A Novel Approach of Wind Turbine Blade Root Load Monitoring Based on Bearings Vibrational Data and a New AE‐LSTM‐Attention Neural Network

• DOI: 10.1002/we.70006
• Publication Date: 2025-04-14T02:23:06Z
• Authors: Feiyun Cong; Chen Liang; Weifei Hu; Tongzhou Zhang; Feng Tang; Qihao Zhou
• Keywords: Aero engine

ABSTRACT The monitoring of wind turbine blade root load (BRL) has always been a valuable yet highly challenging problem. BRL is often measured directly using strain sensors, which are expensive to install and poses safety risks. Thus, it necessary explore an indirect measurement method that establishes mapping relationship between easily measurable data. This paper presents non‐intrusive based on the vibrational signals at bearings (including main bearing, gearbox generator bearing) new proposed AE‐LSTM‐Attention neural network. network incorporates autoencoder (AE) pre‐training process, long short‐term memory (LSTM) network, channel attention mechanism overcome difficulties in extracting fusing long‐term, high‐frequency, multichannel signal features. Additionally, considering transmission mechanism, optimal LSTM step number selection criterion proposed. turbine's physical features innovatively bridged with training parameters, e.g., setting ratio as steps, by most cost‐effective prediction performance can be achieved. Experiments were conducted data from actual operating turbines, results show successfully achieve characteristics outperform traditional methods terms predictive performance. vibration established efficiently accurately.