Large scale test of vibration based icing detection for wind turbines

• Published in: Journal of physics. Conference series Vol. 2647; no. 19; pp. 192008 - 192017
• Main Authors: Weijtjens, Wout, Junior, Adelmo Fernandes de Oliveira, Cloet, Bram, Yilmaz, Özlem Ceyhan, Devriendt, Christof
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.06.2024
• Subjects: Accelerometers; Chambers; Ice formation; Resonance; Resonant frequencies; Strategy; Temperature dependence; Turbine blades; Turbines; Uniqueness; Vibration; Vibration monitoring; Wind turbines
• ISSN: 1742-6588; 1742-6596
• DOI: 10.1088/1742-6596/2647/19/192008

Abstract The build-up of ice on the blades is a known issue for wind turbine operators. In particularly in northernly countries continued operation under icing can lead to significant performance losses and increased wear of key components. Aside from the operational concerns the ice also poses a potential safety issue, as both falling ice or ice-throw can result in damages to nearby property or worse personal injury. In particular in densely populated areas, this means that turbines need to be shut down as soon as a significant risk of icing occurs. To meet these operational concerns several icing solutions have found industry wide adoption. These solutions can be divided into several categories based on the working principles to identify and quantify the icing risk. The most common icing detection solutions rely on the detection of the environmental conditions that can potentially result in icing (e.g. by detecting ice growing on a nacelle-mounted sensor). These so-called indirect methods are fairly cheap to install, yet are generally considered less sensitive to actual icing on the blades, resulting in both false negatives and positives. Direct icing methods try to identify ice as it builds up on the blade directly and are generally considered to be more reliable. They either rely on sensors glued to the surface of the blade or through monitoring the vibrations of the blade and computing the natural frequencies of the turbine’s rotor. The latter group of vibration-based techniques is the subject of this contribution. These techniques detect icing through the assumption that icing will act as an additional mass to the blade. This additional mass will then lead to a decrease of the resonance frequencies of the blade(s). When this decrease is detected, icing is inferred and an alarm is given. Note that this also requires a model to compensate for the temperature (and load) dependency of the resonance frequencies. While this strategy has found industry adoption little controlled experiments are found. This is easily explained as the method relies on the structural dynamics of the whole turbine and it is therefore difficult to perform controlled experiments. To better understand the potential and limitations of vibration based icing detection strategy this contribution presents the results of a unique experiment. A 12m wind turbine blade was instrumented with precision accelerometers and loaded into OWI-lab’s Large climate chamber. The chamber was cooled down to -10°C and icing was applied in a controlled manner using a setup designed to simulate the natural formation of ice on a wind turbine. Data was then processed using automated OMA to monitor the evolution of the natural frequencies over time and a normalization model was used to compensate for the natural variability of the resonance frequencies to temperature. The controlled nature of the experiment allows for a more in-depth quantification on the sensitivity to icing and temperature and moreover it offers a unique dataset to validate future concepts of vibration based icing detection.