SUSTAINABLE POWER CONVERSION LAB
Laboratory
Laboratory
Battery and Energy Storage
In the foreseeable future, batteries and emerging energy storage systems will be integral to nearly all industrial sectors. Our research focuses on cutting-edge battery and energy storage technologies closely linked to power electronics, including advanced modeling of lithium-ion and sodium-ion cells, modules, and packs, as well as hybrid and all-solid-state battery systems. We are also developing intelligent battery management systems (BMSs) and on-board battery chargers (OBCs) that leverage AI for both hardware and software optimization. Our mission is to create practical, deployable battery technologies that enhance safety, reliability, and cost-effectiveness, ultimately contributing to societal and industrial advancement.
New Charging Scheme for Li-ion Battery Cell Based on An AI-Enhanced Electro-Thermal-Degradation Model
We propose a hybrid modeling and control framework that combines physical interpretability with data-driven intelligence. The Random Forest–enhanced electro-thermal-degradation model accurately captures nonlinear electro-thermal behaviors, while a constraint-aware Bayesian Optimization scheme optimizes charging speed and degradation suppression under safety limits. Implemented through a dual-loop structure with temperature and current regulation, the strategy enables secure, efficient, and real-time charging on embedded microcontrollers with minimal computational cost. Experimental results show substantial improvements in charging time, thermal safety, and cycle life over conventional CC-CV and temperature-regulated methods. This work has been published in the IEEE Transactions on Power Electronics under the title “A Secure-Sustainable-Fast Charging Strategy for Lithium-ion Batteries based on a Random Forest-Enhanced Electro-Thermal-Degradation Model”. A concise conference version of this work was presented here.
New Control Scheme for Li-ion Battery Modules (Multiple Cells) Based on A Physics-Informed Neural Network
We propose a hybrid modeling and predictive control framework designed to mitigate cell-to-cell temperature nonuniformity in Li-ion battery modules. A two-dimensional Thermal Network Model (2D-TNM) is developed for fast and accurate multi-cell temperature prediction, augmented by a Physics-Informed Neural Network (PINN) that captures the nonlinear internal resistance behavior under physical constraints. To efficiently identify model parameters, Bayesian Optimization (BO) is employed. Building on this model, a Model Predictive Temperature Balancing Control (MPTBC) strategy is implemented to dynamically regulate individual cell currents through a Single-Input Multi-Output Switched-Capacitor (SIMO-SC) converter, achieving real-time thermal balancing with low computational complexity. Experimental validation demonstrates that the proposed method reduces cell-to-cell temperature differences to below 0.5°C, enhances thermal uniformity, and mitigates degradation compared to conventional equal-current charging. This work has been published in the IEEE Transactions on Industrial Electronics under the title “Physics-Informed Neural Network-Enhanced Model Predictive Temperature Balancing Control for Li-ion Battery Modules”. A concise conference version of this work was presented here.
