The early prediction of lithium-ion battery remaining useful life using a novel Long Short-Term Memory network

Accurate prediction of the lithium-ion battery remaining useful life can effectively manage the lithium-ion battery health. Using the early cycle data to predict the remaining useful life can reduce consumption and detect battery failures earlier, but it is still a great challenge due to weak and high dimensional nonlinear feature data of the early cycle. In order to solve this issue, this paper proposes a Long Short-Term Memory (LSTM) model that combines the idea of Broad Learning System (BLS), called BLS-LSTM, to accurately forecast the lithium-ion battery remaining useful life by using early cycle data. Firstly, according to the BLS idea, more effective feature nodes are obtained by performing mapping operations and enhancement operations on input features. Secondly, the characteristic nodes are input into the LSTM as new input nodes to predict the remaining useful life of the lithium-ion battery. Finally, the proposed model is validated with different early cycle data and compared with other methods. The results show that the BLS-LSTM model has better prediction performance and higher accuracy in the early prediction of the remaining useful life.