System Performance, Availability

and Reliability Trend Analysis

Portfolio Review
2021/22
Introduction
What is SPARTA? What is included in this report?
SPARTA is an offshore wind farm performance benchmarking This report is split into 3 sections:
tool, run by industry for industry. Standing for ‘System
Performance, Availability and Reliability Trend Analysis’, 1. The Year in Review
this tool allows owner/operators of offshore wind farms to The report gives highlights of benchmarks from the 2021/22
compare key performance indicators (KPIs) for their farms to financial year, showing the trends of metrics such as
aggregated and anonymised benchmarks. The SPARTA Joint capacity factor, production-based availability and turbine
Industry Project (JIP) is sponsored by The Crown Estate and transfers. The year is compared to previous years in order
the Offshore Renewable Energy (ORE) Catapult. to evaluate how the industry is changing. We also draw out
more general insights from the set, including a further look
Offshore wind performance benchmarks are available at the effect of the pandemic and at component failures.
from January 2014. In total, owner/operators can supply a
maximum of 159 KPIs and then have access to over 500 2. Grid Curtailment and Availability Loss
benchmarks every month, including derived values, covering
Examining a key factor during the year that led to lower
four main areas:
availability figures, we highlight the impact that forced
• Availability curtailment from system operators has on production.

• Production and Lost Production 3. Operations and Performance Throughout the Life of
• Reliability a Turbine
• Operations Delving into trends across the age of windfarms in the set,
the review questions whether the common bathtub curve
of reliability is apparent in the portfolio. It examines metrics
such as availability, forced outages and major repairs
across the lifespan of a turbine.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/2022 Portfolio Review 2
Introduction
Who is Involved? Principle of SPARTA Why is Benchmarking Important?

All major owner/operators with offshore wind farms in UK The SPARTA platform has been designed based on the Benchmarking with SPARTA allows wind farms to compare
waters are participating in the 2020/21 SPARTA Portfolio following principles, which have helped establish SPARTA their performance to an industry “norm”. This allows a
Review. The SPARTA group aims to continue gathering as the industry-leading performance benchmark provider number of potential benefits:
members across Europe in order to maximise system data for offshore wind:
and produce more robust benchmarks for industry. • Identify underperformance: Find periods where your
• Anonymity: Generation of benchmarks requires sensitive wind farm is not performing as well as the industry and
operational data. To ensure operational KPIs are not be armed with the tools to ask why and perform more
Sponsoring Organisations shared, SPARTA aggregates metrics and securely in-depth analysis.
uploads them into an anonymised data pool.
• Identify good practice: When your wind farm is one of
• Transparency: There is complete transparency in the higher performing wind farms, have the resources
definitions and methodologies used and these are available to first identify this period and be able to review
published in a Metric Handbook. Consequently, results what made this period so good.
are clear, comprehensive and consistent.
• Future planning: By filtering on certain dimensions see
• Quality: Extremely high quality and reliable outputs are how older wind farms are performing and have the ability
Participating Owner Operators achieved through continuous metric assurance and to compare these. This can then be used to plan what
verification activity. can be expected as your wind farm ages.

• Representative data volume: SPARTA benchmarks • Industry collaboration: Be part of the future and help
are based on a representative population, with over the industry improve performance, reduce failures and
50% of all offshore wind farms in UK waters providing optimise transfers, together. By getting industry to work
performance data on a monthly basis for over 6 years. together, SPARTA aims to help tackle climate change by
improving renewables.
• Industry-Led: The SPARTA system was designed by
owner/operators for owner/operators and is continuously
improved to ensure it reflects industry needs.

