Vous êtes sur la page 1sur 12

Reliability comparison of direct-drive and geared-

drive wind turbine concepts


H. Arabian-Hoseynabadi
1
, P. J. Tavner
2
and H. Oraee
3
1
Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran
2
School of Engineering, Durham University, Durham, UK
3
Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran
ABSTRACT
This paper proposes for wind turbines (WTs) an analytical reliability method, used on other engineering systems, to
compare the reliability of different turbine concepts. The main focus of the paper is to compare the reliability of geared
generator and direct-drive concept WTs. Modication methods are also recommended for improving the availability
of WTs and geared generator concept incorporating doubly fed induction generator. Copyright 2009 John Wiley &
Sons, Ltd.
KEYWORDS
wind turbine; direct-drive concept; reliability model
Correspondence
H. Arabian-Hoseynabadi, Sharif University, Electrical Engineering, Tehran, Iran, Islamic Republic of.
E-mail: arabian@ee.sharif.edu
Received 18 October 2008; Revised 08 June 2009; Accepted 08 June 2009
WIND ENERGY
Wind Energ. 2010; 13:6273
Published online 28 July 2009. DOI: 10.1002/we.357
1. INTRODUCTION
Wind power is the fastest growing renewable source of
energy, which can also be considered as distributed gen-
eration. In the last 20 years, wind turbines (WTs) have
increased in power by a factor of more than 100, the cost
of energy from wind power has reduced by a factor of 10
and the industry has moved from an idealistic fringe activ-
ity to a mainstream power generation source.
Because of the competitive environment in electric
power generation, the industry will prefer the most produc-
tive and economic concepts of WTs. However, long-term
cost analysis, including operation and maintenance
(O&M), as well as the rst investment costs, would result
in a better choice of WT concept. This is only possible if
that analysis included a comprehensive review of the reli-
ability of the different technologies.
The reliability of WTs as a part of a large power system
is assessed in many references.
17
Almost all such studies
consider the wind as a stochastic process, using an appro-
62 Copyright 2009 John Wiley & Sons, Ltd.
priate time series to model the wind, and combining it with
the powerspeed curve of the WT.
There have been few studies on the reliability of a WT
as an isolated system rather than a part in a large power
system.
8
But there have been a number of references com-
paring the productivity of different WT concepts.
9,10

