The influence of wind farm control on

optimal wind farm layout

M. M. Pedersen, G.C. Larsen

Outline

• Introduction
• Tools and wake control procedure
• Three wind turbine row
• The Lillgrund wind farm
• Conclusions

2 DTU Wind Energy, Technical University of Denmark 4 July 2019

Introduction

• Wind farm optimization

– Increase the overall farm production
– Lower costs

• Previous work:
– Turbine layout optimization
– Wake steering via wind farm yaw control
– Combined layout and wake steering (Fleming et al 2016, Gebraad et al 2017)
– Wake control via derating
– Past work has had limited focus on other objectives like load reduction and grid
services

• Research objective: Investigate effect of combined layout / derating optimization wrt.

power production and load reductions

3 DTU Wind Energy, Technical University of Denmark 4 July 2019

Topfarm - DTU Wind Energy’s wind-farm optimizer
• Open source python package
developed by DTU Wind Energy
(gitlab.windenergy.dtu.dk/TOPFARM/
TopFarm2)
• Intended to make windfarm
optimization easier
• Based on the framework
(openmdao.org) Design Optimizer/
Constraints
variables Driver

State
TopFarm Cost
Problem component
Cost
Optimize

Cost, state,
recoder

4 DTU Wind Energy, Technical University of Denmark 4 July 2019

Topfarm - DTU Wind Energy’s wind-farm optimizer
• Open source python package
developed by DTU Wind Energy
(gitlab.windenergy.dtu.dk/TOPFARM/
• Position
TopFarm2)
• Intended to make windfarm • Type
optimization easier • Control
• Based on the framework parameters
(openmdao.org) • … Design Constraints
Optimizer/
variables Driver

State
TopFarm Cost
Problem component
Cost
Optimize

Cost, state,
recoder

5 DTU Wind Energy, Technical University of Denmark 4 July 2019

Topfarm - DTU Wind Energy’s wind-farm optimizer
• Open source python package
developed by DTU Wind Energy
(gitlab.windenergy.dtu.dk/TOPFARM/
• Position • Spacing
TopFarm2)
• Type • Boundary
• Intended to make windfarm
optimization easier • Control • Variable
parameters limits
• Based on the framework
(openmdao.org) • … Design • …
Constraints
Optimizer/
variables Driver

State
TopFarm Cost
Problem component
Cost
Optimize

Cost, state,
recoder

6 DTU Wind Energy, Technical University of Denmark 4 July 2019

Topfarm - DTU Wind Energy’s wind-farm optimizer
• Open source python package
developed by DTU Wind Energy
(gitlab.windenergy.dtu.dk/TOPFARM/
• Position • Spacing • SLSQP
TopFarm2)
• Type • Boundary • IPOPT
• Intended to make windfarm
optimization easier • Control • Variable • Random
parameters limits search
• Based on the framework
(openmdao.org) • … Design • …
Constraints
Optimizer/
• …
variables Driver

State
TopFarm Cost
Problem component
Cost
Optimize

Cost, state,
recoder

7 DTU Wind Energy, Technical University of Denmark 4 July 2019

Topfarm - DTU Wind Energy’s wind-farm optimizer
• Open source python package
developed by DTU Wind Energy
(gitlab.windenergy.dtu.dk/TOPFARM/
• Position • Spacing • SLSQP
TopFarm2)
• Type • Boundary • IPOPT
• Intended to make windfarm
optimization easier • Control • Variable • Random
parameters limits search
• Based on the framework
(openmdao.org) • … Design • …
Constraints
Optimizer/
• …
variables Driver

• AEP (PyWake,
FusedWake,
State Floris, …)
TopFarm Cost
Problem • LCOE
component
Cost
• TopFarmProblem
Optimize
• …
Cost, state,
recoder

8 DTU Wind Energy, Technical University of Denmark 4 July 2019

PyWake
• Open source python package developed by
DTU Wind Energy WT positions (x,y)
(gitlab.windenergy.dtu.dk/TOPFARM/PyWake) WT types (t)

• AEP calculator for wind farms

• Included wake models
WindTurbines
– NOJ
– Bastankhah Power
WS_eff
– Fuga (Commercial software required type
CT
D, h
to generate input data) Ref ws, wd and ti
Local ws and wd
Position, (x, y, z)
• Flexible and customizable WT positions
Site AEPCalculator WakeModel
Local ws, wd, ti
Power
P(wd, ws) CT,
WS_eff
Distances D, h
WS_eff WS,
distances
etc.

