You are here:   Home > RT-MBDyn > Wind Turbine

Wind Turbine

RT-MBDyn contributes to renewable energy

Authors: L. Cavagna, L. Dozio, P. Mantegazza, P. Masarati, M. Morandini
Affiliation: Dipartimento di Ingegneria Aerospaziale - Politecnico di Milano
Date: June 19, 2007

TARGET AND MOTIVATIONS

The goal of this work is to demonstrate the rapid feasibility of the implementation of innovative, efficient and low-cost solutions for Fast-Prototyping and customization of controlled mechanical systems.

A controlled constant-speed wind-turbine [9] is considered as one of the possible applications of the developed methodologies; the free RTOS RTAI [1] and the free multibody software MBDyn [2], both originating from research at the Dipartimento di Ingegneria Aerospaziale of Politecnico di Milano - Italy, are at the core of the present work.

Fig. 1 shows a sketch of the controlled model under analysis [4].

 


Fig. 1: sketch of the controlled model.

WIND-TURBINE DESCRIPTION

The CART (Controls Advanced Research Turbine) wind-turbine (Fig. 2) is located at the National Wind Technology Center (NWTC) in Colorado as state of the art test-bench for controls research in wind-engineering [511]. It is an upwind machine with a nacelle tilt of 3.8 deg and two teetering blades with zero precone.

The rotor diameter and the hub height are respectively 43.3 m and 36.6 m. Power energy is rated at 600 kW and generator speed through a gearbox with a ratio of 43.165 is rated at 1800 rpm (the rotor is thus rated at Ωrated = 41.7 rpm).


Fig. 2: the Controls Advanced Research Turbine (CART)

BASELINE CONTROLLER

The baseline controller is composed by independent electric torque and collective pitch algorithms. Both controllers use the rotor angular speed measurement as sole input.

The task of the control-system is to maximize power capture below, and regulate a constant speed above, the rated operating point. At the moment no effort is undertaken to regulate nacelle yaw (in the real wind turbine it is limited to only 0.5 deg and is simply used for tracking relatively small wind-changes) and the high speed generator shaft brake.

Generator commands are calculated by means of a piece-wise function. Below cut-in speed of 10 rpm (Region 1) no electric torque is generated to let the wind accelerate the rotor, i.e. provide maximum angular acceleration. The quadratic region (Region 2) is designed to keep the tip-speed ratio at the optimal value for maximum power. Above 99% of the rated rotor speed Ωrated a constant torque of 3524 Nm is required. Between 98% and 99% of Ωrated the transition is linear, equivalent to a slip of 5% (Region 3) [678].

Fig. 3 shows the piece-wise working function for the electric generator and its block-diagram model.

 



Fig. 3: piece-wise working function for the electric generator and block-diagram model.

The full-span collective blade pitch angle commands are computed by means of a PID controller on the error of the rated angular speed with saturation on the integral term to limit wind-up. Special care is taken to avoid working in post-stall region during the initial acceleration phase. This feature enables to use the same controller both for rotor acceleration starting from null angular speed and constant-speed maintenance at the rated speed.

Fig. 4 shows the final controlled system in the typical Simulink/Scicos/Matrixx environment. The blue box represents the MBDyn model output to the control system: only the angular velocity of the rotor is required for the basic control system developed. The red box represents the inputs to MBDyn which are: free-stream wind velocity, electric resistant torque, and controlled pitch command.

The controlled CART model can be run both in batch mode by means of MBDyn-Simulink interfaces and in real time using controller code automatically generated through Simulink/Scicos/Matrixx from the very same model, and run on the RTAI operating system, with MBDyn executing in real time mode to emulate the real wind-turbine. In the real time mode, the support of RTAILab can be used to monitor the controller, either locally or remotely, with the possibility of tuning gains and other system parameters, as will be shown later on.

 


Fig. 4: controlled system in the typical Simulink/Scicos/Matrixx environment.

MULTIBODY MODEL

The multibody model consists in:

  • a deformable tower, made of 5 three-node finite-volume beam elements [10], rigidly clamped to the ground
  • a rigid nacelle, connected to the tower by a very stiff spring, since no yaw control is implemented
  • a rigid low-speed shaft, connected to the nacelle by ideal bearings; the rotational inertia includes that of the high-speed shaft, accounting for the low- to high-speed shaft gear ratio
  • an ideal generator, consisting in a torque applied to the low-speed shaft, whose value is computed by the control task
  • a rigid teetering body, connected to the low-speed shaft by an ideal teeter joint
  • a pair of deformable blades, each made of 5 three-node finite-volume beam elements [10], including blade element aerodynamics augmented by an induced flow model.

