Multiscale Wavelet-Driven Graph Convolutional Network for Blade Icing Detection of Wind Turbines

Abstract

Blade icing detection is critical to maintaining the health of wind turbines, especially in cold climates. Rapid and accurate icing detection allows proper control of wind turbines, including shutting down and clearing the ice, thus ensuring turbine safety. This paper presents a wavelet-driven multiscale graph convolutional network (MWGCN), which is a supervised deep learning model for blade icing detection. The proposed model first uses wavelet decomposition to capture multivariate information in the time and frequency domains, then employs a temporal graph convolutional network to model the intervariable correlations of the decomposed multiscale wavelets, as well as their temporal dynamics. In addition, this paper introduces scale attention to the MWGCN for a further improvement of the model and proposes the method to address the class imbalance problem of the training data sets. Finally, the paper conducts comprehensive experiments to evaluate the proposed model, and the results demonstrate the effectiveness of the model in blade icing detection and its better performance over eight state-of-the-art algorithms, with 17.2% and 11.3% higher F1 scores over the best state-of-the-art baseline on the labeled datasets.

A Review of Wind Turbine Icing and Anti/De-Icing Technologies

“…Traditional methods of blade icing detection rely heavily on dedicated sensors, which require additional installation costs. According to various studies, it is not necessary to directly measure the thickness of icing, and the icing thickness can be analyzed through the relevant data already obtained by the wind turbine, which belongs to indirect measurement [81,82]. Gantasala et al [83] proposed that the mass and natural frequency of blades will change due to a certain amount of icing, and analyzed the relationship between the natural frequency and the icing mass on the blade, so icing can be detected indirectly through vibration-based methods.…”

“…Recently, some artificial intelligence methods have also been introduced to measure wind turbine icing [82,83,[89][90][91][92][93][94][95]. Yi et al [96] used a large number of data and relevant information of time series, combined the original data, features extracted by a Stacked Auto Encoder (SAE) and a residual vector to obtain discriminant features, tested the operating state of wind turbines, and proposed a fault detection scheme based on discriminant feature learning.…”

LightCTS: A Lightweight Framework for Correlated Time Series Forecasting

“…One significant application occurs in the predictive maintenance of wind turbines [11], which are often deployed in remote and harsh locations. Accurate and instant forecasts of a turbine's operating status, e.g., covering pitch speed and active power, can enable identification of potential failures, thereby enabling timely maintenance, and thus reducing regular maintenance costs, decreases in generated power, and potential safety hazards [33]. Hence, forecasting has attracted extensive research attention.…”