Transcript EOG - EWEA

INVESTIGATION OF EOG COMBINED WITH GRID
FAILURE IN THE CASE OF OFFSHORE WIND
TURBINES
Niels Jacob Tarp-Johansen
Presented at EWEC2007, 7-10 May 2007, Milan, Italy
Outline
 Motivation
 Background of present load case
 Presentation of an alternative approach
 A preliminary example
2
Motivation
 The load case with EOG combined with electrical failure is a design driver for
foundations for offshore wind turbines – DLC 2.3
 This is at least valid for modern MW pitch-regulated machines
 It might be different for stall turbines, but these are not considered at state-of-theart for offshore conditions is pitch-regulated turbines
3
Background for present load case for grid failure
 The gust size seems to be obtained by distributions independent of 10-min mean
wind but conditional on start wind speed
 In other words: it seems that an appropriate weighting of distribution of start wind
speed, gust size and e.g. 10-min mean wind speed is missing.
 Response has been considered to some extend, as the slope of the gust has
been regarded.
 The shape has support in data. The appropriateness of the full coherence
appears not to have been validate and has been accepted as conservative.
 Finally some change of gust size has taken place to ensure that together with the
probability of operational events like starts and stops a 50-yr return period event
was obtained
4
Alternative approach: Aim and Idea of the method
1/2

Aim: to formulate a general method which considers response. The method
shall determine wind, and sea state conditions to apply in combination with the
event of electrical failure

Idea:
1. The starting point is statistics of the rate of electrical failures and the jpdf. Of climate
parameters: U, I, Hs, …
2. Then one determines the ‘normal’ combinations of electrical failure and climate
parameters
3. One finalizes by performing stochastic response simulations for these normal
conditions

So: the method does not aim at determining abnormal combinations of gusts
and electrical failure
5
Alternative approach: Aim and Idea of the method

Three effects are included in the proposed method relative to DLC 2.3
 Replacing coherent gust with simulation of turbulent field
 (Influence of eigenfrequency relative to gust duration)
 Relaxing demand on worst case phasing of gust with electrical failure
 Determining climate parameters in situations with electrical failure by different
rationale
6
Alternative approach – Theory: Poisson model

1/2
It is assumed that the events of electrical failure follows a Poisson process
with constant intensity
ef  the rate of electrical faults per unit time

An electrical failure may potentially lead to structural failure
psf |ef  Pr{structural failure | electrical fault}

Consequently the events where electrical failure leads to structural failure
constitute a Poisson process too with constant intensity
sf  psf |ef ef
7
Alternative approach – Theory: Poisson model

An event with return period T is determined by requiring
sf 

2/2
1
T

psf |ef 
1
ef T
This gives us the target probability of structural failure for characteristic values
to be used in design against electrical failure
8
Alternative approach – Theory: Failure probability

1/2
The structural failure probability depends on
 Loads and strengths
 Loads in turn depends on climate, design, and the applied control strategy in case of
electrical failure.

This may be expressed through
psf |ef 

Pr{structural failure | V , I , H s ,
, R1 , R2 , } 
V , I , H s , , R1 , R2 ,
f (V , I , H s ,

, R1 , R2 , )dVdIdH s
dR1dR2
Dependence between climate and electrical failure is accounted for by the
choice for f (V,I,Hs,…)
9
Alternative approach – Theory: Failure probability

2/2
Solve the equation with respect to design variables z
psf |ef 

, R1 , R2 , } 
Pr{structural failure | V , I , H s ,
V , I , H s , , R1 , R2 ,
f (V , I , H s ,
dR1dR2
0.5
This can be done iteratively by use of
normalized response

, R1 , R2 , )dVdIdH s
 Extrapolation
 FORM or IFORM
 …
0.4
0.3
0.2
0.1
4
3
2
1
10
 [m/s]
0
0
5
10
15
V [m/s]
20
25
30
Standardisation – how?

Extrapolation may be allowed – this is probably less problematic than it
sounds

An alternative similar to the idea behind ETM + stochastic response
simulations may be developed

An alternative EOG dependent on turbine characteristics may be devised

The two first options are the preferred ones

Potentially the EOG should be included in extra/alternative load cases that
aims at verifying the control system behaviour
11
A preliminary example
1/2

We focus on the wind only: IEC 61400-1 model IB

Use IFORM extended to include response
 NB most distributions are close to normal implying expectably small error

Compare results with DLC 2.3

Assumptions: electrical failure on the one side and wind and sea state
conditions on the other side are statistically independent
 storm events that potentially lead to grid loss (overhead lines clashing) will be
geographically separated from offshore farms
 ship dragging an anchor over the sea bottom may cause damage to the cable

Pitch speed: 7.5 deg/sec, tower frequency = 0.38 Hz
12
A preliminary example

2/2
Simulations are performed in this way:
 For a number of combinations of U and I
 For each combination 100 simulations are carried out
 The same 100 seeds are used for each U, I.
 Simulation until transients have died out
 Extremes after electrical failure is detected
 Overturning moment.
13
A preliminary example: conclusions

The proposed method yields characteristic response about a factor of 2
smaller than DLC 2.3 - practically independent of the rate of electrical failure
(4 – 50 yr-1)

BUT ... Safety level …

Safety factors must be added
 Electrical failure + EOG is abnormal, that is gf = 1.1
 For the method presented here the situation is normal. Thus probably gf = 1.35 (but
this remains the be proven)

Consequently one has FEOG, design ≈ 1.5 Felec. failure, design
14
Further work …

Simulation with other turbines:
 Control strategy
 Tower frequencies

Separate the three effects
 Replacing coherent gust with simulation of turbulent field
 Relaxing demand on worst case phasing of gust with electrical failure
 Determining climate parameters in situations with electrical failure by different
rationale

Safety factor assessment

Discuss modelling of dependence between grid failure and climate
parameters
15
Acknowledgements

Discussions with colleagues inside and outside DONG Energy

Public Service Obligation founds from EnergiNet.dk
16