Online state of health estimation on NMC cells based on predictive analytics

Abstract Accurate on board state of health estimation is a key battery management system function to provide optimal management of the battery system under control. In this regard, this paper presents an extensive study and comparison of three of commonly used supervised learning methods for state of health estimation in Graphite/Nickel Manganese Cobalt oxide cells. The three methods were based from the study of both incremental capacity and differential voltage curves. According to the ageing evolution of both curves, features were extracted and used as inputs for the estimation techniques. Ordinary Least Squares, Multilayer Perceptron and Support Vector Machine were used as the estimation techniques and accurate results were obtained while requiring a low computational effort. Moreover, this work allows a deep comparison of the different estimation techniques in terms of accuracy, online estimation and BMS applicability. In addition, estimation can be developed by partial charging and/or partial discharging, reducing the required maintenance time.