Research On Modeling Method Of Condition Monitoring Of Wind Turbine

Wind turbine is a complex nonlinear coupled generation system,its operating environment is complex and operating conditions is poor,resulting in a higher frequency of the unit itself,increasing the operation and maintenance costs.Therefore,the condition monitoring of wind turbine can be found in time to fault,and take the corresponding measures.Wind Turbine Supervisory Control and Data Acquisition(SCADA)records a lot of information about the operating status of the unit and the data is easier to obtain.In this paper,based on SCADA data of wind turbine,the key sub-component(spindle,generator and so on)condition monitoring model of wind turbine is established and corresponding fault diagnosis is carried out.1 Introduce the SCADA System structure of wind turbine and analyze its function.Statistics on the possible failure of key sub-components of wind turbines and analyze the cause of the malfunction.2 The Gaussian Process Model of spindle temperature is established of wind turbine.As a connecting part between the impeller and gearbox,spindle has the function of transmitting torque and energy in the transmission chain of the unit.The condition monitoring of spindle can effectively improve the unit operating efficiency.Aiming at the characteristics of strong randomness and high noise of the wind turbine operating data,using the Gaussian process regression method establish the input variable of the model.Because SCADA Data has the characteristics of large amount of data,multiple correlation,non-linear and randomness of data distribution,it makes the modeling process complex and takes a long time.To reduce the complexity of modeling,the fuzzy kernel clustering method is used to filter the original operation data,and the redundant is established to construct a compact and effective modeling sample set.In order to improve the sensitivity and reliability of the spindle abnormally early warning,using the double moving window calculate the statistic properties of the residual sequence.If the moving residual mean or standard deviation exceeds the set fault alarm threshold,an alarm message will be issued.3 The generator fault identification model is established of wind turbine.Generator as one of the key subassemblies of the wind turbine,it operating condition directly affects the overall performance and power efficiency of the unit.Monitoring its operating status can effectively reduce the operation and maintenance costs.Aiming at the characteristics of strong randomness and high noise of the wind turbine operating data,the wind speed is taken as a reference variable to preprocess the raw data by using the box method to eliminate interference and extract valid data information.There are high correlations and coupling between the variables of the wind turbine.For this characteristic,the method of partial least squares is used to analyze the generator bear temperature.By calculating the projection importance index of each relevant variable,the influence of each correlation variable on the bearing temperature of the generator is determined.Based on the pre-processed SCADA data,the condition monitoring model of the generator bearing temperature is established and the corresponding condition monitoring criterion is calculated.It has the ability in tracing fault further deterioration.