• Monthly Benchmarks: New benchmarks are made

available to members every month. This reveals seasonal
variations and can inform detailed optimisation of
operations and modelling of new wind farms.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/2022 Portfolio Review 3
The Portfolio

Turbine Rating Turbine Count

12 12

10 10
1607
Number of Windfarms

Number of Windfarms
8 8

6 6
Turbines
4 4

2 2

25
0 0
≤3MW Between 3-5MW ≥5MW < 50 50-100 > 100

16
Original Equipment Manufactures
12
Distance to Shore Windfarms
14
10
12
Number of Windfarms

Number of Windfarms

8
10

6348MW
8 6

6
4
4

2
2
Capacity
0 0
OEM 1 OEM 2 OEM 3 < 14km 14 to 25km > 25km

Figure 1 Number of windfarms in the SPARTA portfolio, categorised by rating, turbine count, OEM and distance to shore.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/2022 Portfolio Review 4
The Year
in Review
Annual Performance 2021/22

36.11% 19.84
Average Capacity Crew Transfer Vessel
Days per Turbine
71.94
Factor Transfers per
from 15.11
from 38.86%* Turbine

0.21
from 92.52

Major Repairs
per Turbine
from 0.09
297 MWh
Average Lost

3135 MWh Production per MW

Average Generation from 192 MWh

per MW
91.94%
28.47
from 3380 MWh
Average Production
Based Availability Outages per
Turbine
from 94.58%
from 23.35

*Comparison figures are the average from datapoints before 2021/22

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 6
Capacity Factor
Capacity factor can be treated as the achieved percentage
of the total possible production assuming there were perfect Mean Capacity Factor and Hub Height Wind Speed in 2021/22
environmental conditions. As conditions are seldom perfect, 13
this figure is driven primarily by the wind speed at site, as 70 Mean Capacity Factor
Capacity Factor P25 and P75 12
is highlighted by the closely linked trends of capacity factor
Mean Hub Height Wind Speed
and wind speed throughout the year. However, the financial

Hub Height Wind Speed (m/s)

60 11
year did have a particularly low capacity factor given

Capacity Factor (%)

modest wind speeds, registering as the lowest rating year 50 10

since 2016/17 (where wind conditions were worse). 9

40
8
30
7

20
6

May 21 Jul 21 Sep 21 Nov 21 Jan 22 Mar 22

Yearly Average Capacity Factor and Wind Speed

45
9.0
What is Capacity Factor?

Mean Hub Height Wind Speed (m/s)

8.5
40
Capacity Factor is a measure of how much power a
8.0
Capacity Factor (%)

turbine is producing compared to its rated capacity.

Generally, this is reported over a period of time for a wind 35 7.5
farm, so is a measure of how well the farm is producing
7.0
on average compared to its rated capacity.
30
6.5
Example:
6.0
25
A 500MW wind farm produces 219,000 MWh for a month.
5.5
For a capacity of 500MW for a month (730 hours), the
farm had the potential to produce 365,000 MWh. 20 5.0
2016/17 2017/18 2018/19 2019/20 2020/21 2021/22

219,000 MWh / 364,000 MWh = 60%

Figure 2 Mean monthly capacity factor and wind speed over 2021/22 (top) and annual capacity factor and wind speed from
2016-2022 (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 7
Production Based Availability
PBA, which takes wind speed into account in its definitions
of possible production, is generally accepted to be a more Mean Production Based Availability 2021/22
100
meaningful measure of availability than capacity factor.
Mean PBA
Average PBA was at its lowest in the set over the 2021/22 98 PBA P25 and P75

Production Based Availability (%)

period and was below 94% for all months apart from May. Mean PBA exc. 3 Lowest
The average sits on the lower end of the inner-quartile 96
range, showing that this low PBA may be driven by a
few farms with significantly low PBA dragging down the 94

average. This is in large part driven by larger curtailment

92
from Electricity System Operators (ESOs) which presents
major problem for the development of offshore wind – this is 90
explored more in the following section on Grid curtailment in
2020/21. It shows that despite the relatively good capacity 88
factor owing to high wind speeds, the portfolio experienced
its worst year in the set, in terms of availability. 86
May 21 Jul 21 Sep 21 Nov 21 Jan 22 Mar 22

Yearly Average Production Based Availability (PBA)

98
What is production based availability?
96
Production Based Availability, or PBA, is a measure of
how well a turbine is using the wind resource available
to it. Unlike capacity factor, PBA does not punish for low 94
winds, as it measures how well the turbine is performing
PBA (%)

compared to its power curve, given the wind speeds that

92
occur at that site.

