With the development of audio carrier equipment,the loudspeaker in the trend of small volume,large power development,long time working in large signal drive when the voice coil heating serious,there will be a voice coil break and other thermal damage problems,through the prediction of the loudspeaker electrical parameters data,can reduce the power amplifier power and other methods to protect the voice coil,prolong the service life.Aiming at the time sequence characteristics of electrical parameter data of loudspeaker in power test,the prediction accuracy is improved.A multi-step prediction method for loudspeaker electrical parameters is proposed based on variational mode decomposition(VMD)and sequence to sequence(Seq2seq)model.First,VMD isused to decompose the original data to reduce the non-stationarity of the data.Then,the decomposed data is used to build a training set and the Seq2seq network model is used for training and multi-step prediction.The simulation results show that in the case of single step prediction,the model evaluation index RMSE is 0.044,MAPE is 0.15%,R²is 0.94,and in the case of five-step prediction,the model evaluation index RMSE is 0.05,MAPE is 0.17%,R²is 0.92,allofwhich are better than other comparison models.It shows that the accuracy of the proposed model is higher