Abstract

Rechargeable batteries have proven to be the most efficient energy storage solution for the intermittent renewable energy sources. Currently, the lithium-ion battery (LIB) is regarded as the leading energy storage system because of its power and energy density. Nevertheless, LIBs are facing issues such as the limited availability of lithium precursor materials and the cost attributed to this issue. Thus, sodium ion batteries (SIBs) are regarded as a highly promising alternative to LIBs, but they are still under research. Battery research and especially battery material design is challenging because of the complex structure-property relationships, where different materials interact unpredictably and uncontrollably. Hence, artificial intelligence (AI) can be used to accelerate the process of battery electrode material discovery. There are many parameters that determine whether the battery electrode material is suitable or not, such as the voltage and the specific capacity which both contribute to the overall energy density. Additionally, the volume change throughout the cycles contribute to the safety, efficiency and cyclability of the battery. Thus, several AI techniques such as decision trees, support vector machine, random forest and deep neural network (DNN) models have been used to predict the average voltage, maximum specific capacity, and volume change of the battery. The results showed that the DNN model performed best with an average mean absolute error of around 0.164. Based on the results obtained, machine learning has proven to be a crucial tool towards revolutionizing the battery material industry.