State of Charge Estimation of Lithium-Ion Batteries Using Long Short-Term Memory and Bi-directional Long Short-Term Memory Neural Networks

This research proposes a data-driven method for estimating the state of charge of lithium-ion batteries using two neural networks, namely long short-term memory (LSTM) and bidirectional LSTM. The two schemes are computationally evaluated for various temperatures, changing the quantum of input samples, with different training and testing datasets, with additional chemistry battery, and introducing a dropout layer. Statistical error indices, namely RMSE and MAE, are calculated for three training optimization algorithms, namely SGDM, RMSProp, and ADAM. These results show the potential of the NNs in estimating the SoC. The accuracy is also compared with existing well-established data-driven methods employing DNN, CNN-GRU, and GRU NNs. It is observed that the proposed NNs have simple topologies and that the SoC estimate findings are reasonably accurate. © 2023, The Institution of Engineers (India).