Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'Advanced control concepts' taking place on Wednesday, 12 March 2014 at 11:15-12:45. The meet-the-authors will take place in the poster area.

Nikola Hure Faculty of Electrical Engineering and Computing, Croatia
Nikola Hure (1) F P Mario Vašak (1) Mate Jelavić (2) Nedjeljko Perić (1)
(1) Faculty of Electrical Engineering and Computing, Zagreb, Croatia (2) KONČAR - ELECTRICAL ENGINEERING INSTITUTE Inc., Zagreb, Croatia

Printer friendly version: printer.gif Print


Wind turbine overspeed protection based on robust invariant sets


Reduction of loads experienced by a wind turbine together with a production criterion are the essential objectives of any successful wind turbine control concept. Although the fatigue is accumulated through a long period of a wind turbine operation, the most critical structural loads are often experienced in extreme situations, such as overspeed. Respecting the designed limits, in literature often referred as operating envelope protection, can be systematically achieved using the invariant set framework. Alongside the wind turbine speed limitations, any other imposed constraint can be incorporated in the invariant set-based protection synthesis to prevent structural damage of the wind turbine.


In this work, overspeed protection is achieved by integrating robust controlled invariant set (RCIS) [1] in a wind turbine control system. RCIS is contained in a bounded space of system state variables x and control inputs u. If the state variables and inputs are within such a set, there is a firm guarantee that the state will always remain in the set projection onto the state space, while respecting the given input constraints (e.g. maximal pitch rate) and taking into account uncertainty of the model (e.g. due to uncertain wind).

RCIS is computed offline in an iterative fashion. The computation algorithm takes as inputs: (i) discrete-time state-space model of a plant, (ii) imposed control system constraints, (iii) model uncertainty and states estimation error in a form of an additive bounded disturbance.

Once computed, the RCIS can be integrated in the existing control system like a supervising controller, irrespective of the control scheme used. The structural scheme of a wind turbine control system utilizing a robust invariant set-based protection is conceptually depicted in Fig. 1.

The protection does not utilize the information about the upcoming rotor-effective wind speed. It corrects the controller outputs M'_(g,ref) and β'_(ref) to maintain system state variables and inputs in the RCIS using available measurements and system state estimate at the moment. Online implementation boils down to a search for the closest point in the RCIS for the given state estimate x ̂ (Fig. 2). The protection system operates in a discrete time with the selected time period.

Apparently, the overspeed protection cannot be used to overcome any wind gust. The wind gust evolution model considered for the RCIS computation is a step response of the first order system. The amplitude of the step response is a modifiable parameter of the protection system. The higher is the amplitude of the presumed wind gust, the higher is the activity of the pitch actuation system and the generator torque fluctuations. The appropriate wind amplitude parameter selection is a compromise between the protection conservativeness and the blade bearing fatigue [2].

Main body of abstract

There are two main negative side-effects of the overspeed occurrence: production loss and excessive loads experienced by the wind turbine. Possible overspeed detection initiates the emergency shutdown procedure. Blades are pitched to full feather and the generator torque is kept at the nominal value to stop the wind turbine.

Loads experienced by the wind turbine during the shutdown depend mostly on the dimensioning of the pitch system. Both too high and too low pitch velocities will increase the bending moments in the tower and foundations [3]. Moreover, high pitch velocities increase the cost of the pitching system design and hardware. Inert pitch will result with a cost-effective dimensioning of the pitch system, but will make the wind turbine vulnerable to wind gusts.

The wind turbine considered in this work is a 1 [MW] turbine with a direct drive. Its trigger for the emergency shutdown is set to 31.3 [rpm] while rotor speed should not exceed 33 [rpm] at any circumstance since this could lead to excessive structural loading. The task of the overspeed protection system developed in this work is to reduce incidence of the emergency shutdown what will result with an increased production and reduced fatigue of the wind turbine.

Of very large importance for the protection system is an estimate of the rotor-effective aerodynamic torque Ma that acts on the turbine. Precise knowledge of Ma is a prerequisite for the accurate prediction of the wind turbine speed trajectory given the assumed wind gust model. Therefore, it greatly determines the performance of the protection system.

