Beyond the CP-curve in Model-based Control of Wind Turbines

1
Beyond the CP-curve in Model-based Control of
Wind Turbines

• Lars Christian Henriksen, DTU Wind Energy
• Morten Hartvig Hansen, DTU Wind Energy
• Niels Kjølstad Poulsen, DTU Informatics

2
Outline

• Control Design Model
• Model Components
• Aerodynamic model Dynamic inflow (DI) vs.
  Quasi-steady (QS)
• Controller
• Extended Kalman Filter (EKF) and disturbance
  estimation
• Model Predictive Control (MPC)
• Partial and full load operation and switching the
  two
• Results
• PI vs. MPC1-EKF(Q.S/D.I.) vs. MPC2-EKF(Q.S/D.I.)
• Concluding Remarks

3
Control Design ModelWind Turbine Model Components

• Tower fore-aft DOF
• Low speed drive-shaft torsional DOF
• Pitch actutator
• Generator torque actuator
• Wind turbulence model
• Aerodynamic model Q.S. or D.I.
• Cyclic/individual pitch using the
  Coleman-transformation is possible in the
  presented model. However, only collective pitch
  is considered in the presented work

4
Control Design ModelBlade Element Momentum Theory
Quasi-steady model

• Simplified dynamic inflow model

5
Control Design Model Dynamic vs Quasi-steady
Inflow
6
Control Design ModelBode plots From col. pitch
to gen. speed
8 m/s

• 16 m/s

7
ControllerController Setup

• Extended Kalman Filter (EKF) estimates states and
  disturbances using avaliable sensors
• Model Predictive Controller (MPC) computes
  control action based on estimated states and
  disturbances
• Switching mechanism assigns control objectives
  for partial and full load operation,
  respectively. Furthermore, smooth switching
  between modes of operation is achieved.

8
ResultsControllers Compared

• Controllers
• PI Controller (black line, o) Jonkman
• MPC controllers based on either Q.S. or D.I.
  aerodynamic models (blue and green lines,
  respectively)
• MPC1 () weights on generator speed and generator
  power tracking and frequency dependant weights on
  control signals
• MPC2 (x) like MPC1 but with lower weight on
  generator speed.
• Comparison metrics
• Root mean square (RMS) error of generator speed
  and generator power with regards to their nominal
  values
• Fatigue Drive-train and tower bottom
• Control activity Pitch travel and standard
  deviation (std) of generator torque rate

9
ResultsSimulation in HAWC2
10
ResultsSimulation in HAWC2
11
Concluding Remarks

• A full wind speed range model-based controller
  has been presented with cyclic pitch available if
  needed/desired.
• Dynamic inflow has an importance for model-based
  controllers, especially around rated wind speed.
• For advanced model-based control setups where
  pitch actuation is active during partial load
  operation dynamic inflow should be included in
  the control design model. This could especially
  be relevant for preview-based (e.g. Lidars)
  control strategies attempting to mitigate loads
  via pitch actuation during partial load operation.

12
Thank you for your attention!