Remaining useful life and state of health prediction for lithium batteries based on empirical mode decomposition and a long and short memory neural network

Accurate estimation and prediction of the state of health (SOH) and remaining useful life (RUL) are crucial for battery management systems, which have an important role in the field of new energy. This work combined the empirical mode decomposition (EMD) method and backpropagation long-short-term memory (B-LSTM) neural network (NN) to develop SOH estimation and RUL prediction models. The BLSTM NN of the many-to-one structure uses easily available battery parameters, such as current and voltage, to estimate the SOH. SOH data are processed through the EMD method-to reduce the impact of capacity regeneration and other situations-after which the backpropagation of the one-to-one structure NN performs a RUL prediction. Compared with the current data-driven forecasting model, the model has a simple structure and high accuracy. For SOH estimation, the average root mean square error was 0.02, which was nearly four times lower than that of a simple recurrent NN. For the RUL prediction model, EMD effectively removed noise signals and improved prediction accuracy. The prediction results of the model for different batteries showed good accuracy, indicating that this combined model has high robustness, good accuracy, and applicability. (c) 2021 Elsevier Ltd. All rights reserved.