IEEE PEDS 2005

A Calculation Method of the Total Efficiency

of Wind Generator
Akira Inoue, Mohd. Hasan Ali, Rion Takahashi, Mamoru Kimura, Moto-o Futami, Masaya Ichinose,
Toshiaki Murata, and Junji Tamura and Kazumasa Ide
Dept. of Electrical and Electronic Engineering Hitachi Laboratory
Kitami Institite of Technology Hitachi,Ltd., 1-1 Omika-cho,7-chome,Litachi-shi,Ibaragi-
Kitami, 090-8507, Hokkaido, Japan ken, 319-1292, Japan
rotl 5pullout.elec.kitami-it.ac.jp
Abstrac This paper presents a method to represent various
losses in wind generator as a function of wind speed, which is II. CALCULATION METHOD
based on the steady state analysis. By using the proposed method,
wind turbine power, generated power, copper loss, iron loss, A. Outline of the calculation method
stray load loss, mechanical losses and energy efficiency can be Induction generator is widely used as WG due to its low cost,
calculated quickly. Calculation was done using a wind speed data low maintenance and direct grid connection. So, in this paper
expressed by Probability Density Function in order to predict we consider induction generator for efficiency calculation.
capacity factor and total efficiency of a wind farm for one year. However, there are several problems regarding the induction
Finally it is concluded that the proposed prediction calculation generator as given below.
method is effective, and can contribute to a wind generator
design and construction planning of a wind farm. * Usually the input, output and loss conditions of
induction generator can be determined from rotational
Keywords-Windc generator;
Ef.ficiency; capacityfactor.
induction genetoror; power loss; speed (slip). However, it is difficult to determine slip
from wind turbine input torque.
* Generator input torque is reduced by mechanical
I. INTRODUCTION losses, but mechanical losses are a function of
Recently Wind Generator (WG) is getting increasing rotational speed (slip). It is difficult to determine
emphasis in the world, and its capacity is becoming larger. mechanical losses and slip at the same time.
However, WG output power and losses are dependent on * It is hard to measure stray load loss and iron loss.
wind. In order to capture more energy from wind, it is
important to analyze loss characteristics of wind generator, * It is difficult to evaluate gear loss analytically as a
which can be determined from wind speed. Furthermore, since function of rotational speed.
many non-linear losses occur in WG, making prediction profit In this paper, we propose a method of calculating the
by using average wind speed may cause many errors. This efficiency of WG correctly, taking into account the points
paper presents a method to represent various losses in wind mentioned above. Table I shows the losses of Wind Generator.
generator as a function of wind speed, which is based on the The equivalent circuit of the induction generator used in the
steady state analysis. By using the proposed method, wind proposed method is shown in Fig. 1. The input torque and
turbine power, generated power, copper loss, iron loss, stray copper losses are calculated by solving the circuit equations
load loss, mechanical losses and energy efficiency can be (1).
calculated quickly. However, the proposed method cannot take
a transient state into account. Generally,* since the change of TABLE I. LOSSES OF WIND GENERATOR
the generated power of wind generator is large, it is thought
that a transient state occurs. In order to check the effect of a Gear box losses
transient state, and how much differences appear, same Mechanical Windage loss
calculation was performed by using PSCAD/EMTDC1]. It is loss
seen that the difference between the two methods is negligible. Ball bearing loss
By using a wind speed data expressed by the probability Primarywindingcopperloss
Copperary
Cop
density function, this method can predict energy produced by winding copper loss
WG and also estimate total efficiency and capacity factor. As
a whole, it can be concluded that the proposed prediction Iron loss
Eday wind loss
Eddycurrentloss
method can contribute to te design as well as construction Hysteresis loss
planning of wind farm properly. Stray load lossl

0-7803-9296-5/05/$20.00 ©2005IEEE 1595

 2) Generator input: Generator input power can be
rl xI r2 ^ X2 calculated from the equivalent circuit of Fig.1 as shown
below:
1-s
I2 ( x r2) [W] (5)

v]I Is| |> xm < rm I l-s g3) 2 Copper losses and iron loss: Copper losses are
resistance losses occuring in winding coil and can be
JE } I J l calculated using the equivalent circuit resistances ri and r2' as
Wcopper =r1X11I +r2x12[W] (6)
r,=stator resistance, r2'=rotor resistance, Generally, iron loss is expressed by the parallel resistance in
x1=stator leakage reactance, X2'=rotor leakage reactance, the equivalent circuit. However, iron loss is the loss produced
rm =iron loss resistance, xm=magnetizing
r=iron loss reactance,x'agnetizing
resistance, reby the flux change, and it consists of eddy current loss and
hysteresis loss. In this paper, at first we calculate actual iron
s(slip)= (Ns-N)/Ns, N=rotor speed, Ns=synchronous speed. loss per unit volume, wf , using the flux density, as shown
Figure 1. Equivalent circuit of induction generator below.

