. From a novel classification of the battery state of charge estimators toward a conception of an ideal one. Journal of Power Sources, Elsevier, 2015, 279, pp.694 -706. 10 b s t r a c tAn efficient estimation of the State of Charge (SoC) of an electrical battery in a real-time context is essential for the development of an intelligent management of the battery energy. The main performance limitations of a SoC estimator originate in limited Battery Management System hardware resources as well as in the battery behavior cross-dependence on the battery chemistry and its cycling conditions. This paper presents a review of methods and models used for SoC estimation and discusses their concept, adaptability and performances in real-time applications. It introduces a novel classification of SoC estimation methods to facilitate the identification of aspects to be improved to create an ideal SoC model. An ideal model is defined as the model that provides a reliable SoC for any battery type and cycling condition, online. The benefits of the machine learning methods in providing an online adaptive SoC estimator are thoroughly detailed. Remaining challenges are specified, through which the characteristics of an ideal model can emerge.

An efficient estimation of the State of Charge (SoC) of a battery is a challenging issue in the electric vehicle domain. The battery behavior depends on its chemistry and uncontrolled usage conditions, making it very difficult to estimate the SoC. This paper introduces a new model for SoC estimation given instantaneous measurements of current and voltage using a Switching Markov State-Space Model. The unknown parameters of the model are batch learned using a Monte Carlo approximation of the EM algorithm. Validation of the proposed approach on an electric vehicle real data is encouraging and shows the ability of this new model to accurately estimate the SoC for different usage conditions.