Research On Health Assessment And Prediction Method Of Degradation Trend Of Wind Turbine Bearings

Wind turbines operating environment is poor.It is prone to the degradation performance and state by the weather and other uncertainties.The failure of key components makes maintenance time longer and increases wind farm operation and maintenance costs.Bearing is the critical component of wind turbine,the running condition of the bearing plays an important role in the reliability of the whole equipment.Based on the data of Supervisory Control and Data Acquisition(SCADA)system of wind turbine,the bearings are studied from both health assessment model and the state degradation trend prediction.The model of bearing health degradation assessment and trend forecasting is established.This paper takes the bearing temperature as research object.Considering the influence of bearing temperature on wind speed and power,Bin method is used to divide the working conditions.Using the relative evaluation criteria is to determine the bearing health sample set.The model is proposed to evaluate the state of the bearing temperature parameters of the wind turbine based on the least squares curve fitting.The degradation degree is obtained by combining the upper and lower thresholds of the actual operating state.Considering at the nonlinear problem of the wind turbine bearing degradation trend,the forecasting model based on time series neural network is established,which is validated and compared with other models.Application of the previous model assessment,there are still wind turbine bearing degradation trend instability,which will affect the forecast results.Degradation trend with non-steady characteristics is decomposed by Ensemble Empirical Mode Decomposition(EEMD)to obtain several relatively steady components before the prediction,and then the prediction is performed to each component by time series neural network.The predicted results of all the components are added to obtain final prediction result.The results show that the time series neural network prediction model has certain advantages for non-linear data,which can meet the requirement of monitoring parameters of wind turbine generators,and has good practicability for detecting the potential faults of early generating units.For the time series with strong non-linearity and non-stability,the combined forecasting model of this paper can more effectively follow the tracks of bearing health degradation trend for wind turbine,and can significantly improve the prediction accuracy.