Lithium-ion batteries represent a central part of an electric vehicle, providing the electric energy storage in replacement for the fuel tank of conventional vehicles. Due to its design and manufacturing complexity, as well as complex physical processes involved, they are still expensive and may be prone to accelerated aging. Therefore, in order to achieve maximal utilisation of a vehicle battery pack, a great amount of effort has been made in the area of battery management systems (BMS) development. The BMS ensures an optimal management of the battery pack, so that it can provide high output power with minimum influence on the battery lifetime. It relies on a number of estimation algorithms that observe relevant and hardto-measure battery state variables and parameters such as state-of-charge (SoC), remaining capacity and internal resistance, as well as the remaining useful life by predicting the battery aging. The thesis proposes a comprehensive approach to state and parameter estimation based on a numerically-efficient equivalent-circuit model (ECM), as well as a data-driven aging model based on naturalistic driving cycles and physically simulated battery aging. First, an estimator of the SoC and ECM impedance parameters, based on the dual extended Kalman filter (DEKF), is set up, followed by an observability analysis of its impedance parameters. The DEKF is then extended to estimate the open circuit voltage characteristic. Due to the pronounced nonlinearity and slow dynamics of the charge capacity calculation model, the capacity estimator is based on a separate, sigma-point Kalman filter (SPKF) that is executed on a two orders of magnitude slower time scale. The special emphasis is paid to the design and analytical parameterization of the adaptive coupling of DEKF and SPKF estimators, in order to ensure the accuracy and fast convergence of the slow estimator. The proposed estimators of state variables and impedance parameters are tested experimentally by using the recorded driving cycle data of an electric scooter equipped with a NMC battery. Finally, a method of data-driven battery aging modelling is proposed, which is based on simulated aging and load cycles synthesized from the recorded data. The aging simulation is carried out by using the high-precision physical models of solid-electrolyte interphase (SEI) growth and lithium plating (LPL), both available within the GT-Suite/AutoLion simulation environment. From the wide set of proposed load cycle features, the most statistically relevant ones are selected by using the LASSO regression method. The selected features are used to create and parameterize linear regression (LR) and neural network (NN) aging models. The proposed aging model in its final form consists of an LR model of SEI layer growth, a NN classifier for estimating LPL activation, and an LR model of capacity loss due to lithium plating. The main aim of the thesis is to design real-time estimators of the key battery states and parameters, and build up a data-driven aging model for predicting the remaining battery lifetime in real time.