Fault Diagnosis And Prediction Of Wind Turbine Bearing Based On Optimized Deep Confidence Network

With the increasing demand for energy in society,the pollution caused by fossil energy has become increasingly serious.As an important component of clean energy,the development of wind energy has gradually received worldwide attention.With the increase of wind turbine assembly capacity,the operation and maintenance of related units and equipment has become a subject of widespread concern.The surrounding environment where wind turbines are installed is often very remote and harsh,making it difficult to carry out convenient maintenance.In harsh working environments,unit components gradually wear and degrade,leading to equipment failures and even outages,resulting in huge resource waste and economic losses.In wind turbines,the bearings in the gearbox have the most frequent failures,and the rolling bearings are key components in the equipment.Therefore,fault diagnosis and prediction for them have important theoretical research value and engineering significance.Taking the rolling bearing of wind turbine as the research object,this paper proposes a fault diagnosis model based on the combination of comprehensive learning particle swarm optimization(CLPSO)and improved deep trust network(IDBN)in order to solve the problems of weak fault signal characteristics,difficulty in extraction,and low diagnosis efficiency caused by rolling bearing.In this model,an internal optimization strategy is added to the traditional Deep Belief Network(DBN)and combined with CLPSO to optimize its key parameters,thereby achieving the extraction and classification of fault features.The effectiveness of this model in fault diagnosis has been verified through experiments,and the training speed,accuracy,and other model efficiency have been improved.In order to identify fault symptoms through vibration signals,determine possible faults earlier,and perform predictive diagnosis on units in operation,thereby achieving pre maintenance of equipment and reducing equipment wear,a fault prediction model based on Long Short Term Memory(LSTM)and IDBN is proposed,utilizing the powerful processing ability of LSTM for time series information,Make up for the shortcomings of IDBN in longterm prediction,so as to achieve prediction and classification of possible future failures of wind power bearings.The effectiveness of the model is verified through experiments and comparisons.In order to improve the overall reliability of wind turbine equipment,research was conducted on the residual life prediction of wind turbine bearings.A residual life prediction model based on IDBN-LSTM was proposed.With the powerful feature extraction ability of IDBN,deeper features were extracted while preserving the original data information as much as possible.Then,the feature signals were used to predict the residual life using LSTM.Through experiments and comparisons,the effectiveness of this model in residual life prediction was verified,And improvement in prediction accuracy.