Research On Fault Warning Method Of Wind Turbine Based On Wind Field Data Mining

Wind energy is one of the renewable clean energy,with the advantages of wide distribution,easy access,no pollution and so on,so it has a good development prospect.However,due to the influence of the working environment,the long time operation of the wind motor is likely to lead to some failures or accidents,which will eventually bring serious impact to the wind farm.Therefore,it is necessary to study the fault warning strategy of wind turbine,which can not only monitor the overall operation state of the unit in real time,but also find out some potential faults and hidden dangers in time,so as to avoid equipment damage and unplanned shutdown.In this paper,combined with the wind field data,the gearbox bearing faults of wind turbines are studied,and the corresponding fault warning strategies are analyzed and designed.The main research contents are as follows:In view of the research object in this paper,SCADA data of wind turbine is preprocessed and a deep learning model based on long and short-term memory neural network(LSTM)is established to realize early warning for bearing faults.In order to avoid false alarm,the box diagram is used to deal with the predicted residual of the model and set the alarm rules.The experimental results show that the method can detect the device abnormality about 27.5 hours in advance,and realize the early warning for the potential fault,which verifies the effectiveness of the method.Wind energy has strong randomness and volatility,which makes the operating conditions of the unit complex and changeable.In order to improve the strain capacity of the warning model for the change of working conditions,a fault warning model based on fuzzy C-means clustering and long and short-term memory neural network(FCM-LSTM)is established,which is one of the innovations of this paper.The FCM algorithm is used to carry out cluster analysis on the sample data,and the corresponding prediction model is established according to different sub-conditions.The final results show that the method can detect the device abnormality about 28.9 hours in advance,which has improved the fault warning time to some extent.In order to further improve the early warning effect of the early warning model for gearbox bearing faults,a fault early warning model based on convolutional neural network and long and short-term memory neural network(CNN-LSTM)is established.This model can effectively combine the advantages of CNN and LSTM networks to perform sample data In-depth feature extraction,mining deeper feature information.It can also avoid the occurrence of overfitting,which is one of the innovations of this paper.The verification results show that this method can detect equipment abnormalities approximately 30.5hours in advance,and is significantly better than the previous two methods in terms of early warning time.It proves that the deep extraction of data features by CNN can further effectively improve the fault warning effect.