm ax can be estimated, which can, for in- excess energy should be sold to the balance market at a price
stance, be based on previous meteorological data. The kth hour lower or equal to the spot market price if the system is down-
wind power forecast Pk∗ can now be immediately converted to regulated or up-regulated, respectively [28]. In the absence of
a forecasted maximum wind power Pm∗ ax,k , where the production ceiling from the WF (Plim ,k ), the WF imbalance
Pimb,k could have been more positive than it is. Now, if the
Pm∗ ax,k = min{PR , Pk∗ + ∆Pm ax }. (33)
system was on up-regulation at hour k, then the WF should
The available reactive power from the WF at the kth hour of receive the price ρk , while if down-regulation was ordered, the
operation is Qav,k , considering the hourly wind variation, as WF should receive ρdn,k .
shown in Fig. 10. Now, if the reactive power demand from the For the other situation when the WF imbalance Pimb,k is
ISO is Qr,k , which is in excess of Qav,k , then it will impose a negative, i.e., the WF produces less energy than scheduled, the
668 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

TABLE IV tion cost is 5% of the computer and sensor cost, i.e., 0.27% of
ENERGY PRICE ρk TO BE USED IN LOC ACCOUNT SETTLEMENT
the total investment cost (from Section III-A). Considering the
recently completed WF project Lillgrund WF as an example (see
Table II), such a modification would cost around 6750 $/MW
installed capacity of the WF. Assuming an economic lifetime
of 20 years and the interest rate of 5% per year, the cost of
such a modification is around 0.06 $/MW installed/h (based on
continuous compounding [29]).
So, in this case, the cost components for reactive power ser-
vice from the wind turbine consist of a small fixed cost compo-
nent a0 covering the software modification cost, a cost of loss
component covering the cost of increased active power losses
in the converter switches due to the increased reactive power
production, and an opportunity cost if the WF has to reduce
active power generation, and hence forgo market opportunity
in order to fulfil the ISO request for additional reactive power.
The cost function for case 1 (C1 ), which the ISO can use during
Fig. 11. (a) Fixed cost and cost of loss of reactive power production from the the reactive power procurement phase to determine the possible
WF as a function of produced reactive power within the available reactive power
limit. (b) LOC together with the fixed cost and the cost of loss as a function
payment toward WFs for the reactive power service, becomes
of produced reactive power in excess of the available capacity. The figures are
drawn for P k∗ = 0.5 p.u. and ∆P m a x = 0.2 p.u. assuming V g = 1.0 p.u. and C1 = a0 + ρk ∆Ploss,k + ρk ∆Pk (37)
ρk = 60 $/MW·h, when Q m a n = 0.0 and the design pf is 0.95.
where a0 is the fixed cost (in dollars per hour) that the WF will
spend on software modification, ∆Ploss,k is the increased active
additional energy should be bought from the regulation market power losses in the converter switches due to the increased
at a price higher than or equal to the spot market price if the reactive power production above the mandatory requirement
system is up-regulated or down-regulated, respectively. In the Qm an [can be found from (30)], ∆Pk is the lost active power
absence of the production ceiling from the WF (Plim ,k ), the WF production of the WF to fulfill the ISO request for additional
imbalance Pimb,k could have been less negative than it is. If the reactive power support, and ρk is the electricity market price (in
ISO calls for up-regulation, then the WF should pay the price dollars per megawatt hour). At the end of the hour in question,
ρup,k . If down-regulation is ordered by the ISO during that hour the LOC payment can be made using the method described in
of operation, the WF should pay ρk . These are summarized in Section III-C.
Table IV.
Fig. 11 shows the total cost of reactive power production from B. Case 2: Mandatory 0.95 pf Requirement From ISO
the WF for different reactive power production levels for one
particular case. Fig. 11(a) shows the cost of loss component The reactive power capability of a WF is shown in Fig. 5,
together with the fixed cost within the range of Qm an ≤ Q ≤ when the grid code requires the WF to operate at leading/lagging
Qav,k , and Fig. 11(b) shows the total cost taking the LOC com- 0.95 pf at rated operation. As can be seen from the figure,
ponent into account. Note the large LOC component that is additional reactive power production from the WF is possible
due to the relative flatness of the capability diagram of the WF during low wind speed conditions with control modifications.
in the converter-voltage-limited zone (see Fig. 8). It is important Although hardware modifications (enhancing the GSC power
to point out that the calculation of LOC component based on rating, see Table I) are needed in this case, as compared to
the “forecasted maximum wind power (Pm∗ ax,k )” may result in case 1, the cost will not appear in the reactive power cost model.
