Fast Multilateral Response Method Based on Transfer Learning
In order to solve the time cost problem of obtaining all edge responses through transient simulation in fast time domain simulation based on multilateral response method,a multi response prediction method based on transfer learning was proposed.Firstly,based on the response characteristics of nonlinear links,the trailing re-sponse was divided into linear and nonlinear regions to be processed separately to reduce the required measure-ment response length;Secondly,a pre trained model network was constructed using the"0 level"response as the source domain data;Then,the network parameters of the pre-training model were used to train the other three response transfer learning models;Finally,all nonlinear region responses were plotted within a unit inter-val,and the accumulated voltage in the linear region was extended to the eye contour.The results indicated that this method predicted the remaining responses more comprehensively and accurately with fewer response data,thereby improving the calculation speed of the fast multilateral response method..
NETL
NSTL
万方数据
维普