However, no accurate reliability analysis has yet been
included in these comparisons, although there has been a
general study on the reliability of WTs.
11
This paper sets out to ll this gap using Markov model-
ing, which has not been used before for this analysis. It is
expected that the method introduced in this paper will
assist in the comparison between different WT technolo-
gies from a reliability point of view.
The paper is arranged as follows: Different WT Con-
cepts gives a brief progress in WT technology and pro-
poses the concepts of WTs to be studied. Subassembly
Reliability Models reviews related concepts of reliability
modeling theory, and shows the subdivision considered for
WT. A reliability model for selected subassemblies of WT
RESEARCH ARTICLE
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
63
is suggested to be used in the next section. WT Reliabil-
ity Model builds WT reliability models based on the
reliability models of subassemblies in Subassembly Reli-
ability Models. Special types of reliability block diagram
(RBD) considering the design improvements for WTs are
proposed. Recommended Modications for Geared Gen-
erator WT with DFIG introduces a modied concept for
geared generator WTs using a doubly fed induction gen-
erator (DFIG). It is claimed that such a modication could
improve the availability of this type of WTs. Numerical
Results uses reliability eld data and the WT reliability
models derived in WT Reliability Model and Recom-
mended Modications for Geared Generator WT with
DFIG to calculate a typical availability index for selected
WTs. Conclusions presents the conclusions.
2. DIFFERENT WT CONCEPTS
In general, WT systems are subdivided into two main
categories: xed and variable speed. Early xed-speed
WTs, produced up to the late 1990s with the ratings
below 1 MW, used a multistage gearbox and a standard
squirrel-cage induction generator, directly connected to
the grid.
Some improvements were made to the system, termed
semi-variable-speed systems, using two winding genera-
tors with different pole numbers, resulting in a two-speed
system, or a wound rotor induction generator using an
external resistor connected via brushes and slip rings, or the
OptiSlip technology as another example.
12
From the late 1990s, new variable-speed WTs from the
approximately 1 MW power level were introduced in
response to the shortcomings of xed-speed system,
including low energy yield, poor power quality, signicant
audible noise and difculties in braking the turbine. The
rst generation of variable-speed WT systems used a mul-
tistage gearbox, a relatively low-cost standard DFIG and
a power electronic converter feeding the rotor winding
with a power rating of approximately 30% of the rated
power of the turbine.
Since 1991, there have also been variable-speed WT
systems using gearless generators, so-called direct-drive
systems. A fully rated power electronic converter is then
necessary for the grid connection from the direct-drive
generator.
Most direct-drive turbines being produced at the moment
have synchronous generators with a wound rotor electrical
excitation.
9
However, permanent magnet excitation offers
some reliability benets as it eliminates winding and exci-
tation losses.
9
The main reason for using the direct-drive concept
rather than the geared generator variable-speed system is
to eliminate gearbox failure and downtime effects. A reli-
ability comparison between geared and direct-drive
concept WT could verify this reason.
Two geared generator and two direct-drive WT systems
are studied in this paper for comparison of the global reli-
ability and evaluation of typical availability indices for
each of their technologies.
Vestas V39/500 and Enercon E40 are, respectively, a
geared generator and direct-drive turbine in the 500 kW
range.
Tacke TW 1.5s and Enercon E66/15.66 are, respec-
tively, a geared generator and direct-drive turbine in the
1.5 MW range.
The characteristics of these four WTs and the number
of turbine years considered in the data are summarized in
Table I.
13
There is currently a great variety in WT manu-
facturers, and 500 and 1500 kW turbines may now be
considered rather small or of limited interest in the future.
However, this selection is based on the availability of data
from the survey.
14
All reliability analysis is of necessity
backward looking; therefore, in a time of increasing turbine
ratings, it must consider turbines smaller than those cur-
rently being installed, one can still learn valuable lessons
from these important data, and in this case it does allow
us to consider two different power levels.
The contribution of this paper will be to introduce an
analytical method for comparing different WT systems
from a reliability point of view. Reliability modeling and
the availability index calculation procedure have been
used on other engineering systems, but there is no record
of such an analysis being used on WTs.
3. SUBASSEMBLY RELIABILITY
MODELS
Reliability is the probability of a system to perform its
purpose adequately under the operating conditions encoun-
tered for the period of time intended.
15
It is common to
divide systems, from a reliability point of view, into two
categories: mission orientated and repairable.
Table I. Characteristics of the selected WTs.
Vestas V39/500 Tacke TW 1.5s Enercon E40 Enercon E66/15.66
Technology Geared Geared Direct drive Direct drive
Turbine years considered 804 96 900 154
Power (kW) 500 1500 500 1500
Rotor diameter (m) 39 70.5 40 66
Rotor speed (rev min
1
) 30 (rated) 1625 1234 1229
Control technology Pitch-regulated,
active stall
Pitch-regulated,
variable speed
Pitch-regulated,
variable speed
Pitch-regulated,
variable speed
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
64
Reliability comparison WT concepts H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee
For mission-orientated systems, the rst failure is the
most interesting, and the reliability, the probability to be
in the up state, is the most appropriate parameter to
calculate.
On the other hand, for repairable systems, the avail-
ability, i.e. the probability to nd the system in an up
state, is the most appropriate index to calculate.
In the study on the reliability of systems with several
subassemblies, the rst step is the subdivision of the
system into those subassemblies. The accuracy needed for
a reliability study indicates the depth to which the system
should be subdivided. After this subdivision, a reliability
model for each subassembly can be derived, and thereafter,
a system reliability model could be constructed by combin-
ing the reliability models of the subassemblies.
Data about WT subassembly reliability have become
available in recent years from the following surveys:
WindStats, a quarterly public newsletter about WT,
14
and
LWK and WMEP, two reliability projects in Germany.
16,17

There is no unique and standard subdivision for WT subas-
semblies in these studies, but they are similar. One of the
authors has analysed them,
11
and Table II shows an aggre-
gated subassembly subdivision from that reference.
Although other subassemblies are signicant in WT
reliability, this paper concentrates on comparing geared
generator and direct-drive concept WTs, and so focuses on
the following subassemblies: blade, gearbox, generator,
converter, pitch and yaw system. Figure 1 shows these
subassemblies in a typical geared generator WT.
According to reliability theory,
15
failure and repair rates
can be dened respectively as: l, the number of failures
in a given period of time divided by total period of time
the component was operated, and m, the number of repairs
in a given period of time divided by total period of time
the component was being repaired.
The failure rate function of almost all systems obeys
from the bath-tube curve,
15
and Tavner et al.
11
suggested
that it is reasonable to consider that most WT subassem-
blies lie in the bottom of this curve, i.e. that they have a
xed failure rate. So Markov modeling is the best candi-
date to use for reliability modeling.
The useful life period assumption for equipment allows
the denition of these transition rates as the reciprocal of
the average duration of operation and repair,

