State of health estimation for lithium-ion batteries based on data driven techniques

Abstract

Lithium-Ion batteries (LIBs) have an increasingly critical role in the daily lives of people with their applications in renewable and non-renewable energy systems as an energy storage solution. As a result, the importance of accurate on-board estimations of LIBs has increased in criticality. Due to the complex ageing mechanism of LIBs, this report presents a simple data-driven technique involving Gaussian Process Regression (GPR), which estimates the battery capacities using time-series voltage measurements over a period of galvanostatic operation. The GPR operation is applied to 8 cells from the University of Oxford dataset with 3 variables, the duration of galvanostatic operation, number of datapoints and the lower limit of galvanostatic operation voltage. The final selected model parameters have a model Root Mean Squared Percentage Error (RMSPE) or between 0.33-0.6%.