Li-ion Battery Electrode Health Diagnostics using Machine Learning

Abstract

Diagnostic information of a battery allows for its maximum utilization while avoiding unfavorable or even dangerous operations. Model-based approaches have been proposed to identify the state of health (SOH) related parameters in lithium-ion (Li-ion) batteries; however, high computational cost for solving optimization-based parameter identification makes these approaches difficult to be implemented in onboard applications. To address this issue, this paper proposes a machine learning-based approach using a neural network (NN) model for identifying electrode-level degradation of Li-ion batteries. For the diagnosis of electrode-level degradation (i.e., loss of active material (LAM) for each electrode and loss of lithium inventory (LLI)), electrochemical features are extracted from both incremental capacity (IC) curve and differential voltage (DV) curve. The developed NN model trained with the proposed electrochemical features shows strong potential in identifying each degradation mode accurately: the RMSE of all degradation modes is less than 0.1.