O&M Cost Modelling, Technical Losses and Associated

Uncertainties

Axel AlbersDipl.-Phys.

Deutsche WindGuard Consulting GmbHOldenburger Straße 65, D26316 Varel

a.albers@windguard.d

testing- and calibration laboratory with

quality management system according EN ISO/IEC 17025:2005

DAP-PL-3565.99

for power curve measurements,wind measurements,wind resource assessments

for power curve measurements

Contents• Modelling of O&M cost development in time• Comparison of modelled O&M cost with real wind farm data• Development of WT availability in time• Combination of uncertainties of wind resource, expected technical losses and O&M cost

O&M Cost Issues• Questions:

1. How large are maintenance and repair cost?

2. How do maintenance and repair cost develop with the age of WT’s?• Problems

A) analysis of future O&M-cost on component by component basis often impossible

B) accurate public data on O&M cost is rare

C) only limited number of wind farms in high age (15-20a)

D) rapid development of technology and size of WT in last decade• Several studies show rise in Q&M-cost with WT age (BWE, WMEP, University of Durham and TU Delft)

Operational Costs by WT Age According to WMEP 2006

operating year

repair maintenance insurance property other

Source: WMEP 2006

Warranty Period

Rise of Repair and Maintenance Cost BWE 2002 Study

• In second decade cost twice as high as in first decade (BWE study 1999)• 12 €/MWh/a average cost over 20 years (corresponds to BWE 2002 numbers if 2000 full load hours are assumed), well in line with cost of most full service contracts• Often assumed by wind farm developers or financiers:

- no rise with age

- step functions: after 10a or steps every 5th year

0

2

4

6

8

10

12

14

16

18

0 2 4 6 8 10 12 14 16 18 20

Operating Year [a]

O&

M C

ost

[€/M

Wh/

a]

1.8% of invetsment cost per year

3.6% of invetsment cost per year

Improved Approach for Rise in Repair and Maintenance Cost

• There is no reason to assume a step function.• Integrated cost increase in 2nd decade compared to 1st decade overtaken

from BWE 2002 study (double cost in 2nd decade)

0

5

10

15

20

25

30

0 2 4 6 8 10 12 14 16 18 20

Operating Year [a]

O&

M C

ost

C [

€/M

Wh

/a]

acc. BWE 2002 study exponential cost increase physical model

Physical Model for Rise in Repair and Maintenance Cost

• Damage increase per time is inversely proportional to remaining lifetime L-t

- L: total lifetime

- t : age• Damage increase is proportional to increase of cost C

tL

1

dt

dC

21 const)tLln(constC

tL

LlnconstC 00tC condition 1

a2.26L Cdt2

1Cdt condition

a20t

a10t

a10t

0t

Fit of Model for Rise of Repair and Maintenance Cost to Observations

• In single years large outliers are observed, but the cumulated cost is fitted well by model.• Always the same cost rise has been assumed (twice as much cost in 2nd decade than in 1st

decade).

0

5

10

15

20

25

30

35

0 2 4 6 8 10 12

Operating Year [a]

O&

M C

ost

[€/M

Wh

/a]

1.8MW direct drive, pitch, 7€/MWh/a in 20a

1.5MW gearbox, stall, 12€/MWh/a in 20a

2.0MW gearbox, pitch, 12€/MWh/a in 20a

1.3MW gearbox, active stall, 17€/MWh/a in 20a

2.0MW nearshore, Middelgrunden, 24€/MWh/a

gearbox damage

Extrapolation of Repair and Maintenance Cost

• standard approach : 0.012€/kWh/a averaged over 20a assumed (value from BWE-study 2002)

• Estimated standard uncertainty of standard approach: 50% of modelled cost

0

5

10

15

20

25

30

35

0 2 4 6 8 10 12 14 16 18 20

Operating Year [a]

O&

M C

ost

[€/M

Wh

/a]

1.8MW direct drive, pitch, 7€/MWh/a in 20a1.5MW gearbox, stall, 12€/MWh/a in 20a2.0MW gearbox, pitch, 12€/MWh/a in 20a1.3MW gearbox, active stall, 17€/MWh/a in 20a2.0MW nearshore, Middelgrunden, 24€/MWh/astandard approach onshore, 0.012€/kWh/a in 20a

gearbox damage

Observations in Old Wind Farms In Respect to Availability

• Experience based on:

