Optimized charging of lithium-ion batteries with physico-chemical models
Ringbeck, Florian; Sauer, Dirk Uwe (Thesis advisor); Krewer, Ulrike (Thesis advisor)
Aachen : RWTH Aachen University (2021)
Book, Dissertation / PhD Thesis
In: Aachener Beiträge des ISEA 159
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2021
For the usability of lithium-ion batteries in many applications, their fast charging capability plays an important role. However, a too high charging current can permanently damage the battery and pose a safety risk. Therefore, charging patterns with an optimal current are required for many applications. Physico-chemical battery models, sometimes also called electrochemical battery models, allow deep insights into the internal states and processes of lithium-ion batteries. Therefore, they can be a valuable tool to analyze existing charging schemes and develop new charging methods that allow optimal battery charging. This thesis uses such battery models for the analysis of existing pulse charging regimes and the development of a model-based charging controller for determining the optimal charging current. The goal is to charge the batteries as fast as possible under given environmental conditions with a minimal impact on safety and cycle life. A detailed analysis was conducted to assess the usefulness of pulse charging schemes in terms of speed, efficiency, and impact on aging. It is shown by measurements and physico-chemical simulations, that such schemes offer no immediate benefit during charging. While also for the capacity retention, no benefits can be proven, pulse charging slows down the resistance increase of the tested cells.Lithium plating is one of the primary aging mechanisms occurring during fast charging, especially at low temperatures. As the anode potential is a clear indicator for plating, a model-based observer estimating this potential can be used to implement a charge controller. As the charging controller has to run in real-time, a reduction of the complexity of the applied physico-chemical model is necessary. State-of-the-art order reduction methods are discussed regarding their benefits and drawbacks. Orthogonal-collocation is a promising and scalable technique to reduce computational effort while not losing physical meaning. Transfer function and subspace identification techniques are a more performant alternative at the cost of flexibility and meaningfulness of model states. A reduced single-particle model with coarse discretization is then used to implement a state observer, to allow the model-based controller to track the battery state accurately. The observer, an unscented Kalman filter, allows a conservative estimation of the battery’s anode potential even in the presence of inevitable initialization, measurement, and parameterization errors. The occurrence of lithium plating has to be detected to validate such a charging controller. Therefore a post-mortem study was conducted to assess the applicability of electrical plating detection methods. The research shows that electrical indicators of lithium stripping are sufficient but not necessary criteria for the occurrence of lithium plating. Concluding validation measurements show that the model-based charging controller can mitigate the occurrence of plating at low temperatures. The current profile applied is superior to a constant current profile with the same average current rate and operates at the limit of lithium-plating occurrence.