Robust control for variable-speed

two-bladed horizontal-axis wind

turbines via chattering control

L. Acho, Y. Vidal, and F. Pozo

CoDAlab, www-ma3.upc.edu/codalab
Departament de Matemàtica Aplicada III
Escola Universitària d’Enginyeria Tècnica Industrial de Barcelona
Universitat Politècnica de Catalunya
Barcelona, Spain
 Introduction
There are two types of wind control for turbines
◮ constant speed control.
◮ variable speed control.
 Introduction
There are two types of wind control for turbines
◮ constant speed control.
◮ variable speed control.

Constant speed rotors

◮ are designed to deflect high wind gust loads.
 Introduction
There are two types of wind control for turbines
◮ constant speed control.
◮ variable speed control.

Constant speed rotors

◮ are designed to deflect high wind gust loads.

Variable wind turbines

◮ are designed to control strong and gusty winds.
A robust controller is required to mitigate the effects of external
perturbations. This work presents a robust control design for
variable speed control (rotor speed controller).
 Introduction

Principal turbine
configurations
◮ doubly-fed induction
generator with voltage
source converter.
◮ grid-connected
double-output
induction generator.
◮ two-bladed horizontal
axis similar to DOE
MOD-0.
 Introduction
NASA/DOE MOD-0
Principal turbine
configurations
◮ doubly-fed induction
generator with voltage
source converter.
◮ grid-connected
double-output
induction generator.
◮ two-bladed horizontal
axis similar to DOE
MOD-0.
 Nonlinear model

Subsystems of the wind-to-electric power converter model

wind rotor electric
speed speed generator/converter power
wind model rotor model
model

generator
speed
speed controller
 Wind model
Harmonic oscillator

ẍ + ε[(x − x0 )2 − ρ20 ]ẋ + µ2 (x − x0 ) = 0

5
dx/dt

0.5 1 1.5 2 2.5 3 3.5

x

3
x(t)

0 5 10 15
Time[sec.]

Simulations when x(0) = ẋ(0) = 2, µ = 2π, ρ0 = 0.5, x0 = 2, and ε = 10.

 Wind model
Harmonic oscillator

ẍ + ε[(x − x0 )2 − ρ20 ]ẋ + µ2 (x − x0 ) = 0

It does not take into account the

stochastic nature of the turbulence.
5
dx/dt

0.5 1 1.5 2 2.5 3 3.5

x

3
x(t)

0 5 10 15
Time[sec.]

Simulations when x(0) = ẋ(0) = 2, µ = 2π, ρ0 = 0.5, x0 = 2, and ε = 10.

 Wind model
Stochastic harmonic oscillator
Harmonic oscillator

ẍ + ε[(x − x0 )2 − ρ20 ]ẋ + µ2 (x − x0 ) = 0 ẋ1 = x2 + dW(t)

ẋ2 = −ε[(x1 − x0 )2 − ρ20 ]x2 + µ2 (x1 − x0 )
It does not take into account the
stochastic nature of the turbulence.
where W(t) is a Wiener process.
10
5
5
dx/dt

dx/dt
0
0

0.5 1 1.5 2 2.5 3 3.5

x 0.5 1 1.5 2 2.5 3 3.5
x

3 4

3
x(t)

x(t)
2
1
1
0 5 10 15
Time[sec.] 0
0 5 10 15
Time[sec.]

Simulations when x(0) = ẋ(0) = 2, µ = 2π, ρ0 = 0.5, x0 = 2, and ε = 10.

