Suppression of wind turbine noise from seismological data using nonlinear thresholding and denoising autoencoder
Open Access
- 2 July 2022
- journal article
- research article
- Published by Springer Nature in Journal of Seismology
- Vol. 26 (5) , 913-934
Abstract
Seismologists found a significant deterioration in station quality after installation of wind turbines (WTs), which led to conflicts between WT operators and seismic services. We compare different techniques to reduce the disturbing signals from WTs at seismological stations by selection of an affected station. WT noise and earthquake signals have overlapping frequency bands, and thus spectral filtering cannot be used. As a first method, we apply the continuous wavelet transform on our data to obtain a time-scale representation. From this representation, we estimate a noise threshold function either from noise before the theoretical P-arrival or using a noise signal from the past with similar ground velocity conditions at the surrounding WTs. As a second method, we use a denoising autoencoder (DAE) that learns mapping functions to distinguish between noise and signal. In our tests, the threshold function performs well when the event is visible in the raw or spectral filtered data, but it fails when WT noise dominates. The use of the threshold function and pre-noise can be applied immediately to real-time data and has low computational cost. Using a noise model from our prerecorded database at the seismological station does not improve the result and is more time consuming. In contrast, the DAE is able to remove WT noise even when the event is completely covered by noise. However, the DAE must be trained with typical noise samples and high signal-to-noise ratio events to distinguish between signal and interfering noise.Keywords
Funding Information
- Europäische Fonds für regionale Entwicklung
- Ruhr-Universität Bochum
This publication has 22 references indexed in Scilit:
- Matrix Profile I: All Pairs Similarity Joins for Time Series: A Unifying View That Includes Motifs, Discords and Shapelets2017
- Matrix Profile II: Exploiting a Novel Algorithm and GPUs to Break the One Hundred Million Barrier for Time Series Motifs and Joins2016
- Influence of Wind Turbines on Seismic Records of the Gräfenberg ArraySeismological Research Letters, 2016
- Automatic microseismic denoising and onset detection using the synchrosqueezed continuous wavelet transformGeophysics, 2016
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- Seismic Noise by Wind Farms: A Case Study from the Virgo Gravitational Wave Observatory, ItalyBulletin of the Seismological Society of America, 2011
- ObsPy: A Python Toolbox for SeismologySeismological Research Letters, 2010
- De-noising by soft-thresholdingIEEE Transactions on Information Theory, 1995
- WAVELET TRANSFORMS AND THEIR APPLICATIONS TO TURBULENCEAnnual Review of Fluid Mechanics, 1992
- Filtering noise from images with wavelet transformsMagnetic Resonance in Medicine, 1991