Enhancement of AI-based Prediction Capability of Electrodes in High-Temperature High-Pressure Conditions

Abstract

Rechargeable batteries have emerged as the most efficient energy storage solution for intermittent renewable energy sources. Currently, lithium-ion batteries (LIBs) dominate the energy storage landscape due to their high power and energy density. However, LIBs face challenges such as limited availability and rising costs of lithium precursor materials. In response, sodium-ion batteries (SIBs) have emerged as a promising alternative to LIBs, though they remain under active research. Battery research, particularly in material design, is complex due to unpredictable interactions between different materials and their properties. Conventional methods rely on experimental data and physics-based models, employing a forward trial-anderror approach that can be time-consuming. Artificial Intelligence (AI) offers a transformative solution to expedite the discovery of battery electrode materials. Parameters such as voltage, specific capacity, and volume change throughout cycles are crucial for determining the suitability and performance of battery electrode materials under various operating conditions, including high-temperature and pressure. This research leverages AI to accelerate the discovery of battery electrode materials by focusing on predicting three key performance metrics: specific capacity, volume change percentage, and average voltage. Four AI models were developed: decision tree, random forest, Support Vector Machine (SVM), and Deep Neural Network (DNN). Among these, the DNN demonstrated exceptional predictive accuracy for these metrics, achieving accuracy levels ranging from 84.5% to 95.4%. To enhance efficiency and effectiveness, the study integrates the DNN model with the NSGA2 multi-objective optimization algorithm. This approach identifies Pareto-optimal solutions that balance volume and capacity considerations, facilitating precise adjustments to material properties. The findings underscore the transformative potential of machine learning in revolutionizing battery material research and development. By seamlessly integrating advanced predictive modeling with optimization strategies, this research accelerates the discovery of high-performance sodium-ion battery electrodes.

Date of Award	3 Jul 2024
Original language	American English
Supervisor	Ali Alhammadi (Supervisor)