Anomaly Detection of Wind Turbine Time Series using Variational Recurrent Autoencoders

Alan Preciado-Grijalva,Victor Rodrigo Iza-Teran

DOI: https://doi.org/10.48550/arXiv.2112.02468

2021-12-05

Abstract:Ice accumulation in the blades of wind turbines can cause them to describe anomalous rotations or no rotations at all, thus affecting the generation of electricity and power output. In this work, we investigate the problem of ice accumulation in wind turbines by framing it as anomaly detection of multi-variate time series. Our approach focuses on two main parts: first, learning low-dimensional representations of time series using a Variational Recurrent Autoencoder (VRAE), and second, using unsupervised clustering algorithms to classify the learned representations as normal (no ice accumulated) or abnormal (ice accumulated). We have evaluated our approach on a custom wind turbine time series dataset, for the two-classes problem (one normal versus one abnormal class), we obtained a classification accuracy of up to 96$\%$ on test data. For the multiple-class problem (one normal versus multiple abnormal classes), we present a qualitative analysis of the low-dimensional learned latent space, providing insights into the capacities of our approach to tackle such problem. The code to reproduce this work can be found here <a class="link-external link-https" href="https://github.com/agrija9/Wind-Turbines-VRAE-Paper" rel="external noopener nofollow">this https URL</a>.

Machine Learning

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the abnormal rotation or non - rotation phenomenon on wind turbine blades caused by ice accumulation, which will affect power generation and power output. The author studies the ice accumulation problem on wind turbine blades by formulating this problem as anomaly detection in multivariate time series. Specifically, the paper proposes a method, which is mainly divided into two parts:

1. **Learning low - dimensional representations**: Use Variational Recurrent Autoencoder (VRAE) to learn low - dimensional representations from time - series data. VRAE can capture non - linear relationships in time series and compress high - dimensional time series into a low - dimensional latent space.

2. **Classification and clustering**: Use unsupervised clustering algorithms to classify the learned low - dimensional representations and mark them as normal (no ice accumulation) or abnormal (with ice accumulation). Three clustering algorithms, KMeans++, Hierarchical Clustering and Density - Based Spatial Clustering of Applications with Noise (DBSCAN), are used in the paper.

Through the above methods, the paper is evaluated on a self - defined wind turbine time - series data set. For the binary classification problem (one normal category versus one abnormal category), the classification accuracy of the test data reaches 96%. For the multi - classification problem (one normal category versus multiple abnormal categories), the paper provides a qualitative analysis of the low - dimensional latent space, demonstrating the ability of its method to handle such problems.