Aerodynamic torque estimate could be reached from wind turbine speed derivative, blade root loading and drivetrain torsion [5]. Depending on the approach, various wind turbine models can be used for the RCIS computation.

The system control constraints are given in Table 1. Forcing the generator torque above the rated value is not allowed.

It is crucial that overspeed protection system respects given constraints and fully exploits them in moments of threat. Advanced predictive control methods often suffer from the negative side-effects of the discretisation interval. Because of their step-like pitch angle reference and considerable discretisation interval, it is impossible to achieve a ramp-like blade pitch response with the maximal pitch rate. That could significantly reduce the size of the invariant set used for the overspeed protection synthesis. Therefore is for the RCIS computation purpose used a mathematical model of the turbine with a pitch rate reference control input and pitch position control is entrusted to a separate subsystem (Fig. 3).

The comparative results of the wind turbine control system response with/without the protection patch are shown in Figs. 5-6. The rotor effective wind speed used for the simulation is given in Fig. 4 -- it is composed of a turbulent wind superimposed by step-like wind gusts. The wind turbine baseline controller is based on a gain-scheduling PI above the nominal wind speed and a quadratic optimal torque control with the PI controllers at the borders of the operating region below the nominal wind speed [4]. The simulation results are obtained in the aero-elastic code GH Bladed.

Note: emergency shutdown procedure is turned off for the simulation.

In the given simulation scenario the overspeed protection system reduces the number of the maximal turbine speed violations from 3 to only 1. In Table 2 quantitative measures of the overshoot average are given along with the the pitch bearing fatigue extracted from the simulation results.

It should be noted that if the emergency shutdown procedure was included in the simulation, the pitch bearing fatigue would be by far greater for the wind turbine control configuration without the overspeed protection.


In this work the overspeed protection system based on robust invariant sets is developed. The selected approach is independent of the applied wind turbine control system and can be easily implemented for the real-time control.

The presented overspeed protection is model-based and operates without the measurements of the upcoming wind. It takes into consideration model uncertainties as well as the system state estimation error. All of the system control constraints are included during the overspeed protection synthesis.

The success of a RCIS-based overspeed protection system is conditioned with a mathematical model of the wind turbine as well as the quality of the system state estimation. Good mathematical model throughout the whole operating region will result with less stress on the control system and accurate system state estimation is of vital importance for the timely reaction of the protection system.

The presented results of the wind turbine operation with the overspeed protection system show it can greatly reduce the number of overspeed occurrences with insignificant increase of the pitch bearing fatigue. Power loss and structural loads arising from the emergency shutdown procedure due to overspeed are mitigated. Moreover, the protection system does not interfere with the existing wind turbine controller during normal operation and harmless wind gusts.

Finally, the presented overspeed protection approach based on invariant sets can offer an answer to the following question: “How far can we go with the overspeed protection without the knowledge about the upcoming wind?”. The answer depends greatly on the available measurement signals and blade pitch system design, but for a concrete set-up it can be firmly given.

Learning objectives
A model-based ultimate method for operating envelope protection of wind turbines.

Implementation of the overspeed protection system using robust invariant set-based framework and simulation results.

[1] F. Blanchini, S. Miani. “Set-Theoretic Methods in Control”, Birkhäuser Boston, 2008.
[2] T. Burton, D. Sharpe, N. Jenkins, and E. Bossanyi. “Wind Energy Handbook”, Wiley, 2001.
[3] L. Frøyd, O. G. Dahlahug. “Effect of pitch and safety system design on dimensioning loads for offshore wind turbines during grid fault”, Energy Procedia, 2012.
[4] M. Jelavić, ”Wind turbine control for structural dynamic loads reduction”, Ph.D. dissertation, Faculty of Electrical Engineering and Computing, University of Zagreb, 2009.
[5] J. Suarez, E. Azagra, Y. Urroz, O. H. Mascarell. “Application of wind speed estimation for power production increase”, EWEA 2012.
[6] K. Hammerum. “A Fatigue Approach to Wind Turbine Control”, Master’s thesis,Technical University of Denmark, 2006.