Prediction of Supercapacitor Performance for Cobalt Oxide and Its Derivatives using Machine Learning Techniques

Abstract Supercapacitors' high-power density and quick charge-discharge times have made them stand out as potential energy storage technologies. The electrochemical properties of cobalt oxide and its derivatives have sparked substantial interest in their application as electrode materials for supercapacitors. This study aims to predict the performance of supercapacitors utilizing cobalt oxide and its derivatives through different machine learning models. The prediction model integrates material properties, electrode morphology, and electrolyte characteristics to forecast key performance metrics such as specific capacitance, energy density, power density and cyclic stability. By employing advanced computational tools, this research contributes to the understanding and optimization of cobalt oxide-based supercapacitors, facilitating their use in many different systems that stores energy. The experimental results such as high TPR values of all the grades (0.87 for grade A, 0.91 for grade B, 0.62 for grade C and 0.79 for grade D) on applying RF model confirm the predictive methodology outlined in this study. Furthermore, according to this study, MLP model combined with bagging (Bagging + MLP) is a better model as it gives least MAE and RMSE of 0.15 and 0.30 respectively with a precision of 0.642, recall of 0.562, accuracy of 0.836 and F1 score of 0.579.