Open Access

ARTICLE

Fault Diagnosis of Wind Turbine Generator with Stacked Noise Reduction Autoencoder Based on Group Normalization

Sihua Wang^1,2, Wenhui Zhang^1,2,*, Gaofei Zheng^1,2, Xujie Li^1,2, Yougeng Zhao^1,2

1 School of Automation and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou, 730070, China
2 Rail Transit Electrical Automation Engineering Laboratory of Gansu Province (Lanzhou Jiaotong University), Lanzhou, 730070, China

* Corresponding Author: Wenhui Zhang. Email:

(This article belongs to the Special Issue: Wind Energy Development and Utilization)

Energy Engineering 2022, 119(6), 2431-2445. https://doi.org/10.32604/ee.2022.020779

Received 11 December 2021; Accepted 08 March 2022; Issue published 14 September 2022

Abstract

In order to improve the condition monitoring and fault diagnosis of wind turbines, a stacked noise reduction autoencoding network based on group normalization is proposed in this paper. The network is based on SCADA data of wind turbine operation, firstly, the group normalization (GN) algorithm is added to solve the problems of stack noise reduction autoencoding network training and slow convergence speed, and the RMSProp algorithm is used to update the weight and the bias of the autoenccoder, which further optimizes the problem that the loss function swings too much during the update process. Finally, in the last layer of the network, the softmax activation function is used to classify the results, and the output of the network is transformed into a probability distribution. The selected wind turbine SCADA data was substituted into the pre-improved and improved stacked denoising autoencoding (SDA) networks for comparative training and verification. The results show that the stacked denoising autoencoding network based on group normalization is more accurate and effective for wind turbine condition monitoring and fault diagnosis, and also provides a reference for wind turbine fault identification.

---

Keywords

---