Enabling high-accuracy lithium-ion battery status prediction via machine learning-integrated perovskite sensors
Monitoring the safety and reliability of lithium-ion batteries (LIBs) under extreme operating conditions faces significant challenges, due to the difficulty of conventional battery management systems (BMS) in capturing internal dynamic changes in real-time. To address this challenge, this paper proposes an implantable flexible sensor based on ZnO-passivated CsPbBr 3 quantum dots encapsulated in EPDM, which realizes real-time monitoring inside lithium batteries. Experiments demonstrate that the sensor resistance exhibits high synchronization with battery voltage changes (response time < 5 s), and the capacity fade rate difference after sensor implantation is only 14.2 % over 15 days. Theoretical calculations and in situ X-ray diffraction (XRD) reveal the energy band modulation mechanism of Li + adsorption - induced. During LIBs charging/discharging, changes in Li + concentration around the sensor alter the band structure and density of states (DOS) of the sensing material, leading to changes in its electrical signal. Thus, the working state of the LIBs can be monitored and predicted by tracking changes in the sensor's electrical signal. Further integration of an LSTM-multitask learning model enables advanced prediction of the LIBs operating state based on sensor signals (5000 s ahead), with a mean absolute error (MAE) of 6.10 %, and charge/discharge rate recognition with 96.7 % accuracy.
危险品化学品经营许可证（带存储）