关键词: 风电机组, 宽度学习算法, 集合经验模态分解, 交叉熵

Abstract: To achieve predictive maintenance of the gearbox of wind turbines,a fusion early warning model based on Ensemble Empirical Mode Decision (EEMD) and broad learning algorithm was proposed aiming at the over temperature fault of gearbox oil temperature.Using the gearbox data in a healthy state as a criterion,the gearbox oil temperature timing signal were decomposed by EEMD to obtain the time-frequency characteristics,and then the Supervisory Control and Data Acquisition (SCADA) data was used to model the gearbox oil temperature with broad learning algorithm.Mahalanobis distance and the correlation between the fitting result and the true value were used as indicators to evaluate the health of gearbox from the time dimension and the related variable dimension.By calculating the cross entropy of two algorithms,the early warning results of the two were fused,so as to take into account the accuracy and rapidity of the early warning method.The data recorded before and after the gearbox oil temperature overtemperature fault in the actual wind field were analyzed,and the results verified that the fusion method of EEMD change point and width learning algorithm was feasible in the early fault warning of gearbox oil temperature over temperature.

Key words: wind turbine, width learning algorithm, ensemble empirical mode decomposition, cross entrop