 Rotor model
A simplified rotor model is used in
H. Battista, P. Puleston, R. Mantz and C. Christiansen
Sliding mode control of wind energy systems with DOIG-power
efficiency and torsional dynamics optimization
IEEE Transactions on Power Systems, 15(2), 728–734, 2000.
J. Slootweg, H. Polinder and W. Kling
Dynamic modelling of a wind turbine with doubly fed inductor
generator
Proc. of the IEEE Power Engineering Society Summer Meeting,
644–649, 2001.
Y. Song, B. Dhinakaran and X. Bao
Variable speed control of wind turbines using nonlinear and
adaptive algorithms
Journal of Wind Engineering and Industrial Aerodynamics, 85(3),
293–308, 2000.
 Rotor model
This model assumes the algebraic relation
1
Pm (u) = Cp (λ, β)ρπR2 u3
2

Pm Mechanical power (W) β Pitch angle (deg)

ρ Air density (kg/m3 ) λ Tip-speed ratio, λ = Rw/u
R Rotor radius (m) w Rotor speed (rad/sec)
u Wind speed (m/s) Cp Power coefficient
 Rotor model
This model assumes the algebraic relation
1
Pm (u) = Cp (λ, β)ρπR2 u3
2

Pm Mechanical power (W) β Pitch angle (deg)

ρ Air density (kg/m3 ) λ Tip-speed ratio, λ = Rw/u
R Rotor radius (m) w Rotor speed (rad/sec)
u Wind speed (m/s) Cp Power coefficient

For control design the following parametrization is used

1 R5
Pm (w) = Cp (λ, β)ρπ 3 w3
|2 {z λ}
kw
 Rotor model
The following Cp coefficient [Slootweg et al., 2003] is used
 
151
Cp (λ, β) = 0.73 − 0.58β − 0.002β 2.14 − 13.2 e−18.4/λa
λa
1
where λa = 1
− 0.003 .
λ−0.002β β 3 +1

0.5
Power coeff. (C )
p

β=1
β=5
0 β=10

2 4 6 8 10
Tip speed ratio (λ)

J. Slootweg, S. Haan, H. Polinder and W. Kling

General model for representing variable speed wind turbines in power system
dynamics simulations
IEEE Transactions on Power Systems, 18(1), 144–151, 2003.
 Generator and converter model

Generator/converter system
Tm
T Tp
Tm Input wind torque
w Rotor speed
Generator Te Te Torque produced by the generator
we Angular velocity of the generator
w
we T Torque before the gear box
Tp Torque after the gear box
Je Jm Inertia moment of the turbine
Gearbox Je Inertia moment of the generator
Jm

Aeroturbine
 Generator and converter model
Dynamic model of this system
[Song et al., 2000]:
Tm
T Tp Jẇ + Bw + Kθ = Tm − γTe ,
Generator Te

w
we
J = Jm + γ 2 Je , B =Bm + γ 2 Be ,
we
Je
K = Km + γ 2 Ke , γ= ,
Gearbox w
Z t
Jm
θ(t) = w(τ )dτ,
Aeroturbine
0
where Bm , Km , Be and Ke are the friction and torsional constants of the
aero-turbine and generator, respectively.
Y. Song, B. Dhinakaran and X. Bao
Variable speed control of wind turbines using nonlinear and adaptive algorithms
Journal of Wind Engineering and Industrial Aerodynamics, 85:(3), 293–308, 2000.
 Generator and converter model
The electric power
generated by the system is [Song et al., 2000]: Pe = KΦ we If ,
Kφ Machine constant
If Current field obtained by Lİf + If Rf = uf , where
L Inductance of the circuit
Rf Resistance of the rotor field
uf Control voltage field
Hypothesis: The system operates in the non-saturated magnetic
range [Song et al., 2000].
 Generator and converter model
The electric power
generated by the system is [Song et al., 2000]: Pe = KΦ we If ,
Kφ Machine constant
If Current field obtained by Lİf + If Rf = uf , where
L Inductance of the circuit
Rf Resistance of the rotor field
uf Control voltage field
Hypothesis: The system operates in the non-saturated magnetic
range [Song et al., 2000].

Then, the dynamic model can be rewritten as

Pm Pe
Jẇ + Bw + Kθ = −γ ⇒ Jẇ + Bw + Kθ = kw w2 − γKΦ If .
w we
 Rotor speed controller

The variable speed control objective

is to design a control voltage uf such that the rotor speed w closely
tracks the desired speed (w∗ ).

