Data Driven Estimation of Electric Vehicle Battery State of Charge Informed By Multi-Physics Modeling

Abstract

State of charge (SOC) estimation in lithium-ion batteries (LIBs) is a crucial task of the battery management system (BMS) in electric vehicle (EV) applications. An accurate evaluation of the remaining capacity in a battery is cumbersome due to the non-linear and coupled behavior of the physical processes involved in the battery operation, in particular when LIBs is integrated in dynamic systems with many components such as in EV. With the continued progress in the development of data driven models, different machine learning (ML) and, more specifically, deep learning (DL) techniques have been proposed to predict the SOC in EV.1–4 While these models have been proven to be effective in estimating the SOC variation, the reliability of learning-based algorithms is highly dependent on the data collected for the training and validation purposes. The most straightforward methodology for data collection consists of discharging a lithium-ion cell in a laboratory by applying a discharge current which aims to represent the vehicle’s operation. However, such an approach could not allow to generate data representing realistic driving conditions since the dynamics of components, such as the electric motor and the powertrain system, which have an impact on the SOC and external conditions such as the speed of the wind are not taken into account. In this work, we propose a modeling framework based upon Matlab/Simulink automotive simulations of EV in order to generate a dataset reflecting practical driving conditions. The generated datasets have been used to train ML and DL models for the final SOC estimation. In particular, the most accurate estimation of SOC is obtained with the Long Short Term Memory (LSTM) network, a Recurrent Neural Network (RNN) designed for times series predictions with long term dependencies. Furthermore, employing a multi-physics modeling of the LIBs’ operation comprising the electric motor, the powertrain system and the overall vehicle dynamic, allows to investigate the effect of EV’s driving on the electrochemical processes and reactions occurring inside the battery of different chemistries. For this purpose, we use an electrochemical model of LIBs developed in Comsol Multiphysics combined with EV simulation, in order to study the degradation of the battery as a result of multiple charge and discharge cycles representing the vehicle operation. In particular, formation/decomposition of the solid electrolyte interphase (SEI) at the negative electrode has been considered as a possible degradation phenomenon. Thus, in addition to the development of training data for learning based techniques for SOC estimation informed by EV simulations, the proposed modeling approach allows the investigation of the battery functionality and degradation under realistic driving conditions.(1) Álvarez Antón, J. C.; García Nieto, P. J.; de Cos Juez, F. J.; Sánchez Lasheras, F.; González Vega, M.; Roqueñí Gutiérrez, M. N. Battery State-of-Charge Estimator Using the SVM Technique. Appl. Math. Model. 2013, 37 (9), 6244–6253. https://doi.org/10.1016/j.apm.2013.01.024.(2) Hu, C.; Jain, G.; Schmidt, C.; Strief, C.; Sullivan, M. Online Estimation of Lithium-Ion Battery Capacity Using Sparse Bayesian Learning. J. Power Sources 2015, 289, 105–113. https://doi.org/10.1016/j.jpowsour.2015.04.166.(3) Kang, L. W.; Zhao, X.; Ma, J. A New Neural Network Model for the State-of-Charge Estimation in the Battery Degradation Process. Appl. Energy 2014, 121, 20–27. https://doi.org/10.1016/j.apenergy.2014.01.066.(4) Chemali, E.; Kollmeyer, P. J.; Preindl, M.; Ahmed, R.; Emadi, A. Long Short-Term Memory Networks for Accurate State-of-Charge Estimation of Li-Ion Batteries. IEEE Trans. Ind. Electron. 2018, 65 (8), 6730–6739. https://doi.org/10.1109/TIE.2017.2787586.