Condition Monitoring Of Wind Turbine Generator Based On Spark And Neural Network

Wind energy,as a renewable energy,inexhaustible,clean and environmental protection,has become an important part of the sustainable development strategy of many countries,therefore,wind power has been a rapid development.The working environment of the wind turbine is poor,and the environment factors such as the normal and extreme temperature,rainfall,snow,dust,solar radiation and so on have been affected,so each component is bound to be inevitable with the change of running time and aging,reliability decline,leading to the failure occurred,affecting the safety and stability of the wind farm.Wind power generator is the components of the unit with high failure rate,so real-time monitoring of the timely detection of fault symptom and determining the reasonable maintenance plan is of great significance to reduce maintenance costs and improve the reliability.At present,the wind turbine collects the important parameters through the sensor in real time,which will make the storage data from GB to TB level,or even PB.In the context of large data,how to quickly deal with the growing mass of state monitoring data,and accurately analyze the current situation of the operation of the wind turbine has become a new topic.Under this background,using the method of temperature trend analysis,this paper studies the above problems.(1)Based on the real-time monitoring data of wind generator,a wavelet neural network model for temperature prediction of wind turbine is established.Through the correlation coefficient method to analyze the influence factors of the wind generator temperature,the network input is determined,and the number of hidden layer neurons is obtained by trial and error method,so we determine the network structure.(2)In view of the slow convergence speed and easy to fall into local optimum when using monitoring data of wind turbine to train wavelet neural network,this paper uses the improved flower pollination algorithm to optimize the parameters of wavelet neural network,including the weight,the expansion factor and the translation factor.By introducing chaos sequence and t distribution variation,flower pollination algorithm(FPA)has better optimization ability.Improved FPA is used to optimize the uncertain parameters of wavelet neural network that improves the training speed and precision of wavelet neural network.(3)According to the massive state monitoring data of wind turbine,we present a model that improved flower pollination algorithm optimizes wavelet neural network(CITDMFPA-WNN).The model is deployed on the Spark platform,and the parameters of the optimized parameters are used to predict the temperature.Through the introduction of parallel,the calculation speed is improved,and the algorithm has the ability to deal with massive data.(4)Use the above model and real-time monitoring data of wind power generator,predict the temperature of wind generator,then the residual temperature by using a sliding window statistical method,that is the predicted temperature difference with the actual temperature value,is analyzed to determine the mean and standard required for wind generator abnormal monitoring time difference threshold,so as to determine the real-time operation state of wind power generator and achieve the purpose of on-line monitoring.Finally,the comparative experiment and example analysis are carried out.The real operation data of a wind farm in Inner Mongolia is used,a cloud computing cluster is set up in the laboratory,and the performance test of the proposed algorithm and the condition monitoring and verification of wind power generator are presented in this paper.The experimental results show that the algorithm has good accuracy and parallelism,and can be used in the condition monitoring of wind power generator.