=
1
MTTF
(1)

=
1
MTTR
(2)
where MTTF represents mean time to failure, and MTTR
stands for the mean time to repair. Recording the periods
of system operation and repair, and taking the average of
up- and downtime give numerical values for MTTF and
MTTR.
In Markov modeling, all the performance states of the
system are considered. Using the rate of transition between
these states could result in the probability of residence in
each state. In the simplest form, a two-state up and down
model could be considered, although there is no restriction
on the state numbers except that they must be limited and
identiable.
15
In this paper, a simple two-state model is
assumed for each subassembly, but the extra derated
state, indicating non-optimal electrical energy production,
is also considered for the whole WT system.
Considering each subassembly as a system in its own
right, its components must be considered, and combining
these reliability models could result in a related subas-
sembly reliability model. The reliability model for a DFIG
is considered here, and for other subassemblies the same
procedure could be conducted to nd their Markov
models.
All components of the generator could be categorized
into the two main categories of electrical or mechanical.
The electrical category consists of the stator and rotor
cores, and windings and terminal boxes. The mechanical
category consists of the casing and mounting, shaft,
bearing, slip rings and brushes.
18
Thereafter, the generator
could be treated as a two-component system. Representing
the electrical by E and mechanical by M, three different
states could be considered as shown in Figure 2. In this
gure, a rectangle around E or M means an electrical
or mechanical failure in generator, respectively. It is
assumed that after a generator electrical or mechanical
failure, it will be disconnected from the system, and failure
in other parts will not occur, so the state which shows both
electrical and mechanical failure is omitted.
The transition rate between state 1 and state 2 in Figure
2 is the aggregation of all the electrical components con-
sidered in series from the reliability point of view. The
same procedure is considered for mechanical components.
Therefore, in Figure 2, l
e
represents the summation of
electrical component failure rates, and m
e
could be calcu-
lated by frequency balance approach,
15
with the assump-
tion of series connection of all electrical components. If
the repair rate of each component is greater than its failure
rate, which is normal in practice, the approximation
method
15
could be considered to calculate the equivalent
Table II. WT subassembly subdivision.
WT Subassemblies
Rotor blades
Mechanical brake
Air brake
Main shaft
Gearbox
Generator
Yaw system
Electrical controls
Hydraulics
Grid or electrical system
Pitch control system
Other
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
65
repair rate. As equation (2) shows the reciprocal of repair
rate is the mean time to repair, if r can be considered to
represent this value,

1

e
e
e
= =

r
r
i i
i
(3)
in which l
i
and m
i
shows the ith electrical component
failure and repair rates. A similar procedure can be con-
sidered for mechanical failure and repair rates.
Using the frequency balance approach,
15
the three-state
model of Figure 2 could be summarized to a two-state
model shown in Figure 3.
The generator equivalent transition rates, l
g
and m
g
, are
calculated based on electrical and mechanical transition
rates, and the fact that aggregation of states 2 and 3 in
Figure 2 would result in down state (state 2) in Figure 3.

g e m
= + (4)