VI r,+ x + jrx,,II+r +jr, I2 ~~~ 2

wf(lW=B CH
I +
1\lOU~~~~~~~~~~2
0 aEd (100) [W/kg] (7)
r
.x r2 ' Where, B: flux density [T], 0H: hysteresis loss coefficient,
m I + J mx m ++2j
r
0=_ Jm I2 OE: eddy current loss coefficient, f : frequency[lHz], and
rm +JXmtrm+ IXm s d: thickness of iron core steel plate[mm].
Generally, flux and intemal voltage can be related to eq.(8).
Therefore, if the number of turns of a coil is fixed, a
B. Equations for analysis proportionality holds between the flux density and the internal
1) Wind turbine power: The power captured from the voltage.
wind can be expressed as (2). In this paper, MOD-2 model(2) is E= 4.44xf xkw xwxo [V] (8)
used, whose power coefficient curve is shown in Fig. 2. This (kw: Winding coefficient, w: Number of tums, (p: flux)
turbine characteristic is non-linear expressed as eq.(4). B = Bo X T (9)
Pw,b Pwtb= =IPC ,(A2/
i)rR 2V
2 PCp(R,,B)2rR2Vw3 wI[W ](2)
] (2) (B o
(Eo : Nominal internal voltage)
)w,bR And then the iron loss resistance can be obtained with respect
Vw (3) to the internal voltage E determined by the flux density as
shown below. Wf is the total iron loss which is deternined
0.022/32 5.6)e-7r
Cp (2, /) = 0.5(r-0.022g2- (4)
5.6)e~'7~ using eq.(7) and the iron core weight.
In eqs.(2) to (4), Plb=turbine output power [W], p =air E2
density [kg/m3], Cp=Power coefficient, A =Tip speed ratio, r = (10)
R=Radius of the blade [m], Vw=wind speed [m/s], Wf
w,b=Wind turbine angular speed[rad/s], and = blade pitch
angle [deg]. 4) Bearing loss, windage loss and stray load loss:
Bearing loss is a mechanical friction loss due to the rotation of
0.5- tdie rotor, which is shown below.
r 0.4 *|=,B= deg - ([1)
Wb -KBm-[WI
2 = \ --20
I de Where KB is a paramneter concerning the rotor weight, the
.3 .

diameter of an axis, and the rotational speed of the axis.

U 0.2 .. Windage loss is a friction loss that occurs between the rotor
3 r-'-"----T and the air, and expressed as follows.
0.1R; -
"--4-- Wm =Kw ojm [M] (12)
0.0 Where Kw is a parameter determined by the rotor shape, its
0i5Tip Speed Ratio,
10
A
15 20 length, and the rotational speed.
' ~~~~~~~Stray
qrvIn
load loss is expressed as follows.
A1V'

Figure 2. Power coefficient vs Tip Speed Ratio characteristics p2

Ws= 0.005-p[W] (13)

1596
 Where, P is generated power [W], and Pn is rated power
[WI.
[W]. ... ...... ...... .........~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....
........
. E ... ,,=.. \. . ........'.. . . . ..... . .'.
.0 -

5) Gear box lossf3'41: Gear box losses are primarily due to

tooth contact losses and viscous oil losses. In general, these
losses are difficult to predict. However, tooth contact losses
are very small compared with viscos losses, and at fixed . - . - r
rotational speed, viscous losses do not vary strongly with -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
transmitted torque. Therefore, simple approximation of Slip
gearbox efficiency can be obtained by neglecting the tooth Figure 4. Slip-Torque curve
losses and assuming that the viscous losses are constant (a
fixed percentage of the rated power). A viscous loss of 1% of
rated power per step is a reasonable assumption. Thus the Windturbine Electricpower
efficiency of a gearbox with "q" steps can be computed using Conversion
eq.(14). Generally, the maximum gear ratio per step is m
mechfromni
approximately 6:1, so two or three steps of gears are typically gar energy toI
required. Where Pt is gear box output power, Pm is turbine wtb geelectricaenergy V
power and P,R is rated tubine power.

