Research On The Wind Turbines Fault Diagnosis Technology Based On Hadoop 2.0 Cloud Platform

With the rapid development of wind energy industry,the use scale of wind turbines is growing.More and more people are concerned about how to maintain the safe operation of the wind turbine generator system.Among the various wind turbine fault types,the fault of the transmission system parts often leads to the longest maintenance downtime for wind turbine generator system.In the meanwhile,it is also one of the important reasons caused high maintenance costs of wind turbine.The timely and accurate fault diagnosis is very important.Therefore,this thesis takes the drive system bearing of wind turbine generator system as an example,and proposes a fault diagnosis algorithm based on the deep belief network(DBN)for the wind turbine bearing.We would improve the processing speed of the algorithm by parallelizing it on the cloud computing platform.This thesis will be mainly expounded from the following two aspects:On the one hand,the diagnostic technique uses the deep learning network as the characteristic representation of the fault data and the deep belief network(DBN)technology as the deep structure of network construction.It can highly fit the characteristics of any nonlinear signal,and learn and obtain the higher order correlation characteristics from each hidden layer of input data,and finally process output information by connecting to a feature extractor in the output cascade.In this thesis,the experimental results show that the DBN has excellent performance in data processing.The fault diagnosis algorithm based on this network model for the wind turbine bearing can achieve the unsupervised learning by normalizing the original vibration data.After completing entire DBN training,it can achieve the supervised fault recognition on the top by using the label layer information.It can avoid the complex situation of artificial experience participation in fault characteristics,and make the whole diagnosis more intelligent.On the other hand,the traditional signal analysis method will encounter the bottleneck of computing power when processing the vibration data which is growing sustainably.The cloud computing has excellent performance in big data computing and network storage.Based on Spark,it can achieve the parallel processing methods for vibration data on the Hadoop platform.It also has a good fault tolerance and automatic balance load characteristics,can effectively solve some problems,such as the slow processing speed of the vibration data collected by the wind turbine generator system,incomplete data usage,and the serious loss of valid information.This thesis achieves the fault diagnosis algorithm parallel computation based on Spark for the wind turbine bearing.Experiments on the Hadoop 2.0 cloud platform show that the parallel indexes of the algorithm are good.The algorithm can meet the fault diagnosis processing task requirements.