overcompensation for the WF owner since active power produc- The hardware modifications are needed to fulfill the grid code,
tion may need to be reduced only for a fraction of an hour in and hence, this cost will appear in the active power generation
order to produce enough reactive power. cost of the WF, and eventually affects its energy market bids,
or sell/purchase contract prices. Only the control modification
IV. COST MODEL OF REACTIVE POWER SUPPLY FROM WFS cost will appear as the fixed cost component of reactive power,
as in case 1. The other components of cost, as described in case
A. Case 1: Zero Reactive Power Requirement From ISO 1, are still applicable here. Therefore, the cost function C2 can
In this case, the wind turbine is able to inject reactive power be expressed as
into the grid, in addition to fulfilling the grid requirement, with-
C2 = a0 + ρk ∆Ploss,k + ρk ∆Pk . (38)
out any significant hardware modifications (see Fig. 4). It only
requires to modify the control of the converter, i.e., the software
modification, which is not a significant cost, as mentioned ear- C. Case 3: Enhanced WF Reactive Power Capability
lier. In Section III-A, it is mentioned that the converter control In order to enhance the reactive power capability over and
system cost is embedded as a small part of the computer and above the leading/lagging 0.95 pf grid requirement, additional
sensor cost. Let us assume that the control/software modifica- hardware cost is required compared to case 2. If the GSCs
ULLAH et al.: WIND FARMS AS REACTIVE POWER ANCILLARY SERVICE PROVIDERS—TECHNICAL AND ECONOMIC ISSUES 669

where f0 is the fixed cost offer covering the necessary software

and hardware modification costs (equivalent to a0 and b0 ) in
dollars per hour, m2 is the cost of loss offer in the region Qm an <
Q ≤ Qav,k in $/mega volt ampere reactive-hour (MVAr-h), and
m3 is the LOC offer in the region Qav,k <Q in $/MVAr2 -h.
Coefficients m2 and m3 can be determined from Fig. 11. As
can be seen from the figure, although the cost of losses and
Fig. 12. Composite offer structure from a WF including fixed cost, cost of
the LOC are not exactly linear and quadratic function of the
loss, and LOC offers. reactive power, respectively, but they can be fitted with such
functions, and coefficients m2 and m3 can be determined. Based
on these values, the WF prepares its reactive power offer in the
are designed to fulfill the leading/lagging 0.95 pf requirement, deregulated electricity market. In the absence of a reactive power
then the GSC power rating needs to be overrated by 37% (see market, the proposed cost model can be used to calculate the
Table I). The overrated GSC costs around 0.05 million (M) possible payment toward a WF for the reactive power support.
$/MW installed capacity. Now, if the converters are designed to
provide enhanced reactive power (leading/lagging 0.9 pf opera- V. INTEGRATION OF THE WF Q-CAPABILITY IN SHORT-TERM
tion, case 3), then the converter cost increases to 0.055M $/MW SYSTEM OPERATIONS
installed capacity. Therefore, the difference in the cost—
Issues related to dependance of active and reactive power gen-
5000 $/MW installed—is due to the enhanced reactive power
eration on wind speeds in the dispatch stage and the provision
capability of the GSCs above leading/lagging 0.95 pf require-
of reactive power as an ancillary service need to be examined in
ment. This converts to 0.05 $/MW installed capacity/h (contin-
a comprehensive modeling framework using an optimal power
uous compounding assuming an economic lifetime of 20 years
flow (OPF) type model.
and 5% per year interest rate). Similarly, if the mandatory re-
Various reactive power cost components were formulated in
active power requirement from the ISO is zero and the WF is
Section IV, applicable to WFs supplying reactive power to the
designed at pf = 0.95, then the required hardware modification
ISO while connected to the grid with different connectivity stan-
cost becomes 0.06$/MW installed capacity/h. In addition, if in-
dards. These cost components will result in payment functions
creased reactive power capability is to be utilized during low
that the ISO will use to formulate mechanisms for financial
wind speed conditions, the previously mentioned software cost
compensation for reactive power provision to the WFs.
