Machine learning technique-based data-driven model of exploring effects of electrolyte additives on LiNi_(0.6)Mn_(0.2)Co_(0.2)O_2/graphite cell

摘要

Recently, state-of-the-art and perspective Li-ion batteries have been focused on NMC622 cathode for energy densities and power densities enhancement. This work introduces a novel approach of combining experimental data generation and computational power for understanding the effects of electrolyte additives including vinylene carbonate (VC), lithium bis(oxalate) borate (LiBOB), and fluoroethylene carbonate (FEC) in the carbonate-based electrolyte (a mixture of EC: EMC = 3:7), which caused on NMC622/graphite cell. Furthermore, a screening study of negative to the positive ratio (N/P ratio) confirmed that the best capacity matching of N/P similar to 1.07 was taken for full-cell assembly. It was demonstrated that the presence of additives improved only initial capacity but suppress remarkably capacity fading. With 1 %wt. FEC, the cell delivered 155.1 mAh g(-1) and remained 47.9% of initial capacity after 100 cycles compared to 140.5 mAh g(-1) concentration and 9.3% after 30 cycles, respectively. However, using high FEC as a co-solvent of carbonate-based electrolyte is to be promising in improving long-cycling performance, even in a dual-additive electrolyte with LiBOB. To understand the additive's impact on cell performance, a simulation of an artificial neural network (ANN) was applied. Based on the ANN simulation, the cell could obtain 160 mAh g(-1) and 65% capacity retention after 100 cycles when using 1.1 M LiPF6 in FEC: EC: EMC (0.7: 2.8:6.5) (%v) with 0.6 %wt. LiBOB and 0.01 %wt. VC. Therefore, the robust prediction of the technique for diagnostics and prognostics of LIBs is effective since it helps reduce experiment cost and save the time of conducting experiments.