- hundreds of WT’s in age 12-20a in Eastern-Frisia (backyard of WindGuard)

- due diligence in the frame of sales of wind farms

- technical management of wind farms• Availability normally high, but within year 5 to 15 single events with long standstills likely• One event with 3 months standstill leads to 2.5% additional non-availability over 10 years• Consequence: 97% availability hardly possible in 2nd decade in case of only a single extraordinary event

Model for Increase of Availability Losses in Time

• The initialisation is treated case dependent:

- adjustment according to warranties or

- adjustment according to availability of past operating period or

- adjustment according to experience with wind turbine type

0

2

4

6

8

10

12

14

0 2 4 6 8 10 12 14 16 18 20

Operating Year [a]

Exp

ecte

d L

osse

s d

ue

to N

on-A

vaila

bili

ty [

%]

Standard approach initialised with 3% average availability losses in first 10 years.Resulting losses averaged over first 20 years: 4.5%

Example for Increase of Availability Losses in Time

• Model adjusted to observed non-availability losses in past (after initial project stage with teething problems)

0

2

4

6

8

10

12

0 2 4 6 8 10 12 14 16 18 20

Operating Year [a]

Non

-Ava

ilab

ilit

y L

osse

s [%

]

observed non-availability losses, running annual mean fitted model

teet

hing

pha

se: 7

.6%

loss analysed period: 2.2% loss extrapolation period: 4.3% loss

Combination of Uncertainties of Wind Resource (Revenue) and O&M-Cost

• Problem 1: The lower the wind resource, the lower the wear (repair cost)Solution:i) calculate P50-value of O&M-cost and standard uncertainty of O&M- cost on the basis of Px-value of the production estimateii) combine standard uncertainty of wind resource (revenue) and new standard uncertainty of O&M-cost as independent uncertaintiesiii) consider normal distribution of difference of revenue and O&M- cost with the combined standard uncertainty and calculate Px-value of this distribution

0123456789

0 100 200 300 400 500 600

revenue, earnings [k€]

f[%

]

f-revenue f-O&M-cost at P-revenue=50%

f-O&M-cost at P-revenue=90% f of earnings (revenue - O&M-cost)

Combination of Uncertainties of Non-Availability Losses (Revenue) and O&M-Cost

• Problem 2: increase of O&M-cost with age highly correlated with increase of non-availability losses• Solution:

combine standard uncertainty of O&M-cost and standard uncertainty of revenue due to non-availability losses linearly

0

100

200

300

400

500

0 5 10 15 20

age [a]

reve

nue,

O&

M-c

ost,

ear

ning

s [k

€]

revenue with standard uncertainty due to non-availability(bars), no uncertainty of wind resource consideredO&M cost with standard uncertainty (bars)

earnings with combined standard uncertainty (bars)

Result of Combination of Uncertainties

• final result: risk assessment of net earnings

0

50

100

150

200

250

300

350

0 5 10 15 20Operating Year [a]

Dif

fere

nce

Ear

nin

gs -

O&

M C

ost

[k€]

P50 P75 P90

Example:2MW turbine23% capacity factor0.1€/kWh tariff0.012€/kWh/a O&M cost15% other cost14% standard uncertainty wind resource97% average availability in year 1-10

2% other losses than availability

moderate wind resource, high tariff high wind resource, low tariff

0

50

100

150

200

250

0 5 10 15 20

Operating Year [a]

Dif

fere

nce

Ear

nin

gs -

O&

M C

ost

[k€]

P50 P75 P90

Example:2MW turbine34% capacity factor0.05€/kWh tariff0.012€/kWh/a O&M cost15% other cost14% standard uncertainty wind resource97% avergae availability in year 1-10

2% other losses than availability

Conclusions• Individual modelling of expected O&M cost and availability over project lifetime is recommended

• Long-term O&M cost often underestimated in planning phase:- often about 85% EBITDA-margin expected over 20a- latest study of BWE from 2009: 76% EBITDA-margin (average of 66 wind farms)

• Cost modelling often results in positive earnings even after 20 years

• Problem: type certificate valid only 20 years- building permit of WT may lose validity - extension of type certificate to longer period in most cases not possible

i) high costii) WT design often not conform with latest revision of IEC 61400-1