Abstract：The natural frequency design of wind turbine tower structure is the basis of wind power generation structure system. To overcome the disadvantages of traditional theoretical and finite element method (FEM), a new approach based on BP neural network algorithm is proposed to efficiently predict the natural frequency of the newly designed hybrid Steel-Concrete cylinder tower. First, the features and labels of the training model are determined by FEM calculation and analysis. And then, 32 FEM calculation samples are used train the frequency prediction model of hybrid Steel-Concrete cylinder tower, using the BP neural network algorithm. It has been verified that this method has a high accuracy of the first-order frequency prediction, whose error is limited within only about 1.0E-3. On the other hand, all of the prediction results using the models trained with different sample sets agree well with the FEM calculation, implying the algorithm has a high stability. In addition, the BP neural network algorithm training method can also be used to predict the multi-order frequencies of hybrid Steel-Concrete cylinder tower, which still has the high accuracy. In addition, compared with frequency calculation based on FEM method, this new approach has outstanding efficiency. From the above aspects, it is proved the frequency prediction model based on the BP neural network is efficient and workable, which can provide important guidance for the design of wind power generation structure system.
2021, Vol. 42 