Example:
90

The wind at site is 6m/s and the power curve ‘says’ the
turbine should be generating 1000kW but the turbine is 88
only producing 700kW. This would give the turbine a PBA 2016/17 2017/18 2018/19 2019/20 2020/21 2021/22
of 700kW/1000kW, so 70%.
Figure 3 Mean monthly PBA and PBA excluding 3 lowest performers over 2021/22 (top) and annual PBA from
700kW / 1000kW = 70% 2016-2022 (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 8
Transfers
As reported in previous reviews, the number of turbine
transfers has been on a decreasing trend over the last
decade and has continued its levelled trend over the last Transfers per Turbine Over Time
few years. A lower number of transfers is preferred both 16
for cost and Health & Safety. Turbine transfers peak when Transfers pT
12 Month Rolling Average
operators carry out scheduled maintenance during low 14

Monthly Number of Transfers per Turbine

production periods in summer and fall through the winter
when production is high and non-access days are more 12
likely. The rolling yearly average of this trend has levelled off
in recent years, showing that industry has either reached a 10
minimum feasible number of transfers or has struggled to
continue making improvements in this area. 8

2014 2015 2016 2017 2018 2019 2020 2021 2022

What is a turbine transfer?

Transfers Across the Year
A turbine transfer is defined as the number of completed
transfers of technicians from a vessel onto a turbine or
3.72 3.31 6.3 7.71 7.68 8.73 9.52 9.18 8.97 5.1 3.99 2.94
substation.

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
A technician transferring onto and then subsequently off
of a turbine counts as one transfer.

A single technician can transfer onto several turbines in

a day and a vessel can transport several technicians to a Figure 4 Mean monthly turbine transfers over time with rolling annual average (top) and average transfers across
singular turbine. the year (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 9
COVID and Operations
The 2020/21 portfolio review examined the impact of the
pandemic on operations and maintenance, noting slightly
reduced availability and a higher number of chartered Vessel Usage Throughout the Pandemic
vessels, perhaps motivated by social distancing. Figure
10 shows the usual seasonal variation in CTV rentals 1.4
throughout the year, which closely matches the seasonal
trend of turbine transfers. However, this trend was much
flatter throughout the pandemic, implying that windfarms 1.2
retained CTVs through the winter to accommodate for

CTV Days per Turbine

greater social distancing offshore. Following easing of
restrictions in Europe after July 2021, the seasonal trend 1.0
looks to have returned. There is a case to say that the
effects of reduced maintenance during the pandemic will
have long term effects on reliability, but that does not yet 0.8

appear to be justified.
Pre-Pandemic (Up to Feb 20)
0.6 Pandemic (Mar 20- Jun 21)
Post-Pandemic (Jul 21-Aug 22)

Jan Feb Mar Apr May Jun Jul Aug Sep Dec
What is a Vessel Day? Oct Nov

The number of vessel days in a month is the total number

of available vessels multiplied by how many days those
vessels were available for. SPARTA collects this metric
for crew transfer vessels (CTVs), special operation
CTV Days per Turbine Across the Year
vessels (SOVs) and helicopters but for the purposes of
this report, only CTV vessel days are analysed.
0.98 0.87 1.06 1.11 1.36 1.39 1.42 1.39 1.22 1.13 1.03 1.0
If additional CTVs are chartered in for only part of the
month then these are included.
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Example:

If 5 vessels are each chartered for 10 days of a month in

a windfarm this equates to 50 vessel days, whether the Figure 5 Average CTV Days per turbine across the year before during and after the pandemic (top) and average CTV Days
vessels were used or not. across the year (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 10
Component Outages
In order to understand what parts of the turbine require the most
attention, the turbine is broken down into several components and
sub-components. The group uses the non-vendor specific component Average Monthly Component Outages per Turbine
taxonomy created by the Reference Designation System for Power
Plants (RDS-PP) Renewables Best Practice group (formerly known as 0.5
300
the RDS-PP Nordic group) - a group of Turbine OOs and OEMs who

Lost Production per Outage (MWh)

agree common identifications for wind turbine subcomponents.