will also apply. The cost model will be similar to (38), except
for another fixed cost term, which denotes the cost of supple- A. System Operation Assumptions
mentary hardware for enhanced Q-capability. The cost function
C3 in this case can be expressed as In this paper, it is assumed that the market is dominated by
bilateral contracts and the generators are self-dispatched. There
C3 = a0 + b0 + ρk ∆Ploss,k + ρk ∆Pk (39) also exists a power exchange (Nord Pool) for power trading from
day-ahead to hour-ahead operations. Generators and customers
where b0 (in dollars per hour) is the cost of supplementary can optionally participate in Nord Pool. This is similar to the
hardware for enhanced Q-capability. Swedish and other Nordic market structures [28]. However, this
model can be easily adopted to the pool-type markets (including
D. Generalized Structure locational marginal pricing (LMP) markets) found in several
countries, without much difficulty. The ISO receives the dispatch
Three different cost components have been identified in this
information from the gencos, and these are held constant by
section associated with the reactive power support provision
the ISO during the real-time operation. The ISO ensures that
from WFs. A generalized structure of the cost model based
the system is secure and that a state of balance between the
on these components is shown in Fig. 12 in order to be able
production and the generation exists throughout the network.
to formulate a reactive power offer structure from WFs that
Moreover, the ISO is also responsible for compensating the
suits the deregulated markets. A reactive power offer structure
system transmission losses. Over and above this, the ISO seeks
from generators in the deregulated market is presented in [30]
to manage these responsibilities in a cost-effective manner. In
and [31], where their offers represent different cost components
this arrangement, it is assumed that the WF dispatch is cleared in
associated with the reactive power service. The reactive power
the Nord Pool market auction. Note that, due to the high degree
offer structure from WFs based on different cost components can
of variability of wind power, WFs will bid in the market at zero
be fitted within such a framework (see Fig. 12). The composite
price to clear all the produced energy [1].
offer structure for reactive power service can be expressed as

C = f0 ∀0.0 ≤ Q ≤ Qm an B. ISO Operation and Q-Dispatch From WF

= f0 + m2 (Q − Qm an ) ∀Qm an < Q ≤ Qav,k In real time, the ISO’s objective is to minimize its total elec-
tricity transmission cost while maintaining network reliability
= f0 + m2 (Qav,k − Qm an )
and security. To do so, the ISO minimizes the transmission losses
+ m3 (Q − Qav,k )2 ∀Qav,k < Q (40) by properly scheduling bus voltages/reactive power exchange,
670 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

However, in systems with high wind power penetration, they

will make difference. At this forecasted wind, the actual reactive
power Qac,k is 0.39 p.u. [from (15)]. However, the available
reactive power (Qav,k ) considering the maximum hourly wind
power variations is 0.37 p.u. when ∆Pm ax is 0.2 p.u., and is
0.3 p.u. when ∆Pm ax is 0.5 p.u. In either case, the required
reactive power dispatch request from the ISO (Qr,k ) is less than
the available reactive power capability of the WF. The WF will
receive a payment covering the fixed cost amounting 300a0 $/h
(18 $/h) and a payment covering the cost of losses amounting
300ρk ∆Ploss,k $/h (6 $/h), i.e., 24 $/h in total for this service.
These calculations are also done for a high-load scenario.
When the ISO procures the reactive power service from the
WF, then the losses are minimized and the WF voltage is
scheduled at 0.98 p.u. The WF injects 0.3 p.u. reactive power.
The total system losses are 438 MW. The actual reactive power
Fig. 13. Cigré Nordic 32-bus test system augmented with a WF at the 400 kV
capability of the WF is 0.32 p.u. However, considering 0.2 p.u.
grid at bus 4045 (only a section of the system is shown in the figure). maximum hourly wind power variation (which is a site-specific
value), the available capability reduces to 0.3 p.u. So, the WF
dispatches the scheduled reactive power without reducing the
utilizing the available reactive power resources including gen-
active power production. The expected payment toward the WF
erators, shunt capacitors, WFs, etc.
for this service is 18 $/h for the fixed cost and 36 $/h for the
The ISO’s objective function J can be defined as follows:
 cost of losses, i.e., 54 $/h in total. Note the increased cost of
J = 0.5 (Gi,j × (Vi2 + Vj2 − 2Vi Vj cos(δj − δi ))) (41) losses due to the increased reactive power production compared
i,j to the previous case.
