To monitor wind turbine vibrations, normal behaviour models are built to predict tower\ntop accelerations and drive-train vibrations. Signal deviations from model prediction are labelled as\nanomalies and are further investigated. In this paper we assess a stochastic approach to reconstruct\nthe 1 Hz tower top acceleration signal, which was measured in a wind turbine located at the wind\nfarm Alpha Ventus in the German North Sea. We compare the resulting data reconstruction with\nthat of a model based on a neural network, which has been previously reported as a data-mining\nalgorithm suitable for reconstructing this signal. Our results present evidence that the stochastic\napproach outperforms the neural network in the high frequency domain (1 Hz). Although neural\nnetwork retrieves accurate step-forward predictions, with low mean square errors, the stochastic\napproach predictions better preserve the statistics and the frequency components of the original signal,\nretaining high accuracy levels. The implementation of our stochastic approach is available as open\nsource code and can easily be adapted for other situations involving stochastic data reconstruction.\nBased on our findings we argue that such an approach could be implemented in signal reconstruction\nfor monitoring purposes or for abnormal behaviour detection.
Loading....