Battery state of health and state of charge estimation: comparison between classical and machine learning techniques

The topics covered in this thesis work are related to the field of electric vehicles (EVs) optimization. Thanks to their attractive properties, the majority of EVs adopt lithium-ion batteries as main energy source introducing new challenges in the car manufacturer's world. In order to guarantee the optimal management and the safety of the operations performed on the battery, a vehicle subsystem, called Battery Management System (BMS), has to estimate the state of the battery through two fundamental parameters: the State of Charge (SoC) and the State of Health (SoH). Precisely knowing these quantities in a real driving context is actually a challenging task and, for the remarkable industrial value, it has become a hot research topic in the last decade. In the first part of the thesis report the key aspects of the problem are introduced with a top-down approach, and a state of the art analysis is performed by describing the most relevant SoC and SoH estimation approaches that exist in the literature. In the second part, the thesis work is presented: in the context of MATLAB environment, a first principles approach (classical) estimation technique, based on Extended Kalman Filter, and a machine learning approach are developed under simulation. Finally, both the SoC and SoH estimators are validated and compared using experimental data.