Forced Outages per Turbine

0.4 250

The component that triggers alarms the most is that of the transmission
200
system, the electrical section that transmits the energy provided by 0.3

the generator system into the medium voltage grid. However, the yaw
150
system contributes to more lost production all around due to its higher 0.2
impact on the turbine. The second most frequently triggered group of 100
alarms is that of the control and protection system (CPS). This is akin to
0.1
the central nervous system of the turbine, allowing the exchange of data 50
from turbine control units to the central park communication network.
Failures in the CPS are often incorrectly identified as the root cause as it 0.0 0
is the principal point of measurement.

em

em

em

em

em

em

em

em

em

m
te

te

te

te
st

st

st

st

st

st

st

st

st
ys

ys

ys

ys
Sy

Sy

Sy

Sy

Sy

Sy

Sy

Sy

Sy
rS

bs

tS

bs
on

or

in

ry

n
The 3 components that were most serious upon failing were the balance

Su

Su
an
to

ic

ur
tio

io
Ya

lla
at
si

Ro

ul

Tr

at
os

Pl
n

er
ec

er

ci
is

ra

ric
e

cl

th
of
m

An
en
of plant system (BoP), the drive train system and the central lubrication

ot

yd

no

riv

En

ub

O
ns

Pr

e
G
lH

D
nk

nc

lL
a

ry
nd
Tr

U
tra
system. None of them – particularly BoP and the lubrication system

la

tra
ne
la

Ba
en

hi

en
tro

ac
C

C
– have failed often in the set, but the failures that did occur had large

on

M
C

d
an
consequences.

e
ur
ct
ru
St
What is a forced outage? Figure 6 Average monthly forced outages per turbine and lost production per outage, by component.

A forced outage is defined as when an immediate action to disable

the generating function of the wind turbine is required as unforeseen
damage, faults, failures or alarms are detected. The SPARTA
methodology dictates that members should take the first in a cluster of
alarms as the one that is the most likely root cause of the failure. Forced
Outages do not include major repairs – instances where a jack up barge
is required for significant maintenance or replacement.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 11
Level B Components
Further following the RDS-PP taxonomy, 4 of the turbine’s
components can be split down to another level of detail. Transmission System Failure Breakdown Rotor System Failure Breakdown
Some of these alarms cannot be identified fully as the cause
of failure may be ambiguous and are therefore marked as
‘unknown’.

In the transmission system, the component which is flagged

Hub
the most, the convertor system is flagged as the main root Unknown 33%
Converter Blade Adjustment
Subsystem
cause, constituting 95% of the known transmission failures 44% Subsystem System
53% 37%
at this ‘Level B’.

For the blade, it is the system controlling the pitch that Blade
Other
leads to the highest amount of failures, though many 12%
Subsystem
happen in the hub. In the generator system, the generator 10%
itself fails more often than its cooling system and additional Other Subsystem Transformer System Unknown Subsystem
2% 1% 8%
elements. In the drive train, the gearbox is the most likely
candidate for those turbines that have it.

Generator System Failure Breakdown Drive Train System Failure Breakdown

Unknown
Subsystem
20%
Generator
65% Generator Gearbox
Cooling System Other System
26% Subsystem 52%
19%

Other Main Shaft &

Subsystem Bearing System
Unknown Subsystem 9%
3% 6%

Figure 7 Breakdown of percentage of forced outages for ‘Level B’ subcomponents.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 12
Grid Curtailment and
Availability Loss
Grid Curtailment and Lost Production
Data from the UK Government shows just how important
Offshore Wind is as a portion of the energy mix - as much Share of Electricity Produced by Renewables
as 13% in 2020 and 11.5% in 2021. This presents a new 50

set of challenges for Electricity System Operators (ESOs) 45

UK Electricity Generated (%)

as they try to facilitate the integration of a variable energy 40
35
source in the mix. In its current state, the electricity system
30
is not fit to utilise all of this energy. Particularly in the
25
windiest periods, when supply greatly surpasses local
20
demand and there are no means to transport or store the 13.0
15 11.5
energy, ESOs may have to request that certain wind farms 8.0 9.9
10 6.2
limit their output so that the grid does not get overloaded. 5.1 4.8
3.2 4.0
5 2.1
0
This process occurs through the ‘Bid-Offer-Acceptance
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
(BOA) mechanism, in which an ESO requests ahead of
Offshore Wind Onshore Wind Shoreline wave / tidal
time that an energy producer either limits or increases Solar Photovoltaics Hydro Bioenergy & Waste
their output. In Offshore Wind, where output is generally
limited rather than increased, the producer will usually get
Grid Curtailment in Affected SPARTA Farms
compensated for their losses.
13

As of 2021, National Grid has started publishing records 40 12

of BOAs, allowing us to see their effect on offshore wind.