The blade pitch is controlled by simultaneously rotating the blades at the root node by an amount that is determined by the controller task. An angular velocity sensor measures the low-speed shaft velocity and feeds it into the control task.

 

 
 
Fig. 5: images of the CART model.

The wind-turbine pictures in Fig. 5 are generated by an enhanced version of the free visualization software EasyAnim [3].

The structural model of the tower and of the blades is even too refined, considering the very low rotational velocity and the bandwidth of interest. This leaves room for modeling wind turbines with more blades.


Fig. 6: sketch of a generic distributed real-time simulation layout.
 

REAL-TIME SIMULATION

Fig. 6 shows a fairly broad layout of the real-time simulation setup, where the simulation and the controller are located on different computers connected by a hard real-time network via NetRPC, while multiple observers monitor the output of the controller and of the simulation, and optionally modify the parameters of the controller, using soft real-time connections.
Fig. 7 shows the result in terms of rotor angular speed (black line) and pitch command (red line) of a simulation in correspondence of a growing wind speed rated at an average level of 12 m/s, with a random disturbance of 20% of the wind velocity magnitude. The resulting error on the final rated angular speed is less than 0.5%. The sample rate is 100Hz.

Fig. 7: rotor angular speed and pitch command for random gust.
Fig. 8 shows the result in terms of rotor angular speed (black line) and pitch command (red line) of a simulation in correspondence of a growing wind speed rated at 12 m/s with a sinusoidal disturbance of 5% of the wind velocity magnitude (blue line). The resulting error on the final rated angular speed is less than 0.5%. The sample rate is 100Hz.

Figs. 9a-f show the internal forces and moments in a blade during the same transient.


Fig. 8: rotor angular speed and pitch command for harmonic gust.

Fig. 9a: blade axial force, N

Fig. 9b: blade torsional moment, Nm

Fig. 9c: blade in-plane shear force, N

Fig. 9d: blade out-of-plane bending moment, Nm

Fig. 9e: blade out-of-plane shear force, N

Fig. 9f: blade in-plane bending moment, Nm

Fig. 10 shows the output of the controlled CART model within the RTAILab environment.

 


Fig. 10: output of the controlled CART model within the RTAILab environment.

What has been implemented so far can be essentially considered a test bench to prove the feasibility of the implementation of innovative, efficient and low-cost solutions for Fast-Prototyping and customization of controlled mechanical systems. Further details can be introduced to enhance system modeling with features that have not been considered so far as the primary target of this work. Further efforts are planned for the near future to reach this last target.

ACKNOWLEDGMENTS

The present work has been partially funded by the SI PARTE! project (in Italian), which focuses on the development of innovative solutions for Embedded Real Time Applications.
 

Partners of the project are:

REFERENCES

[1] Real-Time Application Interface (RTAI): http://www.rtai.org/
[2] MultiBody Dynamics (MBDyn): http://www.mbdyn.org/
[3] EasyAnim: http://mecara.fpms.ac.be/EasyDyn/
[4] J. Jonkman, NREL/NWTC report February 15 2007-06-08
[5] L. J. Fingersh, K. Johnson, Controls Advanced Research Turbine (CART) Commissioning and Baseline Data Collection, National Renewable Energy Laboratory, NREL/TP-500-32879 October 2002
[6] K. A. Stol, L. J. Fingersh, Wind Turbine Field Testing of State-Space Control Designs, National Renewable Energy Laboratory, NREL/SR-500-35061 August 25, 2003 November 30, 2003
[7] T. P. Fuglseth, Modelling a 2.5MW direct driven wind turbine with permanent magnet generator, Department of Electrical Power Engineering, Norwegian University of Science and Technology
[8] M. H. Hansen, A. Hansen, T. B. Larsen, S. Oye, P. Sorensen, P. Fuglsang, Control Design for a pitch regulated, variable speed wind turbine, Riso National Laboratory, Denmark, January 2005
[9] R. Garsch, J. Twele, Wind Power Plants, Fundamentals, Design, Construction and Operation, James and James, 2002
[10] G.L. Ghiringhelli, P. Masarati, P. Mantegazza, A Multi-Body Implementation of Finite Volumes C0 Beams, AIAA Journal, Vol.38(1) January 2000, pp. 131-138
[11] K.A. Stol, Geometry and Structural Properties for the Controls Advanced Research Turbine (CART) from Model Tuning, National Renewable Energy Laboratory, NREL/SR-500-32087, August 25, 2003 November 30, 2003

Powered by CMSimple|