State-of-charge estimation over full battery lifespan under diverse fast-charging protocols: A lightweight base-error joint modeling framework

Abstract

Accurate state-of-charge (SOC) online estimation during various multi-stage constant current (MCC) fast-charging protocols over battery entire lifespan holds significant importance. In this work, we develop a lightweight-training oriented data-driven base-error joint modeling framework to fill this research gap. Through deep learning-based initial-cycle data training and lightweight machine learning-based typical-cycle data training, we only extract approximately 1 % of whole battery data for data-driven base-error joint modeling. With consideration of SOC time-dependency, short-term Ampere-hour is further combined via a simple filter structure to guarantee final SOC estimation accuracy. The validation, derived from a public battery degradation dataset comprising 8 different MCC fast-charging protocols from 46 cells, demonstrates that our framework allows rapid data-driven base-error joint modeling with training time only about l min, where both average mean absolute error and average root mean square error of SOC estimation during various MCC fast-charging protocols over battery entire lifespan are roughly below 0.3 %. Our work, for the first time, reveals the possibility of joint data-driven model trained via extremely few data on accurate SOC online estimation with consideration of various MCC fast-charging protocols and battery degradation status, and also offers a pretty concise but efficient solution for multi-scenario battery aging diagnosis and voltage dynamics forecast. The code accompanying this work is available at https://github.com/szzhang96/A-light-weighted-training-oriented-data-driven-base-error-joint-modeling-method-for-SOC-estimation.