Condition Monitoring And Fault Diagnosis Of Wind Turbine Based On Support Vector Machine

In recent years,with the rapid development of wind power industry,the scale of wind farm cluster has been expanding.The emergence of SCADA system has changed the operation and maintenance mode of wind farms,the staff is gradually getting rid of the harsh working environment,the cost of operation and maintenance of wind farm has also dropped.At the same time a series of problems have also emerged: a large number of high-dimensional and multi-class data collected by the wind farm control center are not fully utilized and developed,only stay in real-time data monitoring,historical data display,and report statistics,etc.Caused low utilization of data resources,and “big data” failed to play a corresponding role.In view of the above problems,based on the full utilization of the wind turbine data,this paper mainly conducts condition monitoring and fault diagnosis of the wind turbine from the following aspects.Firstly,based on the characteristics of wind power curve,a method of wind turbine performance and fault monitoring is proposed.According to the historical data set of SCADA system,the k-means algorithm was adopted to screen and clean the power characteristic curve of the wind turbine.Then,the multivariate kurtosis and multivariate statistics are introduced and applied to the power curve,the degree of deviation of reference point is used as an index to evaluate the performance of the wind turbine.The state of unit is divided into normal state,sub-health state,and abnormal state by this indicator,and a multi-variable T2 control chart is used to monitor the abnormal state time of the turbine.Secondly,the abnormal state monitoring model of wind turbine is established based on support vector regression algorithm.Since most of the wind farms have limited power measures,many factors need to be considered.The active power,tower acceleration,gear box oil temperature,pitch angle,and rotor speed are selected as output vectors of the model,and the characteristic parameters are selected as input vector.Cross validation and genetic algorithm are combined to optimize the parameters of SVM regression model.According to the confidence interval method of 0.05 confidence level,the alarm mechanism of the abnormal state of turbine is determined,which can effectively monitor the abnormal condition of the turbine.Finally,support vector machine classification model is applied to the fault diagnosis problem of wind turbines.For the problem that support vector machines is onl y suitable for two categories,the “one-versus-all” and “one-versus-one” multi-classifi cation algorithm of the support vector machine are adopted.And the method of grid search is used to optimize the parameters of SVM,then a multi-class fault analysis model based on support vector machine is established to realize multi-fault classific ation of wind turbines.