NOJ/Fuga/
Bastankhah

ws_eff
AEP
Wake map
AEP map
9 DTU Wind Energy, Technical University of Denmark 4 July 2019
PyWake - Fuga
Simple engineering approach

For each wind direction

For each wind speed
For each turbine (down wind order)
ws_eff = free_ws – sum(deficits)
deficits = f(ws_eff, Ct, dw_dist, cw_dist)

10 DTU Wind Energy, Technical University of Denmark 4 July 2019

Fuga
• Commerical flow solver developed by DTU
Wind Energy (wasp.dk/fuga)
• Linearized CFD RANS solver
• Mixed spectral domain and look-up tables
• Very fast compared to traditional RANS solvers
(PyWake 48 wt, 360 wd, 23 ws: <0.5s)

11 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Derating
– Derate upstream turbines
• Pitch angle: f(wt, wd, ws)
• Rotor speed: g(wt, wd, ws)
– Increase total wind farm power

12 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Derating
– Derate upstream turbines Assumption:
• Pitch angle: f(wt, wd, ws)
Wind direction and ambient wind speed known
• Rotor speed: g(wt, wd, ws) and constant over the wind farm area
– Increase total wind farm power

13 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Power and thrust coefficients for range of rotor speeds and pitch angles
• Calculated using HAWCSTAB2 (steady-state aeroelastic code
http://www.hawcstab2.vindenergi.dtu.dk/)

14 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Power and thrust coefficients for range of rotor speeds and pitch angles
• Calculated using linear steady-state aeroelastic code (HAWCSTAB2)

15 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Greedy operation
• Max Cp
• Constant Ct

16 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Power limit
• Decrease Cp
• Minimize Ct

17 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Power and Rotor speed limits
• Decrease Cp
• Minimize Ct

18 DTU Wind Energy, Technical University of Denmark 4 July 2019

Active wake control - derating
• Power and Rotor speed limits
• Decrease Cp
• Minimize Ct

19 DTU Wind Energy, Technical University of Denmark 4 July 2019

 Single row, 3 wind turbines

WT1 WT2 WT3

10m/s

4D 4D

20 DTU Wind Energy, Technical University of Denmark 4 July 2019

 Single row, 3 wind turbines

WT1 WT2 WT3

10m/s

4D 4D

21 DTU Wind Energy, Technical University of Denmark 4 July 2019

 Single row, 3 wind turbines

WT1 WT2 WT3

10m/s

4D 4D

22 DTU Wind Energy, Technical University of Denmark 4 July 2019

 Single row, 3 wind turbines

WT1 WT2 WT3

10m/s

4D 4D

23 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

WT1 WT2 WT3

12m/s

24 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

WT1 WT2 WT3

12m/s

25 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

WT1 WT2 WT3

26 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

WT1 WT2 WT3

27 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

WT1 WT2 WT3

28 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

Sequential optimization

Layout PyWake
Optimizer

Optimized layout

WT1 WT2 WT3

29 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

Sequential optimization

Layout PyWake
Optimizer

Optimized layout

Control
PyWake
Optimizer
Optimized layout
WT1 WT2 WT3 Optimized control

30 DTU Wind Energy, Technical University of Denmark 4 July 2019

Single row, 3 wind turbines

Nested optimization

Layout Control PyWake

Optimizer Optimizer

Optimized layout
Optimized control
Including ”cocktail” effects

WT1 WT2 WT3

31 DTU Wind Energy, Technical University of Denmark 4 July 2019

Nested optimization

Layout Control PyWake

Optimizer Optimizer

Optimized layout
Optimized control
Including cocktail effects

32 DTU Wind Energy, Technical University of Denmark 4 July 2019

Nested optimization

Layout optimizer:
- Random search
- Gradient based Layout Control PyWake
(finite difference) Optimizer Optimizer

Optimized layout
Optimized control
Including cocktail effects

33 DTU Wind Energy, Technical University of Denmark 4 July 2019

Nested optimization

Layout optimizer:
- Random search
- Gradient based Layout Control PyWake
(finite difference) Optimizer Optimizer

Control optimizer:
- Gradient based Optimized layout
(finite difference) Optimized control
Including cocktail effects

