Remaining Useful Life Prediction of Lithium Batteries Based on ISW and Optimized VMD-LSTM
Aiming at the problem that the remaining useful life(RUL)of lithium battery is difficult to predict accurately due to the capacity regeneration and the continuous curve fluctuation during capacity decline,a long short-term memory(LSTM)neural network prediction model based on variational mode decomposition(VMD)and improved sliding window(ISW)is proposed.Firstly,VMD is used to decompose the capacity data and distinguish the trend of main degradation and the capacity regeneration.Then,ISW is used to capture the curve fluctuation dynamically to improve the prediction accuracy.Finally,LSTM is used to model.The parameters of LSTM and VMD are optimized using Bayesian optimization(BO).The experiment uses NASA data set and is further verified on CALCE data set.Compared with SW-LSTM and ISW-LSTM models,the results show that the proposed method has smaller prediction error and higher stability,which can effectively eliminate the influence of capacity regeneration and curve fluctuation,and has generalization performance and real-time processing capability.
万方数据
维普
