Research On Real-time Fault Detection And Early Warning Of Wind Turbines

The gearbox is the most important equipment in the wind turbine,so it is of great significance to the fault detection and early warning research of the gearbox.At present,the commonly used fault detection methods for gearbox faults include: vibration analysis,fault diagnosis based on performance parameter analysis in operating data,and Bayesian network processing methods based on uncertain information.However,the above methods are prone to missed or misjudged due to inaccurate threshold settings.In this context,this paper proposes to use neural network algorithm to extract and train the fault information of wind turbine gearbox by taking advantage of its high accuracy and strong learning ability,and finally realize fault detection and early warning.The content of the thesis is as following aspects:1 In view of the poor accuracy of the previous fault detection methods,misjudgments and missed judgments are prone to occur,this article combines big data,deep learning and other knowledge to propose a recurrent neural network prediction model based on the attention mechanism.Through training and adjustment of the model,and comparing the results with other related models,the model with the best prediction results is finally obtained and applied to the fault detection of wind turbine gearboxes,practice shows that the fault detection and prediction effect of the model is good.2 Through the research on the wind turbine fault detection and early warning system,combined with actual work experience,this paper designs the overall solution of the fault detection and early warning system,including the design of data collection service,data receiving and forwarding service,etc.,and realizes the characteristic data of the wind turbine gearbox.Extraction and collection,as well as fan fault detection and early warning related functions.