Adaptive model-based battery management - Predicting energy and power capability
The battery is the limiting system for automotive electrication due to cost, size, and uncertain degradation. To be competitive the battery must therefore be used optimally. This thesis address the on-line battery management problem, with primary objectives to:
(i) enable optimal usage of the battery by providing accurate estimates of its power and energy capability, while
(ii) ensuring durability by keeping the battery inside predefined operating limits at all times.
This means translating measurable information of current, voltage, and temperature into cell related quantities such as state-of-charge (SOC) and state-of health (SOH), and vehicle related quantities such as power capability and available energy.
The main difficulty of battery management is that battery cells have complex, non-linear dynamics that changes with both operating conditions, usage history, and age. This thesis and he appended papers proposes a system of adaptive algorithms for on-line battery estimation. Several aspects are considered, from modelling and parameter estimation to estimation of SOC, energy, and power. Recursive algorithms are proposed for estimation of parameters and SOC, while power and energy are estimated using algebraic expressions derived from equivalent circuit battery models. The algorithms are evaluated on lithium-ion battery cell data collected laboratory tests. For the cell chemistries considered, the evaluation indicates that accuracy within 2% can be achieved for both SOC and power, also in cases with limited prior information about the cell.