A Model Predictive Control Scheme for Fast Charging via Accurate Quadratic Battery Models

Performance and degradation prediction along with control of lithium ion batteries is a critical tool to advance electrification of transportation across the world. In this work, we devise a quadratic battery model which retains the structure of higher fidelity models, thereby allowing for implementation of constraints on internal states to ensure safety and to limit degradation. We validate our model against a pseudo-2-dimensional partial-differential-equation based model, and demonstrate that our model can achieve high accuracy in spite of its simple nature. Finally, we implement the proposed MPC algorithm in a high-fidelity simulator, demonstrating its ability to charge nearly twice as fast as the most commonly used charging protocols whilst respecting constraints on anode potential, temperature, current, and state of charge. Calculation of the MPC protocol takes less than 0.15 s, meaning it is capable of running in real time.