ECN publicatie:
Titel:
Load monitoring for wind turbines. Fibre optic sensing and data processing.
 
Auteur(s):
 
Gepubliceerd door: Publicatie datum:
ECN Windenergie 6-1-2010
 
ECN publicatienummer: Publicatie type:
ECN-E--09-071 ECN rapport
 
Aantal pagina's: Volledige tekst:
41 Download PDF  (3175kB)

Samenvatting:
Condition monitoring of wind turbine components is of growing importance. For bearings, gearboxes and other rotating equipment, techniques are available from other applications. For condition monitoring of rotor blades, no suitable techniques are available up to now. However the interest for it is increasing in order to get more insight in the occurring loads and to reduce these loads by applying other control techniques. The developments with respect to application of more advanced control techniques and condition monitoring are hindered by the fact that no reliable and accurate measurement system for measuring the blade loads is available.

For the load measurements, optical techniques are often mentioned as the most promising because of the long life time of the fibers, the insensitivity for electrical interference and stability of the performance, which makes regular re-calibration superfluous. Optical measurement systems are in use for several years now, but the results are not always convincing and require further developments. The systems which have been used initially by ECN showed a lot of failures, so that measurements have been often interrupted. The installation of the sensors appeared to be very time consuming and had to be done by special skilled employees. The connection of the sensors itself also failed sometimes and the repair is laborious, because special skills and bonding techniques are required.

Another general disadvantage of condition monitoring systems is, that a lot of data is generated, which requires a lot of effort to analyze. Special knowledge is required for interpretation of the results. This implies additional work load and costs.

For the current project, the following objectives have been chosen in order to improve the applicability of load measurements and monitoring of the blades:
• Selection of an interrogator which meets the requirements for load monitoring.
• Selection or development of a sensor which is suitable for this application, including handling and installation.
• Development of a processing tool, which operates automatically and generates only the required information.

For the selection of interrogators, an inventory followed by a first selection was made in a desk study. This resulted in three candidates, which matched with the user requirements to a large extend. These interrogators have been tested at ECN and resulted in the choice of a device which meet the requirements to a large extend and which is also considered as suitable for further experiments and development tests for sensors. The accuracy and resolution are sufficient for the application, but also the selected interrogator showed strong sensitivity for polarization. The price level is still above target.

For the sensors, also a desk study has been performed. This showed that no sensor was compliant with the user requirements. The strain was often not measured directly, while installation and/or replacement were difficult. However, the desk study also resulted in a new idea with respect to a new approach. A sensor assembly was designed, built and tested. This sensor meets the following requirements:
• The sensor can be easily installed in the rotor blades. The installation of a sensor assembly takes 15 to 30 minutes, which implies that a turbine can be equipped with a load measurement system within a day. This minimizes the down time of the turbine.
• The sensor is designed for easy maintenance. A failure can easily be detected and the sensor can be replaced within five minutes. Calibration of the sensor can be done during production. On site calibration is not necessary, which implies “Plug-and Play”.
• Temperature compensation is included with respect to sensor properties.
• Installation and maintenance can be done by technicians that are trained for regular wind turbine maintenance. No special skills are required.
• The sensor is also compliant with respect to resolution and accuracy and does not show drift.

A request for patent is pending on this moment. Several parties have shown their interest for this sensor.

Load measurement systems produce large amount of data, which is difficult to handle by the operators. The project resulted in a concept for automatic data processing and providing key figures to the operator to make decisions for maintenance. This improves the cost effectiveness of the maintenance. The concept is described in a specification and the software is in development.
The software has the following features:
• The measurement system converts the strain measurements into bending moments. A well defined interface between measurement system and data processing enables the use of the software in combination with other measurement systems.
• The software automatically generates load spectra and equivalent loads in accordance with the relevant IEC-61400 standards and taking into account the operational conditions.
• The measurements are validated automatically.
• The measurements can be processed directly, which implies that storage of large amounts of measurement data is not necessary.
• The software allows comparison of loads between the individual turbines in order to prioritize maintenance and inspection schemes.
• Although an interface between the wind turbine and the measurement system is required with respect to the PLC-signals, the data processing is suitable for all types of turbines.
• The data processing also includes automatic warning and detailed event analysis.

The software development did not yet result in version ready for application. The requirements are available for a wide extend. The development is hindered by the complexity of large variations in loading in combination with unreliability of measurement data. Although the measurement system should generate reliable data, faults can have strong effects on analysis results. This implies that a lot of effort has been laid in this part of the data processing.


Terug naar overzicht.