(max. 2000 char.): The three different controller designs presented herein are similar and all based on PI-regulation of rotor speed and power through the collective blade pitch angle and generator moment. The aeroelastic and electrical modelling used for the time-domain analysis of these controllers are however different, which makes a directly quantitative comparison difficult. But there are some observations of similar behaviours should be mentioned: • Very similar step responses in rotor speed, pitch angle, and power are seen for simulations with steps in wind speed. • All controllers show a peak in power for wind speed step-up over rated wind speed, which can be almost removed by changing the parameters of the frequency converter. • Responses of rotor speed, pitch angle, and power for different simulations with turbulent inflow are similar for all three controllers. Again, there seems to be an advantage of tuning the parameters of the frequency converter to obtain a more constant power output. The dynamic modelling of the power controller is an important result for the inclusion of generator dynamics in the aeroelastic modelling of wind turbines. A reduced dynamic model of the relation between generator torque and generator speed variations is presented; where the integral term of the inner PI-regulator of rotor current is removed because the time constant is very small compared to the important aeroe-lastic frequencies. It is shown how the parameters of the transfer function for the remaining control system with the outer PI-regulator of power can be derived from the generator data sheet. The main results of the numerical optimisation of the control parameters in the pitch PI-regulator performed in Chapter 6 are the following: • Numerical optimization can be used to tune controller parameters, especially when the optimization is used as refinement of a qualified initial guess. • The design model used to calculate the initial value parameters, as described in Chapter 3, could not be refined much in terms of performance related to the flapwise blade root moment (1-2 %) and tilt tower base moment (2-3 %). • Numerical optimization of control parameters is not well suited for tuning from scratch. If the initial parameters are too far off track the simulation might not come through, or a not representative local maximum obtained. The last problem could very well be related to the chosen optimization method, where more future work could be done.
[1]
P. Sørensen,et al.
Simulation of interaction between wind farm and power system
,
2002
.
[2]
Torben J. Larsen.
Description of the DLL regulation interface in HAWC
,
2001
.
[3]
P. Sørensen,et al.
Wind models for simulation of power fluctuations from wind farms
,
2002
.
[4]
J. Mann.
Wind field simulation
,
1998
.
[5]
D. E. Goldberg,et al.
Genetic Algorithms in Search, Optimization & Machine Learning
,
1989
.
[6]
David E. Goldberg,et al.
Genetic Algorithms in Search Optimization and Machine Learning
,
1988
.
[7]
Poul Ejnar Sørensen,et al.
Wind models for prediction of power fluctuations from wind farms
,
2001
.
[8]
J. Thirstrup Petersen.
The aeroelastic code HawC - model and comparisons
,
1996
.