Wind turbine control
Modern wind turbines in MW class
operate in wide spectra of environmental conditions from calm to stormy winds. Wind
power (mechanical power of moving air) increases with third power of wind speed
so wind turbine characteristics change drastically with change of wind speed.
Turbine operation can be divided in two very distinct areas: below and above
nominal wind speed.
Below nominal wind speed wind
power is lower than turbine nominal power output. In this region control system
has to assure maximization of wind energy capture through optimal control
strategies. Modern wind turbines are connected to electrical grid through
frequency converter which allows them to operate at variable rotational speed. At each
wind speed there is only one rotational speed that will result in maximal
energy capture. Controlling rotational speed to its optimal value at each wind
speed becomes difficult because wind speed can not be measured fast and
Above nominal wind speed wind
power is greater than turbine nominal power output and it increases rapidly. In
this region it is necessary to limit the power in order to keep the turbine
running. Modern turbines for this task use blade pitching (turning the blades
around its axis) so this control concept is called Pitch control. Blade's
aerodynamic properties worsen with increase of pitch angle which results in
lower energy capture. Thus it becomes possible to limit the turbine power and rotational
speed to desired value at wide range of wind speeds. Since the turbine
electrical power is proportional to its rotational speed, power control is
carried out by controlling the rotational speed. Wind turbine dynamics are
nonlinearly dependent on wind speed, rotational speed and blade pitch angle. To
be able to control the turbine rotational speed it is necessary to allow
control system to adapt in order to account for changes in turbine's dynamics.
Our first approach to this
problem was to apply linear PID controller that uses gain scheduling dependent
on current operating point. Linear models were found on various wind speeds
using system identification methods and controller for each operating point was
designed. Since experimenting possibilities on real turbine are limited we used
professional simulation package Garrad Hassan Bladed to simulate wind turbine
system. GH Bladed shown in the picture bellow uses very complex wind turbine model which allows modeling
turbine's aerodynamic and structural properties and simulations of their
dynamical behavior when wind turbine is excited with time varying winds. Its simulation
results are very well correlated with actual measurements done on real wind
turbines so world's largest standardization and certification companies (among
others Germanischer Lloyd) recognize Bladed results as reliable as those obtained
on actual plant.
In this way we could treat Bladed
model as actual plant and perform experimental system identification on it.
This work has been done at Končar – Electrical Engineering
Designed PID controller with gain scheduling showed satisfying
performance in entire operating region. An example of simulated wind turbine behavior is shown in the picture bellow.
Described controller has a
drawback that gain scheduling introduces discontinuities in control signal.
Also sudden changes of wind speed drive system into regions that are away from
operating points at which system identification was carried out. Another problem
that is related to pitch control deserves attention. It increases loads on the
wind turbine structure as well as nodding of the tower. Excessive loads lead to
pronounced fatigue and premature structural failure.
Taking all said into account our
current research efforts focus on the following:
Finding wind turbine mathematical model that will be suitable
for controller design.
Finding suitable adaptation law for controller parameters that
will account for system nonlinearities.
Estimating structural loads and tower motion and modifying control
system in order to keep these loads in desired bounds.
Another important step in wind
turbine design is modeling wind properties specific for wind turbine location.
Wind speed varies in space and time what is commonly called turbulent behavior.
Today there are mathematical models that describe wind turbulent nature. Using
these models it is possible to create wind speed time series in discrete points
that cover turbine rotor as shown in the picture bellow. Those time series have predefined autocorrelation and
cross correlation that resembles actual turbulence observed in nature.
All those models have certain
number of parameters that denote nature of turbulent wind. In our research we
tried to model turbulent wind that would be specific for chosen wind turbine
site. We analyzed spectral properties of measured wind speeds and tried to fit
them with chosen mathematical model by varying its parameters. In that way we
could obtain needed turbulence parameters that can be used later on for