Compensated Lost Production

Mean Hub Height Wind Speed

11
In Figure 8 we can see that grid curtailment has been

(MWh per MW Installed)

30
significant throughout the Winter when wind speed was at 10
its highest. Other factors such as increases in capacity and
9
demand are also important, as can be seen for comparison 20
of March 2021 (with very little compensated production) and 8
March 2022.
10 7

6
0
5
Mar 21 May 21 Jul 21 Sep 21 Nov 21 Jan 22 Mar 22

Figure 8 Share of electricity produced by renewables in the UK (top) and compensated lost production from BOAs according
to Elexon and mean wind speed (bottom).
Reference:
Data from UK Government (Energy Trends: UK renewables - GOV.UK (www.gov.uk)) and Elexon Portal https://www.elexonportal.co.uk/

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 14
Grid Curtailment and Availability
In the SPARTA methodology, lost production is calculated
for turbines that are classified as down or in partial Yearly Lost Production By Type
performance. This means that curtailment from BOAs will
automatically register as lost production for a farm and Downtime
20 Partial Performance
consequently have an impact on its availability figures.

Lost Production per MW

BOAs may have had a role in increasing lost production 15
figures in past years and is certainly doing so now. In
2020/21, we remarked at lower PBA values likely impacted 10
by disruption to operations through the pandemic. While
this is likely the case, grid curtailment is sure to also play
5
a part. In 2021/22 this seems to be the case more than
ever, with over 10% of values registering less than 80%
0
availability and approximately 10% more that were 80-90% 2017/18 2018/19 2019/20 2020/21 2021/22
available.

The cumulative distribution graph shows the share of Cumulative Distribution of PBA for Different Financial Years
datapoints that were submitted at or below each availability. 1.0
This means that higher lines are actually worse due to the 2017
high share that achieved lower availability figures. The latest 2018
0.8 2019
2 years in the set are the ones with the lowest availability 2020
figures. 2021

0.6

0.4

0.2

0.0
80.0 82.5 85.0 87.5 90.0 92.5 95.0 97.5 100.0
Production Based Availability (%)

Figure 9 Lost production by downtime and partial performance from 2016-2022 (top) and cumulative distribution of PBA for
each financial year (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 15
Accounting for BOA Losses
In order to take account of BOA losses in availability, we
must come up with an alternative definition for availability Availability with and without BOAs
or for lost production. For the purposes of this report, we
will remove estimated BOA losses from the lost production 98
figure. These values are not provided by SPARTA members
but sourced publicly from National Grid and Elexon.
96

As can be seen from Figure 10, PBA has significantly

Availability (%)
dipped throughout the winter of 2021/22, and it seems to 94
be driven by a small number of low producing farms given
the shaded P25 and P75 areas. When this is compared to 92
the figure without losses from BOAs, it’s clear that these
high losses are driven mainly by grid curtailment and the
90 Production Based Availability
true availability is actually quite consistent. In fact, Figure PBA w/o BOA Losses
10 shows that without these losses, the year’s PBA actually Interquartile Ranges Shaded

bested the PBA of recent years. 88

May 21 Jul 21 Sep 21 Nov 21 Jan 22 Mar 22

How can we factor grid curtailment Yearly Average Production Based Availability (PBA)

into availability? 98

In SPARTA, Production Based Availability is calculated 96

using lost production and generation through the
following equation: Availability (%)
94
Generation
PBA = 100 ×
Generation + Lost Production
92

