Toward more realistic microgrid optimization: Experiment and high-efficient model of Li-ion battery degradation under dynamic conditions

Accurate and high-efficient battery life prediction is critical for microgrid optimization and control problems. Extracted from EV (electric vehicle)-PV(photovoltaics)-battery-based microgrid working profiles, five sets of accelerated aging experiments are conducted on LFP (graphite-LiFePO4) cells to reflect the effect of different energy storage capacities on battery degradation. Apart from SEI (solid electrolyte interface) growth and LAM (loss of active materials), lithium plating is likely to occur after prolonged cycling under the DOD (depth of discharge) range of over 80% when there are current pulses of around 1.5C. As for the degradation modelling, to efficiently solve while maintaining prediction accuracy, this work develops a reduced-order semi-empirical model (ROSEM) to compute fastly for control algorithms. The model reveals physical mechanisms and thus can generalize across actual conditions compared with another four semi-empirical models. The proposed ROSEM can precisely predict nonlinear battery degradation with less than 1.6% relative errors, and the computation duration for 500 cycles is less than 5s. The margin of simulation errors can be further narrowed to 0.6% with the correction of ANN (artificial neural network).