7gear Pm-(O01)qI,R 0%
100[%]
1)ls
(14)
Mechanical Stray load loss Iron loss Copper loss

Fig. 3 shows the gear box efficiency for three gear steps. In Figure 5. Expression of power flow in the proposed method
this paper, three steps are assumed, according to a large-sized Since mechanical losses and stray load loss cannot be
wind generator in recent years. expressed in a generator equivalent circuit, they have been
deducted from the wind turbine output. Fig. 6 shows the flow
chart ofthe proposed method, which is described below.
1.0 -
- ==-.= .- 7 (i) Wind velocity is taken as the input value, and from
(7/ '.~ this wind velocity all states of wind generator are
g 0.8- , calculated.

| 0.6
j '*
2 step
tep
D( Wind turbine output is calculated from eq.(2). The
synchronous angular velocity is taken as the initial
v : 3 step value of the angular velocity and wind turbine power
is multiplied- by the gear efficiency, n gear.
0.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 (X) Ball bearing loss and windage loss which are
Turbine output [pul mechanical losses are deducted from the wind turbine
3. Gear box efficiency
Figure output calculated in step (), and stray load loss is also
deducted. These losses are assumed to be zero in the
initial calculation.
C. Calculation Method
The efficutiencMethof
The efficiency of aa generator is determined using
gnertorisdetrmied te loss
sig the lss (4) At this step the slip is changed using the characteristic
ofFg. ni giin th aegnrtdpwrahthe
analysis method described above. The input, output and loss
conditions of induction generator can be determined from
of Fi.4ulated
powercalculatedistep(a. powereas
rotational speed (slip). However, it is difficult to determine slip ( By using the slip calculated in step ® and using
from wind turbine input torque. Therefore, we adopt an eq.(1), the currents in the equivalent circuit can be
iterative calculation to obtain a slip which produces torque determined, and consequently the output power,
equal to the wind turbine torque from a slip-torque curve as copper loss and iron loss can be calculated. Next, loss
shown in Fig. 4. Furthermore, it is difficult to determine Wf is calculated from the flux density using the iron
mechanical losses and slip at the same time, because loss calculation method mentioned above, and the iron
mechanical losses are a function of rotational speed (slip). loss resistance, rm, which produces the same loss as Wf
Mechanical loss can also be obtained in the iterative is also determined.
calculation. The power transfer relation in the proposed
method is shown in Fig. 5. (6j) Ball bearing loss and windage loss are calculated by
using eqs.(11) and (12), and the rotational slip of the

1597
 generator determined in step (). And stray load loss thus IG draws more reactive power and internal voltage of IG
is calculated from eq.(13). decreases. As a result, flux density and iron loss decrease.
(7) If the calculated losses converge, the calculation will
stop, otherwise it will return to step (Th TABLE II. GENERATOR PARAMETERS
rated power 5[MVA] rated voltage 6600[V]
r lwind speed input. rated frequency 60[Hz] pole number 6
stator resistance 0.0051 [pu] stator leakage 0.088[pu]
reactance
(2)Calculate power produced by wind rotor resistance 0.0091 [pu] rotor leakage 0.125[pu]
turbine, and multiply gear efficiency 77 gear reactance
iron resistance 1377.4 [pu] magnetizing
reactance
4.776[pu]
(3)Mechanical losses and stray load loss are| l
deducted from wind turbine power, 5M
resulting wind turbine output power Ptout. 4M

