Predicting Temperatures of Wind Turbine Gearbox By a Variable-Weight Combined Model

Abstract

Predicting the temperature variables of the wind turbine gearbox precisely including the axis temperature and the oil temperature can evaluate the gearbox status in real time effectively. Concerning the limitations of a single prediction model, this paper proposes a variable-weight combined model to predict gearbox temperature based on the theory of grey relational degree. Firstly, Principal Component Analysis (PCA) is used to reduce the dimension of the gearbox temperature related factors, and the time series is selected to analyze the internal structure of the gearbox temperature. Then, to analyze the gray correlation degree between the four single models and the actual temperature series, eliminate a certain dynamically model and to update the remaining models weights dynamically. Compared the variable-weight combined model with the equal-weight combined model and each single model, it is shown that the variable-weight combined prediction model has higher prediction accuracy, which is of great significance for further condition monitoring of the gearbox.