To factor in grid curtailment, we must remove estimated

BOA losses (according to National Grid) from lost 90
production and effectively disregard any curtailment that
occurred. As it is already included in lost production, we 88
may subtract it from the denominator of PBA. 2017/18 2018/19 2019/20 2020/21 2021/22 PBA w/o BOAs
2021/22
Generation
PBA w/o BOA Losses = 100× Generation + Lost Production
– Grid Curtailment Loss
Figure 10 Availability with and without BOA periods counted as potential production, with P25 and P75 range shaded (top) and
PBA without BOAs in annual context (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 16
Regional Grid Curtailment
According to data from Elexon’s Balancing Mechanism
Reporting Service (BMRS), over 1,200 GWh of potential
production was lost from the system due to balancing
Percentage of Potential
Production Curtailed by BOA
constraints throughout the year. This equates to
approximately 2.7% of the year’s total potential production
in the UK. Out of 30 farms identified in the BMRS set 25

25
(including those not in SPARTA), 21 had some level of
grid curtailment during the financial year. Those that were
curtailed in SPARTA were affected on average 4.1 months
out of the year.

Percentage of Potential Production Curtailed by BOA

20

20
As shown by the heat map of grid curtailment across the
country there are 3 farms that are affected more than any
other during the period. The 2 curtailed most often were
15

15
Beatrice and Moray East in the North East of Scotland,
which lost 16% and 23.5% of potential production,
respectively. The third farm was the Walney Extension,
North West of England, which lost 5% of potential
10

10
production. All other farms lost less than 1% of their
potential production due to grid curtailment.

The significant lost production in the North East shows the

truly regional nature of the problem – with constraints on 5

5
transmission and no means to efficiently store offshore wind
power, the grid simply can’t cope with so much supply. An
owner/operator might not complain about this situation as
they’ll get paid anyway, but it highlights a clear mismatch 0

0
WalneyExtension

Beatrice

MorayEast
Extension

Beatrice

East
between national goals and capabilities. In late 2022,
Ofgem approved significant investments in transmission

Moray
infrastructure including substantial investments in
Walney

connections from the North East of Scotland to England.

Figure 11 Percentage of potential production curtailed by BOAs for 3 farms that lost the most (left) and map of losses for farms in
Elexon set (right).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 17
 Edinburgh
Glasgow

Additional Factors
Newcastle upon
Curtailment from ESOs occurred on farms big and small, Tyne
UNITED
though the 3 biggest losses did come from farms in the top K INGDOM
10 in terms of capacity. N orthern Belfast
I rel a nd

The map to the right shows the losses which represented

less than 1% of potential production, with the largest being Douglas

0.75% in Hornsea 1. In total, the losses from all these farms The
accounts for over 83GWh of electricity. Pennines Leeds
Iri sh S ea

In general, grid curtailment losses are more likely to occur in Manchester

Dublin
the winter, when production is highest. A heat map over the
year shows that losses were significantly higher between
November and March. This is true for both the farms in the IRELAND
Nottingham

North of Scotland and those that were curtailed less. The

time of day did not significantly affect curtailment, though E ngl a nd
Birmingham
losses were slightly more common throughout the evening
W a l es
when demand was lowest.

English
Cardiff Bristol London C hannel

C el ti c S ea

Lille

Potential Production Lost (%) Over Year

4.85 5.43 4.22 0.23 0.09 0.05 0.09 0.02 0.71 2.03 4.61 3.25
English
C hannel

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
St Helier

Figure 12 Map of farms with minor losses from BOAs (top) and potential lost production over year for all farms (bottom).
Paris

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review Rennes 18
Operations and
Performance Throughout
the Life of a Turbine
The Lifetime of a Turbine
It is well known in industry that failure rates are likely to
follow a bathtub curve trajectory through the life of an asset, The Bathtub Curve
driven by early ‘infant’ failures and late ‘wear out’ failures.
This corresponds with availability figures of the wind farms
in SPARTA, which appear slightly lower in early life and
lower still in later life.

Failure Rate
The SPARTA fleet contains windfarms from 0 to over 17
years old, but discerning the impact of an ageing system is
no simple task. As industry develops technology develops
with it, meaning faults in older farms may be addressed
and new issues might emerge. Developers are also pushing
further and further offshore, where harsh conditions put Time
systems at higher risk of failure and require more complex
Observed Failure Rate Early Life Failure Wear Out Failure Random Failures
solutions. The farms in early and late life in the set are
therefore operating under various different conditions.

