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 operation & maintenance' taking place on Thursday, 13 March 2014 at 11:15-12:45. The meet-the-authors will take place in the poster area.

Michael Beyer Windrad Engineering GmbH, Germany
M Beyer (1) F P U Ritschel (1) C Mehler (2) M Joost (2) B Orlik (2)
(1) Windrad Engineering GmbH, Bad Doberan, Germany (2) IALB U Bremen, Bremen, Germany

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Dr. Michael Beyer is Managing Director of Windrad Engineering GmbH, an engineering consultant for wind industry for more than 10 years. Previously he was Head of Simulation and Loads in the same company. He received his PhD at the University of Mainz and was a post-doc at several research institutions and Universities with a strong focus on numerical simulation and phenomenology. In 1997 he was nominated Privatdozent at the University of Rostock.


Observer structures in advanced power electronics for load analysis and power control


Wind Energy Converters (WEC) belong to technical structures with highest number of load cycles and, also typical, frequent occurrence of high peak loading. As a consequence, e.g., the gear box suffers from high load and possibe damage [1]. Failure leads to unwanted costs and downtimes [2]. Monitoring of loads during operation that could give hints on the remaining fatigue life is not part of a standard operational or monitoring system. We have developed an observer structure that allows monitoring of torsional loads by using standard sensors available in the control system of the wind turbine and/or power electronics.


The approach is based on an observer that is a reduced model of the real system. The focus is on the drive train of the WEC. The observer is a software development that can be embedded in the control system of the turbine. It is a multi-body model of the drive train solved in the rotational degree of freedom. The number of bodies depends on the desired degree of resolution. E.g. modelling of a gear box requires more information than modelling of a direct drive. In order to have the necessary resolution this particular observer is embedded in the controller of the power electronics control that is operating in the kHz regime. The software is fast enough to run in real time parallel to the real drive train of the turbine. It measures the torque on the drive train (or parts of it) depending on the specific wind conditions of the turbine site. Drifting of the observer is prevented by feeding proper state signals of the real turbine such as, e.g., measured generator speed to the observer. The torque can be analysed by standard rain flow counting so that site specific fatigue damage can be accumulated. This fatigue analysis can be compared to the design loads. Another possibility is to use the observed torque as an active drive train damping. The observed torque is used control variable, i.e., the torque demand is corrected by the measured torque signal. The model has been demonstrated on the test stand of the IALB at the University of Bremen and in a standard WEC simulation environment at the Windrad Engineering GmbH.

Main body of abstract

In principle an observer is a model of the real system, in this case the drive train of a WEC, which is calculated in real time parallel to the drive train. This model is subject to the same input as the real system, so if the model is accurate, the initial states are the same and if no perturbation would be present, all the intrinsic states of the drive train can be calculated via the model. Unfortunately this is not the case, as unknown, uncontrolled influences exist. Also the initial conditions are not known to the required degree.This leads to an estimation error of the model. In order to remedy this, estimated values of those states that can be measured are subtracted from measured values and this estimation error is fed back via a correctional vector into the model. By selecting a suitable correctional vector the estimation error is minimized. This is shown in the following figure.
We investigate a one body system several two body systems and a three body system.

In the machine laboratory at IALB of the University of Bremen a test stand was set up in order to test new control concepts for wind energy converters. The test stand emulates the behavior of a WEC. An induction motor driven by a frequency converter provides the torque and simulates the wind driven rotor blades. The mechanical to electrical energy transformation is done by a doubly fed induction generator. Additionally the drive train comprises a single-stage spur gear box. The following figure shows the test stand and gives an overview over its components.

As a part of this project an energy recovery capable frequency converter was developed at IALB and set up at the test stand. The used power semiconductor modules are controlled by a Texas Instruments digital signal processor TMS320F28335. This DSP features free programmability and is an important prerequisite for the use of the frequency converter in implementing and testing new control concepts and even the proposed observer structures at the test stand. In order to evaluate observer and control concepts the test stand is fitted with additional sensory equipment, so that dynamical loads can be measured under operating conditions. For example both sides of the gearbox are fitted with a torque sensing flange.

After successful simulations were made, the observer for a one-mass-system was implemented at the test stand in order to proof the functionality of the structure in a typical application. The estimated values are accessible through analog / digital converters controlled by the DSP and can be directly compared to the measured values provided by the torque sensing flange. A very good agreement between the measured and the estimated torque at the generator side of the gear has been fund. So the observer shows reasonable results with minimal implementation effort. The estimated torque is of such a quality, that it can be used for further load calculations concerning the drive train

The aeroelastic simulation is done with a code that simulates the aerodynamic forces on the rotor and the remaining parts of the turbine, and from that derives the response of the wind turbine structure in a self-consistent way. To this end a typical 2 MW turbine has been implemented. The solution is based on modal analysis. Standard Kaimal turbulent wind fields common for certification have been utilized. Also, all turbulence classes and all standard wind speed classes have been investigated.

One of our results is shown in the above figure. Here the specific contribution of different wind bins to the damage equivalent load of the drive train is shown. From comparison it is seen that the modified observer is able to predict the damage equivalent loads for the torsional loads. Even for the relative contributions given for different wind bins a reasonable agreement can be achieved. Furtheron, it would be possible to derive quantities such as life time consumption and in addition through on-line evaluation open the possibly to determine early detection of damage.

Another example of application is the use as active damping for the drive train. This has been realized for the one-body observer.


We have developed a new software that allows a determination of torsional loading of the wind turbine with normally on-board available sensor equipment. The notion of this loading is helpful to prevent unwanted down times of the turbine. This is in particular interesting for off-shore technology, where demands on reliability are much higher than on-shore. But also on-shore unexpected down times of a turbine is an unwanted scenario. The basis of this software is an observer that has to be implemented at best in the power electronics control system. In this case the resolution is given by the sample rate in the power electronics control that is high enough to resolve fast processes in the gear box. In this approach electrical quantities such as voltage, current and speed of the generator of the wind turbine are measured and analyzed. The multi-body part of the observer has to be adapted to the particular turbine drive train. The observed torque can be used for load analysis but also for turbine control. The observer has been implemented in a test stand of a small scale wind turbine at the IALB at the University of Bremen and proven successful. In order to study the observer in a turbulent condition of a wind field it has also been implemented in an industrial aeroelastic simulation environment usually used for turbine design. After few adjustments the observer has proven succesfull also in this environment. A further application is the use of the observer as a device for active damping that is in many cases realized by filters or combinations thereof.

Learning objectives
The notion of loading in the drive train during operation may be helpful to determine unexpected wear and tear. Such a measurement campaign is usually not part of a CMS of a WEC. The observer developed in this collaboration is capable to determine torsional loads in the drive train. These loads can be utilized 1) to recognize unusual fatigue loading and/or 2) to reduce drive train oscillations through active damping.

[1] S. Faustich, B. Hahn - “Schadensdaten-banken - Fehlerhäufigkeitsanalyse und Prognose von technischen Problemen“, BWE-Fachtagung „Service, Wartung und Betrieb“, Hamburg, 2009.
[2] “Windenergiereport Deutschland 2008”, ISET, Kassel, 2009