If the maximum hourly wind power variation is taken as
where i, j are the indices for buses, G is the conductance of line
0.5 p.u. (a site with a high degree of wind variations), then the
(in p.u.), V is the bus voltage (in p.u.), and δ is the voltage angle
available reactive power capability reduces to 0.27 p.u. which
(in radians). The other important constraints associated with the
is less than the requested value. In this case, the WF has to put
ISO’s real-time operation are as follows:
a ceiling on its production (Plim ,k ), which is 0.62 p.u. The ex-
1) reactive power generation limit from WF (18)–(22);
pected active power production loss due to this production ceil-
2) active and reactive power balance equations for each bus,
ing is ∆Pk = 0.25 + 0.5 − 0.62 = 0.13 p.u. So, in this case, the
i.e., the load flow equations;
WF will receive an LOC payment, in addition to the fixed cost
3) voltage limits at all buses;
and the cost of losses payment, which is 2340 $/h. The sig-
4) active and reactive power limits for conventional
nificantly higher payment to the WF is because of the higher
generators.
wind power forecasting error, i.e., larger hourly wind power
variations.
C. Case Study
The well-known Cigre Nordic 32-bus test system [32] is used
VI. CONCLUSION
in this paper to demonstrate a case study (see Fig. 13). A WF of
300 MW installed capacity is connected at bus 4045, which is Reactive power/voltage control support, temporary primary
typically in the southeastern part of Sweden where several large frequency control support, and power oscillation damping sup-
offshore WFs are being planned. port are some of the functions that can be provided by modern
In this case study, it is assumed that the mandatory reactive WFs equipped with VSWTs with power electronic converters.
power requirement from the ISO is zero and the WF design pf This paper presents a new approach to model the reactive power
= 1.0. The optimization model, which is a nonlinear program- capability of WFs connected to the grid under different grid
ming problem, is modeled in general algebraic modeling system codes taking the wind variations into account. The cost func-
(GAMS) and solved using mixed integer non-linear optimiza- tions developed in each case have different components for fixed
tion solver (MINOS) [33]. cost, cost of loss, and an opportunity component. A new method
When the ISO does not procure the reactive power service for calculating the LOC for a WF is proposed that includes the
from the WF, i.e., the ISO does not schedule the WF bus voltage, effect of wind speed variations. The developed cost model is
then the system losses are 432.55 MW. The forecasted wind generic, and can be used to formulate the reactive power offer
power Pk∗ is 0.25 p.u. When the ISO utilizes the reactive power structure from the WF that fits any market structure. Finally,
capability of the WF, then the scheduled voltage at the WF bus the integration of the reactive power capability model into the
becomes 0.97 p.u. and the WF injects 0.13 p.u. reactive power. short-term system operations is demonstrated using the Cigré
The losses reduce to 432.51 MW. The reduction in losses is not Nordic 32-bus test system.
so significant due to the fact that a single reactive power source It is concluded that GSCs designed to handle only rated active
may not make a large difference to losses in a large system. power cost around 1.5% of the total investment cost of the WF.
ULLAH et al.: WIND FARMS AS REACTIVE POWER ANCILLARY SERVICE PROVIDERS—TECHNICAL AND ECONOMIC ISSUES 671

However, a 50% overrated GSC would cost around 2.25% of [11] K. Bhattacharya and J. Zhong, “Reactive power as an ancillary service,”
the total investment cost of the WF, and would be capable of IEEE Trans. Power Syst., vol. 16, no. 2, pp. 294–300, May 2001.
[12] M. Ahlstrom, L. Jones, R. Zavadil, and W. Grant, “The future of wind
providing 0.65 p.u. reactive power at the grid connection point forecasting and utility operations,” IEEE Power Energy Mag., vol. 3,
under nominal conditions. no. 6, pp. 57–64, Nov./Dec. 2005.
Among the three cost components associated with the reac- [13] B. Ernst, B. Oakleaf, M. L. Ahlstrom, M. Lange, C. Moehrlen, B. Lange,
U. Focken, and K. Rohrig, “Predicting the wind,” IEEE Power Energy
tive power supply from the WFs, the cost of loss component may Mag., vol. 5, no. 6, pp. 78–89, Nov./Dec. 2007.
exceed the fixed cost component depending on the amount of re- [14] P. Kundur, Power System Stability and Control. New York: McGraw-
active power production. However, due to the relative flatness of Hill, 1993.
[15] Grid code: High and extra high voltage, E.ON Netz GmbH, Bayreuth,
the capability curve of the WF in the converter-voltage-limited Germany, Tech. Rep., Apr. 2006 (Status 1).
region, the LOC component is relatively larger than the other [16] Nordic grid code 2007 (Nordic collection of rules), Nordel, Oslo, Norway,
two components. It is also due to the fact that the WF does Tech. Rep., Jan. 2004 (updated 2007).