With these diverse conditions in mind, this section PBA Over the Age of a Turbine
examines the SPARTA data through the lens of a turbine
98
going through the course of its first 10 years of life. We
group the dataset based on the age of the turbine at the
time of the instance and examine how these values change 96
across the age of the windfarms.

Through this first decade of a turbine’s life it will go through Average PBA
94
the early ‘infant mortality’ period where system faults will
be diagnosed and resolved. The farm will then see the end
of its warranty period with its manufacturer - a significant 92
landmark for turbine, particularly if the owner/operator
Average PBA
decides to change providers afterwards. Although the farm
P25 and P75
will likely not be in its last phase of life, certain components 90
will also begin to wear out, along with random failures that 0-1 2-3 4-5 6-7 8-9 >=10
will be unrelated to age. Age (Years)

Figure 13 Illustration of bathtub curve (top) and average PBA by year of operation with P25 and P75 (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 20
Outages Over the Life of a Turbine
Examining forced outages, which exclude major
Recorded Outages Over the Age of a Turbine
replacements, there appears to be slightly more failures
in the first years of the farm compared to the next few Mean Forced Outages
years. On average failures decreased in the first 7 years P25 and P75
of life from an initial failure rate in year 0 of 2.6 monthly 2.0

Mean Monthly Forced Outages

outages per turbine. However, the large spread of failure
rates throughout these years makes the application of the
1.5
bathtub curve difficult.

The set contained multiple farms with higher failures in

1.0
years 8-9 of their life. In particular, one farm contained
a significantly higher failure rate during these years that
appeared to relate to a farm-wide issue. For the purposes 0.5
of more general analysis of the set, this outlier has been
removed from these graphs.
0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)
Components are not necessarily built to last the full life of a
turbine. The failure of some minor components reaching the Downtime per Outage Over the Age of a Turbine
end of their lifetime up towards year 10 may also contribute
towards a peak during the 8-9 year period. 20 Mean Downtime per Outage
P25 and P75

Mean Downtime per Outage (Hours)

The length of outages are on average lower before 5 years
of operation. While the majority of outages last less than 5 15
hours long in early life, requiring a mere remote reset, some
outages will require physical maintenance and pull the mean
up. This appears to become more common as the turbine 10

ages and the mean downtime per outage rises above 10

hours.
5

0
0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)

Figure 14 Average number of forced outage by year of operation with P25 and P75 (top) and average downtime per
outage (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 21
Age and Component Failure
Broken down by component, transmission and yaw system
Expected Forced Outages By Component
failures appear to become much less common throughout
the first decade of a turbine’s life. Generator failures have Rotor System
followed a bathtub curve, albeit with a lower rate of failure Transmission System

Monthly Forced Outages per Turbine

0.8 Generator System
than most other components.
Yaw System

It is expected that some components will be replaced 0.6

throughout their lifetime.

Major repairs – those that require a jack-up vessel to fix – 0.4

appear to increase with age. However, there is also a clearly
significant peak in major repairs around the 4-5 years mark.
0.2
This is directly before the end of the warranty period, so it’s
possible that owners will be looking for major replacements
to be dealt with before this ends. The most common major 0.0
repairs in this period are Blades (65.6%), followed by 0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)
Electrical replacements (17.8%). Removing blade repairs
from this set diminishes the peak in failures and shows the Major Repairs Over the Age of a Turbine
large impact of blade repairs in general. The peak in blade
failures is driven by a small number of farms that had blade Major Repairs pT
Major Repairs pT exc. Blades

Mean Monthly Major System Repairs pT

repair campaigns throughout these years. 0.20

Major repairs are rare occurrences compared to other types

of maintenance, and the 75th-percentile is 0 at all ages. 0.15

0.10

0.05

0.00
0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)

Figure 15 Average monthly number of forced outage per turbine by year of operation, per component (top) and average
major repairs per turbine with and without blade repairs (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 22
Changing Operations
Where operations are concerned, it appears that turbine
Recorded Transfers Over the Age of a Turbine
transfers are also more frequent in the first years of life.
However, the large spread of turbine visits makes it difficult 12
Mean Transfers pT
to accurately determine any trend. P25 and P75

Monthly Transfers per Turbine

10
Meanwhile, the CTVs used to transport those technicians
follow a different trend. Farms have hired more CTVs
throughout the warranty period, but have also hired CTVs 8
for longer the older they get. Apart from the first year of
operation, where CTV days are lower, the trend of chartered
6
vessels is also a bathtub curve.

