Accurate Estimation for State of Charge (SoC) using different topologies of LSTM

Abstract

Abstract For the best charge control of batteries and the forecast of operation range of electric vehicle, accurate estimation State of Charge (SoC) is a need. The accuracy of SoC estimation has a direct impact on these cars' operating range and safety. Accurate SoC estimation becomes a challenge due to environmental alterations, change in temperatures, and interference of electromagnetic fields. There are a lot of technologies depends on Machine Learning (ML) and Artificial Neural Network (ANN), the proposed model is using two cascaded Long Short-Term Memories (LSTM) networks that reduced the Mean Square Error (MSE). There are other models have been simulated such as Nonlinear Auto Regressive models with Exogenous input neural network (NARX) with LSTM, and a standard Long Short-Term Memories (LSTM). The proposed algorithm has reduced the error compared to a LSTM by 55% and has reduced the error compared to NARX with LSTM by 12%.