Research Group Helge Stein

Efforts to enhance the state of health (SOH) estimation for lithium-ion batteries have increasingly focused on diverse machine learning methods, especially with the promising artificial intelligence of things (AIoT) technologies. However, direct SOH prediction using electrical signals as inputs lacks an explanation of the aging mechanism evolution. This study proposes an explainable AIoT framework based on a physics-informed sequence-to-sequence (PIISeq2Seq) model to accurately estimate SOH. The obtained electrical signals are transmitted to the AIoT cloud system for parameter identification and SOH estimation. Moreover, the degradation pathways with different potential mechanisms for lab batteries with LiNiO2 cathodes and Si@C anodes are investigated by quantifying polarization processes. The robustness of PIISeq2Seq across different temperatures is further validated using a public data set of commercial batteries, achieving an mean absolute percentage error of 1.11% on lab cells and 1.06%–1.14% on commercial batteries. This work offers valuable insights for developing explainable approaches to battery SOH estimation, especially the explainable AIoT for real-time evaluation of the lithium-ion batteries in industrial automation.

Accurate prediction of battery behavior under different dynamic operating conditions is critical for both fundamental research and practical applications. However, the diversity of emerging materials and cell architectures presents significant challenges to the generalizability of conventional prognostic approaches. Here, we propose a novel physics-informed battery modeling network (PIBMN) that integrates data-driven learning with physical priors, enabling continuous parameter adaptation and broad applicability across cell formats and chemistries. PIBMN effectively captures both fast and slow dynamic responses under a wide range of load profiles, applicable to both commercial and laboratory-scale cells. By maintaining nonlinear expressivity while ensuring numerical stability, the model yields high-fidelity, interpretable representations of internal electrochemical states. Beyond conventional health prognostics, PIBMN introduces a novel capability to decouple complex electrochemical kinetics and concurrently track terminal voltage in real time, enabling mechanistic diagnostics with high resolution. As such, PIBMN establishes a versatile and scalable framework for in-line quality control, adaptive cell-specific battery management, and data-informed optimization of next-generation battery manufacturing processes.

Bespoke formation of Batteries offers improved lifetime and performance but is generally associated with long processing times, high cost, and large floorspace. Facile strategies like heating or increasing the formation current, as well as current alterations during formation have their limits in speed up and efficiency. We present pulsed formation on graphitic anode full cells as an accelerated formation strategy and investigate its influence on various quality parameters. Optimized pulsed charging is demonstrated herein to reduce the formation time by more than 50% whilst maintaining or improving all other cell quality parameters including discharge capacity. The newly discovered protocol is scaled up to 25Ah prismatic cells in the PHEV1 format that confirm the accelerated and improved pulsed formation strategy. We attribute the accelerated and improved formation to an apt balance of surface and bulk diffusion which results in thinner, more homogenous SEI. Dynamics of pulsed formation also allow for the extraction of new quality markers while formation is happening.

Accelerated formation protocols that utilize pulsed charging offer an unprecedented wealth of electrochemical data. Herein we present methods to extract diagnostic data relating to a pseudo-diffusion coefficients, internal resistance, and others that give live insight to the solid electrolyte interphase (SEI) growth. Specifically, we present a pure mathematical method to track formation progression at near-real time and chart a path towards incorporation of adjusting pulse parameters towards targeted SEI synthesis. The method and analysis performed on 3 mAh cells but can also be applied to higher capacity cells.

High-performance batteries need accelerated discovery and optimization of new anode materials. Herein, we explore the Si─Ge─Sn ternary alloy system as a candidate fast-charging anode materials system by utilizing a scanning droplet cell (SDC) as an autonomous electrochemical characterization tool with the goal of subsequent upscaling. As the SDC is performing experiments sequentially, an exploration of the entire ternary space is unfeasible due to time constraints. Thus, closed-loop optimization, guided by real-time data analysis and sequential learning algorithms, is utilized to direct experiments. The lead material identified is scaled up to a coin cell to validate the findings from the autonomous millimeter-scale thin-film electrochemical experimentation. Explainable machine learning (ML) models incorporating data from high-throughput Raman spectroscopy and X-ray diffraction (XRD) are used to elucidate the effect of short and long-range ordering on material performance.