D-optimal design of experiments applied to lithium battery for ageing model calibration

This paper exploits an optimal design of experiments to calibrate an ageing model for a graphite/LMO-NMC battery. The study is based on experimental data from the MOBICUS project. It features a D-optimal design for power cycling tests and separate calendar tests controlling temperature (0-60 degrees C), state-of charge (SoC, 0-100%), as well as charge/discharge current separately (0, C/3, 1C). Altogether, 27 conditions were executed. The modeling approach makes an analogy between chemical kinetics and battery degradation which allows to identify a rate of capacity fade for constant ageing conditions. Statistical influence of each ageing factors on the degradation rate are studied. Quadratic effects of temperature and SoC are significant on calendar degradation, while an interaction effect between temperature and current describes the coupling to cycling fade. The double quadratic with interaction expression accurately models both calendar and cycling data with R-2 = 0.96. The coupled model is validated on six tests alternating calendar and cycling periods with 1.08% absolute error after 200 days. Feedback is given on the initial design and a new D-optimal of only 10 runs is proposed. The method allows to reduce experimental cost to calibrate an ageing model, which can be used for degradation management of battery-based storage systems. (C) 2017 Elsevier Ltd. All rights reserved.