Predictive model based battery constraints for electric motor control within EV Powertrains

This paper presents a method of predicting the maximum power capability of a Li-Ion battery, to be used for electric motor control within automotive powertrains. As maximum power is highly dependent on battery state, the method consists of a pack level state observer coupled with a predictive battery model. Results indicate that the battery state estimation algorithm can estimate a cell State-of-Charge (SoC) within 3%, while pack level simulations show how this method can be enhanced to provide battery pack level estimates, correctly capturing the spread in terms of State of Charge of the cells within the pack, which is essential for accurate maximum power prediction. Tests show that the maximum battery power varies significantly with SoC. At an ambient temperature of 20°C, as much as a three-fold decrease in power capability is measured for charging power, at SoC values above 90%, and discharging power, at SoC values under 20%. The maximum power prediction algorithm presented in this study is able to correctly predict the maximum battery power over the complete operating range of SoC, at 20°C. Low temperature maximum discharging power tests were carried out, to investigate electric vehicle cold start scenarios. The tests show a strong impact of temperature on the power which can be withdrawn from the battery. At 35% SoC, 2.5 times less power can be withdrawn from the battery at a temperature of 0°C, compared to 20°C.