Research On Health Condition Monitoring Of Wind Turbine Based On Data Mining

With the shortage of the world's energy.Wind energy as a kind of green new energy without pollution has received more and more attention.Wind power generation technology has also been widely used all over the world.However,due to poor operating environment,wind turbine frequent faults.The difficulty of maintenance and operation and maintenance costs are high.This poses a serious challenge to the development of wind power generation.It is of great significance to monitor the health status of wind turbines and realize the early warning of early failures to ensure the safe and stable operation of the generating units and reduce the operating costs.This paper first introduces the structure of wind turbine,the common faults of wind turbine and the method of early warning,which l ays a foundation for the research.And then,this paper analyzes the operating characteristics of wind turbines theoretically.After wind farm SCADA system data analysis and pretreatment,select the appropriate parameters.The fuzzy C-means algorithm is used to partition the operational space of wind turbines.Four operating sub conditions are obtained,which are the same as the theoretical analysis.Then,using the state of health SCADA historical data,using the GMM modeling method,established a wind turbine health status monitoring model.Calculating the average health decline index introduced as an evaluation index,by observing the changes in the average health decline index,can effectively predict the occurrence of failure in advance.Finally,the method of health monitoring of wind turbines based on working condition identification and GMM is proposed.The different working conditions obviously affect the accuracy of fault warning.The data are modeled separately under four operating conditions,and the test data is also divided into the corresponding model of the condition input.The average health decline index was calculated by this method.The model was verified using SCADA data from the actual wind farm.After the division of working conditions,failure warning can effectively reduce the false alarm rate.The gearbox fault data and generator fault data are analyzed,it is found that this method can reflect the state of health of wind turbines in time for different faults.Verify the comprehensiven ess of the model.