Decoding Li-Ion Battery Longevity: Domain-Driven Approach with Advanced Feature Engineering and Machine Learning for RUL and SOH Estimation

Lithium-ion (Li-Ion) batteries have been utilized in several industrial products, ranging from consumer gadgets to electric vehicles (EVs). When handling these batteries, assessing the battery’s state of health (SOH) is crucial to enhance their safety. This research presents a technique for forecasting the remaining useful life (RUL) and SOH using the incremental capacity (IC) curve, constant current time (CC-Time), constant current value, and other relevant data, including the XGBoost model (extreme gradient descent). Throughout the constant current (CC) charging process, the current and voltage measurements of the batteries are transformed into incremental capacity curve (IC-curves). To set up the non-linear relations between SOH/RUL fall and IC-Curve modification and constant current phase by XGBoost, aging research data (conducted by Stanford University) were employed. Experimental results show that there is good accuracy (R2 > 0.91), with a less mean absolute error. With the SOH/RUL estimation technique shown in this chapter, it is possible to accomplish highly accurate SOH/RUL digital estimation.