Embedded real-time state observer implementation for lithium-ion cells using an electrochemical model and extended Kalman filter

Accurate knowledge of the current state of lithium-ion battery cells is crucial for enhanced operational management of battery electric vehicles. Electrochemistry-based models provide the unique ability to predict cell internal states for advanced state monitoring, such as the anode potential, being of great interest for preventing lithium plating during fast charging. Furthermore, they are extendable by electrochemistry-based descriptions of cell degradation mechanisms. However, unlike the widely used equivalent circuit models, they are more computationally expensive and therefore have been scarcely used in embedded systems for online applications. This work addresses an efficient implementation of the popular pseudo-two-dimensional (p2D) electrochemical model in combination with a nonlinear filtering algorithm on a state-of-the-art microcontroller for automotive applications. For the first time, it is shown that with a suitable discretization scheme for the finite-difference approximation, a state observer convergence duration below 40 ms can be achieved for constant current, constant voltage charging at a current rate of 2C, with fast recovery of even local states. The mean relative error of the model internal states is less than 2% compared to an about 15 times slower implementation. This works therefore contributes valuably to demonstrating fast and robust performance of online state estimation using electrochemical models.

Automated summary

Accurate knowledge of the current state of lithium-ion battery cells is crucial for enhanced operational management of battery electric vehicles. This study demonstrates the fast and robust performance of online state estimation using an efficient implementation of a popular pseudo-two-dimensional electrochemical model and a nonlinear filtering algorithm.