Fault Diagnosis Of Wind Power Transmission System Based On Deep Learning

Today,wind power accounts for an increasing proportion of all kinds of renewable energy development,and its maintenance needs are also widely concerned.As the core device of wind power transmission system,planetary gearbox can improve the reliability,safety and efficiency of the transmission system.However,due to the complexity of working conditions,it is difficult to diagnose different failure modes of planetary gearboxes accurately.In order to solve this problem,based on the vibration data of planetary gearbox,deep learning method is adopted to study the intelligent fault diagnosis technology of wind turbine transmission system,so as to obtain a network model with diagnosis accuracy meeting the requirements.The main contents of the study include:(1)Firstly,the empirical mode decomposition(EMD)method is used to filter and de-noise fault signals,which improves the generalization of neural network diagnosis.Then,the reconstructed signal is transformed into the data type suitable for input into the recurrent neural network,at the same time,the direct method and Gram transformation method are used to transform the image format,which is convenient for the input of the convolutional neural network.(2)A recurrent neural network structure,LSTM,is constructed to train and diagnose fault data.The optimal network structure is obtained by adjusting and optimizing various parameters.The neural network is tested by the test set,and the results prove the effectiveness of the optimized neural network.(3)In order to avoid gradient disappearance or gradient explosion,a ResNet network model based on convolutional neural network is constructed,and the data sets of two kinds of graphical processing are adjusted and trained.The results show that the gram image processing is superior,and the test set is used to test it.(4)The fault diagnosis results based on recurrent neural network and convolutional neural network are summarized and analyzed,and the deficiencies in the research are pointed out,and some improvement measures are put forward.In this paper,LSTM and ResNet network structures are built to realize the fault diagnosis of planetary gearbox of wind turbine.After testing and comparison,the ResNet network trained by Gram transformation method is more accurate and effective for the data set of filtering data.The method also enriches the related research on intelligent fault diagnosis of wind turbine transmission system and has certain theoretical and practical significance.