[17] K. Zhou and D. Wang, “Relationship between space-vector modulation
not save any fuel cost by reducing its active power production and three-phase carrier-based PWM: A comprehensive analysis,” IEEE
unlike the conventional generators. The relative sizes of the cost Trans. Ind. Electron., vol. 49, no. 1, pp. 186–196, Feb. 2002.
components depend on the forecasted wind power, wind speed [18] M. B. Zaaijer, H. B. Hendriks, S. Herman, A. Winnemuller, T. Topper,
R. van der Berg, and W. op den Velde, “Optimization through conceptual
prediction error, and the grid voltage at the point of connection. variation of a baseline wind farm,” presented at the Marine Renewable
Another finding from the study is that higher wind speed Energy Conf. (MAREC), Newcastle Upon Tyne, U.K., Sep. 2002.
prediction errors, i.e., a WF site with high degree of wind vari- [19] S. Herman, “DOWEC cost model—Implementation,” Energy Res. Centre
The Netherlands, Petten, The Netherlands, Tech. Rep. ECN-CX-02-048,
ations, may result in higher payments to WFs for the reactive Dec. 2002.
power service, mainly due to the increased LOC component. [20] J. T. G. Pierik, M. Pavlovsky, J. Bozelie, P. Bauer, and S. W. H. de Haan,
For the investigated case considered, the expected payment to “DOWEC electrical system baseline design,” Energy Res. Centre The
Netherlands, Petten, The Netherlands, Tech. Rep. ECN-CX-02-021, Feb.
the WF for a 0.3 p.u. reactive power service is 54 $/h consid- 2002.
ering 0.2 p.u. wind power prediction error, while the payment [21] (2008, Jan.). Facts about The Horns Rev Wind Farm [Online]. Available:
increases to 2394 $/h when the prediction error is 0.5 p.u. http://www.hornsrev.dk/Engelsk/default_ie.htm
[22] (2008, Jan.). Fakta om Lillgrund (in Swedish) [Online]. Available: http://
www.vattenfall.se/
ACKNOWLEDGMENT [23] (2008, Jan.) Våre vindparker [Online]. Available: http://www.statkraft.no/
pub/vindkraft/Kraftverkene/index.asp
The authors would like to express their gratitude to D. Ander- [24] N. Mohan, T. Undeland, and W. Robbins, Power Electronics Converters,
sson, Å. Juntti, and C. Andersson from E.ON Elnät Sverige AB, Applications and Design, 3rd ed. Hoboken, NJ: Wiley, 2003.
[25] A. Petersson, “Analysis, modeling and control of doubly-fed induction
J. Svensson from E.ON Vind Sverige AB, Prof. O. Samuelsson generators for wind turbines,” Ph.D. dissertation, Dept. Energy Environ.,
from Lund University, Faculty of Engineering, Lund, Sweden, Chalmers Univ. Technol., Göteborg, Sweden, 2005.
and D. Karlsson from Gothia Power AB, Göteborg, Sweden, for [26] (2008, Jan.). SKiiP 1803GB172-3DW power module data sheet [Online].
Available: http://www.semikron.com/internet/ds.jsp?file=1264.html
their valuable technical assistance. [27] J. Zhong and K. Bhattacharya, “Reactive power management in dereg-
ulated electricity markets—A review,” in Proc. IEEE Power Eng. Soc.
REFERENCES Winter Meeting, New York, Jan. 2002, pp. 1287–1292.
[28] (2007, Jan.). [Online]. The Swedish electricity market and the role of
[1] E. A. DeMeo, W. Grant, M. R. Milligan, and M. J. Schuerger, “Wind Svenska Kraftnät, Vällingby, Sweden. Available: www.svk.se
plant integration,” IEEE Power Energy Mag., vol. 3, no. 6, pp. 38–46, [29] R. A. Brealey, S. C. Myers, and F. Allen, Corporate Finance, 8th ed.
Nov./Dec. 2005. New York/Homewood, IL: McGraw-Hill/Irwin, 2006.
[2] R. Piwko, N. Miller, J. Sanchez-Gasca, X. Yuan, R. Dai, and J. Lyons, “In- [30] J. Zhong and K. Bhattacharya, “Toward a competitive market for reactive
tegrating large wind farms into weak power grids with long transmission power,” IEEE Trans. Power Syst., vol. 17, no. 4, pp. 1206–1215, Nov.
lines,” in Proc. IEEE/PES Transmiss. Distrib. Conf. Exhib.: Asia Pacific, 2002.
