Abstract

This paper presents the load-dependent power loss in a wind turbine gearbox under real-time operating wind speed for three different oil formulations. The gear power loss was determined using mathematical models from the values of gear loss factors and specific film thickness experimentally determined by other researchers. The bearing power loss was determined using the new SKF calibrated model. Wind data from Bafoussam, a town in Cameroon was used to validate the model. A back propagation neural network with different numbers of hidden neurons was designed for power loss modeling and prediction. The achieved results reveal that the load-dependent power loss in a wind turbine gearbox is greatly influenced by wind speed and oil type. Finally, it is shown that the predictive performance of the neural network is also influenced by the number of neurons in the hidden layer.