Response Prediction of the Overhead Transmission Line Galloping Based on the POD-ANNS Method
The finite element method(FEM)is usually used to obtain the galloping response of the iced 8-buddle conductor under the tower line system.However,the FEM often takes a long time and the time cost of obtaining the galloping response is huge.The finite model is difficult to build and has many types,and it must consider thorny questions such as the calculation non-convergence in the kinetic solution,so it is difficult to obtain the galloping response under different parameters and working conditions.Based on this,this article proposes a hybrid proxy model that can achieve fast prediction of the galloping response based on proper orthogonal decomposition(POD)and artificial neural networks(ANNS).First,the galloping response obtained by the FEM is combined into a snapshot matrix,and then the POD modal basis coefficients and POD modes of the snapshot matrix are obtained based on the POD method.Second,three proxy models(BPNN,RBFNN and CNN)are used to predict the POD modal basis coefficients,and a linear combination of the predicted POD modal basis coefficients and POD modes is used to achieve rapid prediction of the galloping response.Finally,the prediction error and prediction time of three hybrid models(POD-BPNN,POD-RBFNN and POD-CNN)are compared.The results show that all hybrid models can quickly and accurately obtain the target response,and the error of the POD-RBFNN hybrid model is smaller;The number of snapshots reaching 60 can basically characterize the characteristics of all samples.
proper orthogonal decompositionneural networkfinite element analysisgalloping characteristicstower line system
NETL
NSTL
万方数据
