Residual Life Prediction Research For Mechanical Drive Components Of Windturbine Based On Extreme Learning Machine

With the rapid development of economic society and the increasing of people’s living standard, the countries in the world demand for energy are continuing to increase. And as a renewable green energy, wind energy has got more and more attention. However, with the developing of wind power industry, the ability to accurately and timely estimate whether wind turbines operating condition change, assess current running state, predict development trend, estimate the residual useful life of important components, provide guidance for maintenance are problem to be solved at present. In this paper, the residual life prediction problem of wind turbine mechanical drive components is studied. Principal Component Analysis(PCA) and Extreme Learning Machine(ELM) are used to predict the running tendency and remaining useful life of components. And the paper chooses the bearings of mechanical drive components to carry out experimental verification as the object of study. This paper provides a basis for the trend prediction and the residual life prediction of wind turbine mechanical drive components. The specific research ways are as follows:(1) The trend prediction analysis of time series based on ELM is studied. The parameters of ELM need to be optimal selected, and the trend prediction model based on ELM is built. The simulation data and the whole lifetime accelerated vibration data are carried out to validate the feasibility and the effectiveness of the proposed prediction model. Meanwhile, the BP and Support Vector Machine(SVM) algorithm are used for comparative study.(2) Considering that the residual useful life of mechanical drive components are influenced by a variety of factors, the residual life prediction model based on PCA and multivariable ELM is proposed. The proposed method uses PCA to eliminate the redundant data of multiple features. And multivariable ELM is utilized to estimate the remaining useful life. After that, the whole lifetime accelerated vibration data are employed to verify the effectiveness.(3) The prediction system of wind turbine mechanical drive components which is designed by Lab VIEW software is built. Demand analyses of the system and overall design research are processed to achieve the vibration data acquisition, trend prediction of time domain and frequency domain features, and remaining useful life prediction of the components.