Model-Inversion-Based Fast Charging Control of Lithium-Ion Batteries Considering Parameter Uncertainty

Licentiatavhandling, 2024

The lithium-ion battery is a key technology for achieving sustainable mobility. However, due to its limited energy density, one of the main obstacles to replacing fossil-fueled vehicles with lithium-ion battery-powered electric vehicles is range anxiety. Ultra-fast charging is one way of resolving this issue. Unfortunately, simply increasing the charging current rates to reduce the charging time can lead to accelerated aging and shortened service life if the internal conditions of the battery cell are ignored. To achieve health-aware fast charging, an electrochemical model can provide valuable information for observing the internal states of batteries. A well-designed charging algorithm is needed to balance the trade-off between charging time and the rate of degradation. However, this is a challenging task due to uncertainties arising from various high-dimensional modeling and measurement errors.

This thesis investigates the influence of various uncertainties in designing the fast-charging control algorithms of lithium-ion batteries, such as current sensor bias, structural model differences, and errors in identified parameters. The study starts by spatially discretizing the pseudo two-dimensional (P2D) model, the most widely used electrochemical modeling framework for lithium-ion batteries. One key finding is that in the presence of parameter uncertainties, increasing the system order of the discretized model does not necessarily yield meaningful improvements. These uncertainties are often inherent due to difficulties in measurements or lack of clear physical interpretations.

To address the influence of parameter uncertainty during fast charging, a method for calculating a suitable safety margin to avoid lithium plating is developed by inverting the single particle model (SPM) of lithium-ion batteries. With knowledge of the range of parameter biases, the sensitivity of the safety margin with respect to these biases can be calculated, and the range of the safety margin can be determined. The minimum constant safety margin enabling lithium-plating-free fast charging is calculated based on this. An analysis shows that the required charging time is heavily dependent on the set safety margin. To achieve optimized performance, a method for calculating a time-varying safety margin is therefore developed, which speeds up the charging process by determining the maximum possible charging current based on the range of given parameter uncertainties at each time instant. Based on this method, an online strategy is proposed to further reduce the charging time by adaptively updating the learned information about the uncertainties.

To conclude, this thesis contributes to the field by analyzing previously overlooked factors affecting aging-aware fast-charging design based on electrochemical models. Building on this analysis, methods to determine both constant and dynamic safety margins with online parameter uncertainty reduction are derived. The proposed methods ensure that shortened charging times can be achieved without inducing lithium plating, even under various model uncertainties, which is promising for future health-aware charging of electric vehicles.