Research On Health Assessment And Fault Diagnosis Of Key Components Of Wind Turbine Based On Data Driven

As a clean,low-carbon renewable energy source,wind energy has gradually become one of the most widely used and developed renewable energy sources in recent years.However,because wind turbines often suffer from the effects of irregular and complex alternating loads,and their installation locations are generally remote,as a result,its key components are prone to failure and maintenance is difficult after failure.In order to understand the operating status of wind turbines in real time so that pre-maintenance and operation and maintenance strategies can be developed in advance,health assessment and fault diagnosis technologies are particularly important.First,the basic structure of the wind turbine is introduced,and the key components of the wind turbine wich the gear box and the pitch system are explained in detail.At the same time,the common faults and their causes are analyzed,which lays the foundation for the health assessment and fault diagnosis later.The collected data of SCADA(Supervisory Control And Data Acquisition)are listed and preprocessed the data.In order to mine valid signs with category sensitivity from the collected data and prevent too many data dimensions from interfering with health assessment and fault diagnosis.In this paper,a distance-based feature selection method is used to filter the effective features of the data to reduce the dimension of the sample data.Then a fault diagnosis method based on the combination of s-AE(Stacked Auto Encoder),sparsity limitation and Gaussian noise reduction is proposed.Through the representation learning capability of the stacked autoencoder neural network,the hidden associations between features are learned and faults are classified.Aiming at the gradient dissipation phenomenon that is easy to appear in multi-layer neural networks,a batch standardized algorithm is used to suppress it.Using the actual recorded data of SCADA to diagnose and verify common faults in the gearbox and pitch system of wind turbines,and the experimental results show that compared with classification models such as support vector machines,stack autoencoders,and random forests,the fault diagnosis method proposed in this paper which has higher diagnostic accuracy and performance.Finally,a wind turbine health assessment method is proposed that use IAGA(Improved Adaptive Genetic Algorithm,).to optimize the AANN(Auto-Associative Neural Network,)The residual distribution of the operating data of the wind turbine under normal conditions was obtained through AANN,and the deviation from the residual distribution at the moment of failure was calculated using JS(Jensen-Shannon)divergence to determine the health of the wind turbine status and whether it is faulty.Using the historical data recorded by SCADA to experimentally verify the health of the gearbox and pitch system of the wind turbine.The results show that the health assessment method in this paper can predict the potential failure of wind turbines in advance and accurately grasp the health of key components.