Fault Diagnosis Of The Main Components Of The Wind Turbine Drive System Based On Deep Learning Network

Wind turbine drive system is the core component of energy conversion of wind turbine.The factors like tough working environment,uncertainty of wind speed and the changing load make the wind turbine drive system frequently malfunction.The fault diagnosis of the wind turbines‘ main components will be based on deep learning network in this paper,and the fault diagnosis system of the wind turbine drive system will be designed.The main research works are as follows:(1)The obtained experimental data of the vibration of the bearing and gear is preprocessed.The linear trend is eliminated by least square method and the cubical smoothing algorithm with five-point approximation is used to smoothly process the data,which lays the foundation for the construction of the fault diagnosis model.(2)The data of bearing and the gear that is pre-processed is called raw data,which can help to establish the fault diagnosis model of the bearing and gear of the wind turbine based on deep learning network.The key parameters of the deep learning network are set for the different hidden layers,different input sample lengths and different nodes in different hidden layers when the model is designed and analyzed.The result of case analysis shows that the feasibility and effectiveness of the fault diagnosis model.(3)After the feature parameters are extracted,the faults of the bearing and the gear are able to be diagnosed based on the fault diagnosis model of the wind turbine drive system established with the deep learning network.A comparative experiment with the BP neural network and s_kohonen clustering algorithm are made to diagnose the fault of the bearing and the gear,which demonstrates that this fault diagnosis model has a high diagnostic accuracy.(4)Based on the deep learning network,the fault diagnosis system of the main components of the wind turbine driving system are established and various models of the system are analyzed in the paper.The system includes data preprocessing model,feature extraction model,deep learning model.The experimental results verify the effectiveness of the established system.