Research On The Fault Diagnosis Methods Of Wind Turbine Gearbox Under The Big Data

In recent years,with the rapid development of wind power,more and more wind farms have been built,a large number of wind turbines have been put into operation.Due to the wind farm is usually located in Gobi and other areas,resulting in a perennial wind turbine in a very harsh environment,the operation is very easy to run in fault.Among them,the gear box is the highest failure probability of the whole wind turbine components,according to statistics,more than 60% of the wind turbine failure occurred in the parts of the gear box.So it is important to diagnose the fault of gear box quickly and accurately to reduce the operation cost,improve the economic benefit of the wind farm and improve the reliability of the operation of the wind turbine.Firstly,in order to improve the accuracy of fault diagnosis,an algorithm based on BP neural network optimized by artificial bee colony is presented in this paper.The artificial bee colony algorithm is introduced into the traditional BP neural network,and the global search ability of artificial bee colony is used to improve the sensitivity of BP neural network to initial parameters.Secondly,according to the traditional BP neural network algorithm performs low efficiency and the practical application scenarios,the extreme learning machine algorithm is introduced into the field of fault diagnosis of gear box.And the extreme learning machine algorithm is optimized by the firefly algorithm to improve the accuracy of fault diagnosis.Finally,the parallel design of the above two fault diagnosis models is realized on the Spark platform to improve the ability of dealing with massive data.Finally,do some experiment tests.The real operation data of a wind farm is chosen to test the performance of the proposed algorithms on the cloud computing cluster which is set up in the lab and compared with the traditional fault diagnosis experiment.And the experimental resaults show that the proposed models are superior.The experimental results demonstrate the effectiveness of the designed algorithms and the good parallel performance.