2
0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)

Recorded CTV Days Over the Age of a Turbine

2.0

1.8

Monthly CTV Days per Turbine

1.6

1.4

1.2

1.0

0.8

0.6 Mean CTV Days pT

P25 and P75
0.4
0-1 2-3 4-5 6-7 8-9 >=10
Age (Years)

Figure 16 Average monthly transfers per turbine by year of operation with P25 and P75 (top) and average monthly CTV
days (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 23
O&M Providers
Two fifths of farms in the SPARTA set choose not to be fully
maintained by the OEM after warranty. This means that they Forced Outages by O&M Strategy
3.0
are either fully not maintained by the OEM or they have
Maintained by OEM
some other mixed strategy, in which some elements of the Other Arrangement
2.5
farm are maintained by a third party.

Monthly Forced Outages per Turbine

The dataset exhibits a larger number of Forced Outages 2.0
for farms not fully maintained by OEMs after the warranty
period. While farms still maintained by the OEM keep failure 1.5
rates consistent after warranty, other arrangements may
lead to a high number of outages following the period. This 1.0
could be a result of the different strategy itself, leading to
a higher number of forced outages which have a smaller
0.5
impact per failure.

These extra stops therefore do not necessarily result 0.0

Up to 6 Years (During Warranty) Over 6 Years (After Warranty)
in lower availability. As the PBA shows for non-OEM
maintained farms, high outages do not necessarily result in
PBA by O&M Strategy
lower production. 100
Maintained by OEM
The average transfers per turbine between the strategies 98 Other Arrangement

Average Production Based Availability

was not significantly different.
96

94

92

90

88

86

84
Up to 6 Years (During Warranty) Over 6 Years (After Warranty)

Figure 17 Average number of monthly forced outages per turbine by O&M strategy before and after typical warranty
period (6 years) (top) and corresponding average of PBA (bottom).

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/22 Portfolio Review 24
Summary

In general, the offshore wind industry had another momentous year of

production, with 3135 MWh generated for every MW installed in the SPARTA
portfolio. In spite of this lofty production the year also brought its own
challenges, particularly with increasing intervention from network operators
to limit production in certain farms. Through an analysis of public datasets in
combination with the SPARTA data, we investigated the effect on availability
of curtailments from BOAs. As expected, the main factor in grid curtailment
is region, and until the appropriate transmission capacity is in place then the
country will continue to lose potential renewable production.

The review also put the spotlight on ageing farms, investigating different metrics
compared by age including availability, forced outages and major repairs.
Availability did follow something of a bathtub curve during the first 10 years of
life, with lower PBA in the first 2 years and then also in later life. As for forced
outages, this failure rate appeared to decrease across the lifespan of the set
while the downtime from those outages seemed to increase. Different options
exist for maintenance strategy and the set does seem to suggest that changing
O&M provider has not hindered performance. However, any change of strategy
is very dependent on circumstance and must be taken with caution.

System Performance, Availability and Reliability Trend Analysis – SPARTA: 2021/2022 Portfolio Review 25
Membership
Owner/operators not currently involved in the SPARTA programme
are invited to join the group through the members’ collaborative
agreement, to add to the anonymised benchmarking data set and
benefit quickly from an analysis of their performance against their
peers.

Participation in SPARTA also provides owner/operators with the

opportunity to work with seasoned professionals in the field of
offshore wind farm O&M performance measurement.

Applications or enquiries for new members may be made at any

time by contacting the SPARTA team:

Callum Reid
Business Development Manager

[email protected]

Andrew Yardley
SPARTA Technical Lead

[email protected]