g
e m e m
e m m e
=
+ ( )
+
(5)
For all other subassemblies of WT, a two-state model,
similar to generator reliability model in Figure 3, could be
constructed with equivalent failure and repair rates, calcu-
lated based on their components and the type of inter-
connection between these subassemblies, e.g. series or
parallel.
The subdivision of the other ve subassemblies consid-
ered in this paper is based on the data availability and the
Figure 1. WT main subassemblies.
Figure 2. Generator simple Markov model.
Figure 3. Reduced generator Markov model.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
66
Reliability comparison WT concepts H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee
depth of study to be considered. The subassembly blade
could be subdivided into components, but no reliability
data have been gathered for blade components. This may
also be true for other components of subassemblies, and
as the main issue is consideration of the overall WT reli-
ability, it seems reasonable to adopt the subassembly sub-
division: (i) blade; (ii) gearbox (case and mounting; low-,
mid-, high-speed drivetrain; and lubrication system); (iii)
converter (machine-side inverter, grid-side inverter, DC
link and control unit); (iv) pitch system (control system
and actuation system); and (v) yaw system (control system
and actuation system).
The two-state Markov models for the six selected subas-
semblies are used in the next section to build up reli-
ability models for direct-drive and geared concept WTs.
Each subassembly is assigning with its equivalent failure
and repair rates.
4. WT RELIABILITY MODEL
The rst step in system reliability modeling is the analysis
of system performance, which results in the identication
of the interaction between the main subassemblies. The
best method to show this interaction between subassem-
blies is the RBD. RBDs show each subassembly as a
separate block, and these blocks could be in series, paral-
lel or seriesparallel combinations.
In this section, at rst, the RBDs for existing models of
WTs are derived based on the subassembly subdivision in
the previous sections, and then special reliability diagrams
are proposed for recommended models of WTs with design
improvements in pitch and yaw systems for normal wind
conditions.
As one of the main duties of pitch and yaw systems is
to prevent the interruption of operation in adverse weather
or high-wind conditions, they must be considered in series
to other subassemblies in an RBD for a WT. An RBD for
a geared generator concept WT with selected subassem-
blies as described in the last section is shown in Figure 4.
It is evident that for a direct-drive model, the RBD would
be the same with the gearbox block omitted.
The aggregation of series connected of these six subas-
semblies results in a simple two-state updown Markov
model of WT similar to the generator in Figure 3. The
equivalent failure rate of such a model would be the sum-
mation of each subassembly failure rate in Figure 4, and the
equivalent repair rate could be calculated based on equation
(3) using these subassembly failure and repair rates.
The function of the pitch system in normal wind condi-
tion is to optimize the WT performance based on C
p
(l, b)
curve by controlling parameter b; also the function of the
yaw system performance is to adjust the nacelle position
taking account of the position of the WT in a wind farm,
depending on the prevailing wind direction. These two
systems can be considered as special parallel blocks of an
RBD when design improvements are considered.
Based on the assumption of a WT being able to continue
operating after minor failures of pitch and yaw systems in
normal wind conditions, special reliability diagrams for
the geared generator and direct-drive concept WTs are
proposed in the following sections, and the quantitative
improvements in availability index caused by these modi-
cations will be presented in Numerical Results.
It is possible to model the WT from the reliability point
of view in each wind regime separately. These models will
be linked by wind condition transition rate; Castro Sayas
and Allan
8
suggested such a modeling for WTs in different
weather conditions. The reliability models considered here
are based on normal weather conditions and WT design
improvements.
4.1. Geared generator concept WT
All of the six subassemblies which were considered in
Subassembly Reliability Models are used in this type of
WT; the converter is partially rated and the generator is a
DFIG.
As the pitch and yaw subassemblies are assumed to
optimize the electrical energy production of WT, they are
considered in parallel to other subassemblies, which are
series connected. This means that after a pitch or yaw
system failure, the WT could continue to capture electri-
cal energy from the wind energy, but in a suboptimal
situation.
It is worth mentioning that such a reliability diagram,
as shown in Figure 5, is valid until the failure modes
affecting only the pitch or yaw systems are considered. If
root causes affecting subassemblies other than pitch or
yaw have to be considered, then these two subassemblies
must also be connected in series to the other subassem-
blies. For example, a hydraulic system failure could affect
not only the pitch and yaw system, but also other subas-
semblies such as the mechanical brake. However, in the
WT, the hydraulic system could be considered as a sepa-
rate subassembly.
11
In Figure 5, NC stands for a normally closed switch,
which means in normal WT operation, the pitch and yaw
systems are used to optimize electrical energy production.
It is important to note that the reliability diagram shown
here is different from a standard RBD in which system
Figure 4. Geared generator concept WT (existing system) RBD.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
67
success equals the existing path between in and out
point of diagram.
The main path of Figure 5 consists of the four subas-
semblies (blade, gearbox, generator and converter), is the
only path for system success and the parallel paths created
by pitch and yaw systems can only optimize WT perfor-
mance in normal wind condition.
As the four subassemblies in the main path of reliability
diagram in Figure 5 are series connected, they could be
aggregated into one block with equivalent failure and
repair rates. This equivalent block is shown as GGWT
in Figure 6; this stands for geared generator WT.
The equivalent failure rate of the series connected
blocks is the summation of their failure rates. If the repair
rate is much greater than failure rate, which is the case in
many situations, the equivalent repair rate could be calcu-
lated approximately as
1

ggwt
ggwt
ggwt
=

r
r
i i
i
(6)
in which l
ggwt
, the equivalent failure rate, is the summation
of the individual failure rates:


ggwt
=
i
i
(7)
The summation index i could be related to blade,
gearbox, generator and converter.
Based on the reliability diagram derived for the geared
generator concept WT, a three-component system could
be considered for this type of WT. As for an n component
system, 2
n
states show all possible performance states; the
eight states of such a system are shown in Figure 7.
In this diagram, failure of each subassembly is shown
by a rectangle around it.
The eight states in Figure 7 could be categorized in
three states of up, derated and down. It is evident that
only if these three components of system are in their up
state would result in system up state too. As long as the
pitch or yaw system, or both, are in their down state only,
the WT could continue its non-optimal performance,
which is named derated state here.
Any malfunction in a GGWT component, which is the
aggregation of other four subassemblies, would result in
down state of the WT system, regardless of the failure
or operation of yaw and pitch systems. This is the reason
that transition rates between some states are omitted, as
any failure after GGWT failure is irrelevant.
Similar to the procedure for reliability modeling of gen-
erators in previous sections, using the frequency balance
approach,
15
this eight-state model could be reduced to the
three-state model shown in Figure 8. For this to be done,
we need the transition rates between states to nd the
equivalent state probabilities. The capital letters shown in
Figure 7, e.g. Y or P or G, represent these transitions
and their origins: the yaw, geared generator and pitch
systems, respectively. These letters represent the failure or
repair rates, l and m, of those respective systems in the
relevant direction.
The reduced statespace reliability model for geared
generator WT is shown in Figure 8, which is a three-state
model based on six selected subassemblies. The state vari-
able of (Bl., Gear., Gen., Con., Pitch, Yaw) is considered
to represent the state of each subassembly, one for up
and zero for down states, in the system statespace
diagram. The other state variables, which are not shown,
belong to the system down state in Figure 8.
The probability of residence in each of the three states
in Figure 8 could be calculated by frequency balanced
method.
15
Also, aggregation of up and derated states
could give a good estimate of WT availability.
4.2. Direct-drive WT
The same procedure could be conducted for direct-drive
concept WT, as shown in Figure 9. In this type of WTs,
Figure 5. Modied geared generator concept WT reliability diagram.
Figure 6. Simplied reliability diagram of modied geared
generator WT.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
68
Reliability comparison WT concepts H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee
there is no gearbox, but the power electronic converter is
fully rated, which is shown with different block symbol in
Figure 9, and a synchronous multipole generator is used,
which could be wound rotor or permanent magnet
excited.
The same conditions considered in the modied geared
generator concept are considered here, which means that
the main path consists of blade, generator and fully rated
converter. The other two paths, created by the pitch and
yaw systems, are only used to optimize the WT perfor-
mance in normal wind condition.
The aggregation of components in the main path could
be named DDWT, which shows direct-drive WT. The
three-component system consists of DDWT, pitch and yaw
systems; would have eight-state diagram, which could be
reduced to a three-state model shown in Figure 10.
The reduced state variable reliability model for the
modied direct-drive WT concept includes the ve selected
subassemblies. Frequency balance methods could be used
to calculate the probability of residence in each of these
three states for direct-drive WT.
The rst interpretation is that, because of a reduced
number of components resulting from the absence of the
gearbox, the direct-drive concept has a better availability
index than the geared generator concept. But account must
be taken of the dependency on other factors, including
generator and power electronic converter reliabilities in
the two concepts.
The gearbox and partially rated converter combination
should be compared with the fully rated converter in the
two concepts, as should the synchronous multipole gen-
erator be compared with the DFIG.
The next section will introduce a modication method
to increase the availability of the geared generator WT.
This modication will result in a change of the reliability
diagram, and therefore, in its statespace model. Numer-
ical Results uses these ve reliability models and eld
data to compare the two concepts and the effect of the
modications in availability improvement.
5. RECOMMENDED
MODIFICATIONS FOR GEARED
GENERATOR WT WITH DFIG
The main goal of the proposed modication is to reduce
system downtime caused by converter failure. This is
achievable because the converter associated with a DFIG
does not interface directly between the generator and grid,
as it does in the direct-drive concept. In the geared-drive
Figure 7. Statespace diagram of modied geared generator WT.
Figure 8. Reduced statespace diagram of modied geared
generator WT.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
69
concept, the converter is connected to rotor winding
through brushes and slip rings. After a converter fault, the
rotor windings could be shorted and changed into a xed-
speed system, continuing to feed the grid until the con-
verter is repaired.
Two alternative arrangements can be recognized: (i)
star connection of rotor windings (in this mode, the wind-
ings must be shorted after a converter fault); and (ii) delta
connection of rotor windings (in this mode, the connection
between the converter and windings must be opened after
a converter fault).
The recommended concept is shown in Figure 11. As
the controller senses a fault in the converter system, in any
part (inverters, dc link or control unit), it would operate
switches S
1
and S
2
. S
1
must short the outgoing connections
from the rotor windings if they are star connected, or leave
them open if they are delta connected.
Operation of S
1
would change the generator to a wound
rotor with shorted windings, like a squirrel cage machine.
S
2
would open the grid-side connection. The system
remaining after the operation of the switches is a xed-
speed WT that can continue to work during the repair time
of the converter.
With this modication, the reliability diagram derived
for geared generator WT, with a DFIG, in Figure 5 changes
to Figure 12.
Figure 9. Modied direct-drive WT reliability diagram.
Figure 10. Reduced statespace diagram for modied direct-
drive WT.
Figure 11. GGWT modied system (converter failure).
To simplify this block diagram, initially the two-compo-
nent DFIG system, consisting of generator and converter,
could be aggregated. If the time required for controller to
sense the fault in converter and actuation of switches to
isolate the converter is negligible compared to the converter
repair time, which is reasonable, converter failure could not
result in a down state for DFIG, and this system would only
have failed after generator failure. This strategy can be
shown in the statespace diagram shown in Figure 13.
Using the frequency balance method could reduce this
three-state model to a simple two-state model in Figure 14,
with equivalent failure and repair rate equations (8) and (9).



