Remaining Useful Life Prediction of Lithium-Ion Battery Based on LSTM and GRU

Remaining useful life (RUL) prediction of lithium-ion battery remains a challenging problem. Battery failure can occur when there is an abnormal capacity or power degradation that would lead to system downtime and catastrophic occurrence. Thus, it is necessary to build an accurate prediction model to ensure the battery is reliable and safe. Data-driven method using machine learning has drawn much attention in this research area. This study addresses the battery RUL prediction based on long short-term memory (LSTM) and gated recurrent unit (GRU). A comparison was made between LSTM and GRU model performance and complexity. Experimental results show that GRU could achieve better performance compared to LSTM with almost 20% less than the number of parameters of LSTM.