Operating Condition Identification of Complete Wind Turbine using DBN and Improved DDPG-SOM

Although there have been a lot of researches on identifying the condition of components in wind turbine, such as blade, gearbox, bearings, etc., maintenance immediately after some component faults is not an optimal choice in practical use, since it leads to high-frequency downtime and large unnecessary cost. Therefore, the condition identification of complete wind turbine is required. In this paper, we propose a novel method based on data-driven techniques and information fusion model. Firstly, a DBN is used for capturing information about the condition of components in wind turbine. Then, a new information fusion model is presented for integrating the information about conditions of components based on an improved DDPG by incorporating a random module and SOM. Based on the output of the improved DDPG-SOM, a status index is calculated representing the condition of the whole machine. A case study based on real SCADA data is conducted to show the effectiveness and superiority of the proposed method.