eq
g c
c c
=
+
i
(8)

eq g
=
i
(9)
It is emphasized here that this equivalent model is only
valid as long as the actuation time of controller introduced
is negligible, compared to the converter repair time. Also,
the controller and switches are assumed to be fully reli-
able. The modied concept therefore has the same MTTR
as a system with an induction generator only (m
eq
= m
ig
).
The failure rate is slightly reduced compared to a system
with the induction generator alone.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
70
Reliability comparison WT concepts H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee
Figure 12. Modied RBD for WT with a DFIG.
Figure 13. Statespace diagram for modied DFIG system.
Figure 14. Reduced statespace diagram for modied
DFIG system.
As in WT Reliability Model, the three subsystems of
blade, gearbox and equivalent generator and converter
could be aggregated, and a three-state Markov model,
deduced from the eight states statespace diagram, could
be produced as shown in Figure 15.
It is evident, from comparison between Figures 8 and 15,
that this modication transfers state variables from the
down to the derated state, increasing the number of state
variables in the derated state. It must be said that operat-
ing a variable-speed WT at xed speed during wind gusts
may cause the acceptable mechanical torque limit, for
which it is designed, to be exceeded. Variable speed allows
the rotor inertia to absorb some of the violent torques during
wind gusts. More detailed feasibility studies and control
scheme modications would be necessary to implement
such theoretical modications in a WT with a DFIG.
6. NUMERICAL RESULTS
Data about the failure rates and repair times of WTs and
their subassemblies have been gathered from European
surveys.
14,16,17
WindStats, which is a quarterly public domain newslet-
ter about WT, gathers reliability data.
16
An analysis of
Danish and German WTs over 11 years has been done,
based on WindStats data
11
showing that the useful life
assumption for WT and its subassemblies is a reasonable
hypothesis. Another survey in Germany, LWK,
14
is the
main source for eld data used in this section, which
included almost 600 WTs for a 12 year period. An advan-
tage of LWK data, compared to WindStats, is that the WT
population is nearly constant in the period of study, and it
distinguishes between WTs of different concept.
The LWK data consist of the failures and downtime of
the main subassemblies during the period of study. It is
then possible to calculate the failure and repair rate for
each subassembly.
15
The number of turbines changed
slightly in the period of study, so the failure rate is calcu-
lated separately for each year and then averaged for each
subassembly at the end.
The reliability data for the two selected geared genera-
tor concept WTs, introduced in Different WT Concepts,
based on LWK eld data,
14
are tabulated in Table III.
It is important to notice that the presence of the power
electronic converter will result in a smoother interaction
between gearbox parts, which could extend their MTTF,
although different modes of WT operation may shorten the
converter MTTF. This is the reason for the difference
between such assembly failure rates in WTs compared to
gearboxes and converters in other industries.
The same procedure was performed for direct-drive
concept WTs of Different WT Concepts and is shown in
Table IV.
Using these data and the reliability models derived for
existing and modied geared generator and direct-drive
concept WTs in WT Reliability Model and the recom-
mendations for geared generator concept WT in Recom-
mended Modications for Geared Generator WT with
DFIG, the probability of residences in each of the up,
derated and down states could be calculated. As in the
existing system, reliability model of WTs all selected sub-
assemblies are series connected; there is not a derated
state for this model. The aggregation of the up and
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
71
Figure 15. Reduced statespace diagram for modied WT with a DFIG.
Table III. Geared generator WT reliability data.
Vestas V39/500 Tacke TW 1.5s
l (f/year) m (rep./y) l (f/year) m (rep./y)
Blade 0.162 265.3 0.214 523.0
Gearbox 0.168 269.2 0.554 60.2
Generator 0.085 170.7 0.268 89.9
Converter 0.254 508.1 0.357 415.2
Pitch 0.095 559.9 0.500 226.3
Yaw 0.097 436.7 0.161 433.2
Turbine years considered 804 96
Table IV. Direct-drive WT reliability data.
Enercon E40 Enercon E66/15.66
l (f/year) m (rep./y) l (f/year) m (rep./y)
Blade 0.240 133.0 0.167 620.8
Generator 0.354 143.7 0.120 270.5
Converter 0.317 430.7 0.287 291.7
Pitch 0.292 512.0 0.417 533.4
Yaw 0.116 348.3 0.139 428.0
Turbine years considered 900 154
derated states could give a good sense of availability.
These results are tabulated in Table V, and the two rows
marked with an asterisk show the WT systems using the
modications recommended in Recommended Modica-
tions for Geared Generator WT with DFIG, using a fully
reliable controller.
It is evident from these results that the rst guess of
better availability for direct-drive concept WTs, because
of the absence of a gearbox, is not true in all circum-
stances. Taking into account the reliability data of Tables
III and IV, the generator and converters in these four con-
cepts have signicantly different failure and repair rates.
The geared drive shows better availability than the
direct drive for smaller WTs, but this is reversed for the
larger direct-drive WTs concept. This shows the necessity
for detailed reliability studies rather than making rule of
thumb, deterministic assumptions. The calculations here
take no account of the different maintenance manning and
logistic arrangements that may have been made for these
different sizes and concepts of WTs. Also, the approach
to the design of WT gearboxes has been changed over the
last few years, especially after the introduction of the
AGMA 6006 technical standard for WT gearboxes,
19

