Matching contrastive learning: An effective and intelligent method for wind turbine fault diagnosis with imbalanced SCADA data

Sun, Shilin; Hu, Wen‐Yang; Liu, Yuekai; Wang, Tianyang; Chu, Fulei

doi:10.1016/j.eswa.2023.119891

Cited by 19 publications

(5 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Layer normalization has the characteristic of weight scaling invariance, which returns the distribution of input features to the non-saturated region of the activation function. It can effectively mitigate gradient disappearance and explosions, and has an accelerated training effect under the training strip of a single sample; thus, it is widely used in deep-learning model training [18,34,35]. The specific operation process of layer normalization includes the calculation and normalization of the mean and variance of a single sample's features.…”

Section: Layer Normalizationmentioning

confidence: 99%

“…Jiang et al [17] added Gaussian noise to monitoring data and trained them with a denoising autoencoder (DAE) to achieve intelligent fault detection of wind turbines under noisy conditions. Data-driven fault-detection methods based on deep learning are often applied to early warning systems in wind turbine gearboxes, owing to their ability to deeply extract hidden fault features in the gearbox [18].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A multi-head self-attention autoencoder network for fault detection of wind turbine gearboxes under random loads*

Yu,

Zhang,

Tang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Wind turbine gearboxes work under random load for extended periods of time, and the fault detection indicator constructed by the existing deep learning models fluctuate constantly due to the load, which is easy to cause frequent false alarms. Therefore, a multihead self-attention autoencoder network is proposed and combined with a dynamic alarm threshold to detect faults in a wind turbine gearbox subjected to random loads. The multiheaded attention mechanism layer enhances the feature-extraction capability of the proposed network by extracting global and local features from input data. Furthermore, to suppress the influence of the random load, a dynamic warning threshold was designed based on the reconstruction error between the inputs and outputs of the proposed network. Finally, the effectiveness of the proposed method was verified using the vibration data of wind turbine gearboxes from an actual wind farm.

show abstract

Section: Layer Normalizationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A multi-head self-attention autoencoder network for fault detection of wind turbine gearboxes under random loads*

Yu,

Zhang,

Tang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…However, as stated in [24], these have not yet been used for the condition monitoring (CM) of WTs. Other studies that motivated this work were [25][26][27][28][29][30][31][32].…”

Section: State Of the Art And Motivationmentioning

confidence: 99%

Classification of Highly Imbalanced Supervisory Control and Data Acquisition Data for Fault Detection of Wind Turbine Generators

Maldonado-Correa,

Valdiviezo-Condolo,

Artigao

et al. 2024

Energies

View full text Add to dashboard Cite

It is common knowledge that wind energy is a crucial, strategic component of the mix needed to create a green economy. In this regard, optimizing the operations and maintenance (O&M) of wind turbines (WTs) is key, as it will serve to reduce the levelized cost of electricity (LCOE) of wind energy. Since most modern WTs are equipped with a Supervisory Control and Data Acquisition (SCADA) system for remote monitoring and control, condition-based maintenance using SCADA data is considered a promising solution, although certain drawbacks still exist. Typically, large amounts of normal-operating SCADA data are generated against small amounts of fault-related data. In this study, we use high-frequency SCADA data from an operating WT with a significant imbalance between normal and fault classes. We implement several resampling techniques to address this challenge and generate synthetic generator fault data. In addition, several machine learning (ML) algorithms are proposed for processing the resampled data and WT generator fault classification. Experimental results show that ADASYN + Random Forest obtained the best performance, providing promising results toward wind farm O&M optimization.

show abstract

“…In recent years, data-driven machine learning methods have been applied in load identification with the development of artificial intelligence and the rapid improvement of computing power [21][22][23]. The complex convolution mapping relationship between impact load and response, which is hard to solve, can be replaced with well-trained machine learning model.…”

Section: Introductionmentioning

confidence: 99%

Impact load identification and localization method on thin-walled cylinders using machine learning

Guo¹,

Jiang²,

Yang³

et al. 2023

Smart Mater. Struct.

View full text Add to dashboard Cite

In this paper, a novel impact load identification and localization method on actual engineering structures using machine learning is proposed. Three machine learning models, including a Gradient Boosting Decision Tree (GBDT) model based on ensemble learning, a Convolutional Neural Network (CNN) model and a Bidirectional Long Short-Term Memory (BLSTM) model based on deep learning, are trained to directly identify and locate impact loads according to dynamic response. The GBDT model and the CNN model can reversely identify force peak and location of impact loads. The BLSTM model can reconstruct the time history of impact loads. The method is verified on a thin-walled cylinder with obvious nonlinearity. The result shows that the method can accurately identify impact loads and its location. The characteristics of the three models are compared and the influence of structural boundary conditions on the accuracy of identification is discussed. The method proposed in the paper has the potential to be applied to various engineering structures and multiple load types.

show abstract

Matching contrastive learning: An effective and intelligent method for wind turbine fault diagnosis with imbalanced SCADA data

Cited by 19 publications

References 39 publications

A multi-head self-attention autoencoder network for fault detection of wind turbine gearboxes under random loads*

A multi-head self-attention autoencoder network for fault detection of wind turbine gearboxes under random loads*

Classification of Highly Imbalanced Supervisory Control and Data Acquisition Data for Fault Detection of Wind Turbine Generators

Impact load identification and localization method on thin-walled cylinders using machine learning

Contact Info

Product

Resources

About