Analysis of cell-level abnormality diagnosis based on battery pack voltage information

Efficient and secure battery management is essential to optimize the performance and life of battery-powered systems. The key to achieving this goal is to accurately estimate the current state of the battery, which traditionally relies on data collected by the Battery Management System (BMS) from individual cells. However, certain BMS configurations collect data only at the pack level, which obscures insights into the state of individual cells and is likely to overlook significant cell-level anomalies. This restriction requires a new method to estimate the internal state of individual cells using only pack-level data. This paper resolves this gap by leveraging pack-level data and proposing an innovative approach to indirectly estimate the internal state of the cells in the battery pack using neural network algorithms without the need to physically decompose the battery pack. Our method will leverage the power of machine learning to significantly improve the granularity and accuracy of battery state estimation, paving the way for more efficient and reliable battery management solutions. The proposed method also provides a cost-effective and non-disturbing alternative to traditional cell-level data collection methods, making it a powerful option for battery management in a variety of applications.