Fig. 1: A data-driven experimental workflow for additive discovery and optimization for lithium-ion batteries. | Nature Communications

Fig. 1: A data-driven experimental workflow for additive discovery and optimization for lithium-ion batteries.

From: Data-driven design of electrolyte additives supporting high-performance 5 V LiNi0.5Mn1.5O4 positive electrodes

Fig. 1

a Schematic representation of a machine learning (ML)-guided design of experiment workflow for electrolyte additive discovery, consisting of additive synthesis, battery assembly, performance testing, ML model training, new formula selection, and repeat. b Sequential method for developing ML models to predict electrolyte additives starts with initial data collection followed by feature generation and selection for ML models. Then, trained ML models are used to predict performance metrics of unknown candidates, of which the predicted top candidates are suggested for experimental validation. ASI, ΔASI, and Q denote final area-specific impedance, impedance rise, and final specific capacity, respectively.

Back to article page