Fault Diagnosis Of Wind Turbine Blades Based On Deep Learning

With the rapid development of modern society,the demand for industrial electricity and domestic electricity is increasing day by day,and various types of power generation technologies have emerged accordingly.Wind power generation is a relatively mature power generation method with large-scale development and commercialization prospects among new energy power generation technologies.Most of China’s wind farms are built in high-latitude and high-altitude northern regions.They are in extremely cold weather all year round,and with China’s installed capacity continuing to increase,the probability of equipment fault is getting higher and higher.Among them,Wind turbine blade icing fault often occur.However,blades are the key part of wind turbine energy receiving and conversion,and their safe operation is a prerequisite for ensuring stable operation of wind turbines.Blade icing fault not only affects the power generation efficiency of wind turbines,but also easily causes changes in the structural performance of the blades.Fault will accumulate if not be maintained,the blade may break and cause casualties in severe cases.Therefore,condition monitoring and fault diagnosis of Wind turbine blade are particularly important.Traditional fault diagnosis technology based on signal processing technology and machine learning is no longer suitable for processing complex and massive wind turbine operating status data.The execution efficiency and data processing capabilities of traditional fault monitoring technology cannot meet actual application requirements.Therefore,intelligent diagnosis technology based on deep learning has become a research hotspot.Combining the characteristics of deep learning to process massive data,this paper proposes a fault diagnosis method based on the fusion of convolutional neural networks and bidirectional gated recurrent unit,obtains the wind turbine blade icing detection model,and the real data set collected by the wind farm SCADA system is used to verify the feasibility and effectiveness of the model proposed in this paper.First of all,there are problems of unbalanced categories and unstable data in the experimental data set.The data is preprocessed,the sliding window method is adopted to increase the number of fault data,and the magnitude difference between the data variables is eliminated through standardized processing,which lays the foundation for the subsequent construction of deep learning models.Secondly,there are special characteristics such as variable correlation and time dependence on experimental data.The wind turbine blade fault diagnosis method based on CNN-BiGRU fusion is proposed.It uses the correlation characteristics between the adaptive learning variables of the convolutional neural network in deep learning,selects the gated recurrent unit in the recurrent neural network to obtain the timing dependence,The two-way structure is used to introduce the information of the future moments to increase the sensitivity to time.The feature maps extracted by the two models are merged to comprehensively obtain the feature information of whether the wind turbine blade are icing or not.Finally,in order to achieve high-efficiency operation of the model,appropriate model parameters are selected through experiments,and the model is trained and tested to complete the fault classification task.The experimental results show that the deep learning fault diagnosis method proposed in this paper can efficiently process massive amounts of wind turbine operating state data,It can also improve the accuracy and timeliness of fault diagnosis.