基于ICEEMDAN与BiLSTM的交直流混联电网故障识别OA

To address the problems of difficult fault feature extraction and low accuracy of fault type identification in hybrid AC-DC power grids,a fault type identification method of AC-DC hybrid grids has been proposed,which is based on improved complete ensemble empirical modal decomposition with adaptive noise(ICEEMDAN)and honey badger algorithm,using to optimize the hyper parameters of bidirectional long and short term memory neural network.Firstly,ICEEMDAN is used to decompose the line fault voltage and current.Secondly,based on the joint feature index of variance contribution ratio and correlation coefficient,the modal components are selected and their peak-to-peak values are obtained,and the peak-to-peak values of the selected modal components are inputted as eigenvectors into the optimised bi-directional long and short term memory network of the honey badger algorithm for fault type identification.Finally,a hybrid AC-DC grid model is built in PSCAD/EMTDC to validate the proposed method.The experimental results show that the proposed algorithm can achieve 99.5%fault identification accuracy,which is less affected by different fault lines,fault transition resistance,fault location and data loss,and the fault identification accuracy for noise-containing signals can achieve 95.9%under 10 dB noise interference,which has strong adaptability.