Artificial neural network-enabled approaches toward mass balancing and cell optimization of lithium dual ion batteries

Abstract

Dual ion batteries (DIBs) are gaining prominence as alternate energy storage devices in recent years due to their high operating potential and low cost. For an all-carbon based DIB, owing to the low cathode capacity, the anode:cathode mass ratio has a much more significant effect on the overall cell performance. Achieving a generalized model towards optimization of active mass balancing in these devices still remains a challenge. With an aim to accelerate the development of dual ion batteries, herein, we attempt to make an experimental data-driven deep neural network model towards mass balancing and cell optimization of a graphite-graphite lithium dual ion battery. We designed appropriate experimental conditions with maximum ranges of control variables and performed detailed electrochemical studies. The data generated through such a minimum number of experiments are used to predict the performance characteristics of the graphite/graphite-based lithium dual ion full cell under a wide range of active electrode mass ratios and current densities, through a machine learning approach. We demonstrate that our model can predict the charge-discharge profiles of the full cell at any given current density in the range 50 mA g−1 and 1 A g−1. Further, using these charge-discharge profiles, we find the discharge capacities of the cell for varying mass ratios and different mass loading. Though we address here the case of graphite/graphite lithium dual ion battery cells, the approach can easily be extended to other battery chemistries to achieve accelerated development of advanced energy storage devices.