2005, pp. 1–7. [31] I. El-Samahy, K. Bhattacharya, C. Cañizares, M. Anjos, and J. Pan, “A
[3] A. Yazdani and R. Iravani, “A neutral-point clamped converter system procurement market model for reactive power services considering system
for direct-drive variable-speed wind power unit,” IEEE Trans. Energy security,” IEEE Trans. Power Syst., vol. 23, no. 1, pp. 137–149, Feb. 2008.
Convers., vol. 21, no. 2, pp. 596–607, Jun. 2006. [32] K. Walve, “Nordic32-A CIGRÉ test system for simulation of transient
[4] M. Chinchilla, S. Arnaltes, and J. C. Burgos, “Control of permanent- stability and long term dynamics,” Svenska Kraftnät, Vällingby, Sweden,
magnet generators applied to variable-speed wind-energy systems con- Tech. Rep., 1993.
nected to the grid,” IEEE Trans. Energy Convers., vol. 21, no. 1, pp. 130– [33] A User’s Guide, GAMS Development Corporation, Washington, DC,
135, Mar. 2006. GAMS release 2.5, 1999.
[5] G. N. Bathurst, J. Weatherill, and G. Strbac, “Trading wind generation
in short term energy markets,” IEEE Trans. Power Syst., vol. 17, no. 3,
pp. 782–789, Aug. 2002.
[6] J. Matevosyan and L. Söder, “Minimization of imbalance cost trading Nayeem Rahmat Ullah (S’04–M’09) received the
wind power on the short-term power market,” IEEE Trans. Power Syst., B.Sc. degree in electrical and electronic engineering
vol. 21, no. 3, pp. 1396–1404, Aug. 2006. in 2002 from Bangladesh University of Engineering
[7] N. R. Ullah and T. Thiringer, “Variable speed wind turbines for power and Technology (BUET), Dhaka, Bangladesh, and
system stability enhancement,” IEEE Trans. Energy Convers., vol. 22, the M.Sc. and Ph.D. degrees in electrical engineering
no. 1, pp. 52–60, Mar. 2007. in 2004 and 2008, respectively, from Chalmers Uni-
[8] N. R. Ullah, T. Thiringer, and D. Karlsson, “Voltage and transient stability versity of Technology, Göteborg, Sweden.
support by wind farms complying with the E.ON Netz grid code,” IEEE During summer 2007, he was a Visiting Scholar at
Trans. Power Syst., vol. 22, no. 4, pp. 1647–1656, Nov. 2007. the University of Waterloo, Waterloo, ON, Canada.
[9] N. R. Ullah, T. Thiringer, and D. Karlsson, “Temporary primary frequency After completing his Ph.D., he joined the Asset Man-
control support by variable speed wind turbines—Potential and applica- agement Department of E.ON Elnät AB, Malmö,
tions,” IEEE Trans. Power Syst., vol. 23, no. 2, pp. 601–612, May 2008. Sweden, as a Network Planning Engineer. His current research interests in-
[10] S. Hao and A. Papalexopoulas, “Reactive power pricing and management,” clude grid integration of wind energy converters.
IEEE Trans. Power Syst., vol. 12, no. 1, pp. 95–104, Feb. 1997. Dr. Ullah is a member of the IEEE Power and Energy Society.
672 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 3, SEPTEMBER 2009

Kankar Bhattacharya (M’95–SM’01) received the Torbjörn Thiringer (S’92–M’96) received the Ph.D.
Ph.D. degree in electrical engineering from the Indian degree from Chalmers University of Technology,
Institute of Technology-Delhi, New Delhi, India, in Göteborg, Sweden, in 1996.
1993. He is currently is a Professor in the Division of
During 1993–1998, he was a faculty member at Electric Power Engineering, Department of Energy
Indira Gandhi Institute of Development Research, and Environment, Chalmers University of Technol-
Mumbai, India. During 1998–2002, he was with the ogy, Göteborg, Sweden. His current research inter-
Department of Electric Power Engineering, Chalmers ests include power electronic applications for wind
University of Technology, Göteborg, Sweden. Since energy, automotive, and telecom applications as well
January 2003, he has been with the Department of as grid integration of wind energy converters. Prof.
Electrical and Computer Engineering, University of Thiringer is a member of the Cigré B4-39 Group
Waterloo, Waterloo, ON, Canada, where he is currently a Professor. His current “Large-scale integration of wind energy using power electronics and HVDC.”
research interests include power system dynamics, stability and control, eco-
nomic operations planning, electricity pricing, and electric utility deregulation.