Abstract:
|
The fast development of the wind power technology is leading to larger and more
expensive wind turbines which require increasingly advance control systems to
achieve optimal or near optimal operation. However, the increased optimality of
complex operation schemes, like real-time optimization approaches, incur in high
implementation and maintenance costs. Furthermore, while the wind speed is a key
variable for the wind turbine control, using it as a direct input leads to a poor response
of the power control. This makes the industry focus on simpler control structures,
considering wind as a disturbance [1]. Nevertheless, baseline control laws, which
perform a deterministic control, require that complex aerodynamic properties are wellknown
to achieve the desired performance. But in practice, variability bounds the
efficiency of the energy capture. Thus, a constrained self-optimizing control is
proposed to regulate the wind turbine operation coping with wind speed uncertainty.
A data-driven self-optimizing control is proposed for the wind turbine control region
where power is maximized (region 2). Operational data is extracted from a model offline
to examine the structure of the optimal solution. This insight is then transformed
into a simple control structure capable of keeping the wind turbine to an optimal
operation, in terms of maximizing power output. However, at high wind speed, wind
turbine power output has to be maintained at its nominal rate. Thus, a cascade control
structure for self-optimizing and constrained control is incorporated. The control
structure is implemented in Simulink using as a model FAST v8 5MW onshore wind
turbine model.
The proposed self-optimizing control learns the structure of the optimal solution offline
and then performs the optimization strategy, adjusting both torque and pitch to
maximize energy capture. This control approach leads to an increase in power output
when comparing it with the deterministic baseline control. Moreover, this heuristic
control approach has the potential to take into account a higher number of inputs
without compromising reliability. This property allows its future application for more
advance control strategies. |