Optimizing Electric Vehicle Performance With AI-Driven Battery Management Systems

Abstract

Decreasing battery functionality is the primary roadblock to the widespread adoption of electric vehicles (EVs). Hence, solutions are required to optimize the safety, performance, and cycle life of lithium-ion batteries. To address this issue, we present the first AI-driven battery management system (BMS) capable of model-free prediction of state-of-charge, state-of-health, and likely failure dynamics in EV batteries. We utilize industrial X-ray computed tomography to inspect the internal electrodes and separator quality and state-of-charge, and electrochemical impedance spectroscopy to quantify cell state-of-health. Our model-free approach tackled both experimental and industrial EV-relevant data; we demonstrated ground-breaking prediction accuracy and showed neither calibration nor any commercial tool assistance was required. The approach offers a qualitatively fundamentally novel perspective on battery performance that will enable its ultimate understanding and optimized design. Our approach directly supports sustainability and the low-cost driving of electric vehicles.The increasing pace of vehicle electrification and hybridization necessitates accelerating advances in lithium-ion battery performance and safety, which are mainly reliant on sophisticated embedded battery management systems. Specifically, lifelong accurate tracking of the state-of-charge (SoC) and state-of-health (SoH) of the individual cells is of cardinal importance. The influence of underperformance in these capabilities will cause, amongst others, EVs to be stranded at the side of the highway, downtime of large-scale electricity energy buffers, reduced overall EV battery pack use, and early frequent costly degradation and replacement. Decreased reliability affects not only the promise for hardware-in-the-loop research but is a direct consequence of EV industry proliferation. Numerous problems can arise with battery characteristics alone, and the consensus is that the issues will only become more severe. To strongly reduce this risk and accommodate the electrification evolution, battery service life needs to be prolonged by pursuing advanced machine learning algorithms focusing on battery monitoring, modeling, and management.