34 DTU Wind Energy, Technical University of Denmark 4 July 2019

Nested optimization

For all wind directions (360):

Layout optimizer:
for all wind speeds (23):
- Random search
- Gradient based Layout Control PyWake
(finite difference) Optimizer Optimizer

Control optimizer:
- Gradient based Optimized layout
(finite difference) Optimized control
Including cocktail effects

35 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund
• 48 Siemens wind turbines
• 2.3MW
• 93m rotor

3.3D

4.3D

36 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund
• 8 wind turbines
• 3.3D spacing

3.3D

4.3D

37 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

38 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

Initial layout,
44.10 (+8.0%)
optimized control

39 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

Initial layout,
44.10 (+8.0%)
optimized control

Optimized layout,
41.44 (+1.4%)
greedy control

40 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

Initial layout,
44.10 (+8.0%)
optimized control

Optimized layout,
41.44 (+1.4%)
greedy control

Optimized layout,
44.558 (+9.1%)
optimized control (sequential)

41 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

Initial layout,
44.10 (+8.0%)
optimized control

Optimized layout,
41.44 (+1.4%)
greedy control

Optimized layout,
44.558 (+9.1%)
optimized control (sequential)

Optimized layout and control

44.560 (+9.1%)
(nested)

42 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund row
AEP [Gwh]

Initial layout,
40.85
greedy control

Initial layout,
44.10 (+8.0%)
optimized control

Optimized layout,
41.44 (+1.4%)
greedy control

Optimized layout,
44.558 (+9.1%)
optimized control (sequential)

Optimized layout and control

Infeasible for full wind farm with current setup/models
44.560 (+9.1%)
(nested)

43 DTU Wind Energy, Technical University of Denmark 4 July 2019

Lillgrund, initial layout

Initial layout

Control AEP [GWh]

Greedy 345.2

Optimized 349.5 (+1.3%)

44 DTU Wind Energy, Technical University of Denmark 4 July 2019

Lillgrund optimized layout

Initial layout

Control AEP [GWh]

Greedy 345.2

Optimized 349.5 (+1.3%)

Optimized layout
(sequential)
Control AEP [GWh]

Greedy 354.9 (+2.8%)

Optimized 358.7 (+4.0%)

45 DTU Wind Energy, Technical University of Denmark 4 July 2019

Conclusions

• Production gains found

• Layout optimization (+2.8% for Lillgrund)
• Control optimization (+1.3% for Lillgrund)
• Sequential layout and control optimization (+4.0% for Lillgrund)
• Nested layout and control optimization of full wind farm infeasible with
current setup/model
• Cocktail effects observed for 8-WT row, but insignificant

• Future work:
• Include uncertainty of wind direction and wind speed
• Include yaw control and wake steering
• Integrate loads surrogates to address reliability and overall LCOE

46 DTU Wind Energy, Technical University of Denmark 4 July 2019

 Thank you for your attention

This study is funded by

TotalControl is an EU Horizon2020 funded project running from 2018-2021.

47 DTU Wind Energy, Technical University of Denmark 4 July 2019

Lillgrund – AEP increase

48 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund 3x3

Initial layout

Control AEP [GWh]

Greedy 73.28

Optimized 73.61 (+0.5%)

49 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund 3x3

Initial layout Optimized layout

(First layout then control)

Control AEP [GWh] Control AEP [GWh]

Greedy 73.28 Greedy 73.86 (+0.8%)

Optimized 73.61 (+0.5%) Optimized 74.18 (+1.2%)

50 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund 3x3

Initial layout Optimized layout Optimized layout

(Sequential) (Nested layout and control)

Control AEP [GWh] Control AEP [GWh] Control AEP [GWh]

Greedy 73.28 Greedy 73.86 (+0.8%) Optimized 74.15 (+1.18%)

Optimized 73.61 (+0.5%) Optimized 74.18 (+1.23%)

51 DTU Wind Energy, Technical University of Denmark 4 July 2019

LillGrund 3x3

Initial layout Optimized layout Optimized layout

(First layout then control) (Combined layout and control)

Infeasible with current setup

Control AEP [GWh] Control AEP [GWh] Control AEP [GWh]

Greedy 73.28 Greedy 73.86 (+0.8%) Optimized 74.15 (+1.18%)

Optimized 73.61 (+0.5%) Optimized 74.18 (+1.23%)

52 DTU Wind Energy, Technical University of Denmark 4 July 2019