Stacked bidirectional long short-term memory networks for state-of-charge estimation of lithium-ion batteries

Abstract

State-of-charge (SOC) estimation of lithium-ion batteries based on deep learning techniques, especially recurrent neural networks (RNNs), has recently garnered much attention. However, the potential of RNNs in SOC estimation has not been fully exploited in terms of the capture of temporal dependencies and the depth of model structure. In this paper, a stacked bidirectional long short-term memory (SBLSTM) neural network is proposed for SOC estimation. In contrast to unidirectional RNN-based methods, the proposed model employs bidirectional LSTM layers that enable it to capture battery temporal information in both forward and backward directions and summarize long-term dependencies from past and future contexts. Furthermore, the bidirectional LSTM layers are stacked to construct a deep structure that enables the model to characterize the non-linear and dynamic relationship between the input battery measurements and the output SOC on a layer-by-layer basis. By introducing the stacked multilayer and bidirectional recurrent structure, SBLSTM can completely utilize the battery temporal information to estimate SOC value. The experiments were conducted using two public battery datasets to evaluate the validity and applicability of the SBLSTM, and our findings indicate that it can achieve good SOC estimation accuracy for different battery types at various ambient temperature conditions.