fl r
~3M
il /
_
~~~~~~~~~~~~~~~~~Generated
power
2M-
(Search slip from Fig. 4 and determine 1M G
generator input power Pgin l
4 6 8 10 12 14 16 18 20 22
XChange slip Wind speed [m/s]
( (g) ;n=Ptout ? . ) ]No gin'_::PtNot80k -q Copper loss
-- Iron loss
Yes
- -......Stray road loss
60k - Windage loss
Bearing loss
(5)Calculate copper loss and iron loss from --------------------

equivalent circuit with iron loss resistance 20k-

adjusted.
0-
s 4, 6 8 10 12 14 16 18 20 22
(s)Calculate mechanical losses and stray load Wind speed [m/s]
loss with state values converged.
L~~~~~~~~~~~~~~~~~~~~T I5I- 551
Bearing loss
550 __/
t (2)Calculated losses converge? 3 [ 6 8 10
_Wind 12 14
speed [m/s] 16 18 20 22
Ys
1200-1
END ~~~~~1190-1_
01180 K 1~~~~ 1180 ~~~~~~Windage loss
3 11701
Figure 6. Flow chart of the proposed method 4 6 8 10 12 14 16 18 20 22
Wind speed [m/s]

III. SIMULATION RESULTS F35.4k- E335Iokls Iron loss

The parameters of the wind generator used in this paper are 0 35.2k-\
shown in Table II. A 5MW WG was assumed. The cut-in and ' 35.0k
rated wind velocities are 5.8[m/s] and 12.0[m/s] respectively. 4 6 8 10 12 14 16 18 20 22
Moreover, the generated power of induction generator is Wind speed [m/s]
controlled by pitch controller when the wind speed is over the Figure 7. Power and various losses of induction generator
rated wind speed. Fig.7 shows the results of power and various
losses of induction generator, in which the curves for the IV COMPARSON OF SIMULATION RESULT WITH TRANSIENT
windage loss, bearing loss and iron loss are enlarged for a clear
and easy understanding. From the figures it is clear that all
PHANON CONSRED
PHENOMENON CONSIDERED
losses are non-linear with respect to the wind speed. And iron The proposed method cannot take a transient state into
loss decreases with the increase of wind speed. When wind account. Since the change of the generated power of wind
speed increases, induction generator real power increases, and generator is large, a transient state may occur. In order to check

1598
the effect of a transient state on the results, same calculation in Table III. Even if the large change of wind velocity like
was performed by using PSCAD/EMTDC. Wind velocity is wind3 of Fig. 8 is used, the difference is negligible. Moreover,
approximated by the sine waves as shown in Fig. 8, where even if the rotor resistance is small, the difference is not large.
windl indicates the wave with a time period of 20 seconds, As a result, the difference between two methods is so small that
wind2 consists of two waves with periods of 20 seconds and 5 it can be ignored. Therefore, it can be said that the proposed
seconds, and wind3 consists of three waves with time periods method of efficiency calculation of wind generator has
of 20 seconds, 5 seconds and 1 second. Moreover, in order to sufficient accuracy.
lengthen the time constant of the generator circuit calculation
was also performed by setting the resistance value as one half
and one fifth of the original value as shown in Table II. Since TABLE III CALCULATION RESULTS, (I)
gear loss is a fixed loss, it is not included here. Fig. 9 is a result Wind turbine Generated Energy
of the calculation. energy[MJl power[MJ] efficiency[%]
- 12_ PSCAD 43.88 42.40 96.64
l

8 ,indl 8
~ ~ ~ Windi 43.88
~ ~~~~~_method438
Proposed 4242
424 9667
967

PSCAD 42.04 40.07 96.81

10 ~~20[sec]
time
30 40 id
0 10 Wind2 Proposed 42.03 40.06 96.78
method
'7 12-
--Io' wind2 PSCAD 40.96 39.70 96.9
8 W Proed 39.67 96.87
g 6
t
method 40.95
0 10 20 30 40 Wind3 PSCAD 40.96 39.80 97.14
time [sec]
(r2'-F1/2) Pr40.95 39.77 97.12
Io'3M- t
3 124 \/\n / - P8CAD Wind.3
method
PSCAD 40.96 39.85
u 8 97a29
(2-/)Propoe
method
40.95 39.85 97.26
0-) 10 20 30 40
timne [see]
Figure 8. Wind speed variations V. EFFICIENCY CALCULATION USING A PROB3ABILITY
5Mme2[osec
using this functio DENSITY FUNCTION

,~t4M -.efcec
'~~~~~3M ~~~~~~PSCAD If real wind information is available, the efcec
2M calculation of wind generator can be denelytione. However,
~~~~~iM ~~~~~~~~~~~~it is difficult to express the real wind data as a function of time.
010 Therefore, the Weibull function['1 is used in this paper. By
2 3

using this function ahd the proposed method, it is possible to

,5M
04M
__
d
calculate the amount of annual
mpoder generationspefor a cific
tw
0 Figure9. Oututpowero
10 ~~~~im20sc 30 ...area.
40C
The WeibullCC function can be expressed by
2M-

time [e is w ic sale i and is wind

The time [sec] Where,
W tis shape factor, C is thefactor, and v is thespeed.
by PSCAD From the above equation, the status of wind speed can be
s3M .1 r expressed. Although the annual average wind speed for two
PSCAD2andtheMpropoed me re almostthe same AInfI
od .Aareas are the same, the probability functions for two areas can

