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

• DOI: 10.32604/ee.2022.020779
• Publication Date: 2022-09-14T10:10:02Z
• Authors: Sihua Wang; Wenhui Zhang; Gaofei Zheng; Xujie Li; Yougeng Zhao
• Keywords: Normalization; SCADA; Softmax function

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 SCADA data turbine operation, firstly, (GN) algorithm added solve problems stack training slow convergence speed, RMSProp used update weight bias autoenccoder, which further optimizes problem that loss function swings too much during process. Finally, last layer network, softmax activation classify results, output transformed into probability distribution. selected was substituted pre-improved improved denoising (SDA) networks for comparative verification. results show more accurate effective diagnosis, also provides reference identification.