which should improve the reliability indices for future
geared generator WTs.
Table V results show an approximately 0.02% improve-
ment in availability for the geared-drive concept using the
modications introduced in Recommended Modications
for Geared Generator WT with DFIG. This may seem an
insignicant improvement, but annualizing this improve-
ment with the WT rating will result in a signicant increase
of electrical energy production.
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
72
Reliability comparison WT concepts H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee
Table V. State probabilities in statespace model of WT systems and their availability values.
Turbine concept Turbine type Reliability model Up Derated Availability Down
Geared generator Vestas V39/500 Existing system 0.9974 0.9974 0.0026
Modied turbine 0.9976 0.00039 0.9980 0.0020
Modied generator* 0.9976 0.00059 0.9982 0.0018
TackeTW 1.5s Existing system 0.9840 0.9840 0.0160
Modied turbine 0.9914 0.00256 0.9940 0.0060
Modied generator* 0.9914 0.00276 0.9942 0.0058
Direct drive Enercon E40 Existing system 0.9940 0.9940 0.0060
Modied turbine 0.9951 0.00090 0.9960 0.0040
Enercon E66 Existing system 0.9972 0.9972 0.0028
Modied turbine 0.9969 0.00110 0.9980 0.0020
* WT systems using the recommended modications for geared generator WT with DFIG, using a fully reliable controller.
High values for availability obtained in Table V are
caused by the limited number of subassembly selection in
each WT. It is evident that if all subassemblies were con-
sidered, these availability indices would be reduced.
7. CONCLUSIONS
It is not possible to conclude denitively that either geared-
drive or direct-drive concept WTs are more reliable, but
the proposed analysis method shows a way to compare
concepts from the reliability point of view. The analysis
indicates the importance of subassembly reliability data in
calculating overall availability.
Based on these reliability models and data, a more
detailed sensitivity analysis could be done to improve
future designs, and O&M procedures. As an example, if
one subassembly has a signicant share in system unavail-
ability, redesign or more frequent inspection during oper-
ation is recommended.
It can be seen from the reliability diagrams for both WT
concepts that the gearbox and partially rated converter of
the geared drive can be compared with the fully rated
converter of the direct drive. The generators in these two
concepts clearly have different impacts on the reliability
because of their different failure and repair rates.
Thereafter, a detailed reliability analysis, like the
method suggested in this paper, could be used to assess the
reliability of different WT congurations and concepts.
However, it must not be forgotten that WT concept can
affect other subassembly performance, and a true value of
availability for different WTs could be obtained by con-
sidering all subassemblies.
The reliability analysis of WTs stresses the importance
of gathering and processing reliability data from wind
farms around the world. As more reliability eld data
become available, more accurate judgments could be made
about different concepts, design improvements and main-
tenance strategies.
Numerical results from four selected WTs have shown
the availability improvement possible for geared-drive
WTs, by using the modications proposed in the paper. A
new generator technology has been proposed for WTs, the
brushless doubly fed generator (BDFG),
20
from which slip
rings and brushes are omitted. This analysis could be used
to show the availability improvement, which a BDFG
could bestow compared to a DFIG, in a WT.
REFERENCES
1. Gavanidou ES, Bakirtzis AG, Dokopoulos PS. A
probabilistic method for the evaluation of the perfor-
mance of wind-diesel energy systems. IEEE Transac-
tions on Energy Conversion 1992; 7: 418425.
2. Billinton R, Chowdhury AA. Incorporation of wind
energy conversion systems in conventional generat-
ing capacity adequacy assessment. IEE Proceedings
Generation, Transmission & Distribution 1992; 139:
4756.
3. Billinton R, Chen H. Assessment of risk-based capac-
ity benet factors associated with wind energy conver-
sion systems. IEEE Transactions on Power Systems
1998; 13: 11911196.
4. Karki SH, Chedid RB, Ramadan R. Probabilistic per-
formance assessment of wind energy conversion
systems. IEEE Transactions on Energy Conversion
1999; 14: 217224.
5. Bakirtzis AGA. Probabilistic method for the evalua-
tion of the reliability of stand alone wind energy
systems. IEEE Transactions on Energy Conversion
1992; 7: 99107.
6. Mohammed H, Nwankpa CO. Stochastic analysis and
simulation of grid-connected wind energy conversion
system. IEEE Transactions on Energy Conversion
2000; 15: 8590.
7. Karki R, Billinton R. Cost-effective wind energy uti-
lization for reliable power supply. IEEE Transac-
tions on Energy Conversion 2004; 19: 435440.
8. Castro Sayas F, Allan RN. Generation availabili ty
assessment of wind farms. IEE Proceedings
Wind Energ. 2010; 13:6273 2009 John Wiley & Sons, Ltd.
DOI: 10.1002/we
H. Arabian-Hoseynabadi, P. J. Tavner and H. Oraee Reliability comparison WT concepts
73
Generation, Transmission & Distribution 1996; 143:
507518.
9. Polinder H, van der Pijl FFA, de Vilder GJ, Tavner
PJ. Comparison of direct-drive and geared generator
concepts for wind turbines. IEEE Transactions on
Energy Conversion 2006; 21: 725733.
10. Petersson A. Analysis, modeling and control of
doubly-fed induction generators for wind turbines.
PhD Thesis, Chalmers University of Technology,
2005.
11. Tavner PJ, Xiang J, Spinato F. Reliability analysis for
wind turbines. Wind Energy 2007; 10: 118. Published
online 12 July 2006 in Wiley Interscience (www.
interscience.wiley.com) DOI: 10.1002/we.204.
12. V47-660 kW, Vestas pitch regulated wind turbine
with OptiTip

