Ensuring high performances and lifetime of battery packs has critical importance, because of the transition toward electric mobility. Therefore, correct estimation of the battery state with ad-hoc designed Battery Management Systems (BMS) is pivotal to address this challenge. In this context, application of Machine Learning (ML) is gaining increasing research interest as it includes data-driven algorithms that enable accurate and fast predictions of the battery state. For this reason, this paper aims to contribute with: (i) a survey of the newest contributions to the prediction of the State of Charge/Health (SoC/SoH), and (ii) by schematizing a methodology that uses simulated data to train state-of-the-art types of neural networks (NNs) for SoC and SoH estimation of a LiNMC battery cell. Research papers considered in this review included applications of deep NN, and other ML algorithms. The impact of the training dataset on the performances of the ML models and their capability to generalize is remarked throughout the paper. For this reason, a validated electro-thermal model is used to generate data that accounts for different temperatures and current loads to simulate scenarios with different environmental conditions and driving cycles.
Keywords Battery Management System (BMS), Machine Learning (ML), Neural Network (NN), State of Charge (SoC) estimation, State of Health (SoH) estimation