Because of external perturbations and the need to extract the optimal

value of the wind power, the set point w∗ is variable in time; that is,
w∗ = w∗ (t).

Hypothesis: It is assumed that w∗ , ẇ∗ ,and ẅ∗ are bounded.

 Rotor speed controller

The variable speed control objective

is to design a control voltage uf such that the rotor speed w closely
tracks the desired speed (w∗ ).

Because of external perturbations and the need to extract the optimal

value of the wind power, the set point w∗ is variable in time; that is,
w∗ = w∗ (t).

Hypothesis: It is assumed that w∗ , ẇ∗ ,and ẅ∗ are bounded.

In this work, the pitch angle control is ignored.

 Robust control design
Hypothesis: For the purpose of control design it is assumed that uf is
constant (due to the fast response of the electric part in comparison to
the mechanical part).
Thus, the solution of Lİf + If Rf = uf , is
 
uf Rf Rf
If (t) = 1 − e− L t + If (0)e− L t , t ≥ 0.
Rf
Furthermore, because of the fast converge to zero of the exponential
terms, it can be set uf
If ≃ .
Rf
Then for the purpose of control design, a simplified dynamic model
can be written as:
uf
Jẇ + Bw + Kθ = kw w2 − γKφ .
Rf
 Robust control design
From the simplified dynamic model
uf
Jẇ + Bw + Kθ = kw w2 − γKφ ,
Rf

and taking the robust variable speed voltage control law as

Rf
uf = − [J(ẇ∗ + (w∗ − w) + (θ∗ − θ) + ks sgn(w∗ − w)) + Bw + Kθ − kw w2 ],
γKφ
a robust globally asymptotically stable error dynamic is obtained[Orlov et al,
2003]
ë + ė + e + ks sgn(ė) = 0,

where ks is a control parameter, sgn(·) is the sign function, e = θ∗ − θ (where

Rt
θ∗ = 0 w∗ dτ,), ė = w∗ − w and ë = ẇ∗ − ẇ.
Y. Orlov, J. Álvarez, L. Acho and L. Aguilar
Global position regulation of friction manipulators via switched chattering
control
International Journal of Control, 76:(14), 1146–1452, 2003.
 Simulation results
The following control law is considered:

R̄f
uf = − [J̄(ẇ∗ + (w∗ − w) + (θ∗ − θ) + ks sgn(w∗ − w))
γ K¯φ
+ B̄w + K̄θ − kw w2 ],

where R̄f , K¯φ , J̄, B̄, and K̄ are the nominal values.
Rf = 0.02 Ω, J = 16 kg m2 , B = 52, K = 52, and Kφ = 1.7. (Real)
R̄f = 0.02 Ω, J̄ = 18 kg m2 , B̄ = 50, K̄ = 50, and K¯φ = 1.8. (Nominal)
In addition, we set L = 0.001 H, β = 1 deg., R = 5 m, ρ = 1.21 kg/m3 ,
γ = 37.5 and w∗ (t) = 2 + sin (t) rad/s.
The wind signal u(t) is described by the stochastic harmonic
oscillator.
 Simulation results
Controller performance
w(0) = 1rad/s, If (0) = 0, θ∗ (0) = 0, and θ(0) = 0.
rotor speed and desired speed (rad/s) 3.5

3 w*(t)
2.5
ks=5
ks=1
2 ks=0

1.5

0.5
0 5 10 15
Time (s)
 Conclusions

Conclusions: This work has...

◮ proposed a stochastic harmonic oscillator as a source of
wind speed signal.

◮ introduced a commuted robust controller.

◮ has numerically tested the controller robustness in front of

stochastic variations of the wind and parametric
uncertainties.
 Robust control for variable-speed
two-bladed horizontal-axis wind
turbines via chattering control

L. Acho, Y. Vidal, and F. Pozo

CoDAlab, www-ma3.upc.edu/codalab
Departament de Matemàtica Aplicada III
Escola Universitària d’Enginyeria Tècnica Industrial de Barcelona
Universitat Politècnica de Catalunya
Barcelona, Spain