0- ~~~~~~~~~Ibedifeen.Therefore, making predictionbaeonnul

0o1heck0 20 30 40 energy profit by using average wind speed may cause many
time [see] errors. In this paper, three areas are chosen, with different
Figure 9. Output power of IG Weibull function parameters, which are shown in Table IV,
Thedoutputpowed lines shownecycle2
iseond. The resultis shobine where A is the weak wind area in inland, B is the windy area in
the portion of a cape, and C is the windy area at the slope
Th dotted lineshich
by PCADin hic th ianFigt
show trasiet9represet timresultseobtaine
apeas a thme t= se du to parttip
of a mountain. In Table IV, c and k have been taken from
the switching process. However, after 1 sec the results of both LAWEPS(Local Area Wind Energy Prediction System)16]of
PSCAD and the proposed method are almost the same. In order NEO [M] L.AwEPS. can give usacuae wiinapxr;4fnddta A-nd dAt-a
shown in Fig. 10, and it is clear that the wind conditions of factor of area B is higher than that of area C. So, the amount of
these areas are different. In this case, the cutout wind velocity power generation strongly depends on probability density
is assumed to be 20 m/s [2]. distribution. The difference of copper loss and stray load loss
among the three areas are large, because they depend not only
on the generated power but also on the operating time, which is
related with probability density distribution. The differences of
area Ave.wind speed k mechanical losses among the areas are not so large, because
area_________ [m/s] c
c
they mainly depend on operating time. The influence of
A Kyoto city 4.2 6.0 1.9 gearbox loss is comparatively greater in weak wind area, such
B Erimo cape 9.0 | 10.0 | 2.2 | as in area A. Therefore, the total efficiency of area A is less
C Mt.Fuji's slope 10.5
±
9.5 1.6
~~~~~than90%.

0.14-
VI. CONCLUSION
0.12- In this work, we can express the losses and output power of
0.10 Erimo wind generator as a function of wind speed. We adopt Weibull
008 ---- Mt.Fuji's slope function to express the status of wind speed in the calculation
.. Kyoto city of annual energy in wind generation. From the simulation
8 0.06- results it is seen that by using the proposed method it is
0.04- .' / "$<possible to calculate generated power, the losses, total energy
0.02 .. / -- efficiency and capacity factor quickly. It is also clear that the
0.00 [/ . . > : total efficiency of a WG depends not only on the parameters of
10 1d5 20 the generator but also on the situation of wind. The proposed
25 method is effective for improving wind generator design and
construction planning ofwind farm.
Figure 10. Probability density distribution In our future study we would like to establish efficiency
The amount of annual power generation can be calculated
calculation method of wind generator using Permanent Magnet
Synchronous Generator (PMSG).
from the following equation using the probability distribution
function.
REFERENCES
1' max
E oal = Pg (v)x f(v)x 8670dv [1] http://wvw.pscad.com/index.html [online]
I' (16) [2] P.M.Anderson, Anjan Bose,: "Stability Simulation of Wind Turbine
Where, E is annual generated power [Wh], Pg is generated
Where, ia ag rSystems", IEEE Transaction on Power Apparatus and Systems,
Vol.PAS-102,No.12,pp.3791-3795,Dec 1983.
power [W], V,,. is cut out wind speed [m/s], and Vmin is cut in [3] Gary L. Johnson: WIND ENERGY SYSTEMS Electronic Edition
wind speed [m/s]. Moreover, the loss energy and the wind http://w-ww.rpc.com.au/products/windturbines/wind-
turbine energy can be calculated by inserting each equation into book/WindTOC.html [online]
eq.(16). Furthermore, the capacity factor and the total [4] J. Cotrell : "A Preliminary Evaluation of a Multiple-Generator
efficiency are calculated from the following formulas. Drivetrain Configuration for Wind Turbines" , National Renewable
Energy Laboratory, NREL/CP-500-31 178, January (2002).
Total Efficiency = E,1,,,, [Wh] x100[%] [5] Application J.F.Manwell, J.G.McGowan, A.L.Rogers, "WIND
Wind Turbine Energy [Wh] (17) ENERGY EXPLAINED Theory Design and Application", JOHN
WILEY & SONS,LTD., 2002
E,otal [Wh] [6] NEDO LAWEPS, http://www2.infoc.nedo.gojp/nedo/top.html [online]
CaPacit Factr
=Rated Power [W ]x 365 x 24[h ] [100%] (18) [7] NEDO "The New Energy and Industrial Technology Development
Organization" http://www.nedo.gojp/english/introducing/what.html
Table V shows the simulation results. Though the average [online]
wind speed of area C is higher than that of area B, the capacity

TABLE V. SIMULATION RESULT

Wind Total
turbine Iron loss Copper Windage Bearing Stray load Gear box Generated energy Capacity
Energy loss Loss Loss Loss Loss Loss Power Efficiency Factor
[MWh] [MWhL [MWhi [MWh] JMWh] [MWh] [MWh] [MWh] [%] [%]
A 6669.42 123.76 58.61 4.10 1.92 34.66 522.37 5923.75 88.82 13.52
B 20696.07 227.38 252.39 7.63 3.56
1 4 1.78 965.05 19098.01A 92.28 43.60
C 1719i8.42T 186.69 214.09 1 6.26 T2.92 120.11 1792.65 1 5875.47 192.31 136.251

1600