and OptiSlip

. [Online]. Available:
http://www.vestas.com/en/dis/copy-of-about-vestas/
history/2000-2002.aspx. (Accessed 11 July 2009).
13. Vestas, GE and Enercon Data sheets. Available from
manufacturers websites as follows: http://www.vestas.
com/en/wind-power-solutions/wind-turbines.aspx;
http://www.ge-energy.com/businesses/ge_wind_energy/
en/index.htm; http://www.enercon.de/en/_home.htm.
(All accesed 11 July 2009).
14. Landwirtschaftskammer (LWK), Schleswig-Holstein,
Germany. [Online]. Available: http://www.lwksh.de/
cms. (Accessed 29 June 2009).
15. Billinton R, Allan RN. Reliability Evaluation of Engi-
neering Systems (2nd edn). Plenum Press: New York,
1996.
16. WindStats (WS). [Online]. Available: http://www.
windstats.com. (Accessed 29 June 2009).
17. Enlin C, Fller G, Hahn B, Hoppe-Kitpper M, Rohrig
K. The scientic measurement and evaluation pro-
gramme in the German 250 MW Wind Programme
(WMEP). European Wind Energy Conference, EWEC
1993.
18. Tavner PJ. Predicting the design life of high
integrity rotating electrical machines. IEE 9th Inter-
national EMD Conference, Canterbury, 1999; 286
290.
19. ANSI/AGMA/AWEA 6006-A03 Design and speci-
cation of gearboxes for wind turbines. Published by
the American National Standards Institute.
20. McMahon RA, Wang X, Abdi-Jalebi E, Tavner PJ,
Roberts PJ, Jagiela M. The BDFM as a generator in
wind turbines. European Power Electronics Confer-
ence, Portoroz, Slovenia, 2006.

Vous aimerez peut-être aussi